Datasets:

---

bigcode

/

jupyter-parsed

like 4

Follow

BigCode 1.4k

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "<a href=\"https://colab.research.google.com/github/julianovale/pythonparatodos/blob/main/M%C3%B3dulo07Aula24.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>", "_____no_output_____" ], [ "PDF da apresentação:\r\n\r\nhttps://drive.google.com/file/d/1k5xfHrWnjPJJc05kB3op_05PAZpZpAfO/view\r\n", "_____no_output_____" ] ], [ [ "def ma(lista):\r\n soma = 0\r\n cont = 0\r\n for i in lista:\r\n soma = soma + i\r\n cont = cont + 1\r\n return soma/cont\r\n\r\ndef mg(lista):\r\n prod = 1\r\n cont = 0\r\n for i in lista:\r\n prod = prod*i\r\n cont = cont + 1\r\n return prod**(1/cont)\r\n\r\ndef mediaAG(lista, tip = 0):\r\n if (tip == 0):\r\n r = ma(lista)\r\n else:\r\n r = mg(lista)\r\n return r", "_____no_output_____" ], [ "l1 = [7,8,9]\r\nt = 0\r\nm1 = mediaAG(tip = t, lista = l1)\r\nprint(\"A média aritmética é: {0:.2f}\".format(m1))\r\n\r\nt = 1\r\nm2 = mediaAG(tip = t, lista = l1)\r\nprint(\"A média geométrica é: {0:.2f}\".format(m2))", "A média aritmética é: 8.00\nA média geométrica é: 7.96\n" ] ] ]

[ "markdown", "code" ]

[ [ "markdown", "markdown" ], [ "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Timeseries\n\nCe notebook présente quelques étapes simples pour une série temporelle. La plupart utilise le module [statsmodels.tsa](https://www.statsmodels.org/stable/tsa.html#module-statsmodels.tsa).", "_____no_output_____" ] ], [ [ "from jyquickhelper import add_notebook_menu\nadd_notebook_menu()", "_____no_output_____" ], [ "%matplotlib inline", "_____no_output_____" ] ], [ [ "## Données\n\nLes données sont artificielles mais simulent ce que pourraient être le chiffre d'affaires d'un magasin de quartier, des samedi très forts, une semaine morne, un Noël chargé, un été plat.", "_____no_output_____" ] ], [ [ "from ensae_teaching_cs.data import generate_sells\nimport pandas\ndf = pandas.DataFrame(generate_sells())\ndf.head()", "_____no_output_____" ] ], [ [ "## Premiers graphiques\n\nLa série a deux saisonnalités, hebdomadaire, mensuelle.", "_____no_output_____" ] ], [ [ "import matplotlib.pyplot as plt\nfig, ax = plt.subplots(1, 2, figsize=(14, 4))\ndf.iloc[-30:].set_index('date').plot(ax=ax[0])\ndf.set_index('date').plot(ax=ax[1])\nax[0].set_title(\"chiffre d'affaire sur le dernier mois\")\nax[1].set_title(\"chiffre d'affaire sur deux ans\");", "_____no_output_____" ] ], [ [ "Elle a une vague tendance, on peut calculer un tendance à l'ordre 1, 2, ...", "_____no_output_____" ] ], [ [ "from statsmodels.tsa.tsatools import detrend\nnotrend = detrend(df.value, order=1)\ndf[\"notrend\"] = notrend\ndf[\"trend\"] = df['value'] - notrend\nax = df.plot(x=\"date\", y=[\"value\", \"trend\"], figsize=(14,4))\nax.set_title('tendance');", "_____no_output_____" ] ], [ [ "Autocorrélations...", "_____no_output_____" ] ], [ [ "from statsmodels.tsa.stattools import acf\ncor = acf(df.value)\ncor", "_____no_output_____" ], [ "fig, ax = plt.subplots(1, 1, figsize=(14,2))\nax.plot(cor)\nax.set_title(\"Autocorrélogramme\");", "_____no_output_____" ] ], [ [ "La première saisonalité apparaît, 7, 14, 21... Les autocorrélations partielles confirment cela, plutôt 7 jours.", "_____no_output_____" ] ], [ [ "from statsmodels.tsa.stattools import pacf\nfrom statsmodels.graphics.tsaplots import plot_pacf\nplot_pacf(df.value, lags=50);", "_____no_output_____" ] ], [ [ "Comme il n'y a rien le dimanche, il vaut mieux les enlever. Garder des zéros nous priverait de modèles multiplicatifs.", "_____no_output_____" ] ], [ [ "df[\"weekday\"] = df.date.dt.weekday\ndf.head()", "_____no_output_____" ], [ "df_nosunday = df[df.weekday != 6]\ndf_nosunday.head(n=10)", "_____no_output_____" ], [ "fig, ax = plt.subplots(1, 1, figsize=(14,2))\ncor = acf(df_nosunday.value)\nax.plot(cor)\nax.set_title(\"Autocorrélogramme\");", "_____no_output_____" ], [ "plot_pacf(df_nosunday.value, lags=50);", "_____no_output_____" ] ], [ [ "On décompose la série en tendance + saisonnalité. Les étés et Noël apparaissent.", "_____no_output_____" ] ], [ [ "from statsmodels.tsa.seasonal import seasonal_decompose\nres = seasonal_decompose(df_nosunday.value, freq=7)\nres.plot();", "_____no_output_____" ], [ "plt.plot(res.seasonal[-30:])\nplt.title(\"Saisonnalité\");", "_____no_output_____" ], [ "cor = acf(res.trend[5:-5]);\nplt.plot(cor);", "_____no_output_____" ] ], [ [ "On cherche maintenant la saisonnalité de la série débarrassée de sa tendance herbdomadaire. On retrouve la saisonnalité mensuelle.", "_____no_output_____" ] ], [ [ "res_year = seasonal_decompose(res.trend[5:-5], freq=25)\nres_year.plot();", "_____no_output_____" ] ], [ [ "## Test de stationnarité\n\nLe test [KPSS](https://en.wikipedia.org/wiki/KPSS_test) permet de tester la stationnarité d'une série.", "_____no_output_____" ] ], [ [ "from statsmodels.tsa.stattools import kpss\nkpss(res.trend[5:-5])", "c:\\python372_x64\\lib\\site-packages\\statsmodels\\tsa\\stattools.py:1276: InterpolationWarning: p-value is smaller than the indicated p-value\n warn(\"p-value is smaller than the indicated p-value\", InterpolationWarning)\n" ] ], [ [ "Comme ce n'est pas toujours facile à interpréter, on simule une variable aléatoire gaussienne donc sans tendance.", "_____no_output_____" ] ], [ [ "from numpy.random import randn\nbruit = randn(1000)\nkpss(bruit)", "c:\\python372_x64\\lib\\site-packages\\statsmodels\\tsa\\stattools.py:1278: InterpolationWarning: p-value is greater than the indicated p-value\n warn(\"p-value is greater than the indicated p-value\", InterpolationWarning)\n" ] ], [ [ "Et puis une série avec une tendance forte.", "_____no_output_____" ] ], [ [ "from numpy.random import randn\nfrom numpy import arange\nbruit = randn(1000) * 100 + arange(1000) / 10\nkpss(bruit)", "_____no_output_____" ] ], [ [ "Une valeur forte indique une tendance et la série en a clairement une.", "_____no_output_____" ], [ "## Prédiction\n\nLes modèles *AR*, *ARMA*, *ARIMA* se concentrent sur une série à une dimension. En machine learning, il y a la série et plein d'autres informations. On construit une matrice avec des séries décalées.", "_____no_output_____" ] ], [ [ "from statsmodels.tsa.tsatools import lagmat\nlag = 8\nX = lagmat(df_nosunday[\"value\"], lag)\nlagged = df_nosunday.copy()\nfor c in range(1,lag+1):\n lagged[\"lag%d\" % c] = X[:, c-1]\nlagged.tail()", "_____no_output_____" ] ], [ [ "On ajoute ou on réécrit le jour de la semaine qu'on utilise comme variable supplémentaire.", "_____no_output_____" ] ], [ [ "lagged[\"weekday\"] = lagged.date.dt.weekday", "_____no_output_____" ], [ "X = lagged.drop([\"date\", \"value\", \"notrend\", \"trend\"], axis=1)\nY = lagged[\"value\"]\nX.shape, Y.shape", "_____no_output_____" ], [ "from numpy import corrcoef\ncorrcoef(X)", "_____no_output_____" ] ], [ [ "Etrange autant de grandes valeurs, cela veut dire que la tendance est trop forte pour calculer des corrélations, il vaudrait mieux tout recommencer avec la série $\\Delta Y_t = Y_t - Y_{t-1}$. Bref, passons...", "_____no_output_____" ] ], [ [ "X.columns", "_____no_output_____" ] ], [ [ "Une régression linéaire car les modèles linéaires sont toujours de bonnes baseline et pour connaître le modèle simulé, on ne fera pas beaucoup mieux.", "_____no_output_____" ] ], [ [ "from sklearn.linear_model import LinearRegression\nclr = LinearRegression()\nclr.fit(X, Y)", "_____no_output_____" ], [ "from sklearn.metrics import r2_score\nr2_score(Y, clr.predict(X))", "_____no_output_____" ], [ "clr.coef_", "_____no_output_____" ] ], [ [ "On retrouve la saisonnalité, $Y_t$ et $Y_{t-6}$ sont de mèches.", "_____no_output_____" ] ], [ [ "for i in range(1, X.shape[1]):\n print(\"X(t-%d)\" % (i), r2_score(Y, X.iloc[:, i]))", "X(t-1) -0.5799344874123231\nX(t-2) -1.0271995693021703\nX(t-3) -1.2451181610331497\nX(t-4) -1.0497781280554928\nX(t-5) -0.6300696727754616\nX(t-6) 0.8212380818165589\nX(t-7) -0.643903114304551\nX(t-8) -1.09689904382877\n" ] ], [ [ "Auparavant (l'année dernière en fait), je construisais deux bases, apprentissage et tests, comme ceci :", "_____no_output_____" ] ], [ [ "n = X.shape[0]\nX_train = X.iloc[:n * 2//3]\nX_test = X.iloc[n * 2//3:]\nY_train = Y[:n * 2//3]\nY_test = Y[n * 2//3:]", "_____no_output_____" ] ], [ [ "Et puis *scikit-learn* est arrivée avec [TimeSeriesSplit](https://scikit-learn.org/stable/modules/generated/sklearn.model_selection.TimeSeriesSplit.html).", "_____no_output_____" ] ], [ [ "from sklearn.model_selection import TimeSeriesSplit\ntscv = TimeSeriesSplit(n_splits=5)\nfor train_index, test_index in tscv.split(lagged):\n data_train, data_test = lagged.iloc[train_index, :], lagged.iloc[test_index, :]\n print(\"TRAIN:\", data_train.shape, \"TEST:\", data_test.shape)", "TRAIN: (107, 13) TEST: (104, 13)\nTRAIN: (211, 13) TEST: (104, 13)\nTRAIN: (315, 13) TEST: (104, 13)\nTRAIN: (419, 13) TEST: (104, 13)\nTRAIN: (523, 13) TEST: (104, 13)\n" ] ], [ [ "Et on calé une forêt aléatoire...", "_____no_output_____" ] ], [ [ "import warnings\nfrom sklearn.ensemble import RandomForestRegressor\nclr = RandomForestRegressor()\n\ndef train_test(clr, train_index, test_index):\n data_train = lagged.iloc[train_index, :]\n data_test = lagged.iloc[test_index, :]\n clr.fit(data_train.drop([\"value\", \"date\", \"notrend\", \"trend\"], \n axis=1), \n data_train.value)\n r2 = r2_score(data_test.value,\n clr.predict(data_test.drop([\"value\", \"date\", \"notrend\",\n \"trend\"], axis=1).as_matrix()))\n return r2\n\nwarnings.simplefilter(\"ignore\")\nlast_test_index = None\nfor train_index, test_index in tscv.split(lagged):\n r2 = train_test(clr, train_index, test_index) \n if last_test_index is not None:\n r2_prime = train_test(clr, last_test_index, test_index) \n print(r2, r2_prime)\n else:\n print(r2)\n last_test_index = test_index", "0.6795844303954097\n0.8643514472109561 0.8438674497139955\n0.889424015517347 0.8566080912412136\n0.8108804128996021 0.652299214153693\n0.9082580468326908 0.9146950904017541\n" ] ], [ [ "2 ans coupé en 5, soit tous les 5 mois, ça veut dire que ce découpage inclut parfois Noël, parfois l'été et que les performances y seront très sensibles.", "_____no_output_____" ] ], [ [ "from sklearn.metrics import r2_score\nr2 = r2_score(data_test.value,\n clr.predict(data_test.drop([\"value\", \"date\", \"notrend\",\n \"trend\"], axis=1).as_matrix()))\nr2", "_____no_output_____" ] ], [ [ "On compare avec le $r_2$ avec le même $r_2$ obtenu en utilisant $Y_{t-1}$, $Y_{t-2}$, ... $Y_{t-d}$ comme prédiction.", "_____no_output_____" ] ], [ [ "for i in range(1, 9):\n print(i, \":\", r2_score(data_test.value, data_test[\"lag%d\" % i]))", "1 : -0.9357733133623056\n2 : -1.4886358005577112\n3 : -1.8674091365311156\n4 : -1.5617646397661757\n5 : -0.9531085434874935\n6 : 0.8567402931376816\n7 : -0.9271996269480662\n8 : -1.5219232924913277\n" ], [ "lagged[:5]", "_____no_output_____" ] ], [ [ "En fait le jour de la semaine est une variable catégorielle, on crée une colonne par jour.", "_____no_output_____" ] ], [ [ "from sklearn.compose import ColumnTransformer\nfrom sklearn.preprocessing import OneHotEncoder", "_____no_output_____" ], [ "cols = ['lag1', 'lag2', 'lag3',\n 'lag4', 'lag5', 'lag6', 'lag7', 'lag8']\nct = ColumnTransformer(\n [('pass', \"passthrough\", cols),\n (\"dummies\", OneHotEncoder(), [\"weekday\"])])\npred = ct.fit(lagged).transform(lagged[:5])\npred", "_____no_output_____" ] ], [ [ "On met tout dans un pipeline parce que c'est plus joli, plus pratique aussi.", "_____no_output_____" ] ], [ [ "from sklearn.pipeline import make_pipeline\nfrom sklearn.decomposition import PCA, TruncatedSVD \ncols = ['lag1', 'lag2', 'lag3',\n 'lag4', 'lag5', 'lag6', 'lag7', 'lag8']\nmodel = make_pipeline(\n make_pipeline(\n ColumnTransformer(\n [('pass', \"passthrough\", cols),\n (\"dummies\", make_pipeline(OneHotEncoder(), \n TruncatedSVD(n_components=2)), [\"weekday\"])]),\n LinearRegression()))\nmodel.fit(lagged, lagged[\"value\"])", "_____no_output_____" ] ], [ [ "C'est plus facile à voir visuellement.", "_____no_output_____" ] ], [ [ "from mlinsights.plotting import pipeline2dot\ndot = pipeline2dot(model, lagged)\nfrom jyquickhelper import RenderJsDot\nRenderJsDot(dot)", "_____no_output_____" ], [ "r2_score(lagged['value'], model.predict(lagged))", "_____no_output_____" ] ], [ [ "## Templating\n\nComplètement hors sujet mais utile.", "_____no_output_____" ] ], [ [ "from jinja2 import Template\ntemplate = Template('Hello {{ name }}!')\ntemplate.render(name='John Doe')", "_____no_output_____" ], [ "template = Template(\"\"\"\n{{ name }}\n{{ \"-\" * len(name) }}\nPossède :\n{% for i in range(len(meubles)) %}\n- {{meubles[i]}}{% endfor %}\n\"\"\")\nmeubles = ['table', \"tabouret\"]\nprint(template.render(name='John Doe Doe', len=len,\n meubles=meubles))", "\nJohn Doe Doe\n------------\nPossède :\n\n- table\n- tabouret\n" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[]

[]

[]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# ETL Processes\nUse this notebook to develop the ETL process for each of your tables before completing the `etl.py` file to load the whole datasets.", "_____no_output_____" ] ], [ [ "import os\nimport glob\nimport psycopg2\nimport pandas as pd\nfrom sql_queries import *", "_____no_output_____" ], [ "conn = psycopg2.connect(\"host=127.0.0.1 dbname=studentdb user=student password=student\")\ncur = conn.cursor()", "_____no_output_____" ], [ "conn = psycopg2.connect(\"host=127.0.0.1 dbname=sparkifydb user=student password=student\")\ncur = conn.cursor()", "_____no_output_____" ], [ "#YD: set auto commit. pring connection \n\nconn.set_session(autocommit=True)\nprint(conn)", "<connection object at 0x11d86a048; dsn: 'user=student password=xxx dbname=sparkifydb host=127.0.0.1', closed: 0>\n" ], [ "#YD: get all files. Use os.walk()\n\ndef get_files(filepath):\n all_files = []\n for root, dirs, files in os.walk(filepath):\n files = glob.glob(os.path.join(root,'*.json'))\n for f in files :\n all_files.append(os.path.abspath(f))\n \n return all_files", "_____no_output_____" ], [ "#YD: read json file\n\ndf1 = pd.read_json('data/log_data/2018/11/2018-11-01-events.json', lines=True)", "_____no_output_____" ], [ "df1.head()", "_____no_output_____" ], [ "df2 = pd.read_json('data/song_data/A/A/A/TRAAAAW128F429D538.json', lines=True)", "_____no_output_____" ], [ "df2.head()", "_____no_output_____" ] ], [ [ "# Process `song_data`\nIn this first part, you'll perform ETL on the first dataset, `song_data`, to create the `songs` and `artists` dimensional tables.\n\nLet's perform ETL on a single song file and load a single record into each table to start.\n- Use the `get_files` function provided above to get a list of all song JSON files in `data/song_data`\n- Select the first song in this list\n- Read the song file and view the data", "_____no_output_____" ] ], [ [ "song_files = get_files('data/song_data')", "_____no_output_____" ], [ "print(song_files[0])", "/Users/yumengdong/E_Drive/13Fall/Rochester_AD_415/For SSN/0_Data Science/Data Engineer/Data Modeling/Project_data modeling with postgres/data/song_data/A/A/A/TRAAAEF128F4273421.json\n" ], [ "len(song_files)", "_____no_output_____" ], [ "filepath = song_files[0]", "_____no_output_____" ], [ "df = pd.read_json(filepath, lines=True)\ndf.head()", "_____no_output_____" ] ], [ [ "## #1: `songs` Table\n#### Extract Data for Songs Table\n- Select columns for song ID, title, artist ID, year, and duration\n- Use `df.values` to select just the values from the dataframe\n- Index to select the first (only) record in the dataframe\n- Convert the array to a list and set it to `song_data`", "_____no_output_____" ] ], [ [ "# song_id, title, artist_id, year, duration\n#YD: Pandas tolist() is used to convert a series to list.\n\nsong_data = df[['song_id','title','artist_id','year','duration']].values.tolist()[0]\nsong_data", "_____no_output_____" ], [ "song_data = df[['song_id','title','artist_id','year','duration']].values[0]\nsong_data", "_____no_output_____" ] ], [ [ "#### Insert Record into Song Table\nImplement the `song_table_insert` query in `sql_queries.py` and run the cell below to insert a record for this song into the `songs` table. Remember to run `create_tables.py` before running the cell below to ensure you've created/resetted the `songs` table in the sparkify database.", "_____no_output_____" ] ], [ [ "cur.execute(song_table_insert, song_data)\nconn.commit()", "_____no_output_____" ] ], [ [ "Run `test.ipynb` to see if you've successfully added a record to this table.", "_____no_output_____" ] ], [ [ "# cur.close()\n# conn.close()", "_____no_output_____" ] ], [ [ "## #2: `artists` Table\n#### Extract Data for Artists Table\n- Select columns for artist ID, name, location, latitude, and longitude\n- Use `df.values` to select just the values from the dataframe\n- Index to select the first (only) record in the dataframe\n- Convert the array to a list and set it to `artist_data`", "_____no_output_____" ] ], [ [ "artist_data = df[['artist_id','artist_name','artist_location','artist_latitude','artist_longitude']].values.tolist()[0]\nartist_data", "_____no_output_____" ], [ "type(artist_data)", "_____no_output_____" ], [ "artist_data = df[['artist_id','artist_name','artist_location','artist_latitude','artist_longitude']].values[0]\nartist_data", "_____no_output_____" ], [ "type(artist_data)", "_____no_output_____" ] ], [ [ "#### Insert Record into Artist Table\nImplement the `artist_table_insert` query in `sql_queries.py` and run the cell below to insert a record for this song's artist into the `artists` table. Remember to run `create_tables.py` before running the cell below to ensure you've created/resetted the `artists` table in the sparkify database.", "_____no_output_____" ] ], [ [ "cur.execute(artist_table_insert, artist_data)\nconn.commit()", "_____no_output_____" ] ], [ [ "Run `test.ipynb` to see if you've successfully added a record to this table.", "_____no_output_____" ], [ "# Process `log_data`\nIn this part, you'll perform ETL on the second dataset, `log_data`, to create the `time` and `users` dimensional tables, as well as the `songplays` fact table.\n\nLet's perform ETL on a single log file and load a single record into each table.\n- Use the `get_files` function provided above to get a list of all log JSON files in `data/log_data`\n- Select the first log file in this list\n- Read the log file and view the data", "_____no_output_____" ] ], [ [ "log_files = get_files('data/log_data')", "_____no_output_____" ], [ "filepath = log_files[0]", "_____no_output_____" ], [ "df = pd.read_json(filepath, lines=True)\ndf.head()", "_____no_output_____" ] ], [ [ "## #3: `time` Table\n#### Extract Data for Time Table\n- Filter records by `NextSong` action\n- Convert the `ts` timestamp column to datetime\n - Hint: the current timestamp is in milliseconds\n- Extract the timestamp, hour, day, week of year, month, year, and weekday from the `ts` column and set `time_data` to a list containing these values in order\n - Hint: use pandas' [`dt` attribute](https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.Series.dt.html) to access easily datetimelike properties.\n- Specify labels for these columns and set to `column_labels`\n- Create a dataframe, `time_df,` containing the time data for this file by combining `column_labels` and `time_data` into a dictionary and converting this into a dataframe", "_____no_output_____" ] ], [ [ "df = df[df['page'] == 'NextSong']\ndf.head()", "_____no_output_____" ], [ "#YD: note that to_datetime default is 'ns'\nt = pd.to_datetime(df['ts'], unit='ms')\nt.head()", "_____no_output_____" ], [ "t = df.copy()\nt['ts'] = pd.to_datetime(df['ts'], unit='ms')\nt.head()", "_____no_output_____" ], [ "time_data = t['ts'].tolist()", "_____no_output_____" ], [ "len(time_data)", "_____no_output_____" ], [ "time_data[72].year", "_____no_output_____" ], [ "time_data[72].day_name()", "_____no_output_____" ], [ "time_data[72].weekday()", "_____no_output_____" ], [ "#YD: use day_name() to return weekday name\n\ntime_data_list = (t['ts'], t['ts'].dt.hour, t['ts'].dt.day, t['ts'].dt.week, t['ts'].dt.month, t['ts'].dt.year, t['ts'].dt.day_name())", "_____no_output_____" ], [ "column_labels = ['start_time','hour','day','week','month', 'year','weekday']", "_____no_output_____" ], [ "# time_df = pd.DataFrame(time_data_list,column_labels)\n# time_df.head(10)", "_____no_output_____" ], [ "time_df = pd.DataFrame(dict(zip(column_labels, time_data_list)))\ntime_df.head(10)", "_____no_output_____" ], [ "df['ts'] = time_df['start_time']", "_____no_output_____" ], [ "# start_time = []\n# time_hour = []\n# time_day = []\n# time_week = []\n# time_month = []\n# time_year = []\n# time_weekday = []\n\n# for i in time_data:\n# start_time = i\n# time_hour.append(i.hour)\n# time_day.append(i.day)\n# time_week.append(i.week)\n# time_month.append(i.month)\n# time_year.append(i.year)\n# time_weekday.append(i.weekday())", "_____no_output_____" ], [ "# column_labels = pd.DataFrame({\n# 'start_time': start_time,\n# 'hour': time_hour\n# 'day': time_day\n# 'week': time_week\n# 'month': time_month\n# 'year': time_year\n# 'weekday': time_weekday \n# })\n# column_", "_____no_output_____" ], [ "# time_df = pd.DataFrame({\n# 'start_time': start_time,\n# 'hour': time_hour,\n# 'day': time_day,\n# 'week': time_week,\n# 'month': time_month,\n# 'year': time_year,\n# 'weekday': time_weekday \n# })\n# time_df.head()", "_____no_output_____" ] ], [ [ "#### Insert Records into Time Table\nImplement the `time_table_insert` query in `sql_queries.py` and run the cell below to insert records for the timestamps in this log file into the `time` table. Remember to run `create_tables.py` before running the cell below to ensure you've created/resetted the `time` table in the sparkify database.", "_____no_output_____" ] ], [ [ "#YD: how to insert rows one by one\n\nfor i, row in time_df.iterrows():\n cur.execute(time_table_insert, list(row))\n conn.commit()", "_____no_output_____" ], [ "# cur.close()\n# conn.close()", "_____no_output_____" ] ], [ [ "Run `test.ipynb` to see if you've successfully added records to this table.", "_____no_output_____" ], [ "## #4: `users` Table\n#### Extract Data for Users Table\n- Select columns for user ID, first name, last name, gender and level and set to `user_df`", "_____no_output_____" ] ], [ [ "user_df = df[['userId','firstName','lastName','gender','level']]\nuser_df.head()", "_____no_output_____" ] ], [ [ "#### Insert Records into Users Table\nImplement the `user_table_insert` query in `sql_queries.py` and run the cell below to insert records for the users in this log file into the `users` table. Remember to run `create_tables.py` before running the cell below to ensure you've created/resetted the `users` table in the sparkify database.", "_____no_output_____" ] ], [ [ "for i, row in user_df.iterrows():\n cur.execute(user_table_insert, row)\n conn.commit()", "_____no_output_____" ] ], [ [ "Run `test.ipynb` to see if you've successfully added records to this table.", "_____no_output_____" ], [ "## #5: `songplays` Table\n#### Extract Data and Songplays Table\nThis one is a little more complicated since information from the songs table, artists table, and original log file are all needed for the `songplays` table. Since the log file does not specify an ID for either the song or the artist, you'll need to get the song ID and artist ID by querying the songs and artists tables to find matches based on song title, artist name, and song duration time.\n- Implement the `song_select` query in `sql_queries.py` to find the song ID and artist ID based on the title, artist name, and duration of a song.\n- Select the timestamp, user ID, level, song ID, artist ID, session ID, location, and user agent and set to `songplay_data`\n\n#### Insert Records into Songplays Table\n- Implement the `songplay_table_insert` query and run the cell below to insert records for the songplay actions in this log file into the `songplays` table. Remember to run `create_tables.py` before running the cell below to ensure you've created/resetted the `songplays` table in the sparkify database.", "_____no_output_____" ] ], [ [ "df.head(1)", "_____no_output_____" ], [ "df['ts'] = df['ts']", "_____no_output_____" ], [ "songplay_table_insert", "_____no_output_____" ], [ "song_select", "_____no_output_____" ], [ "import pdb", "_____no_output_____" ], [ "for index, row in df.iterrows():\n cur.execute(song_select, (row.song,row.artist,row.length))\n results = cur.fetchone()\n \n if results:\n songid, artistid = results\n else:\n songid, artistid = None, None\n# pdb.set_trace()\n\n songplay_data = (str(row['ts']), row['userId'], row['level']\n ,songid, artistid, row['sessionId'], row['location'],\n row['userAgent'])\n \n cur.execute(songplay_table_insert, songplay_data)\n conn.commit()", "_____no_output_____" ] ], [ [ "Run `test.ipynb` to see if you've successfully added records to this table.", "_____no_output_____" ], [ "# Close Connection to Sparkify Database", "_____no_output_____" ] ], [ [ "conn.close()", "_____no_output_____" ] ], [ [ "# Implement `etl.py`\nUse what you've completed in this notebook to implement `etl.py`.", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "class Customer:\n def __init__(self, name, zip_code):\n self.name = name\n self.zip_code = zip_code", "_____no_output_____" ], [ "customer1 = Customer(\"Jane\", '90210')\ncustomer2 = Customer(\"Joe\", '29402')\ncustomer3 = Customer('Harold','39233')", "_____no_output_____" ], [ "class Complex:\n def __init__(self, real, imaginary):\n self.r = real\n self.i = imaginary\n def add(self, other_complex):\n self.r += other_complex.r\n self.i += other_complex.i\n\n", "_____no_output_____" ], [ "x = Complex(3.0, 21)\ny = Complex(0.3, 93)", "_____no_output_____" ], [ "x.add(y)", "_____no_output_____" ], [ "class Boat:\n def __init__(self, boat_name, boat_price, boat_type):\n self.name = boat_name\n self.price = boat_price\n self.type = boat_type\n\n", "_____no_output_____" ], [ "boat1 = Boat('Giligans', 2000, 'Sailboat')", "_____no_output_____" ], [ "boat1.type", "_____no_output_____" ], [ "class Carclub:\n def __init__(self, membername):\n self.name = NameError\n self.car_count = 1\n \n def add_car(self):\n self.car_count += 1\n\n def sold_car(self):\n self.car_count -= 1\nclass Cars:\n def __init__(self, brand, year, model):\n self.brand = str(brand)\n self.year = int(year)\n self.model = model\n def sound(self):\n return \"Vroom!\"\n\nclass Porsche(Cars):\n def __init__(self, year, model):\n super().__init__(year,model, model)\n self.brand = 'Porsche'\n self.year = int(year)\n self.model = model\n def sound():\n return \"Buzz\"", "_____no_output_____" ], [ "car1 = Porsche(2020, '911')", "_____no_output_____" ], [ "import Cars", "_____no_output_____" ], [ "class Rectangle:\n def __init__(self, length, width):\n self.length = length\n self.width = width\n\n def area(self):\n return self.length * self.width\n\n def perimeter(self):\n return 2 * self.length + 2 * self.width\n\nclass Square:\n def __init__(self, length):\n self.length = length\n\n def area(self):\n return self.length * self.length\n\n def perimeter(self):\n return 4 * self.length", "_____no_output_____" ], [ "shape1 = Square(4)\nshape1.area()", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "<a href=\"https://cognitiveclass.ai\"><img src = \"https://s3-api.us-geo.objectstorage.softlayer.net/cf-courses-data/CognitiveClass/Logos/organization_logo/organization_logo.png\" width = 400> </a>\n", "_____no_output_____" ], [ "This notebook is designed to run in a IBM Watson Studio default runtime (NOT the Watson Studio Apache Spark Runtime as the default runtime with 1 vCPU is free of charge). Therefore, we install Apache Spark in local mode for test purposes only. Please don't use it in production.\n\nIn case you are facing issues, please read the following two documents first:\n\n<https://github.com/IBM/skillsnetwork/wiki/Environment-Setup>\n\n<https://github.com/IBM/skillsnetwork/wiki/FAQ>\n\nThen, please feel free to ask:\n\n[https://coursera.org/learn/machine-learning-big-data-apache-spark/discussions/all](https://coursera.org/learn/machine-learning-big-data-apache-spark/discussions/all?cm_mmc=Email_Newsletter-_-Developer_Ed%2BTech-_-WW_WW-_-SkillsNetwork-Courses-IBMDeveloperSkillsNetwork-ML0201EN-SkillsNetwork-20647446&cm_mmca1=000026UJ&cm_mmca2=10006555&cm_mmca3=M12345678&cvosrc=email.Newsletter.M12345678&cvo_campaign=000026UJ)\n\nPlease make sure to follow the guidelines before asking a question:\n\n<https://github.com/IBM/skillsnetwork/wiki/FAQ#im-feeling-lost-and-confused-please-help-me>\n\nIf running outside Watson Studio, this should work as well. In case you are running in an Apache Spark context outside Watson Studio, please remove the Apache Spark setup in the first notebook cells.\n", "_____no_output_____" ] ], [ [ "from IPython.display import Markdown, display\ndef printmd(string):\n display(Markdown('# <span style=\"color:red\">'+string+'</span>'))\n\n\nif ('sc' in locals() or 'sc' in globals()):\n printmd('<<<<<!!!!! It seems that you are running in a IBM Watson Studio Apache Spark Notebook. Please run it in an IBM Watson Studio Default Runtime (without Apache Spark) !!!!!>>>>>')\n", "_____no_output_____" ], [ "!pip install pyspark==2.4.5", "Looking in indexes: https://pypi.org/simple, https://indervir.singh.banipal%40ibm.com:****@na.artifactory.swg-devops.com/artifactory/api/pypi/wcp-nlp-pypi-virtual/simple, https://indervir.singh.banipal%40ibm.com:****@na.artifactory.swg-devops.com/artifactory/api/pypi/wcp-nlp-snapshot-pypi-local/simple\r\nRequirement already satisfied: pyspark==2.4.5 in /usr/local/lib/python3.7/site-packages (2.4.5)\r\nRequirement already satisfied: py4j==0.10.7 in /usr/local/lib/python3.7/site-packages (from pyspark==2.4.5) (0.10.7)\r\n" ], [ "try:\n from pyspark import SparkContext, SparkConf\n from pyspark.sql import SparkSession\nexcept ImportError as e:\n printmd('<<<<<!!!!! Please restart your kernel after installing Apache Spark !!!!!>>>>>')", "_____no_output_____" ], [ "sc = SparkContext.getOrCreate(SparkConf().setMaster(\"local[*]\"))\n\nspark = SparkSession \\\n .builder \\\n .getOrCreate()", "_____no_output_____" ] ], [ [ "Welcome to exercise two of week three of “Apache Spark for Scalable Machine Learning on BigData”. In this exercise we’ll work on clustering.\n\nLet’s create our DataFrame again:\n", "_____no_output_____" ] ], [ [ "# delete files from previous runs\n!rm -f hmp.parquet*\n\n# download the file containing the data in PARQUET format\n!wget https://github.com/IBM/coursera/raw/master/hmp.parquet\n \n# create a dataframe out of it\ndf = spark.read.parquet('hmp.parquet')\n\n# register a corresponding query table\ndf.createOrReplaceTempView('df')", "--2020-12-09 04:39:14-- https://github.com/IBM/coursera/raw/master/hmp.parquet\nResolving github.com (github.com)... 140.82.114.4\nConnecting to github.com (github.com)|140.82.114.4|:443... connected.\nHTTP request sent, awaiting response... 301 Moved Permanently\nLocation: https://github.com/IBM/skillsnetwork/raw/master/hmp.parquet [following]\n--2020-12-09 04:39:14-- https://github.com/IBM/skillsnetwork/raw/master/hmp.parquet\nReusing existing connection to github.com:443.\nHTTP request sent, awaiting response... 302 Found\nLocation: https://raw.githubusercontent.com/IBM/skillsnetwork/master/hmp.parquet [following]\n--2020-12-09 04:39:15-- https://raw.githubusercontent.com/IBM/skillsnetwork/master/hmp.parquet\nResolving raw.githubusercontent.com (raw.githubusercontent.com)... 151.101.64.133, 151.101.128.133, 151.101.192.133, ...\nConnecting to raw.githubusercontent.com (raw.githubusercontent.com)|151.101.64.133|:443... connected.\nHTTP request sent, awaiting response... 200 OK\nLength: 932997 (911K) [application/octet-stream]\nSaving to: ‘hmp.parquet’\n\nhmp.parquet 100%[===================>] 911.13K 4.21MB/s in 0.2s \n\n2020-12-09 04:39:15 (4.21 MB/s) - ‘hmp.parquet’ saved [932997/932997]\n\n" ] ], [ [ "Let’s reuse our feature engineering pipeline.\n", "_____no_output_____" ] ], [ [ "from pyspark.ml.feature import OneHotEncoder, StringIndexer, VectorAssembler, Normalizer\nfrom pyspark.ml.linalg import Vectors\nfrom pyspark.ml import Pipeline\n\nindexer = StringIndexer(inputCol=\"class\", outputCol=\"classIndex\")\nencoder = OneHotEncoder(inputCol=\"classIndex\", outputCol=\"categoryVec\")\nvectorAssembler = VectorAssembler(inputCols=[\"x\",\"y\",\"z\"],\n outputCol=\"features\")\nnormalizer = Normalizer(inputCol=\"features\", outputCol=\"features_norm\", p=1.0)\n\npipeline = Pipeline(stages=[indexer, encoder, vectorAssembler, normalizer])\nmodel = pipeline.fit(df)\nprediction = model.transform(df)\nprediction.show()", "+---+---+---+--------------------+-----------+----------+--------------+----------------+--------------------+\n| x| y| z| source| class|classIndex| categoryVec| features| features_norm|\n+---+---+---+--------------------+-----------+----------+--------------+----------------+--------------------+\n| 22| 49| 35|Accelerometer-201...|Brush_teeth| 6.0|(13,[6],[1.0])|[22.0,49.0,35.0]|[0.20754716981132...|\n| 22| 49| 35|Accelerometer-201...|Brush_teeth| 6.0|(13,[6],[1.0])|[22.0,49.0,35.0]|[0.20754716981132...|\n| 22| 52| 35|Accelerometer-201...|Brush_teeth| 6.0|(13,[6],[1.0])|[22.0,52.0,35.0]|[0.20183486238532...|\n| 22| 52| 35|Accelerometer-201...|Brush_teeth| 6.0|(13,[6],[1.0])|[22.0,52.0,35.0]|[0.20183486238532...|\n| 21| 52| 34|Accelerometer-201...|Brush_teeth| 6.0|(13,[6],[1.0])|[21.0,52.0,34.0]|[0.19626168224299...|\n| 22| 51| 34|Accelerometer-201...|Brush_teeth| 6.0|(13,[6],[1.0])|[22.0,51.0,34.0]|[0.20560747663551...|\n| 20| 50| 35|Accelerometer-201...|Brush_teeth| 6.0|(13,[6],[1.0])|[20.0,50.0,35.0]|[0.19047619047619...|\n| 22| 52| 34|Accelerometer-201...|Brush_teeth| 6.0|(13,[6],[1.0])|[22.0,52.0,34.0]|[0.20370370370370...|\n| 22| 50| 34|Accelerometer-201...|Brush_teeth| 6.0|(13,[6],[1.0])|[22.0,50.0,34.0]|[0.20754716981132...|\n| 22| 51| 35|Accelerometer-201...|Brush_teeth| 6.0|(13,[6],[1.0])|[22.0,51.0,35.0]|[0.20370370370370...|\n| 21| 51| 33|Accelerometer-201...|Brush_teeth| 6.0|(13,[6],[1.0])|[21.0,51.0,33.0]|[0.2,0.4857142857...|\n| 20| 50| 34|Accelerometer-201...|Brush_teeth| 6.0|(13,[6],[1.0])|[20.0,50.0,34.0]|[0.19230769230769...|\n| 21| 49| 33|Accelerometer-201...|Brush_teeth| 6.0|(13,[6],[1.0])|[21.0,49.0,33.0]|[0.20388349514563...|\n| 21| 49| 33|Accelerometer-201...|Brush_teeth| 6.0|(13,[6],[1.0])|[21.0,49.0,33.0]|[0.20388349514563...|\n| 20| 51| 35|Accelerometer-201...|Brush_teeth| 6.0|(13,[6],[1.0])|[20.0,51.0,35.0]|[0.18867924528301...|\n| 18| 49| 34|Accelerometer-201...|Brush_teeth| 6.0|(13,[6],[1.0])|[18.0,49.0,34.0]|[0.17821782178217...|\n| 19| 48| 34|Accelerometer-201...|Brush_teeth| 6.0|(13,[6],[1.0])|[19.0,48.0,34.0]|[0.18811881188118...|\n| 16| 53| 34|Accelerometer-201...|Brush_teeth| 6.0|(13,[6],[1.0])|[16.0,53.0,34.0]|[0.15533980582524...|\n| 18| 52| 35|Accelerometer-201...|Brush_teeth| 6.0|(13,[6],[1.0])|[18.0,52.0,35.0]|[0.17142857142857...|\n| 18| 51| 32|Accelerometer-201...|Brush_teeth| 6.0|(13,[6],[1.0])|[18.0,51.0,32.0]|[0.17821782178217...|\n+---+---+---+--------------------+-----------+----------+--------------+----------------+--------------------+\nonly showing top 20 rows\n\n" ] ], [ [ "Now let’s create a new pipeline for kmeans.\n", "_____no_output_____" ] ], [ [ "from pyspark.ml.clustering import KMeans\nfrom pyspark.ml.evaluation import ClusteringEvaluator\n\nkmeans = KMeans(featuresCol=\"features\").setK(14).setSeed(1)\npipeline = Pipeline(stages=[vectorAssembler, kmeans])\nmodel = pipeline.fit(df)\npredictions = model.transform(df)\n\nevaluator = ClusteringEvaluator()\n\nsilhouette = evaluator.evaluate(predictions)\nprint(\"Silhouette with squared euclidean distance = \" + str(silhouette))", "Silhouette with squared euclidean distance = 0.4020618068072166\n" ] ], [ [ "We have 14 different movement patterns in the dataset, so setting K of KMeans to 14 is a good idea. But please experiment with different values for K, do you find a sweet spot? The closer Silhouette gets to 1, the better.\n\n[https://en.wikipedia.org/wiki/Silhouette\\_(clustering)](https://en.wikipedia.org/wiki/Silhouette_(clustering)?cm_mmc=Email_Newsletter-_-Developer_Ed%2BTech-_-WW_WW-_-SkillsNetwork-Courses-IBMDeveloperSkillsNetwork-ML0201EN-SkillsNetwork-20647446&cm_mmca1=000026UJ&cm_mmca2=10006555&cm_mmca3=M12345678&cvosrc=email.Newsletter.M12345678&cvo_campaign=000026UJ)\n", "_____no_output_____" ] ], [ [ "# please change the pipeline the check performance for different K, feel free to use a loop", "_____no_output_____" ] ], [ [ "Now please extend the pipeline to work on the normalized features. You need to tell KMeans to use the normalized feature column and change the pipeline in order to contain the normalizer stage as well.\n", "_____no_output_____" ] ], [ [ "kmeans = KMeans($$).setK(14).setSeed(1)\npipeline = $$\nmodel = pipeline.fit(df)\n\npredictions = model.transform(df)\n\nevaluator = ClusteringEvaluator()\n\nsilhouette = evaluator.evaluate(predictions)\nprint(\"Silhouette with squared euclidean distance = \" + str(silhouette))\n", "_____no_output_____" ] ], [ [ "Sometimes, inflating the dataset helps, here we multiply x by 10, let’s see if the performance inceases.\n", "_____no_output_____" ] ], [ [ "from pyspark.sql.functions import col\ndf_denormalized = df.select([col('*'),(col('x')*10)]).drop('x').withColumnRenamed('(x * 10)','x')", "_____no_output_____" ], [ "kmeans = KMeans(featuresCol=\"features\").setK(14).setSeed(1)\npipeline = Pipeline(stages=[vectorAssembler, kmeans])\nmodel = pipeline.fit(df_denormalized)\npredictions = model.transform(df_denormalized)\n\nevaluator = ClusteringEvaluator()\n\nsilhouette = evaluator.evaluate(predictions)\nprint(\"Silhouette with squared euclidean distance = \" + str(silhouette))", "_____no_output_____" ] ], [ [ "Apache SparkML can be used to try many different algorithms and parametrizations using the same pipeline. Please change the code below to use GaussianMixture over KMeans. Please use the following link for your reference.\n\n[https://spark.apache.org/docs/latest/ml-clustering.html#gaussian-mixture-model-gmm](https://spark.apache.org/docs/latest/ml-clustering.html#gaussian-mixture-model-gmm?cm_mmc=Email_Newsletter-_-Developer_Ed%2BTech-_-WW_WW-_-SkillsNetwork-Courses-IBMDeveloperSkillsNetwork-ML0201EN-SkillsNetwork-20647446&cm_mmca1=000026UJ&cm_mmca2=10006555&cm_mmca3=M12345678&cvosrc=email.Newsletter.M12345678&cvo_campaign=000026UJ)\n", "_____no_output_____" ] ], [ [ "from pyspark.ml.clustering import GaussianMixture\n\ngmm = $$\npipeline = $$\n\nmodel = pipeline.fit(df)\n\npredictions = model.transform(df)\n\nevaluator = ClusteringEvaluator()\n\nsilhouette = evaluator.evaluate(predictions)\nprint(\"Silhouette with squared euclidean distance = \" + str(silhouette))", "_____no_output_____" ] ], [ [ "### Thank you for completing this lab!\n\nThis notebook was created by <a href=\"https://linkedin.com/in/romeo-kienzler-089b4557\"> Romeo Kienzler </a> I hope you found this lab interesting and educational. Feel free to contact me if you have any questions!\n", "_____no_output_____" ], [ "## Change Log\n\n| Date (YYYY-MM-DD) | Version | Changed By | Change Description |\n| ----------------- | ------- | ---------- | ----------------------------------------------------------- |\n| 2020-09-29 | 2.0 | Srishti | Migrated Lab to Markdown and added to course repo in GitLab |\n\n<hr>\n\n## <h3 align=\"center\"> © IBM Corporation 2020. All rights reserved. <h3/>\n", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# NumPy Operations", "_____no_output_____" ], [ "## Arithmetic\n\nYou can easily perform array with array arithmetic, or scalar with array arithmetic. Let's see some examples:", "_____no_output_____" ] ], [ [ "import numpy as np\narr = np.arange(0,10)", "_____no_output_____" ], [ "arr + arr", "_____no_output_____" ], [ "arr * arr", "_____no_output_____" ], [ "arr - arr", "_____no_output_____" ], [ "# Warning on division by zero, but not an error!\n# Just replaced with nan\narr/arr", "/Users/marci/anaconda/lib/python3.5/site-packages/ipykernel/__main__.py:1: RuntimeWarning: invalid value encountered in true_divide\n if __name__ == '__main__':\n" ], [ "# Also warning, but not an error instead infinity\n1/arr", "/Users/marci/anaconda/lib/python3.5/site-packages/ipykernel/__main__.py:1: RuntimeWarning: divide by zero encountered in true_divide\n if __name__ == '__main__':\n" ], [ "arr**3", "_____no_output_____" ] ], [ [ "## Universal Array Functions\n\nNumpy comes with many [universal array functions](http://docs.scipy.org/doc/numpy/reference/ufuncs.html), which are essentially just mathematical operations you can use to perform the operation across the array. Let's show some common ones:", "_____no_output_____" ] ], [ [ "#Taking Square Roots\nnp.sqrt(arr)", "_____no_output_____" ], [ "#Calcualting exponential (e^)\nnp.exp(arr)", "_____no_output_____" ], [ "np.max(arr) #same as arr.max()", "_____no_output_____" ], [ "np.sin(arr)", "_____no_output_____" ], [ "np.log(arr)", "/Users/marci/anaconda/lib/python3.5/site-packages/ipykernel/__main__.py:1: RuntimeWarning: divide by zero encountered in log\n if __name__ == '__main__':\n" ] ], [ [ "# Great Job!\n\nThat's all we need to know for now!", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown" ], [ "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Copyright 2020 IITK EE604A Image Processing. All Rights Reserved.\n# \n# Licensed under the MIT License. Use and/or modification of this code outside of EE604 must reference:\n#\n# © IITK EE604A Image Processing \n# https://github.com/ee604/ee604_assignments\n#\n# Author: Shashi Kant Gupta, Chiranjeev Prachand and Prof K. S. Venkatesh, Department of Electrical Engineering, IIT Kanpur", "_____no_output_____" ] ], [ [ "# Task 2: Image Enhancement II: Spatial Smoothing\n\nIn this task, we will implement average, gaussian, and median spatial filter.", "_____no_output_____" ] ], [ [ "%%bash\npip install git+https://github.com/ee604/ee604_plugins", "_____no_output_____" ], [ "# Importing required libraries\n\nimport cv2\nimport numpy as np\nimport matplotlib.pyplot as plt\n\nfrom ee604_plugins import download_dataset, cv2_imshow\ndownload_dataset(assignment_no=2, task_no=2) # download data for this assignment", "_____no_output_____" ], [ "def avgFilter(img, kernel_size=7):\n '''\n Write a program to implement average filter. You have to assume square kernels.\n \n Inputs:\n + img - grayscaled image of size N x N\n - values between [0, 255] - 'uint8'\n + kernel_size - size of the kernel window which should be used for averaging.\n \n Ouputs:\n + out_img - smoothed grayscaled image of size N x N\n - values between [0, 255] - 'uint8'\n \n Allowed modules:\n + Basic numpy operations\n + cv2.filter2D() to perform 2D convolution\n \n Hint:\n + Not needed.\n \n '''\n \n #############################\n # Start your code from here #\n #############################\n \n # Replace with your code...\n \n #############################\n # End your code here ########\n ############################# \n \n return out_img", "_____no_output_____" ], [ "def gaussianFilter(img, kernel_size=7, sigma=3):\n '''\n Write a program to implement gaussian filter. You have to assume square kernels.\n \n Inputs:\n + img - grayscaled image of size N x N\n - values between [0, 255] - 'uint8'\n + kernel_size - size of the kernel window which should be used for smoothing. \n + sigma - sigma parameter for gaussian kernel\n \n Ouputs:\n + out_img - smoothed grayscaled image of size N x N\n - values between [0, 255] - 'uint8'\n \n Allowed modules:\n + Basic numpy operations\n + cv2.filter2D() to perform 2D convolution\n + cv2.getGaussianKernel(). Note that this will give you 1D gaussian.\n \n Hint:\n + Not needed.\n \n '''\n \n #############################\n # Start your code from here #\n #############################\n \n # Replace with your code...\n \n #############################\n # End your code here ########\n ############################# \n \n return out_img", "_____no_output_____" ], [ "def medianFilter(img, kernel_size=7):\n '''\n Write a program to implement median filter. You have to assume square kernels.\n \n Inputs:\n + img - grayscaled image of size N x N\n - values between [0, 255] - 'uint8'\n + kernel_size - size of the kernel window which should be used for smoothing.\n \n Ouputs:\n + out_img - smoothed grayscaled image of size N x N\n - values between [0, 255] - 'uint8'\n \n Allowed modules:\n + Basic numpy operations\n + np.median()\n \n Hint:\n + Not needed.\n \n '''\n \n #############################\n # Start your code from here #\n #############################\n \n # Replace with your code...\n \n #############################\n # End your code here ########\n #############################\n \n return out_img", "_____no_output_____" ] ], [ [ "### Test\n\n---\nYour observation should compare the different methods for different images. Must include a sentence on which method + kernel size worked best in each case.", "_____no_output_____" ] ], [ [ "# Do not change codes inside this cell\n# Add your observations in next to next cell\n# Your observation should compare the different methods for different images\n\nlena_orig = cv2.imread('data/lena_gray.jpg', 0)\nlena_noisy_1 = cv2.imread('data/lena_noisy_1.jpg', 0)\nlena_noisy_2 = cv2.imread('data/lena_noisy_2.jpg', 0)\nlena_noisy_3 = cv2.imread('data/lena_noisy_3.jpg', 0)\n\ndef plot_frame(gridx, gridy, subplot_id, img, name):\n plt.subplot(gridx, gridy, 1 + int(subplot_id))\n plt.imshow(np.uint8(img), cmap=\"gray\", vmin=0, vmax=255)\n plt.axis(\"off\")\n plt.title(name)", "_____no_output_____" ], [ "# Do not change codes inside this cell\n# Add your observations in next cell\n\nimg_arr = [lena_noisy_1, lena_noisy_2, lena_noisy_3]\nimg_caption = [\"Noisy 1\", \"Noisy 2\", \"Noisy 3\"]\n\nfor i in range(3):\n for kernel_size in [5, 7, 9]:\n print(\"\\n-------------------------------------\")\n print(\"# Lena\", img_caption[i], \"| kernel:\", kernel_size, \"x\", kernel_size)\n print(\"-------------------------------------\")\n\n plt.figure(figsize=(20, 13))\n plot_frame(1, 5, 0, lena_orig, \"Original\")\n plot_frame(1, 5, 1, img_arr[i], \"Noisy\")\n\n tmp_img = avgFilter(np.copy(img_arr[i]), kernel_size=kernel_size)\n plot_frame(1, 5, 2, tmp_img, \"Avg.\")\n\n tmp_img = gaussianFilter(np.copy(img_arr[i]), kernel_size=kernel_size, sigma=int(kernel_size/5))\n plot_frame(1, 5, 3, tmp_img, \"Gaussian.\")\n\n tmp_img = medianFilter(np.copy(img_arr[i]), kernel_size=kernel_size)\n plot_frame(1, 5, 4, tmp_img, \"Median.\")\n\n plt.show()", "_____no_output_____" ], [ "your_observation = \"\"\"\nReplace this with your observations.\n\"\"\"\n\nprint(your_observation)", "_____no_output_____" ], [ "# Submission >>>>>>>>>>>>>>>>>>>>>\n# Do not change codes inside this cell.\n\ngen_imgs = []\nimg_arr = [lena_noisy_1, lena_noisy_2, lena_noisy_3]\n\nfor i in range(3):\n for kernel_size in [5, 7, 9]:\n tmp_img = avgFilter(np.copy(img_arr[i]), kernel_size=kernel_size)\n gen_imgs.append(tmp_img)\n \n tmp_img = gaussianFilter(np.copy(img_arr[i]), kernel_size=kernel_size, sigma=int(kernel_size/5))\n gen_imgs.append(tmp_img)\n \n tmp_img = medianFilter(np.copy(img_arr[i]), kernel_size=kernel_size)\n gen_imgs.append(tmp_img)\n\ntask2_submission = np.array(gen_imgs)", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code" ]

[ [ "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ] ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ [ [ "# A simple notebook to do MIRI coordinate transforms. #", "_____no_output_____" ], [ "Some functionality depends on having the JWST pipeline and/or pysiaf module installed", "_____no_output_____" ] ], [ [ "import numpy as np\nimport pdb as pdb\nfrom astropy.modeling import models\nfrom asdf import AsdfFile\nfrom jwst import datamodels\nfrom jwst.assign_wcs import miri\nimport pysiaf", "_____no_output_____" ] ], [ [ "### Imager transforms using standalone code plus pysiaf ###", "_____no_output_____" ], [ "Import the miricoord standalone code:", "_____no_output_____" ] ], [ [ "import miricoord.imager.mirim_tools as mt", "_____no_output_____" ] ], [ [ "Read the MIRI apertures from the SIAF", "_____no_output_____" ] ], [ [ "siaf = pysiaf.Siaf('MIRI')#,basepath='/Users/dlaw/jwcode/pysiaf/pysiaf/pre_delivery_data/MIRI')", "_____no_output_____" ] ], [ [ "Get the MIRIM_FULL x,y reference location from the SIAF", "_____no_output_____" ] ], [ [ "xref,yref=siaf['MIRIM_FULL'].XDetRef,siaf['MIRIM_FULL'].YDetRef", "_____no_output_____" ] ], [ [ "Note that these are in the SIAF 1-indexed reference frame; in order to use them we'll first have to transform them to the 0-indexed frame used by all MIRI coordinates code (and the JWST pipeline):", "_____no_output_____" ] ], [ [ "xref,yref=xref-1,yref-1", "_____no_output_____" ], [ "xref,yref", "_____no_output_____" ] ], [ [ "Transform them to v2,v3 for filter 'F770W'", "_____no_output_____" ] ], [ [ "v2ref,v3ref=mt.xytov2v3(xref,yref,'F770W')", "_____no_output_____" ], [ "v2ref,v3ref", "_____no_output_____" ] ], [ [ "This should be showing that the v2,v3 reference point of MIRIM_FULL (for which F770W is the reference filter) is -453.559, -373.814 (note that this changed in CDP-7)", "_____no_output_____" ], [ "We can also convert a given location to RA,DEC if we assume a few JWST attitude keywords. First import the miricoord telescope tools module:", "_____no_output_____" ] ], [ [ "import miricoord.tel.tel_tools as teltools", "_____no_output_____" ] ], [ [ "Let's pretend that the telescope pointing had the reference point looking at RA=312.5, DEC=-76.0, and had spacecraft roll 73 degrees", "_____no_output_____" ] ], [ [ "raref,decref,rollref=312.5,-76.0,73.0", "_____no_output_____" ] ], [ [ "Given that, we want to know where the location v2,v3=(-400,-420) is (this is somewhere in the coronagraphs):", "_____no_output_____" ] ], [ [ "v2,v3=[-400.],[-420.]", "_____no_output_____" ], [ "ra,dec,newroll=teltools.jwst_v2v3toradec(v2,v3,v2ref=v2ref,v3ref=v3ref,raref=raref,decref=decref,rollref=rollref)", "_____no_output_____" ] ], [ [ "The RA,dec of this point is:", "_____no_output_____" ] ], [ [ "ra,dec", "_____no_output_____" ] ], [ [ "And the local roll at this new location is:", "_____no_output_____" ] ], [ [ "newroll", "_____no_output_____" ] ], [ [ "Note that if we instead had a FITS header with the appropriate keywords, we could have passed that to jwst_v2v3toradec instead of individual values.", "_____no_output_____" ], [ "### Now let's do an imager transform using the JWST pipeline code ###", "_____no_output_____" ], [ "Import the miricoord pipeline access code:", "_____no_output_____" ] ], [ [ "import miricoord.imager.mirim_pipetools as mpt", "_____no_output_____" ], [ "v2ref,v3ref=mpt.xytov2v3(xref,yref,'F770W')", "_____no_output_____" ], [ "v2ref,v3ref", "_____no_output_____" ] ], [ [ "This should be the same answer as before, but under the hood it used the JWST pipeline!", "_____no_output_____" ], [ "We can also access the pipeline distortion model directly:", "_____no_output_____" ] ], [ [ "model=mpt.xytov2v3model('F770W')", "_____no_output_____" ] ], [ [ "And use that to do forward transforms:", "_____no_output_____" ] ], [ [ "model(xref,yref)", "_____no_output_____" ] ], [ [ "And backward transforms:", "_____no_output_____" ] ], [ [ "model.inverse(v2ref,v3ref)", "_____no_output_____" ] ], [ [ "### Now do a conversion to Ideal coordinates using the SIAF apertures: ###", "_____no_output_____" ], [ "Let's work out where v2,v3=-415.069, -400.576 is for the LRS slit", "_____no_output_____" ] ], [ [ "v2,v3=-415.069, -400.576", "_____no_output_____" ], [ "xideal,yideal=mt.v2v3toIdeal(v2,v3,'MIRIM_SLIT')", "SIAF version: PRDOPSSOC-031\n" ], [ "xideal,yideal", "_____no_output_____" ] ], [ [ "It's 0,0, which makes sense since this was the MIRIM_SLIT reference point.", "_____no_output_____" ], [ "Now see what the lower-left corner of the LRS slit corresponds to in the SIAF:", "_____no_output_____" ] ], [ [ "xideal,yideal=siaf['MIRIM_SLIT'].XIdlVert1,siaf['MIRIM_SLIT'].YIdlVert1", "_____no_output_____" ], [ "v2,v3=mt.Idealtov2v3(xideal,yideal,'MIRIM_SLIT')", "SIAF version: PRDOPSSOC-031\n" ], [ "xideal,yideal,v2,v3", "_____no_output_____" ], [ "siaf['MIRIM_SLIT'].plot()", "_____no_output_____" ] ], [ [ "As another example, APT requires Ideal coordinate offsets from the reference point", "_____no_output_____" ], [ "If we wanted to see where an offset of XIdeal,YIdeal=10,0 in filter F2300C would land a star on the imager detector compared to the nominal Lyot coronagraph reference point in F770W:", "_____no_output_____" ] ], [ [ "xideal,yideal=10,0", "_____no_output_____" ], [ "v2,v3=mt.Idealtov2v3(xideal,yideal,'MIRIM_CORONLYOT')", "SIAF version: PRDOPSSOC-031\n" ], [ "x,y=mt.v2v3toxy(v2,v3,'F2300C')", "_____no_output_____" ], [ "print(x,y)", "[231.28115409] [885.2137205]\n" ], [ "siaf['MIRIM_CORONLYOT'].plot()", "_____no_output_____" ] ], [ [ "### Now we'll do an MRS transform using standalone code plus pysiaf ###", "_____no_output_____" ] ], [ [ "import miricoord.mrs.mrs_tools as mrst", "_____no_output_____" ] ], [ [ "Get the MRS v2,v3 reference point from the SIAF", "_____no_output_____" ] ], [ [ "v2ref,v3ref=siaf['MIRIFU_CHANNEL1A'].V2Ref,siaf['MIRIFU_CHANNEL1A'].V3Ref", "_____no_output_____" ], [ "v2ref,v3ref", "_____no_output_____" ], [ "siaf['MIRIFU_CHANNEL1A'].plot()", "_____no_output_____" ] ], [ [ "Figure out what alpha,beta this is in Channel 1A", "_____no_output_____" ] ], [ [ "alpha,beta=mrst.v2v3toab(v2ref,v3ref,'1A')", "_____no_output_____" ], [ "alpha,beta", "_____no_output_____" ] ], [ [ "By design, it's zero,zero since this was the reference point", "_____no_output_____" ], [ "Now find out where pixels 50,60 55,60 and 60,60 on the SHORT detector would be for Ch1A", "_____no_output_____" ] ], [ [ "x,y=[50,55,60],[60,60,60]", "_____no_output_____" ], [ "temp=mrst.xytoabl(x,y,'1A',trim=1)", "_____no_output_____" ], [ "temp", "_____no_output_____" ] ], [ [ "Note that here the return is actually a dictionary of information, and that it is only 2 elements long. This is because we specified trim=1, which will remove any values that do not correspond to a light-sensitive slice.", "_____no_output_____" ] ], [ [ "v2,v3=mrst.abtov2v3(temp['alpha'],temp['beta'],'1A')", "_____no_output_____" ], [ "v2,v3", "_____no_output_____" ] ], [ [ "### Now we'll do an MRS transform using the pipeline code ###", "_____no_output_____" ] ], [ [ "import miricoord.mrs.mrs_pipetools as mrspt", "_____no_output_____" ], [ "x,y=30.31,511.0", "_____no_output_____" ], [ "a,b,l=mrspt.xytoabl(x,y,'1A')", "_____no_output_____" ], [ "print(a,b,l)", "0.4612954052724527 0.0 5.343680519509376\n" ] ], [ [ "Be warned: using the pipeline code in this way can give strange results if you try to transform a pixel that doesn't land on a slice in your channel specified!! (The pipeline itself has code elsewhere to deal with this, but here we're hooking directly into the transform modules).", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import os\nos.environ['CUDA_VISIBLE_DEVICES']='4,5,6,7'", "_____no_output_____" ], [ "%run -p ../train_cnf_disentangle_rl_2cond_multiscale_beta.py --data colormnist --dims 64,64,64 --strides 1,1,1,1 --num_blocks 2 --layer_type concat --multiscale True --rademacher True --batch_size 900 --test_batch_size 500 --save ../experiments_published/infocnf_conditional_disentangle_colormnist_bs900_sratio_1_4th_drop_0_5_rl_stdscale_6_2cond_linear_multiscale_beta_10_run1 --seed 1 --lr 0.001 --conditional True --controlled_tol False --train_mode semisup --log_freq 10 --weight_y 0.5 --condition_ratio 0.25 --dropout_rate 0.5 --scale_fac 1.0 --scale_std 6.0 --cond_nn linear --y_color 10 --y_class 10 --beta 10.0\n#", "/tancode/repos/tan-ffjord/train_cnf_disentangle_rl_2cond_multiscale_beta.py\nfrom __future__ import print_function\n\nimport argparse\nimport os\nimport time\nimport numpy as np\n\nimport torch\nimport torch.optim as optim\nimport torchvision.datasets as dset\nimport torchvision.transforms as tforms\nfrom torchvision.utils import save_image\n\nimport torch.utils.data as data\nfrom torch.utils.data import Dataset\n\nfrom PIL import Image\nimport os.path\nimport errno\nimport codecs\n\nimport lib.layers as layers\nimport lib.utils as utils\nimport lib.multiscale_parallel as multiscale_parallel\nimport lib.modules as modules\nimport lib.thops as thops\n\nfrom train_misc import standard_normal_logprob\nfrom train_misc import set_cnf_options, count_nfe, count_parameters, count_total_time, count_nfe_gate\nfrom train_misc import add_spectral_norm, spectral_norm_power_iteration\nfrom train_misc import create_regularization_fns, get_regularization, append_regularization_to_log\n\nfrom tensorboardX import SummaryWriter\n\n# go fast boi!!\ntorch.backends.cudnn.benchmark = True\nSOLVERS = [\"dopri5\", \"bdf\", \"rk4\", \"midpoint\", 'adams', 'explicit_adams']\nGATES = [\"cnn1\", \"cnn2\", \"rnn\"]\n\nparser = argparse.ArgumentParser(\"Continuous Normalizing Flow\")\nparser.add_argument(\"--data\", choices=[\"colormnist\", \"mnist\", \"svhn\", \"cifar10\", 'lsun_church'], type=str, default=\"mnist\")\nparser.add_argument(\"--dims\", type=str, default=\"8,32,32,8\")\nparser.add_argument(\"--strides\", type=str, default=\"2,2,1,-2,-2\")\nparser.add_argument(\"--num_blocks\", type=int, default=1, help='Number of stacked CNFs.')\n\nparser.add_argument(\"--conv\", type=eval, default=True, choices=[True, False])\nparser.add_argument(\n \"--layer_type\", type=str, default=\"ignore\",\n choices=[\"ignore\", \"concat\", \"concat_v2\", \"squash\", \"concatsquash\", \"concatcoord\", \"hyper\", \"blend\"]\n)\nparser.add_argument(\"--divergence_fn\", type=str, default=\"approximate\", choices=[\"brute_force\", \"approximate\"])\nparser.add_argument(\n \"--nonlinearity\", type=str, default=\"softplus\", choices=[\"tanh\", \"relu\", \"softplus\", \"elu\", \"swish\"]\n)\n\nparser.add_argument(\"--seed\", type=int, default=0)\n\nparser.add_argument('--solver', type=str, default='dopri5', choices=SOLVERS)\nparser.add_argument('--atol', type=float, default=1e-5)\nparser.add_argument('--rtol', type=float, default=1e-5)\nparser.add_argument(\"--step_size\", type=float, default=None, help=\"Optional fixed step size.\")\n\nparser.add_argument('--gate', type=str, default='cnn1', choices=GATES)\nparser.add_argument('--scale', type=float, default=1.0)\nparser.add_argument('--scale_fac', type=float, default=1.0)\nparser.add_argument('--scale_std', type=float, default=1.0)\nparser.add_argument('--eta', default=0.1, type=float,\n help='tuning parameter that allows us to trade-off the competing goals of' \n 'minimizing the prediction loss and maximizing the gate rewards ')\nparser.add_argument('--rl-weight', default=0.01, type=float,\n help='rl weight')\n\nparser.add_argument('--gamma', default=0.99, type=float,\n help='discount factor, default: (0.99)')\n\nparser.add_argument('--test_solver', type=str, default=None, choices=SOLVERS + [None])\nparser.add_argument('--test_atol', type=float, default=None)\nparser.add_argument('--test_rtol', type=float, default=None)\n\nparser.add_argument(\"--imagesize\", type=int, default=None)\nparser.add_argument(\"--alpha\", type=float, default=1e-6)\nparser.add_argument('--time_length', type=float, default=1.0)\nparser.add_argument('--train_T', type=eval, default=True)\n\nparser.add_argument(\"--num_epochs\", type=int, default=500)\nparser.add_argument(\"--batch_size\", type=int, default=200)\nparser.add_argument(\n \"--batch_size_schedule\", type=str, default=\"\", help=\"Increases the batchsize at every given epoch, dash separated.\"\n)\nparser.add_argument(\"--test_batch_size\", type=int, default=200)\nparser.add_argument(\"--lr\", type=float, default=1e-3)\nparser.add_argument(\"--warmup_iters\", type=float, default=1000)\nparser.add_argument(\"--weight_decay\", type=float, default=0.0)\nparser.add_argument(\"--spectral_norm_niter\", type=int, default=10)\nparser.add_argument(\"--weight_y\", type=float, default=0.5)\nparser.add_argument(\"--annealing_std\", type=eval, default=False, choices=[True, False])\nparser.add_argument(\"--y_class\", type=int, default=10)\nparser.add_argument(\"--y_color\", type=int, default=10)\n\nparser.add_argument(\"--add_noise\", type=eval, default=True, choices=[True, False])\nparser.add_argument(\"--batch_norm\", type=eval, default=False, choices=[True, False])\nparser.add_argument('--residual', type=eval, default=False, choices=[True, False])\nparser.add_argument('--autoencode', type=eval, default=False, choices=[True, False])\nparser.add_argument('--rademacher', type=eval, default=True, choices=[True, False])\nparser.add_argument('--spectral_norm', type=eval, default=False, choices=[True, False])\nparser.add_argument('--multiscale', type=eval, default=False, choices=[True, False])\nparser.add_argument('--parallel', type=eval, default=False, choices=[True, False])\nparser.add_argument('--conditional', type=eval, default=False, choices=[True, False])\nparser.add_argument('--controlled_tol', type=eval, default=False, choices=[True, False])\nparser.add_argument(\"--train_mode\", choices=[\"semisup\", \"sup\", \"unsup\"], type=str, default=\"semisup\")\nparser.add_argument(\"--condition_ratio\", type=float, default=0.5)\nparser.add_argument(\"--dropout_rate\", type=float, default=0.0)\nparser.add_argument(\"--cond_nn\", choices=[\"linear\", \"mlp\"], type=str, default=\"linear\")\n\n\n# Regularizations\nparser.add_argument('--l1int', type=float, default=None, help=\"int_t ||f||_1\")\nparser.add_argument('--l2int', type=float, default=None, help=\"int_t ||f||_2\")\nparser.add_argument('--dl2int', type=float, default=None, help=\"int_t ||f^T df/dt||_2\")\nparser.add_argument('--JFrobint', type=float, default=None, help=\"int_t ||df/dx||_F\")\nparser.add_argument('--JdiagFrobint', type=float, default=None, help=\"int_t ||df_i/dx_i||_F\")\nparser.add_argument('--JoffdiagFrobint', type=float, default=None, help=\"int_t ||df/dx - df_i/dx_i||_F\")\n\nparser.add_argument(\"--time_penalty\", type=float, default=0, help=\"Regularization on the end_time.\")\nparser.add_argument(\n \"--max_grad_norm\", type=float, default=1e10,\n help=\"Max norm of graidents (default is just stupidly high to avoid any clipping)\"\n)\n\nparser.add_argument(\"--begin_epoch\", type=int, default=1)\nparser.add_argument(\"--resume\", type=str, default=None)\nparser.add_argument(\"--save\", type=str, default=\"experiments/cnf\")\nparser.add_argument(\"--val_freq\", type=int, default=1)\nparser.add_argument(\"--log_freq\", type=int, default=1)\n\n# for disentanglement\nparser.add_argument('--beta', default=0.01, type=float, help='disentanglement weight')\n\nargs = parser.parse_args()\n\nimport lib.odenvp_conditional_rl_2cond_multiscale as odenvp\n \n# set seed\ntorch.manual_seed(args.seed)\nnp.random.seed(args.seed)\n\n# logger\nutils.makedirs(args.save)\nlogger = utils.get_logger(logpath=os.path.join(args.save, 'logs'), filepath=os.path.abspath(__file__)) # write to log file\nwriter = SummaryWriter(os.path.join(args.save, 'tensorboard')) # write to tensorboard\n\nif args.layer_type == \"blend\":\n logger.info(\"!! Setting time_length from None to 1.0 due to use of Blend layers.\")\n args.time_length = 1.0\n\nlogger.info(args)\n\nclass ColorMNIST(data.Dataset):\n \"\"\"\n ColorMNIST\n \"\"\"\n urls = [\n 'http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz',\n 'http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz',\n 'http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz',\n 'http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz',\n ]\n raw_folder = 'raw'\n processed_folder = 'processed'\n training_file = 'training.pt'\n test_file = 'test.pt'\n\n def __init__(self, root, train=True, transform=None, target_transform=None, download=False):\n self.root = os.path.expanduser(root)\n self.transform = transform\n self.target_transform = target_transform\n self.train = train # training set or test set\n\n if download:\n self.download()\n\n if not self._check_exists():\n raise RuntimeError('Dataset not found.' +\n ' You can use download=True to download it')\n\n if self.train:\n self.train_data, self.train_labels = torch.load(\n os.path.join(self.root, self.processed_folder, self.training_file))\n \n self.train_data = np.tile(self.train_data[:, :, :, np.newaxis], 3)\n else:\n self.test_data, self.test_labels = torch.load(\n os.path.join(self.root, self.processed_folder, self.test_file))\n \n self.test_data = np.tile(self.test_data[:, :, :, np.newaxis], 3)\n \n self.pallette = [[31, 119, 180],\n [255, 127, 14],\n [44, 160, 44],\n [214, 39, 40],\n [148, 103, 189],\n [140, 86, 75],\n [227, 119, 194],\n [127, 127, 127],\n [188, 189, 34],\n [23, 190, 207]]\n\n def __getitem__(self, index):\n \"\"\"\n Args:\n index (int): Index\n\n Returns:\n tuple: (image, target) where target is index of the target class.\n \"\"\"\n if self.train:\n img, target = self.train_data[index].copy(), self.train_labels[index]\n else:\n img, target = self.test_data[index].copy(), self.test_labels[index]\n \n # doing this so that it is consistent with all other datasets\n # to return a PIL Image\n y_color_digit = np.random.randint(0, args.y_color)\n c_digit = self.pallette[y_color_digit]\n \n img[:, :, 0] = img[:, :, 0] / 255 * c_digit[0]\n img[:, :, 1] = img[:, :, 1] / 255 * c_digit[1]\n img[:, :, 2] = img[:, :, 2] / 255 * c_digit[2]\n \n img = Image.fromarray(img)\n\n if self.transform is not None:\n img = self.transform(img)\n\n if self.target_transform is not None:\n target = self.target_transform(target)\n\n return img, [target,torch.from_numpy(np.array(y_color_digit))]\n\n def __len__(self):\n if self.train:\n return len(self.train_data)\n else:\n return len(self.test_data)\n\n def _check_exists(self):\n return os.path.exists(os.path.join(self.root, self.processed_folder, self.training_file)) and \\\n os.path.exists(os.path.join(self.root, self.processed_folder, self.test_file))\n\n def download(self):\n \"\"\"Download the MNIST data if it doesn't exist in processed_folder already.\"\"\"\n from six.moves import urllib\n import gzip\n\n if self._check_exists():\n return\n\n # download files\n try:\n os.makedirs(os.path.join(self.root, self.raw_folder))\n os.makedirs(os.path.join(self.root, self.processed_folder))\n except OSError as e:\n if e.errno == errno.EEXIST:\n pass\n else:\n raise\n\n for url in self.urls:\n print('Downloading ' + url)\n data = urllib.request.urlopen(url)\n filename = url.rpartition('/')[2]\n file_path = os.path.join(self.root, self.raw_folder, filename)\n with open(file_path, 'wb') as f:\n f.write(data.read())\n with open(file_path.replace('.gz', ''), 'wb') as out_f, \\\n gzip.GzipFile(file_path) as zip_f:\n out_f.write(zip_f.read())\n os.unlink(file_path)\n\n # process and save as torch files\n print('Processing...')\n\n training_set = (\n read_image_file(os.path.join(self.root, self.raw_folder, 'train-images-idx3-ubyte')),\n read_label_file(os.path.join(self.root, self.raw_folder, 'train-labels-idx1-ubyte'))\n )\n test_set = (\n read_image_file(os.path.join(self.root, self.raw_folder, 't10k-images-idx3-ubyte')),\n read_label_file(os.path.join(self.root, self.raw_folder, 't10k-labels-idx1-ubyte'))\n )\n with open(os.path.join(self.root, self.processed_folder, self.training_file), 'wb') as f:\n torch.save(training_set, f)\n with open(os.path.join(self.root, self.processed_folder, self.test_file), 'wb') as f:\n torch.save(test_set, f)\n\n print('Done!')\n\n def __repr__(self):\n fmt_str = 'Dataset ' + self.__class__.__name__ + '\\n'\n fmt_str += ' Number of datapoints: {}\\n'.format(self.__len__())\n tmp = 'train' if self.train is True else 'test'\n fmt_str += ' Split: {}\\n'.format(tmp)\n fmt_str += ' Root Location: {}\\n'.format(self.root)\n tmp = ' Transforms (if any): '\n fmt_str += '{0}{1}\\n'.format(tmp, self.transform.__repr__().replace('\\n', '\\n' + ' ' * len(tmp)))\n tmp = ' Target Transforms (if any): '\n fmt_str += '{0}{1}'.format(tmp, self.target_transform.__repr__().replace('\\n', '\\n' + ' ' * len(tmp)))\n return fmt_str\n\n\ndef add_noise(x):\n \"\"\"\n [0, 1] -> [0, 255] -> add noise -> [0, 1]\n \"\"\"\n if args.add_noise:\n noise = x.new().resize_as_(x).uniform_()\n x = x * 255 + noise\n x = x / 256\n return x\n\n\ndef update_lr(optimizer, itr):\n iter_frac = min(float(itr + 1) / max(args.warmup_iters, 1), 1.0)\n lr = args.lr * iter_frac\n for param_group in optimizer.param_groups:\n param_group[\"lr\"] = lr\n\n \ndef update_scale_std(model, epoch):\n epoch_frac = 1.0 - float(epoch - 1) / max(args.num_epochs + 1, 1)\n scale_std = args.scale_std * epoch_frac\n model.set_scale_std(scale_std)\n\n\ndef get_train_loader(train_set, epoch):\n if args.batch_size_schedule != \"\":\n epochs = [0] + list(map(int, args.batch_size_schedule.split(\"-\")))\n n_passed = sum(np.array(epochs) <= epoch)\n current_batch_size = int(args.batch_size * n_passed)\n else:\n current_batch_size = args.batch_size\n train_loader = torch.utils.data.DataLoader(\n dataset=train_set, batch_size=current_batch_size, shuffle=True, drop_last=True, pin_memory=True\n )\n logger.info(\"===> Using batch size {}. Total {} iterations/epoch.\".format(current_batch_size, len(train_loader)))\n return train_loader\n\n\ndef get_dataset(args):\n trans = lambda im_size: tforms.Compose([tforms.Resize(im_size), tforms.ToTensor(), add_noise])\n\n if args.data == \"mnist\":\n im_dim = 1\n im_size = 28 if args.imagesize is None else args.imagesize\n train_set = dset.MNIST(root=\"../data\", train=True, transform=trans(im_size), download=True)\n test_set = dset.MNIST(root=\"../data\", train=False, transform=trans(im_size), download=True)\n if args.data == \"colormnist\":\n im_dim = 3\n im_size = 28 if args.imagesize is None else args.imagesize\n train_set = ColorMNIST(root=\"../data\", train=True, transform=trans(im_size), download=True)\n test_set = ColorMNIST(root=\"../data\", train=False, transform=trans(im_size), download=True)\n elif args.data == \"svhn\":\n im_dim = 3\n im_size = 32 if args.imagesize is None else args.imagesize\n train_set = dset.SVHN(root=\"../data\", split=\"train\", transform=trans(im_size), download=True)\n test_set = dset.SVHN(root=\"../data\", split=\"test\", transform=trans(im_size), download=True)\n elif args.data == \"cifar10\":\n im_dim = 3\n im_size = 32 if args.imagesize is None else args.imagesize\n train_set = dset.CIFAR10(\n root=\"../data\", train=True, transform=tforms.Compose([\n tforms.Resize(im_size),\n tforms.RandomHorizontalFlip(),\n tforms.ToTensor(),\n add_noise,\n ]), download=True\n )\n test_set = dset.CIFAR10(root=\"../data\", train=False, transform=trans(im_size), download=True)\n elif args.data == 'celeba':\n im_dim = 3\n im_size = 64 if args.imagesize is None else args.imagesize\n train_set = dset.CelebA(\n train=True, transform=tforms.Compose([\n tforms.ToPILImage(),\n tforms.Resize(im_size),\n tforms.RandomHorizontalFlip(),\n tforms.ToTensor(),\n add_noise,\n ])\n )\n test_set = dset.CelebA(\n train=False, transform=tforms.Compose([\n tforms.ToPILImage(),\n tforms.Resize(im_size),\n tforms.ToTensor(),\n add_noise,\n ])\n )\n elif args.data == 'lsun_church':\n im_dim = 3\n im_size = 64 if args.imagesize is None else args.imagesize\n train_set = dset.LSUN(\n '../data', ['church_outdoor_train'], transform=tforms.Compose([\n tforms.Resize(96),\n tforms.RandomCrop(64),\n tforms.Resize(im_size),\n tforms.ToTensor(),\n add_noise,\n ])\n )\n test_set = dset.LSUN(\n '../data', ['church_outdoor_val'], transform=tforms.Compose([\n tforms.Resize(96),\n tforms.RandomCrop(64),\n tforms.Resize(im_size),\n tforms.ToTensor(),\n add_noise,\n ])\n ) \n elif args.data == 'imagenet_64':\n im_dim = 3\n im_size = 64 if args.imagesize is None else args.imagesize\n train_set = dset.ImageFolder(\n train=True, transform=tforms.Compose([\n tforms.ToPILImage(),\n tforms.Resize(im_size),\n tforms.RandomHorizontalFlip(),\n tforms.ToTensor(),\n add_noise,\n ])\n )\n test_set = dset.ImageFolder(\n train=False, transform=tforms.Compose([\n tforms.ToPILImage(),\n tforms.Resize(im_size),\n tforms.ToTensor(),\n add_noise,\n ])\n )\n \n data_shape = (im_dim, im_size, im_size)\n if not args.conv:\n data_shape = (im_dim * im_size * im_size,)\n\n test_loader = torch.utils.data.DataLoader(\n dataset=test_set, batch_size=args.test_batch_size, shuffle=False, drop_last=True\n )\n return train_set, test_loader, data_shape\n\n\ndef compute_bits_per_dim(x, model):\n zero = torch.zeros(x.shape[0], 1).to(x)\n\n # Don't use data parallelize if batch size is small.\n # if x.shape[0] < 200:\n # model = model.module\n \n z, delta_logp, atol, rtol, logp_actions, nfe = model(x, zero) # run model forward\n\n logpz = standard_normal_logprob(z).view(z.shape[0], -1).sum(1, keepdim=True) # logp(z)\n logpx = logpz - delta_logp\n\n logpx_per_dim = torch.sum(logpx) / x.nelement() # averaged over batches\n bits_per_dim = -(logpx_per_dim - np.log(256)) / np.log(2)\n\n return bits_per_dim, atol, rtol, logp_actions, nfe\n\ndef compute_bits_per_dim_conditional(x, y, y_color, model):\n zero = torch.zeros(x.shape[0], 1).to(x)\n y_onehot = thops.onehot(y, num_classes=model.module.y_class).to(x)\n y_onehot_color = thops.onehot(y_color, num_classes=model.module.y_color).to(x)\n\n # Don't use data parallelize if batch size is small.\n # if x.shape[0] < 200:\n # model = model.module\n \n z, z_unsup, delta_logp, atol, rtol, logp_actions, nfe = model(x, zero) # run model forward\n \n z_unsup = torch.cat(z_unsup, 1)\n \n z_sup_class = [o[:,:int(np.prod(o.size()[1:])*0.5)] for o in z]\n z_sup_class = torch.cat(z_sup_class,1)\n \n z_sup_color = [o[:,int(np.prod(o.size()[1:])*0.5):] for o in z]\n z_sup_color = torch.cat(z_sup_color,1)\n \n # prior\n mean, logs = model.module._prior(y_onehot)\n mean_color, logs_color = model.module._prior_color(y_onehot_color)\n\n logpz_sup = modules.GaussianDiag.logp(mean, logs, z_sup_class).view(-1,1) # logp(z)_sup\n beta_logpz_sup = logpz_sup * (1.0 - args.beta * torch.exp(logpz_sup) / torch.tensor(model.module.y_class).to(logpz_sup))\n \n logpz_color_sup = modules.GaussianDiag.logp(mean_color, logs_color, z_sup_color).view(-1,1) # logp(z)_color_sup\n beta_logpz_color_sup = logpz_color_sup * (1.0 - args.beta * torch.exp(logpz_color_sup) / torch.tensor(model.module.y_color).to(logpz_color_sup))\n \n logpz_unsup = standard_normal_logprob(z_unsup).view(z_unsup.shape[0], -1).sum(1, keepdim=True)\n logpz = beta_logpz_sup + beta_logpz_color_sup + logpz_unsup\n logpx = logpz - delta_logp\n\n logpx_per_dim = torch.sum(logpx) / x.nelement() # averaged over batches\n bits_per_dim = -(logpx_per_dim - np.log(256)) / np.log(2)\n \n # dropout\n if args.dropout_rate > 0:\n z_sup_class = model.module.dropout(z_sup_class)\n z_sup_color = model.module.dropout_color(z_sup_color)\n \n # compute xentropy loss\n y_logits = model.module.project_class(z_sup_class)\n loss_xent = model.module.loss_class(y_logits, y.to(x.get_device()))\n y_predicted = np.argmax(y_logits.cpu().detach().numpy(), axis=1)\n \n y_logits_color = model.module.project_color(z_sup_color)\n loss_xent_color = model.module.loss_class(y_logits_color, y_color.to(x.get_device()))\n y_color_predicted = np.argmax(y_logits_color.cpu().detach().numpy(), axis=1)\n\n return bits_per_dim, loss_xent, loss_xent_color, y_predicted, y_color_predicted, atol, rtol, logp_actions, nfe\n\ndef create_model(args, data_shape, regularization_fns):\n hidden_dims = tuple(map(int, args.dims.split(\",\")))\n strides = tuple(map(int, args.strides.split(\",\")))\n\n if args.multiscale:\n model = odenvp.ODENVP(\n (args.batch_size, *data_shape),\n n_blocks=args.num_blocks,\n intermediate_dims=hidden_dims,\n nonlinearity=args.nonlinearity,\n alpha=args.alpha,\n cnf_kwargs={\"T\": args.time_length, \"train_T\": args.train_T, \"regularization_fns\": regularization_fns, \"solver\": args.solver, \"atol\": args.atol, \"rtol\": args.rtol, \"scale\": args.scale, \"scale_fac\": args.scale_fac, \"scale_std\": args.scale_std, \"gate\": args.gate},\n condition_ratio=args.condition_ratio * 2.,\n dropout_rate=args.dropout_rate,\n cond_nn=args.cond_nn,\n y_class = args.y_class,\n y_color = args.y_color)\n elif args.parallel:\n model = multiscale_parallel.MultiscaleParallelCNF(\n (args.batch_size, *data_shape),\n n_blocks=args.num_blocks,\n intermediate_dims=hidden_dims,\n alpha=args.alpha,\n time_length=args.time_length,\n )\n else:\n if args.autoencode:\n\n def build_cnf():\n autoencoder_diffeq = layers.AutoencoderDiffEqNet(\n hidden_dims=hidden_dims,\n input_shape=data_shape,\n strides=strides,\n conv=args.conv,\n layer_type=args.layer_type,\n nonlinearity=args.nonlinearity,\n )\n odefunc = layers.AutoencoderODEfunc(\n autoencoder_diffeq=autoencoder_diffeq,\n divergence_fn=args.divergence_fn,\n residual=args.residual,\n rademacher=args.rademacher,\n )\n cnf = layers.CNF(\n odefunc=odefunc,\n T=args.time_length,\n regularization_fns=regularization_fns,\n solver=args.solver,\n )\n return cnf\n else:\n\n def build_cnf():\n diffeq = layers.ODEnet(\n hidden_dims=hidden_dims,\n input_shape=data_shape,\n strides=strides,\n conv=args.conv,\n layer_type=args.layer_type,\n nonlinearity=args.nonlinearity,\n )\n odefunc = layers.ODEfunc(\n diffeq=diffeq,\n divergence_fn=args.divergence_fn,\n residual=args.residual,\n rademacher=args.rademacher,\n )\n cnf = layers.CNF(\n odefunc=odefunc,\n T=args.time_length,\n train_T=args.train_T,\n regularization_fns=regularization_fns,\n solver=args.solver,\n )\n return cnf\n\n chain = [layers.LogitTransform(alpha=args.alpha)] if args.alpha > 0 else [layers.ZeroMeanTransform()]\n chain = chain + [build_cnf() for _ in range(args.num_blocks)]\n if args.batch_norm:\n chain.append(layers.MovingBatchNorm2d(data_shape[0]))\n model = layers.SequentialFlow(chain)\n return model\n\n\nif __name__ == \"__main__\":\n\n # get deivce\n device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")\n cvt = lambda x: x.type(torch.float32).to(device, non_blocking=True)\n\n # load dataset\n train_set, test_loader, data_shape = get_dataset(args)\n\n # build model\n regularization_fns, regularization_coeffs = create_regularization_fns(args)\n model = create_model(args, data_shape, regularization_fns)\n\n if args.spectral_norm: add_spectral_norm(model, logger)\n set_cnf_options(args, model)\n\n logger.info(model)\n logger.info(\"Number of trainable parameters: {}\".format(count_parameters(model)))\n \n writer.add_text('info', \"Number of trainable parameters: {}\".format(count_parameters(model)))\n\n # optimizer\n optimizer = optim.Adam(model.parameters(), lr=args.lr, weight_decay=args.weight_decay)\n \n # set initial iter\n itr = 1\n \n # set the meters\n time_epoch_meter = utils.RunningAverageMeter(0.97)\n time_meter = utils.RunningAverageMeter(0.97)\n loss_meter = utils.RunningAverageMeter(0.97) # track total loss\n nll_meter = utils.RunningAverageMeter(0.97) # track negative log-likelihood\n xent_meter = utils.RunningAverageMeter(0.97) # track xentropy score\n xent_color_meter = utils.RunningAverageMeter(0.97) # track xentropy score\n error_meter = utils.RunningAverageMeter(0.97) # track error score\n error_color_meter = utils.RunningAverageMeter(0.97)\n steps_meter = utils.RunningAverageMeter(0.97)\n grad_meter = utils.RunningAverageMeter(0.97)\n tt_meter = utils.RunningAverageMeter(0.97)\n\n # restore parameters\n if args.resume is not None:\n checkpt = torch.load(args.resume, map_location=lambda storage, loc: storage)\n model.load_state_dict(checkpt[\"state_dict\"])\n if \"optim_state_dict\" in checkpt.keys():\n optimizer.load_state_dict(checkpt[\"optim_state_dict\"])\n # Manually move optimizer state to device.\n for state in optimizer.state.values():\n for k, v in state.items():\n if torch.is_tensor(v):\n state[k] = cvt(v)\n args.begin_epoch = checkpt['epoch'] + 1\n itr = checkpt['iter'] + 1\n time_epoch_meter.set(checkpt['epoch_time_avg'])\n time_meter.set(checkpt['time_train'])\n loss_meter.set(checkpt['loss_train'])\n nll_meter.set(checkpt['bits_per_dim_train'])\n xent_meter.set(checkpt['xent_train'])\n xent_color_meter.set(checkpt['xent_train_color'])\n error_meter.set(checkpt['error_train'])\n error_color_meter.set(checkpt['error_train_color'])\n steps_meter.set(checkpt['nfe_train'])\n grad_meter.set(checkpt['grad_train'])\n tt_meter.set(checkpt['total_time_train'])\n\n if torch.cuda.is_available():\n model = torch.nn.DataParallel(model).cuda()\n\n # For visualization.\n if args.conditional:\n fixed_y = torch.from_numpy(np.arange(model.module.y_class)).repeat(model.module.y_class).type(torch.long).to(device, non_blocking=True)\n fixed_y_onehot = thops.onehot(fixed_y, num_classes=model.module.y_class)\n \n fixed_y_color = torch.from_numpy(np.arange(model.module.y_color)).repeat(model.module.y_color).type(torch.long).to(device, non_blocking=True)\n fixed_y_onehot_color = thops.onehot(fixed_y_color, num_classes=model.module.y_color)\n with torch.no_grad():\n mean, logs = model.module._prior(fixed_y_onehot)\n mean_color, logs_color = model.module._prior_color(fixed_y_onehot_color)\n fixed_z_sup = modules.GaussianDiag.sample(mean, logs)\n fixed_z_color_sup = modules.GaussianDiag.sample(mean_color, logs_color)\n dim_unsup = np.prod(data_shape) - np.prod(fixed_z_sup.shape[1:]) - np.prod(fixed_z_color_sup.shape[1:])\n fixed_z_unsup = cvt(torch.randn(model.module.y_class**2, dim_unsup))\n \n a_sup = fixed_z_sup.shape[1] // (2**(model.module.n_scale - 1))\n a_color_sup = fixed_z_color_sup.shape[1] // (2**(model.module.n_scale - 1))\n a_unsup = fixed_z_unsup.shape[1] // (2**(model.module.n_scale - 1))\n \n fixed_z = []\n start_sup = 0; start_color_sup = 0; start_unsup = 0\n for ns in range(model.module.n_scale, 1, -1):\n end_sup = start_sup + (2**(ns-2))*a_sup\n end_color_sup = start_color_sup + (2**(ns-2))*a_color_sup\n end_unsup = start_unsup + (2**(ns-2))*a_unsup\n \n fixed_z.append(fixed_z_sup[:,start_sup:end_sup])\n fixed_z.append(fixed_z_color_sup[:,start_color_sup:end_color_sup])\n fixed_z.append(fixed_z_unsup[:,start_unsup:end_unsup])\n \n start_sup = end_sup; start_color_sup = end_color_sup; start_unsup = end_unsup\n \n end_sup = start_sup + a_sup\n end_color_sup = start_color_sup + a_color_sup\n end_unsup = start_unsup + a_unsup\n \n fixed_z.append(fixed_z_sup[:,start_sup:end_sup])\n fixed_z.append(fixed_z_color_sup[:,start_color_sup:end_color_sup])\n fixed_z.append(fixed_z_unsup[:,start_unsup:end_unsup])\n \n # for i_z in range(len(fixed_z)): print(fixed_z[i_z].shape)\n \n fixed_z = torch.cat(fixed_z,1)\n else:\n fixed_z = cvt(torch.randn(100, *data_shape))\n \n\n if args.spectral_norm and not args.resume: spectral_norm_power_iteration(model, 500)\n\n best_loss_nll = float(\"inf\")\n best_error_score = float(\"inf\")\n best_error_score_color = float(\"inf\")\n \n for epoch in range(args.begin_epoch, args.num_epochs + 1):\n start_epoch = time.time()\n model.train()\n if args.annealing_std:\n update_scale_std(model.module, epoch)\n \n train_loader = get_train_loader(train_set, epoch)\n for _, (x, y_all) in enumerate(train_loader):\n start = time.time()\n \n y = y_all[0]\n y_color = y_all[1]\n \n update_lr(optimizer, itr)\n optimizer.zero_grad()\n\n if not args.conv:\n x = x.view(x.shape[0], -1)\n\n # cast data and move to device\n x = cvt(x)\n \n # compute loss\n if args.conditional:\n loss_nll, loss_xent, loss_xent_color, y_predicted, y_color_predicted, atol, rtol, logp_actions, nfe = compute_bits_per_dim_conditional(x, y, y_color, model)\n if args.train_mode == \"semisup\":\n loss = loss_nll + args.weight_y * 0.5 * (loss_xent + loss_xent_color)\n elif args.train_mode == \"sup\":\n loss = 0.5 * (loss_xent + loss_xent_color)\n elif args.train_mode == \"unsup\":\n loss = loss_nll\n else:\n raise ValueError('Choose supported train_mode: semisup, sup, unsup')\n error_score = 1. - np.mean(y_predicted.astype(int) == y.numpy()) \n error_score_color = 1. - np.mean(y_color_predicted.astype(int) == y_color.numpy())\n \n else:\n loss, atol, rtol, logp_actions, nfe = compute_bits_per_dim(x, model)\n loss_nll, loss_xent, loss_xent_color, error_score, error_score_color = loss, 0., 0., 0., 0.\n \n if regularization_coeffs:\n reg_states = get_regularization(model, regularization_coeffs)\n reg_loss = sum(\n reg_state * coeff for reg_state, coeff in zip(reg_states, regularization_coeffs) if coeff != 0\n )\n loss = loss + reg_loss\n total_time = count_total_time(model)\n loss = loss + total_time * args.time_penalty\n\n # re-weight the gate rewards\n normalized_eta = args.eta / len(logp_actions)\n \n # collect cumulative future rewards\n R = - loss\n cum_rewards = []\n for r in nfe[::-1]:\n R = -normalized_eta * r.view(-1,1) + args.gamma * R\n cum_rewards.insert(0,R)\n \n # apply REINFORCE\n rl_loss = 0\n for lpa, r in zip(logp_actions, cum_rewards):\n rl_loss = rl_loss - lpa.view(-1,1) * args.rl_weight * r\n \n loss = loss + rl_loss.mean()\n \n loss.backward()\n \n grad_norm = torch.nn.utils.clip_grad_norm_(model.parameters(), args.max_grad_norm)\n\n optimizer.step()\n\n if args.spectral_norm: spectral_norm_power_iteration(model, args.spectral_norm_niter)\n \n time_meter.update(time.time() - start)\n loss_meter.update(loss.item())\n nll_meter.update(loss_nll.item())\n if args.conditional:\n xent_meter.update(loss_xent.item())\n xent_color_meter.update(loss_xent_color.item())\n else:\n xent_meter.update(loss_xent)\n xent_color_meter.update(loss_xent_color)\n error_meter.update(error_score)\n error_color_meter.update(error_score_color)\n steps_meter.update(count_nfe_gate(model))\n grad_meter.update(grad_norm)\n tt_meter.update(total_time)\n \n for idx in range(len(model.module.transforms)):\n for layer in model.module.transforms[idx].chain:\n if hasattr(layer, 'atol'):\n layer.odefunc.after_odeint()\n \n # write to tensorboard\n writer.add_scalars('time', {'train_iter': time_meter.val}, itr)\n writer.add_scalars('loss', {'train_iter': loss_meter.val}, itr)\n writer.add_scalars('bits_per_dim', {'train_iter': nll_meter.val}, itr)\n writer.add_scalars('xent', {'train_iter': xent_meter.val}, itr)\n writer.add_scalars('xent_color', {'train_iter': xent_color_meter.val}, itr)\n writer.add_scalars('error', {'train_iter': error_meter.val}, itr)\n writer.add_scalars('error_color', {'train_iter': error_color_meter.val}, itr)\n writer.add_scalars('nfe', {'train_iter': steps_meter.val}, itr)\n writer.add_scalars('grad', {'train_iter': grad_meter.val}, itr)\n writer.add_scalars('total_time', {'train_iter': tt_meter.val}, itr)\n\n if itr % args.log_freq == 0:\n for tol_indx in range(len(atol)):\n writer.add_scalars('atol_%i'%tol_indx, {'train': atol[tol_indx].mean()}, itr)\n writer.add_scalars('rtol_%i'%tol_indx, {'train': rtol[tol_indx].mean()}, itr)\n \n log_message = (\n \"Iter {:04d} | Time {:.4f}({:.4f}) | Bit/dim {:.4f}({:.4f}) | Xent {:.4f}({:.4f}) | Xent Color {:.4f}({:.4f}) | Loss {:.4f}({:.4f}) | Error {:.4f}({:.4f}) | Error Color {:.4f}({:.4f}) |\"\n \"Steps {:.0f}({:.2f}) | Grad Norm {:.4f}({:.4f}) | Total Time {:.2f}({:.2f})\".format(\n itr, time_meter.val, time_meter.avg, nll_meter.val, nll_meter.avg, xent_meter.val, xent_meter.avg, xent_color_meter.val, xent_color_meter.avg, loss_meter.val, loss_meter.avg, error_meter.val, error_meter.avg, error_color_meter.val, error_color_meter.avg, steps_meter.val, steps_meter.avg, grad_meter.val, grad_meter.avg, tt_meter.val, tt_meter.avg\n )\n )\n if regularization_coeffs:\n log_message = append_regularization_to_log(log_message, regularization_fns, reg_states)\n logger.info(log_message)\n writer.add_text('info', log_message, itr)\n\n itr += 1\n \n # compute test loss\n model.eval()\n if epoch % args.val_freq == 0:\n with torch.no_grad():\n # write to tensorboard\n writer.add_scalars('time', {'train_epoch': time_meter.avg}, epoch)\n writer.add_scalars('loss', {'train_epoch': loss_meter.avg}, epoch)\n writer.add_scalars('bits_per_dim', {'train_epoch': nll_meter.avg}, epoch)\n writer.add_scalars('xent', {'train_epoch': xent_meter.avg}, epoch)\n writer.add_scalars('xent_color', {'train_epoch': xent_color_meter.avg}, epoch)\n writer.add_scalars('error', {'train_epoch': error_meter.avg}, epoch)\n writer.add_scalars('error_color', {'train_epoch': error_color_meter.avg}, epoch)\n writer.add_scalars('nfe', {'train_epoch': steps_meter.avg}, epoch)\n writer.add_scalars('grad', {'train_epoch': grad_meter.avg}, epoch)\n writer.add_scalars('total_time', {'train_epoch': tt_meter.avg}, epoch)\n \n start = time.time()\n logger.info(\"validating...\")\n writer.add_text('info', \"validating...\", epoch)\n losses_nll = []; losses_xent = []; losses_xent_color = []; losses = []\n total_correct = 0\n total_correct_color = 0\n \n for (x, y_all) in test_loader:\n y = y_all[0]\n y_color = y_all[1]\n if not args.conv:\n x = x.view(x.shape[0], -1)\n x = cvt(x)\n if args.conditional:\n loss_nll, loss_xent, loss_xent_color, y_predicted, y_color_predicted, atol, rtol, logp_actions, nfe = compute_bits_per_dim_conditional(x, y, y_color, model)\n if args.train_mode == \"semisup\":\n loss = loss_nll + args.weight_y * 0.5 * (loss_xent + loss_xent_color)\n elif args.train_mode == \"sup\":\n loss = 0.5 * (loss_xent + loss_xent_color)\n elif args.train_mode == \"unsup\":\n loss = loss_nll\n else:\n raise ValueError('Choose supported train_mode: semisup, sup, unsup')\n total_correct += np.sum(y_predicted.astype(int) == y.numpy())\n total_correct_color += np.sum(y_color_predicted.astype(int) == y_color.numpy())\n else:\n loss, atol, rtol, logp_actions, nfe = compute_bits_per_dim(x, model)\n loss_nll, loss_xent, loss_xent_color = loss, 0., 0.\n losses_nll.append(loss_nll.cpu().numpy()); losses.append(loss.cpu().numpy())\n if args.conditional: \n losses_xent.append(loss_xent.cpu().numpy())\n losses_xent_color.append(loss_xent_color.cpu().numpy())\n else:\n losses_xent.append(loss_xent)\n losses_xent_color.append(loss_xent_color)\n \n loss_nll = np.mean(losses_nll); loss_xent = np.mean(losses_xent); loss_xent_color = np.mean(losses_xent_color); loss = np.mean(losses)\n error_score = 1. - total_correct / len(test_loader.dataset)\n error_score_color = 1. - total_correct_color / len(test_loader.dataset)\n time_epoch_meter.update(time.time() - start_epoch)\n \n # write to tensorboard\n test_time_spent = time.time() - start\n writer.add_scalars('time', {'validation': test_time_spent}, epoch)\n writer.add_scalars('epoch_time', {'validation': time_epoch_meter.val}, epoch)\n writer.add_scalars('bits_per_dim', {'validation': loss_nll}, epoch)\n writer.add_scalars('xent', {'validation': loss_xent}, epoch)\n writer.add_scalars('xent_color', {'validation': loss_xent_color}, epoch)\n writer.add_scalars('loss', {'validation': loss}, epoch)\n writer.add_scalars('error', {'validation': error_score}, epoch)\n writer.add_scalars('error_color', {'validation': error_score_color}, epoch)\n \n for tol_indx in range(len(atol)):\n writer.add_scalars('atol_%i'%tol_indx, {'validation': atol[tol_indx].mean()}, epoch)\n writer.add_scalars('rtol_%i'%tol_indx, {'validation': rtol[tol_indx].mean()}, epoch)\n \n log_message = \"Epoch {:04d} | Time {:.4f}, Epoch Time {:.4f}({:.4f}), Bit/dim {:.4f}(best: {:.4f}), Xent {:.4f}, Xent Color {:.4f}. Loss {:.4f}, Error {:.4f}(best: {:.4f}), Error Color {:.4f}(best: {:.4f})\".format(epoch, time.time() - start, time_epoch_meter.val, time_epoch_meter.avg, loss_nll, best_loss_nll, loss_xent, loss_xent_color, loss, error_score, best_error_score, error_score_color, best_error_score_color)\n logger.info(log_message)\n writer.add_text('info', log_message, epoch)\n \n for name, param in model.named_parameters():\n writer.add_histogram(name, param.clone().cpu().data.numpy(), epoch)\n \n \n utils.makedirs(args.save)\n torch.save({\n \"args\": args,\n \"state_dict\": model.module.state_dict() if torch.cuda.is_available() else model.state_dict(),\n \"optim_state_dict\": optimizer.state_dict(),\n \"epoch\": epoch,\n \"iter\": itr-1,\n \"error\": error_score,\n \"error_color\": error_score_color,\n \"loss\": loss,\n \"xent\": loss_xent,\n \"xent_color\": loss_xent_color,\n \"bits_per_dim\": loss_nll,\n \"best_bits_per_dim\": best_loss_nll,\n \"best_error_score\": best_error_score,\n \"best_error_score_color\": best_error_score_color,\n \"epoch_time\": time_epoch_meter.val,\n \"epoch_time_avg\": time_epoch_meter.avg,\n \"time\": test_time_spent,\n \"error_train\": error_meter.avg,\n \"error_train_color\": error_color_meter.avg,\n \"loss_train\": loss_meter.avg,\n \"xent_train\": xent_meter.avg,\n \"xent_train_color\": xent_color_meter.avg,\n \"bits_per_dim_train\": nll_meter.avg,\n \"total_time_train\": tt_meter.avg,\n \"time_train\": time_meter.avg,\n \"nfe_train\": steps_meter.avg,\n \"grad_train\": grad_meter.avg,\n }, os.path.join(args.save, \"epoch_%i_checkpt.pth\"%epoch))\n \n torch.save({\n \"args\": args,\n \"state_dict\": model.module.state_dict() if torch.cuda.is_available() else model.state_dict(),\n \"optim_state_dict\": optimizer.state_dict(),\n \"epoch\": epoch,\n \"iter\": itr-1,\n \"error\": error_score,\n \"error_color\": error_score_color,\n \"loss\": loss,\n \"xent\": loss_xent,\n \"xent_color\": loss_xent_color,\n \"bits_per_dim\": loss_nll,\n \"best_bits_per_dim\": best_loss_nll,\n \"best_error_score\": best_error_score,\n \"best_error_score_color\": best_error_score_color,\n \"epoch_time\": time_epoch_meter.val,\n \"epoch_time_avg\": time_epoch_meter.avg,\n \"time\": test_time_spent,\n \"error_train\": error_meter.avg,\n \"error_train_color\": error_color_meter.avg,\n \"loss_train\": loss_meter.avg,\n \"xent_train\": xent_meter.avg,\n \"xent_train_color\": xent_color_meter.avg,\n \"bits_per_dim_train\": nll_meter.avg,\n \"total_time_train\": tt_meter.avg,\n \"time_train\": time_meter.avg,\n \"nfe_train\": steps_meter.avg,\n \"grad_train\": grad_meter.avg,\n }, os.path.join(args.save, \"current_checkpt.pth\"))\n \n if loss_nll < best_loss_nll:\n best_loss_nll = loss_nll\n utils.makedirs(args.save)\n torch.save({\n \"args\": args,\n \"state_dict\": model.module.state_dict() if torch.cuda.is_available() else model.state_dict(),\n \"optim_state_dict\": optimizer.state_dict(),\n \"epoch\": epoch,\n \"iter\": itr-1,\n \"error\": error_score,\n \"error_color\": error_score_color,\n \"loss\": loss,\n \"xent\": loss_xent,\n \"xent_color\": loss_xent_color,\n \"bits_per_dim\": loss_nll,\n \"best_bits_per_dim\": best_loss_nll,\n \"best_error_score\": best_error_score,\n \"best_error_score_color\": best_error_score_color,\n \"epoch_time\": time_epoch_meter.val,\n \"epoch_time_avg\": time_epoch_meter.avg,\n \"time\": test_time_spent,\n \"error_train\": error_meter.avg,\n \"error_train_color\": error_color_meter.avg,\n \"loss_train\": loss_meter.avg,\n \"xent_train\": xent_meter.avg,\n \"xent_train_color\": xent_color_meter.avg,\n \"bits_per_dim_train\": nll_meter.avg,\n \"total_time_train\": tt_meter.avg,\n \"time_train\": time_meter.avg,\n \"nfe_train\": steps_meter.avg,\n \"grad_train\": grad_meter.avg,\n }, os.path.join(args.save, \"best_nll_checkpt.pth\"))\n \n if args.conditional:\n if error_score < best_error_score:\n best_error_score = error_score\n utils.makedirs(args.save)\n torch.save({\n \"args\": args,\n \"state_dict\": model.module.state_dict() if torch.cuda.is_available() else model.state_dict(),\n \"optim_state_dict\": optimizer.state_dict(),\n \"epoch\": epoch,\n \"iter\": itr-1,\n \"error\": error_score,\n \"error_color\": error_score_color,\n \"loss\": loss,\n \"xent\": loss_xent,\n \"xent_color\": loss_xent_color,\n \"bits_per_dim\": loss_nll,\n \"best_bits_per_dim\": best_loss_nll,\n \"best_error_score\": best_error_score,\n \"best_error_score_color\": best_error_score_color,\n \"epoch_time\": time_epoch_meter.val,\n \"epoch_time_avg\": time_epoch_meter.avg,\n \"time\": test_time_spent,\n \"error_train\": error_meter.avg,\n \"error_train_color\": error_color_meter.avg,\n \"loss_train\": loss_meter.avg,\n \"xent_train\": xent_meter.avg,\n \"xent_train_color\": xent_color_meter.avg,\n \"bits_per_dim_train\": nll_meter.avg,\n \"total_time_train\": tt_meter.avg,\n \"time_train\": time_meter.avg,\n \"nfe_train\": steps_meter.avg,\n \"grad_train\": grad_meter.avg,\n }, os.path.join(args.save, \"best_error_checkpt.pth\"))\n \n if error_score_color < best_error_score_color:\n best_error_score_color = error_score_color\n utils.makedirs(args.save)\n torch.save({\n \"args\": args,\n \"state_dict\": model.module.state_dict() if torch.cuda.is_available() else model.state_dict(),\n \"optim_state_dict\": optimizer.state_dict(),\n \"epoch\": epoch,\n \"iter\": itr-1,\n \"error\": error_score,\n \"error_color\": error_score_color,\n \"loss\": loss,\n \"xent\": loss_xent,\n \"xent_color\": loss_xent_color,\n \"bits_per_dim\": loss_nll,\n \"best_bits_per_dim\": best_loss_nll,\n \"best_error_score\": best_error_score,\n \"best_error_score_color\": best_error_score_color,\n \"epoch_time\": time_epoch_meter.val,\n \"epoch_time_avg\": time_epoch_meter.avg,\n \"time\": test_time_spent,\n \"error_train\": error_meter.avg,\n \"error_train_color\": error_color_meter.avg,\n \"loss_train\": loss_meter.avg,\n \"xent_train\": xent_meter.avg,\n \"xent_train_color\": xent_color_meter.avg,\n \"bits_per_dim_train\": nll_meter.avg,\n \"total_time_train\": tt_meter.avg,\n \"time_train\": time_meter.avg,\n \"nfe_train\": steps_meter.avg,\n \"grad_train\": grad_meter.avg,\n }, os.path.join(args.save, \"best_error_color_checkpt.pth\"))\n \n\n # visualize samples and density\n with torch.no_grad():\n fig_filename = os.path.join(args.save, \"figs\", \"{:04d}.jpg\".format(epoch))\n utils.makedirs(os.path.dirname(fig_filename))\n generated_samples, atol, rtol, logp_actions, nfe = model(fixed_z, reverse=True)\n generated_samples = generated_samples.view(-1, *data_shape)\n for tol_indx in range(len(atol)):\n writer.add_scalars('atol_gen_%i'%tol_indx, {'validation': atol[tol_indx].mean()}, epoch)\n writer.add_scalars('rtol_gen_%i'%tol_indx, {'validation': rtol[tol_indx].mean()}, epoch)\n save_image(generated_samples, fig_filename, nrow=10)\n if args.data == \"mnist\":\n writer.add_images('generated_images', generated_samples.repeat(1,3,1,1), epoch)\n else:\n writer.add_images('generated_images', generated_samples.repeat(1,1,1,1), epoch)\nNamespace(JFrobint=None, JdiagFrobint=None, JoffdiagFrobint=None, add_noise=True, alpha=1e-06, annealing_std=False, atol=1e-05, autoencode=False, batch_norm=False, batch_size=900, batch_size_schedule='', begin_epoch=1, beta=10.0, cond_nn='linear', condition_ratio=0.25, conditional=True, controlled_tol=False, conv=True, data='colormnist', dims='64,64,64', divergence_fn='approximate', dl2int=None, dropout_rate=0.5, eta=0.1, gamma=0.99, gate='cnn1', imagesize=None, l1int=None, l2int=None, layer_type='concat', log_freq=10, lr=0.001, max_grad_norm=10000000000.0, multiscale=True, nonlinearity='softplus', num_blocks=2, num_epochs=500, parallel=False, rademacher=True, residual=False, resume=None, rl_weight=0.01, rtol=1e-05, save='../experiments_published/infocnf_conditional_disentangle_colormnist_bs900_sratio_1_4th_drop_0_5_rl_stdscale_6_2cond_linear_multiscale_beta_10_run1', scale=1.0, scale_fac=1.0, scale_std=6.0, seed=1, solver='dopri5', spectral_norm=False, spectral_norm_niter=10, step_size=None, strides='1,1,1,1', test_atol=None, test_batch_size=500, test_rtol=None, test_solver=None, time_length=1.0, time_penalty=0, train_T=True, train_mode='semisup', val_freq=1, warmup_iters=1000, weight_decay=0.0, weight_y=0.5, y_class=10, y_color=10)\nODENVP(\n (transforms): ModuleList(\n (0): StackedCNFLayers(\n (chain): ModuleList(\n (0): LogitTransform()\n (1): CNF_Gate(\n (gate_net): FeedforwardGateI(\n (maxpool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n (conv1): Conv2d(3, 10, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n (bn1): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n (relu1): ReLU(inplace)\n (conv2): Conv2d(10, 10, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n (bn2): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n (relu2): ReLU(inplace)\n (avg_layer): AdaptiveAvgPool2d(output_size=(1, 1))\n (linear_layer): Conv2d(10, 1, kernel_size=(1, 1), stride=(1, 1))\n (sigmoid): Sigmoid()\n )\n (odefunc): RegularizedODEfunc(\n (odefunc): ODEfunc_rl(\n (diffeq): ODEnet(\n (layers): ModuleList(\n (0): ConcatConv2d(\n (_layer): Conv2d(4, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (1): ConcatConv2d(\n (_layer): Conv2d(65, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (2): ConcatConv2d(\n (_layer): Conv2d(65, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (3): ConcatConv2d(\n (_layer): Conv2d(65, 3, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n )\n (activation_fns): ModuleList(\n (0): Softplus(beta=1, threshold=20)\n (1): Softplus(beta=1, threshold=20)\n (2): Softplus(beta=1, threshold=20)\n )\n )\n )\n )\n )\n (2): CNF_Gate(\n (gate_net): FeedforwardGateI(\n (maxpool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n (conv1): Conv2d(3, 10, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n (bn1): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n (relu1): ReLU(inplace)\n (conv2): Conv2d(10, 10, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n (bn2): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n (relu2): ReLU(inplace)\n (avg_layer): AdaptiveAvgPool2d(output_size=(1, 1))\n (linear_layer): Conv2d(10, 1, kernel_size=(1, 1), stride=(1, 1))\n (sigmoid): Sigmoid()\n )\n (odefunc): RegularizedODEfunc(\n (odefunc): ODEfunc_rl(\n (diffeq): ODEnet(\n (layers): ModuleList(\n (0): ConcatConv2d(\n (_layer): Conv2d(4, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (1): ConcatConv2d(\n (_layer): Conv2d(65, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (2): ConcatConv2d(\n (_layer): Conv2d(65, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (3): ConcatConv2d(\n (_layer): Conv2d(65, 3, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n )\n (activation_fns): ModuleList(\n (0): Softplus(beta=1, threshold=20)\n (1): Softplus(beta=1, threshold=20)\n (2): Softplus(beta=1, threshold=20)\n )\n )\n )\n )\n )\n (3): SqueezeLayer()\n (4): CNF_Gate(\n (gate_net): FeedforwardGateI(\n (maxpool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n (conv1): Conv2d(12, 10, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n (bn1): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n (relu1): ReLU(inplace)\n (conv2): Conv2d(10, 10, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n (bn2): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n (relu2): ReLU(inplace)\n (avg_layer): AdaptiveAvgPool2d(output_size=(1, 1))\n (linear_layer): Conv2d(10, 1, kernel_size=(1, 1), stride=(1, 1))\n (sigmoid): Sigmoid()\n )\n (odefunc): RegularizedODEfunc(\n (odefunc): ODEfunc_rl(\n (diffeq): ODEnet(\n (layers): ModuleList(\n (0): ConcatConv2d(\n (_layer): Conv2d(13, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (1): ConcatConv2d(\n (_layer): Conv2d(65, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (2): ConcatConv2d(\n (_layer): Conv2d(65, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (3): ConcatConv2d(\n (_layer): Conv2d(65, 12, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n )\n (activation_fns): ModuleList(\n (0): Softplus(beta=1, threshold=20)\n (1): Softplus(beta=1, threshold=20)\n (2): Softplus(beta=1, threshold=20)\n )\n )\n )\n )\n )\n (5): CNF_Gate(\n (gate_net): FeedforwardGateI(\n (maxpool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n (conv1): Conv2d(12, 10, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n (bn1): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n (relu1): ReLU(inplace)\n (conv2): Conv2d(10, 10, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n (bn2): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n (relu2): ReLU(inplace)\n (avg_layer): AdaptiveAvgPool2d(output_size=(1, 1))\n (linear_layer): Conv2d(10, 1, kernel_size=(1, 1), stride=(1, 1))\n (sigmoid): Sigmoid()\n )\n (odefunc): RegularizedODEfunc(\n (odefunc): ODEfunc_rl(\n (diffeq): ODEnet(\n (layers): ModuleList(\n (0): ConcatConv2d(\n (_layer): Conv2d(13, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (1): ConcatConv2d(\n (_layer): Conv2d(65, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (2): ConcatConv2d(\n (_layer): Conv2d(65, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (3): ConcatConv2d(\n (_layer): Conv2d(65, 12, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n )\n (activation_fns): ModuleList(\n (0): Softplus(beta=1, threshold=20)\n (1): Softplus(beta=1, threshold=20)\n (2): Softplus(beta=1, threshold=20)\n )\n )\n )\n )\n )\n )\n )\n (1): StackedCNFLayers(\n (chain): ModuleList(\n (0): CNF_Gate(\n (gate_net): FeedforwardGateI(\n (maxpool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n (conv1): Conv2d(6, 10, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n (bn1): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n (relu1): ReLU(inplace)\n (conv2): Conv2d(10, 10, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n (bn2): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n (relu2): ReLU(inplace)\n (avg_layer): AdaptiveAvgPool2d(output_size=(1, 1))\n (linear_layer): Conv2d(10, 1, kernel_size=(1, 1), stride=(1, 1))\n (sigmoid): Sigmoid()\n )\n (odefunc): RegularizedODEfunc(\n (odefunc): ODEfunc_rl(\n (diffeq): ODEnet(\n (layers): ModuleList(\n (0): ConcatConv2d(\n (_layer): Conv2d(7, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (1): ConcatConv2d(\n (_layer): Conv2d(65, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (2): ConcatConv2d(\n (_layer): Conv2d(65, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (3): ConcatConv2d(\n (_layer): Conv2d(65, 6, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n )\n (activation_fns): ModuleList(\n (0): Softplus(beta=1, threshold=20)\n (1): Softplus(beta=1, threshold=20)\n (2): Softplus(beta=1, threshold=20)\n )\n )\n )\n )\n )\n (1): CNF_Gate(\n (gate_net): FeedforwardGateI(\n (maxpool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n (conv1): Conv2d(6, 10, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n (bn1): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n (relu1): ReLU(inplace)\n (conv2): Conv2d(10, 10, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n (bn2): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n (relu2): ReLU(inplace)\n (avg_layer): AdaptiveAvgPool2d(output_size=(1, 1))\n (linear_layer): Conv2d(10, 1, kernel_size=(1, 1), stride=(1, 1))\n (sigmoid): Sigmoid()\n )\n (odefunc): RegularizedODEfunc(\n (odefunc): ODEfunc_rl(\n (diffeq): ODEnet(\n (layers): ModuleList(\n (0): ConcatConv2d(\n (_layer): Conv2d(7, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (1): ConcatConv2d(\n (_layer): Conv2d(65, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (2): ConcatConv2d(\n (_layer): Conv2d(65, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (3): ConcatConv2d(\n (_layer): Conv2d(65, 6, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n )\n (activation_fns): ModuleList(\n (0): Softplus(beta=1, threshold=20)\n (1): Softplus(beta=1, threshold=20)\n (2): Softplus(beta=1, threshold=20)\n )\n )\n )\n )\n )\n (2): SqueezeLayer()\n (3): CNF_Gate(\n (gate_net): FeedforwardGateI(\n (maxpool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n (conv1): Conv2d(24, 10, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n (bn1): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n (relu1): ReLU(inplace)\n (conv2): Conv2d(10, 10, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n (bn2): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n (relu2): ReLU(inplace)\n (avg_layer): AdaptiveAvgPool2d(output_size=(1, 1))\n (linear_layer): Conv2d(10, 1, kernel_size=(1, 1), stride=(1, 1))\n (sigmoid): Sigmoid()\n )\n (odefunc): RegularizedODEfunc(\n (odefunc): ODEfunc_rl(\n (diffeq): ODEnet(\n (layers): ModuleList(\n (0): ConcatConv2d(\n (_layer): Conv2d(25, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (1): ConcatConv2d(\n (_layer): Conv2d(65, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (2): ConcatConv2d(\n (_layer): Conv2d(65, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (3): ConcatConv2d(\n (_layer): Conv2d(65, 24, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n )\n (activation_fns): ModuleList(\n (0): Softplus(beta=1, threshold=20)\n (1): Softplus(beta=1, threshold=20)\n (2): Softplus(beta=1, threshold=20)\n )\n )\n )\n )\n )\n (4): CNF_Gate(\n (gate_net): FeedforwardGateI(\n (maxpool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n (conv1): Conv2d(24, 10, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n (bn1): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n (relu1): ReLU(inplace)\n (conv2): Conv2d(10, 10, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n (bn2): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n (relu2): ReLU(inplace)\n (avg_layer): AdaptiveAvgPool2d(output_size=(1, 1))\n (linear_layer): Conv2d(10, 1, kernel_size=(1, 1), stride=(1, 1))\n (sigmoid): Sigmoid()\n )\n (odefunc): RegularizedODEfunc(\n (odefunc): ODEfunc_rl(\n (diffeq): ODEnet(\n (layers): ModuleList(\n (0): ConcatConv2d(\n (_layer): Conv2d(25, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (1): ConcatConv2d(\n (_layer): Conv2d(65, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (2): ConcatConv2d(\n (_layer): Conv2d(65, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (3): ConcatConv2d(\n (_layer): Conv2d(65, 24, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n )\n (activation_fns): ModuleList(\n (0): Softplus(beta=1, threshold=20)\n (1): Softplus(beta=1, threshold=20)\n (2): Softplus(beta=1, threshold=20)\n )\n )\n )\n )\n )\n )\n )\n (2): StackedCNFLayers(\n (chain): ModuleList(\n (0): CNF_Gate(\n (gate_net): FeedforwardGateI(\n (maxpool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n (conv1): Conv2d(12, 10, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n (bn1): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n (relu1): ReLU(inplace)\n (conv2): Conv2d(10, 10, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n (bn2): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n (relu2): ReLU(inplace)\n (avg_layer): AdaptiveAvgPool2d(output_size=(1, 1))\n (linear_layer): Conv2d(10, 1, kernel_size=(1, 1), stride=(1, 1))\n (sigmoid): Sigmoid()\n )\n (odefunc): RegularizedODEfunc(\n (odefunc): ODEfunc_rl(\n (diffeq): ODEnet(\n (layers): ModuleList(\n (0): ConcatConv2d(\n (_layer): Conv2d(13, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (1): ConcatConv2d(\n (_layer): Conv2d(65, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (2): ConcatConv2d(\n (_layer): Conv2d(65, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (3): ConcatConv2d(\n (_layer): Conv2d(65, 12, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n )\n (activation_fns): ModuleList(\n (0): Softplus(beta=1, threshold=20)\n (1): Softplus(beta=1, threshold=20)\n (2): Softplus(beta=1, threshold=20)\n )\n )\n )\n )\n )\n (1): CNF_Gate(\n (gate_net): FeedforwardGateI(\n (maxpool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n (conv1): Conv2d(12, 10, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n (bn1): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n (relu1): ReLU(inplace)\n (conv2): Conv2d(10, 10, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n (bn2): BatchNorm2d(10, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n (relu2): ReLU(inplace)\n (avg_layer): AdaptiveAvgPool2d(output_size=(1, 1))\n (linear_layer): Conv2d(10, 1, kernel_size=(1, 1), stride=(1, 1))\n (sigmoid): Sigmoid()\n )\n (odefunc): RegularizedODEfunc(\n (odefunc): ODEfunc_rl(\n (diffeq): ODEnet(\n (layers): ModuleList(\n (0): ConcatConv2d(\n (_layer): Conv2d(13, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (1): ConcatConv2d(\n (_layer): Conv2d(65, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (2): ConcatConv2d(\n (_layer): Conv2d(65, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n (3): ConcatConv2d(\n (_layer): Conv2d(65, 12, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n )\n )\n (activation_fns): ModuleList(\n (0): Softplus(beta=1, threshold=20)\n (1): Softplus(beta=1, threshold=20)\n (2): Softplus(beta=1, threshold=20)\n )\n )\n )\n )\n )\n )\n )\n )\n (project_ycond): LinearZeros(in_features=10, out_features=1176, bias=True)\n (project_ycond_color): LinearZeros(in_features=10, out_features=1176, bias=True)\n (project_class): LinearZeros(in_features=588, out_features=10, bias=True)\n (project_color): LinearZeros(in_features=588, out_features=10, bias=True)\n (dropout): Dropout(p=0.5)\n (dropout_color): Dropout(p=0.5)\n)\nNumber of trainable parameters: 951104\n===> Using batch size 900. Total 66 iterations/epoch.\n/tancode/repos/tan-ffjord/lib/layers/odefunc_rl.py:291: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).\n t = torch.tensor(t).type_as(y)\n/opt/conda/lib/python3.6/site-packages/torch/nn/parallel/_functions.py:61: UserWarning: Was asked to gather along dimension 0, but all input tensors were scalars; will instead unsqueeze and return a vector.\n warnings.warn('Was asked to gather along dimension 0, but all '\nIter 0010 | Time 6.7172(28.9773) | Bit/dim 25.4505(27.3297) | Xent 2.2869(2.3009) | Xent Color 2.3007(2.3023) | Loss 48.6142(51.7970) | Error 0.8744(0.8882) | Error Color 0.9067(0.8898) |Steps 296(293.37) | Grad Norm 260.8726(275.0037) | Total Time 0.00(0.00)\nIter 0020 | Time 7.2747(23.2512) | Bit/dim 20.0909(26.0735) | Xent 2.2513(2.2928) | Xent Color 2.2917(2.3003) | Loss 39.1978(49.5869) | Error 0.8644(0.8869) | Error Color 0.8811(0.8910) |Steps 290(296.46) | Grad Norm 216.4315(264.8294) | Total Time 0.00(0.00)\nIter 0030 | Time 7.2844(19.0315) | Bit/dim 13.8656(23.5615) | Xent 2.1974(2.2735) | Xent Color 2.2745(2.2954) | Loss 27.6801(45.0974) | Error 0.5411(0.8307) | Error Color 0.7822(0.8687) |Steps 302(298.98) | Grad Norm 159.2173(243.6388) | Total Time 0.00(0.00)\nIter 0040 | Time 8.5218(16.1508) | Bit/dim 8.9832(20.2265) | Xent 2.1211(2.2412) | Xent Color 2.2517(2.2866) | Loss 18.9711(39.0977) | Error 0.3322(0.7163) | Error Color 0.7511(0.8413) |Steps 362(309.32) | Grad Norm 95.5069(211.8952) | Total Time 0.00(0.00)\nIter 0050 | Time 8.9191(14.2288) | Bit/dim 6.6884(16.8755) | Xent 2.0425(2.1962) | Xent Color 2.2351(2.2744) | Loss 14.8061(33.0821) | Error 0.3333(0.6080) | Error Color 0.7467(0.8130) |Steps 392(327.39) | Grad Norm 33.8638(171.6100) | Total Time 0.00(0.00)\nIter 0060 | Time 9.3227(12.9358) | Bit/dim 6.0965(14.1054) | Xent 1.9660(2.1429) | Xent Color 2.1978(2.2582) | Loss 13.8245(28.1125) | Error 0.3422(0.5346) | Error Color 0.7800(0.8032) |Steps 368(346.20) | Grad Norm 25.0495(132.5938) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0001 | Time 70.6257, Epoch Time 660.3164(660.3164), Bit/dim 5.7409(best: inf), Xent 1.8894, Xent Color 2.1737. Loss 6.7567, Error 0.2505(best: inf), Error Color 0.7276(best: inf)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 0070 | Time 9.1856(11.9171) | Bit/dim 5.5568(11.9159) | Xent 1.8885(2.0823) | Xent Color 2.1676(2.2383) | Loss 12.6809(24.6215) | Error 0.3100(0.4738) | Error Color 0.6700(0.7873) |Steps 368(355.91) | Grad Norm 18.3093(103.8220) | Total Time 0.00(0.00)\nIter 0080 | Time 8.5462(11.0740) | Bit/dim 4.9704(10.1646) | Xent 1.8138(2.0166) | Xent Color 2.1314(2.2142) | Loss 11.5146(21.3208) | Error 0.3022(0.4245) | Error Color 0.6000(0.7364) |Steps 386(361.58) | Grad Norm 14.4526(80.3129) | Total Time 0.00(0.00)\nIter 0090 | Time 8.7664(10.4519) | Bit/dim 4.5811(8.7587) | Xent 1.7558(1.9542) | Xent Color 2.1022(2.1889) | Loss 10.9431(18.6679) | Error 0.2756(0.3892) | Error Color 0.6278(0.7121) |Steps 398(365.35) | Grad Norm 13.4106(63.1016) | Total Time 0.00(0.00)\nIter 0100 | Time 8.5721(10.0347) | Bit/dim 4.1664(7.6068) | Xent 1.7182(1.8965) | Xent Color 2.0705(2.1608) | Loss 9.9141(16.4906) | Error 0.2700(0.3599) | Error Color 0.5889(0.6806) |Steps 350(367.33) | Grad Norm 9.0484(49.1766) | Total Time 0.00(0.00)\nIter 0110 | Time 8.8333(9.7371) | Bit/dim 3.7622(6.6452) | Xent 1.7104(1.8495) | Xent Color 2.0338(2.1304) | Loss 9.2627(14.6923) | Error 0.2900(0.3392) | Error Color 0.5344(0.6460) |Steps 344(368.86) | Grad Norm 8.0415(38.5385) | Total Time 0.00(0.00)\nIter 0120 | Time 8.8842(9.5004) | Bit/dim 3.3653(5.8283) | Xent 1.7375(1.8180) | Xent Color 2.0034(2.1006) | Loss 8.6042(13.1561) | Error 0.3022(0.3279) | Error Color 0.5533(0.6173) |Steps 398(369.11) | Grad Norm 8.2501(30.6077) | Total Time 0.00(0.00)\nIter 0130 | Time 8.7716(9.3606) | Bit/dim 3.0526(5.1349) | Xent 1.7686(1.8017) | Xent Color 1.9625(2.0678) | Loss 7.8607(11.8673) | Error 0.3089(0.3252) | Error Color 0.5411(0.5919) |Steps 362(371.64) | Grad Norm 6.3208(24.4518) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0002 | Time 56.7973, Epoch Time 664.9273(660.4547), Bit/dim 2.9671(best: 5.7409), Xent 1.7650, Xent Color 1.9373. Loss 3.8927, Error 0.2727(best: 0.2505), Error Color 0.4008(best: 0.7276)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 0140 | Time 9.3806(9.3302) | Bit/dim 2.7925(4.5470) | Xent 1.8212(1.8013) | Xent Color 1.9172(2.0341) | Loss 7.5780(11.2172) | Error 0.3511(0.3309) | Error Color 0.4789(0.5642) |Steps 398(378.14) | Grad Norm 5.6835(19.5810) | Total Time 0.00(0.00)\nIter 0150 | Time 9.7713(9.3318) | Bit/dim 2.6174(4.0578) | Xent 1.8334(1.8090) | Xent Color 1.8795(1.9970) | Loss 7.2931(10.2060) | Error 0.3667(0.3395) | Error Color 0.4844(0.5414) |Steps 392(382.29) | Grad Norm 4.3354(15.6803) | Total Time 0.00(0.00)\nIter 0160 | Time 9.5154(9.3397) | Bit/dim 2.4789(3.6578) | Xent 1.8511(1.8204) | Xent Color 1.8173(1.9568) | Loss 7.0443(9.3975) | Error 0.3756(0.3494) | Error Color 0.4633(0.5223) |Steps 416(387.01) | Grad Norm 3.2786(12.5222) | Total Time 0.00(0.00)\nIter 0170 | Time 9.5460(9.3467) | Bit/dim 2.4090(3.3358) | Xent 1.8710(1.8333) | Xent Color 1.7650(1.9131) | Loss 6.8228(8.7430) | Error 0.4067(0.3620) | Error Color 0.4389(0.5023) |Steps 410(390.85) | Grad Norm 2.7683(10.0176) | Total Time 0.00(0.00)\nIter 0180 | Time 9.0851(9.3864) | Bit/dim 2.3311(3.0782) | Xent 1.8909(1.8464) | Xent Color 1.7012(1.8651) | Loss 6.7733(8.2210) | Error 0.4244(0.3752) | Error Color 0.3933(0.4826) |Steps 398(392.75) | Grad Norm 2.4628(8.0934) | Total Time 0.00(0.00)\nIter 0190 | Time 9.8556(9.4278) | Bit/dim 2.2726(2.8714) | Xent 1.8944(1.8565) | Xent Color 1.6574(1.8142) | Loss 6.5912(7.7909) | Error 0.4322(0.3885) | Error Color 0.4000(0.4623) |Steps 392(394.52) | Grad Norm 1.9854(6.5953) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0003 | Time 59.4674, Epoch Time 703.2980(661.7400), Bit/dim 2.2406(best: 2.9671), Xent 1.8488, Xent Color 1.5590. Loss 3.0926, Error 0.3303(best: 0.2505), Error Color 0.3205(best: 0.4008)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 0200 | Time 9.4039(9.4423) | Bit/dim 2.2369(2.7060) | Xent 1.8804(1.8628) | Xent Color 1.5552(1.7596) | Loss 6.4890(7.9800) | Error 0.4267(0.3986) | Error Color 0.3600(0.4415) |Steps 422(397.15) | Grad Norm 2.9059(5.6168) | Total Time 0.00(0.00)\nIter 0210 | Time 8.9537(9.3514) | Bit/dim 2.1969(2.5768) | Xent 1.8212(1.8583) | Xent Color 1.4897(1.7008) | Loss 6.2960(7.5526) | Error 0.4022(0.4033) | Error Color 0.3600(0.4234) |Steps 392(397.56) | Grad Norm 1.7889(4.7524) | Total Time 0.00(0.00)\nIter 0220 | Time 9.1212(9.3087) | Bit/dim 2.1935(2.4774) | Xent 1.7891(1.8448) | Xent Color 1.4049(1.6344) | Loss 6.1313(7.2172) | Error 0.3756(0.4026) | Error Color 0.3744(0.4061) |Steps 380(397.62) | Grad Norm 1.9538(4.2179) | Total Time 0.00(0.00)\nIter 0230 | Time 9.4057(9.3153) | Bit/dim 2.1735(2.4017) | Xent 1.7253(1.8199) | Xent Color 1.3348(1.5587) | Loss 6.2112(6.9530) | Error 0.3633(0.3967) | Error Color 0.3722(0.3903) |Steps 404(399.36) | Grad Norm 15.3468(4.9750) | Total Time 0.00(0.00)\nIter 0240 | Time 9.2129(9.3003) | Bit/dim 2.1952(2.3445) | Xent 1.6687(1.7879) | Xent Color 1.2999(1.4933) | Loss 6.2078(6.7462) | Error 0.3622(0.3885) | Error Color 0.3678(0.3874) |Steps 374(398.80) | Grad Norm 38.6927(10.2800) | Total Time 0.00(0.00)\nIter 0250 | Time 9.3261(9.2917) | Bit/dim 2.1615(2.2988) | Xent 1.6233(1.7444) | Xent Color 1.1964(1.4226) | Loss 6.0412(6.5681) | Error 0.3689(0.3747) | Error Color 0.3122(0.3710) |Steps 392(398.79) | Grad Norm 17.9464(13.0578) | Total Time 0.00(0.00)\nIter 0260 | Time 9.2764(9.2878) | Bit/dim 2.1560(2.2610) | Xent 1.5278(1.6948) | Xent Color 1.1005(1.3497) | Loss 5.8918(6.4088) | Error 0.3200(0.3614) | Error Color 0.2467(0.3444) |Steps 404(399.47) | Grad Norm 5.7904(13.4234) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0004 | Time 59.1054, Epoch Time 692.9184(662.6754), Bit/dim 2.2640(best: 2.2406), Xent 1.4234, Xent Color 1.9069. Loss 3.0966, Error 0.2461(best: 0.2505), Error Color 0.7540(best: 0.3205)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 0270 | Time 9.0536(9.2732) | Bit/dim 2.1902(2.2433) | Xent 1.4021(1.6347) | Xent Color 1.1097(1.3395) | Loss 5.8966(6.7792) | Error 0.2878(0.3472) | Error Color 0.3044(0.3785) |Steps 404(400.46) | Grad Norm 29.4089(27.0024) | Total Time 0.00(0.00)\nIter 0280 | Time 9.1345(9.2491) | Bit/dim 2.1653(2.2253) | Xent 1.3247(1.5696) | Xent Color 1.0664(1.2790) | Loss 5.9299(6.5529) | Error 0.2567(0.3333) | Error Color 0.2189(0.3610) |Steps 410(399.54) | Grad Norm 21.5437(28.0124) | Total Time 0.00(0.00)\nIter 0290 | Time 9.1795(9.1947) | Bit/dim 2.1553(2.2092) | Xent 1.2923(1.5015) | Xent Color 0.9834(1.2108) | Loss 5.7691(6.3521) | Error 0.2689(0.3191) | Error Color 0.1867(0.3251) |Steps 398(399.75) | Grad Norm 12.1269(25.2689) | Total Time 0.00(0.00)\nIter 0300 | Time 9.1232(9.1806) | Bit/dim 2.1806(2.1968) | Xent 1.1452(1.4301) | Xent Color 0.8935(1.1420) | Loss 5.7087(6.1823) | Error 0.2422(0.3059) | Error Color 0.1556(0.2911) |Steps 398(397.63) | Grad Norm 6.1207(21.4241) | Total Time 0.00(0.00)\nIter 0310 | Time 9.2053(9.1935) | Bit/dim 2.1731(2.1918) | Xent 1.1081(1.3515) | Xent Color 0.8773(1.0706) | Loss 5.6488(6.0548) | Error 0.2400(0.2904) | Error Color 0.2133(0.2650) |Steps 410(398.16) | Grad Norm 17.8320(18.8727) | Total Time 0.00(0.00)\nIter 0320 | Time 9.1081(9.1686) | Bit/dim 2.3436(2.2107) | Xent 1.0477(1.2709) | Xent Color 3.3597(1.2907) | Loss 7.0553(6.1134) | Error 0.2422(0.2780) | Error Color 0.8400(0.3280) |Steps 392(398.99) | Grad Norm 217.2945(46.1337) | Total Time 0.00(0.00)\nIter 0330 | Time 8.5102(9.1415) | Bit/dim 2.2318(2.2186) | Xent 1.0155(1.2024) | Xent Color 1.2981(1.3370) | Loss 5.8035(6.0780) | Error 0.2356(0.2679) | Error Color 0.5267(0.3948) |Steps 374(396.60) | Grad Norm 41.3241(51.2425) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0005 | Time 57.1491, Epoch Time 681.8088(663.2494), Bit/dim 2.1932(best: 2.2406), Xent 0.9300, Xent Color 1.1729. Loss 2.7189, Error 0.1822(best: 0.2461), Error Color 0.4311(best: 0.3205)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 0340 | Time 9.3889(9.1476) | Bit/dim 2.1497(2.2083) | Xent 1.0450(1.1533) | Xent Color 1.1686(1.2932) | Loss 5.8075(6.4107) | Error 0.2489(0.2618) | Error Color 0.3900(0.3965) |Steps 404(395.94) | Grad Norm 26.2196(45.2528) | Total Time 0.00(0.00)\nIter 0350 | Time 9.3942(9.0989) | Bit/dim 2.1385(2.1904) | Xent 0.9926(1.1168) | Xent Color 1.0802(1.2370) | Loss 5.6914(6.2140) | Error 0.2256(0.2577) | Error Color 0.3122(0.3728) |Steps 410(394.40) | Grad Norm 11.9567(37.0845) | Total Time 0.00(0.00)\nIter 0360 | Time 8.8042(9.0569) | Bit/dim 2.1227(2.1759) | Xent 0.9006(1.0708) | Xent Color 0.9718(1.1790) | Loss 5.4301(6.0503) | Error 0.2300(0.2523) | Error Color 0.2300(0.3397) |Steps 368(394.10) | Grad Norm 9.2982(29.8413) | Total Time 0.00(0.00)\nIter 0370 | Time 8.7372(9.0536) | Bit/dim 2.1365(2.1642) | Xent 0.8315(1.0143) | Xent Color 0.9206(1.1199) | Loss 5.5160(5.9123) | Error 0.2356(0.2434) | Error Color 0.2100(0.3076) |Steps 386(394.04) | Grad Norm 4.8088(23.3643) | Total Time 0.00(0.00)\nIter 0380 | Time 8.9203(9.0468) | Bit/dim 2.1290(2.1573) | Xent 0.8045(0.9584) | Xent Color 0.8965(1.0625) | Loss 5.3339(5.7925) | Error 0.2089(0.2356) | Error Color 0.1800(0.2782) |Steps 380(394.12) | Grad Norm 5.0422(18.5704) | Total Time 0.00(0.00)\nIter 0390 | Time 9.6733(9.0798) | Bit/dim 2.1331(2.1505) | Xent 0.8247(0.9170) | Xent Color 0.8518(1.0075) | Loss 5.4388(5.7002) | Error 0.2433(0.2295) | Error Color 0.1978(0.2537) |Steps 392(394.66) | Grad Norm 7.0343(15.7348) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0006 | Time 58.0401, Epoch Time 679.6995(663.7429), Bit/dim 2.1330(best: 2.1932), Xent 0.6577, Xent Color 0.7270. Loss 2.4792, Error 0.1498(best: 0.1822), Error Color 0.0746(best: 0.3205)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 0400 | Time 9.5213(9.1470) | Bit/dim 2.1461(2.1435) | Xent 0.7502(0.8785) | Xent Color 0.7788(0.9524) | Loss 5.4233(6.1132) | Error 0.1978(0.2235) | Error Color 0.1689(0.2321) |Steps 398(394.49) | Grad Norm 9.6663(13.7149) | Total Time 0.00(0.00)\nIter 0410 | Time 9.4314(9.1417) | Bit/dim 2.1297(2.1368) | Xent 0.7200(0.8438) | Xent Color 0.7489(0.9007) | Loss 5.4094(5.9191) | Error 0.2056(0.2188) | Error Color 0.1456(0.2131) |Steps 410(394.66) | Grad Norm 12.3953(13.3133) | Total Time 0.00(0.00)\nIter 0420 | Time 9.0920(9.1527) | Bit/dim 2.1761(2.1342) | Xent 0.6716(0.8066) | Xent Color 1.4224(0.8880) | Loss 5.6537(5.7845) | Error 0.1989(0.2123) | Error Color 0.6189(0.2252) |Steps 386(395.06) | Grad Norm 142.6906(22.1804) | Total Time 0.00(0.00)\nIter 0430 | Time 9.4062(9.1654) | Bit/dim 2.0998(2.1298) | Xent 0.6981(0.7777) | Xent Color 0.6841(0.8974) | Loss 5.3709(5.6903) | Error 0.1889(0.2067) | Error Color 0.1678(0.2502) |Steps 410(396.21) | Grad Norm 4.9823(31.4163) | Total Time 0.00(0.00)\nIter 0440 | Time 9.1139(9.1420) | Bit/dim 2.0973(2.1218) | Xent 0.7159(0.7677) | Xent Color 0.6583(0.8468) | Loss 5.2336(5.5906) | Error 0.2167(0.2065) | Error Color 0.1422(0.2334) |Steps 386(396.01) | Grad Norm 2.4519(28.0055) | Total Time 0.00(0.00)\nIter 0450 | Time 9.1725(9.1428) | Bit/dim 2.1241(2.1149) | Xent 0.6861(0.7504) | Xent Color 0.6117(0.7930) | Loss 5.2659(5.5104) | Error 0.2067(0.2051) | Error Color 0.1222(0.2073) |Steps 404(396.49) | Grad Norm 8.9610(24.1920) | Total Time 0.00(0.00)\nIter 0460 | Time 9.2673(9.1450) | Bit/dim 2.0895(2.1095) | Xent 0.7110(0.7286) | Xent Color 0.5501(0.7383) | Loss 5.2467(5.4377) | Error 0.2044(0.2018) | Error Color 0.0989(0.1823) |Steps 404(398.34) | Grad Norm 5.9507(21.1386) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0007 | Time 59.2969, Epoch Time 686.7517(664.4332), Bit/dim 2.0983(best: 2.1330), Xent 0.5478, Xent Color 0.5069. Loss 2.3620, Error 0.1357(best: 0.1498), Error Color 0.0459(best: 0.0746)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 0470 | Time 9.4887(9.1887) | Bit/dim 2.0731(2.1032) | Xent 0.6185(0.7068) | Xent Color 0.5381(0.6898) | Loss 5.1876(5.8212) | Error 0.1789(0.1979) | Error Color 0.0922(0.1615) |Steps 398(399.24) | Grad Norm 4.4478(18.8473) | Total Time 0.00(0.00)\nIter 0480 | Time 8.9057(9.2101) | Bit/dim 2.0673(2.0977) | Xent 0.6631(0.6911) | Xent Color 0.4830(0.6432) | Loss 5.1382(5.6402) | Error 0.1967(0.1962) | Error Color 0.0867(0.1440) |Steps 410(400.13) | Grad Norm 11.1923(15.9550) | Total Time 0.00(0.00)\nIter 0490 | Time 9.3357(9.2346) | Bit/dim 2.0761(2.0928) | Xent 0.6305(0.6788) | Xent Color 0.5262(0.6059) | Loss 5.0455(5.5159) | Error 0.1744(0.1946) | Error Color 0.1367(0.1333) |Steps 404(401.53) | Grad Norm 27.7882(16.3739) | Total Time 0.00(0.00)\nIter 0500 | Time 9.3203(9.2283) | Bit/dim 2.0709(2.0864) | Xent 0.6408(0.6658) | Xent Color 1.0010(0.6060) | Loss 5.3432(5.4281) | Error 0.2000(0.1923) | Error Color 0.4322(0.1506) |Steps 404(402.79) | Grad Norm 128.9114(26.3958) | Total Time 0.00(0.00)\nIter 0510 | Time 9.0984(9.2196) | Bit/dim 2.0598(2.0856) | Xent 0.6175(0.6543) | Xent Color 0.5491(0.6588) | Loss 5.1286(5.3819) | Error 0.1833(0.1891) | Error Color 0.1556(0.1832) |Steps 398(401.93) | Grad Norm 19.0915(37.1310) | Total Time 0.00(0.00)\nIter 0520 | Time 8.9585(9.1894) | Bit/dim 2.0503(2.0760) | Xent 0.6164(0.6534) | Xent Color 0.5226(0.6338) | Loss 4.9709(5.3120) | Error 0.1800(0.1888) | Error Color 0.1033(0.1737) |Steps 380(400.84) | Grad Norm 18.5146(33.6722) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0008 | Time 59.0021, Epoch Time 690.7191(665.2217), Bit/dim 2.0525(best: 2.0983), Xent 0.4770, Xent Color 0.4204. Loss 2.2769, Error 0.1287(best: 0.1357), Error Color 0.0557(best: 0.0459)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 0530 | Time 9.4225(9.2271) | Bit/dim 2.0289(2.0667) | Xent 0.5710(0.6465) | Xent Color 0.5675(0.6031) | Loss 4.9024(5.7885) | Error 0.1756(0.1879) | Error Color 0.1822(0.1627) |Steps 404(401.99) | Grad Norm 42.6756(30.9025) | Total Time 0.00(0.00)\nIter 0540 | Time 9.1718(9.2343) | Bit/dim 2.0199(2.0599) | Xent 0.5650(0.6336) | Xent Color 0.4578(0.5697) | Loss 5.0246(5.5880) | Error 0.1733(0.1857) | Error Color 0.1000(0.1511) |Steps 398(401.78) | Grad Norm 19.2245(28.7212) | Total Time 0.00(0.00)\nIter 0550 | Time 9.3834(9.2499) | Bit/dim 2.0297(2.0522) | Xent 0.6291(0.6221) | Xent Color 0.3987(0.5301) | Loss 4.9850(5.4283) | Error 0.1744(0.1822) | Error Color 0.0633(0.1355) |Steps 404(402.88) | Grad Norm 8.7679(25.3062) | Total Time 0.00(0.00)\nIter 0560 | Time 9.4572(9.2707) | Bit/dim 2.1703(2.0513) | Xent 0.5655(0.6127) | Xent Color 4.3467(0.6392) | Loss 7.0091(5.3909) | Error 0.1767(0.1821) | Error Color 0.8111(0.1596) |Steps 380(403.58) | Grad Norm 248.6860(34.1775) | Total Time 0.00(0.00)\nIter 0570 | Time 9.1423(9.2506) | Bit/dim 2.0738(2.0761) | Xent 0.7124(0.6236) | Xent Color 0.7875(0.8889) | Loss 5.3442(5.4804) | Error 0.2222(0.1881) | Error Color 0.2867(0.2423) |Steps 398(402.40) | Grad Norm 41.5933(46.4471) | Total Time 0.00(0.00)\nIter 0580 | Time 8.5995(9.1970) | Bit/dim 2.0546(2.0708) | Xent 0.6698(0.6517) | Xent Color 0.6495(0.8420) | Loss 5.0626(5.4015) | Error 0.2044(0.1974) | Error Color 0.1978(0.2429) |Steps 380(398.69) | Grad Norm 17.5373(39.7830) | Total Time 0.00(0.00)\nIter 0590 | Time 9.2272(9.1438) | Bit/dim 2.0165(2.0621) | Xent 0.6385(0.6487) | Xent Color 0.5056(0.7693) | Loss 5.0389(5.3103) | Error 0.1911(0.1970) | Error Color 0.1356(0.2216) |Steps 398(395.55) | Grad Norm 8.5801(31.7753) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0009 | Time 58.0060, Epoch Time 686.8772(665.8714), Bit/dim 2.0364(best: 2.0525), Xent 0.4601, Xent Color 0.3912. Loss 2.2493, Error 0.1255(best: 0.1287), Error Color 0.0549(best: 0.0459)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 0600 | Time 9.1449(9.1441) | Bit/dim 2.0257(2.0514) | Xent 0.6099(0.6353) | Xent Color 0.4471(0.6925) | Loss 4.9638(5.6913) | Error 0.2089(0.1943) | Error Color 0.1100(0.1940) |Steps 404(395.72) | Grad Norm 9.6668(25.6121) | Total Time 0.00(0.00)\nIter 0610 | Time 9.6455(9.1792) | Bit/dim 2.0256(2.0436) | Xent 0.5043(0.6160) | Xent Color 0.3961(0.6144) | Loss 4.8792(5.4914) | Error 0.1511(0.1884) | Error Color 0.0900(0.1657) |Steps 386(396.20) | Grad Norm 3.8504(20.2711) | Total Time 0.00(0.00)\nIter 0620 | Time 8.8903(9.2134) | Bit/dim 1.9888(2.0326) | Xent 0.5913(0.6060) | Xent Color 0.3353(0.5466) | Loss 4.8844(5.3381) | Error 0.1700(0.1856) | Error Color 0.0600(0.1423) |Steps 404(398.27) | Grad Norm 8.6997(16.8191) | Total Time 0.00(0.00)\nIter 0630 | Time 9.2710(9.2189) | Bit/dim 2.0170(2.0247) | Xent 0.5323(0.5984) | Xent Color 0.3095(0.4888) | Loss 4.9071(5.2141) | Error 0.1700(0.1842) | Error Color 0.0622(0.1217) |Steps 398(396.72) | Grad Norm 7.3644(14.3718) | Total Time 0.00(0.00)\nIter 0640 | Time 9.2847(9.2266) | Bit/dim 1.9812(2.0166) | Xent 0.5717(0.5927) | Xent Color 0.2686(0.4385) | Loss 4.8180(5.1147) | Error 0.1633(0.1821) | Error Color 0.0489(0.1059) |Steps 374(394.98) | Grad Norm 7.3085(12.7881) | Total Time 0.00(0.00)\nIter 0650 | Time 9.1105(9.2200) | Bit/dim 2.0091(2.0090) | Xent 0.5506(0.5830) | Xent Color 0.2471(0.3933) | Loss 4.8428(5.0406) | Error 0.1622(0.1789) | Error Color 0.0489(0.0910) |Steps 368(395.67) | Grad Norm 1.3976(10.4449) | Total Time 0.00(0.00)\nIter 0660 | Time 9.5099(9.2618) | Bit/dim 1.9760(2.0011) | Xent 0.5797(0.5731) | Xent Color 0.2764(0.3546) | Loss 4.8831(4.9795) | Error 0.1878(0.1761) | Error Color 0.0556(0.0788) |Steps 404(398.82) | Grad Norm 26.5622(9.3783) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0010 | Time 61.5212, Epoch Time 695.7308(666.7672), Bit/dim 1.9873(best: 2.0364), Xent 0.4134, Xent Color 0.2015. Loss 2.1410, Error 0.1191(best: 0.1255), Error Color 0.0214(best: 0.0459)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 0670 | Time 9.4487(9.2806) | Bit/dim 1.9724(1.9930) | Xent 0.5625(0.5711) | Xent Color 0.2049(0.3238) | Loss 4.7675(5.3341) | Error 0.1844(0.1765) | Error Color 0.0333(0.0703) |Steps 398(400.04) | Grad Norm 10.6748(10.0485) | Total Time 0.00(0.00)\nIter 0680 | Time 9.1763(9.2918) | Bit/dim 1.9614(1.9849) | Xent 0.4529(0.5608) | Xent Color 0.2836(0.2985) | Loss 4.7054(5.1797) | Error 0.1356(0.1729) | Error Color 0.0667(0.0634) |Steps 392(400.26) | Grad Norm 45.3951(12.0547) | Total Time 0.00(0.00)\nIter 0690 | Time 9.6863(9.2823) | Bit/dim 2.4790(2.1575) | Xent 0.8430(0.5829) | Xent Color 1.5044(1.4873) | Loss 6.4157(5.9438) | Error 0.2789(0.1799) | Error Color 0.5856(0.2093) |Steps 398(400.74) | Grad Norm 19.2891(48.4319) | Total Time 0.00(0.00)\nIter 0700 | Time 8.8773(9.2739) | Bit/dim 2.1898(2.1726) | Xent 0.7311(0.6872) | Xent Color 0.9471(1.4220) | Loss 5.5616(5.9115) | Error 0.2322(0.2159) | Error Color 0.3900(0.2775) |Steps 356(396.67) | Grad Norm 15.6015(40.7727) | Total Time 0.00(0.00)\nIter 0710 | Time 9.0732(9.1932) | Bit/dim 2.0809(2.1576) | Xent 0.6330(0.6813) | Xent Color 0.9156(1.3055) | Loss 5.2688(5.7770) | Error 0.1900(0.2144) | Error Color 0.3356(0.3062) |Steps 386(393.00) | Grad Norm 8.5618(33.5252) | Total Time 0.00(0.00)\nIter 0720 | Time 9.1195(9.1243) | Bit/dim 2.0419(2.1299) | Xent 0.6402(0.6746) | Xent Color 0.8017(1.1845) | Loss 5.0754(5.6253) | Error 0.1956(0.2116) | Error Color 0.3033(0.3090) |Steps 350(390.36) | Grad Norm 6.2612(27.3116) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0011 | Time 58.3225, Epoch Time 686.1206(667.3478), Bit/dim 2.0283(best: 1.9873), Xent 0.4357, Xent Color 0.5437. Loss 2.2731, Error 0.1243(best: 0.1191), Error Color 0.1362(best: 0.0214)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 0730 | Time 8.7026(9.0884) | Bit/dim 2.0347(2.1004) | Xent 0.5862(0.6566) | Xent Color 0.6284(1.0539) | Loss 5.0554(5.9545) | Error 0.1733(0.2050) | Error Color 0.2078(0.2907) |Steps 380(388.79) | Grad Norm 8.5212(22.0441) | Total Time 0.00(0.00)\nIter 0740 | Time 8.8387(9.0506) | Bit/dim 2.0098(2.0769) | Xent 0.5509(0.6400) | Xent Color 0.5142(0.9263) | Loss 4.9784(5.6903) | Error 0.1700(0.1997) | Error Color 0.1489(0.2638) |Steps 392(386.59) | Grad Norm 3.8025(17.4328) | Total Time 0.00(0.00)\nIter 0750 | Time 9.3532(9.0600) | Bit/dim 2.0143(2.0566) | Xent 0.5931(0.6181) | Xent Color 0.4427(0.8097) | Loss 4.9312(5.4847) | Error 0.1778(0.1928) | Error Color 0.1356(0.2350) |Steps 350(386.22) | Grad Norm 2.8505(13.6355) | Total Time 0.00(0.00)\nIter 0760 | Time 9.0506(9.0744) | Bit/dim 1.9839(2.0391) | Xent 0.5358(0.6008) | Xent Color 0.3637(0.7037) | Loss 4.8744(5.3164) | Error 0.1644(0.1872) | Error Color 0.0989(0.2038) |Steps 398(386.38) | Grad Norm 4.2409(10.9902) | Total Time 0.00(0.00)\nIter 0770 | Time 9.3725(9.1241) | Bit/dim 1.9972(2.0253) | Xent 0.4860(0.5874) | Xent Color 0.2982(0.6077) | Loss 4.8649(5.1938) | Error 0.1644(0.1823) | Error Color 0.0744(0.1745) |Steps 410(389.36) | Grad Norm 2.4446(9.1092) | Total Time 0.00(0.00)\nIter 0780 | Time 9.3258(9.1627) | Bit/dim 1.9846(2.0133) | Xent 0.6120(0.5788) | Xent Color 0.2460(0.5215) | Loss 4.8556(5.0897) | Error 0.2011(0.1795) | Error Color 0.0589(0.1469) |Steps 404(391.47) | Grad Norm 1.8936(7.9804) | Total Time 0.00(0.00)\nIter 0790 | Time 9.4605(9.1706) | Bit/dim 1.9565(2.0020) | Xent 0.4803(0.5631) | Xent Color 0.2402(0.4471) | Loss 4.7149(4.9970) | Error 0.1522(0.1746) | Error Color 0.0678(0.1242) |Steps 422(395.10) | Grad Norm 1.4454(6.8745) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0012 | Time 61.1946, Epoch Time 686.8557(667.9330), Bit/dim 1.9769(best: 1.9873), Xent 0.3842, Xent Color 0.1415. Loss 2.1084, Error 0.1056(best: 0.1191), Error Color 0.0123(best: 0.0214)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 0800 | Time 9.0690(9.1844) | Bit/dim 1.9928(1.9931) | Xent 0.5582(0.5535) | Xent Color 0.4535(0.3976) | Loss 4.8590(5.3691) | Error 0.1778(0.1709) | Error Color 0.1844(0.1113) |Steps 380(396.99) | Grad Norm 55.9886(9.9093) | Total Time 0.00(0.00)\nIter 0810 | Time 9.1171(9.1723) | Bit/dim 1.9829(2.0024) | Xent 0.4543(0.5580) | Xent Color 0.4551(0.5369) | Loss 4.8335(5.3183) | Error 0.1400(0.1733) | Error Color 0.1556(0.1621) |Steps 398(396.03) | Grad Norm 18.2568(20.6431) | Total Time 0.00(0.00)\nIter 0820 | Time 8.7196(9.1322) | Bit/dim 2.0183(1.9996) | Xent 0.4991(0.5617) | Xent Color 0.3924(0.5085) | Loss 4.8489(5.2000) | Error 0.1533(0.1749) | Error Color 0.1356(0.1592) |Steps 404(395.06) | Grad Norm 13.9879(19.2338) | Total Time 0.00(0.00)\nIter 0830 | Time 8.7758(9.0919) | Bit/dim 1.9692(1.9902) | Xent 0.5009(0.5557) | Xent Color 0.2398(0.4511) | Loss 4.6502(5.0816) | Error 0.1678(0.1734) | Error Color 0.0667(0.1375) |Steps 410(393.94) | Grad Norm 6.1257(16.0344) | Total Time 0.00(0.00)\nIter 0840 | Time 8.6898(9.1062) | Bit/dim 1.9555(1.9822) | Xent 0.5122(0.5444) | Xent Color 0.1836(0.3856) | Loss 4.6352(4.9828) | Error 0.1456(0.1695) | Error Color 0.0433(0.1127) |Steps 386(393.98) | Grad Norm 2.9694(13.0906) | Total Time 0.00(0.00)\nIter 0850 | Time 9.0588(9.1164) | Bit/dim 1.9506(1.9724) | Xent 0.5105(0.5342) | Xent Color 0.1626(0.3293) | Loss 4.6384(4.8949) | Error 0.1611(0.1666) | Error Color 0.0400(0.0934) |Steps 416(394.51) | Grad Norm 3.0437(10.5216) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0013 | Time 60.4818, Epoch Time 685.7198(668.4666), Bit/dim 1.9471(best: 1.9769), Xent 0.3693, Xent Color 0.0975. Loss 2.0638, Error 0.1070(best: 0.1056), Error Color 0.0095(best: 0.0123)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 0860 | Time 9.3641(9.1482) | Bit/dim 1.9290(1.9638) | Xent 0.4822(0.5289) | Xent Color 0.1484(0.2826) | Loss 4.5541(5.3621) | Error 0.1356(0.1655) | Error Color 0.0300(0.0767) |Steps 380(394.81) | Grad Norm 3.0546(8.4811) | Total Time 0.00(0.00)\nIter 0870 | Time 9.1043(9.1583) | Bit/dim 1.9578(1.9546) | Xent 0.4984(0.5228) | Xent Color 0.1165(0.2438) | Loss 4.6591(5.1646) | Error 0.1544(0.1630) | Error Color 0.0222(0.0632) |Steps 380(397.37) | Grad Norm 1.2352(6.7387) | Total Time 0.00(0.00)\nIter 0880 | Time 9.1356(9.1862) | Bit/dim 1.9213(1.9477) | Xent 0.5333(0.5161) | Xent Color 0.1196(0.2129) | Loss 4.6060(5.0257) | Error 0.1867(0.1614) | Error Color 0.0256(0.0532) |Steps 404(397.81) | Grad Norm 1.1387(5.4732) | Total Time 0.00(0.00)\nIter 0890 | Time 9.0929(9.2180) | Bit/dim 1.9106(1.9421) | Xent 0.5084(0.5124) | Xent Color 0.1044(0.1858) | Loss 4.5874(4.9172) | Error 0.1578(0.1604) | Error Color 0.0156(0.0440) |Steps 410(398.64) | Grad Norm 4.2379(4.5422) | Total Time 0.00(0.00)\nIter 0900 | Time 9.4961(9.2743) | Bit/dim 1.9293(1.9363) | Xent 0.4574(0.5083) | Xent Color 0.1061(0.1643) | Loss 4.6209(4.8241) | Error 0.1533(0.1587) | Error Color 0.0244(0.0379) |Steps 404(397.89) | Grad Norm 1.4704(4.1748) | Total Time 0.00(0.00)\nIter 0910 | Time 9.2397(9.3108) | Bit/dim 1.9109(1.9298) | Xent 0.5186(0.5024) | Xent Color 0.0875(0.1469) | Loss 4.5440(4.7581) | Error 0.1678(0.1570) | Error Color 0.0189(0.0329) |Steps 404(399.04) | Grad Norm 3.6815(3.9061) | Total Time 0.00(0.00)\nIter 0920 | Time 9.1589(9.2879) | Bit/dim 1.9186(1.9231) | Xent 0.4982(0.5008) | Xent Color 0.0842(0.1317) | Loss 4.5277(4.7066) | Error 0.1478(0.1542) | Error Color 0.0122(0.0283) |Steps 404(398.89) | Grad Norm 4.0143(3.4989) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0014 | Time 60.4269, Epoch Time 697.6148(669.3411), Bit/dim 1.9074(best: 1.9471), Xent 0.3536, Xent Color 0.0475. Loss 2.0076, Error 0.1032(best: 0.1056), Error Color 0.0032(best: 0.0095)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 0930 | Time 9.6231(9.3157) | Bit/dim 1.8977(1.9168) | Xent 0.4540(0.4983) | Xent Color 0.0755(0.1198) | Loss 4.5132(5.1282) | Error 0.1489(0.1543) | Error Color 0.0144(0.0253) |Steps 404(400.66) | Grad Norm 4.6427(3.9062) | Total Time 0.00(0.00)\nIter 0940 | Time 9.6831(9.2965) | Bit/dim 1.8851(1.9110) | Xent 0.4422(0.4884) | Xent Color 0.0802(0.1095) | Loss 4.4898(4.9689) | Error 0.1456(0.1513) | Error Color 0.0122(0.0225) |Steps 416(404.43) | Grad Norm 5.3138(4.3038) | Total Time 0.00(0.00)\nIter 0950 | Time 9.4883(9.2934) | Bit/dim 1.8723(1.9052) | Xent 0.4868(0.4916) | Xent Color 0.0826(0.1031) | Loss 4.5530(4.8612) | Error 0.1489(0.1522) | Error Color 0.0144(0.0215) |Steps 416(404.87) | Grad Norm 14.4166(5.7711) | Total Time 0.00(0.00)\nIter 0960 | Time 9.1306(9.2467) | Bit/dim 1.8979(1.8985) | Xent 0.5749(0.4880) | Xent Color 0.0645(0.0959) | Loss 4.5138(4.7637) | Error 0.1822(0.1513) | Error Color 0.0078(0.0197) |Steps 404(403.77) | Grad Norm 8.5314(6.1008) | Total Time 0.00(0.00)\nIter 0970 | Time 9.1473(9.2605) | Bit/dim 1.8876(1.8938) | Xent 0.5040(0.4897) | Xent Color 0.0739(0.0890) | Loss 4.4940(4.6939) | Error 0.1689(0.1524) | Error Color 0.0156(0.0180) |Steps 392(403.65) | Grad Norm 13.0868(6.2384) | Total Time 0.00(0.00)\nIter 0980 | Time 9.2549(9.2795) | Bit/dim 1.8844(1.8883) | Xent 0.4150(0.4862) | Xent Color 0.0951(0.0882) | Loss 4.5102(4.6486) | Error 0.1344(0.1508) | Error Color 0.0233(0.0188) |Steps 404(406.12) | Grad Norm 21.7056(9.0341) | Total Time 0.00(0.00)\nIter 0990 | Time 9.0893(9.2506) | Bit/dim 1.8608(1.8813) | Xent 0.4447(0.4872) | Xent Color 0.0669(0.0833) | Loss 4.4198(4.6010) | Error 0.1422(0.1516) | Error Color 0.0156(0.0177) |Steps 410(405.68) | Grad Norm 9.7601(9.0233) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0015 | Time 61.4996, Epoch Time 695.4272(670.1237), Bit/dim 1.8708(best: 1.9074), Xent 0.3374, Xent Color 0.0398. Loss 1.9651, Error 0.0986(best: 0.1032), Error Color 0.0047(best: 0.0032)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 1000 | Time 9.0699(9.3324) | Bit/dim 2.3956(1.9209) | Xent 0.5016(0.4876) | Xent Color 2.7520(0.6848) | Loss 6.6072(5.3392) | Error 0.1589(0.1514) | Error Color 0.5533(0.0678) |Steps 374(406.49) | Grad Norm 110.2073(32.6068) | Total Time 0.00(0.00)\nIter 1010 | Time 10.8736(9.5948) | Bit/dim 2.2263(2.0558) | Xent 0.9779(0.5655) | Xent Color 2.2289(1.2679) | Loss 6.5132(5.7851) | Error 0.3222(0.1780) | Error Color 0.6200(0.2152) |Steps 512(422.95) | Grad Norm 48.4332(38.4317) | Total Time 0.00(0.00)\nIter 1020 | Time 11.2847(10.0230) | Bit/dim 2.1151(2.0802) | Xent 0.5581(0.5990) | Xent Color 1.0769(1.2977) | Loss 5.7828(5.8314) | Error 0.1844(0.1899) | Error Color 0.4233(0.2914) |Steps 524(445.79) | Grad Norm 18.8597(34.8306) | Total Time 0.00(0.00)\nIter 1030 | Time 11.1110(10.3452) | Bit/dim 2.0076(2.0673) | Xent 0.5589(0.6089) | Xent Color 0.8675(1.2136) | Loss 5.3677(5.7608) | Error 0.1756(0.1937) | Error Color 0.3322(0.3134) |Steps 458(455.42) | Grad Norm 9.8981(29.3010) | Total Time 0.00(0.00)\nIter 1040 | Time 10.9590(10.4383) | Bit/dim 1.9554(2.0451) | Xent 0.6056(0.5988) | Xent Color 0.6453(1.0885) | Loss 5.1025(5.6166) | Error 0.1922(0.1909) | Error Color 0.2378(0.3031) |Steps 452(460.43) | Grad Norm 5.7269(23.8483) | Total Time 0.00(0.00)\nIter 1050 | Time 10.0779(10.4113) | Bit/dim 1.9754(2.0231) | Xent 0.5564(0.5900) | Xent Color 0.4602(0.9367) | Loss 4.9695(5.4534) | Error 0.1789(0.1881) | Error Color 0.1589(0.2682) |Steps 434(459.21) | Grad Norm 3.6072(18.5817) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0016 | Time 66.4438, Epoch Time 784.8283(673.5648), Bit/dim 1.9604(best: 1.8708), Xent 0.3865, Xent Color 0.2256. Loss 2.1135, Error 0.1115(best: 0.0986), Error Color 0.0281(best: 0.0032)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 1060 | Time 10.1463(10.3948) | Bit/dim 1.9573(2.0048) | Xent 0.5829(0.5853) | Xent Color 0.3091(0.7883) | Loss 4.8385(5.8491) | Error 0.1789(0.1846) | Error Color 0.0844(0.2266) |Steps 470(458.57) | Grad Norm 4.2805(14.6382) | Total Time 0.00(0.00)\nIter 1070 | Time 10.4849(10.3174) | Bit/dim 1.9164(1.9854) | Xent 0.5039(0.5729) | Xent Color 0.2587(0.6548) | Loss 4.7758(5.5692) | Error 0.1689(0.1802) | Error Color 0.0589(0.1871) |Steps 434(455.54) | Grad Norm 4.3223(11.7948) | Total Time 0.00(0.00)\nIter 1080 | Time 10.1013(10.2491) | Bit/dim 1.9398(1.9710) | Xent 0.5117(0.5652) | Xent Color 0.1877(0.5389) | Loss 4.7566(5.3512) | Error 0.1678(0.1780) | Error Color 0.0344(0.1515) |Steps 440(450.92) | Grad Norm 1.5271(9.3968) | Total Time 0.00(0.00)\nIter 1090 | Time 10.0345(10.1756) | Bit/dim 1.8993(1.9567) | Xent 0.5515(0.5547) | Xent Color 0.1800(0.4451) | Loss 4.6080(5.1747) | Error 0.1800(0.1746) | Error Color 0.0400(0.1217) |Steps 446(449.98) | Grad Norm 5.1568(7.8671) | Total Time 0.00(0.00)\nIter 1100 | Time 9.8408(10.0937) | Bit/dim 1.8862(1.9446) | Xent 0.4684(0.5368) | Xent Color 0.1607(0.3722) | Loss 4.6977(5.0361) | Error 0.1522(0.1696) | Error Color 0.0367(0.0995) |Steps 470(448.02) | Grad Norm 1.9512(6.6309) | Total Time 0.00(0.00)\nIter 1110 | Time 10.0474(10.0635) | Bit/dim 1.8848(1.9338) | Xent 0.5305(0.5306) | Xent Color 0.1321(0.3138) | Loss 4.6870(4.9353) | Error 0.1689(0.1678) | Error Color 0.0300(0.0822) |Steps 464(446.28) | Grad Norm 2.0182(6.1817) | Total Time 0.00(0.00)\nIter 1120 | Time 10.4883(10.0633) | Bit/dim 1.8636(1.9204) | Xent 0.4987(0.5226) | Xent Color 0.1371(0.2686) | Loss 4.5417(4.8475) | Error 0.1433(0.1649) | Error Color 0.0344(0.0685) |Steps 464(445.79) | Grad Norm 4.9133(5.9002) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0017 | Time 65.4989, Epoch Time 749.9574(675.8566), Bit/dim 1.8936(best: 1.8708), Xent 0.3526, Xent Color 0.0725. Loss 1.9999, Error 0.1054(best: 0.0986), Error Color 0.0055(best: 0.0032)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 1130 | Time 10.1681(10.0673) | Bit/dim 1.8864(1.9103) | Xent 0.5354(0.5150) | Xent Color 0.1192(0.2322) | Loss 4.6361(5.2581) | Error 0.1567(0.1621) | Error Color 0.0256(0.0579) |Steps 458(442.59) | Grad Norm 6.3032(5.7457) | Total Time 0.00(0.00)\nIter 1140 | Time 10.0575(10.0400) | Bit/dim 1.8893(1.9014) | Xent 0.5804(0.5121) | Xent Color 0.0897(0.2016) | Loss 4.6937(5.0932) | Error 0.1767(0.1600) | Error Color 0.0122(0.0493) |Steps 440(443.17) | Grad Norm 1.1176(5.2114) | Total Time 0.00(0.00)\nIter 1150 | Time 9.6653(10.0419) | Bit/dim 1.8532(1.8916) | Xent 0.4365(0.5000) | Xent Color 0.1106(0.1765) | Loss 4.5525(4.9604) | Error 0.1233(0.1567) | Error Color 0.0267(0.0418) |Steps 458(444.52) | Grad Norm 3.6062(4.6835) | Total Time 0.00(0.00)\nIter 1160 | Time 9.8308(9.9833) | Bit/dim 1.8513(1.8846) | Xent 0.4282(0.4911) | Xent Color 0.0979(0.1553) | Loss 4.5076(4.8616) | Error 0.1300(0.1523) | Error Color 0.0222(0.0357) |Steps 422(441.02) | Grad Norm 3.0018(4.0295) | Total Time 0.00(0.00)\nIter 1170 | Time 9.6432(9.9606) | Bit/dim 1.8621(1.8756) | Xent 0.4850(0.4915) | Xent Color 0.0931(0.1393) | Loss 4.6277(4.7815) | Error 0.1644(0.1524) | Error Color 0.0144(0.0314) |Steps 428(437.52) | Grad Norm 3.9596(3.9596) | Total Time 0.00(0.00)\nIter 1180 | Time 9.2510(9.9429) | Bit/dim 1.8427(1.8676) | Xent 0.4270(0.4822) | Xent Color 0.0869(0.1264) | Loss 4.5443(4.7215) | Error 0.1322(0.1502) | Error Color 0.0222(0.0287) |Steps 434(437.79) | Grad Norm 7.5967(4.5563) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0018 | Time 67.5054, Epoch Time 746.8974(677.9878), Bit/dim 1.8453(best: 1.8708), Xent 0.3305, Xent Color 0.0427. Loss 1.9387, Error 0.0973(best: 0.0986), Error Color 0.0044(best: 0.0032)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 1190 | Time 10.1824(9.9412) | Bit/dim 1.8388(1.8610) | Xent 0.4431(0.4768) | Xent Color 0.0920(0.1171) | Loss 4.5423(5.2448) | Error 0.1333(0.1486) | Error Color 0.0200(0.0262) |Steps 440(435.70) | Grad Norm 5.6429(5.1003) | Total Time 0.00(0.00)\nIter 1200 | Time 10.3162(9.9635) | Bit/dim 1.8459(1.8539) | Xent 0.4069(0.4691) | Xent Color 0.0817(0.1074) | Loss 4.5155(5.0583) | Error 0.1333(0.1455) | Error Color 0.0167(0.0234) |Steps 452(438.44) | Grad Norm 9.4718(5.2399) | Total Time 0.00(0.00)\nIter 1210 | Time 9.4049(9.9426) | Bit/dim 1.8181(1.8473) | Xent 0.3744(0.4636) | Xent Color 0.0826(0.1021) | Loss 4.3795(4.9052) | Error 0.1222(0.1442) | Error Color 0.0211(0.0223) |Steps 428(438.11) | Grad Norm 8.0320(6.3285) | Total Time 0.00(0.00)\nIter 1220 | Time 9.6611(9.8660) | Bit/dim 1.8051(1.8389) | Xent 0.4537(0.4639) | Xent Color 0.0836(0.0946) | Loss 4.5339(4.7916) | Error 0.1522(0.1447) | Error Color 0.0167(0.0204) |Steps 440(438.35) | Grad Norm 9.5114(6.0232) | Total Time 0.00(0.00)\nIter 1230 | Time 10.2137(9.8156) | Bit/dim 1.8186(1.8329) | Xent 0.4690(0.4604) | Xent Color 0.0797(0.0897) | Loss 4.5100(4.7006) | Error 0.1489(0.1436) | Error Color 0.0167(0.0195) |Steps 434(437.53) | Grad Norm 8.3272(6.3474) | Total Time 0.00(0.00)\nIter 1240 | Time 9.8491(9.8270) | Bit/dim 1.8196(1.8253) | Xent 0.4630(0.4579) | Xent Color 0.0835(0.0859) | Loss 4.4431(4.6304) | Error 0.1500(0.1442) | Error Color 0.0200(0.0187) |Steps 434(436.98) | Grad Norm 11.6574(6.9939) | Total Time 0.00(0.00)\nIter 1250 | Time 10.2548(9.8693) | Bit/dim 1.8081(1.8166) | Xent 0.4386(0.4532) | Xent Color 0.0590(0.0801) | Loss 4.4754(4.5750) | Error 0.1300(0.1431) | Error Color 0.0100(0.0172) |Steps 434(435.15) | Grad Norm 4.8559(6.7727) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0019 | Time 65.4746, Epoch Time 738.5180(679.8037), Bit/dim 1.7912(best: 1.8453), Xent 0.3014, Xent Color 0.0278. Loss 1.8735, Error 0.0888(best: 0.0973), Error Color 0.0015(best: 0.0032)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 1260 | Time 10.0549(9.8854) | Bit/dim 1.7608(1.8070) | Xent 0.4680(0.4506) | Xent Color 0.0685(0.0753) | Loss 4.3793(5.0409) | Error 0.1500(0.1425) | Error Color 0.0189(0.0161) |Steps 446(433.84) | Grad Norm 6.3717(6.2720) | Total Time 0.00(0.00)\nIter 1270 | Time 9.7888(9.8830) | Bit/dim 1.7779(1.8004) | Xent 0.4289(0.4457) | Xent Color 0.0578(0.0717) | Loss 4.3557(4.8629) | Error 0.1400(0.1404) | Error Color 0.0122(0.0155) |Steps 434(432.70) | Grad Norm 5.0653(6.3611) | Total Time 0.00(0.00)\nIter 1280 | Time 9.5061(9.8549) | Bit/dim 1.7995(1.7930) | Xent 0.3734(0.4442) | Xent Color 0.9662(0.0987) | Loss 4.7890(4.7492) | Error 0.1244(0.1399) | Error Color 0.3156(0.0251) |Steps 404(432.00) | Grad Norm 130.9594(11.0568) | Total Time 0.00(0.00)\nIter 1290 | Time 11.0816(10.0268) | Bit/dim 2.0478(1.8870) | Xent 0.8469(0.5529) | Xent Color 1.5619(1.0244) | Loss 5.8629(5.3159) | Error 0.2833(0.1781) | Error Color 0.5467(0.1731) |Steps 494(441.78) | Grad Norm 19.8707(28.9034) | Total Time 0.00(0.00)\nIter 1300 | Time 12.2225(10.6073) | Bit/dim 1.9553(1.9086) | Xent 0.5780(0.5967) | Xent Color 0.7770(1.0044) | Loss 5.4686(5.3834) | Error 0.1833(0.1930) | Error Color 0.2989(0.2229) |Steps 560(470.64) | Grad Norm 53.1587(26.8076) | Total Time 0.00(0.00)\nIter 1310 | Time 12.0484(11.0618) | Bit/dim 1.8772(1.9088) | Xent 0.5537(0.5983) | Xent Color 0.5335(0.9004) | Loss 5.1145(5.3524) | Error 0.1822(0.1913) | Error Color 0.1889(0.2231) |Steps 536(495.55) | Grad Norm 7.8364(37.4357) | Total Time 0.00(0.00)\nIter 1320 | Time 11.3798(11.3357) | Bit/dim 1.8520(1.8943) | Xent 0.5194(0.5889) | Xent Color 0.3734(0.7793) | Loss 4.9295(5.2653) | Error 0.1833(0.1887) | Error Color 0.1189(0.2043) |Steps 518(504.86) | Grad Norm 3.2076(31.2316) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0020 | Time 69.4055, Epoch Time 819.7647(684.0025), Bit/dim 1.8551(best: 1.7912), Xent 0.3654, Xent Color 0.2492. Loss 2.0087, Error 0.1116(best: 0.0888), Error Color 0.0431(best: 0.0015)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 1330 | Time 10.6615(11.1700) | Bit/dim 1.8045(1.8773) | Xent 0.4502(0.5709) | Xent Color 0.2625(0.6639) | Loss 4.7234(5.7595) | Error 0.1422(0.1824) | Error Color 0.0578(0.1789) |Steps 488(500.31) | Grad Norm 2.8765(24.2814) | Total Time 0.00(0.00)\nIter 1340 | Time 10.5896(10.9899) | Bit/dim 1.8041(1.8571) | Xent 0.5435(0.5552) | Xent Color 0.2278(0.5572) | Loss 4.5962(5.4667) | Error 0.1744(0.1773) | Error Color 0.0522(0.1503) |Steps 482(493.56) | Grad Norm 3.5998(18.8485) | Total Time 0.00(0.00)\nIter 1350 | Time 10.5911(10.8870) | Bit/dim 1.7709(1.8403) | Xent 0.4719(0.5388) | Xent Color 0.1830(0.4608) | Loss 4.4392(5.2259) | Error 0.1556(0.1727) | Error Color 0.0367(0.1224) |Steps 470(487.62) | Grad Norm 2.3257(14.4280) | Total Time 0.00(0.00)\nIter 1360 | Time 10.4566(10.7605) | Bit/dim 1.7677(1.8226) | Xent 0.4449(0.5235) | Xent Color 0.1667(0.3856) | Loss 4.4318(5.0222) | Error 0.1444(0.1674) | Error Color 0.0422(0.1020) |Steps 464(481.56) | Grad Norm 8.8613(12.3380) | Total Time 0.00(0.00)\nIter 1370 | Time 10.0449(10.6168) | Bit/dim 1.7133(1.8041) | Xent 0.4770(0.5099) | Xent Color 0.1748(0.3257) | Loss 4.3153(4.8523) | Error 0.1411(0.1627) | Error Color 0.0511(0.0857) |Steps 458(474.30) | Grad Norm 11.2445(11.0540) | Total Time 0.00(0.00)\nIter 1380 | Time 10.8412(10.6133) | Bit/dim 1.7427(1.7873) | Xent 0.4293(0.4906) | Xent Color 0.1217(0.2772) | Loss 4.3659(4.7122) | Error 0.1333(0.1567) | Error Color 0.0233(0.0722) |Steps 476(468.75) | Grad Norm 5.8912(10.1031) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0021 | Time 61.4770, Epoch Time 779.7184(686.8740), Bit/dim 1.7337(best: 1.7912), Xent 0.3153, Xent Color 0.0677. Loss 1.8294, Error 0.0944(best: 0.0888), Error Color 0.0085(best: 0.0015)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 1390 | Time 10.8641(10.6235) | Bit/dim 1.7295(1.7722) | Xent 0.5226(0.4853) | Xent Color 0.1130(0.2364) | Loss 4.3443(5.1321) | Error 0.1589(0.1546) | Error Color 0.0300(0.0609) |Steps 440(464.38) | Grad Norm 5.1180(8.8831) | Total Time 0.00(0.00)\nIter 1400 | Time 9.9560(10.5498) | Bit/dim 1.7113(1.7564) | Xent 0.4259(0.4797) | Xent Color 0.1044(0.2036) | Loss 4.1957(4.9019) | Error 0.1411(0.1519) | Error Color 0.0278(0.0515) |Steps 446(460.26) | Grad Norm 1.7798(7.8331) | Total Time 0.00(0.00)\nIter 1410 | Time 10.7834(10.5279) | Bit/dim 1.7041(1.7434) | Xent 0.4087(0.4672) | Xent Color 0.0964(0.1748) | Loss 4.2710(4.7269) | Error 0.1400(0.1484) | Error Color 0.0200(0.0438) |Steps 470(456.10) | Grad Norm 3.7130(6.7916) | Total Time 0.00(0.00)\nIter 1420 | Time 10.0972(10.4721) | Bit/dim 1.6948(1.7304) | Xent 0.4419(0.4604) | Xent Color 0.0894(0.1537) | Loss 4.2427(4.5957) | Error 0.1522(0.1467) | Error Color 0.0267(0.0383) |Steps 446(452.59) | Grad Norm 8.7802(6.8365) | Total Time 0.00(0.00)\nIter 1430 | Time 10.4856(10.4040) | Bit/dim 1.6552(1.7162) | Xent 0.4129(0.4586) | Xent Color 0.0707(0.1351) | Loss 4.2130(4.4936) | Error 0.1400(0.1462) | Error Color 0.0156(0.0331) |Steps 458(451.09) | Grad Norm 4.0481(6.1251) | Total Time 0.00(0.00)\nIter 1440 | Time 9.9936(10.3703) | Bit/dim 1.6634(1.7023) | Xent 0.4256(0.4498) | Xent Color 0.0811(0.1217) | Loss 4.1582(4.4027) | Error 0.1389(0.1432) | Error Color 0.0200(0.0295) |Steps 392(445.91) | Grad Norm 3.0063(6.4069) | Total Time 0.00(0.00)\nIter 1450 | Time 10.3397(10.3513) | Bit/dim 1.6315(1.6873) | Xent 0.4142(0.4410) | Xent Color 0.1114(0.1096) | Loss 4.0053(4.3222) | Error 0.1411(0.1407) | Error Color 0.0278(0.0261) |Steps 458(443.64) | Grad Norm 14.5687(6.7656) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0022 | Time 62.9821, Epoch Time 767.9363(689.3059), Bit/dim 1.6432(best: 1.7337), Xent 0.2891, Xent Color 0.0274. Loss 1.7223, Error 0.0881(best: 0.0888), Error Color 0.0020(best: 0.0015)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 1460 | Time 10.0633(10.3080) | Bit/dim 1.6214(1.6738) | Xent 0.3947(0.4296) | Xent Color 0.0704(0.0998) | Loss 4.0569(4.7327) | Error 0.1189(0.1373) | Error Color 0.0167(0.0236) |Steps 404(439.19) | Grad Norm 10.4894(7.3675) | Total Time 0.00(0.00)\nIter 1470 | Time 10.2523(10.2567) | Bit/dim 1.6127(1.6594) | Xent 0.4291(0.4252) | Xent Color 0.0591(0.0912) | Loss 4.0157(4.5524) | Error 0.1378(0.1358) | Error Color 0.0133(0.0214) |Steps 422(436.19) | Grad Norm 7.1890(7.5960) | Total Time 0.00(0.00)\nIter 1480 | Time 10.3116(10.2423) | Bit/dim 1.8385(1.6586) | Xent 0.4860(0.4296) | Xent Color 3.2933(0.3747) | Loss 5.9462(4.5680) | Error 0.1433(0.1366) | Error Color 0.6222(0.0653) |Steps 464(433.35) | Grad Norm 91.6240(20.7798) | Total Time 0.00(0.00)\nIter 1490 | Time 11.8057(10.6166) | Bit/dim 1.8785(1.7307) | Xent 0.5243(0.4681) | Xent Color 0.9022(0.5822) | Loss 5.2026(4.7681) | Error 0.1578(0.1488) | Error Color 0.3767(0.1562) |Steps 554(456.20) | Grad Norm 22.0677(23.9255) | Total Time 0.00(0.00)\nIter 1500 | Time 11.3599(10.7758) | Bit/dim 1.7272(1.7389) | Xent 0.5089(0.4800) | Xent Color 0.5630(0.5955) | Loss 4.5945(4.7638) | Error 0.1767(0.1527) | Error Color 0.2100(0.1806) |Steps 488(467.88) | Grad Norm 10.8173(20.7590) | Total Time 0.00(0.00)\nIter 1510 | Time 10.8252(10.8340) | Bit/dim 1.6442(1.7203) | Xent 0.4569(0.4777) | Xent Color 0.3407(0.5549) | Loss 4.3068(4.6805) | Error 0.1422(0.1519) | Error Color 0.1244(0.1777) |Steps 440(468.06) | Grad Norm 3.6537(17.0549) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0023 | Time 67.3688, Epoch Time 799.2027(692.6028), Bit/dim 1.6290(best: 1.6432), Xent 0.2977, Xent Color 0.1813. Loss 1.7487, Error 0.0886(best: 0.0881), Error Color 0.0381(best: 0.0015)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 1520 | Time 11.3099(10.8290) | Bit/dim 1.6265(1.6966) | Xent 0.4106(0.4750) | Xent Color 0.2763(0.4919) | Loss 4.2799(5.2049) | Error 0.1411(0.1503) | Error Color 0.0967(0.1608) |Steps 488(468.47) | Grad Norm 4.7597(13.8233) | Total Time 0.00(0.00)\nIter 1530 | Time 10.8410(10.8382) | Bit/dim 1.6060(1.6725) | Xent 0.4574(0.4631) | Xent Color 0.2480(0.4292) | Loss 4.2548(4.9491) | Error 0.1422(0.1466) | Error Color 0.0844(0.1409) |Steps 476(470.99) | Grad Norm 7.3784(11.4601) | Total Time 0.00(0.00)\nIter 1540 | Time 11.2530(10.8414) | Bit/dim 1.5601(1.6494) | Xent 0.5047(0.4562) | Xent Color 0.1769(0.3684) | Loss 4.1286(4.7385) | Error 0.1578(0.1437) | Error Color 0.0500(0.1205) |Steps 482(472.19) | Grad Norm 5.0254(9.5706) | Total Time 0.00(0.00)\nIter 1550 | Time 11.0349(10.8389) | Bit/dim 1.5544(1.6267) | Xent 0.4614(0.4505) | Xent Color 0.1278(0.3138) | Loss 4.0877(4.5768) | Error 0.1544(0.1422) | Error Color 0.0278(0.1012) |Steps 482(474.68) | Grad Norm 1.3833(7.7774) | Total Time 0.00(0.00)\nIter 1560 | Time 10.9531(10.8457) | Bit/dim 1.5228(1.6052) | Xent 0.4508(0.4443) | Xent Color 0.1380(0.2674) | Loss 4.0479(4.4392) | Error 0.1356(0.1405) | Error Color 0.0411(0.0847) |Steps 476(474.92) | Grad Norm 3.3626(6.3929) | Total Time 0.00(0.00)\nIter 1570 | Time 11.2369(10.8472) | Bit/dim 1.5309(1.5862) | Xent 0.4138(0.4362) | Xent Color 0.1078(0.2287) | Loss 4.0365(4.3280) | Error 0.1189(0.1374) | Error Color 0.0233(0.0702) |Steps 482(472.45) | Grad Norm 5.7928(5.9460) | Total Time 0.00(0.00)\nIter 1580 | Time 11.0123(10.8581) | Bit/dim 1.5204(1.5683) | Xent 0.4126(0.4241) | Xent Color 0.0845(0.1963) | Loss 4.0103(4.2320) | Error 0.1267(0.1339) | Error Color 0.0211(0.0588) |Steps 476(470.47) | Grad Norm 4.0500(5.8972) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0024 | Time 63.7600, Epoch Time 802.1201(695.8883), Bit/dim 1.5125(best: 1.6290), Xent 0.2608, Xent Color 0.0421. Loss 1.5882, Error 0.0776(best: 0.0881), Error Color 0.0030(best: 0.0015)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 1590 | Time 10.6728(10.8014) | Bit/dim 1.5085(1.5527) | Xent 0.3607(0.4155) | Xent Color 0.0803(0.1678) | Loss 3.9470(4.6703) | Error 0.1200(0.1323) | Error Color 0.0144(0.0480) |Steps 482(469.44) | Grad Norm 5.9171(5.6315) | Total Time 0.00(0.00)\nIter 1600 | Time 10.4735(10.7862) | Bit/dim 1.4794(1.5390) | Xent 0.3817(0.4025) | Xent Color 0.0723(0.1432) | Loss 3.8686(4.4731) | Error 0.1233(0.1278) | Error Color 0.0156(0.0400) |Steps 446(467.56) | Grad Norm 1.8761(4.9943) | Total Time 0.00(0.00)\nIter 1610 | Time 10.1747(10.7077) | Bit/dim 1.4652(1.5236) | Xent 0.3509(0.3999) | Xent Color 0.0823(0.1249) | Loss 3.7517(4.3085) | Error 0.1233(0.1269) | Error Color 0.0267(0.0341) |Steps 458(467.18) | Grad Norm 8.0624(4.7682) | Total Time 0.00(0.00)\nIter 1620 | Time 10.9256(10.6540) | Bit/dim 1.4709(1.5091) | Xent 0.3612(0.3918) | Xent Color 0.0674(0.1088) | Loss 3.8012(4.1789) | Error 0.1244(0.1251) | Error Color 0.0111(0.0286) |Steps 482(466.70) | Grad Norm 5.8153(4.5728) | Total Time 0.00(0.00)\nIter 1630 | Time 10.7002(10.6515) | Bit/dim 1.4628(1.4958) | Xent 0.3770(0.3854) | Xent Color 0.0584(0.0955) | Loss 3.7908(4.0771) | Error 0.1178(0.1226) | Error Color 0.0111(0.0244) |Steps 470(464.78) | Grad Norm 6.7987(4.8117) | Total Time 0.00(0.00)\nIter 1640 | Time 10.6101(10.6416) | Bit/dim 1.4471(1.4829) | Xent 0.3592(0.3813) | Xent Color 0.0488(0.0845) | Loss 3.7891(3.9944) | Error 0.1100(0.1214) | Error Color 0.0067(0.0208) |Steps 458(462.82) | Grad Norm 4.5723(5.0824) | Total Time 0.00(0.00)\nIter 1650 | Time 10.2190(10.6700) | Bit/dim 1.4268(1.4712) | Xent 0.3277(0.3747) | Xent Color 0.0628(0.0770) | Loss 3.6863(3.9394) | Error 0.1078(0.1196) | Error Color 0.0156(0.0181) |Steps 446(461.89) | Grad Norm 12.0591(5.7447) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0025 | Time 63.7553, Epoch Time 787.8342(698.6467), Bit/dim 1.4406(best: 1.5125), Xent 0.2303, Xent Color 0.0277. Loss 1.5051, Error 0.0681(best: 0.0776), Error Color 0.0030(best: 0.0015)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 1660 | Time 10.5589(10.6289) | Bit/dim 1.4272(1.4605) | Xent 0.3613(0.3744) | Xent Color 0.0648(0.0715) | Loss 3.6763(4.3508) | Error 0.1244(0.1184) | Error Color 0.0167(0.0165) |Steps 428(459.72) | Grad Norm 11.9885(6.5228) | Total Time 0.00(0.00)\nIter 1670 | Time 10.1024(10.5641) | Bit/dim 1.4198(1.4512) | Xent 0.3248(0.3658) | Xent Color 0.0599(0.0652) | Loss 3.6408(4.1827) | Error 0.1033(0.1156) | Error Color 0.0167(0.0146) |Steps 446(459.62) | Grad Norm 15.9009(7.3713) | Total Time 0.00(0.00)\nIter 1680 | Time 10.7272(10.5560) | Bit/dim 1.8146(1.5225) | Xent 0.7990(0.3806) | Xent Color 1.0064(0.6126) | Loss 5.1858(4.4722) | Error 0.2667(0.1212) | Error Color 0.4122(0.1198) |Steps 488(461.33) | Grad Norm 36.3684(32.1239) | Total Time 0.00(0.00)\nIter 1690 | Time 11.3191(10.7113) | Bit/dim 1.7268(1.5876) | Xent 0.3950(0.3836) | Xent Color 0.4712(0.6232) | Loss 4.5729(4.5445) | Error 0.1311(0.1229) | Error Color 0.1811(0.1553) |Steps 482(468.62) | Grad Norm 8.6082(27.3734) | Total Time 0.00(0.00)\nIter 1700 | Time 11.3540(10.8664) | Bit/dim 1.5930(1.6046) | Xent 0.4256(0.3929) | Xent Color 0.3205(0.5645) | Loss 4.2546(4.5090) | Error 0.1411(0.1271) | Error Color 0.1078(0.1522) |Steps 494(474.41) | Grad Norm 3.3526(21.6519) | Total Time 0.00(0.00)\nIter 1710 | Time 10.6015(10.8816) | Bit/dim 1.5068(1.5875) | Xent 0.3495(0.3851) | Xent Color 0.2025(0.4872) | Loss 3.9034(4.4031) | Error 0.1167(0.1242) | Error Color 0.0622(0.1347) |Steps 470(474.75) | Grad Norm 2.3386(17.1110) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0026 | Time 67.4678, Epoch Time 801.4857(701.7319), Bit/dim 1.4864(best: 1.4406), Xent 0.2286, Xent Color 0.0956. Loss 1.5675, Error 0.0692(best: 0.0681), Error Color 0.0107(best: 0.0015)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 1720 | Time 11.0920(10.8559) | Bit/dim 1.4809(1.5625) | Xent 0.3310(0.3795) | Xent Color 0.1505(0.4076) | Loss 3.9369(4.9067) | Error 0.1067(0.1214) | Error Color 0.0422(0.1136) |Steps 464(475.10) | Grad Norm 1.8869(13.4196) | Total Time 0.00(0.00)\nIter 1730 | Time 10.3236(10.8365) | Bit/dim 1.4444(1.5360) | Xent 0.3329(0.3725) | Xent Color 0.1251(0.3359) | Loss 3.7861(4.6297) | Error 0.1122(0.1198) | Error Color 0.0333(0.0928) |Steps 458(471.87) | Grad Norm 1.6695(10.5350) | Total Time 0.00(0.00)\nIter 1740 | Time 10.4816(10.7692) | Bit/dim 1.4468(1.5119) | Xent 0.3627(0.3621) | Xent Color 0.0943(0.2774) | Loss 3.8147(4.4211) | Error 0.1100(0.1166) | Error Color 0.0222(0.0759) |Steps 476(470.79) | Grad Norm 3.2942(8.4806) | Total Time 0.00(0.00)\nIter 1750 | Time 10.6355(10.6659) | Bit/dim 1.4132(1.4893) | Xent 0.3393(0.3515) | Xent Color 0.0956(0.2308) | Loss 3.7851(4.2445) | Error 0.1056(0.1123) | Error Color 0.0200(0.0620) |Steps 458(465.40) | Grad Norm 2.6602(6.8658) | Total Time 0.00(0.00)\nIter 1760 | Time 10.1077(10.5862) | Bit/dim 1.4208(1.4690) | Xent 0.3121(0.3407) | Xent Color 0.0718(0.1922) | Loss 3.6106(4.0970) | Error 0.1089(0.1086) | Error Color 0.0144(0.0504) |Steps 410(459.29) | Grad Norm 1.4382(5.6283) | Total Time 0.00(0.00)\nIter 1770 | Time 9.7391(10.4586) | Bit/dim 1.4020(1.4530) | Xent 0.3177(0.3335) | Xent Color 0.0767(0.1624) | Loss 3.6610(3.9738) | Error 0.1033(0.1058) | Error Color 0.0233(0.0419) |Steps 440(453.36) | Grad Norm 1.9866(5.1364) | Total Time 0.00(0.00)\nIter 1780 | Time 10.1746(10.3801) | Bit/dim 1.3849(1.4371) | Xent 0.2866(0.3286) | Xent Color 0.0721(0.1373) | Loss 3.5932(3.8799) | Error 0.1022(0.1041) | Error Color 0.0122(0.0348) |Steps 446(446.86) | Grad Norm 1.1129(4.6004) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0027 | Time 63.6583, Epoch Time 774.0552(703.9016), Bit/dim 1.3922(best: 1.4406), Xent 0.1944, Xent Color 0.0257. Loss 1.4473, Error 0.0609(best: 0.0681), Error Color 0.0012(best: 0.0015)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 1790 | Time 10.0713(10.2612) | Bit/dim 1.3834(1.4258) | Xent 0.3025(0.3269) | Xent Color 0.0648(0.1186) | Loss 3.5570(4.2913) | Error 0.1011(0.1051) | Error Color 0.0133(0.0295) |Steps 404(441.42) | Grad Norm 3.3275(4.2114) | Total Time 0.00(0.00)\nIter 1800 | Time 9.6213(10.1681) | Bit/dim 1.3835(1.4136) | Xent 0.3010(0.3214) | Xent Color 0.0581(0.1038) | Loss 3.5690(4.1040) | Error 0.1022(0.1031) | Error Color 0.0111(0.0252) |Steps 422(438.90) | Grad Norm 6.0030(4.7781) | Total Time 0.00(0.00)\nIter 1810 | Time 10.1876(10.1509) | Bit/dim 1.3674(1.4015) | Xent 0.2601(0.3114) | Xent Color 0.0443(0.0911) | Loss 3.5528(3.9599) | Error 0.0956(0.1008) | Error Color 0.0078(0.0217) |Steps 458(439.51) | Grad Norm 2.6926(4.4417) | Total Time 0.00(0.00)\nIter 1820 | Time 10.1891(10.1695) | Bit/dim 1.3553(1.3913) | Xent 0.2625(0.3068) | Xent Color 0.0495(0.0805) | Loss 3.5431(3.8535) | Error 0.0811(0.0985) | Error Color 0.0089(0.0183) |Steps 434(437.52) | Grad Norm 2.6302(3.9918) | Total Time 0.00(0.00)\nIter 1830 | Time 10.4139(10.1755) | Bit/dim 1.3534(1.3829) | Xent 0.2804(0.3035) | Xent Color 0.0524(0.0720) | Loss 3.5033(3.7728) | Error 0.0900(0.0981) | Error Color 0.0100(0.0160) |Steps 452(437.09) | Grad Norm 1.7018(3.4445) | Total Time 0.00(0.00)\nIter 1840 | Time 10.2529(10.1567) | Bit/dim 1.3574(1.3741) | Xent 0.2895(0.2980) | Xent Color 0.0383(0.0645) | Loss 3.5120(3.7048) | Error 0.0956(0.0966) | Error Color 0.0067(0.0139) |Steps 428(436.18) | Grad Norm 1.6553(3.2773) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0028 | Time 62.7857, Epoch Time 752.2048(705.3507), Bit/dim 1.3478(best: 1.3922), Xent 0.1787, Xent Color 0.0163. Loss 1.3966, Error 0.0562(best: 0.0609), Error Color 0.0009(best: 0.0012)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 1850 | Time 10.2656(10.1637) | Bit/dim 1.3447(1.3675) | Xent 0.3255(0.2988) | Xent Color 0.0450(0.0593) | Loss 3.4642(4.2204) | Error 0.0933(0.0960) | Error Color 0.0111(0.0125) |Steps 446(435.38) | Grad Norm 3.5596(3.4322) | Total Time 0.00(0.00)\nIter 1860 | Time 9.8071(10.1092) | Bit/dim 1.3372(1.3599) | Xent 0.2692(0.2958) | Xent Color 0.0347(0.0544) | Loss 3.5086(4.0254) | Error 0.0944(0.0955) | Error Color 0.0033(0.0110) |Steps 428(433.73) | Grad Norm 2.9107(3.4930) | Total Time 0.00(0.00)\nIter 1870 | Time 10.1905(10.1021) | Bit/dim 1.3362(1.3519) | Xent 0.2703(0.2939) | Xent Color 0.0516(0.0508) | Loss 3.5183(3.8801) | Error 0.0933(0.0949) | Error Color 0.0133(0.0104) |Steps 440(434.21) | Grad Norm 3.6545(3.4229) | Total Time 0.00(0.00)\nIter 1880 | Time 9.9582(10.0967) | Bit/dim 1.3271(1.3451) | Xent 0.2890(0.2901) | Xent Color 0.0263(0.0476) | Loss 3.3198(3.7657) | Error 0.0867(0.0936) | Error Color 0.0022(0.0092) |Steps 410(431.71) | Grad Norm 2.5046(3.2449) | Total Time 0.00(0.00)\nIter 1890 | Time 9.9263(10.1187) | Bit/dim 1.3073(1.3375) | Xent 0.2611(0.2825) | Xent Color 0.0337(0.0453) | Loss 3.3953(3.6773) | Error 0.0833(0.0915) | Error Color 0.0044(0.0087) |Steps 422(429.37) | Grad Norm 4.1807(3.3581) | Total Time 0.00(0.00)\nIter 1900 | Time 10.2864(10.1006) | Bit/dim 1.3201(1.3322) | Xent 0.2803(0.2799) | Xent Color 0.0276(0.0425) | Loss 3.3719(3.6100) | Error 0.0878(0.0901) | Error Color 0.0022(0.0081) |Steps 428(429.71) | Grad Norm 7.6358(3.8557) | Total Time 0.00(0.00)\nIter 1910 | Time 10.9099(10.0990) | Bit/dim 1.3051(1.3258) | Xent 0.2642(0.2778) | Xent Color 0.0273(0.0415) | Loss 3.5011(3.5605) | Error 0.0844(0.0892) | Error Color 0.0033(0.0080) |Steps 434(429.39) | Grad Norm 2.2276(4.9871) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0029 | Time 63.8571, Epoch Time 751.2987(706.7291), Bit/dim 1.3038(best: 1.3478), Xent 0.1675, Xent Color 0.0152. Loss 1.3495, Error 0.0515(best: 0.0562), Error Color 0.0008(best: 0.0009)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 1920 | Time 9.7226(10.0579) | Bit/dim 1.3090(1.3196) | Xent 0.2643(0.2737) | Xent Color 0.0361(0.0397) | Loss 3.3950(3.9939) | Error 0.0878(0.0887) | Error Color 0.0056(0.0074) |Steps 434(428.69) | Grad Norm 3.0889(5.2995) | Total Time 0.00(0.00)\nIter 1930 | Time 9.8496(9.9875) | Bit/dim 1.3059(1.3135) | Xent 0.2405(0.2704) | Xent Color 0.0378(0.0385) | Loss 3.4114(3.8296) | Error 0.0678(0.0866) | Error Color 0.0056(0.0071) |Steps 416(428.78) | Grad Norm 3.8538(5.4746) | Total Time 0.00(0.00)\nIter 1940 | Time 9.8119(9.9506) | Bit/dim 1.2835(1.3055) | Xent 0.2457(0.2670) | Xent Color 0.0376(0.0364) | Loss 3.3472(3.7024) | Error 0.0833(0.0850) | Error Color 0.0078(0.0064) |Steps 428(425.51) | Grad Norm 4.9931(4.9850) | Total Time 0.00(0.00)\nIter 1950 | Time 9.9121(9.9499) | Bit/dim 1.2805(1.2985) | Xent 0.2258(0.2617) | Xent Color 0.0401(0.0351) | Loss 3.2639(3.6088) | Error 0.0756(0.0844) | Error Color 0.0100(0.0062) |Steps 386(423.97) | Grad Norm 5.3077(5.0814) | Total Time 0.00(0.00)\nIter 1960 | Time 9.4864(9.9164) | Bit/dim 1.2645(1.2917) | Xent 0.2581(0.2644) | Xent Color 0.0322(0.0335) | Loss 3.2996(3.5392) | Error 0.0889(0.0862) | Error Color 0.0044(0.0059) |Steps 386(422.43) | Grad Norm 5.8611(4.7377) | Total Time 0.00(0.00)\nIter 1970 | Time 9.8230(9.9165) | Bit/dim 1.2660(1.2843) | Xent 0.2619(0.2617) | Xent Color 0.0316(0.0329) | Loss 3.3205(3.4792) | Error 0.0800(0.0844) | Error Color 0.0056(0.0056) |Steps 434(421.80) | Grad Norm 6.7399(5.5887) | Total Time 0.00(0.00)\nIter 1980 | Time 9.7297(9.9146) | Bit/dim 1.2857(1.2797) | Xent 0.2365(0.2611) | Xent Color 0.0361(0.0330) | Loss 3.3130(3.4364) | Error 0.0744(0.0841) | Error Color 0.0078(0.0058) |Steps 410(423.34) | Grad Norm 13.7749(7.3325) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0030 | Time 62.4518, Epoch Time 736.2992(707.6162), Bit/dim 1.2690(best: 1.3038), Xent 0.1576, Xent Color 0.0132. Loss 1.3117, Error 0.0517(best: 0.0515), Error Color 0.0010(best: 0.0008)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 1990 | Time 10.0191(9.8896) | Bit/dim 1.2652(1.2739) | Xent 0.2751(0.2586) | Xent Color 0.0437(0.0327) | Loss 3.3332(3.8301) | Error 0.0889(0.0832) | Error Color 0.0122(0.0060) |Steps 428(423.14) | Grad Norm 15.5137(7.9520) | Total Time 0.00(0.00)\nIter 2000 | Time 10.3006(9.8920) | Bit/dim 1.2300(1.2666) | Xent 0.2296(0.2543) | Xent Color 0.0315(0.0312) | Loss 3.2665(3.6866) | Error 0.0678(0.0809) | Error Color 0.0056(0.0058) |Steps 422(421.63) | Grad Norm 4.5016(7.5746) | Total Time 0.00(0.00)\nIter 2010 | Time 9.8735(9.8718) | Bit/dim 1.2190(1.2569) | Xent 0.2115(0.2468) | Xent Color 0.0258(0.0298) | Loss 3.1929(3.5644) | Error 0.0678(0.0791) | Error Color 0.0056(0.0052) |Steps 416(421.14) | Grad Norm 5.6267(7.1701) | Total Time 0.00(0.00)\nIter 2020 | Time 10.2935(9.8959) | Bit/dim 1.2076(1.2477) | Xent 0.2436(0.2473) | Xent Color 0.0225(0.0284) | Loss 3.1533(3.4762) | Error 0.0756(0.0792) | Error Color 0.0044(0.0050) |Steps 452(421.52) | Grad Norm 1.7587(6.4590) | Total Time 0.00(0.00)\nIter 2030 | Time 9.8719(9.8883) | Bit/dim 1.2075(1.2363) | Xent 0.2401(0.2423) | Xent Color 0.0228(0.0275) | Loss 3.1591(3.3974) | Error 0.0811(0.0779) | Error Color 0.0033(0.0048) |Steps 410(420.00) | Grad Norm 4.2834(5.7865) | Total Time 0.00(0.00)\nIter 2040 | Time 9.7017(9.8615) | Bit/dim 1.2631(1.2320) | Xent 0.2464(0.2439) | Xent Color 0.0790(0.0296) | Loss 3.3610(3.3549) | Error 0.0867(0.0788) | Error Color 0.0278(0.0057) |Steps 416(419.46) | Grad Norm 33.9761(7.6979) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0031 | Time 64.5348, Epoch Time 737.3258(708.5075), Bit/dim 2.1069(best: 1.2690), Xent 2.4425, Xent Color 3.5489. Loss 3.6047, Error 0.6689(best: 0.0515), Error Color 0.7436(best: 0.0008)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 2050 | Time 11.4719(9.9931) | Bit/dim 1.9940(1.3855) | Xent 1.0502(0.4537) | Xent Color 1.6618(0.7301) | Loss 5.8807(4.5456) | Error 0.3300(0.1322) | Error Color 0.5100(0.1272) |Steps 500(425.61) | Grad Norm 33.3734(30.4530) | Total Time 0.00(0.00)\nIter 2060 | Time 10.2090(10.1409) | Bit/dim 1.8117(1.5093) | Xent 0.5017(0.5160) | Xent Color 0.4190(0.7052) | Loss 4.6304(4.6476) | Error 0.1633(0.1546) | Error Color 0.1567(0.1599) |Steps 446(433.14) | Grad Norm 7.4566(25.2362) | Total Time 0.00(0.00)\nIter 2070 | Time 10.3240(10.1685) | Bit/dim 1.5631(1.5445) | Xent 0.4775(0.5068) | Xent Color 0.2740(0.6075) | Loss 4.0571(4.5508) | Error 0.1533(0.1550) | Error Color 0.0978(0.1497) |Steps 410(431.71) | Grad Norm 3.4849(19.9980) | Total Time 0.00(0.00)\nIter 2080 | Time 9.9788(10.1222) | Bit/dim 1.4324(1.5320) | Xent 0.3659(0.4740) | Xent Color 0.1649(0.5016) | Loss 3.7485(4.3854) | Error 0.1233(0.1466) | Error Color 0.0500(0.1274) |Steps 416(429.50) | Grad Norm 3.0758(15.7535) | Total Time 0.00(0.00)\nIter 2090 | Time 9.5606(10.0531) | Bit/dim 1.3566(1.4958) | Xent 0.3238(0.4381) | Xent Color 0.1235(0.4081) | Loss 3.6169(4.2009) | Error 0.1044(0.1364) | Error Color 0.0367(0.1048) |Steps 428(426.15) | Grad Norm 1.9801(12.3836) | Total Time 0.00(0.00)\nIter 2100 | Time 9.4048(10.0201) | Bit/dim 1.3151(1.4536) | Xent 0.3779(0.4089) | Xent Color 0.1071(0.3313) | Loss 3.5368(4.0317) | Error 0.1156(0.1272) | Error Color 0.0322(0.0853) |Steps 422(426.71) | Grad Norm 2.6107(9.9385) | Total Time 0.00(0.00)\nIter 2110 | Time 10.1181(10.0133) | Bit/dim 1.2643(1.4107) | Xent 0.2563(0.3807) | Xent Color 0.0892(0.2692) | Loss 3.4213(3.8790) | Error 0.0878(0.1192) | Error Color 0.0211(0.0689) |Steps 446(426.75) | Grad Norm 4.1994(8.0581) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0032 | Time 61.7884, Epoch Time 755.0395(709.9034), Bit/dim 1.2772(best: 1.2690), Xent 0.1802, Xent Color 0.0368. Loss 1.3315, Error 0.0571(best: 0.0515), Error Color 0.0025(best: 0.0008)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 2120 | Time 9.7413(9.9806) | Bit/dim 1.2557(1.3734) | Xent 0.2998(0.3551) | Xent Color 0.0703(0.2199) | Loss 3.3859(4.2346) | Error 0.1067(0.1121) | Error Color 0.0089(0.0552) |Steps 434(427.18) | Grad Norm 1.9258(6.6215) | Total Time 0.00(0.00)\nIter 2130 | Time 9.8337(9.9295) | Bit/dim 1.2421(1.3418) | Xent 0.2442(0.3325) | Xent Color 0.0568(0.1780) | Loss 3.3162(3.9967) | Error 0.0778(0.1040) | Error Color 0.0078(0.0437) |Steps 404(423.78) | Grad Norm 1.5098(5.3820) | Total Time 0.00(0.00)\nIter 2140 | Time 9.8488(9.9175) | Bit/dim 1.2251(1.3125) | Xent 0.2706(0.3137) | Xent Color 0.0598(0.1470) | Loss 3.2969(3.8133) | Error 0.0978(0.0985) | Error Color 0.0078(0.0351) |Steps 440(423.02) | Grad Norm 2.0878(4.4894) | Total Time 0.00(0.00)\nIter 2150 | Time 10.1331(9.9692) | Bit/dim 1.2060(1.2871) | Xent 0.3043(0.3039) | Xent Color 0.0489(0.1219) | Loss 3.2867(3.6719) | Error 0.1067(0.0967) | Error Color 0.0078(0.0281) |Steps 440(424.48) | Grad Norm 2.3548(3.8127) | Total Time 0.00(0.00)\nIter 2160 | Time 9.8118(9.9331) | Bit/dim 1.1912(1.2646) | Xent 0.2563(0.2912) | Xent Color 0.0397(0.1032) | Loss 3.1634(3.5499) | Error 0.0722(0.0928) | Error Color 0.0044(0.0232) |Steps 386(420.95) | Grad Norm 2.6257(3.5553) | Total Time 0.00(0.00)\nIter 2170 | Time 10.1594(9.8912) | Bit/dim 1.1858(1.2456) | Xent 0.2944(0.2870) | Xent Color 0.0429(0.0882) | Loss 3.1830(3.4572) | Error 0.0978(0.0916) | Error Color 0.0067(0.0193) |Steps 416(419.79) | Grad Norm 4.1198(3.6987) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0033 | Time 62.2389, Epoch Time 735.6644(710.6763), Bit/dim 1.1836(best: 1.2690), Xent 0.1539, Xent Color 0.0148. Loss 1.2258, Error 0.0475(best: 0.0515), Error Color 0.0005(best: 0.0008)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 2180 | Time 9.6602(9.8639) | Bit/dim 1.1927(1.2310) | Xent 0.2147(0.2774) | Xent Color 0.0465(0.0759) | Loss 3.1710(3.9397) | Error 0.0733(0.0895) | Error Color 0.0133(0.0161) |Steps 416(419.56) | Grad Norm 3.9995(3.4793) | Total Time 0.00(0.00)\nIter 2190 | Time 9.6172(9.8038) | Bit/dim 1.1707(1.2173) | Xent 0.3076(0.2704) | Xent Color 0.0371(0.0668) | Loss 3.1910(3.7318) | Error 0.0844(0.0866) | Error Color 0.0067(0.0137) |Steps 416(418.65) | Grad Norm 3.1594(3.6499) | Total Time 0.00(0.00)\nIter 2200 | Time 9.2424(9.7881) | Bit/dim 1.1614(1.2038) | Xent 0.2110(0.2622) | Xent Color 0.0347(0.0596) | Loss 3.1255(3.5777) | Error 0.0700(0.0843) | Error Color 0.0067(0.0121) |Steps 422(418.48) | Grad Norm 4.1251(3.8501) | Total Time 0.00(0.00)\nIter 2210 | Time 9.6080(9.7318) | Bit/dim 1.1753(1.1931) | Xent 0.2813(0.2587) | Xent Color 0.0400(0.0536) | Loss 3.1512(3.4610) | Error 0.0811(0.0832) | Error Color 0.0067(0.0105) |Steps 428(418.06) | Grad Norm 7.1129(3.8984) | Total Time 0.00(0.00)\nIter 2220 | Time 9.8741(9.7708) | Bit/dim 1.1551(1.1824) | Xent 0.2290(0.2540) | Xent Color 0.0344(0.0488) | Loss 3.1393(3.3708) | Error 0.0656(0.0809) | Error Color 0.0056(0.0095) |Steps 434(418.54) | Grad Norm 2.0300(3.8230) | Total Time 0.00(0.00)\nIter 2230 | Time 9.3988(9.7393) | Bit/dim 1.1466(1.1727) | Xent 0.1765(0.2438) | Xent Color 0.0313(0.0448) | Loss 3.1235(3.2942) | Error 0.0533(0.0773) | Error Color 0.0022(0.0086) |Steps 422(416.93) | Grad Norm 4.3794(3.9318) | Total Time 0.00(0.00)\nIter 2240 | Time 9.7513(9.7288) | Bit/dim 1.1286(1.1645) | Xent 0.2263(0.2360) | Xent Color 0.0262(0.0407) | Loss 3.0467(3.2322) | Error 0.0678(0.0745) | Error Color 0.0011(0.0074) |Steps 416(415.82) | Grad Norm 4.6442(3.6141) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0034 | Time 61.8092, Epoch Time 723.9513(711.0745), Bit/dim 1.1346(best: 1.1836), Xent 0.1396, Xent Color 0.0106. Loss 1.1722, Error 0.0450(best: 0.0475), Error Color 0.0003(best: 0.0005)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 2250 | Time 9.9605(9.7246) | Bit/dim 1.1364(1.1557) | Xent 0.2008(0.2344) | Xent Color 0.0316(0.0391) | Loss 3.0265(3.6982) | Error 0.0667(0.0746) | Error Color 0.0044(0.0075) |Steps 398(416.67) | Grad Norm 4.0524(3.7066) | Total Time 0.00(0.00)\nIter 2260 | Time 9.6750(9.7137) | Bit/dim 1.1302(1.1492) | Xent 0.2307(0.2295) | Xent Color 0.0348(0.0373) | Loss 3.1043(3.5280) | Error 0.0811(0.0736) | Error Color 0.0111(0.0071) |Steps 422(415.88) | Grad Norm 7.5547(3.9539) | Total Time 0.00(0.00)\nIter 2270 | Time 9.4189(9.7306) | Bit/dim 1.1122(1.1423) | Xent 0.2488(0.2300) | Xent Color 0.0310(0.0356) | Loss 3.0152(3.4018) | Error 0.0767(0.0733) | Error Color 0.0078(0.0066) |Steps 428(416.51) | Grad Norm 4.8857(4.2189) | Total Time 0.00(0.00)\nIter 2280 | Time 9.9815(9.7714) | Bit/dim 1.1134(1.1368) | Xent 0.2462(0.2274) | Xent Color 0.0393(0.0344) | Loss 3.0603(3.3115) | Error 0.0611(0.0714) | Error Color 0.0089(0.0061) |Steps 416(416.31) | Grad Norm 8.6953(4.6368) | Total Time 0.00(0.00)\nIter 2290 | Time 9.6117(9.8010) | Bit/dim 1.0977(1.1291) | Xent 0.2516(0.2247) | Xent Color 0.0319(0.0328) | Loss 2.9586(3.2290) | Error 0.0689(0.0705) | Error Color 0.0056(0.0057) |Steps 392(414.90) | Grad Norm 5.7328(4.9206) | Total Time 0.00(0.00)\nIter 2300 | Time 9.2555(9.7443) | Bit/dim 1.1092(1.1225) | Xent 0.2274(0.2219) | Xent Color 0.0265(0.0308) | Loss 2.9835(3.1649) | Error 0.0711(0.0698) | Error Color 0.0056(0.0051) |Steps 416(415.99) | Grad Norm 2.5049(4.3441) | Total Time 0.00(0.00)\nIter 2310 | Time 9.7926(9.7119) | Bit/dim 1.0949(1.1164) | Xent 0.2084(0.2182) | Xent Color 0.0283(0.0297) | Loss 2.9188(3.1132) | Error 0.0667(0.0688) | Error Color 0.0033(0.0049) |Steps 410(413.98) | Grad Norm 6.2273(4.2906) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0035 | Time 61.2910, Epoch Time 727.2759(711.5606), Bit/dim 1.0973(best: 1.1346), Xent 0.1300, Xent Color 0.0080. Loss 1.1318, Error 0.0440(best: 0.0450), Error Color 0.0002(best: 0.0003)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 2320 | Time 9.7029(9.6985) | Bit/dim 1.1043(1.1114) | Xent 0.1727(0.2165) | Xent Color 0.0235(0.0287) | Loss 3.0047(3.5032) | Error 0.0522(0.0684) | Error Color 0.0067(0.0049) |Steps 422(414.30) | Grad Norm 3.8727(4.1444) | Total Time 0.00(0.00)\nIter 2330 | Time 9.7642(9.6916) | Bit/dim 1.0843(1.1055) | Xent 0.1947(0.2134) | Xent Color 0.0253(0.0281) | Loss 2.9849(3.3619) | Error 0.0667(0.0675) | Error Color 0.0022(0.0050) |Steps 410(413.99) | Grad Norm 7.4909(4.5104) | Total Time 0.00(0.00)\nIter 2340 | Time 9.6672(9.7357) | Bit/dim 1.0827(1.1002) | Xent 0.1915(0.2106) | Xent Color 0.0189(0.0277) | Loss 2.9263(3.2569) | Error 0.0611(0.0664) | Error Color 0.0022(0.0050) |Steps 440(416.48) | Grad Norm 2.7897(4.9824) | Total Time 0.00(0.00)\nIter 2350 | Time 10.2639(9.7361) | Bit/dim 1.0793(1.0947) | Xent 0.1893(0.2079) | Xent Color 0.0280(0.0271) | Loss 2.9536(3.1687) | Error 0.0678(0.0663) | Error Color 0.0067(0.0049) |Steps 422(416.25) | Grad Norm 4.8626(4.8281) | Total Time 0.00(0.00)\nIter 2360 | Time 9.8634(9.7864) | Bit/dim 1.0734(1.0896) | Xent 0.1882(0.2029) | Xent Color 0.0358(0.0263) | Loss 2.8794(3.0991) | Error 0.0567(0.0642) | Error Color 0.0089(0.0048) |Steps 404(416.22) | Grad Norm 4.6684(4.4615) | Total Time 0.00(0.00)\nIter 2370 | Time 9.7149(9.7711) | Bit/dim 1.0682(1.0846) | Xent 0.2175(0.2025) | Xent Color 0.0222(0.0254) | Loss 2.9243(3.0560) | Error 0.0744(0.0644) | Error Color 0.0033(0.0044) |Steps 416(414.89) | Grad Norm 5.3931(4.2988) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0036 | Time 69.9213, Epoch Time 737.2601(712.3316), Bit/dim 2.9427(best: 1.0973), Xent 0.5492, Xent Color 14.6926. Loss 6.7531, Error 0.1682(best: 0.0440), Error Color 0.7722(best: 0.0002)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 2380 | Time 9.5014(9.8285) | Bit/dim 1.9872(1.2046) | Xent 0.7397(0.2502) | Xent Color 1.3971(0.7714) | Loss 5.4572(4.1613) | Error 0.2389(0.0789) | Error Color 0.3889(0.0862) |Steps 410(417.78) | Grad Norm 61.7220(26.8109) | Total Time 0.00(0.00)\nIter 2390 | Time 10.4311(9.8520) | Bit/dim 1.9182(1.4052) | Xent 0.3951(0.3123) | Xent Color 0.5192(0.7595) | Loss 4.9314(4.4002) | Error 0.1089(0.0995) | Error Color 0.1978(0.1356) |Steps 452(421.42) | Grad Norm 10.1583(26.6980) | Total Time 0.00(0.00)\nIter 2400 | Time 10.3912(9.8856) | Bit/dim 1.7745(1.5139) | Xent 0.2988(0.3238) | Xent Color 0.3245(0.6611) | Loss 4.3863(4.4445) | Error 0.0922(0.1037) | Error Color 0.1167(0.1375) |Steps 422(423.75) | Grad Norm 4.1629(21.3450) | Total Time 0.00(0.00)\nIter 2410 | Time 10.0214(9.9454) | Bit/dim 1.6951(1.5686) | Xent 0.2300(0.3099) | Xent Color 0.1556(0.5414) | Loss 4.1379(4.3911) | Error 0.0767(0.0992) | Error Color 0.0433(0.1167) |Steps 446(428.29) | Grad Norm 2.1595(16.5409) | Total Time 0.00(0.00)\nIter 2420 | Time 10.5438(10.0628) | Bit/dim 1.6092(1.5851) | Xent 0.2432(0.2995) | Xent Color 0.0984(0.4309) | Loss 3.9837(4.2988) | Error 0.0722(0.0959) | Error Color 0.0256(0.0939) |Steps 458(432.77) | Grad Norm 2.5642(12.8995) | Total Time 0.00(0.00)\nIter 2430 | Time 10.4977(10.1604) | Bit/dim 1.4924(1.5714) | Xent 0.2036(0.2842) | Xent Color 0.0980(0.3435) | Loss 3.7266(4.1798) | Error 0.0633(0.0911) | Error Color 0.0278(0.0759) |Steps 452(437.30) | Grad Norm 1.6580(10.0502) | Total Time 0.00(0.00)\nIter 2440 | Time 10.6080(10.2186) | Bit/dim 1.3888(1.5341) | Xent 0.2313(0.2671) | Xent Color 0.1049(0.2785) | Loss 3.6111(4.0484) | Error 0.0711(0.0856) | Error Color 0.0244(0.0623) |Steps 470(439.60) | Grad Norm 2.5435(8.0721) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0037 | Time 62.6210, Epoch Time 758.2331(713.7086), Bit/dim 1.3559(best: 1.0973), Xent 0.1340, Xent Color 0.0395. Loss 1.3993, Error 0.0422(best: 0.0440), Error Color 0.0038(best: 0.0002)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 2450 | Time 9.7957(10.1705) | Bit/dim 1.2751(1.4784) | Xent 0.2217(0.2552) | Xent Color 0.0890(0.2290) | Loss 3.2939(4.3592) | Error 0.0656(0.0817) | Error Color 0.0256(0.0519) |Steps 440(440.54) | Grad Norm 3.8020(6.6138) | Total Time 0.00(0.00)\nIter 2460 | Time 9.9927(10.1526) | Bit/dim 1.1973(1.4172) | Xent 0.1985(0.2436) | Xent Color 0.0646(0.1880) | Loss 3.2729(4.0868) | Error 0.0767(0.0793) | Error Color 0.0178(0.0429) |Steps 440(441.01) | Grad Norm 1.8684(5.5110) | Total Time 0.00(0.00)\nIter 2470 | Time 10.4066(10.1404) | Bit/dim 1.1649(1.3578) | Xent 0.2289(0.2402) | Xent Color 0.0789(0.1563) | Loss 3.1673(3.8584) | Error 0.0733(0.0775) | Error Color 0.0200(0.0360) |Steps 434(441.26) | Grad Norm 4.0790(4.7856) | Total Time 0.00(0.00)\nIter 2480 | Time 9.9581(10.1688) | Bit/dim 1.1552(1.3071) | Xent 0.1931(0.2307) | Xent Color 0.0440(0.1286) | Loss 3.0893(3.6761) | Error 0.0589(0.0741) | Error Color 0.0100(0.0293) |Steps 428(438.83) | Grad Norm 2.6073(4.2315) | Total Time 0.00(0.00)\nIter 2490 | Time 10.4849(10.2052) | Bit/dim 1.1345(1.2639) | Xent 0.2375(0.2250) | Xent Color 0.0496(0.1075) | Loss 3.1602(3.5246) | Error 0.0789(0.0723) | Error Color 0.0100(0.0242) |Steps 458(437.58) | Grad Norm 4.4413(4.0263) | Total Time 0.00(0.00)\nIter 2500 | Time 9.8055(10.1243) | Bit/dim 1.1100(1.2278) | Xent 0.1843(0.2180) | Xent Color 0.0449(0.0905) | Loss 3.0341(3.4047) | Error 0.0689(0.0703) | Error Color 0.0100(0.0201) |Steps 422(434.34) | Grad Norm 4.2116(3.6959) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0038 | Time 62.0665, Epoch Time 752.1846(714.8629), Bit/dim 1.1058(best: 1.0973), Xent 0.1180, Xent Color 0.0124. Loss 1.1384, Error 0.0391(best: 0.0422), Error Color 0.0003(best: 0.0002)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 2510 | Time 9.8949(10.1124) | Bit/dim 1.1024(1.1962) | Xent 0.2335(0.2159) | Xent Color 0.0443(0.0771) | Loss 2.9929(3.8502) | Error 0.0667(0.0697) | Error Color 0.0089(0.0168) |Steps 410(431.30) | Grad Norm 1.5775(3.3313) | Total Time 0.00(0.00)\nIter 2520 | Time 10.1499(10.1120) | Bit/dim 1.0886(1.1702) | Xent 0.1822(0.2077) | Xent Color 0.0291(0.0663) | Loss 2.9475(3.6217) | Error 0.0522(0.0664) | Error Color 0.0033(0.0142) |Steps 428(431.33) | Grad Norm 3.2525(3.0791) | Total Time 0.00(0.00)\nIter 2530 | Time 9.6173(10.0753) | Bit/dim 1.0745(1.1484) | Xent 0.2095(0.2061) | Xent Color 0.0396(0.0582) | Loss 2.9691(3.4555) | Error 0.0656(0.0656) | Error Color 0.0078(0.0122) |Steps 410(429.27) | Grad Norm 4.0222(3.0317) | Total Time 0.00(0.00)\nIter 2540 | Time 9.7201(10.0340) | Bit/dim 1.0793(1.1311) | Xent 0.1788(0.2014) | Xent Color 0.0305(0.0509) | Loss 2.9900(3.3225) | Error 0.0611(0.0644) | Error Color 0.0033(0.0102) |Steps 410(428.77) | Grad Norm 1.7389(3.0723) | Total Time 0.00(0.00)\nIter 2550 | Time 9.6763(10.0286) | Bit/dim 1.0810(1.1162) | Xent 0.1697(0.1976) | Xent Color 0.0365(0.0459) | Loss 2.9829(3.2210) | Error 0.0489(0.0628) | Error Color 0.0078(0.0090) |Steps 422(427.30) | Grad Norm 1.6228(2.8957) | Total Time 0.00(0.00)\nIter 2560 | Time 10.2310(9.9878) | Bit/dim 1.0752(1.1036) | Xent 0.1658(0.1958) | Xent Color 0.0342(0.0414) | Loss 2.9080(3.1402) | Error 0.0567(0.0625) | Error Color 0.0067(0.0082) |Steps 410(425.43) | Grad Norm 4.4101(2.9189) | Total Time 0.00(0.00)\nIter 2570 | Time 9.7499(9.9659) | Bit/dim 1.0425(1.0927) | Xent 0.2005(0.1953) | Xent Color 0.0256(0.0379) | Loss 2.9020(3.0793) | Error 0.0600(0.0622) | Error Color 0.0033(0.0073) |Steps 422(426.17) | Grad Norm 3.2509(2.7828) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0039 | Time 61.5005, Epoch Time 741.1859(715.6526), Bit/dim 1.0577(best: 1.0973), Xent 0.1078, Xent Color 0.0070. Loss 1.0864, Error 0.0353(best: 0.0391), Error Color 0.0004(best: 0.0002)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 2580 | Time 9.6987(9.9560) | Bit/dim 1.0543(1.0832) | Xent 0.1674(0.1957) | Xent Color 0.0278(0.0347) | Loss 2.8852(3.5169) | Error 0.0578(0.0620) | Error Color 0.0089(0.0067) |Steps 416(425.69) | Grad Norm 3.8620(2.7988) | Total Time 0.00(0.00)\nIter 2590 | Time 9.2163(9.9217) | Bit/dim 1.0489(1.0753) | Xent 0.1860(0.1962) | Xent Color 0.0214(0.0320) | Loss 2.8220(3.3469) | Error 0.0578(0.0626) | Error Color 0.0011(0.0060) |Steps 410(422.23) | Grad Norm 1.3866(2.6407) | Total Time 0.00(0.00)\nIter 2600 | Time 10.1422(9.8507) | Bit/dim 1.0577(1.0692) | Xent 0.1967(0.1912) | Xent Color 0.0219(0.0299) | Loss 2.9217(3.2207) | Error 0.0589(0.0607) | Error Color 0.0044(0.0055) |Steps 416(420.17) | Grad Norm 3.2073(2.6961) | Total Time 0.00(0.00)\nIter 2610 | Time 9.6628(9.7972) | Bit/dim 1.0491(1.0622) | Xent 0.1626(0.1870) | Xent Color 0.0323(0.0289) | Loss 2.8208(3.1223) | Error 0.0556(0.0595) | Error Color 0.0100(0.0054) |Steps 422(418.32) | Grad Norm 5.2766(2.7426) | Total Time 0.00(0.00)\nIter 2620 | Time 9.7170(9.7699) | Bit/dim 1.0177(1.0563) | Xent 0.1689(0.1831) | Xent Color 0.0229(0.0269) | Loss 2.8137(3.0446) | Error 0.0556(0.0588) | Error Color 0.0067(0.0048) |Steps 410(413.78) | Grad Norm 3.9437(3.3058) | Total Time 0.00(0.00)\nIter 2630 | Time 9.6302(9.7749) | Bit/dim 1.0236(1.0506) | Xent 0.1675(0.1807) | Xent Color 0.0168(0.0255) | Loss 2.8068(2.9892) | Error 0.0600(0.0582) | Error Color 0.0000(0.0044) |Steps 416(414.88) | Grad Norm 2.0069(3.5191) | Total Time 0.00(0.00)\nIter 2640 | Time 9.8312(9.7970) | Bit/dim 1.0150(1.0454) | Xent 0.1873(0.1838) | Xent Color 0.0267(0.0245) | Loss 2.7958(2.9495) | Error 0.0533(0.0587) | Error Color 0.0089(0.0044) |Steps 428(416.45) | Grad Norm 3.5151(3.3749) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0040 | Time 62.1752, Epoch Time 729.4202(716.0656), Bit/dim 1.0298(best: 1.0577), Xent 0.0995, Xent Color 0.0060. Loss 1.0561, Error 0.0324(best: 0.0353), Error Color 0.0002(best: 0.0002)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 2650 | Time 9.6004(9.8310) | Bit/dim 1.0346(1.0408) | Xent 0.1604(0.1840) | Xent Color 0.0176(0.0237) | Loss 2.8614(3.3495) | Error 0.0500(0.0577) | Error Color 0.0011(0.0040) |Steps 422(417.77) | Grad Norm 1.9295(3.5032) | Total Time 0.00(0.00)\nIter 2660 | Time 10.1061(9.8660) | Bit/dim 1.0299(1.0373) | Xent 0.1555(0.1779) | Xent Color 0.0230(0.0228) | Loss 2.8260(3.2151) | Error 0.0500(0.0561) | Error Color 0.0044(0.0039) |Steps 434(417.91) | Grad Norm 8.2870(3.9972) | Total Time 0.00(0.00)\nIter 2670 | Time 10.0151(9.8910) | Bit/dim 1.0122(1.0328) | Xent 0.2055(0.1768) | Xent Color 0.0228(0.0224) | Loss 2.7496(3.1075) | Error 0.0622(0.0560) | Error Color 0.0056(0.0038) |Steps 398(418.03) | Grad Norm 4.3161(4.5639) | Total Time 0.00(0.00)\nIter 2680 | Time 10.0023(9.8830) | Bit/dim 1.0155(1.0290) | Xent 0.1779(0.1776) | Xent Color 0.0195(0.0218) | Loss 2.7620(3.0275) | Error 0.0644(0.0562) | Error Color 0.0044(0.0038) |Steps 398(417.47) | Grad Norm 5.4581(4.6408) | Total Time 0.00(0.00)\nIter 2690 | Time 9.8704(9.8873) | Bit/dim 1.0211(1.0252) | Xent 0.1955(0.1774) | Xent Color 0.0173(0.0210) | Loss 2.8317(2.9660) | Error 0.0600(0.0565) | Error Color 0.0011(0.0035) |Steps 428(417.11) | Grad Norm 4.1449(4.4552) | Total Time 0.00(0.00)\nIter 2700 | Time 10.0196(9.9116) | Bit/dim 1.0062(1.0213) | Xent 0.1789(0.1759) | Xent Color 0.0148(0.0204) | Loss 2.7911(2.9187) | Error 0.0556(0.0559) | Error Color 0.0011(0.0033) |Steps 428(419.69) | Grad Norm 6.1872(4.3752) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0041 | Time 61.8295, Epoch Time 740.2248(716.7904), Bit/dim 1.0078(best: 1.0298), Xent 0.1001, Xent Color 0.0037. Loss 1.0338, Error 0.0324(best: 0.0324), Error Color 0.0000(best: 0.0002)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 2710 | Time 10.0076(9.8967) | Bit/dim 1.0142(1.0180) | Xent 0.1568(0.1705) | Xent Color 0.0218(0.0195) | Loss 2.7943(3.4049) | Error 0.0567(0.0544) | Error Color 0.0022(0.0030) |Steps 410(420.86) | Grad Norm 9.0402(4.3940) | Total Time 0.00(0.00)\nIter 2720 | Time 10.3534(9.8738) | Bit/dim 1.0316(1.0172) | Xent 0.1767(0.1709) | Xent Color 0.0509(0.0214) | Loss 2.8781(3.2470) | Error 0.0533(0.0543) | Error Color 0.0156(0.0039) |Steps 446(420.52) | Grad Norm 24.8705(6.5514) | Total Time 0.00(0.00)\nIter 2740 | Time 9.9983(9.9270) | Bit/dim 0.9905(1.0120) | Xent 0.1842(0.1662) | Xent Color 0.0174(0.0218) | Loss 2.6859(3.0307) | Error 0.0633(0.0531) | Error Color 0.0033(0.0041) |Steps 410(421.41) | Grad Norm 6.9738(8.3075) | Total Time 0.00(0.00)\nIter 2750 | Time 10.2656(10.0061) | Bit/dim 0.9974(1.0084) | Xent 0.1319(0.1649) | Xent Color 0.0149(0.0203) | Loss 2.7447(2.9639) | Error 0.0344(0.0521) | Error Color 0.0011(0.0036) |Steps 434(421.19) | Grad Norm 2.0268(7.3544) | Total Time 0.00(0.00)\nIter 2760 | Time 10.4816(10.0375) | Bit/dim 1.0067(1.0060) | Xent 0.1769(0.1651) | Xent Color 0.0147(0.0187) | Loss 2.7996(2.9085) | Error 0.0589(0.0525) | Error Color 0.0011(0.0031) |Steps 428(419.25) | Grad Norm 4.3037(6.6123) | Total Time 0.00(0.00)\nIter 2770 | Time 9.6160(10.0401) | Bit/dim 0.9884(1.0019) | Xent 0.1781(0.1649) | Xent Color 0.0157(0.0175) | Loss 2.7477(2.8628) | Error 0.0544(0.0524) | Error Color 0.0022(0.0027) |Steps 416(419.78) | Grad Norm 2.9880(5.9501) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0042 | Time 63.0289, Epoch Time 747.4804(717.7111), Bit/dim 0.9899(best: 1.0078), Xent 0.0963, Xent Color 0.0036. Loss 1.0149, Error 0.0321(best: 0.0324), Error Color 0.0000(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 2780 | Time 9.8247(10.0020) | Bit/dim 0.9794(0.9987) | Xent 0.1687(0.1607) | Xent Color 0.0177(0.0171) | Loss 2.7128(3.2742) | Error 0.0567(0.0511) | Error Color 0.0044(0.0025) |Steps 398(418.19) | Grad Norm 7.6125(6.1036) | Total Time 0.00(0.00)\nIter 2790 | Time 10.0746(9.9989) | Bit/dim 0.9767(0.9947) | Xent 0.1553(0.1594) | Xent Color 0.0156(0.0164) | Loss 2.7047(3.1330) | Error 0.0444(0.0510) | Error Color 0.0022(0.0024) |Steps 416(419.05) | Grad Norm 6.2967(5.7498) | Total Time 0.00(0.00)\nIter 2800 | Time 9.7509(9.9832) | Bit/dim 0.9943(0.9920) | Xent 0.1184(0.1610) | Xent Color 0.0197(0.0159) | Loss 2.7269(3.0298) | Error 0.0389(0.0507) | Error Color 0.0056(0.0023) |Steps 422(421.20) | Grad Norm 15.8447(5.9318) | Total Time 0.00(0.00)\nIter 2810 | Time 10.0694(9.9408) | Bit/dim 0.9894(0.9921) | Xent 0.1497(0.1598) | Xent Color 0.0223(0.0193) | Loss 2.7834(2.9585) | Error 0.0511(0.0502) | Error Color 0.0044(0.0037) |Steps 428(422.66) | Grad Norm 10.8856(8.3513) | Total Time 0.00(0.00)\nIter 2820 | Time 10.2160(9.9268) | Bit/dim 0.9751(0.9895) | Xent 0.1477(0.1545) | Xent Color 0.0254(0.0191) | Loss 2.7304(2.8964) | Error 0.0533(0.0490) | Error Color 0.0067(0.0038) |Steps 428(421.65) | Grad Norm 12.0320(8.8072) | Total Time 0.00(0.00)\nIter 2830 | Time 10.0070(9.9186) | Bit/dim 0.9722(0.9859) | Xent 0.1635(0.1538) | Xent Color 0.0173(0.0181) | Loss 2.7359(2.8500) | Error 0.0456(0.0490) | Error Color 0.0033(0.0034) |Steps 428(421.00) | Grad Norm 9.0599(8.4845) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0043 | Time 62.3765, Epoch Time 737.9840(718.3193), Bit/dim 0.9723(best: 0.9899), Xent 0.0987, Xent Color 0.0028. Loss 0.9977, Error 0.0320(best: 0.0321), Error Color 0.0000(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 2840 | Time 10.1240(9.8998) | Bit/dim 0.9734(0.9825) | Xent 0.1738(0.1537) | Xent Color 0.0143(0.0171) | Loss 2.7582(3.3402) | Error 0.0489(0.0487) | Error Color 0.0022(0.0031) |Steps 416(419.35) | Grad Norm 9.2707(8.3527) | Total Time 0.00(0.00)\nIter 2850 | Time 10.1689(9.8942) | Bit/dim 0.9680(0.9787) | Xent 0.1208(0.1535) | Xent Color 0.0109(0.0161) | Loss 2.6434(3.1719) | Error 0.0467(0.0486) | Error Color 0.0000(0.0028) |Steps 416(418.73) | Grad Norm 5.2610(7.9240) | Total Time 0.00(0.00)\nIter 2860 | Time 9.8554(9.9284) | Bit/dim 0.9758(0.9767) | Xent 0.1516(0.1541) | Xent Color 0.0149(0.0153) | Loss 2.7611(3.0563) | Error 0.0544(0.0487) | Error Color 0.0011(0.0025) |Steps 434(419.22) | Grad Norm 7.0780(7.3810) | Total Time 0.00(0.00)\nIter 2870 | Time 9.7171(9.8992) | Bit/dim 0.9650(0.9730) | Xent 0.1336(0.1535) | Xent Color 0.0130(0.0146) | Loss 2.6984(2.9593) | Error 0.0500(0.0485) | Error Color 0.0011(0.0023) |Steps 416(419.01) | Grad Norm 9.8421(7.0148) | Total Time 0.00(0.00)\nIter 2880 | Time 10.0441(9.8821) | Bit/dim 0.9712(0.9699) | Xent 0.1691(0.1550) | Xent Color 0.0150(0.0145) | Loss 2.6560(2.8842) | Error 0.0611(0.0492) | Error Color 0.0000(0.0021) |Steps 422(417.88) | Grad Norm 17.8040(7.4188) | Total Time 0.00(0.00)\nIter 2890 | Time 10.5155(9.8903) | Bit/dim 0.9647(0.9692) | Xent 0.1632(0.1520) | Xent Color 0.0160(0.0156) | Loss 2.6506(2.8306) | Error 0.0567(0.0478) | Error Color 0.0022(0.0024) |Steps 386(415.33) | Grad Norm 12.6932(8.9222) | Total Time 0.00(0.00)\nIter 2900 | Time 10.3922(9.9516) | Bit/dim 0.9666(0.9668) | Xent 0.1377(0.1488) | Xent Color 0.0149(0.0158) | Loss 2.6366(2.7941) | Error 0.0500(0.0472) | Error Color 0.0022(0.0025) |Steps 428(418.67) | Grad Norm 8.0344(9.1717) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0044 | Time 62.1809, Epoch Time 740.9018(718.9967), Bit/dim 0.9572(best: 0.9723), Xent 0.0847, Xent Color 0.0024. Loss 0.9790, Error 0.0288(best: 0.0320), Error Color 0.0000(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 2910 | Time 10.0556(10.0122) | Bit/dim 0.9436(0.9640) | Xent 0.1640(0.1492) | Xent Color 0.0126(0.0146) | Loss 2.6887(3.2480) | Error 0.0533(0.0474) | Error Color 0.0000(0.0021) |Steps 428(420.98) | Grad Norm 8.1203(8.5676) | Total Time 0.00(0.00)\nIter 2920 | Time 10.1032(10.0211) | Bit/dim 0.9432(0.9609) | Xent 0.1366(0.1475) | Xent Color 0.0110(0.0138) | Loss 2.6250(3.0995) | Error 0.0433(0.0471) | Error Color 0.0000(0.0018) |Steps 434(422.85) | Grad Norm 6.6158(7.9120) | Total Time 0.00(0.00)\nIter 2930 | Time 9.5882(9.9721) | Bit/dim 0.9506(0.9602) | Xent 0.1265(0.1447) | Xent Color 0.0132(0.0143) | Loss 2.6713(2.9881) | Error 0.0411(0.0465) | Error Color 0.0011(0.0020) |Steps 398(421.34) | Grad Norm 13.3389(9.3025) | Total Time 0.00(0.00)\nIter 2940 | Time 10.1049(9.9406) | Bit/dim 0.9525(0.9588) | Xent 0.1075(0.1427) | Xent Color 0.0126(0.0138) | Loss 2.6758(2.9049) | Error 0.0311(0.0453) | Error Color 0.0011(0.0018) |Steps 422(419.66) | Grad Norm 8.2451(9.1872) | Total Time 0.00(0.00)\nIter 2950 | Time 10.4178(10.0161) | Bit/dim 0.9410(0.9553) | Xent 0.1703(0.1417) | Xent Color 0.0099(0.0138) | Loss 2.6682(2.8434) | Error 0.0511(0.0437) | Error Color 0.0022(0.0018) |Steps 398(420.56) | Grad Norm 7.2887(9.1235) | Total Time 0.00(0.00)\nIter 2960 | Time 9.5779(9.9620) | Bit/dim 0.9437(0.9538) | Xent 0.1530(0.1439) | Xent Color 0.0121(0.0144) | Loss 2.5809(2.7961) | Error 0.0444(0.0443) | Error Color 0.0022(0.0022) |Steps 416(420.39) | Grad Norm 8.7960(10.3386) | Total Time 0.00(0.00)\nIter 2970 | Time 10.0892(9.9201) | Bit/dim 0.9342(0.9500) | Xent 0.1379(0.1413) | Xent Color 0.0079(0.0139) | Loss 2.6642(2.7588) | Error 0.0467(0.0438) | Error Color 0.0000(0.0020) |Steps 440(419.93) | Grad Norm 7.0007(9.7874) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0045 | Time 62.3398, Epoch Time 741.1944(719.6627), Bit/dim 0.9397(best: 0.9572), Xent 0.0841, Xent Color 0.0018. Loss 0.9612, Error 0.0275(best: 0.0288), Error Color 0.0000(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 2980 | Time 10.6627(9.9408) | Bit/dim 0.9257(0.9467) | Xent 0.1462(0.1413) | Xent Color 0.0163(0.0133) | Loss 2.6874(3.1609) | Error 0.0433(0.0443) | Error Color 0.0044(0.0019) |Steps 434(420.47) | Grad Norm 7.7170(9.2322) | Total Time 0.00(0.00)\nIter 2990 | Time 13.6349(10.3682) | Bit/dim 2.6458(1.2452) | Xent 1.4467(0.4841) | Xent Color 7.6559(2.0915) | Loss 10.2100(4.7002) | Error 0.4678(0.1213) | Error Color 0.8144(0.1468) |Steps 596(443.35) | Grad Norm 108.6735(41.6873) | Total Time 0.00(0.00)\nIter 3000 | Time 15.3921(11.8508) | Bit/dim 2.3301(1.5363) | Xent 0.7281(0.7438) | Xent Color 1.4747(2.2121) | Loss 6.6568(5.4237) | Error 0.2489(0.2022) | Error Color 0.5122(0.2721) |Steps 680(504.78) | Grad Norm 1332.5699(473693509.9970) | Total Time 0.00(0.00)\nIter 3010 | Time 13.1624(12.1216) | Bit/dim 2.0117(1.6864) | Xent 0.7244(0.7225) | Xent Color 0.7986(1.9130) | Loss 5.5643(5.5392) | Error 0.2278(0.2054) | Error Color 0.3233(0.3081) |Steps 602(523.91) | Grad Norm 5.6010(349313025.0914) | Total Time 0.00(0.00)\nIter 3020 | Time 12.6405(12.1993) | Bit/dim 1.9104(1.7591) | Xent 0.4362(0.6531) | Xent Color 0.5649(1.5908) | Loss 5.0309(5.4522) | Error 0.1444(0.1905) | Error Color 0.2156(0.2983) |Steps 542(533.46) | Grad Norm 3.8594(257591853.5596) | Total Time 0.00(0.00)\nIter 3030 | Time 11.6638(12.2292) | Bit/dim 1.8741(1.7935) | Xent 0.3787(0.5784) | Xent Color 0.3747(1.2939) | Loss 4.8549(5.3025) | Error 0.1233(0.1712) | Error Color 0.1344(0.2651) |Steps 524(535.72) | Grad Norm 3.8450(189954448.6436) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0046 | Time 72.9126, Epoch Time 918.6792(725.6332), Bit/dim 1.8286(best: 0.9397), Xent 0.2107, Xent Color 0.1766. Loss 1.9255, Error 0.0651(best: 0.0275), Error Color 0.0458(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 3040 | Time 11.9673(12.1137) | Bit/dim 1.8132(1.8003) | Xent 0.3446(0.5117) | Xent Color 0.2435(1.0332) | Loss 4.6631(5.8348) | Error 0.1167(0.1541) | Error Color 0.0733(0.2235) |Steps 518(533.89) | Grad Norm 3.9150(140076994.6635) | Total Time 0.00(0.00)\nIter 3050 | Time 11.2961(11.9906) | Bit/dim 1.7668(1.7935) | Xent 0.2700(0.4576) | Xent Color 0.2044(0.8183) | Loss 4.5129(5.4945) | Error 0.0811(0.1397) | Error Color 0.0544(0.1821) |Steps 530(529.13) | Grad Norm 5.4982(103296156.4381) | Total Time 0.00(0.00)\nIter 3060 | Time 11.4660(11.8552) | Bit/dim 1.6871(1.7731) | Xent 0.2888(0.4148) | Xent Color 0.1819(0.6505) | Loss 4.3406(5.2099) | Error 0.0978(0.1284) | Error Color 0.0556(0.1480) |Steps 518(523.45) | Grad Norm 7.6199(76173079.8367) | Total Time 0.00(0.00)\nIter 3070 | Time 10.9618(11.6676) | Bit/dim 1.6454(1.7441) | Xent 0.2484(0.3768) | Xent Color 0.1711(0.5216) | Loss 4.1335(4.9563) | Error 0.0778(0.1177) | Error Color 0.0556(0.1206) |Steps 482(512.81) | Grad Norm 10.3675(56171868.7777) | Total Time 0.00(0.00)\nIter 3080 | Time 11.1656(11.5233) | Bit/dim 1.5764(1.7075) | Xent 0.2984(0.3476) | Xent Color 0.1128(0.4196) | Loss 4.0736(4.7451) | Error 0.0878(0.1083) | Error Color 0.0222(0.0987) |Steps 488(506.70) | Grad Norm 7.7450(41422493.1034) | Total Time 0.00(0.00)\nIter 3090 | Time 11.5132(11.4432) | Bit/dim 1.5065(1.6683) | Xent 0.2464(0.3196) | Xent Color 0.1138(0.3384) | Loss 3.9514(4.5600) | Error 0.0722(0.1000) | Error Color 0.0278(0.0801) |Steps 512(501.91) | Grad Norm 7.2884(30545947.4533) | Total Time 0.00(0.00)\nIter 3100 | Time 11.0463(11.3805) | Bit/dim 1.4971(1.6299) | Xent 0.2477(0.3006) | Xent Color 0.0909(0.2830) | Loss 3.8950(4.4069) | Error 0.0756(0.0942) | Error Color 0.0200(0.0685) |Steps 458(497.28) | Grad Norm 5.4977(22525321.8390) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0047 | Time 71.2535, Epoch Time 840.7912(729.0879), Bit/dim 1.4997(best: 0.9397), Xent 0.1462, Xent Color 0.0581. Loss 1.5507, Error 0.0476(best: 0.0275), Error Color 0.0070(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 3110 | Time 10.7467(11.3465) | Bit/dim 1.4814(1.5911) | Xent 0.2038(0.2818) | Xent Color 0.0793(0.2399) | Loss 3.8077(4.8396) | Error 0.0656(0.0891) | Error Color 0.0200(0.0595) |Steps 470(492.64) | Grad Norm 10.0661(16610719.7827) | Total Time 0.00(0.00)\nIter 3120 | Time 10.6995(11.3132) | Bit/dim 1.3975(1.5509) | Xent 0.2106(0.2662) | Xent Color 0.0862(0.2012) | Loss 3.6614(4.5636) | Error 0.0589(0.0837) | Error Color 0.0200(0.0497) |Steps 470(490.31) | Grad Norm 11.4004(12249147.9793) | Total Time 0.00(0.00)\nIter 3130 | Time 10.8139(11.2680) | Bit/dim 1.3651(1.5057) | Xent 0.1998(0.2580) | Xent Color 0.0779(0.1756) | Loss 3.6488(4.3394) | Error 0.0644(0.0813) | Error Color 0.0156(0.0449) |Steps 476(489.58) | Grad Norm 10.1359(9032820.8870) | Total Time 0.00(0.00)\nIter 3140 | Time 10.6613(11.1826) | Bit/dim 1.3018(1.4593) | Xent 0.2162(0.2523) | Xent Color 0.0707(0.1505) | Loss 3.4667(4.1379) | Error 0.0711(0.0795) | Error Color 0.0189(0.0385) |Steps 476(486.07) | Grad Norm 9.0679(6661022.7753) | Total Time 0.00(0.00)\nIter 3150 | Time 10.5930(11.1500) | Bit/dim 1.2830(1.4135) | Xent 0.2508(0.2456) | Xent Color 0.0666(0.1284) | Loss 3.4772(3.9681) | Error 0.0767(0.0771) | Error Color 0.0167(0.0323) |Steps 464(485.85) | Grad Norm 5.9934(4912000.2061) | Total Time 0.00(0.00)\nIter 3160 | Time 10.6010(11.1028) | Bit/dim 1.2521(1.3727) | Xent 0.2272(0.2399) | Xent Color 0.0512(0.1109) | Loss 3.4070(3.8265) | Error 0.0711(0.0764) | Error Color 0.0089(0.0271) |Steps 476(482.15) | Grad Norm 8.4584(3622229.0753) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0048 | Time 68.9469, Epoch Time 823.8692(731.9313), Bit/dim 1.2366(best: 0.9397), Xent 0.1272, Xent Color 0.0884. Loss 1.2905, Error 0.0422(best: 0.0275), Error Color 0.0242(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 3170 | Time 10.9526(11.0714) | Bit/dim 1.2271(1.3363) | Xent 0.1963(0.2281) | Xent Color 0.1566(0.1063) | Loss 3.3993(4.3844) | Error 0.0689(0.0733) | Error Color 0.0556(0.0276) |Steps 464(479.51) | Grad Norm 30.7495(2671123.6576) | Total Time 0.00(0.00)\nIter 3180 | Time 11.1038(11.0235) | Bit/dim 1.1958(1.3036) | Xent 0.1809(0.2223) | Xent Color 0.0679(0.0995) | Loss 3.2913(4.1089) | Error 0.0611(0.0709) | Error Color 0.0144(0.0260) |Steps 458(476.29) | Grad Norm 11.7938(1969754.5276) | Total Time 0.00(0.00)\nIter 3190 | Time 10.8111(11.0027) | Bit/dim 1.1809(1.2726) | Xent 0.1651(0.2169) | Xent Color 0.0626(0.0882) | Loss 3.2869(3.8900) | Error 0.0478(0.0689) | Error Color 0.0156(0.0227) |Steps 452(472.73) | Grad Norm 8.7296(1452546.8269) | Total Time 0.00(0.00)\nIter 3200 | Time 11.2599(11.0300) | Bit/dim 1.1519(1.2449) | Xent 0.2000(0.2138) | Xent Color 0.0474(0.0773) | Loss 3.2195(3.7177) | Error 0.0711(0.0679) | Error Color 0.0111(0.0195) |Steps 458(473.24) | Grad Norm 3.6667(1071145.2079) | Total Time 0.00(0.00)\nIter 3210 | Time 11.0787(11.0173) | Bit/dim 1.1374(1.2196) | Xent 0.1647(0.2075) | Xent Color 0.0418(0.0718) | Loss 3.1410(3.5819) | Error 0.0522(0.0664) | Error Color 0.0122(0.0183) |Steps 446(472.26) | Grad Norm 6.5489(789891.4762) | Total Time 0.00(0.00)\nIter 3220 | Time 10.9514(11.0320) | Bit/dim 1.1234(1.1962) | Xent 0.2085(0.2038) | Xent Color 0.0481(0.0642) | Loss 3.2026(3.4723) | Error 0.0822(0.0658) | Error Color 0.0089(0.0155) |Steps 470(470.98) | Grad Norm 8.4709(582487.2583) | Total Time 0.00(0.00)\nIter 3230 | Time 10.7576(11.0309) | Bit/dim 1.1016(1.1738) | Xent 0.1842(0.2011) | Xent Color 0.0485(0.0588) | Loss 3.1488(3.3855) | Error 0.0589(0.0646) | Error Color 0.0144(0.0140) |Steps 464(470.13) | Grad Norm 6.3236(429541.9672) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0049 | Time 70.4602, Epoch Time 818.0482(734.5149), Bit/dim 1.1034(best: 0.9397), Xent 0.1106, Xent Color 0.0112. Loss 1.1338, Error 0.0375(best: 0.0275), Error Color 0.0007(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 3240 | Time 11.1508(11.0093) | Bit/dim 1.0910(1.1549) | Xent 0.1872(0.1970) | Xent Color 0.0706(0.0550) | Loss 3.1752(3.8772) | Error 0.0600(0.0631) | Error Color 0.0222(0.0131) |Steps 440(468.82) | Grad Norm 19.0395(316756.9045) | Total Time 0.00(0.00)\nIter 3250 | Time 10.5911(11.0078) | Bit/dim 1.0956(1.1426) | Xent 0.2183(0.1970) | Xent Color 0.0371(0.0595) | Loss 3.1354(3.6862) | Error 0.0622(0.0626) | Error Color 0.0067(0.0152) |Steps 464(468.86) | Grad Norm 9.3085(233589.3635) | Total Time 0.00(0.00)\nIter 3260 | Time 10.5825(10.9236) | Bit/dim 1.0739(1.1268) | Xent 0.1849(0.1966) | Xent Color 0.0309(0.0554) | Loss 3.0051(3.5176) | Error 0.0611(0.0634) | Error Color 0.0067(0.0138) |Steps 428(465.58) | Grad Norm 4.0804(172256.7120) | Total Time 0.00(0.00)\nIter 3270 | Time 11.5750(10.8557) | Bit/dim 1.0725(1.1117) | Xent 0.1773(0.1908) | Xent Color 0.0227(0.0502) | Loss 3.1032(3.3860) | Error 0.0533(0.0614) | Error Color 0.0022(0.0119) |Steps 482(460.36) | Grad Norm 4.4026(127027.6096) | Total Time 0.00(0.00)\nIter 3280 | Time 10.3573(10.8024) | Bit/dim 1.0709(1.0999) | Xent 0.1539(0.1887) | Xent Color 0.0381(0.0457) | Loss 3.0279(3.2900) | Error 0.0589(0.0610) | Error Color 0.0089(0.0106) |Steps 476(458.61) | Grad Norm 4.3127(93674.5484) | Total Time 0.00(0.00)\nIter 3290 | Time 10.2460(10.7629) | Bit/dim 1.0500(1.0868) | Xent 0.1407(0.1837) | Xent Color 0.0402(0.0423) | Loss 2.9063(3.2100) | Error 0.0411(0.0591) | Error Color 0.0111(0.0093) |Steps 440(455.41) | Grad Norm 5.6571(69079.1887) | Total Time 0.00(0.00)\nIter 3300 | Time 11.3130(10.7662) | Bit/dim 1.0385(1.0769) | Xent 0.1320(0.1793) | Xent Color 0.0357(0.0391) | Loss 2.9339(3.1459) | Error 0.0444(0.0577) | Error Color 0.0078(0.0081) |Steps 434(452.54) | Grad Norm 4.0464(50941.9567) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0050 | Time 68.6077, Epoch Time 801.8720(736.5356), Bit/dim 1.0443(best: 0.9397), Xent 0.1002, Xent Color 0.0088. Loss 1.0716, Error 0.0337(best: 0.0275), Error Color 0.0002(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 3310 | Time 10.4147(10.7321) | Bit/dim 1.0363(1.0666) | Xent 0.2178(0.1778) | Xent Color 0.0442(0.0376) | Loss 2.9881(3.5888) | Error 0.0711(0.0566) | Error Color 0.0111(0.0075) |Steps 458(453.27) | Grad Norm 8.8812(37567.4434) | Total Time 0.00(0.00)\nIter 3320 | Time 10.6438(10.7419) | Bit/dim 1.0360(1.0583) | Xent 0.1510(0.1738) | Xent Color 0.0232(0.0350) | Loss 3.0738(3.4338) | Error 0.0478(0.0555) | Error Color 0.0033(0.0067) |Steps 470(454.51) | Grad Norm 10.5411(27704.8179) | Total Time 0.00(0.00)\nIter 3330 | Time 13.3089(10.9967) | Bit/dim 1.9930(1.2169) | Xent 0.5803(0.2475) | Xent Color 1.1236(0.6551) | Loss 5.5741(3.9394) | Error 0.2156(0.0782) | Error Color 0.4011(0.1007) |Steps 602(467.39) | Grad Norm 18.5798(20451.1492) | Total Time 0.00(0.00)\nIter 3340 | Time 12.5056(11.4567) | Bit/dim 1.9284(1.4116) | Xent 0.3478(0.2760) | Xent Color 0.7798(0.7549) | Loss 5.0710(4.2902) | Error 0.1078(0.0882) | Error Color 0.3233(0.1772) |Steps 524(489.54) | Grad Norm 4.3123(15083.7312) | Total Time 0.00(0.00)\nIter 3350 | Time 12.1846(11.7639) | Bit/dim 1.7462(1.5179) | Xent 0.2827(0.2788) | Xent Color 0.4569(0.7078) | Loss 4.6792(4.4234) | Error 0.0922(0.0891) | Error Color 0.1811(0.1904) |Steps 560(505.02) | Grad Norm 9.7637(11124.7679) | Total Time 0.00(0.00)\nIter 3360 | Time 11.9998(11.8867) | Bit/dim 1.6883(1.5688) | Xent 0.1925(0.2694) | Xent Color 0.2582(0.6126) | Loss 4.4252(4.4478) | Error 0.0567(0.0856) | Error Color 0.0778(0.1729) |Steps 548(514.01) | Grad Norm 4.2425(8204.8276) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0051 | Time 74.2080, Epoch Time 877.7962(740.7734), Bit/dim 1.6196(best: 0.9397), Xent 0.1455, Xent Color 0.1065. Loss 1.6826, Error 0.0443(best: 0.0275), Error Color 0.0135(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 3370 | Time 12.0566(11.9904) | Bit/dim 1.5680(1.5817) | Xent 0.1847(0.2591) | Xent Color 0.1802(0.5098) | Loss 4.0838(5.0818) | Error 0.0622(0.0818) | Error Color 0.0522(0.1458) |Steps 560(522.91) | Grad Norm 2.2570(6051.2279) | Total Time 0.00(0.00)\nIter 3380 | Time 12.0005(11.9557) | Bit/dim 1.5211(1.5737) | Xent 0.1609(0.2457) | Xent Color 0.1256(0.4136) | Loss 3.9361(4.8207) | Error 0.0511(0.0783) | Error Color 0.0389(0.1179) |Steps 518(523.36) | Grad Norm 3.5226(4463.1148) | Total Time 0.00(0.00)\nIter 3390 | Time 12.0140(11.9219) | Bit/dim 1.4834(1.5530) | Xent 0.1907(0.2349) | Xent Color 0.1266(0.3390) | Loss 3.9115(4.5851) | Error 0.0700(0.0753) | Error Color 0.0322(0.0957) |Steps 530(523.09) | Grad Norm 4.0904(3292.2754) | Total Time 0.00(0.00)\nIter 3400 | Time 11.1763(11.8225) | Bit/dim 1.4375(1.5272) | Xent 0.1635(0.2257) | Xent Color 0.0860(0.2764) | Loss 3.7811(4.3819) | Error 0.0533(0.0729) | Error Color 0.0189(0.0767) |Steps 512(518.19) | Grad Norm 2.9146(2428.8246) | Total Time 0.00(0.00)\nIter 3410 | Time 10.5266(11.7425) | Bit/dim 1.3935(1.4961) | Xent 0.1404(0.2159) | Xent Color 0.1007(0.2269) | Loss 3.6348(4.2073) | Error 0.0478(0.0696) | Error Color 0.0267(0.0616) |Steps 488(514.96) | Grad Norm 5.3650(1791.9281) | Total Time 0.00(0.00)\nIter 3420 | Time 10.9844(11.6318) | Bit/dim 1.3736(1.4660) | Xent 0.2045(0.2108) | Xent Color 0.0733(0.1872) | Loss 3.6510(4.0635) | Error 0.0622(0.0672) | Error Color 0.0167(0.0499) |Steps 470(507.14) | Grad Norm 3.1664(1322.1187) | Total Time 0.00(0.00)\nIter 3430 | Time 11.9435(11.4294) | Bit/dim 1.3526(1.4360) | Xent 0.2144(0.2045) | Xent Color 0.0737(0.1596) | Loss 3.6653(3.9264) | Error 0.0644(0.0650) | Error Color 0.0189(0.0424) |Steps 518(499.09) | Grad Norm 7.8452(976.8924) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0052 | Time 69.2212, Epoch Time 852.0209(744.1108), Bit/dim 1.3288(best: 0.9397), Xent 0.1133, Xent Color 0.0280. Loss 1.3642, Error 0.0359(best: 0.0275), Error Color 0.0014(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 3440 | Time 10.5240(11.2677) | Bit/dim 1.2885(1.4053) | Xent 0.1701(0.1979) | Xent Color 0.0726(0.1374) | Loss 3.4280(4.3340) | Error 0.0622(0.0632) | Error Color 0.0156(0.0359) |Steps 446(490.41) | Grad Norm 5.6523(721.9607) | Total Time 0.00(0.00)\nIter 3450 | Time 10.8449(11.1886) | Bit/dim 1.2833(1.3771) | Xent 0.1887(0.1936) | Xent Color 0.0558(0.1176) | Loss 3.3688(4.0985) | Error 0.0611(0.0616) | Error Color 0.0122(0.0301) |Steps 452(487.30) | Grad Norm 1.3953(533.7736) | Total Time 0.00(0.00)\nIter 3460 | Time 11.3564(11.1277) | Bit/dim 1.2720(1.3486) | Xent 0.1472(0.1904) | Xent Color 0.0699(0.1037) | Loss 3.4203(3.9106) | Error 0.0478(0.0603) | Error Color 0.0233(0.0265) |Steps 458(479.29) | Grad Norm 6.5839(395.2360) | Total Time 0.00(0.00)\nIter 3470 | Time 10.4613(11.0747) | Bit/dim 1.2447(1.3227) | Xent 0.1643(0.1874) | Xent Color 0.0712(0.0928) | Loss 3.2866(3.7511) | Error 0.0511(0.0595) | Error Color 0.0167(0.0234) |Steps 446(473.72) | Grad Norm 9.1661(293.2859) | Total Time 0.00(0.00)\nIter 3480 | Time 12.1524(10.9769) | Bit/dim 1.3034(1.3044) | Xent 0.3870(0.1941) | Xent Color 1.7725(0.1757) | Loss 4.3401(3.6792) | Error 0.1189(0.0617) | Error Color 0.4433(0.0491) |Steps 440(465.07) | Grad Norm 62.3594(225.1998) | Total Time 0.00(0.00)\nIter 3490 | Time 12.0924(11.0205) | Bit/dim 1.3149(1.3149) | Xent 0.2314(0.1959) | Xent Color 0.1760(0.2124) | Loss 3.6659(3.6810) | Error 0.0756(0.0620) | Error Color 0.0544(0.0665) |Steps 536(470.61) | Grad Norm 14.3821(171.4624) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0053 | Time 70.6044, Epoch Time 818.9796(746.3569), Bit/dim 1.2388(best: 0.9397), Xent 0.1170, Xent Color 0.0364. Loss 1.2772, Error 0.0362(best: 0.0275), Error Color 0.0023(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 3500 | Time 11.6858(11.1584) | Bit/dim 1.2324(1.3020) | Xent 0.1790(0.1949) | Xent Color 0.0977(0.1889) | Loss 3.3947(4.2743) | Error 0.0611(0.0621) | Error Color 0.0256(0.0581) |Steps 464(477.81) | Grad Norm 5.9221(128.4060) | Total Time 0.00(0.00)\nIter 3510 | Time 11.1504(11.0814) | Bit/dim 1.1996(1.2798) | Xent 0.1823(0.1875) | Xent Color 0.0747(0.1598) | Loss 3.2904(4.0174) | Error 0.0589(0.0598) | Error Color 0.0211(0.0483) |Steps 494(475.08) | Grad Norm 3.8924(95.9343) | Total Time 0.00(0.00)\nIter 3520 | Time 10.3129(10.9419) | Bit/dim 1.1592(1.2541) | Xent 0.1757(0.1809) | Xent Color 0.0599(0.1342) | Loss 3.1495(3.7978) | Error 0.0600(0.0575) | Error Color 0.0089(0.0392) |Steps 452(469.01) | Grad Norm 2.6320(71.5475) | Total Time 0.00(0.00)\nIter 3530 | Time 10.7515(10.8691) | Bit/dim 1.1507(1.2273) | Xent 0.1369(0.1739) | Xent Color 0.0565(0.1143) | Loss 3.0604(3.6128) | Error 0.0433(0.0553) | Error Color 0.0156(0.0322) |Steps 434(461.34) | Grad Norm 1.6801(53.2821) | Total Time 0.00(0.00)\nIter 3540 | Time 10.5627(10.7441) | Bit/dim 1.1228(1.2017) | Xent 0.1582(0.1709) | Xent Color 0.0445(0.0974) | Loss 3.0362(3.4666) | Error 0.0500(0.0546) | Error Color 0.0067(0.0262) |Steps 446(457.12) | Grad Norm 1.7110(39.6987) | Total Time 0.00(0.00)\nIter 3550 | Time 10.6538(10.7325) | Bit/dim 1.1028(1.1788) | Xent 0.1637(0.1675) | Xent Color 0.0476(0.0846) | Loss 3.0433(3.3526) | Error 0.0544(0.0541) | Error Color 0.0089(0.0221) |Steps 464(453.58) | Grad Norm 1.4121(29.8419) | Total Time 0.00(0.00)\nIter 3560 | Time 10.3399(10.5977) | Bit/dim 1.0941(1.1571) | Xent 0.1539(0.1667) | Xent Color 0.0346(0.0731) | Loss 3.0282(3.2488) | Error 0.0489(0.0533) | Error Color 0.0067(0.0184) |Steps 446(449.29) | Grad Norm 3.1637(22.7407) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0054 | Time 64.1016, Epoch Time 784.2161(747.4927), Bit/dim 1.0794(best: 0.9397), Xent 0.0978, Xent Color 0.0130. Loss 1.1071, Error 0.0318(best: 0.0275), Error Color 0.0005(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 3570 | Time 10.4994(10.4946) | Bit/dim 1.0719(1.1359) | Xent 0.1468(0.1640) | Xent Color 0.0426(0.0644) | Loss 2.8893(3.6622) | Error 0.0467(0.0525) | Error Color 0.0078(0.0158) |Steps 416(443.54) | Grad Norm 2.0843(17.4037) | Total Time 0.00(0.00)\nIter 3580 | Time 10.4748(10.4725) | Bit/dim 1.0534(1.1161) | Xent 0.2077(0.1630) | Xent Color 0.0304(0.0578) | Loss 2.9428(3.4614) | Error 0.0611(0.0517) | Error Color 0.0000(0.0135) |Steps 428(440.25) | Grad Norm 1.9345(13.4894) | Total Time 0.00(0.00)\nIter 3590 | Time 10.9252(10.4590) | Bit/dim 1.0500(1.0987) | Xent 0.1416(0.1607) | Xent Color 0.0433(0.0521) | Loss 2.9063(3.3107) | Error 0.0422(0.0503) | Error Color 0.0122(0.0119) |Steps 440(440.86) | Grad Norm 4.3560(10.5524) | Total Time 0.00(0.00)\nIter 3600 | Time 10.3601(10.4601) | Bit/dim 1.0361(1.0831) | Xent 0.1585(0.1601) | Xent Color 0.0400(0.0480) | Loss 2.8735(3.1913) | Error 0.0489(0.0497) | Error Color 0.0089(0.0107) |Steps 452(440.37) | Grad Norm 3.8851(8.6537) | Total Time 0.00(0.00)\nIter 3610 | Time 9.7622(10.4226) | Bit/dim 1.0274(1.0702) | Xent 0.1797(0.1611) | Xent Color 0.0304(0.0437) | Loss 2.8111(3.0945) | Error 0.0622(0.0502) | Error Color 0.0044(0.0095) |Steps 434(440.60) | Grad Norm 2.2262(7.3895) | Total Time 0.00(0.00)\nIter 3620 | Time 11.0030(10.4191) | Bit/dim 1.0338(1.0593) | Xent 0.1379(0.1574) | Xent Color 0.0300(0.0401) | Loss 2.8115(3.0243) | Error 0.0478(0.0489) | Error Color 0.0044(0.0082) |Steps 416(439.55) | Grad Norm 7.9139(6.6959) | Total Time 0.00(0.00)\nIter 3630 | Time 10.4749(10.3750) | Bit/dim 1.0101(1.0495) | Xent 0.1808(0.1566) | Xent Color 0.0269(0.0371) | Loss 2.8219(2.9656) | Error 0.0556(0.0491) | Error Color 0.0044(0.0075) |Steps 428(437.45) | Grad Norm 5.4955(5.9650) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0055 | Time 61.9616, Epoch Time 769.1950(748.1437), Bit/dim 1.0175(best: 0.9397), Xent 0.0885, Xent Color 0.0089. Loss 1.0419, Error 0.0287(best: 0.0275), Error Color 0.0004(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 3640 | Time 10.4229(10.3508) | Bit/dim 0.9976(1.0385) | Xent 0.1401(0.1574) | Xent Color 0.0339(0.0348) | Loss 2.7262(3.3221) | Error 0.0433(0.0495) | Error Color 0.0078(0.0067) |Steps 416(435.75) | Grad Norm 1.6376(5.1961) | Total Time 0.00(0.00)\nIter 3650 | Time 10.0588(10.4122) | Bit/dim 1.0126(1.0311) | Xent 0.1894(0.1565) | Xent Color 0.0239(0.0328) | Loss 2.7918(3.1821) | Error 0.0544(0.0497) | Error Color 0.0022(0.0061) |Steps 434(433.92) | Grad Norm 2.9208(4.7835) | Total Time 0.00(0.00)\nIter 3660 | Time 10.4882(10.4413) | Bit/dim 0.9907(1.0230) | Xent 0.1142(0.1555) | Xent Color 0.0260(0.0319) | Loss 2.7837(3.0728) | Error 0.0400(0.0488) | Error Color 0.0056(0.0059) |Steps 446(433.41) | Grad Norm 1.8756(4.5870) | Total Time 0.00(0.00)\nIter 3670 | Time 10.4793(10.4818) | Bit/dim 0.9929(1.0169) | Xent 0.1397(0.1544) | Xent Color 0.0209(0.0306) | Loss 2.7527(2.9919) | Error 0.0478(0.0488) | Error Color 0.0022(0.0056) |Steps 380(431.13) | Grad Norm 3.0617(4.1896) | Total Time 0.00(0.00)\nIter 3680 | Time 9.8231(10.4785) | Bit/dim 0.9895(1.0093) | Xent 0.1360(0.1468) | Xent Color 0.0253(0.0296) | Loss 2.7055(2.9242) | Error 0.0433(0.0467) | Error Color 0.0089(0.0057) |Steps 440(430.44) | Grad Norm 4.0735(4.2251) | Total Time 0.00(0.00)\nIter 3690 | Time 10.3128(10.4342) | Bit/dim 0.9917(1.0036) | Xent 0.1819(0.1475) | Xent Color 0.0245(0.0283) | Loss 2.7499(2.8722) | Error 0.0556(0.0468) | Error Color 0.0022(0.0053) |Steps 416(430.22) | Grad Norm 3.3080(4.0151) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0056 | Time 63.3388, Epoch Time 775.8958(748.9763), Bit/dim 0.9811(best: 0.9397), Xent 0.0859, Xent Color 0.0066. Loss 1.0042, Error 0.0269(best: 0.0275), Error Color 0.0003(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 3700 | Time 9.6396(10.4209) | Bit/dim 0.9774(0.9981) | Xent 0.1392(0.1460) | Xent Color 0.0226(0.0277) | Loss 2.6368(3.3352) | Error 0.0467(0.0475) | Error Color 0.0044(0.0052) |Steps 410(432.33) | Grad Norm 2.6000(4.1919) | Total Time 0.00(0.00)\nIter 3710 | Time 10.6041(10.4310) | Bit/dim 0.9858(0.9932) | Xent 0.1397(0.1444) | Xent Color 0.0203(0.0259) | Loss 2.7163(3.1743) | Error 0.0433(0.0465) | Error Color 0.0033(0.0046) |Steps 428(430.19) | Grad Norm 4.1839(4.1432) | Total Time 0.00(0.00)\nIter 3720 | Time 10.5802(10.3950) | Bit/dim 0.9736(0.9889) | Xent 0.1349(0.1466) | Xent Color 0.0254(0.0250) | Loss 2.7908(3.0533) | Error 0.0411(0.0464) | Error Color 0.0033(0.0044) |Steps 446(431.07) | Grad Norm 11.0706(4.8902) | Total Time 0.00(0.00)\nIter 3730 | Time 10.2228(10.3915) | Bit/dim 0.9624(0.9844) | Xent 0.1321(0.1428) | Xent Color 0.0235(0.0246) | Loss 2.6426(2.9541) | Error 0.0456(0.0454) | Error Color 0.0033(0.0042) |Steps 422(431.19) | Grad Norm 6.8351(5.4439) | Total Time 0.00(0.00)\nIter 3740 | Time 10.1589(10.3739) | Bit/dim 0.9672(0.9802) | Xent 0.1575(0.1419) | Xent Color 0.0198(0.0239) | Loss 2.6994(2.8840) | Error 0.0489(0.0454) | Error Color 0.0033(0.0041) |Steps 428(431.29) | Grad Norm 5.1485(5.7045) | Total Time 0.00(0.00)\nIter 3750 | Time 10.2240(10.3356) | Bit/dim 0.9671(0.9764) | Xent 0.1278(0.1375) | Xent Color 0.0199(0.0231) | Loss 2.7124(2.8325) | Error 0.0367(0.0442) | Error Color 0.0033(0.0039) |Steps 440(430.54) | Grad Norm 5.3458(5.7747) | Total Time 0.00(0.00)\nIter 3760 | Time 11.1958(10.3106) | Bit/dim 0.9551(0.9719) | Xent 0.1256(0.1354) | Xent Color 0.0190(0.0227) | Loss 2.6821(2.7882) | Error 0.0456(0.0438) | Error Color 0.0022(0.0039) |Steps 446(429.23) | Grad Norm 3.0156(5.6574) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0057 | Time 62.3839, Epoch Time 765.7905(749.4807), Bit/dim 0.9558(best: 0.9397), Xent 0.0808, Xent Color 0.0050. Loss 0.9773, Error 0.0258(best: 0.0269), Error Color 0.0002(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 3770 | Time 9.8027(10.2923) | Bit/dim 0.9670(0.9679) | Xent 0.1116(0.1327) | Xent Color 0.0160(0.0215) | Loss 2.6583(3.2119) | Error 0.0367(0.0427) | Error Color 0.0044(0.0035) |Steps 422(430.16) | Grad Norm 3.7357(5.3542) | Total Time 0.00(0.00)\nIter 3780 | Time 10.5420(10.2561) | Bit/dim 0.9429(0.9630) | Xent 0.1284(0.1326) | Xent Color 0.0275(0.0209) | Loss 2.5919(3.0661) | Error 0.0378(0.0425) | Error Color 0.0044(0.0033) |Steps 434(430.22) | Grad Norm 8.8183(5.2837) | Total Time 0.00(0.00)\nIter 3790 | Time 10.1808(10.2586) | Bit/dim 0.9587(0.9613) | Xent 0.1132(0.1349) | Xent Color 0.0288(0.0210) | Loss 2.6165(2.9622) | Error 0.0356(0.0430) | Error Color 0.0056(0.0034) |Steps 422(430.23) | Grad Norm 11.0751(6.3930) | Total Time 0.00(0.00)\nIter 3800 | Time 9.7919(10.2413) | Bit/dim 0.9526(0.9580) | Xent 0.1377(0.1339) | Xent Color 0.0152(0.0202) | Loss 2.6161(2.8737) | Error 0.0389(0.0422) | Error Color 0.0011(0.0032) |Steps 416(425.94) | Grad Norm 7.7603(6.7672) | Total Time 0.00(0.00)\nIter 3810 | Time 10.1970(10.2660) | Bit/dim 0.9440(0.9547) | Xent 0.1307(0.1331) | Xent Color 0.0188(0.0198) | Loss 2.6716(2.8154) | Error 0.0333(0.0420) | Error Color 0.0022(0.0031) |Steps 428(426.63) | Grad Norm 6.2282(6.9896) | Total Time 0.00(0.00)\nIter 3820 | Time 10.1225(10.2331) | Bit/dim 0.9353(0.9519) | Xent 0.1015(0.1307) | Xent Color 0.0221(0.0198) | Loss 2.5539(2.7673) | Error 0.0356(0.0409) | Error Color 0.0044(0.0032) |Steps 416(424.99) | Grad Norm 9.1407(7.3750) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0058 | Time 64.1946, Epoch Time 760.5858(749.8139), Bit/dim 0.9598(best: 0.9397), Xent 0.0807, Xent Color 0.0153. Loss 0.9838, Error 0.0249(best: 0.0258), Error Color 0.0013(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 3830 | Time 10.2165(10.2232) | Bit/dim 0.9619(0.9514) | Xent 0.1387(0.1326) | Xent Color 0.0166(0.0221) | Loss 2.7204(3.3024) | Error 0.0422(0.0416) | Error Color 0.0011(0.0040) |Steps 446(425.56) | Grad Norm 11.2609(8.5621) | Total Time 0.00(0.00)\nIter 3840 | Time 9.7978(10.1949) | Bit/dim 0.9343(0.9496) | Xent 0.0809(0.1309) | Xent Color 0.0211(0.0244) | Loss 2.6276(3.1347) | Error 0.0267(0.0408) | Error Color 0.0033(0.0050) |Steps 428(425.02) | Grad Norm 7.6154(9.3932) | Total Time 0.00(0.00)\nIter 3850 | Time 10.7020(10.2353) | Bit/dim 0.9447(0.9476) | Xent 0.1191(0.1303) | Xent Color 0.0154(0.0228) | Loss 2.6495(3.0104) | Error 0.0322(0.0399) | Error Color 0.0033(0.0045) |Steps 428(427.10) | Grad Norm 8.0136(9.4152) | Total Time 0.00(0.00)\nIter 3860 | Time 10.3541(10.2567) | Bit/dim 0.9273(0.9445) | Xent 0.1397(0.1318) | Xent Color 0.0152(0.0214) | Loss 2.6673(2.9103) | Error 0.0389(0.0397) | Error Color 0.0022(0.0040) |Steps 446(427.49) | Grad Norm 3.2285(8.8115) | Total Time 0.00(0.00)\nIter 3870 | Time 9.5984(10.2344) | Bit/dim 0.9304(0.9408) | Xent 0.0923(0.1286) | Xent Color 0.0114(0.0198) | Loss 2.5477(2.8325) | Error 0.0333(0.0393) | Error Color 0.0000(0.0035) |Steps 428(427.43) | Grad Norm 4.2072(8.2373) | Total Time 0.00(0.00)\nIter 3880 | Time 9.6286(10.2097) | Bit/dim 0.9156(0.9352) | Xent 0.1374(0.1306) | Xent Color 0.0170(0.0184) | Loss 2.5627(2.7681) | Error 0.0489(0.0404) | Error Color 0.0033(0.0031) |Steps 416(426.58) | Grad Norm 7.0760(7.3780) | Total Time 0.00(0.00)\nIter 3890 | Time 10.1302(10.2126) | Bit/dim 0.9224(0.9317) | Xent 0.1086(0.1291) | Xent Color 0.0112(0.0172) | Loss 2.5150(2.7213) | Error 0.0378(0.0402) | Error Color 0.0011(0.0027) |Steps 398(423.63) | Grad Norm 2.8011(7.2104) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0059 | Time 63.6394, Epoch Time 763.1483(750.2139), Bit/dim 0.9167(best: 0.9397), Xent 0.0743, Xent Color 0.0029. Loss 0.9359, Error 0.0236(best: 0.0249), Error Color 0.0000(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 3900 | Time 10.0011(10.2395) | Bit/dim 0.9180(0.9282) | Xent 0.1420(0.1278) | Xent Color 0.0165(0.0166) | Loss 2.5812(3.1606) | Error 0.0411(0.0401) | Error Color 0.0011(0.0025) |Steps 422(427.40) | Grad Norm 12.5241(7.0922) | Total Time 0.00(0.00)\nIter 3910 | Time 10.2950(10.2179) | Bit/dim 0.9203(0.9263) | Xent 0.1280(0.1267) | Xent Color 0.0116(0.0163) | Loss 2.6571(3.0180) | Error 0.0367(0.0394) | Error Color 0.0011(0.0024) |Steps 410(426.70) | Grad Norm 3.7905(7.7577) | Total Time 0.00(0.00)\nIter 3920 | Time 10.7919(10.2038) | Bit/dim 0.9340(0.9248) | Xent 0.1392(0.1257) | Xent Color 0.0216(0.0167) | Loss 2.6318(2.9092) | Error 0.0533(0.0399) | Error Color 0.0022(0.0026) |Steps 434(427.84) | Grad Norm 13.2191(8.0687) | Total Time 0.00(0.00)\nIter 3930 | Time 10.2034(10.2249) | Bit/dim 0.9201(0.9228) | Xent 0.1284(0.1253) | Xent Color 0.0277(0.0168) | Loss 2.6109(2.8273) | Error 0.0489(0.0401) | Error Color 0.0067(0.0025) |Steps 422(427.32) | Grad Norm 10.5302(8.6332) | Total Time 0.00(0.00)\nIter 3940 | Time 10.3458(10.2727) | Bit/dim 0.9181(0.9214) | Xent 0.1284(0.1262) | Xent Color 0.0184(0.0177) | Loss 2.6110(2.7653) | Error 0.0356(0.0403) | Error Color 0.0022(0.0027) |Steps 410(426.25) | Grad Norm 10.5888(9.2557) | Total Time 0.00(0.00)\nIter 3950 | Time 9.9312(10.2592) | Bit/dim 0.9251(0.9190) | Xent 0.1107(0.1229) | Xent Color 0.0173(0.0170) | Loss 2.5769(2.7184) | Error 0.0400(0.0389) | Error Color 0.0022(0.0027) |Steps 404(424.61) | Grad Norm 10.2038(8.9391) | Total Time 0.00(0.00)\nIter 3960 | Time 10.4438(10.2878) | Bit/dim 0.9042(0.9161) | Xent 0.0928(0.1190) | Xent Color 0.0142(0.0168) | Loss 2.5650(2.6776) | Error 0.0300(0.0383) | Error Color 0.0011(0.0029) |Steps 434(426.04) | Grad Norm 7.3624(8.8840) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0060 | Time 64.3287, Epoch Time 763.5196(750.6131), Bit/dim 0.9080(best: 0.9167), Xent 0.0713, Xent Color 0.0025. Loss 0.9265, Error 0.0231(best: 0.0236), Error Color 0.0000(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 3970 | Time 9.9098(10.2967) | Bit/dim 0.9034(0.9136) | Xent 0.1106(0.1177) | Xent Color 0.0253(0.0165) | Loss 2.5951(3.0807) | Error 0.0411(0.0378) | Error Color 0.0067(0.0028) |Steps 428(426.24) | Grad Norm 12.2552(9.1293) | Total Time 0.00(0.00)\nIter 3980 | Time 10.0607(10.2862) | Bit/dim 0.8952(0.9114) | Xent 0.1203(0.1177) | Xent Color 0.0174(0.0159) | Loss 2.6122(2.9543) | Error 0.0389(0.0376) | Error Color 0.0033(0.0026) |Steps 434(428.30) | Grad Norm 8.7222(9.0709) | Total Time 0.00(0.00)\nIter 3990 | Time 9.9780(10.2427) | Bit/dim 0.9163(0.9107) | Xent 0.0942(0.1178) | Xent Color 0.0620(0.0184) | Loss 2.6509(2.8613) | Error 0.0322(0.0374) | Error Color 0.0200(0.0036) |Steps 434(428.80) | Grad Norm 19.7269(10.1676) | Total Time 0.00(0.00)\nIter 4000 | Time 10.2167(10.2858) | Bit/dim 0.9126(0.9103) | Xent 0.1132(0.1175) | Xent Color 0.0169(0.0179) | Loss 2.6041(2.7887) | Error 0.0333(0.0376) | Error Color 0.0033(0.0033) |Steps 452(429.76) | Grad Norm 11.2636(10.5177) | Total Time 0.00(0.00)\nIter 4010 | Time 9.9297(10.3093) | Bit/dim 0.9067(0.9074) | Xent 0.1283(0.1164) | Xent Color 0.0120(0.0172) | Loss 2.6023(2.7320) | Error 0.0422(0.0374) | Error Color 0.0033(0.0031) |Steps 434(430.90) | Grad Norm 7.6740(9.7399) | Total Time 0.00(0.00)\nIter 4020 | Time 10.0679(10.3174) | Bit/dim 0.9270(0.9051) | Xent 0.1069(0.1152) | Xent Color 0.1103(0.0200) | Loss 2.6593(2.6936) | Error 0.0344(0.0367) | Error Color 0.0433(0.0042) |Steps 404(431.28) | Grad Norm 39.6421(10.5757) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0061 | Time 75.3388, Epoch Time 783.4211(751.5973), Bit/dim 2.2663(best: 0.9080), Xent 1.6442, Xent Color 1.6166. Loss 3.0815, Error 0.4792(best: 0.0231), Error Color 0.5290(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 4030 | Time 13.0602(10.7522) | Bit/dim 2.3469(1.2478) | Xent 0.7853(0.3852) | Xent Color 1.8542(1.5808) | Loss 6.5451(4.7666) | Error 0.2433(0.1148) | Error Color 0.5256(0.1626) |Steps 614(456.64) | Grad Norm 43.5064(36.6952) | Total Time 0.00(0.00)\nIter 4040 | Time 11.5646(11.1290) | Bit/dim 1.9834(1.4683) | Xent 0.5406(0.4450) | Xent Color 0.7042(1.3950) | Loss 5.2255(4.9630) | Error 0.1678(0.1368) | Error Color 0.2833(0.2035) |Steps 536(479.20) | Grad Norm 14.4918(32.6762) | Total Time 0.00(0.00)\nIter 4050 | Time 12.4259(11.4671) | Bit/dim 1.8232(1.5778) | Xent 0.3378(0.4176) | Xent Color 0.3621(1.1489) | Loss 4.7698(4.9413) | Error 0.1111(0.1290) | Error Color 0.1511(0.1967) |Steps 584(502.52) | Grad Norm 7.2378(26.5839) | Total Time 0.00(0.00)\nIter 4060 | Time 12.1982(11.7527) | Bit/dim 1.7114(1.6222) | Xent 0.2446(0.3783) | Xent Color 0.1806(0.9220) | Loss 4.3769(4.8305) | Error 0.0756(0.1171) | Error Color 0.0522(0.1710) |Steps 572(520.78) | Grad Norm 2.2878(21.7750) | Total Time 0.00(0.00)\nIter 4070 | Time 13.0160(12.0282) | Bit/dim 1.6638(1.6359) | Xent 0.2178(0.3422) | Xent Color 0.1465(0.7250) | Loss 4.3259(4.6976) | Error 0.0744(0.1066) | Error Color 0.0433(0.1399) |Steps 566(534.83) | Grad Norm 9.8532(17.7885) | Total Time 0.00(0.00)\nIter 4080 | Time 12.3015(12.1368) | Bit/dim 1.5621(1.6232) | Xent 0.1601(0.3075) | Xent Color 0.1323(0.5718) | Loss 4.0931(4.5583) | Error 0.0533(0.0962) | Error Color 0.0456(0.1133) |Steps 578(541.77) | Grad Norm 8.9148(15.0780) | Total Time 0.00(0.00)\nIter 4090 | Time 11.9137(12.1849) | Bit/dim 1.4915(1.5939) | Xent 0.2104(0.2834) | Xent Color 0.1086(0.4507) | Loss 3.9406(4.4086) | Error 0.0644(0.0892) | Error Color 0.0322(0.0910) |Steps 530(543.64) | Grad Norm 9.3690(12.6961) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0062 | Time 72.2232, Epoch Time 919.5928(756.6372), Bit/dim 1.4712(best: 0.9080), Xent 0.1228, Xent Color 0.0413. Loss 1.5123, Error 0.0415(best: 0.0231), Error Color 0.0019(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 4100 | Time 12.5289(12.1758) | Bit/dim 1.4071(1.5554) | Xent 0.2143(0.2585) | Xent Color 0.0842(0.3597) | Loss 3.8367(4.8438) | Error 0.0644(0.0816) | Error Color 0.0111(0.0739) |Steps 536(541.31) | Grad Norm 4.6716(11.0612) | Total Time 0.00(0.00)\nIter 4110 | Time 11.8140(12.1509) | Bit/dim 1.3700(1.5103) | Xent 0.1753(0.2406) | Xent Color 0.1015(0.2907) | Loss 3.6905(4.5462) | Error 0.0622(0.0758) | Error Color 0.0267(0.0604) |Steps 512(536.23) | Grad Norm 9.2377(10.2293) | Total Time 0.00(0.00)\nIter 4120 | Time 11.2751(12.0387) | Bit/dim 1.3142(1.4669) | Xent 0.2255(0.2295) | Xent Color 0.1534(0.2389) | Loss 3.4760(4.2987) | Error 0.0711(0.0724) | Error Color 0.0533(0.0512) |Steps 488(526.09) | Grad Norm 30.2588(10.8551) | Total Time 0.00(0.00)\nIter 4130 | Time 14.1509(12.2205) | Bit/dim 1.4741(1.4674) | Xent 0.2382(0.2364) | Xent Color 0.4785(0.4842) | Loss 4.1201(4.3168) | Error 0.0733(0.0747) | Error Color 0.1767(0.1173) |Steps 578(533.07) | Grad Norm 11.8423(23.5550) | Total Time 0.00(0.00)\nIter 4140 | Time 11.8213(12.1367) | Bit/dim 1.4200(1.4653) | Xent 0.1946(0.2340) | Xent Color 0.2988(0.4681) | Loss 3.7912(4.2154) | Error 0.0600(0.0740) | Error Color 0.1056(0.1283) |Steps 470(524.22) | Grad Norm 6.2545(20.8490) | Total Time 0.00(0.00)\nIter 4150 | Time 11.7854(12.0718) | Bit/dim 1.3279(1.4359) | Xent 0.2018(0.2225) | Xent Color 0.1776(0.4085) | Loss 3.6127(4.0582) | Error 0.0778(0.0708) | Error Color 0.0544(0.1152) |Steps 506(519.62) | Grad Norm 6.8059(17.0435) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0063 | Time 69.7222, Epoch Time 887.7507(760.5706), Bit/dim 1.2819(best: 0.9080), Xent 0.1132, Xent Color 0.0652. Loss 1.3265, Error 0.0356(best: 0.0231), Error Color 0.0038(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 4160 | Time 11.3275(12.0126) | Bit/dim 1.2792(1.3974) | Xent 0.2013(0.2149) | Xent Color 0.1195(0.3415) | Loss 3.3865(4.5049) | Error 0.0611(0.0686) | Error Color 0.0222(0.0959) |Steps 470(512.72) | Grad Norm 4.0013(14.1803) | Total Time 0.00(0.00)\nIter 4170 | Time 11.7934(11.9280) | Bit/dim 1.2146(1.3559) | Xent 0.1532(0.2074) | Xent Color 0.1127(0.2807) | Loss 3.2744(4.2019) | Error 0.0511(0.0662) | Error Color 0.0311(0.0779) |Steps 506(507.09) | Grad Norm 5.8378(11.9232) | Total Time 0.00(0.00)\nIter 4180 | Time 11.3756(11.8246) | Bit/dim 1.1735(1.3101) | Xent 0.1858(0.1985) | Xent Color 0.0649(0.2287) | Loss 3.2312(3.9448) | Error 0.0567(0.0633) | Error Color 0.0144(0.0623) |Steps 494(503.03) | Grad Norm 3.8928(9.8322) | Total Time 0.00(0.00)\nIter 4190 | Time 10.6993(11.6938) | Bit/dim 1.1316(1.2698) | Xent 0.1606(0.1882) | Xent Color 0.0712(0.1872) | Loss 3.1166(3.7414) | Error 0.0456(0.0599) | Error Color 0.0167(0.0496) |Steps 476(495.35) | Grad Norm 3.3916(8.2851) | Total Time 0.00(0.00)\nIter 4200 | Time 11.3703(11.5818) | Bit/dim 1.1095(1.2309) | Xent 0.1790(0.1831) | Xent Color 0.0545(0.1546) | Loss 3.1607(3.5747) | Error 0.0533(0.0585) | Error Color 0.0133(0.0402) |Steps 512(488.69) | Grad Norm 1.8221(7.1392) | Total Time 0.00(0.00)\nIter 4210 | Time 10.8935(11.4798) | Bit/dim 1.1051(1.1986) | Xent 0.1753(0.1795) | Xent Color 0.0498(0.1282) | Loss 3.1089(3.4489) | Error 0.0644(0.0579) | Error Color 0.0078(0.0324) |Steps 470(483.60) | Grad Norm 3.3441(6.1593) | Total Time 0.00(0.00)\nIter 4220 | Time 10.7786(11.4418) | Bit/dim 1.0710(1.1695) | Xent 0.1370(0.1737) | Xent Color 0.0532(0.1077) | Loss 2.9756(3.3364) | Error 0.0467(0.0556) | Error Color 0.0122(0.0264) |Steps 458(480.45) | Grad Norm 1.3586(4.9300) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0064 | Time 67.0232, Epoch Time 842.5922(763.0312), Bit/dim 1.0785(best: 0.9080), Xent 0.0966, Xent Color 0.0147. Loss 1.1063, Error 0.0310(best: 0.0231), Error Color 0.0005(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 4230 | Time 11.3158(11.4209) | Bit/dim 1.0765(1.1444) | Xent 0.1905(0.1696) | Xent Color 0.0339(0.0917) | Loss 3.0567(3.7985) | Error 0.0511(0.0540) | Error Color 0.0056(0.0221) |Steps 488(476.68) | Grad Norm 4.0325(4.3387) | Total Time 0.00(0.00)\nIter 4240 | Time 11.7808(11.3837) | Bit/dim 1.0677(1.1240) | Xent 0.1168(0.1635) | Xent Color 0.0429(0.0783) | Loss 2.9649(3.5892) | Error 0.0433(0.0525) | Error Color 0.0056(0.0180) |Steps 500(477.72) | Grad Norm 1.7431(3.9788) | Total Time 0.00(0.00)\nIter 4250 | Time 10.8380(11.3375) | Bit/dim 1.0288(1.1048) | Xent 0.1346(0.1600) | Xent Color 0.0381(0.0685) | Loss 2.8681(3.4248) | Error 0.0433(0.0508) | Error Color 0.0067(0.0149) |Steps 440(475.95) | Grad Norm 3.8695(4.2578) | Total Time 0.00(0.00)\nIter 4260 | Time 11.0697(11.3801) | Bit/dim 1.0348(1.0873) | Xent 0.1274(0.1566) | Xent Color 0.0372(0.0608) | Loss 2.9160(3.2940) | Error 0.0444(0.0500) | Error Color 0.0067(0.0131) |Steps 458(474.68) | Grad Norm 3.2015(4.3710) | Total Time 0.00(0.00)\nIter 4270 | Time 11.2525(11.2997) | Bit/dim 1.0396(1.0736) | Xent 0.1684(0.1537) | Xent Color 0.0321(0.0535) | Loss 2.9128(3.1937) | Error 0.0544(0.0489) | Error Color 0.0067(0.0110) |Steps 452(472.12) | Grad Norm 2.5450(4.0373) | Total Time 0.00(0.00)\nIter 4280 | Time 10.9683(11.2572) | Bit/dim 1.0270(1.0614) | Xent 0.1650(0.1517) | Xent Color 0.0365(0.0487) | Loss 2.8877(3.1089) | Error 0.0433(0.0476) | Error Color 0.0067(0.0097) |Steps 446(469.46) | Grad Norm 2.8584(4.0711) | Total Time 0.00(0.00)\nIter 4290 | Time 10.6345(11.1760) | Bit/dim 1.0131(1.0508) | Xent 0.1444(0.1497) | Xent Color 0.0359(0.0446) | Loss 2.8433(3.0472) | Error 0.0422(0.0470) | Error Color 0.0078(0.0088) |Steps 482(469.75) | Grad Norm 5.8081(4.5880) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0065 | Time 68.1525, Epoch Time 829.6735(765.0305), Bit/dim 1.0153(best: 0.9080), Xent 0.0909, Xent Color 0.0099. Loss 1.0404, Error 0.0298(best: 0.0231), Error Color 0.0006(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 4300 | Time 10.9345(11.1023) | Bit/dim 1.0179(1.0414) | Xent 0.1418(0.1501) | Xent Color 0.0342(0.0416) | Loss 2.8522(3.4516) | Error 0.0433(0.0467) | Error Color 0.0056(0.0082) |Steps 440(467.27) | Grad Norm 4.0146(5.0037) | Total Time 0.00(0.00)\nIter 4310 | Time 11.7497(11.1469) | Bit/dim 1.0122(1.0321) | Xent 0.1012(0.1469) | Xent Color 0.0254(0.0384) | Loss 2.8456(3.2917) | Error 0.0367(0.0462) | Error Color 0.0044(0.0073) |Steps 428(465.12) | Grad Norm 6.1178(5.0157) | Total Time 0.00(0.00)\nIter 4320 | Time 10.6128(11.1330) | Bit/dim 1.0071(1.0235) | Xent 0.1090(0.1442) | Xent Color 0.0264(0.0357) | Loss 2.7645(3.1673) | Error 0.0333(0.0448) | Error Color 0.0044(0.0066) |Steps 452(460.28) | Grad Norm 1.9250(4.6220) | Total Time 0.00(0.00)\nIter 4330 | Time 11.4873(11.1470) | Bit/dim 1.0044(1.0157) | Xent 0.1359(0.1409) | Xent Color 0.0278(0.0336) | Loss 2.8876(3.0756) | Error 0.0411(0.0438) | Error Color 0.0056(0.0062) |Steps 500(459.74) | Grad Norm 1.6360(4.4772) | Total Time 0.00(0.00)\nIter 4340 | Time 10.7984(11.0964) | Bit/dim 0.9900(1.0089) | Xent 0.1268(0.1405) | Xent Color 0.0217(0.0313) | Loss 2.7811(2.9994) | Error 0.0422(0.0441) | Error Color 0.0022(0.0057) |Steps 458(459.50) | Grad Norm 8.9695(4.6757) | Total Time 0.00(0.00)\nIter 4350 | Time 10.6590(11.0006) | Bit/dim 0.9624(1.0019) | Xent 0.1074(0.1383) | Xent Color 0.0322(0.0307) | Loss 2.7542(2.9377) | Error 0.0389(0.0432) | Error Color 0.0067(0.0055) |Steps 470(457.03) | Grad Norm 4.9809(4.7116) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0066 | Time 66.3550, Epoch Time 815.2879(766.5382), Bit/dim 0.9777(best: 0.9080), Xent 0.0839, Xent Color 0.0059. Loss 1.0001, Error 0.0272(best: 0.0231), Error Color 0.0000(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 4360 | Time 10.7468(10.9712) | Bit/dim 0.9740(0.9953) | Xent 0.1326(0.1377) | Xent Color 0.0208(0.0287) | Loss 2.7810(3.4360) | Error 0.0411(0.0435) | Error Color 0.0044(0.0048) |Steps 446(454.94) | Grad Norm 3.8491(4.5719) | Total Time 0.00(0.00)\nIter 4370 | Time 10.0285(10.9054) | Bit/dim 0.9837(0.9898) | Xent 0.1414(0.1382) | Xent Color 0.0189(0.0273) | Loss 2.7233(3.2561) | Error 0.0422(0.0434) | Error Color 0.0011(0.0043) |Steps 452(454.90) | Grad Norm 5.4240(4.6667) | Total Time 0.00(0.00)\nIter 4380 | Time 11.3466(10.8847) | Bit/dim 0.9603(0.9842) | Xent 0.1314(0.1377) | Xent Color 0.0216(0.0266) | Loss 2.7727(3.1189) | Error 0.0356(0.0435) | Error Color 0.0022(0.0041) |Steps 470(452.48) | Grad Norm 4.0971(5.0131) | Total Time 0.00(0.00)\nIter 4400 | Time 10.1810(10.8125) | Bit/dim 0.9602(0.9746) | Xent 0.1433(0.1306) | Xent Color 0.0185(0.0238) | Loss 2.7112(2.9381) | Error 0.0456(0.0415) | Error Color 0.0011(0.0036) |Steps 446(449.13) | Grad Norm 6.5543(5.1072) | Total Time 0.00(0.00)\nIter 4410 | Time 11.4532(10.8133) | Bit/dim 0.9675(0.9717) | Xent 0.1501(0.1300) | Xent Color 0.0487(0.0245) | Loss 2.7909(2.8790) | Error 0.0478(0.0415) | Error Color 0.0122(0.0041) |Steps 482(450.22) | Grad Norm 23.6484(6.4219) | Total Time 0.00(0.00)\nIter 4420 | Time 12.0348(10.9460) | Bit/dim 1.5519(1.1056) | Xent 0.5189(0.1905) | Xent Color 0.2821(0.5186) | Loss 4.1580(3.3475) | Error 0.1678(0.0602) | Error Color 0.0900(0.0973) |Steps 506(456.70) | Grad Norm 17.1980(23.3241) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0067 | Time 72.4244, Epoch Time 814.9995(767.9921), Bit/dim 1.5478(best: 0.9080), Xent 0.1209, Xent Color 0.2538. Loss 1.6415, Error 0.0373(best: 0.0231), Error Color 0.0903(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 4430 | Time 12.7597(11.3411) | Bit/dim 1.3768(1.2053) | Xent 0.2270(0.1923) | Xent Color 0.1895(0.4626) | Loss 3.8726(4.0582) | Error 0.0689(0.0610) | Error Color 0.0556(0.1008) |Steps 554(479.10) | Grad Norm 5.2862(19.2754) | Total Time 0.00(0.00)\nIter 4440 | Time 13.2707(11.7320) | Bit/dim 1.2048(1.2234) | Xent 0.1441(0.1963) | Xent Color 0.0809(0.3724) | Loss 3.4277(3.9371) | Error 0.0467(0.0629) | Error Color 0.0256(0.0835) |Steps 554(496.97) | Grad Norm 6.8881(15.7553) | Total Time 0.00(0.00)\nIter 4450 | Time 12.0369(11.8443) | Bit/dim 1.1016(1.2009) | Xent 0.1823(0.1866) | Xent Color 0.0689(0.2935) | Loss 3.1903(3.7545) | Error 0.0556(0.0596) | Error Color 0.0122(0.0659) |Steps 530(503.45) | Grad Norm 2.8424(12.7723) | Total Time 0.00(0.00)\nIter 4460 | Time 11.3261(11.8170) | Bit/dim 1.0613(1.1674) | Xent 0.1372(0.1742) | Xent Color 0.0561(0.2332) | Loss 2.9392(3.5670) | Error 0.0433(0.0558) | Error Color 0.0122(0.0520) |Steps 482(502.58) | Grad Norm 2.5452(10.2159) | Total Time 0.00(0.00)\nIter 4470 | Time 10.8695(11.7132) | Bit/dim 1.0307(1.1330) | Xent 0.1012(0.1635) | Xent Color 0.0487(0.1846) | Loss 2.9171(3.4106) | Error 0.0311(0.0520) | Error Color 0.0089(0.0404) |Steps 470(498.72) | Grad Norm 1.7279(8.1746) | Total Time 0.00(0.00)\nIter 4480 | Time 11.7239(11.6208) | Bit/dim 1.0024(1.1014) | Xent 0.1354(0.1585) | Xent Color 0.0409(0.1483) | Loss 2.8267(3.2722) | Error 0.0422(0.0502) | Error Color 0.0056(0.0321) |Steps 452(491.93) | Grad Norm 3.8098(6.8368) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0068 | Time 69.0079, Epoch Time 876.2288(771.2392), Bit/dim 0.9962(best: 0.9080), Xent 0.0792, Xent Color 0.0128. Loss 1.0192, Error 0.0250(best: 0.0231), Error Color 0.0005(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 4490 | Time 10.5980(11.5114) | Bit/dim 1.0008(1.0753) | Xent 0.1433(0.1532) | Xent Color 0.0514(0.1208) | Loss 2.8222(3.7552) | Error 0.0444(0.0486) | Error Color 0.0089(0.0255) |Steps 476(487.91) | Grad Norm 2.9832(5.8148) | Total Time 0.00(0.00)\nIter 4500 | Time 10.9340(11.5133) | Bit/dim 1.0019(1.0532) | Xent 0.1183(0.1444) | Xent Color 0.0363(0.0992) | Loss 2.7675(3.5065) | Error 0.0444(0.0454) | Error Color 0.0044(0.0205) |Steps 482(486.58) | Grad Norm 3.0086(4.9348) | Total Time 0.00(0.00)\nIter 4510 | Time 11.2328(11.4946) | Bit/dim 0.9838(1.0344) | Xent 0.1326(0.1405) | Xent Color 0.0334(0.0824) | Loss 2.7720(3.3218) | Error 0.0444(0.0444) | Error Color 0.0067(0.0169) |Steps 452(482.14) | Grad Norm 1.8229(4.1758) | Total Time 0.00(0.00)\nIter 4520 | Time 11.5063(11.4605) | Bit/dim 0.9741(1.0184) | Xent 0.1390(0.1372) | Xent Color 0.0317(0.0693) | Loss 2.7473(3.1753) | Error 0.0333(0.0429) | Error Color 0.0044(0.0140) |Steps 470(478.96) | Grad Norm 2.1549(3.5949) | Total Time 0.00(0.00)\nIter 4530 | Time 11.8084(11.4513) | Bit/dim 0.9622(1.0043) | Xent 0.1078(0.1328) | Xent Color 0.0324(0.0593) | Loss 2.7280(3.0594) | Error 0.0322(0.0416) | Error Color 0.0089(0.0119) |Steps 476(476.75) | Grad Norm 2.2620(3.2129) | Total Time 0.00(0.00)\nIter 4540 | Time 11.4347(11.3788) | Bit/dim 0.9701(0.9934) | Xent 0.1134(0.1317) | Xent Color 0.0227(0.0515) | Loss 2.7050(2.9750) | Error 0.0322(0.0411) | Error Color 0.0011(0.0099) |Steps 488(475.74) | Grad Norm 4.0254(3.2858) | Total Time 0.00(0.00)\nIter 4550 | Time 11.3928(11.3700) | Bit/dim 0.9564(0.9844) | Xent 0.1421(0.1316) | Xent Color 0.0309(0.0462) | Loss 2.7125(2.9108) | Error 0.0411(0.0410) | Error Color 0.0056(0.0089) |Steps 440(472.17) | Grad Norm 2.3956(3.3462) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0069 | Time 65.7329, Epoch Time 837.9097(773.2393), Bit/dim 0.9547(best: 0.9080), Xent 0.0795, Xent Color 0.0082. Loss 0.9766, Error 0.0258(best: 0.0231), Error Color 0.0004(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 4560 | Time 11.3289(11.3029) | Bit/dim 0.9504(0.9767) | Xent 0.1053(0.1287) | Xent Color 0.0282(0.0408) | Loss 2.6514(3.3575) | Error 0.0333(0.0401) | Error Color 0.0044(0.0077) |Steps 440(469.83) | Grad Norm 2.6407(3.1781) | Total Time 0.00(0.00)\nIter 4570 | Time 10.5585(11.3317) | Bit/dim 0.9406(0.9698) | Xent 0.1413(0.1285) | Xent Color 0.0277(0.0369) | Loss 2.6857(3.1870) | Error 0.0422(0.0399) | Error Color 0.0044(0.0068) |Steps 458(470.18) | Grad Norm 3.3248(3.1493) | Total Time 0.00(0.00)\nIter 4580 | Time 11.0509(11.3217) | Bit/dim 0.9479(0.9643) | Xent 0.1004(0.1257) | Xent Color 0.0224(0.0336) | Loss 2.5985(3.0546) | Error 0.0333(0.0391) | Error Color 0.0044(0.0061) |Steps 482(470.05) | Grad Norm 2.4012(3.0381) | Total Time 0.00(0.00)\nIter 4590 | Time 11.7037(11.3058) | Bit/dim 0.9350(0.9588) | Xent 0.1339(0.1267) | Xent Color 0.0308(0.0313) | Loss 2.7106(2.9566) | Error 0.0400(0.0400) | Error Color 0.0067(0.0057) |Steps 446(470.05) | Grad Norm 7.3741(3.1742) | Total Time 0.00(0.00)\nIter 4600 | Time 11.6375(11.2933) | Bit/dim 0.9404(0.9540) | Xent 0.1090(0.1237) | Xent Color 0.0395(0.0298) | Loss 2.6727(2.8851) | Error 0.0344(0.0390) | Error Color 0.0167(0.0056) |Steps 482(471.93) | Grad Norm 11.1740(3.9338) | Total Time 0.00(0.00)\nIter 4610 | Time 11.1336(11.2705) | Bit/dim 0.9279(0.9508) | Xent 0.1052(0.1219) | Xent Color 0.0251(0.0277) | Loss 2.6848(2.8373) | Error 0.0333(0.0384) | Error Color 0.0056(0.0051) |Steps 452(471.97) | Grad Norm 5.2054(4.3754) | Total Time 0.00(0.00)\nIter 4620 | Time 11.7781(11.3257) | Bit/dim 0.9336(0.9461) | Xent 0.1209(0.1219) | Xent Color 0.0193(0.0262) | Loss 2.6545(2.7924) | Error 0.0422(0.0385) | Error Color 0.0033(0.0048) |Steps 458(470.85) | Grad Norm 4.2293(4.0046) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0070 | Time 66.2768, Epoch Time 834.8327(775.0871), Bit/dim 0.9327(best: 0.9080), Xent 0.0741, Xent Color 0.0062. Loss 0.9528, Error 0.0234(best: 0.0231), Error Color 0.0004(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 4630 | Time 11.5168(11.3046) | Bit/dim 0.9219(0.9420) | Xent 0.1014(0.1193) | Xent Color 0.0175(0.0252) | Loss 2.6590(3.2070) | Error 0.0267(0.0371) | Error Color 0.0022(0.0047) |Steps 452(468.73) | Grad Norm 2.9576(3.7813) | Total Time 0.00(0.00)\nIter 4640 | Time 11.7715(11.3167) | Bit/dim 0.9236(0.9382) | Xent 0.0893(0.1171) | Xent Color 0.0247(0.0244) | Loss 2.6487(3.0582) | Error 0.0311(0.0369) | Error Color 0.0022(0.0043) |Steps 500(465.38) | Grad Norm 5.5701(3.8985) | Total Time 0.00(0.00)\nIter 4650 | Time 11.2355(11.2397) | Bit/dim 0.9362(0.9359) | Xent 0.1217(0.1171) | Xent Color 0.0266(0.0234) | Loss 2.6962(2.9461) | Error 0.0378(0.0366) | Error Color 0.0078(0.0041) |Steps 458(464.33) | Grad Norm 9.9564(4.9054) | Total Time 0.00(0.00)\nIter 4660 | Time 11.7525(11.2300) | Bit/dim 0.9241(0.9329) | Xent 0.1513(0.1174) | Xent Color 0.0223(0.0224) | Loss 2.7113(2.8642) | Error 0.0467(0.0363) | Error Color 0.0067(0.0041) |Steps 476(462.91) | Grad Norm 4.1602(4.9107) | Total Time 0.00(0.00)\nIter 4670 | Time 11.2572(11.2325) | Bit/dim 0.9074(0.9294) | Xent 0.1162(0.1156) | Xent Color 0.0221(0.0213) | Loss 2.6181(2.7965) | Error 0.0367(0.0357) | Error Color 0.0056(0.0038) |Steps 464(462.04) | Grad Norm 6.0508(4.7545) | Total Time 0.00(0.00)\nIter 4680 | Time 10.6520(11.2070) | Bit/dim 0.9182(0.9264) | Xent 0.1252(0.1177) | Xent Color 0.0157(0.0207) | Loss 2.6537(2.7545) | Error 0.0389(0.0364) | Error Color 0.0033(0.0037) |Steps 446(463.07) | Grad Norm 1.9537(5.1452) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0071 | Time 66.7659, Epoch Time 826.6961(776.6353), Bit/dim 0.9151(best: 0.9080), Xent 0.0696, Xent Color 0.0040. Loss 0.9335, Error 0.0221(best: 0.0231), Error Color 0.0000(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 4690 | Time 10.8511(11.1863) | Bit/dim 0.9190(0.9243) | Xent 0.1296(0.1195) | Xent Color 0.0213(0.0203) | Loss 2.5723(3.2509) | Error 0.0433(0.0370) | Error Color 0.0033(0.0037) |Steps 458(462.66) | Grad Norm 8.9644(5.5918) | Total Time 0.00(0.00)\nIter 4700 | Time 11.2497(11.1719) | Bit/dim 0.9079(0.9219) | Xent 0.1144(0.1166) | Xent Color 0.0277(0.0200) | Loss 2.6076(3.0841) | Error 0.0378(0.0365) | Error Color 0.0067(0.0035) |Steps 440(461.39) | Grad Norm 9.5699(6.0804) | Total Time 0.00(0.00)\nIter 4710 | Time 11.0571(11.1118) | Bit/dim 0.9212(0.9206) | Xent 0.0986(0.1162) | Xent Color 0.0186(0.0206) | Loss 2.6227(2.9574) | Error 0.0344(0.0365) | Error Color 0.0022(0.0036) |Steps 440(456.17) | Grad Norm 14.2705(7.7395) | Total Time 0.00(0.00)\nIter 4720 | Time 11.1768(11.0820) | Bit/dim 0.9167(0.9197) | Xent 0.1404(0.1149) | Xent Color 0.0262(0.0205) | Loss 2.6139(2.8659) | Error 0.0456(0.0367) | Error Color 0.0033(0.0034) |Steps 440(456.94) | Grad Norm 12.5807(8.5553) | Total Time 0.00(0.00)\nIter 4730 | Time 11.9478(11.0933) | Bit/dim 0.9285(0.9218) | Xent 0.1158(0.1153) | Xent Color 0.0378(0.0255) | Loss 2.6953(2.8062) | Error 0.0378(0.0367) | Error Color 0.0100(0.0053) |Steps 512(454.80) | Grad Norm 19.7461(11.4250) | Total Time 0.00(0.00)\nIter 4740 | Time 10.8897(11.0597) | Bit/dim 0.9415(0.9242) | Xent 0.1105(0.1148) | Xent Color 0.0450(0.0314) | Loss 2.6511(2.7704) | Error 0.0333(0.0367) | Error Color 0.0144(0.0077) |Steps 470(454.33) | Grad Norm 28.7228(14.7613) | Total Time 0.00(0.00)\nIter 4750 | Time 10.8620(11.0153) | Bit/dim 0.9124(0.9231) | Xent 0.1101(0.1119) | Xent Color 0.0235(0.0291) | Loss 2.6050(2.7284) | Error 0.0367(0.0353) | Error Color 0.0044(0.0069) |Steps 434(454.80) | Grad Norm 10.7070(14.2929) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0072 | Time 66.3359, Epoch Time 817.1167(777.8498), Bit/dim 0.9106(best: 0.9080), Xent 0.0723, Xent Color 0.0038. Loss 0.9296, Error 0.0237(best: 0.0221), Error Color 0.0001(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 4760 | Time 11.6926(11.0365) | Bit/dim 0.9135(0.9196) | Xent 0.0959(0.1122) | Xent Color 0.0143(0.0252) | Loss 2.6178(3.1622) | Error 0.0356(0.0358) | Error Color 0.0022(0.0055) |Steps 476(452.76) | Grad Norm 6.6925(12.4702) | Total Time 0.00(0.00)\nIter 4770 | Time 10.5181(11.0248) | Bit/dim 0.9102(0.9156) | Xent 0.1130(0.1083) | Xent Color 0.0202(0.0230) | Loss 2.5731(3.0152) | Error 0.0300(0.0346) | Error Color 0.0044(0.0048) |Steps 452(454.85) | Grad Norm 6.3041(11.1132) | Total Time 0.00(0.00)\nIter 4780 | Time 11.1290(11.0975) | Bit/dim 0.8957(0.9120) | Xent 0.1102(0.1093) | Xent Color 0.0152(0.0207) | Loss 2.5628(2.9087) | Error 0.0333(0.0349) | Error Color 0.0033(0.0041) |Steps 476(454.90) | Grad Norm 5.1985(9.6237) | Total Time 0.00(0.00)\nIter 4790 | Time 10.8751(11.0430) | Bit/dim 0.8972(0.9088) | Xent 0.1391(0.1084) | Xent Color 0.0160(0.0197) | Loss 2.5630(2.8161) | Error 0.0444(0.0342) | Error Color 0.0033(0.0039) |Steps 488(451.34) | Grad Norm 3.1934(8.7157) | Total Time 0.00(0.00)\nIter 4800 | Time 10.9269(11.0048) | Bit/dim 0.8969(0.9044) | Xent 0.1029(0.1076) | Xent Color 0.0183(0.0189) | Loss 2.5906(2.7500) | Error 0.0311(0.0343) | Error Color 0.0044(0.0038) |Steps 434(451.27) | Grad Norm 5.5558(7.8922) | Total Time 0.00(0.00)\nIter 4810 | Time 10.5343(10.9960) | Bit/dim 0.8953(0.9011) | Xent 0.0944(0.1062) | Xent Color 0.0118(0.0173) | Loss 2.4820(2.6963) | Error 0.0289(0.0337) | Error Color 0.0011(0.0032) |Steps 434(454.38) | Grad Norm 2.0089(6.8728) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0073 | Time 64.7044, Epoch Time 815.4135(778.9767), Bit/dim 0.8892(best: 0.9080), Xent 0.0671, Xent Color 0.0028. Loss 0.9067, Error 0.0210(best: 0.0221), Error Color 0.0002(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 4820 | Time 10.7092(11.0217) | Bit/dim 0.8936(0.8981) | Xent 0.0865(0.1049) | Xent Color 0.0105(0.0164) | Loss 2.5488(3.2218) | Error 0.0322(0.0334) | Error Color 0.0000(0.0027) |Steps 440(454.99) | Grad Norm 5.3356(6.2989) | Total Time 0.00(0.00)\nIter 4830 | Time 12.0769(11.0452) | Bit/dim 0.8867(0.8954) | Xent 0.0964(0.1043) | Xent Color 0.0140(0.0155) | Loss 2.5820(3.0475) | Error 0.0400(0.0342) | Error Color 0.0011(0.0025) |Steps 506(455.00) | Grad Norm 7.7008(6.0849) | Total Time 0.00(0.00)\nIter 4840 | Time 10.6606(11.0252) | Bit/dim 0.8879(0.8942) | Xent 0.0883(0.1047) | Xent Color 0.0138(0.0149) | Loss 2.5217(2.9185) | Error 0.0300(0.0340) | Error Color 0.0022(0.0024) |Steps 446(453.33) | Grad Norm 8.3886(6.8207) | Total Time 0.00(0.00)\nIter 4850 | Time 10.3349(11.0234) | Bit/dim 0.8885(0.8921) | Xent 0.0996(0.1040) | Xent Color 0.0114(0.0145) | Loss 2.5555(2.8211) | Error 0.0378(0.0338) | Error Color 0.0022(0.0023) |Steps 440(450.77) | Grad Norm 6.0562(7.4238) | Total Time 0.00(0.00)\nIter 4860 | Time 10.4748(11.0193) | Bit/dim 0.8792(0.8893) | Xent 0.1390(0.1052) | Xent Color 0.0120(0.0138) | Loss 2.5516(2.7432) | Error 0.0533(0.0338) | Error Color 0.0011(0.0019) |Steps 440(445.28) | Grad Norm 3.3728(6.9072) | Total Time 0.00(0.00)\nIter 4870 | Time 10.7824(10.9779) | Bit/dim 0.8740(0.8875) | Xent 0.1238(0.1061) | Xent Color 0.0128(0.0138) | Loss 2.4965(2.6904) | Error 0.0400(0.0337) | Error Color 0.0011(0.0020) |Steps 422(442.12) | Grad Norm 8.6185(7.2318) | Total Time 0.00(0.00)\nIter 4880 | Time 10.8104(10.9238) | Bit/dim 0.8793(0.8846) | Xent 0.1039(0.1049) | Xent Color 0.0081(0.0133) | Loss 2.5372(2.6430) | Error 0.0289(0.0331) | Error Color 0.0011(0.0019) |Steps 452(443.22) | Grad Norm 2.4485(6.4766) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0074 | Time 63.6873, Epoch Time 809.2891(779.8861), Bit/dim 0.8942(best: 0.8892), Xent 0.0624, Xent Color 0.0047. Loss 0.9110, Error 0.0209(best: 0.0210), Error Color 0.0005(best: 0.0000)\n===> Using batch size 900. Total 66 iterations/epoch.\nIter 4890 | Time 10.6669(10.8854) | Bit/dim 0.8729(0.8848) | Xent 0.1104(0.1031) | Xent Color 0.0148(0.0140) | Loss 2.5325(3.0945) | Error 0.0322(0.0324) | Error Color 0.0022(0.0021) |Steps 446(442.35) | Grad Norm 6.9647(7.9128) | Total Time 0.00(0.00)\nIter 4900 | Time 10.2333(10.8478) | Bit/dim 0.8686(0.8832) | Xent 0.1032(0.1034) | Xent Color 0.0135(0.0137) | Loss 2.4645(2.9455) | Error 0.0367(0.0324) | Error Color 0.0022(0.0018) |Steps 434(444.96) | Grad Norm 9.5204(8.3229) | Total Time 0.00(0.00)\nIter 4910 | Time 10.3570(10.8855) | Bit/dim 0.8703(0.8824) | Xent 0.1156(0.1030) | Xent Color 0.0141(0.0137) | Loss 2.4807(2.8347) | Error 0.0300(0.0322) | Error Color 0.0022(0.0020) |Steps 440(443.71) | Grad Norm 6.1625(8.9185) | Total Time 0.00(0.00)\nIter 4920 | Time 10.7328(10.8409) | Bit/dim 0.8727(0.8811) | Xent 0.1061(0.1045) | Xent Color 0.0093(0.0145) | Loss 2.5145(2.7575) | Error 0.0322(0.0323) | Error Color 0.0011(0.0023) |Steps 458(446.47) | Grad Norm 5.7268(10.0351) | Total Time 0.00(0.00)\nIter 4930 | Time 11.2225(10.8717) | Bit/dim 0.8772(0.8787) | Xent 0.1192(0.1036) | Xent Color 0.0156(0.0139) | Loss 2.5363(2.6922) | Error 0.0322(0.0319) | Error Color 0.0033(0.0021) |Steps 422(444.01) | Grad Norm 10.3617(9.7026) | Total Time 0.00(0.00)\nIter 4940 | Time 11.1679(10.8871) | Bit/dim 0.9120(0.8816) | Xent 0.0787(0.1025) | Xent Color 0.1628(0.0213) | Loss 2.6559(2.6576) | Error 0.0267(0.0318) | Error Color 0.0622(0.0052) |Steps 440(443.77) | Grad Norm 50.7377(13.1521) | Total Time 0.00(0.00)\nIter 4950 | Time 16.6078(11.8831) | Bit/dim 2.2892(1.2317) | Xent 1.0228(0.5160) | Xent Color 1.3565(1.1235) | Loss 6.4887(4.0077) | Error 0.3256(0.1387) | Error Color 0.4989(0.1512) |Steps 692(480.00) | Grad Norm 13.3803(36.6409) | Total Time 0.00(0.00)\nvalidating...\nEpoch 0075 | Time 90.9583, Epoch Time 864.9465(782.4379), Bit/dim 2.3095(best: 0.8892), Xent 0.5704, Xent Color 0.8185. Loss 2.6567, Error 0.1754(best: 0.0209), Error Color 0.3454(best: 0.0000)\n" ] ] ]

[ "code" ]

[ [ "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# FEEDFORWARD NEURAL NETWORK\n#---------------------------------------------------\nimport numpy as np\nf = lambda x: 1.0/(1.0 + np.exp(-x)) # activation function (use sigmoid)\nx = np.array([[-1], [1], [1]]) # random input vector of three numbers (3x1)\n\nW1 = np.random.uniform(-0.5, 0.5, 12).reshape(4, 3)\nb1 = np.random.uniform(-0.05, 0.05, 4).reshape(4, 1)\nW2 = np.random.uniform(-0.5, 0.5, 16).reshape(4, 4)\nb2 = np.random.uniform(-0.05, 0.05, 4).reshape(4, 1)\nW3 = np.random.uniform(-0.5, 0.5, 4).reshape(1, 4)\n\nprint(x)", "[[-1]\n [ 1]\n [ 1]]\n" ], [ "# Feedforward Neural Networks\nh1 = f(np.dot(W1, x) + b1) # calculate first hidden layer activations (4x1)\nh2 = f(np.dot(W2, h1) + b2) # calculate second hidden layer activations (4x1)\nout = np.dot(W3, h2) # output neuron (1x1)", "_____no_output_____" ], [ "# Feedforward Neural Networks\n#Check the Activation\nif out>0:\n print(\"Activated\")\nelse:\n print(\"Not Activated\")", "Not Activated\n" ], [ "import numpy as np\nfor i in range(0, 10):\n f = lambda x: 1.0/(1.0 + np.exp(-x)) # activation function (use sigmoid)\n x = np.array([[0], [0], [0]]) # if is set to 0 everything depend by the bias\n W1 = np.random.uniform(-0.5, 0.5, 12).reshape(4, 3)\n b1 = np.random.uniform(-0.05, 0.05, 4).reshape(4, 1)\n W2 = np.random.uniform(-0.5, 0.5, 16).reshape(4, 4)\n b2 = np.random.uniform(-0.05, 0.05, 4).reshape(4, 1)\n W3 = np.random.uniform(-0.5, 0.5, 4).reshape(1, 4)\n # Feedforward Neural Networks\n h1 = f(np.dot(W1, x) + b1) # calculate first hidden layer activations (4x1)\n h2 = f(np.dot(W2, h1) + b2) # calculate second hidden layer activations (4x1)\n out = np.dot(W3, h2) # output neuron (1x1)\n # Feedforward Neural Networks\n #Check the Activation\n if out>0:\n print(\"Activated\")\n else:\n print(\"Not Activated\")", "Not Activated\nNot Activated\nActivated\nActivated\nNot Activated\nNot Activated\nNot Activated\nActivated\nActivated\nNot Activated\n" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "##### Copyright 2019 The TensorFlow Hub Authors.\n\nLicensed under the Apache License, Version 2.0 (the \"License\");", "_____no_output_____" ] ], [ [ "# Copyright 2019 The TensorFlow Hub Authors. All Rights Reserved.\n#\n# Licensed under the Apache License, Version 2.0 (the \"License\");\n# you may not use this file except in compliance with the License.\n# You may obtain a copy of the License at\n#\n# http://www.apache.org/licenses/LICENSE-2.0\n#\n# Unless required by applicable law or agreed to in writing, software\n# distributed under the License is distributed on an \"AS IS\" BASIS,\n# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n# See the License for the specific language governing permissions and\n# limitations under the License.\n# ==============================================================================", "_____no_output_____" ], [ "#@title MIT License\n#\n# Copyright (c) 2017 François Chollet\n#\n# Permission is hereby granted, free of charge, to any person obtaining a\n# copy of this software and associated documentation files (the \"Software\"),\n# to deal in the Software without restriction, including without limitation\n# the rights to use, copy, modify, merge, publish, distribute, sublicense,\n# and/or sell copies of the Software, and to permit persons to whom the\n# Software is furnished to do so, subject to the following conditions:\n#\n# The above copyright notice and this permission notice shall be included in\n# all copies or substantial portions of the Software.\n#\n# THE SOFTWARE IS PROVIDED \"AS IS\", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR\n# IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,\n# FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL\n# THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER\n# LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING\n# FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER\n# DEALINGS IN THE SOFTWARE.", "_____no_output_____" ] ], [ [ "# Text Classification with Movie Reviews", "_____no_output_____" ], [ "<table class=\"tfo-notebook-buttons\" align=\"left\">\n <td>\n <a target=\"_blank\" href=\"https://www.tensorflow.org/hub/tutorials/tf2_text_classification\"><img src=\"https://www.tensorflow.org/images/tf_logo_32px.png\" />View on TensorFlow.org</a>\n </td>\n <td>\n <a target=\"_blank\" href=\"https://colab.research.google.com/github/tensorflow/hub/blob/master/examples/colab/tf2_text_classification.ipynb\"><img src=\"https://www.tensorflow.org/images/colab_logo_32px.png\" />Run in Google Colab</a>\n </td>\n <td>\n <a target=\"_blank\" href=\"https://github.com/tensorflow/hub/blob/master/examples/colab/tf2_text_classification.ipynb\"><img src=\"https://www.tensorflow.org/images/GitHub-Mark-32px.png\" />View source on GitHub</a>\n </td>\n <td>\n <a href=\"https://storage.googleapis.com/tensorflow_docs/hub/examples/colab/tf2_text_classification.ipynb\"><img src=\"https://www.tensorflow.org/images/download_logo_32px.png\" />Download notebook</a>\n </td>\n</table>", "_____no_output_____" ], [ "This notebook classifies movie reviews as *positive* or *negative* using the text of the review. This is an example of *binary*—or two-class—classification, an important and widely applicable kind of machine learning problem. \n\nWe'll use the [IMDB dataset](https://www.tensorflow.org/api_docs/python/tf/keras/datasets/imdb) that contains the text of 50,000 movie reviews from the [Internet Movie Database](https://www.imdb.com/). These are split into 25,000 reviews for training and 25,000 reviews for testing. The training and testing sets are *balanced*, meaning they contain an equal number of positive and negative reviews. \n\nThis notebook uses [tf.keras](https://www.tensorflow.org/guide/keras), a high-level API to build and train models in TensorFlow, and [TensorFlow Hub](https://www.tensorflow.org/hub), a library and platform for transfer learning. For a more advanced text classification tutorial using `tf.keras`, see the [MLCC Text Classification Guide](https://developers.google.com/machine-learning/guides/text-classification/).", "_____no_output_____" ], [ "## Setup", "_____no_output_____" ] ], [ [ "import numpy as np\n\nimport tensorflow as tf\nimport tensorflow_hub as hub\nimport tensorflow_datasets as tfds\n\nimport matplotlib.pyplot as plt\n\nprint(\"Version: \", tf.__version__)\nprint(\"Eager mode: \", tf.executing_eagerly())\nprint(\"Hub version: \", hub.__version__)\nprint(\"GPU is\", \"available\" if tf.config.list_physical_devices('GPU') else \"NOT AVAILABLE\")", "_____no_output_____" ] ], [ [ "## Download the IMDB dataset\n\nThe IMDB dataset is available on [TensorFlow datasets](https://github.com/tensorflow/datasets). The following code downloads the IMDB dataset to your machine (or the colab runtime):", "_____no_output_____" ] ], [ [ "train_data, test_data = tfds.load(name=\"imdb_reviews\", split=[\"train\", \"test\"], \n batch_size=-1, as_supervised=True)\n\ntrain_examples, train_labels = tfds.as_numpy(train_data)\ntest_examples, test_labels = tfds.as_numpy(test_data)", "_____no_output_____" ] ], [ [ "## Explore the data \n\nLet's take a moment to understand the format of the data. Each example is a sentence representing the movie review and a corresponding label. The sentence is not preprocessed in any way. The label is an integer value of either 0 or 1, where 0 is a negative review, and 1 is a positive review.", "_____no_output_____" ] ], [ [ "print(\"Training entries: {}, test entries: {}\".format(len(train_examples), len(test_examples)))", "_____no_output_____" ] ], [ [ "Let's print first 10 examples.", "_____no_output_____" ] ], [ [ "train_examples[:10]", "_____no_output_____" ] ], [ [ "Let's also print the first 10 labels.", "_____no_output_____" ] ], [ [ "train_labels[:10]", "_____no_output_____" ] ], [ [ "## Build the model\n\nThe neural network is created by stacking layers—this requires three main architectural decisions:\n\n* How to represent the text?\n* How many layers to use in the model?\n* How many *hidden units* to use for each layer?\n\nIn this example, the input data consists of sentences. The labels to predict are either 0 or 1.\n\nOne way to represent the text is to convert sentences into embeddings vectors. We can use a pre-trained text embedding as the first layer, which will have two advantages:\n* we don't have to worry about text preprocessing,\n* we can benefit from transfer learning.\n\nFor this example we will use a model from [TensorFlow Hub](https://www.tensorflow.org/hub) called [google/tf2-preview/gnews-swivel-20dim/1](https://tfhub.dev/google/tf2-preview/gnews-swivel-20dim/1).\n\nThere are three other models to test for the sake of this tutorial:\n* [google/tf2-preview/gnews-swivel-20dim-with-oov/1](https://tfhub.dev/google/tf2-preview/gnews-swivel-20dim-with-oov/1) - same as [google/tf2-preview/gnews-swivel-20dim/1](https://tfhub.dev/google/tf2-preview/gnews-swivel-20dim/1), but with 2.5% vocabulary converted to OOV buckets. This can help if vocabulary of the task and vocabulary of the model don't fully overlap.\n* [google/tf2-preview/nnlm-en-dim50/1](https://tfhub.dev/google/tf2-preview/nnlm-en-dim50/1) - A much larger model with ~1M vocabulary size and 50 dimensions.\n* [google/tf2-preview/nnlm-en-dim128/1](https://tfhub.dev/google/tf2-preview/nnlm-en-dim128/1) - Even larger model with ~1M vocabulary size and 128 dimensions.", "_____no_output_____" ], [ "Let's first create a Keras layer that uses a TensorFlow Hub model to embed the sentences, and try it out on a couple of input examples. Note that the output shape of the produced embeddings is a expected: `(num_examples, embedding_dimension)`.", "_____no_output_____" ] ], [ [ "model = \"https://tfhub.dev/google/tf2-preview/gnews-swivel-20dim/1\"\nhub_layer = hub.KerasLayer(model, output_shape=[20], input_shape=[], \n dtype=tf.string, trainable=True)\nhub_layer(train_examples[:3])", "_____no_output_____" ] ], [ [ "Let's now build the full model:", "_____no_output_____" ] ], [ [ "model = tf.keras.Sequential()\nmodel.add(hub_layer)\nmodel.add(tf.keras.layers.Dense(16, activation='relu'))\nmodel.add(tf.keras.layers.Dense(1))\n\nmodel.summary()", "_____no_output_____" ] ], [ [ "The layers are stacked sequentially to build the classifier:\n\n1. The first layer is a TensorFlow Hub layer. This layer uses a pre-trained Saved Model to map a sentence into its embedding vector. The model that we are using ([google/tf2-preview/gnews-swivel-20dim/1](https://tfhub.dev/google/tf2-preview/gnews-swivel-20dim/1)) splits the sentence into tokens, embeds each token and then combines the embedding. The resulting dimensions are: `(num_examples, embedding_dimension)`.\n2. This fixed-length output vector is piped through a fully-connected (`Dense`) layer with 16 hidden units.\n3. The last layer is densely connected with a single output node. This outputs logits: the log-odds of the true class, according to the model.", "_____no_output_____" ], [ "### Hidden units\n\nThe above model has two intermediate or \"hidden\" layers, between the input and output. The number of outputs (units, nodes, or neurons) is the dimension of the representational space for the layer. In other words, the amount of freedom the network is allowed when learning an internal representation.\n\nIf a model has more hidden units (a higher-dimensional representation space), and/or more layers, then the network can learn more complex representations. However, it makes the network more computationally expensive and may lead to learning unwanted patterns—patterns that improve performance on training data but not on the test data. This is called *overfitting*, and we'll explore it later.", "_____no_output_____" ], [ "### Loss function and optimizer\n\nA model needs a loss function and an optimizer for training. Since this is a binary classification problem and the model outputs a probability (a single-unit layer with a sigmoid activation), we'll use the `binary_crossentropy` loss function. \n\nThis isn't the only choice for a loss function, you could, for instance, choose `mean_squared_error`. But, generally, `binary_crossentropy` is better for dealing with probabilities—it measures the \"distance\" between probability distributions, or in our case, between the ground-truth distribution and the predictions.\n\nLater, when we are exploring regression problems (say, to predict the price of a house), we will see how to use another loss function called mean squared error.\n\nNow, configure the model to use an optimizer and a loss function:", "_____no_output_____" ] ], [ [ "model.compile(optimizer='adam',\n loss=tf.losses.BinaryCrossentropy(from_logits=True),\n metrics=[tf.metrics.BinaryAccuracy(threshold=0.0, name='accuracy')])", "_____no_output_____" ] ], [ [ "## Create a validation set\n\nWhen training, we want to check the accuracy of the model on data it hasn't seen before. Create a *validation set* by setting apart 10,000 examples from the original training data. (Why not use the testing set now? Our goal is to develop and tune our model using only the training data, then use the test data just once to evaluate our accuracy).", "_____no_output_____" ] ], [ [ "x_val = train_examples[:10000]\npartial_x_train = train_examples[10000:]\n\ny_val = train_labels[:10000]\npartial_y_train = train_labels[10000:]", "_____no_output_____" ] ], [ [ "## Train the model\n\nTrain the model for 40 epochs in mini-batches of 512 samples. This is 40 iterations over all samples in the `x_train` and `y_train` tensors. While training, monitor the model's loss and accuracy on the 10,000 samples from the validation set:", "_____no_output_____" ] ], [ [ "history = model.fit(partial_x_train,\n partial_y_train,\n epochs=40,\n batch_size=512,\n validation_data=(x_val, y_val),\n verbose=1)", "_____no_output_____" ] ], [ [ "## Evaluate the model\n\nAnd let's see how the model performs. Two values will be returned. Loss (a number which represents our error, lower values are better), and accuracy.", "_____no_output_____" ] ], [ [ "results = model.evaluate(test_data, test_labels)\n\nprint(results)", "_____no_output_____" ] ], [ [ "This fairly naive approach achieves an accuracy of about 87%. With more advanced approaches, the model should get closer to 95%.", "_____no_output_____" ], [ "## Create a graph of accuracy and loss over time\n\n`model.fit()` returns a `History` object that contains a dictionary with everything that happened during training:", "_____no_output_____" ] ], [ [ "history_dict = history.history\nhistory_dict.keys()", "_____no_output_____" ] ], [ [ "There are four entries: one for each monitored metric during training and validation. We can use these to plot the training and validation loss for comparison, as well as the training and validation accuracy:", "_____no_output_____" ] ], [ [ "acc = history_dict['accuracy']\nval_acc = history_dict['val_accuracy']\nloss = history_dict['loss']\nval_loss = history_dict['val_loss']\n\nepochs = range(1, len(acc) + 1)\n\n# \"bo\" is for \"blue dot\"\nplt.plot(epochs, loss, 'bo', label='Training loss')\n# b is for \"solid blue line\"\nplt.plot(epochs, val_loss, 'b', label='Validation loss')\nplt.title('Training and validation loss')\nplt.xlabel('Epochs')\nplt.ylabel('Loss')\nplt.legend()\n\nplt.show()", "_____no_output_____" ], [ "plt.clf() # clear figure\n\nplt.plot(epochs, acc, 'bo', label='Training acc')\nplt.plot(epochs, val_acc, 'b', label='Validation acc')\nplt.title('Training and validation accuracy')\nplt.xlabel('Epochs')\nplt.ylabel('Accuracy')\nplt.legend()\n\nplt.show()", "_____no_output_____" ] ], [ [ "In this plot, the dots represent the training loss and accuracy, and the solid lines are the validation loss and accuracy.\n\nNotice the training loss *decreases* with each epoch and the training accuracy *increases* with each epoch. This is expected when using a gradient descent optimization—it should minimize the desired quantity on every iteration.\n\nThis isn't the case for the validation loss and accuracy—they seem to peak after about twenty epochs. This is an example of overfitting: the model performs better on the training data than it does on data it has never seen before. After this point, the model over-optimizes and learns representations *specific* to the training data that do not *generalize* to test data.\n\nFor this particular case, we could prevent overfitting by simply stopping the training after twenty or so epochs. Later, you'll see how to do this automatically with a callback.", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ] ]

[ "CC-BY-4.0" ]

[ "CC-BY-4.0" ]

[ "CC-BY-4.0" ]

[ [ [ "# Sample grouping\nWe are going to linger into the concept of sample groups. As in the previous\nsection, we will give an example to highlight some surprising results. This\ntime, we will use the handwritten digits dataset.", "_____no_output_____" ] ], [ [ "from sklearn.datasets import load_digits\n\ndigits = load_digits()\ndata, target = digits.data, digits.target", "_____no_output_____" ] ], [ [ "We will recreate the same model used in the previous exercise:\na logistic regression classifier with preprocessor to scale the data.", "_____no_output_____" ] ], [ [ "from sklearn.preprocessing import StandardScaler\nfrom sklearn.linear_model import LogisticRegression\nfrom sklearn.pipeline import make_pipeline\n\nmodel = make_pipeline(StandardScaler(), LogisticRegression())", "_____no_output_____" ] ], [ [ "We will use the same baseline model. We will use a `KFold` cross-validation\nwithout shuffling the data at first.", "_____no_output_____" ] ], [ [ "from sklearn.model_selection import cross_val_score, KFold\n\ncv = KFold(shuffle=False)\ntest_score_no_shuffling = cross_val_score(model, data, target, cv=cv,\n n_jobs=-1)\nprint(f\"The average accuracy is \"\n f\"{test_score_no_shuffling.mean():.3f} +/- \"\n f\"{test_score_no_shuffling.std():.3f}\")", "_____no_output_____" ] ], [ [ "Now, let's repeat the experiment by shuffling the data within the\ncross-validation.", "_____no_output_____" ] ], [ [ "cv = KFold(shuffle=True)\ntest_score_with_shuffling = cross_val_score(model, data, target, cv=cv,\n n_jobs=-1)\nprint(f\"The average accuracy is \"\n f\"{test_score_with_shuffling.mean():.3f} +/- \"\n f\"{test_score_with_shuffling.std():.3f}\")", "_____no_output_____" ] ], [ [ "We observe that shuffling the data improves the mean accuracy.\nWe could go a little further and plot the distribution of the testing\nscore. We can first concatenate the test scores.", "_____no_output_____" ] ], [ [ "import pandas as pd\n\nall_scores = pd.DataFrame(\n [test_score_no_shuffling, test_score_with_shuffling],\n index=[\"KFold without shuffling\", \"KFold with shuffling\"],\n).T", "_____no_output_____" ] ], [ [ "Let's plot the distribution now.", "_____no_output_____" ] ], [ [ "import matplotlib.pyplot as plt\nimport seaborn as sns\n\nall_scores.plot.hist(bins=10, edgecolor=\"black\", density=True, alpha=0.7)\nplt.xlim([0.8, 1.0])\nplt.xlabel(\"Accuracy score\")\nplt.legend(bbox_to_anchor=(1.05, 0.8), loc=\"upper left\")\n_ = plt.title(\"Distribution of the test scores\")", "_____no_output_____" ] ], [ [ "The cross-validation testing error that uses the shuffling has less\nvariance than the one that does not impose any shuffling. It means that some\nspecific fold leads to a low score in this case.", "_____no_output_____" ] ], [ [ "print(test_score_no_shuffling)", "_____no_output_____" ] ], [ [ "Thus, there is an underlying structure in the data that shuffling will break\nand get better results. To get a better understanding, we should read the\ndocumentation shipped with the dataset.", "_____no_output_____" ] ], [ [ "print(digits.DESCR)", "_____no_output_____" ] ], [ [ "If we read carefully, 13 writers wrote the digits of our dataset, accounting\nfor a total amount of 1797 samples. Thus, a writer wrote several times the\nsame numbers. Let's suppose that the writer samples are grouped.\nSubsequently, not shuffling the data will keep all writer samples together\neither in the training or the testing sets. Mixing the data will break this\nstructure, and therefore digits written by the same writer will be available\nin both the training and testing sets.\n\nBesides, a writer will usually tend to write digits in the same manner. Thus,\nour model will learn to identify a writer's pattern for each digit instead of\nrecognizing the digit itself.\n\nWe can solve this problem by ensuring that the data associated with a writer\nshould either belong to the training or the testing set. Thus, we want to\ngroup samples for each writer.\n\nHere, we will manually define the group for the 13 writers.", "_____no_output_____" ] ], [ [ "from itertools import count\nimport numpy as np\n\n# defines the lower and upper bounds of sample indices\n# for each writer\nwriter_boundaries = [0, 130, 256, 386, 516, 646, 776, 915, 1029,\n 1157, 1287, 1415, 1545, 1667, 1797]\ngroups = np.zeros_like(target)\nlower_bounds = writer_boundaries[:-1]\nupper_bounds = writer_boundaries[1:]\n\nfor group_id, lb, up in zip(count(), lower_bounds, upper_bounds):\n groups[lb:up] = group_id", "_____no_output_____" ] ], [ [ "We can check the grouping by plotting the indices linked to writer ids.", "_____no_output_____" ] ], [ [ "plt.plot(groups)\nplt.yticks(np.unique(groups))\nplt.xticks(writer_boundaries, rotation=90)\nplt.xlabel(\"Target index\")\nplt.ylabel(\"Writer index\")\n_ = plt.title(\"Underlying writer groups existing in the target\")", "_____no_output_____" ] ], [ [ "Once we group the digits by writer, we can use cross-validation to take this\ninformation into account: the class containing `Group` should be used.", "_____no_output_____" ] ], [ [ "from sklearn.model_selection import GroupKFold\n\ncv = GroupKFold()\ntest_score = cross_val_score(model, data, target, groups=groups, cv=cv,\n n_jobs=-1)\nprint(f\"The average accuracy is \"\n f\"{test_score.mean():.3f} +/- \"\n f\"{test_score.std():.3f}\")", "_____no_output_____" ] ], [ [ "We see that this strategy is less optimistic regarding the model statistical\nperformance. However, this is the most reliable if our goal is to make\nhandwritten digits recognition writers independent. Besides, we can as well\nsee that the standard deviation was reduced.", "_____no_output_____" ] ], [ [ "all_scores = pd.DataFrame(\n [test_score_no_shuffling, test_score_with_shuffling, test_score],\n index=[\"KFold without shuffling\", \"KFold with shuffling\",\n \"KFold with groups\"],\n).T", "_____no_output_____" ], [ "all_scores.plot.hist(bins=10, edgecolor=\"black\", density=True, alpha=0.7)\nplt.xlim([0.8, 1.0])\nplt.xlabel(\"Accuracy score\")\nplt.legend(bbox_to_anchor=(1.05, 0.8), loc=\"upper left\")\n_ = plt.title(\"Distribution of the test scores\")", "_____no_output_____" ] ], [ [ "As a conclusion, it is really important to take any sample grouping pattern\ninto account when evaluating a model. Otherwise, the results obtained will\nbe over-optimistic in regards with reality.", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import numpy as np\nimport matplotlib.pyplot as plt\nfrom matplotlib import pyplot as plt\nimport random\nfrom numba import njit, prange, cuda\nimport math", "_____no_output_____" ], [ "@njit\ndef uniform_samples(n=10000):\n samples=[]\n for i in range (0, n):\n samples.append(random.random())\n return samples;\n\nu_samples = uniform_samples()\n\nplt.figure(figsize=(9, 5))\n\nplt.hlines(y=1, xmin=0, xmax=1, colors='r',zorder=3, lw=1, label=\"PDF\") # f(x)=1/a+b\nplt.hist(u_samples,alpha=.9, rwidth=0.9, bins=20, density=True, label=\"uniform distribution bins:20\")\n\nhistogram, bins = np.histogram(u_samples,bins=100,density=True)\nbin_centers = 0.5*(bins[1:] + bins[:-1])\nplt.plot(bin_centers, histogram, color='black',alpha=.5, label=\"uniform distribution bins:100\")\n\n\nplt.legend()\nplt.show()", "_____no_output_____" ], [ "@njit\ndef u_samples_sum(n=10000):\n total=0\n for i in range (0, n):\n total+=random.random()\n return total;\n\n@njit\ndef get_normal_samples(n=10000):\n samples=[]\n total=0\n for i in range (0, n):\n uniform_sample_total = u_samples_sum()\n samples.append(uniform_sample_total)\n total+=uniform_sample_total\n return samples,total;\n\ndef get_mu_std(normal_samples,total,n):\n mu = total/n\n std =math.sqrt(1/n * sum(pow(x-mu,2) for x in normal_samples))\n return mu,std;\n\n#mu: mean, std: standard di\ndef get_normal_pdf(mu,std):\n pi = math.pi\n e = math.e\n sqrt = math.sqrt\n return lambda x:(1/(std*sqrt(2*pi))) * pow(e,(-1/2)*(pow((x-mu)/std,2)));\n\nn = 10000\n\nnormal_samples,total = get_normal_samples(n)\n\nmu,std = get_mu_std(normal_samples,total,n)\n\npdf = get_normal_pdf(mu,std)\n\nplt.figure(figsize=(9, 5))\n\nplt.hist(normal_samples,rwidth=0.9, bins=20,density=True, label=\"normal distribution bins:20\")\n\nhistogram, bins = np.histogram(normal_samples,bins=500,density=True)\nbin_centers = 0.5*(bins[1:] + bins[:-1])\nplt.plot(bin_centers, histogram,color='black',alpha=.4, label=\"normal distribution bins:500\")\n\npdf_normal_samples=[pdf(x) for x in bin_centers]\n\nplt.plot(bin_centers, pdf_normal_samples,color='red',alpha=1, label=\"normal PDF\")\n\n\nplt.legend()\nplt.show()", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# OSMnx + igraph for faster performance\n\nAuthor: [Geoff Boeing](https://geoffboeing.com/)\n\nNetworkX is ubiquitous, easy to use, and sufficiently fast for most use cases. But it can be slow for analyzing very large graphs because it is pure Python, trading off speed for ease of use. Fortunately, converting OSMnx-created NetworkX graphs to other graph libraries' types is relatively quick and simple and makes analysis much faster for those use cases where it's needed. For example, you might consider converting your NetworkX graph to igraph, graph-tool, or cugraph for analysis. This example demonstrates igraph.\n\nFirst install [igraph](https://igraph.org/python/) or run Jupyter from the [Docker container](https://hub.docker.com/r/gboeing/osmnx) (which already has it installed along with OSMnx and NetworkX).\n\n - [Overview of OSMnx](http://geoffboeing.com/2016/11/osmnx-python-street-networks/)\n - [GitHub repo](https://github.com/gboeing/osmnx)\n - [Examples, demos, tutorials](https://github.com/gboeing/osmnx-examples)\n - [Documentation](https://osmnx.readthedocs.io/en/stable/)\n - [Journal article/citation](http://geoffboeing.com/publications/osmnx-complex-street-networks/)", "_____no_output_____" ] ], [ [ "import operator\n\nimport igraph as ig\nimport networkx as nx\nimport numpy as np\nimport osmnx as ox\n\n%matplotlib inline\nox.config(use_cache=True, log_console=False)\nprint(ox.__version__)\nprint(ig.__version__)\n\nweight = \"length\"", "1.1.2\n0.9.9\n" ] ], [ [ "## Construct graphs", "_____no_output_____" ] ], [ [ "# create networkx graph\nG_nx = ox.graph_from_place(\"Piedmont, CA, USA\", network_type=\"drive\")\nosmids = list(G_nx.nodes)\nG_nx = nx.relabel.convert_node_labels_to_integers(G_nx)\n\n# give each node its original osmid as attribute since we relabeled them\nosmid_values = {k: v for k, v in zip(G_nx.nodes, osmids)}\nnx.set_node_attributes(G_nx, osmid_values, \"osmid\")", "/Users/junhyun/.pyenv/versions/3.8.5/envs/openstreetmap/lib/python3.8/site-packages/osmnx/geocoder.py:110: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.\n gdf = gdf.append(_geocode_query_to_gdf(q, wr, by_osmid))\n" ], [ "%%time\n# convert networkx graph to igraph\nG_ig = ig.Graph(directed=True)\nG_ig.add_vertices(G_nx.nodes)\nG_ig.add_edges(G_nx.edges())\nG_ig.vs[\"osmid\"] = osmids\nG_ig.es[weight] = list(nx.get_edge_attributes(G_nx, weight).values())", "CPU times: user 1.56 ms, sys: 27 µs, total: 1.59 ms\nWall time: 1.6 ms\n" ], [ "assert len(G_nx.nodes()) == G_ig.vcount()\nassert len(G_nx.edges()) == G_ig.ecount()", "_____no_output_____" ] ], [ [ "## Shortest paths", "_____no_output_____" ] ], [ [ "source = list(G_nx.nodes())[0]\ntarget = list(G_nx.nodes())[-1]", "_____no_output_____" ], [ "%%time\npath1 = G_ig.get_shortest_paths(v=source, to=target, weights=weight)[0]", "CPU times: user 168 µs, sys: 19 µs, total: 187 µs\nWall time: 193 µs\n" ], [ "%%time\npath2 = nx.shortest_path(G_nx, source, target, weight=weight)", "CPU times: user 543 µs, sys: 1 µs, total: 544 µs\nWall time: 549 µs\n" ], [ "assert path1 == path2", "_____no_output_____" ], [ "%%time\npath_length1 = G_ig.shortest_paths(source=source, target=target, weights=weight)[0][0]", "CPU times: user 106 µs, sys: 13 µs, total: 119 µs\nWall time: 122 µs\n" ], [ "%%time\npath_length2 = nx.shortest_path_length(G_nx, source, target, weight)", "CPU times: user 482 µs, sys: 0 ns, total: 482 µs\nWall time: 485 µs\n" ], [ "assert path_length1 == path_length2", "_____no_output_____" ] ], [ [ "## Centrality analysis", "_____no_output_____" ] ], [ [ "%%time\ncloseness1 = G_ig.closeness(vertices=None, mode=\"ALL\", cutoff=None, weights=weight, normalized=True)\nmax_closeness1 = np.argmax(closeness1)", "CPU times: user 10.3 ms, sys: 259 µs, total: 10.6 ms\nWall time: 10.6 ms\n" ], [ "%%time\ncloseness2 = nx.closeness_centrality(G_nx, distance=weight, wf_improved=True)\nmax_closeness2 = max(closeness2.items(), key=operator.itemgetter(1))[0]", "CPU times: user 573 ms, sys: 7.44 ms, total: 580 ms\nWall time: 587 ms\n" ], [ "# confirm same node has max closeness in both graphs\nassert G_nx.nodes[max_closeness2][\"osmid\"] == G_ig.vs[max_closeness1][\"osmid\"]", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "<a href=\"https://colab.research.google.com/github/harnalashok/deeplearning-sequences/blob/main/tf%20data%20API.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>", "_____no_output_____" ] ], [ [ "#Last amended: 12th March, 2021\r\n# My folder: harnalashok/github/deeplearning-sequences\r\n# References:\r\n# https://www.tensorflow.org/tutorials/text/text_classification_rnn\r\n# https://colab.research.google.com/github/tensorflow/docs/blob/master/site/en/guide/data.ipynb#scrollTo=m5bz7R1xhX1f\r\n# https://stackoverflow.com/a/49579995/3282777\r\n# https://www.tensorflow.org/tutorials/load_data/text\r\n\r\n# Objectives:\r\n# i) Learning to work with tensors\r\n# ii) Learning to work with tf.data API\r\n# iii) Text Classification--Work in progess", "_____no_output_____" ], [ "# 1.0 Call libraries\r\nimport numpy as np\r\nimport tensorflow_datasets as tfds\r\nimport tensorflow as tf\r\nimport matplotlib.pyplot as plt\r\nimport os", "_____no_output_____" ], [ "# 1.1 More libraries\r\nfrom tensorflow.keras import utils\r\nfrom tensorflow.keras import preprocessing\r\nfrom tensorflow.keras.layers.experimental.preprocessing import TextVectorization\r\n", "_____no_output_____" ], [ "# 1.2 Set numpy decimal printoptions\r\n# Limit display to precision of 3\r\n\r\nnp.set_printoptions(precision=3)", "_____no_output_____" ], [ "# 1.3\r\nfrom IPython.core.interactiveshell import InteractiveShell\r\nInteractiveShell.ast_node_interactivity = \"all\"", "_____no_output_____" ] ], [ [ "# Introduction to tensors\r\nRefer this [page](https://www.tensorflow.org/guide/tensor)\r\n\r\nTensors are (kind of) like np.arrays. All tensors are immutable like Python numbers and strings: you can never update the contents of a tensor, only create a new one. \r\nRank in tensor is akin to dimensions in numpy.", "_____no_output_____" ], [ "### Basics", "_____no_output_____" ] ], [ [ "# 2.0 Rank 0\r\n# A \"scalar\" or \"rank-0\" tensor . \r\n# A scalar contains a single value,\r\n# and no \"axes\".\r\n\r\nprint( tf.constant(4))", "tf.Tensor(4, shape=(), dtype=int32)\n" ], [ "# 2.1 Rank 1\r\n# A \"vector\" or \"rank-1\" tensor is like\r\n# a list of values. A vector has one axis:\r\n\r\nprint(tf.constant([2.0, 3.0, 4.0]))", "tf.Tensor([2. 3. 4.], shape=(3,), dtype=float32)\n" ], [ "# 2.2 Rank 2\r\n# A \"matrix\" or \"rank-2\" tensor has two axes:\r\n\r\ns=tf.constant(\r\n [\r\n [1, 2],\r\n [3, 4],\r\n [5, 6]\r\n ]\r\n )\r\n\r\nprint(s)", "_____no_output_____" ], [ "# 2.3 Rank 3\r\n# There can be an arbitrary number of\r\n# axes (sometimes called \"dimensions\")\r\n\r\nrank_3_tensor = tf.constant(\r\n [\r\n [\r\n [0, 1, 2, 3, 4],\r\n [5, 6, 7, 8, 9]\r\n ],\r\n [\r\n [10, 11, 12, 13, 14],\r\n [15, 16, 17, 18, 19]\r\n ],\r\n [\r\n [20, 21, 22, 23, 24],\r\n [25, 26, 27, 28, 29]\r\n ],\r\n ]\r\n )\r\n\r\nrank_3_tensor\r\n", "_____no_output_____" ] ], [ [ "### Tensor to numpy\r\n\r\nConvert a tensor to numpy array using `np.array` method or using `tensor.numpy` method:", "_____no_output_____" ] ], [ [ "# 3.0 Tensor to numpy\r\nnp.array(s)\r\nprint()\r\ns.numpy()\r\nprint()\r\nnp.array(rank_3_tensor)\r\nprint()\r\nrank_3_tensor.numpy()", "_____no_output_____" ] ], [ [ "### Basic maths on tensors\r\n\r\nBasic math on tensors, including addition, element-wise multiplication, and matrix multiplication.", "_____no_output_____" ] ], [ [ "# 3.1\r\na = tf.constant([[1, 2],\r\n [3, 4]])\r\nb = tf.constant([[1, 2],\r\n [1, 0]])\r\nc = tf.ones([2,3])", "_____no_output_____" ], [ "# 3.2\r\na+b\r\nprint()\r\ntf.add(a,b)\r\n\r\nprint()\r\na * b\r\nprint()\r\ntf.multiply(a,b)\r\n\r\nprint()\r\ntf.matmul(a,b)\r\n", "_____no_output_____" ], [ "# 3.3 This fails. Tensors are very sensitive \r\n# to data types\r\n\r\nr = tf.constant([1.0,2.0], shape = [2,1])\r\np = tf.constant(5)\r\nr * p", "_____no_output_____" ], [ "# 3.4 This also fails\r\n# even though both operands\r\n# are floats\r\n\r\nr = tf.constant([1.0,2.0], shape = [2,1])\r\nr\r\np = tf.constant(5)\r\nr * tf.cast(p, tf.float16)", "_____no_output_____" ], [ "# 3.5 This succeeds\r\n\r\nr = tf.constant([1.0,2.0], shape = [2,1])\r\nr\r\np = tf.constant(5)\r\nr * tf.cast(p, tf.float32)", "_____no_output_____" ] ], [ [ "#### Some operations on tensors", "_____no_output_____" ] ], [ [ "# 4.0\r\nc = tf.constant([[4.0, 5.0], [10.0, 1.0]])\r\n\r\n# 4.1 Find the largest value\r\nprint(tf.reduce_max(c))\r\n\r\n# 4.2 Find the index of the largest value\r\nprint(tf.argmax(c))\r\n\r\n# 4.3 Compute the softmax\r\n# Note that each 'row'\r\n# (or axis 0) or putput \r\n# sums to 1\r\nprint(tf.nn.softmax(c))\r\n", "_____no_output_____" ] ], [ [ "#### Some vocabulary\r\nTensors have shapes. Some vocabulary:\r\n\r\n> **Shape**: The length (number of elements) of each of the axes of a tensor. \r\n> **Rank**: Number of tensor axes. A scalar has rank 0, a vector has rank 1, a matrix is rank 2. \r\n> **Axis** or Dimension: A particular dimension of a tensor. \r\n> **Size**: The total number of items in the tensor, the product shape vector.\r\n", "_____no_output_____" ] ], [ [ "# 5.0\r\nrank_4_tensor = tf.zeros([3, 2, 4, 5])\r\n# axis 0 is 3\r\n# axis -1 is 5", "_____no_output_____" ] ], [ [ "\r\n\r\n\r\n", "_____no_output_____" ] ], [ [ "# 5.2 Like in numpy arrays, we have attributes:\r\n# dtype, ndim, shape\r\n\r\nrank_4_tensor.dtype\r\nprint()\r\nrank_4_tensor.ndim\r\nprint()\r\nrank_4_tensor.shape\r\nprint()\r\nrank_4_tensor.shape[0]\r\nprint()\r\nrank_4_tensor.shape[-1]\r\nprint()\r\ntf.size(rank_4_tensor).numpy()", "_____no_output_____" ] ], [ [ "It is important to keep in mind inherent or implied meaning of each axis. In the above example, here are the implied meanings: \r\n\r\n*Batch size:* 3 \r\n*Depth*: 2 \r\n*Width*: 4 \r\n*Height*: 5 ", "_____no_output_____" ], [ "### Indexing (See [here](https://www.tensorflow.org/guide/tensor#indexing))", "_____no_output_____" ], [ "Single-axis indexing\r\n\r\nTensorFlow follows standard Python indexing rules, similar to indexing a list or a string in Python, and the basic rules for NumPy indexing.\r\n\r\n> indexes start at 0 \r\n> negative indices count backwards from the end \r\n> colons, :, are used for slices: start:stop:step", "_____no_output_____" ] ], [ [ "# 6.0 Sample 1-axes tensor\r\nrank_1_tensor = tf.constant([0, 1, 1, 2, 3, 5, 8, 13, 21, 34])\r\nprint(rank_1_tensor.numpy())\r\n", "[ 0 1 1 2 3 5 8 13 21 34]\n" ], [ "# 6.2 Indexing with a : slice keeps the axis:\r\n\r\n#6.2.1 Everything\r\nrank_1_tensor[:].numpy()\r\nprint()\r\n\r\n#6.2.2 Before 4:\r\nrank_1_tensor[:4].numpy()\r\nprint()\r\n\r\n#6.2.3 From 4 to the end\r\nrank_1_tensor[4:].numpy()\r\nprint()\r\n\r\n#6.2.4 From 2, before 7:\r\nrank_1_tensor[2:7].numpy()\r\nprint()\r\n\r\n#6.2.5 Every other item\r\nrank_1_tensor[::2].numpy()\r\nprint()\r\n\r\n#6.2.6 Reversed\r\nrank_1_tensor[::-1].numpy()\r\nprint()", "_____no_output_____" ] ], [ [ "### Manipulating shapes", "_____no_output_____" ] ], [ [ "# 7.0 Shape returns a `TensorShape` object\r\n# that shows the size along each axis\r\n\r\nx = tf.constant([[1], [2], [3]])\r\nprint(x.shape)\r\n", "_____no_output_____" ] ], [ [ "You can reshape a tensor into a new shape. The tf.reshape operation is fast and cheap as the underlying data does not need to be duplicated.", "_____no_output_____" ] ], [ [ "# 7.1 You can reshape a tensor to a new shape.\r\n# Note that you're passing in a list\r\n\r\nreshaped = tf.reshape(x, [1, 3])\r\nreshaped.shape", "_____no_output_____" ], [ "# 7.2 We created this tensor earlier\r\n\r\nprint(rank_3_tensor)", "_____no_output_____" ] ], [ [ "If you flatten a tensor you can see what order it is laid out in memory.", "_____no_output_____" ] ], [ [ "#7.3 A `-1` passed in the `shape` argument\r\n# says \"Whatever fits\".\r\n\r\nprint(tf.reshape(rank_3_tensor, [-1]))", "_____no_output_____" ] ], [ [ "Typically the only reasonable use of tf.reshape is to combine or split adjacent axes (or add/remove 1s).\r\n\r\nFor this 3x2x5 tensor, reshaping to (3x2)x5 or 3x(2x5) are both reasonable things to do, as the slices do not mix:", "_____no_output_____" ] ], [ [ "# 7.4\r\nprint(tf.reshape(rank_3_tensor, [3*2, 5]), \"\\n\")\r\nprint(tf.reshape(rank_3_tensor, [3, -1]))", "_____no_output_____" ] ], [ [ "#### Reshaping can be a mess\r\n\r\nReshaping will \"work\" for any new shape with the same total number of elements, but it will not do anything useful if you do not respect the order of the axes. It will be a mess.", "_____no_output_____" ] ], [ [ "# 8.0 Bad examples: don't do this\r\n# Only multiply adjacent indices\r\n\r\n# 8.1 You can't reorder axes with reshape.\r\nprint(tf.reshape(rank_3_tensor, [2, 3, 5]), \"\\n\") \r\n\r\n# 8.2 This is a mess\r\nprint(tf.reshape(rank_3_tensor, [5, 6]), \"\\n\")\r\n", "_____no_output_____" ] ], [ [ "### Broadcasting", "_____no_output_____" ], [ "Broadcasting is a concept borrowed from the equivalent feature in NumPy. In short, under certain conditions, smaller tensors are \"stretched\" automatically to fit larger tensors when running combined operations on them.\r\n\r\nThe simplest and most common case is when you attempt to multiply or add a tensor to a scalar. In that case, the scalar is broadcast to be the same shape as the other argument. ", "_____no_output_____" ] ], [ [ "# 9.0\r\nx = tf.constant([1, 2, 3])\r\ny = tf.constant(2)\r\nz = tf.constant([2, 2, 2])", "_____no_output_____" ], [ "# 9.1\r\nx * y\r\nx + z", "_____no_output_____" ] ], [ [ "Likewise, axes with length 1 can be stretched out to match the other arguments. Both arguments can be stretched in the same computation.\r\n\r\nIn this case a 3x1 matrix is element-wise multiplied by a 1x4 matrix to produce a 3x4 matrix. Note how the leading 1 is optional: The shape of y is [4].", "_____no_output_____" ] ], [ [ "# 9.2 These are the same computations\r\nx = tf.reshape(x,[3,1])\r\ny = tf.range(1, 5)\r\n\r\n# 9.2.1\r\nprint(x, \"\\n\")\r\nprint(y, \"\\n\")\r\nprint(tf.multiply(x, y))\r\n", "_____no_output_____" ], [ "# 9.2.2\r\nx = tf.constant([1, 2, 3])\r\ny = tf.constant(2)\r\nz = tf.constant([2, 2, 2])\r\n\r\n# 9.2.3 All of these are the same computation\r\nprint(tf.multiply(x, 2))\r\nprint(x * y)\r\nprint(x * z)\r\n", "_____no_output_____" ] ], [ [ "#### Logical operations", "_____no_output_____" ] ], [ [ "# 9.3 Logical operations on tensors\r\n# Pl refer: tf.math\r\n# https://www.tensorflow.org/api_docs/python/tf/math\r\n\r\nprint(tf.constant([1,2,10]) < 5)\r\nprint(tf.constant([True, False, True,True], dtype = tf.bool))\r\n\r\n# 9.4 A tensor can be reshaped as per its shape argument\r\n\r\nprint(tf.constant(np.arange(20), shape = [5,4])) \r\nx = tf.constant(np.arange(20), shape = [5,4])\r\n# Display Ist two rows and columns\r\nx[: 2, :2]", "_____no_output_____" ], [ "# 9.5 Show first and third rows\r\n# This works as an exception\r\n\r\nx[tf.constant([True, False,True,False, False], dtype = tf.bool)]\r\n\r\n# But the following do not work:\r\n# Valid indices are: integers, slices (`:`),\r\n# ellipsis (`...`), tf.newaxis (`None`)\r\n# and scalar tf.int32/tf.int64 tensors \r\n\r\n#x[\r\n# tf.constant([True, False,True,False, False], dtype = tf.bool),\r\n# tf.constant([True, False,True,False], dtype = tf.bool)\r\n# ]\r\n\r\n#x[\r\n# tf.constant([1,0,0,1,0], dtype = tf.int32),\r\n# tf.constant([1,0,1,0], dtype = tf.int32)\r\n# ]\r\n", "_____no_output_____" ] ], [ [ "### Define custom loss function\r\nRefer: Page 384 of Book by Aurelion", "_____no_output_____" ] ], [ [ "# 10.0 Define a model\r\nimport pandas as pd\r\nfrom tensorflow import keras\r\nfrom sklearn.datasets import load_boston\r\n# 10.1\r\nX,y = load_boston(return_X_y= True)\r\nX.shape # (506, 13)\r\n# 10.2\r\nmodel = keras.models.Sequential(\r\n [\r\n keras.layers.Dense(5, activation = 'relu'),\r\n keras.layers.Dense(1,activation = 'sigmoid')\r\n ]\r\n )\r\n", "_____no_output_____" ] ], [ [ "#### Huber loss\r\n\r\nFor a comparative picture of loss functions being used in Regression, see [here](https://heartbeat.fritz.ai/5-regression-loss-functions-all-machine-learners-should-know-4fb140e9d4b0). Briefly, if data is noisy, use MAE and if their are more outliers or noise, use huber loss. \r\n\r\nHuber loss is less sensitive to outliers in data than the squared error loss. It’s also differentiable at 0 (unlike MAE). It’s basically absolute error, which becomes quadratic when error is small. How small that error has to be to make it quadratic depends on a hyperparameter, 𝛿 (delta), which can be tuned. \r\n\r\n\r\n\r\n\r\n\r\n\r\n", "_____no_output_____" ] ], [ [ "# 10.3 Custom loss function\r\ndef huber_loss(y_true,y_pred):\r\n error = y_true - y_pred\r\n is_small_error = tf.abs(error) < 1.0\r\n squared_loss = tf.square(error) / 2.0\r\n linear_loss = tf.abs(error) - 0.5\r\n return tf.where(is_small_error, squared_loss, linear_loss)\r\n\r\n", "_____no_output_____" ], [ "# 10.4 Compile and run the model with our loss function\r\nmodel.compile(loss= huber_loss, metrics = \"mse\", optimizer = \"nadam\")\r\nmodel.fit(X,y,epochs = 20)", "_____no_output_____" ] ], [ [ "### Generate random tensors\r\nRefer this [article](https://www.tensorflow.org/guide/random_numbers)", "_____no_output_____" ] ], [ [ "# 11.0 Generate random data using\r\n# Generator object\r\ng1 = tf.random.Generator.from_seed(1)\r\n\r\n# 2.0.1 Use object 'g1'\r\ng1.normal([2,3])\r\nprint()\r\n# 2.0.2\r\ng1.uniform([1])\r\nprint()\r\n\r\n# 11.1 Generate random data\r\n# directly\r\ntf.random.uniform([4]) # shape is (4,)\r\nprint()\r\ntf.random.uniform([4]).shape\r\nprint()\r\ntf.random.uniform([4]).numpy()\r\nprint()\r\n\r\n# 11.2\r\ntf.random.normal(shape = (10,4), mean = 3,stddev=1.3)\r\nprint()\r\ntf.random.normal(shape = (10,4), mean = 3,stddev=1.3).shape\r\nprint()\r\n\r\n# 2.2\r\ntf.random.uniform([4, 100], maxval=100, dtype=tf.int32)", "_____no_output_____" ] ], [ [ "# tf.data.Dataset API\r\nUnderstanding Dataset object. The `tf.data.Dataset` API supports writing descriptive and efficient **input pipelines**. Dataset usage follows a common pattern (see [the right-panel here](https://www.tensorflow.org/api_docs/python/tf/data/Dataset) and a colab page with example [here](https://colab.research.google.com/github/tensorflow/docs/blob/master/site/en/guide/data.ipynb#scrollTo=k_5N7CdNGYAa) ):\r\n\r\n> Create a source dataset from your input data.\r\n\r\n> Write dataset transformations to preprocess the data.\r\n\r\n> Iterate over the dataset and process the elements.\r\n\r\nIteration happens in a streaming fashion, so the full dataset does not need to fit into memory. That is processing is performed batch-wise. \r\n\r\nThe simplest way to create a dataset is to create it from a python list:", "_____no_output_____" ], [ "## from_tensors() vs from_tensor_slices()\r\nSee this [stackoverflow answer](https://stackoverflow.com/a/49579995/3282777) \r\n\r\nRead `from_tensor_slices()` as `to_tensor_slices()`", "_____no_output_____" ] ], [ [ "# 1.0 Call libraries\r\nimport numpy as np\r\nimport tensorflow_datasets as tfds\r\nimport tensorflow as tf\r\nimport matplotlib.pyplot as plt\r\nimport os", "_____no_output_____" ], [ "# 1.1 More libraries\r\nfrom tensorflow.keras import utils\r\nfrom tensorflow.keras import preprocessing\r\nimport pathlib\r\nfrom tensorflow.keras.layers.experimental.preprocessing import TextVectorization", "_____no_output_____" ], [ "# 1.2 Set numpy decimal printoptions\r\n# Limit display to precision of 3\r\n\r\nnp.set_printoptions(precision=3)", "_____no_output_____" ], [ "# 1.3\r\nfrom IPython.core.interactiveshell import InteractiveShell\r\nInteractiveShell.ast_node_interactivity = \"all\"", "_____no_output_____" ], [ "# 2.0 Read from_tensor_slices as to_tensor_slices()\r\n# 2.0.1 from_tensor_slices()\r\ndataset1 = tf.data.Dataset.from_tensor_slices([[1,2],[3,4],[5,6]])\r\n# 2.0.2 from_tensors\r\ndataset2 = tf.data.Dataset.from_tensors([[1,2],[3,4],[5,6]])", "_____no_output_____" ], [ "# 2.1 Print the two objects\r\ndataset1\r\nprint()\r\ndataset2", "_____no_output_____" ], [ "# 2.2 Extract each element from\r\n# dataset. dataset is iterable.\r\n# Four elements. Each has two values\r\n\r\nfor elem in dataset1:\r\n print(\"--\")\r\n print(elem.numpy())", "--\n[1 2]\n--\n[3 4]\n--\n[5 6]\n" ], [ "# 2.2.1 Just one element\r\nfor elem in dataset2:\r\n print(\"--\")\r\n print(elem)", "_____no_output_____" ], [ "# 2.3\r\ndataset3 = tf.data.Dataset.from_tensor_slices(\r\n (\r\n tf.random.uniform([4]), # shape [4]\r\n tf.random.normal(\r\n [4, 6],\r\n mean=1.5,\r\n stddev=2.2\r\n ) \r\n )\r\n )", "_____no_output_____" ], [ "# 2.3.1 Print contents\r\nfor elem in dataset3:\r\n # Four elements.\r\n # Each element is a tuple of two tensors\r\n # one from uniform and the other from normal dist\r\n print(elem) \r\n", "(<tf.Tensor: shape=(), dtype=float32, numpy=0.2308768>, <tf.Tensor: shape=(6,), dtype=float32, numpy=array([-1.476, 7.276, 0.228, 3.193, 2.923, 0.521], dtype=float32)>)\n(<tf.Tensor: shape=(), dtype=float32, numpy=0.8270308>, <tf.Tensor: shape=(6,), dtype=float32, numpy=array([ 2.976, -1.021, 1.489, -1.378, 5.806, -1.429], dtype=float32)>)\n(<tf.Tensor: shape=(), dtype=float32, numpy=0.14125514>, <tf.Tensor: shape=(6,), dtype=float32, numpy=array([0.491, 4.323, 2.083, 3.948, 2.935, 2.654], dtype=float32)>)\n(<tf.Tensor: shape=(), dtype=float32, numpy=0.57362044>, <tf.Tensor: shape=(6,), dtype=float32, numpy=array([ 1.875, 1.325, 1.823, -1.774, 5.792, 3.923], dtype=float32)>)\n" ], [ "# 2.4 Next same internal contents but using from_tensors\r\ndataset4 = tf.data.Dataset.from_tensors(\r\n (\r\n tf.random.uniform([4]), # shape [4]\r\n tf.random.normal([4, 6], mean=1.5,stddev=2.2) # [4,100]\r\n )\r\n )", "_____no_output_____" ], [ "# 2.4.1\r\nfor elem in dataset4:\r\n # One element\r\n # This elem is a tuple of two tensors\r\n print()\r\n print(elem) ", "\n(<tf.Tensor: shape=(4,), dtype=float32, numpy=array([0.52 , 0.01 , 0.11 , 0.324], dtype=float32)>, <tf.Tensor: shape=(4, 6), dtype=float32, numpy=\narray([[ 2.189, 1.342, 2.591, -0.121, 0.247, 2.917],\n [ 2.694, 2.623, 1.488, 5.46 , -0.065, 1.36 ],\n [ 0.13 , 2.027, -3.341, 3.226, 4.435, 2.126],\n [ 0.062, -0.098, 0.346, -0.038, 2.953, 2.251]], dtype=float32)>)\n" ], [ "# Two tensors can be combined into one Dataset object.\r\nfeatures = tf.constant([[1, 3], [2, 1], [3, 3]]) # ==> 3x2 tensor\r\nlabels = tf.constant(['A', 'B', 'A']) # ==> 3x1 tensor\r\ndataset = tf.data.Dataset.from_tensor_slices((features, labels))\r\n", "_____no_output_____" ], [ "# Both the features and the labels tensors can be converted\r\n# to a Dataset object separately and combined after.\r\nfeatures_dataset = tf.data.Dataset.from_tensor_slices(features)\r\nlabels_dataset = tf.data.Dataset.from_tensor_slices(labels)\r\ndataset = tf.data.Dataset.zip((features_dataset, labels_dataset))\r\n", "_____no_output_____" ], [ "#help(tfds.load)", "_____no_output_____" ] ], [ [ "## Dataset objects from numpy-arrays in memory", "_____no_output_____" ], [ "If all data is in memory, to create a `Dataset` object, use `Dataset.from_tensor_slices()` on numpy arrays (in memory).", "_____no_output_____" ] ], [ [ "# 3.0 Download numpy arrays to \r\n# two objects (train,test) in memory\r\n\r\n# 3.0.1 First delete any existing folder\r\n\r\n!rm -rf /root/.keras/datasets\r\n!ls -la /root/.keras", "total 16\ndrwxr-xr-x 2 root root 4096 Mar 11 05:52 .\ndrwx------ 1 root root 4096 Mar 11 05:52 ..\n-rw-r--r-- 1 root root 123 Mar 11 05:52 keras.json\n" ], [ "# 3.0.2 Next download data:\r\ntrain, test = tf.keras.datasets.fashion_mnist.load_data()", "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-labels-idx1-ubyte.gz\n32768/29515 [=================================] - 0s 0us/step\nDownloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-images-idx3-ubyte.gz\n26427392/26421880 [==============================] - 1s 0us/step\nDownloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-labels-idx1-ubyte.gz\n8192/5148 [===============================================] - 0s 0us/step\nDownloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-images-idx3-ubyte.gz\n4423680/4422102 [==============================] - 0s 0us/step\n" ], [ "# 3.0.3 Check downloaded files\r\n! ls -la /root/.keras/datasets", "total 12\ndrwxr-xr-x 3 root root 4096 Mar 11 06:07 .\ndrwxr-xr-x 3 root root 4096 Mar 11 06:07 ..\ndrwxr-xr-x 2 root root 4096 Mar 11 06:07 fashion-mnist\n" ], [ "# 4.1 Little more checking:\r\n\r\ntype(train)\r\nprint()\r\ntype(train[0])\r\nprint()\r\ntype(test[0])", "_____no_output_____" ], [ "# 4.2 Separate images and labels:\r\nimages,labels = train\r\n\r\n# 4.2.1 Do something with images\r\nimages = images/255", "_____no_output_____" ], [ "# 4.3 Create Dataset object. \r\n# It is similar to way we did in #3.3 above:\r\n\r\ndataset = tf.data.Dataset.from_tensor_slices((images, labels))\r\ntype(dataset)\r\ndataset", "_____no_output_____" ] ], [ [ "## Dataset objects from textfiles on disk\r\nPlease see this [tutorial](https://www.tensorflow.org/tutorials/load_data/text). We are downloading files from stackoverflow. The file contains questions asked on some subjects. (This section is to be completed.)", "_____no_output_____" ], [ "### Consuming CSV file", "_____no_output_____" ], [ "See [Loading CSV Files](../tutorials/load_data/csv.ipynb), and [Loading Pandas DataFrames](../tutorials/load_data/pandas.ipynb) for more examples. The CSV file format is a popular format for storing tabular data in plain text. For example: \r\n", "_____no_output_____" ] ], [ [ "# 5.0 Download titanic file 'train.csv'\r\n# from a URL:\r\n\r\nimport pandas as pd\r\ntitanic_file = tf.keras.utils.get_file(\r\n \"train.csv\",\r\n \"https://storage.googleapis.com/tf-datasets/titanic/train.csv\"\r\n )", "Downloading data from https://storage.googleapis.com/tf-datasets/titanic/train.csv\n32768/30874 [===============================] - 0s 0us/step\n" ], [ "# 5.1 Read downloaded file:\r\n\r\ndf = pd.read_csv(titanic_file)\r\ndf.head()", "_____no_output_____" ], [ "# 5.2 Transform pandas dataframe to dictionary\r\ndict(df)", "_____no_output_____" ], [ "# 6.0 Create tf.data.Dataset objects:\r\n# tensor_slice: Slices along along first dimension. \r\n# This operation preserves the structure of the input\r\n# tensors, removing the first dimension of each tensor\r\n# and using it as the dataset dimension. \r\n\r\ntitanic_slices = tf.data.Dataset.from_tensor_slices(dict(df))\r\ntype(titanic_slices) # TensorSliceDataset", "_____no_output_____" ], [ "# 6.1 Just extract one row/batch\r\nn = 1\r\nfor feature_batch in titanic_slices.take(n):\r\n for key, value in feature_batch.items():\r\n print(\" {}:{}\".format(key, value))", " survived:0\n sex:b'male'\n age:22.0\n n_siblings_spouses:1\n parch:0\n fare:7.25\n class:b'Third'\n deck:b'unknown'\n embark_town:b'Southampton'\n alone:b'n'\n" ] ], [ [ "A more scalable approach is to load from disk as necessary. The tf.data module provides methods to extract records from one or more CSV files fron disk.\r\n\r\nThe `experimental.make_csv_dataset` function is the high level interface for reading sets of csv files. It supports column type inference and many other features, like batching and shuffling, to make usage simple.", "_____no_output_____" ] ], [ [ "# 6.2\r\nbatch_size = 4\r\ntitanic_batches = tf.data.experimental.make_csv_dataset(\r\n titanic_file,\r\n batch_size=batch_size,\r\n label_name=\"survived\"\r\n )", "_____no_output_____" ], [ "# 6.3 Get just one batch\r\n# Batch size is 4\r\n\r\nhowManyBatches = 1\r\nfor feature_batch, label_batch in titanic_batches.take(howManyBatches):\r\n print(\"'survived': {}\".format(label_batch))\r\n print(\"features:\")\r\n for key, value in feature_batch.items():\r\n print(\" {!r:20s}: {}\".format(key, value))", "'survived': [0 0 1 0]\nfeatures:\n 'sex' : [b'male' b'female' b'female' b'male']\n 'age' : [45. 22. 38. 28.]\n 'n_siblings_spouses': [1 0 0 0]\n 'parch' : [0 0 0 0]\n 'fare' : [ 83.475 10.517 227.525 13. ]\n 'class' : [b'First' b'Third' b'First' b'Second']\n 'deck' : [b'C' b'unknown' b'C' b'unknown']\n 'embark_town' : [b'Southampton' b'Southampton' b'Cherbourg' b'Southampton']\n 'alone' : [b'n' b'y' b'y' b'y']\n" ], [ "# 6.4\r\ntitanic_batches = tf.data.experimental.make_csv_dataset(\r\n titanic_file, batch_size=4,\r\n label_name=\"survived\", select_columns=['class', 'fare', 'survived'])", "_____no_output_____" ], [ "# 6.5\r\nfor feature_batch, label_batch in titanic_batches.take(1):\r\n print(\"'survived': {}\".format(label_batch))\r\n for key, value in feature_batch.items():\r\n print(\" {!r:20s}: {}\".format(key, value))", "_____no_output_____" ] ], [ [ "You can use the select_columns argument if you only need a subset of columns.", "_____no_output_____" ] ], [ [ "# 6.6\r\ntitanic_batches = tf.data.experimental.make_csv_dataset(\r\n titanic_file, batch_size=4,\r\n label_name=\"survived\", select_columns=['class', 'fare', 'survived'])", "_____no_output_____" ], [ "# 6.7\r\nfor feature_batch, label_batch in titanic_batches.take(1):\r\n print(\"'survived': {}\".format(label_batch))\r\n for key, value in feature_batch.items():\r\n print(\" {!r:20s}: {}\".format(key, value))", "'survived': [1 0 0 1]\n 'fare' : [12.288 8.654 52. 7.142]\n 'class' : [b'Third' b'Third' b'First' b'Third']\n" ] ], [ [ "### Consuming text files", "_____no_output_____" ], [ "#### Download file from a url\r\nHow to download a file from a URL.\r\nThis way any file can also be downloaded from *gdrive* without mounting it. Uses [`get_file()`](https://www.tensorflow.org/api_docs/python/tf/keras/utils/get_file) utility of `tf.keras.utils`. See this [example](https://www.tensorflow.org/tutorials/load_data/text).", "_____no_output_____" ] ], [ [ "# 7.0 Which files are where:\r\n\r\n!ls -la /root/.keras/\r\n!ls -la /root/.keras/datasets\r\n\r\n# 7.0.1 Deleting datasets dir\r\n!rm -rf /root/.keras/datasets\r\n", "_____no_output_____" ], [ "# 7.1 Specify file URL from stackoverflow.\r\n# Our file: stack_overflow_16k.tar.gz\r\n\r\ndata_url = 'https://storage.googleapis.com/download.tensorflow.org/data/stack_overflow_16k.tar.gz'\r\n\r\n# 7.1.1 Use keras get_file() utility\r\ndnld = utils.get_file(\r\n 'stack_overflow_16k.tar.gz', \r\n data_url,\r\n untar=True,\r\n )\r\n\r\n# 7.1.2 Where is 'dnld' object?\r\n# Get its path\r\ndnld_dir = pathlib.Path(dnld).parent\r\ndnld_dir # '/root/.keras/datasets'", "Downloading data from https://storage.googleapis.com/download.tensorflow.org/data/stack_overflow_16k.tar.gz\n6053888/6053168 [==============================] - 0s 0us/step\n" ], [ "# 7.1.3\r\n! ls -la /root/.keras/datasets\r\n\r\n# 5.1.4 Result after untarring\r\n! ls -la /root/.keras/datasets/train\r\n! ls -la /root/.keras/datasets/test", "_____no_output_____" ] ], [ [ "#### Create a dataset of text-lines", "_____no_output_____" ] ], [ [ "# 8.0 Files in 'train/python' folder\r\ntrain_dir = \"/root/.keras/datasets/train/\"\r\npython_files = os.listdir(train_dir + \"python\")\r\npython_files[:3]\r\nprint()\r\n\r\n# 8.0.1 Join directory path to file-names\r\nfiles_with_path = [os.path.join(train_dir, file) for file in python_files]\r\nfiles_with_path[:4]", "_____no_output_____" ], [ "# 8.1 tf.data.Dataset of just filepaths\r\n# Prepare a dataset of all files matching\r\n# one or more glob patterns.\r\n\r\nfilepath_dataset_py = tf.data.Dataset.list_files(\"/root/.keras/datasets/train/python/*.txt\")", "_____no_output_____" ], [ "# 8.2 Create a datset comprising lines from one or more text files.\r\n\r\ndata_lines = tf.data.TextLineDataset(filepath_dataset_py)", "_____no_output_____" ], [ "# 8.2.1 Examine 5 files:\r\n\r\nfor line in data_lines.take(5):\r\n print(line.numpy)", "<bound method _EagerTensorBase.numpy of <tf.Tensor: shape=(), dtype=string, numpy=b'\"blank submodule imports using __init__.py i\\'m learning blank, and i can\\'t figure out how imports in __init__.py work...i understand from the blank tutorial that the __init__.py file initializes a package, and that i can import subpackages here...i\\'m doing something wrong, though. could you explain for me (and for future blank-learners) what i\\'m doing wrong?..here\\'s a simplified example of what i\\'m trying to do...this is my file structure:..package. __init__.py. test.py. subpackage. __init__.py. hello_world.py...the contents of hello_world.py:..def do_something():. print \"\"hello, world!\"\"...subpackage/__init__.py is empty...package/__init__.py contains:..import test.submodule.do_something...and finally, test.py contains:..do_something()...this is how i attempt to run hello_world.py using osx terminal and blank 3:..blank test.py...blank then throws the following error:..nameerror: name \\'do_something\\' is not defined\"'>>\n<bound method _EagerTensorBase.numpy of <tf.Tensor: shape=(), dtype=string, numpy=b'\"blank - is there a way to get all the attributes of an object so i have a class called \\'car\\' in another .py file (myclasses.py)..class car:. tires = 4. color = \"\"\"\". speed = 0. engine = \"\"\"\"...and in my main..import myclasses..ferrari = myclasses.car().ferrari.color = \"\"red\"\".ferrari.engine = \"\"fast\"\".ferrari.speed = 200..for prpty in ferrari:. print(prpty)...the output i want is like this:..tires = 4.color = red.speed = 200.engine = fast...i know the format of the output depends on me. only if in my for-loop i can get the property name and its value base on my object.\"'>>\n<bound method _EagerTensorBase.numpy of <tf.Tensor: shape=(), dtype=string, numpy=b'\"sequence switch in blank i am solving a problem in blank. ..the general idea is that given a sequence of numbers, then pick out the numbers form this sequence according to a rule, then this pick-out numbers form a new list...step1, pick out the numbers in odd position (the position count from zero) and append to a new list. ..step2, the remaining numbers form a new sequence...step3, in the next round, pick out the numbers in even position of the new sequence, and append to a new list...step4, after this round, the remaining numbers forms a a new sequences. ..repeat this procedure until all the number in the original sequence have been pick out to for a new list. ..for example, given a sequence [0, 1, 2, 3, 4, 5, 6, 7, 8], ..step1, pick out the numbers in odd position, new_list = [1, 3, 5, 7] ..step2, the remaining numbers form a new sequence is [0, 2, 4, 6, 8]..step3, pick out the numbers in even position of the new sequence, then the numbers [0, 4, 8] are picked out from the new sequence, and append to new list. then new_list = [1, 3, 5, 7, 0, 4, 8] ..step4, the remaining numbers form a new sequence is [2, 6]..repeat step 1, then 6 is picked out append to new_list = [1, 3, 5, 7, 0, 4, 8, 6] .repeat step 2, then 2 is picked out append to new_list = [1, 3, 5, 7, 0, 4, 8, 6, 2], then all numbers in original list are picked out. and the final result is the new_list...i do not know how to implement it in blank. i got stuck in the first step, i want to use a for loop over the sequence and pick out the numbers according to odd or even index. but after picking out the numbers, the sequence changes. then i have to start over a new for loop to loop over this new sequence, if the original sequence is very long. i will be too long and inefficient coding. could someone gives me idea how to deal with this problem?\"'>>\n<bound method _EagerTensorBase.numpy of <tf.Tensor: shape=(), dtype=string, numpy=b'\"input variables in one line i have a problem with reading variables in one line. i used a map(int, input().splite()) when i know the number of variables, but in this case in every run of code this number is different. (blank 3)\"'>>\n<bound method _EagerTensorBase.numpy of <tf.Tensor: shape=(), dtype=string, numpy=b'\"print line in specific order i want to print around 10 lines in such a way that in every sequence 9 lines print with sign \\'-\\'. for example, if x=[1,2,3,4,5]. now i want to print in following sequence,.. 1.-2.-3.-4.-5..-1.2.-3.-4.-5..-1.-2.3.-4.-5...etc. ...please help me to solve this.\"'>>\n" ] ], [ [ "#### Downloading data from gdrive", "_____no_output_____" ] ], [ [ "# 8.3 gdrive download: May also download a shared file from gdrive\r\n# No need to mount gdrive\r\n# Download 'archive.csv.zip' from my gdrive:\r\n\r\ngdrive_url = \"https://drive.google.com/file/d/14gjcWMRJORJ4bQ2QerWaodDpLi5oDcxM/view?usp=sharing\"\r\n\r\n# 8.3.1 Use keras get_file() utility\r\ndnld_gdrive = utils.get_file(\r\n 'archive.csv.zip', \r\n gdrive_url,\r\n )\r\n\r\n# 8.3.2 Where is 'dataset' object?\r\n# Get its path\r\ngdrive_dir = pathlib.Path(dnld_gdrive).parent\r\ngdrive_dir # PosixPath('/root/.keras/datasets')", "_____no_output_____" ], [ "# 8.3.3 Check folders under 'train'\r\n! ls -la /root/.keras/datasets/train", "_____no_output_____" ] ], [ [ "## preprocessing data\r\nThe Keras preprocessing layers API allows developers to build Keras-native input processing pipelines. These input processing pipelines can be used as independent preprocessing code in non-Keras workflows, combined directly with Keras models, and exported as part of a Keras SavedModel. \r\n\r\nWith Keras preprocessing layers, you can build and export models that are truly end-to-end: models that accept raw images or raw structured data as input; models that handle feature normalization or feature value indexing on their own.\r\n\r\nSee this [link](https://www.tensorflow.org/guide/keras/preprocessing_layers) for usage of preprocessing library methods. Pay attention to the panel on the right. \r\n\r\n", "_____no_output_____" ], [ "### preprocessing.text_dataset_from_directory\r\nRefer this [link](https://www.tensorflow.org/api_docs/python/tf/keras/preprocessing/text_dataset_from_directory). \r\n\r\nThe `preprocessing.text_dataset_from_directory()` expects a directory structure as follows. See an example [here](https://www.tensorflow.org/tutorials/keras/text_classification).\r\n\r\nmain_directory/<br>\r\n...class_a/<br>\r\n......a_text_1.txt<br>\r\n......a_text_2.txt<br>\r\n...class_b/<br>\r\n......b_text_1.txt<br>\r\n......b_text_2.txt<br>\r\n\r\n\r\nWhen running a machine learning experiment, it is a best practice to divide your dataset into three splits: train, validation, and test. The Stack Overflow dataset has already been divided into train and test, but it lacks a validation set. Create a validation set using an 80:20 split of the training data by using the validation_split argument below.\r\n\r\n\r\n\r\n\r\n", "_____no_output_____" ] ], [ [ "# 9.0 Our stackoverflow training data is here:\r\ntrain_dir = \"/root/.keras/datasets/train\"\r\n\r\n# 9.0.1 Some constants\r\nbatch_size = 32\r\nseed = 42\r\n\r\n# 9.1 Create a training dataset object\r\nraw_train_ds = preprocessing.text_dataset_from_directory(\r\n train_dir,\r\n labels = \"inferred\",\r\n batch_size=batch_size,\r\n # fraction of data to \r\n # reserve for validation. \r\n validation_split=0.2, \r\n # Return back 'train' subset\r\n # but in batches\r\n subset='training',\r\n seed =2 \r\n )\r\n", "Found 8000 files belonging to 4 classes.\nUsing 6400 files for training.\n" ], [ "# 9.1.1\r\ntype(raw_train_ds) # BatchDataset", "_____no_output_____" ], [ "# 9.2\r\n# Dataset.take(n) object is iterable.\r\n# Iterate over the dataset and print out a few examples,\r\n# to get a feel for the data.\r\n# take(n): Reads n batches\r\n\r\nfor text_batch, label_batch in raw_train_ds.take(1):\r\n print(text_batch.numpy().shape) # 32: batch_size\r\n # Change batch_size to see how it changes\r\n for i in range(10):\r\n print(\"Question: \", text_batch.numpy()[i])\r\n print(\"Label:\", label_batch.numpy()[i])\r\n", "_____no_output_____" ], [ "# 9.3 Get class_names in the dataset:\r\nfor i, label in enumerate(raw_train_ds.class_names):\r\n print(\"Label\", i, \"corresponds to\", label)\r\n", "Label 0 corresponds to csharp\nLabel 1 corresponds to java\nLabel 2 corresponds to javascript\nLabel 3 corresponds to python\n" ], [ "# 10.0 From train_dir pick-up remaining validation\r\n# dataset but in batches:\r\n\r\nraw_validation_ds = preprocessing.text_dataset_from_directory(\r\n train_dir,\r\n batch_size=batch_size,\r\n validation_split=0.2,\r\n # Return validation data\r\n # but in batches\r\n subset='validation',\r\n seed = 2\r\n )\r\n", "Found 8000 files belonging to 4 classes.\nUsing 1600 files for validation.\n" ], [ "# 11.0 Test dataset\r\n\r\ntest_dir = \"/root/.keras/datasets/test\"\r\n\r\n# 11.1\r\nraw_test_ds = preprocessing.text_dataset_from_directory(\r\n test_dir,\r\n batch_size=batch_size\r\n )\r\n", "Found 8000 files belonging to 4 classes.\n" ] ], [ [ "### Text Vectorization\r\nOnce our train, test and validation datasets are ready, we proceed to feed textvectorization layer.\r\n", "_____no_output_____" ], [ "#### Using `preprocessing.TextVectorization` layer. Its [full syntax](https://www.tensorflow.org/api_docs/python/tf/keras/layers/experimental/preprocessing/TextVectorization) is:\r\n\r\n`tf.keras.layers.experimental.preprocessing.TextVectorization(\r\n max_tokens=None, standardize=LOWER_AND_STRIP_PUNCTUATION,\r\n split=SPLIT_ON_WHITESPACE, ngrams=None, output_mode=INT,\r\n output_sequence_length=None, pad_to_max_tokens=True, vocabulary=None, **kwargs\r\n)\r\n`", "_____no_output_____" ], [ "`TextVectorization` layer will standardize, tokenize, and vectorize the data using the preprocessing.TextVectorization layer.\r\n\r\n> Standardization refers to preprocessing the text, typically to remove punctuation or HTML elements to simplify the dataset.\r\n\r\n> Tokenization refers to splitting strings into tokens (for example, splitting a sentence into individual words by splitting on whitespace).\r\n\r\n> Vectorization refers to converting tokens into numbers so they can be fed into a neural network.\r\n\r\nAll of these tasks can be accomplished with this layer. You can learn more about each of these in the API doc.\r\n\r\n> The default standardization converts text to lowercase and removes punctuation.\r\n\r\n> The default tokenizer splits on whitespace.\r\n\r\n> The default vectorization mode is int. This outputs integer indices (one per token). This mode can be used to build models that take word order into account. You can also use other modes, like binary, to build bag-of-word models.\r\n", "_____no_output_____" ] ], [ [ "from tensorflow.keras.layers.experimental import preprocessing ", "_____no_output_____" ] ], [ [ "##### Simple **experiment**", "_____no_output_____" ] ], [ [ "data = [\r\n \"The title of Rachel Levin’s book, Look Big, is\"\r\n \"just about the best two words of advice one can\"\r\n \" give about how to survive most animal encounters.\"\r\n \"In her illustrated service manual, Levin breaks down\"\r\n \"how to handle 50 different kinds of animals common in \"\r\n \"North America, based on expert advice. Let’s look at her\"\r\n \" tips for dealing with five of these creatures and see\"\r\n \" how they stack up with what the experts say—and with\"\r\n \" real-world experience. \"\r\n ]\r\nlayer = preprocessing.TextVectorization()\r\nlayer.adapt(data)\r\nvectorized_text = layer(data)\r\nprint(vectorized_text)\r\n", "tf.Tensor(\n[[ 4 15 2 25 31 53 6 54 35 10 4 55 14 11 2 9 26 51 39 10 7 5 19 29\n 59 45 8 37 21 30 32 52 46 5 38 62 47 34 2 58 50 36 28 61 56 27 43 9\n 33 6 57 8 16 40 48 3 41 2 18 49 60 22 7 17 20 13 3 12 4 42 23 3\n 24 44]], shape=(1, 74), dtype=int64)\n" ], [ "data = [ [\r\n \"The title of Rachel Levin’s book, Look Big, is\"\r\n \"just about the best two words of advice one can\"\r\n \" give about how to survive most animal encounters.\"\r\n \"In her illustrated service manual, Levin breaks down\"\r\n ],\r\n [\"how to handle 50 different kinds of animals common in \"\r\n \"North America, based on expert advice. Let’s look at her\"\r\n \" tips for dealing with five of these creatures and see\"\r\n \" how they stack up with what the experts say—and with\"\r\n \" real-world experience. \"\r\n ]\r\n ]\r\nlayer = preprocessing.TextVectorization()\r\nlayer.adapt(data)\r\nvectorized_text = layer(data)\r\nprint(vectorized_text)\r\n", "tf.Tensor(\n[[ 4 15 2 25 31 53 7 54 35 10 4 55 14 11 2 9 26 51 39 10 5 6 19 29\n 59 45 8 37 21 30 32 52 46 0 0 0 0 0 0 0 0 0]\n [ 5 6 38 62 47 34 2 58 50 36 28 61 56 27 43 9 33 7 57 8 16 40 48 3\n 41 2 18 49 60 22 5 17 20 13 3 12 4 42 23 3 24 44]], shape=(2, 42), dtype=int64)\n" ], [ "# 12.0\r\nVOCAB_SIZE = 10000\r\nMAX_SEQUENCE_LENGTH = 250\r\n\r\n# 12.1 Instantiate TextVectorization class\r\nint_vectorize_layer = TextVectorization(\r\n max_tokens=VOCAB_SIZE,\r\n output_mode='int',\r\n output_sequence_length=MAX_SEQUENCE_LENGTH\r\n )\r\n", "_____no_output_____" ], [ "#binary_vectorize_layer = TextVectorization(\r\n# max_tokens=VOCAB_SIZE,\r\n# output_mode='binary'\r\n# )", "_____no_output_____" ], [ "# 12.2 Make a text-only dataset (without labels), then call adapt\r\n\r\ntrain_text = raw_train_ds.map(lambda text, labels: text)\r\n", "_____no_output_____" ], [ "# 12.3 'adapt' or 'fit' over 'int_vectorize_layer'\r\n# object. 'adpat' fits the state of the preprocessing\r\n# layer to the dataset.\r\n\r\nint_vectorize_layer.adapt(\r\n # The data to train on. \r\n # It can be passed either as\r\n # a tf.data Dataset, as a NumPy array\r\n train_text # \r\n )\r\n#binary_vectorize_layer.adapt(train_text)", "_____no_output_____" ], [ "# 12.3.1\r\nint_vectorize_layer.get_vocabulary()[:4]", "_____no_output_____" ], [ "# 12.4 Retrieve a batch (of 32 reviews and labels) from the dataset\r\n\r\ntext_batch, label_batch = next(iter(raw_train_ds))\r\nfirst_question, first_label = text_batch[0], label_batch[0]\r\nprint(\"Question\", first_question)\r\nprint(\"Label\", first_label)", "Question tf.Tensor(b'\"error loading report i have a report named \"\"relatorio1.rdlc\"\", in windows forms with report viewer...the code have no errors when built in vs studio ..sqlconnection conec1 = tconex.getconnection();..private void verrelatorio_load(object sender, eventargs e).{. sqlcommand cmdreport = new sqlcommand();. sqldatareader drreport;. dataset dsreport = new dsbairros();.. try. {. cmdreport.commandtype = commandtype.text;. cmdreport.connection = conec1;. cmdreport.commandtext = \"\"select codigo,descricao from tblbairros \"\";. drreport = cmdreport.executereader();. dsreport.tables[0].load(drreport);. drreport.close();.. reportviewer1.localreport.reportembeddedresource = \"\"contatroco.relatorio1.rdlc\"\";.. reportdatasource rds = new reportdatasource();.. rds.name = \"\"dsbairros\"\";. rds.value = dsreport.tables[0];. reportviewer1.localreport.datasources.add(rds);. reportviewer1.refreshreport();. }. catch (exception ex). {. messagebox.show(ex.message);. }. finally. {. if (conec1.state == connectionstate.open). { conec1.close(); }. }.}...the reportviewer shows me the name of the fields codigo, descri\\xc3\\xa7ao but does not show me any data. ..dsbairros is a typed dataset. i need to make it this way because i changed the origin connection of the datasource to don\\'t update and now the report does not work.\"\\n', shape=(), dtype=string)\nLabel tf.Tensor(0, shape=(), dtype=int32)\n" ], [ "# 12.5.1 Expand \r\ntype(first_question)\r\nprint(\"\\n--------------\\n\")\r\n# 12.5.2\r\nfirst_question.get_shape()\r\nprint(\"\\n--------------\\n\")\r\n# 12.5.3\r\nt = tf.expand_dims(first_question,-1)\r\nt.get_shape()\r\nprint(\"\\n--------------\\n\")\r\n# 12.5.4\r\nprint(t)\r\nprint(\"\\n--------------\\n\")\r\n", "_____no_output_____" ], [ "12.6\r\nint_vectorize_layer(t)", "_____no_output_____" ], [ "def int_vectorize_text(text, label):\r\n text = tf.expand_dims(text, -1)\r\n return int_vectorize_layer(text) , label", "_____no_output_____" ], [ "# Retrieve a batch (of 32 reviews and labels) \r\n# from the dataset\r\ntext_batch, label_batch = next(iter(raw_train_ds))\r\nfirst_question, first_label = text_batch[0], label_batch[0]\r\nprint(\"Question\", first_question)\r\nprint(\"Label\", first_label)\r\n", "Question tf.Tensor(b'\"simple program not giving desired output i am starting to learn blank and i have a simple program to request two numbers from the user and add them together. when i run my program , i get asked for the first number , i type it in and that\\'s it, the program doesn\\'t go any further. i am running it in visual studio..can anyone tell me what i\\'m doing wrong? ..using system;..namespace coding.{. class program. {. static void main(string[] args). {.. int number1;. int number2;. int result;... console.writeline(\"\"enter first number to be calculated\"\");. number1 = convert.toint32(console.readline());... console.writeline(\"\"enter second number to be calculated\"\");. number2 = convert.toint32(console.readline());.. result = number1 + number2;. console.writeline(\"\"the total is\"\" + \"\" \"\" + result);. }. }.}...result screenshot\"\\n', shape=(), dtype=string)\nLabel tf.Tensor(0, shape=(), dtype=int32)\n" ], [ "print(\"'int' vectorized question:\",\r\n int_vectorize_text(first_question, first_label)[0])", "_____no_output_____" ], [ "print(\"'int' vectorized question:\",\r\n int_vectorize_text(first_question, first_label)[0])\r\n", "_____no_output_____" ], [ "######## I am done ###############", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code", "code", "code" ], [ "markdown", "markdown", "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code", "code", "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Zebrafish Double coiling model\n### Below is the block of code to run double coiling models", "_____no_output_____" ], [ "INITIALIZE - Run before running the models to set the parameters", "_____no_output_____" ] ], [ [ "%matplotlib inline\n\n### FIRST, run this code to set the simulation parameters. Then select from the codes below to run different models\n\n#Declare the duration of the simulation\nTIME_END = 100000\nSKIP = 200\nPLOT_INTERVAL = 10000\nPLOT_RESULT = True\nPRINT_PARAM = False\nSAVE_CSV = True\nSAVE_ANIM = True\n\nimport winsound\nduration = 1000 # milliseconds\nfreq = 440 # Hz", "_____no_output_____" ] ], [ [ "Base Model", "_____no_output_____" ] ], [ [ "\nfrom Double_coiling_model import Double_coil_base\n\nprint(\"Base Model\")\ntest = Double_coil_base(dt = 0.1, stim0 = 35, sigma = 0, \n E_glu = 0, E_gly = -58, cv = 1, nIC = 5, nMN = 10, nV0d = 10, nV0v=10, nV2a=10, nMuscle = 10)\n#test.setWeightParameters(V0d_IC_syn_weight = 2.0, V0d_MN_syn_weight = 2.0, V0d_V2a_syn_weight = 2.0)\ntest.setWeightParameters(V0d_IC_syn_weight = 8.0, V0d_MN_syn_weight = 8.0, V0d_V2a_syn_weight = 8.0)\n\n((VLIC, VRIC), (VLMN, VRMN), (VLV0d, VRV0d), (VLV0v, VRV0v), (VLV2a, VRV2a), \n (VLMuscle, VRMuscle), Time) = test.mainLoop(rand=0, tmax = TIME_END, tskip = SKIP, tplot_interval = PLOT_INTERVAL, plotResult = PLOT_RESULT, printParam = PRINT_PARAM, saveCSV = SAVE_CSV, saveAnim = SAVE_ANIM)\n \n#The code below produces a beeping sound af the code has been executed\nwinsound.Beep(freq, duration) \n", "_____no_output_____" ] ], [ [ "Glycine Null", "_____no_output_____" ] ], [ [ "from Double_coiling_glycine_null import Double_coil_glycine_null\n\nprint(\"Glycine Null Model\")\ntest = Double_coil_glycine_null(dt = 0.1, stim0 = 35, sigma = 0, \n E_glu = 0, E_gly = -58, cv = 1, nIC = 5, nMN = 10, nV0d = 10, nV0v=10, nV2a=10, nMuscle = 10)\ntest.setWeightParameters(V0d_IC_syn_weight=0, V0d_MN_syn_weight=0, V0d_V2a_syn_weight=0)\n((VLIC, VRIC), (VLMN, VRMN), (VLV0d, VRV0d), (VLV0v, VRV0v), (VLV2a, VRV2a), \n (VLMuscle, VRMuscle), Time) = test.mainLoop(rand=0, tmax = TIME_END, tskip = SKIP, tplot_interval = PLOT_INTERVAL, plotResult = PLOT_RESULT, printParam = PRINT_PARAM, saveCSV = SAVE_CSV, saveAnim = SAVE_ANIM)\n\n#The code below produces a beeping sound af the code has been executed\nwinsound.Beep(freq, duration) ", "_____no_output_____" ] ], [ [ "V0v Overexcitation", "_____no_output_____" ] ], [ [ "\nfrom Double_coiling_with_Tonic_drive_to_IC_overexcitation_of_V0v import Double_coil_with_Tonic_drive_to_IC_overexcitation_of_V0v\n\nprint(\"IC Tonic Drive Model\")\ntest = Double_coil_with_Tonic_drive_to_IC_overexcitation_of_V0v(dt = 0.1, stim0 = 35, sigma = 0.1, \n E_glu = 0, E_gly = -58, cv = 1, nIC = 5, nMN = 10, nV0d = 10, nV0v=10, nV2a=10, nMuscle = 10)\ntest.setWeightParameters(V0d_IC_syn_weight = 8.0, V0d_MN_syn_weight = 8.0, V0d_V2a_syn_weight = 8.0)\n((VLIC, VRIC), (VLMN, VRMN), (VLV0d, VRV0d), (VLV0v, VRV0v), (VLV2a, VRV2a), \n (VLMuscle, VRMuscle), Time) = test.mainLoop(rand=0, tmax = TIME_END, tskip = SKIP, tplot_interval = PLOT_INTERVAL, plotResult = PLOT_RESULT, printParam = PRINT_PARAM, saveCSV = SAVE_CSV, saveAnim = SAVE_ANIM)\n\n#The code below produces a beeping sound af the code has been executed\nwinsound.Beep(freq, duration) \n", "_____no_output_____" ] ], [ [ "V2a KO", "_____no_output_____" ] ], [ [ "\nfrom Double_coiling_V2a_KO import Double_coil_V2a_KO\n\nprint(\"V2a Knockout Model\")\ntest = Double_coil_V2a_KO(dt = 0.1, stim0 = 35, sigma = 0, \n E_glu = 0, E_gly = -58, cv = 1, nIC = 5, nMN = 10, nV0d = 10, nV0v=10, nV2a=10, nMuscle = 10)\ntest.setWeightParameters(V0d_IC_syn_weight = 8, V0d_MN_syn_weight = 8, V0d_V2a_syn_weight = 8)\n((VLIC, VRIC), (VLMN, VRMN), (VLV0d, VRV0d), (VLV0v, VRV0v), (VLV2a, VRV2a), \n (VLMuscle, VRMuscle), Time) = test.mainLoop(rand=0, tmax = TIME_END, tskip = SKIP, tplot_interval = PLOT_INTERVAL, plotResult = PLOT_RESULT, printParam = PRINT_PARAM, saveCSV = SAVE_CSV, saveAnim = SAVE_ANIM)\n\n#The code below produces a beeping sound af the code has been executed\nwinsound.Beep(freq, duration) ", "_____no_output_____" ] ], [ [ "No excitatory Synapse", "_____no_output_____" ] ], [ [ "\nfrom Double_coiling_no_excitatory_syn import Double_coil_no_excitatory_syn\n\nprint(\"No Excitatory Synapse Model\")\ntest = Double_coil_no_excitatory_syn(dt = 0.1, stim0 = 35, sigma = 0, \n E_glu = 0, E_gly = -58, cv = 1, nIC = 5, nMN = 10, nV0d = 10, nV0v=10, nV2a=10, nMuscle = 10)\ntest.setWeightParameters(V0d_IC_syn_weight = 2, V0d_MN_syn_weight = 2, V0d_V2a_syn_weight = 2)\n((VLIC, VRIC), (VLMN, VRMN), (VLV0d, VRV0d), (VLV0v, VRV0v), (VLV2a, VRV2a), \n (VLMuscle, VRMuscle), Time) = test.mainLoop(rand=0, tmax = TIME_END, tskip = SKIP, tplot_interval = PLOT_INTERVAL, plotResult = PLOT_RESULT, printParam = PRINT_PARAM, saveCSV = SAVE_CSV, saveAnim = SAVE_ANIM)\n\n#The code below produces a beeping sound af the code has been executed\nwinsound.Beep(freq, duration) ", "_____no_output_____" ] ], [ [ "V0v to IC Null", "_____no_output_____" ] ], [ [ "\nfrom Double_coiling_V0v_to_IC_null import Double_coil_V0v_to_IC_null\n\nprint(\"V0v to IC null Model\")\ntest = Double_coil_V0v_to_IC_null(dt = 0.1, stim0 = 35, sigma = 0.001, \n E_glu = 0, E_gly = -58, cv = 1, nIC = 5, nMN = 10, nV0d = 10, nV0v=10, nV2a=10, nMuscle = 10)\ntest.setWeightParameters(V0d_IC_syn_weight = 2, V0d_MN_syn_weight = 2, V0d_V2a_syn_weight = 2)\n((VLIC, VRIC), (VLMN, VRMN), (VLV0d, VRV0d), (VLV0v, VRV0v), (VLV2a, VRV2a), \n (VLMuscle, VRMuscle), Time) = test.mainLoop(rand=0, tmax = TIME_END, tskip = SKIP, tplot_interval = PLOT_INTERVAL, plotResult = PLOT_RESULT, printParam = PRINT_PARAM, saveCSV = SAVE_CSV, saveAnim = SAVE_ANIM)\n\n#The code below produces a beeping sound af the code has been executed\nwinsound.Beep(freq, duration) ", "_____no_output_____" ] ], [ [ "30 Somites", "_____no_output_____" ] ], [ [ "\nfrom Double_coiling_30_somites import Double_coil_30_somites\n\nprint(\"30 somites Model\")\ntest = Double_coil_30_somites(dt = 0.1, stim0 = 50, sigma = 0, \n E_glu = 0, E_gly = -58, cv = 1, nIC = 5, nMN = 30, nV0d = 30, nV0v=30, nV2a=30, nMuscle = 30)\ntest.setWeightParameters(V0d_IC_syn_weight = 2, V0d_MN_syn_weight = 2, V0d_V2a_syn_weight = 2)\n((VLIC, VRIC), (VLMN, VRMN), (VLV0d, VRV0d), (VLV0v, VRV0v), (VLV2a, VRV2a), \n (VLMuscle, VRMuscle), Time) = test.mainLoop(rand=0, tmax = TIME_END, tskip = SKIP, tplot_interval = PLOT_INTERVAL, plotResult = PLOT_RESULT, printParam = PRINT_PARAM, saveCSV = SAVE_CSV, saveAnim = SAVE_ANIM)\n\n#The code below produces a beeping sound af the code has been executed\nwinsound.Beep(freq, duration) ", "30 somites Model\n" ] ], [ [ "Sigma", "_____no_output_____" ] ], [ [ "from Double_coiling_with_sigmas import Double_coil_with_sigmas\n\nprint(\"Sigma Model\")\ntest = Double_coil_with_sigmas(dt = 0.1, stim0 = 35, sigmaD=0.2, sigmaL = 0, sigmaP = 0, sigmaW=0, \n E_glu = 0, E_gly = -58, cv = 1, nIC = 5, nMN = 10, nV0d = 10, nV0v=10, nV2a=10, nMuscle = 10)\ntest.setWeightParameters(V0d_IC_syn_weight = 2, V0d_MN_syn_weight = 2, V0d_V2a_syn_weight = 2)\n((VLIC, VRIC), (VLMN, VRMN), (VLV0d, VRV0d), (VLV0v, VRV0v), (VLV2a, VRV2a), \n (VLMuscle, VRMuscle), Time) = test.mainLoop(rand=0, tmax = TIME_END, tskip = SKIP, tplot_interval = PLOT_INTERVAL, plotResult = PLOT_RESULT, printParam = PRINT_PARAM, saveCSV = SAVE_CSV, saveAnim = SAVE_ANIM)\n\nwinsound.Beep(freq, duration)", "_____no_output_____" ] ], [ [ "Chemical and Gap Sigmas", "_____no_output_____" ] ], [ [ "from Double_coiling_with_sigmas_chem_syn_and_gap import Double_coil_with_gap_chem_sigma\n\nprint(\"Chemical and Gap Sigmas\")\ntest = Double_coil_with_gap_chem_sigma(dt = 0.1, stim0 = 35, sigma_chem = 0, sigma_gap = 0, \n E_glu = 0, E_gly = -58, cv = 1, nIC = 5, nMN = 10, nV0d = 10, nV0v=10, nV2a=10, nMuscle = 10)\ntest.setWeightParameters(V0d_IC_syn_weight = 2, V0d_MN_syn_weight = 2, V0d_V2a_syn_weight = 2)\n((VLIC, VRIC), (VLMN, VRMN), (VLV0d, VRV0d), (VLV0v, VRV0v), (VLV2a, VRV2a), \n (VLMuscle, VRMuscle), Time) = test.mainLoop(rand=0, tmax = TIME_END, tskip = SKIP, tplot_interval = PLOT_INTERVAL, plotResult = PLOT_RESULT, printParam = PRINT_PARAM, saveCSV = SAVE_CSV, saveAnim = SAVE_ANIM)\n\n#The code below produces a beeping sound af the code has been executed\nwinsound.Beep(freq, duration) ", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Visualizations\n\n## How to visuzalize facial expression data with Feat.\n*Written by Jin Hyun Cheong*", "_____no_output_____" ] ] ]

[ "markdown" ]

[ [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "!pip install pandas-summary", "Collecting pandas-summary\n Downloading https://files.pythonhosted.org/packages/f5/20/865d4785f86e86f7ad7ebd52fc14810ef59ef13cd959f1363a2615d1c665/pandas_summary-0.0.7-py2.py3-none-any.whl\nRequirement already satisfied: numpy in /usr/local/lib/python3.6/dist-packages (from pandas-summary) (1.18.5)\nRequirement already satisfied: pandas in /usr/local/lib/python3.6/dist-packages (from pandas-summary) (1.1.2)\nRequirement already satisfied: python-dateutil>=2.7.3 in /usr/local/lib/python3.6/dist-packages (from pandas->pandas-summary) (2.8.1)\nRequirement already satisfied: pytz>=2017.2 in /usr/local/lib/python3.6/dist-packages (from pandas->pandas-summary) (2018.9)\nRequirement already satisfied: six>=1.5 in /usr/local/lib/python3.6/dist-packages (from python-dateutil>=2.7.3->pandas->pandas-summary) (1.15.0)\nInstalling collected packages: pandas-summary\nSuccessfully installed pandas-summary-0.0.7\n" ], [ "import pandas as pd\n# pip install pandas_summary\nfrom pandas_summary import DataFrameSummary", "_____no_output_____" ] ], [ [ "# Competencia de Kaggle\n\n[ir a Kaggle](https://www.kaggle.com/c/rossmann-store-sales/data)\n\n[3er puesto](https://github.com/entron/entity-embedding-rossmann)", "_____no_output_____" ], [ "# Métrica de la competencia", "_____no_output_____" ], [ "$$\n\\textrm{RMSPE} = \\sqrt{\\frac{1}{n} \\sum_{i=1}^{n} \\left(\\frac{\\hat{y}_i - y_i}{y_i}\\right)^2}\n$$\n\ndonde:\n\n- $y_i$ las ventas de un día particular de un store\n- $\\hat{y}_i$ ventas estimadas por el modelo\n- $n$ es el número de predicciones realizadas", "_____no_output_____" ] ], [ [ "from google.colab import drive\ndrive.mount('/content/drive')", "Mounted at /content/drive\n" ] ], [ [ "## Importamos dataset", "_____no_output_____" ], [ "La competencia permitía agregar datos externos para realizar la predicción", "_____no_output_____" ], [ "The following tables are available in the datasets:\n\n| Archivo | Descripción| Origen de Datos|\n|--------------|--------------------------------------------------------------------|--|\n| train.csv | training set: información del store día a día, ventas, clientes, si es feriado, etc | Kaggle |\n| store.csv | Información general del store, por ejemplo datos del competidor | Kaggle |\n| store_states.csv | Mapea de store a estado - Dato externo| Externos |\n| state_names.csv | Mapea estados a acronimo de estado | Externos |\n| googletrend.csv | Tendencias por semana - Dato externo| Externos|\n| weather.csv | Condiciones meteorológicas por día | Externos|", "_____no_output_____" ] ], [ [ "PATH = '/content/drive/My Drive/Colab Notebooks/kaggle-rossmann-master/rossmann/rossmann/'", "_____no_output_____" ], [ "table_names = ['train', 'store', 'store_states', 'state_names', 'googletrend', 'weather']\ntrain, store, store_states, state_names, googletrend, weather = [pd.read_csv(PATH + fname+'.csv', \n low_memory=False) for fname in table_names]", "_____no_output_____" ], [ "display(train.head())\ndisplay(DataFrameSummary(train).summary())", "_____no_output_____" ] ], [ [ "- Mirar counts que todos tienen la misma cantidad\n- Ninguno tiene missing\n- Los tipos tambien es interesante observar", "_____no_output_____" ] ], [ [ "train['StateHoliday'].value_counts()", "_____no_output_____" ], [ "display(store.head())\ndisplay(DataFrameSummary(store).summary())", "_____no_output_____" ] ], [ [ "Descripción de algunas columnas que quizas no sean tan claras:\n\n- `Customers`: La cantidad de clientes por día\n- `Open`: Indicador si el store estaba abierto o cerrado: 0 = closed, 1 = open\n- `StateHoliday`: Indica feriado en ese estado. a = public holiday, b = Easter holiday, c = Christmas, 0 = None\n- `SchoolHoliday`: Inidica si el store fue afectado por el feriado escolar\n- `StoreType`: Tipos de store: a, b, c, d\n- `Assortment`: Describe el nivel de surtido de la tienda: a = basic, b = extra, c = extended\n- `CompetitionDistance`: Distancia en metros al competidor\n- `CompetitionOpenSince[Month/Year]`: Fecha en que abrío la competencia\n- `Promo`: Si el store esta corriendo una promoción ese día\n- `Promo2`: Promo2 is a continuing and consecutive promotion for some stores: 0 = store is not participating, 1 = store is participating\n- `Promo2Since[Year/Week]`: describes the year and calendar week when the store started participating in Promo2\n- `PromoInterval`: describes the consecutive intervals Promo2 is started, naming the months the promotion is started anew. E.g. \"Feb,May,Aug,Nov\" means each round starts in February, May, August, November of any given year for that store\n", "_____no_output_____" ] ], [ [ "store['StoreType'].value_counts()", "_____no_output_____" ], [ "store['PromoInterval'].value_counts()", "_____no_output_____" ], [ "display(store_states.head(20))\ndisplay(DataFrameSummary(store_states).summary())", "_____no_output_____" ], [ "display(state_names.head(20))\ndisplay(DataFrameSummary(state_names).summary())", "_____no_output_____" ], [ "display(googletrend)\ndisplay(DataFrameSummary(googletrend).summary())", "_____no_output_____" ], [ "display(weather)\ndisplay(DataFrameSummary(weather[['Max_TemperatureC', 'Mean_TemperatureC', 'Min_TemperatureC',\n 'Max_Humidity', 'Mean_Humidity', 'Min_Humidity', 'Max_Wind_SpeedKm_h', \n 'Mean_Wind_SpeedKm_h', 'CloudCover', 'Precipitationmm']]).summary())", "_____no_output_____" ], [ "weather.columns", "_____no_output_____" ], [ "", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "code", "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "<a href=\"https://colab.research.google.com/github/hlab-repo/purity-and-danger/blob/master/Immigration.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>", "_____no_output_____" ], [ "# Creating a Model for Immigration and Outsider Language\n\nThis notebook starts with a baseline system and then provides users the opportunity to attempt to improve performance with their own custom, complete system.", "_____no_output_____" ], [ "## Set-up", "_____no_output_____" ] ], [ [ "%%capture\n!pip install datasets\n!pip install transformers", "_____no_output_____" ], [ "import re\nfrom collections import Counter\nimport datasets\nimport pandas as pd\nimport torch\nimport torch.nn as nn\nimport torch.optim as optim\nfrom sklearn.feature_extraction.text import TfidfVectorizer\nfrom sklearn.metrics import accuracy_score, confusion_matrix, f1_score, precision_score, recall_score\nfrom sklearn.model_selection import train_test_split\nfrom sklearn.naive_bayes import MultinomialNB\nfrom torch.utils.data import DataLoader\nfrom transformers import BertTokenizer, BertForSequenceClassification", "_____no_output_____" ] ], [ [ "## Getting a test dataset", "_____no_output_____" ], [ "We can start with the Common Crawl news corpus (January 2017 - December 2019). See here for details:\n\nhttps://huggingface.co/datasets/cc_news\n\nThis will constitute our test dataset. Note that the pseudolabels were generated from the beginning of this dataset but that the dataset (of 708,241 news articles) was in no way exhausted. You could perhaps skip the first 20,000 or so articles to deal only with new data.", "_____no_output_____" ] ], [ [ "# this could take several minutes\ndataset = datasets.load_dataset('cc_news')", "_____no_output_____" ], [ "dataset", "_____no_output_____" ], [ "# look at the first 10 samples\nfor i, s in enumerate(dataset['train']):\n print(s)\n if i >= 10:\n break", "{'date': '2017-04-17 00:00:00', 'description': \"Officials unsealed court documents Monday (April 17) to reveal details surrounding the first searches of Prince's Paisley Park estate.\", 'domain': '1041jackfm.cbslocal.com', 'image_url': 'https://cbs1041jackfm.files.wordpress.com/2017/04/prince-young-and-sad.jpg?w=946', 'text': 'By Abby Hassler\\nOfficials unsealed court documents Monday (April 17) to reveal details surrounding the first searches of Prince’s Paisley Park estate following his untimely death.\\nRelated: Prince’s Ex-Wife Mayte Garcia Says Memoir is not a Tell-All\\nThe unsealed search warrants don’t confirm the source of the drug, fentanyl, that led to the 57-year-old singer’s accidental, self-administered overdose last April, according to The Star Tribune.\\nInvestigators found no prescriptions in Prince’s name, however, Dr. Michael Todd Schulenberg told detectives he had written a prescription for oxycodone, which is also an opioid, under the name of long-time Prince associate and drummer Kirk Johnson.\\nBetween April 21 and Sept. 19, 2016, Carver County authorities conducted investigations into Prince’s death with a total of 11 search warrants.', 'title': 'Prince Search Warrants Unsealed, Answer Few Questions', 'url': 'http://1041jackfm.cbslocal.com/2017/04/17/prince-search-warrants-unsealed-2/'}\n{'date': '2017-08-14 00:00:00', 'description': '\"The spirit of Green Day has always been about rising above oppression.\"', 'domain': '1041jackfm.cbslocal.com', 'image_url': 'https://cbs1041jackfm.files.wordpress.com/2017/08/billie-joe-armstrong-theo-wargo-getty-images.jpg?w=946', 'text': 'By Abby Hassler\\nGreen Day’s Billie Joe Armstrong has always been outspoken about his political beliefs. Following the tragedy in Charlottesville, Virgina, over the weekend, Armstrong felt the need to speak out against the white supremacists who caused much of the violence.\\nRelated: Billie Joe Armstrong Wins #TBT with Childhood Studio Photo\\n“My heart feels heavy. I feel like what happened in Charlottesville goes beyond the point of anger,” Armstrong wrote on Facebook. “It makes me sad and desperate. shocked. I f—— hate racism more than anything.”\\n“The spirit of Green Day has always been about rising above oppression. and sticking up for what you believe in and singing it at the top of your lungs,” Armstrong continued. “We grew up fearing nuclear holocaust because of the cold war. those days are feeling way too relevant these days. these issues are our ugly past.. and now it’s coming to haunt us. always resist these doomsday politicians. and in the words of our punk forefathers .. Nazi punks f— off.”', 'title': 'Green Day’s Billie Joe Armstrong Rails Against White Nationalists', 'url': 'http://1041jackfm.cbslocal.com/2017/08/14/billie-joe-armstrong-white-nationalists/'}\n{'date': '2017-02-15 00:00:00', 'description': 'The concept, which has broken YouTube records and dominated social media, remains as simple and delightful as ever. Only this time, a rotating cast of celebrities will replace James Corden.', 'domain': '1041jackfm.cbslocal.com', 'image_url': 'https://cbs1041jackfm.files.wordpress.com/2017/02/metallica-carpool.jpg?w=946', 'text': 'By Hayden Wright\\nA trailer has been released for the next sequence of Carpool Karaoke videos, and the lineup is stellar. The CBS late-night segment will stream on Apple Music in a similar format, but with new guests and even crazier moments behind the wheel. The concept, which has broken YouTube records and dominated social media, remains as simple and delightful as ever. Only this time, a rotating cast of celebrities will replace James Corden.\\nRelated: Adele’s ‘Carpool Karaoke’ is the Most Popular Viral Video of 2016\\nHere are the moments we’re most excited for:\\nMetallica making their headbanging mark on Rihanna’s “Diamonds” with Billy Eichner.\\nJohn Legend duetting on with Alicia Keys on her breakout hit, “Falling.”\\nAriana Grande belting the Little Shop of Horrors soundtrack with Seth MacFarlane.\\nJames Corden’s return with Will Smith — they rap the Fresh Prince theme!\\nChelsea Handler slinging whiskey and singing Bon Jovi’s “Living on a Prayer” with Blake Shelton.\\nCorden also presides over an epic, R. Kelly-inspired key change for the Carpool Karaoke franchise — there’s a helicopter.\\nA new installment will drop on Apple Music each week and the premiere is “coming soon.”\\nWatch the preview here:', 'title': 'Upcoming ‘Carpool Karaoke’ Series Looks Epic', 'url': 'http://1041jackfm.cbslocal.com/2017/02/15/carpool-karaoke-series-epic/'}\n{'date': '2017-01-01 00:00:00', 'description': '\"S--- happens,\" said Carey.', 'domain': '1041jackfm.cbslocal.com', 'image_url': 'https://cbs1041jackfm.files.wordpress.com/2017/01/mariah-carey-getty.jpg?w=946', 'text': 'By Brian Ives\\nSo, you’re Mariah Carey, one of the world’s biggest stars for a quarter of a century, and you fail on national TV, big time. How do you handle it?\\nAs it turns out, you handle it with humor.\\nRelated: Mariah Carey Shades Ariana Grande, Demi Lovato\\nFor those just waking up now: in case you were totally absent from social media on New Year’s Eve (and if so, good for you!), Mariah Carey had a disastrous performance—or perhaps, non-performance is a better way to describe it—on Dick Clark’s New Year’s Rockin’ Eve last night. As CBS News reported, Carey “paced the stage without singing. It’s unclear what exactly were the technical difficulties, but the disaster was obvious as it unfolded live.”\\nPredictably, twitter erupted with mockery, including memes about 2016 claiming its final victim (Mariah Carey’s career).\\nCarey’s response was a single [NSFW] tweet, “S— happens. Have a happy and healthy new year everybody! Here’s to making more headlines in 2017.”', 'title': 'Mariah Carey Reacts to Her NYE Performance with Humor', 'url': 'http://1041jackfm.cbslocal.com/2017/01/01/mariah-carey-reacts-to-her-nye-performance-with-humor/'}\n{'date': '2017-10-06 00:00:00', 'description': 'Dubbed the \"Raw Sessions Versions,\" the tracks are prototypes for the songs we know.', 'domain': '1041jackfm.cbslocal.com', 'image_url': 'https://cbs1041jackfm.files.wordpress.com/2017/10/queen-photo-by-rogers-express-getty-images.jpg?w=946', 'text': 'By Hayden Wright\\nLast month, Queen announced a deluxe box set celebrating the 40th anniversary of their 1977 album News of the World. In the press release, the band promised “Every lead vocal is different, as are most of the lead guitar parts and a great many other instrumental details.” Now Queen have revealed the reissue versions of “We Are the Champions” and “We Will Rock You,” two of the band’s best-loved songs.\\nRelated: Queen Detail ‘News of the World’ Deluxe Box Set\\nDubbed the “Raw Sessions Versions,” the tracks are prototypes for the songs we know: Freddie Mercury’s vocals are a bit looser on “We Will Rock You,” which begins with a few warmup bars of singing. Brian May’s guitar solo is quite different, too. You get the sense that Queen were feeling their way through the tracks as they recorded earlier versions. The piano arrangement on “We Are the Champions” is brighter and happier.\\nThe News of the World box set debuts November 17. Listen to the never-before-heard raw sessions here:', 'title': 'Queen Release Alternate Takes of Classic Songs \" 104.1 Jack FM', 'url': 'http://1041jackfm.cbslocal.com/2017/10/06/queen-studio-outtakes-we-will-rock-you-we-are-the-champions/'}\n{'date': '2017-02-14 00:00:00', 'description': 'Katy Perry, Paul McCarteny and Miley Cyrus are just a few of the artists to spread the love.', 'domain': '1041jackfm.cbslocal.com', 'image_url': 'https://cbs1041jackfm.files.wordpress.com/2017/02/grammy-red-carpet-43.jpg?w=946', 'text': \"By Radio.com Staff\\nIt’s Valentine’s day and artists are taking to social media to show love for their fans and significant others.\\nRelated: John Mayer is Cupid’s Secret Weapon\\nKaty Perry, Paul McCartney and Miley Cyrus are just a few of the musicians to spread the love.\\nChris Young had his tongue planted firmly in his cheek when he wished fans a “Happy Taco Tuesday,” and then there was Kesha, who loves her fans, but worries her cats will eat her. Valid concern.\\nCheck out the best Valentine’s Day messages below.\\n❤✨So much love for my #KatyCats on this mushy day! Thanks for keeping me floating and grounded all at the same time… twitter.com/i/web/status/8… —\\nKATY PERRY (@katyperry) February 14, 2017\\nAll we need is love. Happy Valentine's Day. X #ValentinesDay https://t.co/DsxieopcYJ —\\nPaul McCartney (@PaulMcCartney) February 14, 2017\\nHappy Valentine's Day https://t.co/p4VIPntyHx —\\nKim Kardashian West (@KimKardashian) February 14, 2017\\nHave a Happy Hippie Valentimezzzzz! TBTuesday to my date night with @tywrent!!!!!! It's all about L-O-V-E everyday.… twitter.com/i/web/status/8… —\\nMiley Ray Cyrus (@MileyCyrus) February 14, 2017\\nHappy Valentine's Day to all our fans ❤️ https://t.co/IzYZEVsyEX —\\nPearl Jam (@PearlJam) February 14, 2017\\nHappy Valentine's Day ❤😘😍 I love you —\\nAustin Mahone (@AustinMahone) February 14, 2017\\nHappy Valentines Day X Adam https://t.co/eiKiysyNE3 —\\n(@U2) February 14, 2017\\nHappy Valentine's Day, ya creeps. —\\nMark Hoppus (@markhoppus) February 14, 2017\\nI think I'm forgetting something today... Oh yeah! Happy Taco Tuesday everybody! —\\n(@ChrisYoungMusic) February 14, 2017\", 'title': 'Musicians Wish Fans a Happy Valentine’s Day', 'url': 'http://1041jackfm.cbslocal.com/2017/02/14/musicians-happy-valentines-day/'}\n{'date': '2017-08-14 00:00:00', 'description': \"The lineup for the four tour dates will feature Don Henley, Joe Walsh and Timothy B. Schmit with Frey's son Deacon and Vince Gill.\", 'domain': '1041jackfm.cbslocal.com', 'image_url': 'https://cbs1041jackfm.files.wordpress.com/2017/08/the-eagles-kennedy-center.jpg?w=946', 'text': \"By Annie Reuter\\nThe Eagles have announced four new concert dates. The lineup for the short run will feature Don Henley, Joe Walsh and Timothy B. Schmit with Frey’s son, Deacon, and Vince Gill filling in for the late Glenn Frey.\\nRelated: Vince Gill to Join Eagles for Classic East and West Shows\\nAn Evening with the Eagles will stop at the Greensboro Coliseum in North Carolina on October 17 followed by a show a Philips Arena in Atlanta, Georgia. on October 20. The band will return to Louisville, Kentucky on October 24 at KFC Yum! Center before traveling to the late Glenn Frey’s hometown of Detroit, Michigan on October 27 to wrap up the run at Little Caesars Arena.\\nTickets for the four new dates go on sale at Saturday (Aug. 19) at 10 am. An American Express card member pre-sale starts on Tuesday while VIP packages will be available through Eagles.com.\\nThe four new tour dates follow the success of the band’s Classic West and East shows, earlier this summer. Deacon and Gill also played with the band for those shows.\\n“Bringing Deacon in was my idea,” Don Henley told the LA Times. “I think of the guild system, which in both Eastern and Western cultures is a centuries-old tradition of the father passing down the trade to his son, and to me, that makes perfect moral and ethical sense. The primary thing is I think Glenn would be good with it—with both of these guys. I think he’d go, ‘That’s the perfect way to do this.'”\", 'title': 'The Eagles Add Four New Dates to 2017 Tour \" 104.1 Jack FM', 'url': 'http://1041jackfm.cbslocal.com/2017/08/14/the-eagles-2017-tour-dates/'}\n{'date': '2017-02-14 00:00:00', 'description': \"George Michael's manager, Michael Lippman, wanted the three artists as well as host James Corden to perform a mashup of Michael's hits.\", 'domain': '1041jackfm.cbslocal.com', 'image_url': 'https://cbs1041jackfm.files.wordpress.com/2017/02/grammy-show-60.jpg?w=946', 'text': 'By Annie Reuter\\nAdele’s tribute to George Michael on Sunday (Feb. 12) at the GRAMMY Awards was a memorable one despite an early glitch that caused her to start the song over. But the original plans for the performance were far different. At one point the tribute could have included Beyoncé and Rihanna.\\nRelated: Adele Tributes George Michael after Rocky Start at 2017 GRAMMYs\\nGeorge Michael’s manager, Michael Lippman, wanted the three artists as well as host James Corden to perform a mashup of Michael’s hits including “Freedom” and “One More Try,” GRAMMY executive producer Ken Ehrlich told Billboard,\\nThat decision shifted when Lippman found out “how passionate Adele was,” Ehrlich says, “and that she had a vision for what she wanted to do with it.” Adele would go on to perform a ballad version of Michael’s 1996 hit “Fastlove” backed by an orchestra. It was a song she recalls hearing for the first time at the age of 10 and she instantly “heard the vulnerability in that song,” Ehrlich says.\\nBackstage following her emotional GRAMMY performance, Adele raved about Michael, adding that “it was an honor” to show her respects to the singer. Later that night, Adele won GRAMMYs for Record of the Year and Album of the Year.', 'title': 'George Michael GRAMMY Tribute Plans Included Beyoncé, Rihanna', 'url': 'http://1041jackfm.cbslocal.com/2017/02/14/george-michael-grammy-tribute-plans-beyonce-rihanna/'}\n{'date': '2017-02-15 00:00:00', 'description': '\"In these trying times we need some fun. We’re very serious about fun.”', 'domain': '1041jackfm.cbslocal.com', 'image_url': 'https://cbs1041jackfm.files.wordpress.com/2017/02/deborah-harry-getty.jpg?w=946', 'text': 'By Jon Wiederhorn\\nBlondie have released a video for “Fun,” the first single from their upcoming album Pollinator. Directed by Beyoncé collaborator Dikyal Rimmasch, the clip lives up to the song’s title, featuring the band in a space ship, wandering another planet and animated footage of galactic travel contrasted with performance shots and images of people partying on the dance floor.\\nRelated: Blondie and Garbage Announce Co-Headlining Summer Tour\\nGuitarist Chris Stein said the interstellar theme was an effort to escape the stress and turmoil of modern day events. “Quoting Emma Goldman, ‘If I can’t dance I don’t want to be part of your revolution.’ In these trying times we need some fun. We’re very serious about fun,” he told NME.\\n“The video was shot in two places,” he added. “The color stuff was all shot in LA without us and the black and white stuff was shot in New York. It’s got some good cameos—it’s got a pretty big cameo from Raja, who’s one of the Drag Race superstars, and little cameos from Grace McKagan, Duff McKagan’s daughter who has a band called The Pink Slips, and also Tony Maserati, who was a producer and mixer for us.”\\nStein and vocalist Debbie Harry wrote the upbeat, disco-inflected “Fun” with TV On The Radio member Dave Sitek. Pollinator, which is scheduled for release May 5, also features writing by Johnny Marr, Sia, Charli XCX, The Strokes guitarist Nick Valensi and Dev Hynes. It will be Blondie’s first record since 2014’s Ghost of Download.', 'title': 'Blondie Release Galactic Video for New Song ‘Fun’', 'url': 'http://1041jackfm.cbslocal.com/2017/02/15/blondie-video-new-song-fun/'}\n{'date': '2017-06-14 00:00:00', 'description': 'Check out the latest from the Las Vegas rockers.', 'domain': '1041jackfm.cbslocal.com', 'image_url': 'https://s0.wp.com/wp-content/themes/vip/cbs-local/images/global/facebook/facebook-share-260x260.png', 'text': 'The Killers have released a new single titled “The Man” from their forthcoming album, Wonderful Wonderful.\\nRelated: The Killers Perform New Track, ‘Run for Cover’\\nThe track is the first new music from the band since their 2012 album Battle Born (assuming you don’t count their 2016 Christmas album Don’t Waste Your Wishes). Hopefully, this signals a fifth studio effort is imminent. “The Man” was recorded with the producer Jacknife Lee during album sessions in Las Vegas and Los Angeles. The song finds frontman Brandon Flowers looking back on his younger self, the persona from their Grammy-nominated debut Hot Fuss, and reconciling that wide-eyed character with the man he is now.\\nCheck out the latest from The Killers below.', 'title': 'Listen to The Killers’ New Track ‘The Man’', 'url': 'http://1041jackfm.cbslocal.com/2017/06/14/the-killers-the-man-3/'}\n{'date': '2017-06-14 00:00:00', 'description': \"The song is featured in the upcoming film 'The Book of Henry.'\", 'domain': '1041jackfm.cbslocal.com', 'image_url': 'https://cbs1041jackfm.files.wordpress.com/2017/06/stevienickspress.jpg?w=946', 'text': 'By Abby Hassler\\nStevie Nicks debuted a new ballad “Your Hand I Will Never Let Go” today (June 14). The track is featured in the Naomi Wats-drama, The Book of Henry, which will hit theaters this Friday (June 16).\\nRelated: Lana Del Rey Taps Stevie Nicks for New Track: Report\\nWritten by Thomas Barlett and Ryan Miller, Nicks’ song will fall alongside original music composed by Michael Giacchino on the film’s soundtrack.\\n“Drowned in thought and caught in a stare/ Talking to ghosts who were not there,” Nicks sings. “Then you took my hand/ Transformation began/ Commotion where it once was still/ Fireworks explode/ Front row tickets to the show/ This hand I will never let it go.”\\nListen to “Your Hand I Will Never Let It Go” below.', 'title': 'Stevie Nicks Debuts New Single ‘Your Hand I Will Never Let It Go’', 'url': 'http://1041jackfm.cbslocal.com/2017/06/14/stevie-nicks-your-hand-i-will-never-let-it-go/'}\n" ] ], [ [ "## Getting pseudo-labeled data for training\n\n`0` represents viral language, `1` immigration language, and `2` a blend of the two. These categorizations are fuzzy and inexact and are not the result of manual annotations. They should be improved upon during the training process (or adjusted manually) when possible.", "_____no_output_____" ] ], [ [ "df = pd.read_csv('https://www.dropbox.com/s/kfbja23kisimedm/immigration.csv?dl=1')\ndf.head()", "_____no_output_____" ], [ "X_train, X_valid, y_train, y_valid = train_test_split(df['text'], df['target'], train_size=0.7, random_state=42)", "_____no_output_____" ], [ "X_valid, y_valid", "_____no_output_____" ] ], [ [ "# Baseline 1\n\nLet's use Naive Bayes. For the sake of simplicity, I will not add weighting to the classes here (we probably should!), but sklearn wants its weights to correspond to samples in the train dataset (when using the fit method). So you would need to feed in a list of weights the same length as your samples. Think about the weights in a table corresponding to class like this:\n\n| Sample | Class | Weight |\n| --- | --- | --- |\n| sample 1 | 1 | 0.05 |\n| sample 2 | 2 | 0.8 |\n| sample 3 | 1 | 0.05 |\n| sample 4 | 0 | 0.15 |\n| sample 5 | 2 | 0.8 |", "_____no_output_____" ] ], [ [ "vectorizer = TfidfVectorizer()\ntrain_vectorized = vectorizer.fit_transform(X_train)\nvalid_vectorized = vectorizer.transform(X_valid)\ntrain_vectorized", "_____no_output_____" ], [ "naive_bayes = MultinomialNB()\nnaive_bayes.fit(train_vectorized, y_train)", "_____no_output_____" ], [ "predictions = naive_bayes.predict(valid_vectorized)\npredictions", "_____no_output_____" ], [ "print(f'Accuracy: {accuracy_score(y_valid, predictions)}\\n'\n f'Precision: {precision_score(y_valid, predictions, average=None)}\\n'\n f'Recall: {recall_score(y_valid, predictions, average=None)}\\n'\n f'F1 Score: {f1_score(y_valid, predictions, average=None)}\\n')", "Accuracy: 0.9649855444908448\nPrecision: [0.99604156 0.95005945 0. ]\nRecall: [0.92424242 0.99875 0. ]\nF1 Score: [0.95879971 0.97379647 0. ]\n\n" ], [ "# y-axis (rows) == true label and x-axis (columns) == predicted label\nconfusion_matrix(y_valid, predictions)", "_____no_output_____" ] ], [ [ "# Baseline 2", "_____no_output_____" ] ], [ [ "model = BertForSequenceClassification.from_pretrained('bert-large-uncased', num_labels=3)\ntokenizer = BertTokenizer.from_pretrained('bert-large-uncased')", "_____no_output_____" ], [ "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')\nmodel = model.to(device)", "_____no_output_____" ], [ "# the classes are extremely unbalanced; let's generate weights that we can feed to loss function\nunbalanced_weights = 1 / (y_train.value_counts() / len(y_train)).sort_index()\nweights = unbalanced_weights / unbalanced_weights.sum()\nweights", "_____no_output_____" ], [ "# I will exclude datasets, dataloaders, etc. for the sake of simplicity\ncriterion = nn.CrossEntropyLoss(weight=torch.tensor(weights.values).float().to(device))\noptimizer = optim.AdamW(model.parameters(), lr=1e-5)\nfor epoch in range(1): # make this up to 3!\n running_loss = 0.\n for batch_start in range(0, len(X_train), 4):\n X = X_train[batch_start:batch_start + 4].tolist()\n y = torch.tensor(y_train[batch_start:batch_start + 4].values).to(device)\n\n predictions = model(**tokenizer(X, return_tensors='pt', padding=True).to(device))\n loss = criterion(torch.softmax(predictions.logits, dim=-1), y)\n optimizer.zero_grad()\n loss.backward()\n optimizer.step()\n running_loss += loss.item()\n print(f'Finished epoch {epoch} with running loss of {running_loss / len(X_train)}')", "_____no_output_____" ], [ "# make predictions on validation set\nvalid_predictions = torch.zeros_like(torch.tensor(y_valid.values))\nfor batch_start in range(0, len(X_valid), 4):\n X = X_valid[batch_start:batch_start + 4].tolist()\n\n with torch.no_grad():\n predictions = model(**tokenizer(X, return_tensors='pt', padding=True).to(device))\n indices = torch.argmax(torch.softmax(predictions.logits, dim=-1), dim=-1)\n valid_predictions[batch_start:batch_start + 4] = indices", "_____no_output_____" ], [ "print(f'Accuracy: {accuracy_score(y_valid, valid_predictions.numpy())}\\n'\n f'Precision: {precision_score(y_valid, valid_predictions.numpy(), average=None)}\\n'\n f'Recall: {recall_score(y_valid, valid_predictions.numpy(), average=None)}\\n'\n f'F1 Score: {f1_score(y_valid, valid_predictions.numpy(), average=None)}\\n')", "_____no_output_____" ], [ "# y-axis (rows) == true label and x-axis (columns) == predicted label\nconfusion_matrix(y_valid, valid_predictions.numpy())", "_____no_output_____" ] ], [ [ "# Your Original System\n\nImprove upon the baselines above. Feel free to copy cells from one of the baselines above, paste it here, and tweak it for improvements. You have several models to select from from sklearn (both for classification and for vectorization of text). And even just trying different architectures for Basline 2 (such as RoBERTa, distilbert, etc.) would help.", "_____no_output_____" ] ], [ [ "", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "%matplotlib inline", "_____no_output_____" ] ], [ [ "# 592B, Class 11.2 (11/14). Filtering review", "_____no_output_____" ] ], [ [ "import numpy as np\nimport matplotlib.pyplot as plt\nimport scipy.io.wavfile as wavfile \nimport scipy.signal as signal\nfrom scipy import fftpack\n\nfrom ipywidgets import interactive\nfrom IPython.display import Audio, display", "_____no_output_____" ] ], [ [ "## Review question: filtering and \"translating\" to cepstral analysis\n\nRemember this from last week? Here I've chosen a small frequency for `y1` and a higher frequency for `y2`. ", "_____no_output_____" ] ], [ [ "def plot_play_summed_sines(f1 = 50, f2 = 880, t_start = 0, t_stop = 1, fs = 2000, xlim_max = 0.01):\n x = np.linspace(t_start, t_stop, fs * (t_stop - t_start))\n y1 = np.sin(2*np.pi*f1*x)\n y2 = np.sin(2*np.pi*f2*x)\n \n plt.xlim(t_start,xlim_max)\n plt.plot(x , y1, \"-g\", label=\"y1\") \n plt.plot(x , y2, \"-b\", label=\"y2\") \n plt.plot(x , y1 + y2, \"-r\", label=\"y1+y2\") \n plt.legend(loc=\"upper right\") \n plt.xlabel('Time (s)')\n plt.ylabel('Amplitude (dB)')\n plt.title(\"Adding up sines\") \n \n display(Audio(data=y1, rate=fs))\n display(Audio(data=y2, rate=fs))\n display(Audio(data=y1+y2, rate=fs))\n \n return (x, y1, y2) #I've added this return line. What does this do?", "_____no_output_____" ], [ "x, y1, y2 = plot_play_summed_sines()", "_____no_output_____" ] ], [ [ "**1. Modify the code to add an additional sinusoidal signal `y3` in, with frequency `f3`. Set the default frequency of `f3` to be 300 Hz.**", "_____no_output_____" ], [ "**2. Perform a computation so you can separate out and visualize the individual component frequencies contributing to `y1 + y2 + y3`.**", "_____no_output_____" ], [ "**3. Design and implement filters so that you can separate out: (a) the highest frequency component, (b) the lowest frequency component, (c) the middle frequency componenent. What are these different kinds of filters called?** *Hint: you may find [this sample code](https://stackoverflow.com/questions/19122157/fft-bandpass-filter-in-python#19124329) helpful.*", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown" ] ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ [ [ "# 9. Import a time series", "_____no_output_____" ], [ "In this tutorial we are adding a timeseries to the DataStore object. This might be useful if the temperature in one of the calibration baths was measured with an external device. It requires three steps to add the measurement files to the DataStore object:\n1. Load the measurement files (e.g., csv, txt) with pandas into a pandas.Series object\n2. Add the pandas.Series object to the DataStore\n3. Align the time to that of the DTS measurement (required for calibration)", "_____no_output_____" ] ], [ [ "import pandas as pd\nimport os\n\nfrom dtscalibration import read_silixa_files", "/Users/bfdestombe/anaconda3/envs/dts/lib/python3.7/typing.py:847: FutureWarning: xarray subclass DataStore should explicitly define __slots__\n super().__init_subclass__(*args, **kwargs)\n" ] ], [ [ "## Step 1: load the measurement files", "_____no_output_____" ] ], [ [ "filepath = os.path.join('..', '..', 'tests', 'data', \n 'external_temperature_timeseries', \n 'Loodswaternet2018-03-28 02h.csv')\n\n# Bonus:\nprint(filepath, '\\n')\nwith open(filepath, 'r') as f:\n head = [next(f) for _ in range(5)]\nprint(' '.join(head))", "../../tests/data/external_temperature_timeseries/Loodswaternet2018-03-28 02h.csv \n\n\"time\",\"Pt100 2\"\n 2018-03-28 02:00:05, 12.748\n 2018-03-28 02:00:10, 12.747\n 2018-03-28 02:00:15, 12.746\n 2018-03-28 02:00:20, 12.747\n\n" ], [ "ts = pd.read_csv(filepath, sep=',', index_col=0, parse_dates=True, \n squeeze=True, engine='python') # the latter 2 kwargs are to ensure a pd.Series is returned\nts = ts.tz_localize('Europe/Amsterdam') # set the timezone", "_____no_output_____" ], [ "ts.head() # Double check the timezone", "_____no_output_____" ] ], [ [ "Now we quickly create a DataStore from xml-files with Stokes measurements to add the external timeseries to", "_____no_output_____" ] ], [ [ "filepath_ds = os.path.join('..', '..', 'tests', 'data', 'double_ended2')\nds = read_silixa_files(directory=filepath_ds,\n timezone_netcdf='UTC',\n file_ext='*.xml')", "6 files were found, each representing a single timestep\n6 recorded vars were found: LAF, ST, AST, REV-ST, REV-AST, TMP\nRecorded at 1693 points along the cable\nThe measurement is double ended\nReading the data from disk\n" ] ], [ [ "## Step 2: Add the temperature measurements of the external probe to the DataStore.\n\nFirst add the coordinates", "_____no_output_____" ] ], [ [ "ds.coords['time_external'] = ts.index.values", "_____no_output_____" ] ], [ [ "Second we add the measured values", "_____no_output_____" ] ], [ [ "ds['external_probe'] = (('time_external',), ts)", "_____no_output_____" ] ], [ [ "## Step 3: Align the time of the external measurements to the Stokes measurement times\nWe linearly interpolate the measurements of the external sensor to the times we have DTS measurements", "_____no_output_____" ] ], [ [ "ds['external_probe_dts'] = ds['external_probe'].interp(time_external=ds.time)", "_____no_output_____" ], [ "print(ds.data_vars)", "Data variables:\n st (x, time) float64 1.281 -0.5321 ... -43.44 -41.08\n ast (x, time) float64 0.4917 1.243 ... -30.14 -32.09\n rst (x, time) float64 0.4086 -0.568 ... 4.822e+03\n rast (x, time) float64 2.569 -1.603 ... 4.224e+03\n tmp (x, time) float64 196.1 639.1 218.7 ... 8.442 18.47\n acquisitionTime (time) float32 2.098 2.075 2.076 2.133 2.085 2.062\n referenceTemperature (time) float32 21.0536 21.054 ... 21.0531 21.057\n probe1Temperature (time) float32 4.36149 4.36025 ... 4.36021 4.36118\n probe2Temperature (time) float32 18.5792 18.5785 ... 18.5805 18.5723\n referenceProbeVoltage (time) float32 0.121704 0.121704 ... 0.121705\n probe1Voltage (time) float32 0.114 0.114 0.114 0.114 0.114 0.114\n probe2Voltage (time) float32 0.121 0.121 0.121 0.121 0.121 0.121\n userAcquisitionTimeFW (time) float32 2.0 2.0 2.0 2.0 2.0 2.0\n userAcquisitionTimeBW (time) float32 2.0 2.0 2.0 2.0 2.0 2.0\n external_probe (time_external) float64 12.75 12.75 ... 12.76 12.76\n external_probe_dts (time) float64 12.75 12.75 12.75 12.75 12.75 12.75\n" ] ], [ [ "Now we can use `external_probe_dts` when we define sections and use it for calibration", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "\"\"\"Udacity - Self-Driving Engineer Nanodegree\nProject 1: Finding Lane Lines on the Road\nNikko Sadural\"\"\"\n\nimport matplotlib.pyplot as plt\nimport matplotlib.image as mpimg\nimport numpy as np\nimport cv2\nimport math\nfrom moviepy.editor import VideoFileClip\nfrom IPython.display import HTML\n%matplotlib inline\n\ndef grayscale(img):\n \"\"\"Applies the Grayscale transform\n This will return an image with only one color channel\n but NOTE: to see the returned image as grayscale\n (assuming your grayscaled image is called 'gray')\n you should call plt.imshow(gray, cmap='gray')\"\"\"\n return cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)\n # Or use BGR2GRAY if you read an image with cv2.imread()\n # return cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)\n\ndef canny(img, low_threshold, high_threshold):\n \"\"\"Applies the Canny transform\"\"\"\n return cv2.Canny(img, low_threshold, high_threshold)\n\ndef gaussian_blur(img, kernel_size):\n \"\"\"Applies the Gaussian Noise kernel\"\"\"\n return cv2.GaussianBlur(img, (kernel_size, kernel_size), 0)\n\ndef region_of_interest(img, vertices):\n \"\"\"\n Applies an image mask.\n \n Only keeps the region of the image defined by the polygon\n formed from 'vertices'. The rest of the image is set to black.\n 'vertices' should be a numpy array of integer points.\n \"\"\"\n #defining a blank mask to start with\n mask = np.zeros_like(img)\n \n #defining a 3 channel or 1 channel color to fill the mask with depending on the input image\n if len(img.shape) > 2:\n channel_count = img.shape[2]\n ignore_mask_color = (255,) * channel_count\n else:\n ignore_mask_color = 255\n \n #filling pixels inside the polygon defined by \"vertices\" with the fill color\n cv2.fillPoly(mask, vertices, ignore_mask_color)\n \n #returning the image only where mask pixels are nonzero\n masked_image = cv2.bitwise_and(img, mask)\n return masked_image\n\ndef draw_lines(img, lines, top_left_y, top_right_y, color=[255, 0, 0], thickness=10):\n \"\"\"\n NOTE: this is the function you might want to use as a starting point once you want to\n average/extrapolate the lines segments you detect to map out the full\n extent of the lane (going from the result shown in raw-lines-example.mp4\n to that shown in P1_example.mp4.)\n \n Think about things like separating lines segments by their\n slope ((y2-y1)/(x2-x1)) to decide which segments are part of the left\n line vs. the right line. Then, you can average the position of each of\n the lines and extrapolate to the top and bottom of the lane.\n \n This function draws 'lines' with 'color' and 'thickness'.\n Lines are drawn on the image inplace (mutates the image).\n If you want to make the lines semi-transparent, think about combining\n this function with the weighted_img() function below\n \"\"\"\n \n #Initialize variables for holding high/low x/y values and counters\n x1_low1_pos = img.shape[1]\n y1_low1_pos = img.shape[0]\n x1_low2_pos = img.shape[1]\n y1_low2_pos = img.shape[0]\n x2_high1_pos = 0\n y2_high1_pos = 0\n x2_high2_pos = 0\n y2_high2_pos = 0\n x1_low1_neg = img.shape[1]\n y1_low1_neg = img.shape[0]\n x1_low2_neg = img.shape[1]\n y1_low2_neg = img.shape[0]\n x2_high1_neg = 0\n y2_high1_neg = 0\n x2_high2_neg = 0\n y2_high2_neg = 0\n m_sum_pos = 0\n sum_count_pos = 0\n m_sum_neg = 0\n sum_count_neg = 0\n \n for line in lines:\n \n for x1,y1,x2,y2 in line:\n \n #Get slope and print line variables\n m = ((y2-y1)/(x2-x1))\n \n #For positive/negative slopes, get two smallest x1's and two largest x2's \n #(x1_low2 < x1_low1)\n #(x2_high2 > x2_high1) \n \n if m > 0.55 and m < 0.9:\n #Get two smallest x1's\n if x1 < x1_low1_pos and x1 < x1_low2_pos:\n x1_low1_pos = x1_low2_pos\n y1_low1_pos = y1_low2_pos\n x1_low2_pos = x1\n y1_low2_pos = y1\n elif x1 < x1_low1_pos and x1 > x1_low2_pos:\n x1_low1_pos = x1\n y1_low1_pos = y1\n \n #Get two largest x2's\n if x2 > x2_high1_pos and x2 > x2_high2_pos:\n x2_high1_pos = x2_high2_pos\n y2_high1_pos = y2_high2_pos\n x2_high2_pos = x2\n y2_high2_pos = y2\n elif x2 > x2_high1_pos and x2 < x2_high2_pos:\n x2_high1_pos = x2\n y2_high1_pos = y2\n m_sum_pos = m_sum_pos + m\n sum_count_pos = sum_count_pos + 1\n \n elif m < -0.55 and m > -0.9:\n #Get two smallest x1's\n if x1 < x1_low1_neg and x1 < x1_low2_neg:\n x1_low1_neg = x1_low2_neg\n y1_low1_neg = y1_low2_neg\n x1_low2_neg = x1\n y1_low2_neg = y1\n elif x1 < x1_low1_neg and x1 > x1_low2_neg:\n x1_low1_neg = x1\n y1_low1_neg = y1\n \n #Get two largest x2's\n if x2 > x2_high1_neg and x2 > x2_high2_neg:\n x2_high1_neg = x2_high2_neg\n y2_high1_neg = y2_high2_neg\n x2_high2_neg = x2\n y2_high2_neg = y2\n elif x2 > x2_high1_neg and x2 < x2_high2_neg:\n x2_high1_neg = x2\n y2_high1_neg = y2\n m_sum_neg = m_sum_neg + m\n sum_count_neg = sum_count_neg + 1\n \n #Calculate positive line slope from average of detected line slopes\n m_pos = round(m_sum_pos / sum_count_pos,2)\n \n #Calculate endpoints for extrapolated line with positive slope\n x2_calc_pos = int(x1_low2_pos + (1/m_pos)*(img.shape[0] - y1_low2_pos))\n x1_calc_pos = (math.floor(x2_calc_pos - (1/m_pos)*(img.shape[0] - top_right_y)))\n x2_calc_pos = (math.floor(x1_calc_pos + (1/m_pos)*(img.shape[0] - top_right_y)))\n \n #Calculate negative line slope from average of detected line slopes\n m_neg = round(m_sum_neg / sum_count_neg,2)\n \n #Calculate endpoints for extrapolated line with negative slope\n x1_calc_neg = int(x2_high2_neg - (1/m_neg)*(y2_high2_neg - img.shape[0]))\n x2_calc_neg = (math.floor(x1_calc_neg + (1/m_neg)*(top_left_y - img.shape[0])))\n x1_calc_neg = (math.floor(x2_calc_neg - (1/m_neg)*(top_left_y - img.shape[0])))\n \n #Draw extrapolated lines with positive and negative slopes onto image\n new_img = cv2.line(img, (x1_calc_pos, top_right_y), (x2_calc_pos, img.shape[0]), color, thickness)\n cv2.line(new_img, (x1_calc_neg, img.shape[0]), (x2_calc_neg, top_left_y), color, thickness)\n \ndef hough_lines(img, rho, theta, threshold, min_line_len, max_line_gap, top_left_y, top_right_y):\n \"\"\"\n 'img' should be the output of a Canny transform.\n \n Returns an image with hough lines drawn.\n \"\"\"\n lines = cv2.HoughLinesP(img, rho, theta, threshold, np.array([]), minLineLength=min_line_len, maxLineGap=max_line_gap)\n line_img = np.zeros((img.shape[0], img.shape[1], 3), dtype=np.uint8)\n draw_lines(line_img, lines, top_left_y, top_right_y)\n return line_img\n\ndef weighted_img(img, initial_img, α=0.8, β=1., γ=0.):\n \"\"\"\n 'img' is the output of the hough_lines(), an image with lines drawn on it.\n Should be a blank image (all black) with lines drawn on it.\n \n 'initial_img' should be the image before any processing.\n \n The result image is computed as follows:\n \n initial_img * α + img * β + γ\n NOTE: initial_img and img must be the same shape!\n \"\"\"\n return cv2.addWeighted(initial_img, α, img, β, γ)\n\ndef process_image(image):\n #Read in image and convert to grayscale\n gray = grayscale(image)\n\n #Gaussian smoothing to suppress noise/gradients by averaging\n kernel_size = 3\n blur_gray = gaussian_blur(gray, kernel_size)\n\n #Canny edge detection to find edges\n low_threshold = 150\n high_threshold = 250\n masked_edges = canny(blur_gray, low_threshold, high_threshold)\n\n #Define polygon for region masking\n imshape = masked_edges.shape\n top_left_x = 440\n top_left_y = 330\n top_right_x = 530\n top_right_y = 330\n vertices = np.array([[(152,imshape[0]),(top_left_x, top_left_y),(top_right_x, top_right_y),(imshape[1]-70,imshape[0])]],dtype=np.int32)\n masked_image = region_of_interest(masked_edges, vertices)\n \n #Hough to find line segments on edge-detected Canny image\n rho = 1 # distance resolution in pixels of Hough grid\n theta = np.pi/180 # angular resolution in radians of Hough grid\n threshold = 15 # 15 minimum number of votes (Hough grid cell intersections)\n min_line_length = 20 # 20 minimum number of pixels making up a line\n max_line_gap = 1000 # maximum gap between connectable line segments\n lines_image = hough_lines(masked_image, rho, theta, threshold, min_line_length, max_line_gap, top_left_y, top_right_y)\n \n #Draw lines on the edge image\n lines_edges = weighted_img(lines_image, image)\n plt.imshow(lines_edges)\n \n return lines_edges\n\n\nwhite_output = 'test_videos_output/solidWhiteRight.mp4'\nclip1 = VideoFileClip(\"test_videos/solidWhiteRight.mp4\")\nwhite_clip = clip1.fl_image(process_image)\n%time white_clip.write_videofile(white_output, audio=False)\n\nyellow_output = 'test_videos_output/solidYellowLeft.mp4'\nclip2 = VideoFileClip('test_videos/solidYellowLeft.mp4')\nyellow_clip = clip2.fl_image(process_image)\n%time yellow_clip.write_videofile(yellow_output, audio=False)\n\nimage = mpimg.imread('test_images/whiteCarLaneSwitch.jpg')\nprocess_image(image)", "[MoviePy] >>>> Building video test_videos_output/solidWhiteRight.mp4\n[MoviePy] Writing video test_videos_output/solidWhiteRight.mp4\n" ] ] ]

[ "code" ]

[ [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# basic\nimport sys\nimport os\n\n# common\nimport numpy as np\nimport pandas as pd\nimport xarray as xr\nimport matplotlib.pyplot as plt\nfrom datetime import datetime, timedelta\nfrom sklearn.cluster import KMeans\nimport pickle\nimport warnings\nwarnings.filterwarnings('ignore')\nfrom IPython.display import Image\n\n#lib\nfrom lib.validation_methodology_plots import *\n", "_____no_output_____" ], [ "path_p = r'/home/administrador/Documentos/seasonal/seasonal_forecast/new/'", "_____no_output_____" ], [ "df_2021 = pd.read_pickle(path_p+'df_coordinates_pmin_sst_mld_2021.pkl')\nxs = xr.open_dataset(path_p+'xs_index_vars_19822019_2deg_new.nc')\nxds_kma = xr.open_dataset(path_p+'kma_model/xds_kma_index_vars_1b.nc')\nxs_dwt_counts = xr.open_dataset(path_p+'kma_model/xds_count_tcs8.nc')\nxs_dwt_counts_964 = xr.open_dataset(path_p+'kma_model/xds_count_tcs8_964.nc')\nxds_timeM = xr.open_dataset(path_p+'xds_timeM8.nc')\nxds_PCA = xr.open_dataset(path_p+'xds_PCA.nc')\nxds_kma_ord = xr.open_dataset(path_p+'xds_kma_ord.nc')", "_____no_output_____" ] ], [ [ "<br>\n<br>\n<br>\n\n# <font color='navy'>**Model Validation** </font> \n", "_____no_output_____" ], [ ">[Index predictor](#p)<br> <br>\n>[Cluster comparison](#cc)<br> <br>\n>[Predictand computation and plotting](#plv)<br> <br>\n\n\n\n\n\n\n\n\n\n\n\n", "_____no_output_____" ], [ "<br>\n<br>\n\n**After analizing the tailor-made predictor along the hindcast data for the calibration period (1982-2019), the performace of the model will be validated for year 2020, which has not been included in the predictor calibration process.**", "_____no_output_____" ], [ "<br>\n\n\n<div style=\"padding: 15px; border: 1px solid transparent; border-color: transparent; margin-bottom: 20px; border-radius: 4px; color: rgb(0,0,0); background-color: #fcf8e3; border-color: #faebcc; \">\n \n\n**Steps:**\n* **1.** Download and preprocess (file conversion and resolution interpolation) SST and MLD data for the validation time period.\n* **2.** Generation of the index predictor based on the index function obtained at the calibration period.\n* **3.** The fitted Principal Component Analysis for the calibration is used to predict the index principal components in that same temporal-spatial space.\n* **4.** The predicted PCs are assigned to the best match unit group from the fitted K-means clustering -> based on the index predictor a DWT is assigned to each day.\n* **5.** From the DWT the expected daily mean number of TCs in 8x8º cells map in the target area is known.\n \n\n</div>", "_____no_output_____" ], [ "<br />\n<br />\n\n## <font color='royalblue'>**Index predictor and DWTs**</font> <a name=\"p\"></a>", "_____no_output_____" ], [ "\n\n**Download and preprocess (file conversion and resolution interpolation) SST and MLD data for the validation time period.**", "_____no_output_____" ] ], [ [ "path_val = r'/home/administrador/Documentos/seasonal/seasonal_forecast/validation/'\nyear_val = 2020", "_____no_output_____" ], [ "change_sst_resolution_val(path_val,year_val)", "Start time: 2021-04-12 09:57:46.179185\nEnd time: 2021-04-12 10:01:47.651739\n" ] ], [ [ "<br>\n\n**Generation of the index predictor based on the index function obtained at the calibration period.**", "_____no_output_____" ] ], [ [ "xs_val = ds_index_over_time_val(path_val,path_p,year_val)\nxs_val", "_____no_output_____" ] ], [ [ "<br>\n<br>\n\n**The fitted Principal Component Analysis for the calibration is used to predict the index principal components in that same temporal-spatial space and the predicted PCs are assigned to the best match unit group from the fitted K-means clustering -> based on the index predictor a DWT is assigned to each day.**", "_____no_output_____" ] ], [ [ "val_bmus = PCA_k_means_val(path_p,path_val,xs_val)", "K-means order previously obtained in the calibration period: \n\n[11 27 28 45 16 23 47 19 7 14 0 31 46 30 34 35 15 2 8 33 37 41 5 10\n 48 4 40 21 3 20 42 29 38 26 32 43 13 17 18 39 25 22 44 9 1 24 6 12\n 36]\n\n\n DWTs for the validation period and their counting: \n\n(array([ 6, 8, 9, 12, 15, 17, 18, 24, 25, 26, 27, 30, 32, 35, 36, 37, 40,\n 42, 43, 45, 49]), array([31, 1, 37, 11, 33, 18, 18, 8, 13, 30, 8, 5, 1, 21, 1, 10, 27,\n 22, 8, 3, 60]))\n" ] ], [ [ "<br>\n<br>\n\n**Chronology of the DWTs:**", "_____no_output_____" ] ], [ [ "fig_bmus = plot_bmus_chronology(xs_val,val_bmus,year_val)", "_____no_output_____" ] ], [ [ "<br>\n\n**The resulting classification can be seen in the PCs space of the predictor index data. The obtained centroids (black dots), span the wide variability of the data.**", "_____no_output_____" ] ], [ [ "fig = plot_scatter_kmeans(xds_kma_ord, val_bmus, xds_kma_ord.cenEOFs.values, size_l=12, size_h=10);", "_____no_output_____" ] ], [ [ "<br />\n<br />\n\n## <font color='royalblue'>**Cluster comparison**</font> <a name=\"cc\"></a>", "_____no_output_____" ] ], [ [ "fig = plot_bmus_comparison_validation_calibration(xs,xds_kma,xs_val,val_bmus,9,49)", "_____no_output_____" ] ], [ [ "<br />\n<br />\n\n## <font color='royalblue'>**Predictand computation and plotting**</font> <a name=\"plv\"></a>", "_____no_output_____" ], [ "**From the DWT the daily expected mean number of TCs in 8x8º cells in the target area is known for each day and thus maps at different time scales can be computed.**", "_____no_output_____" ], [ "**Daily mean expected number of TCs**", "_____no_output_____" ] ], [ [ "xds_timeline_val,xs_M_val = ds_monthly_probabilities_val(df_2021,val_bmus,xs_val,xs_dwt_counts,xs_dwt_counts_964)", "_____no_output_____" ] ], [ [ "<br>\n\n**Monthly aggregated mean expected number of TCs**", "_____no_output_____" ] ], [ [ "xs_M_val", "_____no_output_____" ], [ "fig_val_year_8 = plot_validation_year(df_2021,xs_M_val,xds_timeline_val,35)", "_____no_output_____" ] ], [ [ "<br>\n\n**Whole period aggregated mean expected number of TCs**", "_____no_output_____" ] ], [ [ "fig_val_year_8 = plot_validation_full_season(df_2021,xs_M_val,xds_timeline_val,35)", "_____no_output_____" ] ], [ [ "<br>\n<br>\n\n<div style=\"padding: 15px; border: 1px solid transparent; border-color: transparent; margin-bottom: 20px; border-radius: 4px; color: rgb(0,0,0); background-color: #fcf8e3; border-color: #faebcc; \">\n \n\n* **The model performs very well when estimating the expected TC activity (number and intensity of TCs), not understimating the threat.** \n \n* **In some cells adjacents to the cells including TC tracks it overstimates TC activity.**\n \n</div>", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "code", "code", "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import numpy as np\nimport matplotlib.pyplot as plt", "_____no_output_____" ] ], [ [ "# Parte 1: Iteração de Rayleigh\n\nVimos que podemos iterar um vetor $v$ pela matriz $A$, obtendo a sequência de vetores $A^nv$, por multiplicações sucessivas, e que isso permite encontrar um autovetor.", "_____no_output_____" ], [ "## Questão 1\n\nImplemente uma função `itera(A,v,tol,debug)` que itera o vetor $v$, normalizando a cada iteração, e que retorna $(v_\\lambda, \\lambda, n)$, respectivamente:\n- uma estimativa do autovetor\n- uma estimativa do autovalor correspondente\n- o número de iterações realizadas até atingir a precisão `tol`.\n\nSe `debug == True`, retorne também a lista dos vetores (unitários) produzidos ao longo do processo.", "_____no_output_____" ] ], [ [ "def itera(A,v, tol=1e-12, maxiter=1000, debug=False):\n v = np.array(v)\n n,m = np.shape(A)\n assert n==m, 'A must be square'\n \n def eigenvector_normalizer(A,v,n=29):\n answ = []\n answ.append(v)\n for i in range(0,n):\n v_next = A @ answ[-1]\n answ.append(v_next/np.linalg.norm(v_next))\n return answ,n\n \n def eigenvalue_picker(v,u):\n v_max = vs[-1]\n print(v_max)\n return v_max/u\n \n vs,it = eigenvector_normalizer(A,v)\n l = np.linalg.norm(A@vs[-1])\n if debug == True:\n return vs[-1],l,it,vs\n else:\n return vs[-1],l,it ", "_____no_output_____" ], [ "# Autovetores conhecidos\nA = [[1,2],[2,1]]\nalvo = np.array([1,1])/np.sqrt(2)\n\nv, l, n = itera(A,[1,2])\n\nassert(abs(l-3) < 1e-15)\nassert(all(abs(v-alvo) < 1e-12))\nassert(n < 30)", "_____no_output_____" ], [ "# Autovetores aleatórios: verificando que satisfaz (aproximadamente) a definição\nnp.random.seed(4444)\nA = np.random.rand(4,4)\nv, l, n = itera(A, np.random.rand(4))\nerr = np.dot(A,v) - l*v\nassert(np.linalg.norm(err) < 1e-12)\nassert(n < 30)", "_____no_output_____" ] ], [ [ "## Questão 2: Convergência\n\nTemos o número de iterações, mas não vimos como o algoritmo \"converge\" para o autovetor.\nAssim, use os vetores intermediários e faça um gráfico da evolução do erro entre os $v$'s produzidos e o autovetor $v_\\lambda$.", "_____no_output_____" ] ], [ [ "ax = None\nv,l,n, vs_intermediarios = itera(A, np.random.rand(4), debug=True)\ntam = len(vs_intermediarios)\nrng = (0,n)\nks = [([v]*tam)/i for i in vs_intermediarios]\n#plt.plot(ks,rng)\n\nax = plt.gca()\nplt.show()", "30\n" ], [ "assert ax.title.get_text() != \"\"\nassert len(ax.lines) == 1\n\nys = ax.lines[0].get_ydata()\n\nassert min(ys) < 1e-12\nassert np.all(ys[:-1] > ys[1:])", "_____no_output_____" ] ], [ [ "O que o último assert quer dizer?", "_____no_output_____" ], [ "Compara se o $ys[i]$ é menor que $ys[i-1]$, ou seja, se a lista está convergindo", "_____no_output_____" ], [ "## Questão 3: Convergência comparada\n\nPara cada um dos vetores `d1` e `d2` abaixo, considere a matriz $A = \\operatorname{diag}(d_i)$ correspondente.", "_____no_output_____" ] ], [ [ "d1 = [1,10,20,30,31,32]\nd2 = [1,10,20,29,30,32]", "_____no_output_____" ] ], [ [ "Qual é o autovetor com o maior autovalor para $A_1$ e $A_2$?", "_____no_output_____" ], [ "YOUR ANSWER HERE", "_____no_output_____" ], [ "Agora, compare a velocidade de convergência do autovetor usando `itera` para cada uma destas matrizes,\nfazendo o gráfico do erro entre os vetores gerados para $A_1$ e $A_2$ no mesmo eixo.", "_____no_output_____" ] ], [ [ "ax = []\n\n_,_,_,l_1 = itera(np.diag(d1), np.ones_like(d1), debug=True)\n_,_,_,l_2 = itera(np.diag(d2), np.ones_like(d2), debug=True)\n\n# YOUR CODE HERE\nraise NotImplementedError()\n\nax = plt.gca()\nplt.show()", "_____no_output_____" ], [ "assert ax.title.get_text() != \"\"\nassert len(ax.lines) == 2\nassert len(ax.legend().texts) == 2", "_____no_output_____" ] ], [ [ "Para qual matriz há convergência mais rápida? Como você explicaria isso?", "_____no_output_____" ], [ "YOUR ANSWER HERE", "_____no_output_____" ], [ "## Questão 4: Convergência?\n\nSejam $\\theta \\in [0,2\\pi]$ e $\\alpha \\in \\mathbb{R}$,\ne considere a matriz \n$$A(\\theta, \\alpha) = \\begin{bmatrix} \n\\cos(\\theta) & \\sin(\\theta) & 0\\\\\n-\\sin(\\theta) &\\cos(\\theta) & 0\\\\\n0 & 0 & \\alpha\\\\\n\\end{bmatrix}.$$\n\nQual a interpretação geométrica dessa matriz?", "_____no_output_____" ], [ "YOUR ANSWER HERE", "_____no_output_____" ], [ "Quais são os autovetores de $A$ (em função de $\\theta$ e $\\alpha$)?", "_____no_output_____" ], [ "YOUR ANSWER HERE", "_____no_output_____" ], [ "Implemente a função abaixo que gera a matriz $A$:", "_____no_output_____" ] ], [ [ "def make_matrix(theta,alpha):\n # YOUR CODE HERE\n raise NotImplementedError()", "_____no_output_____" ], [ "assert np.allclose(make_matrix(0,1),np.eye(3))\nassert np.allclose(make_matrix(np.pi,0.5),[[-1,0,0],[0,-1,0],[0,0,0.5]])", "_____no_output_____" ] ], [ [ "Fixando $\\theta = \\dfrac{\\pi}{4}$,\nfaça um gráfico do número de iterações necessários para calcular o maior autovetor,\nem função de $\\alpha \\in [0.5,1.5]$.", "_____no_output_____" ] ], [ [ "alphas = np.linspace(0.5,1.5,100)\nax = []\n# YOUR CODE HERE\nraise NotImplementedError()\n\nax = plt.gca()\nplt.show()", "_____no_output_____" ], [ "assert ax.title.get_text() != \"\"\nassert len(ax.lines) == 1\nassert ax.get_xlabel() != \"\"\n\nys = ax.lines[0].get_ydata()\n\nassert 100 > ys.min() > 60\nassert ys[55] < 600\nassert ys[50] > 900", "_____no_output_____" ] ], [ [ "Agora, faça o gráfico com a estimativa do autovalor, novamente em função de $\\alpha$.", "_____no_output_____" ] ], [ [ "ax = []\n\n# YOUR CODE HERE\nraise NotImplementedError()\n\nax = plt.gca()\nplt.show()", "_____no_output_____" ], [ "assert ax.title.get_text() != \"\"\nassert len(ax.lines) == 1\nassert ax.get_xlabel() != \"\"\n\nys = ax.lines[0].get_ydata()\n\nassert np.all(0.7 <= ys) and np.all(ys <= 1.5)", "_____no_output_____" ] ], [ [ "Como explicar a variação no número de iterações? O que isso tem a ver com o autovalor retornado?", "_____no_output_____" ], [ "YOUR ANSWER HERE", "_____no_output_____" ], [ "# Parte 2: Generalizando\n\n## Questão 5: Outra iteração, novos limites\n\nEm vez de iterar $A^n v$, é possível iterar $A^{-n} v$.\nAssim, em vez de \"aumentar\" os vetores correspondentes aos autovalores de módulo grande,\nestes serão \"diminuídos\", e sobra o vetor do \"menor\" (de novo, em módulo) autovalor.\n\nMostre que $\\dfrac{A^{-n}v_0}{\\lVert A^{-n}v_0 \\rVert} \\rightarrow v_{min}$, onde $v_{min}$ é o \"menor\" autovalor de $A$.", "_____no_output_____" ], [ "YOUR ANSWER HERE", "_____no_output_____" ], [ "Agora, generalize um pouco mais:\n\nSeja $\\alpha \\in C$ um número complexo qualquer.\nMostre que\n$$\\frac{(A - \\alpha I)^{-n}v_0}{\\lVert (A - \\alpha I)^{-n}v_0 \\rVert} \\rightarrow v_{\\alpha},$$\nonde $v_{\\alpha}$ é o autovetor de $A$ com autovalor mais próximo de $\\alpha$.\n\nEste método é conhecido como \"Iteração inversa deslocada\".", "_____no_output_____" ], [ "YOUR ANSWER HERE", "_____no_output_____" ], [ "## Questão 6: Iteração inversa com deslocamento\n\nImplemente a iteração inversa com deslocamento, com argumentos semelhantes a função `itera`.", "_____no_output_____" ] ], [ [ "def inverse_iteration(A, v, alpha=0, tol=1e-12, maxiter=1000, debug=False):\n\n v = np.array(v)\n n,m = np.shape(A)\n assert n==m, 'A must be square'\n\n # YOUR CODE HERE\n raise NotImplementedError()", "_____no_output_____" ], [ "A = [[1,2],[2,1]]\nans = np.array([-1,1])/np.sqrt(2)\n\nv, l, n = inverse_iteration(A,[1,2])\n\nassert np.allclose(np.linalg.norm(v),1)\nassert np.allclose(v,ans) or np.allclose(v, -ans)\nassert 20 < n < 40", "_____no_output_____" ], [ "A = [[1,2],[2,1]]\nans = np.array([1,1])/np.sqrt(2)\n\nv, l, n = inverse_iteration(A,[1,2], alpha=2, maxiter=50)\n\nassert np.allclose(np.linalg.norm(v),1)\nassert np.allclose(v,ans) or np.allclose(v, -ans)\nassert 20 < n < 40", "_____no_output_____" ], [ "A = [[1,2],[2,1]]\nans = np.array([1,1])/np.sqrt(2)\n\nv, l, n = inverse_iteration(A,[1,2], alpha=2.5, maxiter=50)\n\nassert np.allclose(np.linalg.norm(v),1)\nassert np.allclose(v,ans) or np.allclose(v, -ans)\nassert 10 < n < 20", "_____no_output_____" ] ], [ [ "## Questão 7: Convergência comparada\n\nFaça o gráfico da velocidade de convergência dos autovetores da iteração inversa aplicada à matriz $A$ acima,\npara $\\alpha \\in \\{-2,0,2\\}$.", "_____no_output_____" ] ], [ [ "np.random.seed(1234)\n\nax = []\nv0 = np.random.rand(2)\n\n\n# YOUR CODE HERE\nraise NotImplementedError()\n\nplt.ylabel('Distance to eigenvector')\n\nax = plt.gca()\nplt.show()", "_____no_output_____" ], [ "assert ax.title.get_text() != \"\"\nassert len(ax.lines) == 3\nassert len(ax.legend().texts) == 3\nassert ax.get_xlabel() != \"\"\n\nys = [l.get_ydata() for l in ax.lines]\n\nassert np.isclose(max(max(y) for y in ys),2)\nassert min(min(y) for y in ys) <= 1e-16", "_____no_output_____" ] ], [ [ "Qual valor de $\\alpha$ levou à convergência mais rápida?\nComo você explicaria isso?", "_____no_output_____" ], [ "YOUR ANSWER HERE", "_____no_output_____" ], [ "O que mais você observa neste gráfico?", "_____no_output_____" ], [ "YOUR ANSWER HERE", "_____no_output_____" ], [ "## Questão 8: Zoom da convergência", "_____no_output_____" ], [ "Agora, repita o mesmo gráfico para $\\alpha \\in \\{2, 2.5, 2.9, 2.99 \\}$.", "_____no_output_____" ] ], [ [ "np.random.seed(1234)\n\nax = []\nv0 = np.random.rand(2)\n\n\n# YOUR CODE HERE\nraise NotImplementedError()\n\nplt.ylabel('Distance to eigenvector')\n\nax = plt.gca()\nplt.show()", "_____no_output_____" ], [ "assert ax.title.get_text() != \"\"\nassert len(ax.lines) == 4\nassert len(ax.legend().texts) == 4\nassert ax.get_xlabel() != \"\"\n\nys = [l.get_ydata() for l in ax.lines]\n\nassert min(min(y) for y in ys) <= 1e-16", "_____no_output_____" ] ], [ [ "O que este gráfico sugere quanto à velocidade de convergência da iteração inversa?\nSerá que isso já era possível de \"ver\" no outro gráfico?", "_____no_output_____" ], [ "YOUR ANSWER HERE", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "code" ], [ "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Homework 1: Classification With Naive Bayes\n\n", "_____no_output_____" ], [ "## Problem 1: Diabetes Classification\n\nPoints: 40\n\nA famous collection of data on whether a patient has diabetes, known as the Pima Indians dataset, and originally owned by the National Institute of Diabetes and Digestive and Kidney Diseases can be found at Kaggle. Download this dataset from https://www.kaggle.com/kumargh/pimaindiansdiabetescsv. This data has a set of attributes of patients, and a categorical variable telling whether the patient is diabetic or not. For several attributes in this data set, a value of 0 may indicate a missing value of the variable. There are a total of 767 data-points.\n\n", "_____no_output_____" ], [ "## Part 1A \nBuild a simple naive Bayes classifier to classify this data set. You should use a normal distribution to model each of the class-conditional distributions.\n\nCompute an estimate of the accuracy of the classifier by averaging over 10 test-train splits. Each split should randomly assign 20% of the data to test, and the rest to train.\nYou should write this classifier and the test-train split code yourself (it's quite straight-forward). Libraries can be used to load & hold the data.\n", "_____no_output_____" ], [ "### Answer Part 1A\n\nTo build a somple naive Bayes classifier to classify Pima Indians dataset. We will be using Python 3 in Google Colab.\n\n", "_____no_output_____" ], [ "#### Set up\n\nLoad required libraries\n", "_____no_output_____" ] ], [ [ "!pip install pandas_profiling", "Collecting pandas_profiling\n Downloading https://files.pythonhosted.org/packages/a7/7c/84f15ee705793a3cdd43bc65e6166d65d36f743b815ea517b02582989533/pandas_profiling-1.4.1-py2.py3-none-any.whl\nRequirement already satisfied: six>=1.9 in /Users/sms/anaconda3/envs/insight/lib/python3.6/site-packages (from pandas_profiling) (1.12.0)\nRequirement already satisfied: pandas>=0.19 in /Users/sms/anaconda3/envs/insight/lib/python3.6/site-packages (from pandas_profiling) (0.23.4)\nRequirement already satisfied: jinja2>=2.8 in /Users/sms/anaconda3/envs/insight/lib/python3.6/site-packages (from pandas_profiling) (2.10)\nRequirement already satisfied: matplotlib>=1.4 in /Users/sms/anaconda3/envs/insight/lib/python3.6/site-packages (from pandas_profiling) (3.0.2)\nRequirement already satisfied: python-dateutil>=2.5.0 in /Users/sms/anaconda3/envs/insight/lib/python3.6/site-packages (from pandas>=0.19->pandas_profiling) (2.7.5)\nRequirement already satisfied: pytz>=2011k in /Users/sms/anaconda3/envs/insight/lib/python3.6/site-packages (from pandas>=0.19->pandas_profiling) (2018.7)\nRequirement already satisfied: numpy>=1.9.0 in /Users/sms/.local/lib/python3.6/site-packages (from pandas>=0.19->pandas_profiling) (1.16.0)\nRequirement already satisfied: MarkupSafe>=0.23 in /Users/sms/anaconda3/envs/insight/lib/python3.6/site-packages (from jinja2>=2.8->pandas_profiling) (1.1.0)\nRequirement already satisfied: cycler>=0.10 in /Users/sms/anaconda3/envs/insight/lib/python3.6/site-packages (from matplotlib>=1.4->pandas_profiling) (0.10.0)\nRequirement already satisfied: kiwisolver>=1.0.1 in /Users/sms/anaconda3/envs/insight/lib/python3.6/site-packages (from matplotlib>=1.4->pandas_profiling) (1.0.1)\nRequirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /Users/sms/anaconda3/envs/insight/lib/python3.6/site-packages (from matplotlib>=1.4->pandas_profiling) (2.3.0)\nRequirement already satisfied: setuptools in /Users/sms/anaconda3/envs/insight/lib/python3.6/site-packages (from kiwisolver>=1.0.1->matplotlib>=1.4->pandas_profiling) (40.6.3)\nInstalling collected packages: pandas-profiling\nSuccessfully installed pandas-profiling-1.4.1\n" ], [ "import pandas as pd\nimport numpy as np\nimport math\nfrom scipy.stats import norm\n\nimport pandas_profiling\n\nimport pickle\n\nimport matplotlib.pyplot as plt\n%matplotlib inline\n\nimport warnings\nwarnings.filterwarnings(\"ignore\")", "_____no_output_____" ] ], [ [ "#### Load dataset\n\nAccess https://www.kaggle.com/kumargh/pimaindiansdiabetescsv, download the dataset, it will download pimaindiansdiabetescsv.zip.\n\n\n\n", "_____no_output_____" ], [ "##### Dataset dictionary\n\nFollowing are the datset details as per kaggle.\n\nAbout this file\nThis dataset describes the medical records for Pima Indians and whether or not each patient will have an onset of diabetes within \fve years.\n\nFields description follow:\n\npreg = Number of times pregnant\n\nplas = Plasma glucose concentration a 2 hours in an oral glucose tolerance test\n\npres = Diastolic blood pressure (mm Hg)\n\nskin = Triceps skin fold thickness (mm)\n\ntest = 2-Hour serum insulin (mu U/ml)\n\nmass = Body mass index (weight in kg/(height in m)^2)\n\npedi = Diabetes pedigree function\n\nage = Age (years)\n\nclass = Class variable (1:tested positive for diabetes, 0: tested negative for diabetes)\n\n\n\nColumns, first row is sample data, ignoring to obtain dataset of rows 767, as mentioned in home work assignment link.\n\n6 Pregnancies\n\n148 Glucose\n\n72 BloodPressure\n\n35 SkinThickness\n\n0 Insulin\n\n33.6 BMI\n\n0.627 DiabetesPedigreeFunction\n\n50 Age\n\n1 Class\n\n", "_____no_output_____" ], [ "To access the dataset in Google Colab you can either use Github or Google Drive. We will be accessing dataset via Google Drive. Unzip the pimaindiansdiabetescsv.zip, add pima-indians-diabetes.csv to a known folder in Google Drive, this folder path in drive will be accessed later to load dataset.\n\n\nWe added the pima-indians-diabetes.csv to Google Drive folder /My Drive/UISC-MCS-DS/CS498AML/homework_1/1a/data/.", "_____no_output_____" ], [ "This step is not required if you are using localhost\n* Mount Google Drive to access data Note: This is not required if you are not using Google colab\n\n", "_____no_output_____" ], [ "from google.colab import drive\ndrive.mount('/content/gdrive')", "_____no_output_____" ], [ "Load pima-indians-diabetes.csv Dataset and save it as a pickle object", "_____no_output_____" ] ], [ [ "#pima_indias_diabetes_data = pd.read_csv(\"/content/gdrive/My Drive/UIUC-MCS-DS/CS498AML/homework_1/1a/data/pima-indians-diabetes.csv\")\n#pickle.dump(pima_indias_diabetes_data, open( '/content/gdrive/My Drive/UIUC-MCS-DS/CS498AML/homework_1/1a/data/pima_indias_diabetes_data.pkl','wb'))\n\npima_indias_diabetes_data = pd.read_csv(\"/MS2/academics/MCS-DS-UIUC/Coursera/CS-498-AML/homework 1/1a/data/pima-indians-diabetes.csv\")\npickle.dump(pima_indias_diabetes_data, open('/MS2/academics/MCS-DS-UIUC/Coursera/CS-498-AML/homework 1/1a/data/pima_indias_diabetes_data.pkl','wb'))\n\n\npima_indias_diabetes_data.head()", "_____no_output_____" ] ], [ [ "#### Exploratory Data Analysis\n\n", "_____no_output_____" ], [ "##### Validate dataset", "_____no_output_____" ] ], [ [ "# rename column names\npima_indias_diabetes_data.columns = ['Pregnancies', 'Glucose', 'BloodPressure', 'SkinThickness', 'Insulin', 'BMI', 'DiabetesPedigreeFunction', 'Age', 'Class']\npima_indias_diabetes_data.head()", "_____no_output_____" ], [ "# count number of dataset\nprint(\"There are a total of\", len(pima_indias_diabetes_data),\"data-points\")", "There are a total of 767 data-points\n" ], [ "pima_indias_diabetes_data_features = pima_indias_diabetes_data[['Pregnancies', 'Glucose', 'BloodPressure', 'SkinThickness', 'Insulin', 'BMI', 'DiabetesPedigreeFunction', 'Age']]\npima_indias_diabetes_data_labels = pima_indias_diabetes_data[['Class']]\nprint(pima_indias_diabetes_data_features.shape)\nprint(pima_indias_diabetes_data_labels.shape)", "(767, 8)\n(767, 1)\n" ] ], [ [ "There are 767 observations of 8 different features.\n\n", "_____no_output_____" ], [ "##### Analyse label", "_____no_output_____" ] ], [ [ "count_classes = pd.value_counts(pima_indias_diabetes_data['Class'], sort = True).sort_index()\ncount_classes.plot(kind = 'bar')\nplt.title(\"Pima Indians diabetes class histogram\")\nplt.xlabel(\"Class\")\nplt.ylabel(\"Frequency\")\n\n# 284315 normal transactions (class 0)\n# 492 fraud transactions (class 1)\n\npima_indias_diabetes_data.groupby('Class')['Class'].count()", "_____no_output_____" ] ], [ [ "There are 500 Pima Indians 0: tested negative for diabetes, 267 :tested positive for diabetes. ", "_____no_output_____" ], [ "##### Data distribution analysis for each feature and class label\n\nPlot the data by each feature", "_____no_output_____" ] ], [ [ "axarr = [[]]*len(pima_indias_diabetes_data_features.columns)\ncolumns = 4\nrows = int( np.ceil( len(pima_indias_diabetes_data_features.columns) / columns ) )\nf, fig = plt.subplots( figsize=(columns*3.5, rows*2) )\n\nf.suptitle('Data Distributions by Feature and Class', size=16)\n\nfor i, col in enumerate(pima_indias_diabetes_data_features.columns[:]):\n axarr[i] = plt.subplot2grid( (int(rows), int(columns)), (int(i//columns), int(i%columns)) )\n axarr[i].hist( [ pima_indias_diabetes_data.loc[ pima_indias_diabetes_data.Class == 0, col ], pima_indias_diabetes_data.loc[ pima_indias_diabetes_data.Class == 1, col ] ], label=['tested negative','tested positive'], \n bins=np.linspace( np.percentile(pima_indias_diabetes_data[col],0.1), np.percentile(pima_indias_diabetes_data[col],99.9), 30 ),\n normed=True )\n axarr[i].set_xlabel(col, size=12)\n axarr[i].set_ylim([0,0.8])\n axarr[i].tick_params(axis='both', labelsize=10)\n if i == 0: \n legend = axarr[i].legend()\n legend.get_frame().set_facecolor('white')\n if i%4 != 0 : \n axarr[i].tick_params(axis='y', left='off', labelleft='off')\n else:\n axarr[i].set_ylabel('Fraction',size=12)\n\nplt.tight_layout(rect=[0,0,1,0.95]) # xmin, ymin, xmax, ymax\nplt.show()", "_____no_output_____" ] ], [ [ "#### Classify Dataset - Build a simple naive Bayes classifier ", "_____no_output_____" ], [ "##### Split data\n", "_____no_output_____" ] ], [ [ "\ndef train_test_split(features, labels, test_size=0.2):\n np.random.seed(4)\n id = np.random.rand(len(features))>test_size\n #print(id)\n features_train = features[id]\n labels_train = labels[id]\n features_test = features[np.invert(id)]\n labels_test = labels[np.invert(id)]\n return features_train, labels_train, features_test, labels_test\n\n\nfeatures = pima_indias_diabetes_data.drop('Class', axis = 1)\nlabels = pima_indias_diabetes_data['Class']\n\n\n\n", "_____no_output_____" ] ], [ [ "#### Gaussian Naive Bayes classifier\n\nIn Gaussian Naive Bayes, continuous values associated with each feature are assumed to be distributed according to a Gaussian distribution. A Gaussian distribution is also called Normal distribution. \n\nThe likelihood of the features is assumed to be Gaussian, hence, conditional probability is given by:\n\n >$P(x_i | y) = \\frac{1}{\\sqrt{2\\pi\\sigma _{y}^{2} }} exp \\left (-\\frac{(x_i-\\mu _{y})^2}{2\\sigma _{y}^{2}} \\right ) $\n \n we will use norm.pdf to calculate probablity density function.", "_____no_output_____" ] ], [ [ "class GaussianNaiveBayes():\n \"\"\"The Gaussian Naive Bayes classifier. \"\"\"\n def fit(self,features, labels):\n \"\"\"\n for each label and feature combination we need to calculate the std and mean value from the features & labels.\n \"\"\"\n self.min_std = 0.00000001\n self.ntargets= np.unique(labels).shape[0]\n self.target_labels = np.unique(labels)\n self.nfeatures = features.shape[1]\n self.means = np.zeros((self.ntargets,self.nfeatures))\n self.stds = np.zeros((self.ntargets,self.nfeatures))\n self.priors = np.zeros(self.ntargets)\n for _index in range(self.ntargets):\n # Get the boolean vector to filter for labels = i\n where_label = [label==self.target_labels[_index] for label in labels]\n self.means[_index] = np.nanmean(features[where_label],axis=0)\n # To avoid devide by 0/very small value issue, add a min for standard deviation to min_std = 0.00000001\n self.stds[_index] = np.clip(np.nanstd(features[where_label],axis=0),self.min_std,None)\n #print(self.means[_index], self.stds[_index])\n #Calculate the prior for given label \n self.priors[_index] = np.log(np.sum(where_label)/len(labels))\n \n def predict(self,test_features):\n \"\"\" Classification using Bayes Rule P(Y|X) = P(X|Y)*P(Y)/P(X),\n or Posterior = Likelihood * Prior / Scaling Factor\n P(Y|X) - The posterior is the probability that sample x is of class y given the\n feature values of x being distributed according to distribution of y and the prior.\n P(X|Y) - Likelihood of data X given class distribution Y.\n Gaussian distribution (given by _calculate_likelihood)\n P(Y) - Prior (given by _calculate_prior)\n P(X) - Scales the posterior to make it a proper probability distribution.\n This term is ignored in this implementation since it doesn't affect\n which class distribution the sample is most likely to belong to.\n Classifies the sample as the class that results in the largest P(Y|X) (posterior)\n \"\"\"\n test_samples = test_features.shape[0]\n posterior = np.zeros((test_samples,self.ntargets))\n # Naive assumption (independence):\n # P(x1,x2,x3|Y) = P(x1|Y)*P(x2|Y)*P(x3|Y)\n # Posterior is product of prior and likelihoods (ignoring scaling factor)\n for target_label in range(self.ntargets):\n posterior[:,target_label] = self.priors[target_label] + np.nansum(np.log(norm.pdf(test_features,self.means[target_label],self.stds[target_label])),axis=1)\n label = self.target_labels[np.argmax(posterior, axis=1)]\n return label \n \n \n def score(self,X_test, y_test):\n y_predict = self.predict(X_test)\n return (y_predict == y_test).mean()", "_____no_output_____" ] ], [ [ "Compute an estimate of the accuracy of the classifier by averaging over 10 test-train splits. Each split should randomly assign 20% of the data to test, and the rest to train.", "_____no_output_____" ] ], [ [ "print(features.shape)\n\n\n\ntest_accuracy_iterations =[]\n# for 10 iterations\nfor i in range(10):\n features_train, labels_train, features_test, labels_test = train_test_split(features.values, \n labels.values, test_size=0.2)\n nb = GaussianNaiveBayes()\n\n nb.fit(features_train, labels_train)\n test_accuracy = nb.score(features_test, labels_test)\n test_accuracy_iterations.append(test_accuracy)\n\nprint(test_accuracy_iterations)\nprint(\"Average test accuracy\", np.mean(test_accuracy_iterations))", "(767, 8)\n[0.7724550898203593, 0.7724550898203593, 0.7724550898203593, 0.7724550898203593, 0.7724550898203593, 0.7724550898203593, 0.7724550898203593, 0.7724550898203593, 0.7724550898203593, 0.7724550898203593]\nAverage test accuracy 0.7724550898203593\n" ] ], [ [ "Validate using scikit learn naive bayes", "_____no_output_____" ] ], [ [ "from sklearn import naive_bayes\n\ntest_accuracy_iterations =[]\n# for 10 iterations\nfor i in range(10):\n features_train, labels_train, features_test, labels_test = train_test_split(features.values, labels.values, test_size=0.2)\n snb = naive_bayes.GaussianNB()\n\n snb.fit(features_train, labels_train)\n test_accuracy = snb.score(features_test, labels_test)\n test_accuracy_iterations.append(test_accuracy)\n\nprint(test_accuracy_iterations)\nprint(\"Average test accuracy\", np.mean(test_accuracy_iterations))", "[0.7724550898203593, 0.7724550898203593, 0.7724550898203593, 0.7724550898203593, 0.7724550898203593, 0.7724550898203593, 0.7724550898203593, 0.7724550898203593, 0.7724550898203593, 0.7724550898203593]\nAverage test accuracy 0.7724550898203593\n" ] ], [ [ "## Part 1B\n\nNow adjust your code so that, for attribute 3 (Diastolic blood pressure), attribute 4 (Triceps skinfold thickness), attribute 6 (Body mass index), and attribute 8 (Age), it regards a value of 0 as a missing value when estimating the class-conditional distributions, and the posterior. \n\n* Compute an estimate of the accuracy of the classifier by averaging over 10 test-train splits.\n\n\n", "_____no_output_____" ], [ "### Answer Part 1B\n\nAll the above work done for Part 1A will be reused for Part 1B.\n\nWe will mainly be processing data to remove missing values which are 0.", "_____no_output_____" ], [ "Impute missing values 0 for attributes 3 (Diastolic blood pressure), attribute 4 (Triceps skinfold thickness), attribute 6 (Body mass index), and attribute 8 (Age) as np.NaN.\n\nCheck how many missing values are present.", "_____no_output_____" ] ], [ [ "pima_indias_diabetes_data[['BloodPressure','SkinThickness','BMI','Age']]= pima_indias_diabetes_data[['BloodPressure','SkinThickness','BMI','Age']].replace(0, np.NaN)\n\nprint(pima_indias_diabetes_data.isnull().sum())", "Pregnancies 0\nGlucose 0\nBloodPressure 35\nSkinThickness 227\nInsulin 0\nBMI 11\nDiabetesPedigreeFunction 0\nAge 0\nClass 0\ndtype: int64\n" ], [ "#pima_indias_diabetes_data.dropna(inplace=True)\n\nprint(pima_indias_diabetes_data.isnull().sum())\n", "Pregnancies 0\nGlucose 0\nBloodPressure 35\nSkinThickness 227\nInsulin 0\nBMI 11\nDiabetesPedigreeFunction 0\nAge 0\nClass 0\ndtype: int64\n" ], [ "pima_indias_diabetes_data.shape", "_____no_output_____" ] ], [ [ "After processing the missing data split data and build new model with new train dataset and test he accuracy.", "_____no_output_____" ] ], [ [ "processed_features = pima_indias_diabetes_data.drop('Class', axis = 1)\nprocessed_labels = pima_indias_diabetes_data['Class']\n\n#scaler = StandardScaler()\n#normal_processed_features = scaler.fit_transform(processed_features)", "_____no_output_____" ], [ "processed_test_accuracy_iterations =[]\n# for 10 iterations\nfor i in range(10):\n p_features_train, p_labels_train, p_features_test, p_labels_test = train_test_split(processed_features.values, processed_labels.values, test_size=0.2)\n p_nb = GaussianNaiveBayes()\n\n p_nb.fit(p_features_train, p_labels_train)\n p_test_accuracy = p_nb.score(p_features_test, p_labels_test)\n processed_test_accuracy_iterations.append(p_test_accuracy)\n\nprint(processed_test_accuracy_iterations)\nprint(\"Average test accuracy for processed data\", np.mean(processed_test_accuracy_iterations))", "[0.7425149700598802, 0.7425149700598802, 0.7425149700598802, 0.7425149700598802, 0.7425149700598802, 0.7425149700598802, 0.7425149700598802, 0.7425149700598802, 0.7425149700598802, 0.7425149700598802]\nAverage test accuracy for processed data 0.7425149700598802\n" ] ], [ [ "## References\nFollowing are various resources referred while writing this solution\n\n* Github projects https://github.com/sriharshams/mlnd/\n* Code of codyznash https://github.com/codyznash/GANs_for_Credit_Card_Data\n* Tutorials of https://machinelearningmastery.com/naive-bayes-classifier-scratch-python/\n* [Naive Bayes in Scikit-Learn](https://github.com/scikit-learn/scikit-learn/blob/7389dba/sklearn/naive_bayes.py#L107): Implementation of naive bayes in the scikit-learn library.\n* [ ML from scratch] (https://github.com/eriklindernoren/ML-From-Scratch)\n* [Naive Bayes documentation](https://scikit-learn.org/stable/modules/naive_bayes.html): Scikit-Learn documentation and sample code for Naive Bayes\n* Naive Bayes Classifiers https://www.geeksforgeeks.org/naive-bayes-classifiers/\n* [Naive Bayes Classifier From Scratch](https://chrisalbon.com/machine_learning/naive_bayes/naive_bayes_classifier_from_scratch/)\n* https://chrisalbon.com/machine_learning/naive_bayes/naive_bayes_classifier_from_scratch/\n* Naive Bayes from scratch in python http://kenzotakahashi.github.io/naive-bayes-from-scratch-in-python.html\n* [Applied Machine Learning, D.A. Forsyth, (approximate 18'th draft)](http://luthuli.cs.uiuc.edu/~daf/courses/AML-18-Fall/AMLbook-3-Dec-18.pdf)\n* Piazza & Slack discussions on CS-498 Spring 2019\n\n\n\n", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "%load_ext autoreload\n%autoreload 2\n%matplotlib inline", "_____no_output_____" ], [ "import sys\nimport pathlib", "_____no_output_____" ], [ "try:\n import matplotlib_views as views\nexcept ModuleNotFoundError:\n cwd = pathlib.Path().resolve().parent\n sys.path.append(str(cwd))\n import matplotlib_views as views", "_____no_output_____" ] ], [ [ "## Create some fake data", "_____no_output_____" ] ] ]

[ "code", "markdown" ]

[ [ "code", "code", "code" ], [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import pandas as pd\nimport numpy as np\nfrom glob import glob", "_____no_output_____" ], [ "# Lista arquivos CSV:\ndataPath = '/home/skems/gabinete/dados/siconv/siconv/'\ndataFiles = glob(dataPath+'/*.csv')\ndataFiles.sort()", "_____no_output_____" ], [ "#Carrega um arquivo:\ndata = pd.read_csv(dataFiles[1], sep=';')", "_____no_output_____" ], [ "data", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Understanding Transient Transport Settings", "_____no_output_____" ] ] ]

[ "markdown" ]

[ [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "<a rel=\"license\" href=\"http://creativecommons.org/licenses/by/4.0/\"><img alt=\"Creative Commons License\" style=\"border-width:0\" src=\"https://i.creativecommons.org/l/by/4.0/88x31.png\" /></a><br /><span xmlns:dct=\"http://purl.org/dc/terms/\" property=\"dct:title\"><b>The Knapsack Problem</b></span> by <a xmlns:cc=\"http://creativecommons.org/ns#\" href=\"http://mate.unipv.it/gualandi\" property=\"cc:attributionName\" rel=\"cc:attributionURL\">Stefano Gualandi</a> is licensed under a <a rel=\"license\" href=\"http://creativecommons.org/licenses/by/4.0/\">Creative Commons Attribution 4.0 International License</a>.<br />Based on a work at <a xmlns:dct=\"http://purl.org/dc/terms/\" href=\"https://github.com/mathcoding/opt4ds\" rel=\"dct:source\">https://github.com/mathcoding/opt4ds</a>.", "_____no_output_____" ], [ "**NOTE:** Run the following script whenever running this script on a Google Colab.", "_____no_output_____" ] ], [ [ "import shutil\nimport sys\nimport os.path\n\nif not shutil.which(\"pyomo\"):\n !pip install -q pyomo\n assert(shutil.which(\"pyomo\"))\n\nif not (shutil.which(\"glpk\") or os.path.isfile(\"glpk\")):\n if \"google.colab\" in sys.modules:\n !apt-get install -y -qq glpk-utils\n else:\n try:\n !conda install -c conda-forge glpk \n except:\n pass", "_____no_output_____" ] ], [ [ "# $n$-Queens Problem\nThe $n$-Queens puzzle is the problem of placing eight chess queens on an $n \\times n$ chessboard so that no two queens threaten each other; thus, a solution requires that no two queens share the same row, column, or diagonal (source: [wikipedia](https://en.wikipedia.org/wiki/Eight_queens_puzzle)). \n\nA solution exists for all natural numbers n with the exception of $n = 2$ and $n = 3$.", "_____no_output_____" ], [ " **Example:** For $n=8$, we have the following solution:\n\n```\n1 . . . . . Q . .\n2 . . . Q . . . . \n3 . . . . . . Q .\n4 Q . . . . . . .\n5 . . . . . . . Q\n6 . Q . . . . . .\n7 . . . . Q . . .\n8 . . Q . . . . .\n a b c d e f g h \n```", "_____no_output_____" ], [ "## Integer Linear Programming Model\nThe $n$-Queens problem can be formalized with the following **ILP** model.\n\n**Data:** Size of the board $n\\times n$. Let $I=:\\{1,\\dots,n\\}$ a set of indices.\n\n**Decision Variables:** The variable $x_{ij} \\in \\{0,1\\}$ is equal to 1 if we place a queen in position $(i,j)$ on the chessboard.\n\n**Objective function:** Since the problem is a feasibility problem, we can set the objective function equal to any constant value.\n\n**Constraints:** We need the following linear constraints, which encode the puzzle rules:\n\n1. Each queen appears once per row:\n$$\n \\sum_{j \\in I} x_{ij} = 1, \\forall i \\in I\n$$\n2. Each queen appears once per column:\n$$\n \\sum_{i \\in I} x_{ij} = 1, \\forall j \\in I\n$$\n3. Each queen appears once per main diagonals:\n$$\n \\sum_{(i,j) \\in D_k} x_{ij} \\leq 1, D_k \\mbox{ main diagonals}\n$$\n4. Each queen appears once per off-diagonals:\n$$\n \\sum_{(i,j) \\in O_k} x_{ij} \\leq 1, O_k \\mbox{ off diagonals}\n$$", "_____no_output_____" ], [ "### Main Diagonals $D_k$\nSince we need to specify the pairs of indices that define as a function of $n$, we first defined the following nested loop:", "_____no_output_____" ] ], [ [ "n = 5\nfor j in range(-n+2,n-1):\n for i in range(1, n+1):\n if 0 < j+i <= n:\n print(i, j+i, end='\\t')\n else:\n print(' ', end='\\t')\n print()", "_____no_output_____" ] ], [ [ "### Off Diagonals $_k$\nSimilarly, we can define the off diagonals as follows:", "_____no_output_____" ] ], [ [ "for i in reversed(range(-n+3, n)):\n for j in range(1, n):\n if 0 < n - j+i <= n:\n print(j, n-j+i, end='\\t')\n else:\n print(' ', end='\\t')\n print()", "_____no_output_____" ] ], [ [ "### Full Model defined in Pyomo\nIf we put all the definitions together, we can solve the $n$-Queens problem with the script below.\n\nPlease, note the following Pyomo syntax used to define variable $x_{ij}$ over the [RangeSet](https://pyomo.readthedocs.io/en/stable/library_reference/aml/index.html#pyomo.environ.RangeSet) $I$ and $J$:\n\n```\nmodel.I = RangeSet(1, n)\nmodel.J = RangeSet(1, n)\nmodel.x = Var(model.I, model.J, within=Binary)\n```\n\nNotice also the syntax used to define the row and column constraints, which uses `lambda` function to define constraint rules:\n\n```\nmodel.row = Constraint(model.I, \n rule = lambda mod, i: sum(mod.x[i,j] for j in mod.J) == 1)\n```\n\nFinally, to define the main and of diagonals, we use the [ConstraintList](https://pyomo.readthedocs.io/en/stable/working_models.html) class:\n\n```\nmodel.mainD = ConstraintList()\n#...\n model.mainD.add( expr <= 1 )\n```\n\nThe complete Pyomo script is as follows.", "_____no_output_____" ] ], [ [ "# Import the libraries\nfrom pyomo.environ import ConcreteModel, Var, Objective, Constraint, SolverFactory\nfrom pyomo.environ import maximize, Binary, RangeSet, ConstraintList\n\n\nn = 8\n\n# Create concrete model\nmodel = ConcreteModel()\n\nmodel.I = RangeSet(1, n)\nmodel.J = RangeSet(1, n)\n\n# Variables\nmodel.x = Var(model.I, model.J, within=Binary)\n\n# Objective Function: Maximize Profit\nmodel.obj = Objective(expr = n, sense = maximize)\n\n# 1. Row constraints\ndef VincoloRighe(mod, i):\n return sum(mod.x[i,j] for j in mod.J) == 1\n\nmodel.row = Constraint(model.I, \n rule = VincoloRighe)\n\n# 2. Column constraints\nmodel.column = Constraint(model.J, \n rule = lambda mod, j: sum(mod.x[i,j] for i in mod.I) == 1)\n\n# 3. Main Diagonal constraints\nmodel.mainD = ConstraintList()\n# Build the list of possible pairs\nfor j in range(-n+2,n-1):\n expr = 0\n for i in model.I:\n if 0 < j+i <= n:\n expr += model.x[i, j+i] \n model.mainD.add( expr <= 1 )\n\n# 4. Off Diagonal constraints\nmodel.offD = ConstraintList()\n# Build the list of possible pairs\nfor i in range(-n+3,n+1):\n expr = 0\n for j in model.J:\n if 0 < n-j+i <= n:\n expr += model.x[j, n-j+i] \n model.offD.add( expr <= 1 )", "_____no_output_____" ] ], [ [ "To solve the script, we use a solver factory, specifying the GLPK solver, and we inspect the Solver **status** (infeasible, unbounded, or optimal).", "_____no_output_____" ] ], [ [ "# Solve the model\nsol = SolverFactory('glpk').solve(model)\n\n# Basic info about the solution process\nfor info in sol['Solver']:\n print(info)", "_____no_output_____" ] ], [ [ "We aspect the optimal decision variables (only the positive variables).", "_____no_output_____" ] ], [ [ "# Report solution value\nprint(\"Optimal solution value: z =\", model.obj())\nprint(\"Decision variables:\")\nfor i in model.I:\n for j in model.J:\n if model.x[i,j]() > 0:\n print(\"x({},{}) = {}\".format(i, j, model.x[i,j]()))", "_____no_output_____" ] ], [ [ "And finally, we print a solution on a simplified chessboard $n\\times n$.", "_____no_output_____" ] ], [ [ "print('\\nChessboard Solution:')\nfor i in model.I:\n for j in model.J:\n if model.x[i,j]() > 0:\n print('Q', end=' ')\n else:\n print('.', end=' ')\n print()", "_____no_output_____" ] ], [ [ "## Plotting a solution with a Chessboard", "_____no_output_____" ] ], [ [ "# CREDIT: Solution original appeared on Stackoverflow at:\n# https://stackoverflow.com/questions/60608055/insert-queen-on-a-chessboard-with-pyplot\n\ndef PlotSolution(n, x, size=6):\n import matplotlib.pyplot as plt\n import numpy as np\n\n chessboard = np.zeros((n, n))\n\n chessboard[1::2,0::2] = 1\n chessboard[0::2,1::2] = 1\n\n plt.figure(figsize=(size, size))\n plt.imshow(chessboard, cmap='binary')\n\n for i, j in x:\n if x[i,j]() > 0:\n plt.text(i-1, j-1, '♕', color='darkorange',\n fontsize=56*size/n, fontweight='bold', ha='center', va='center')\n plt.xticks([])\n plt.yticks([])\n plt.show()", "_____no_output_____" ], [ "PlotSolution(n, model.x)", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "**Title**: Upload kaggle chest X-Ray. \n**Date**: 12-Oct-2020 \n**Description**: \nPneumonia accounts for over 15% of all deaths of children under 5 years old internationally. Advanced detection of pneumonia could save thousands of lives a year.\n\nIn 2018 the RSNA Pneumonia Detection Challenge was posted on Kaggle, an organization for machine learning training and purpose-driven competitions in Data Science.\n\nThis notebook downloads the entire RSNA Pneumonia Detection Challenge Dataset (3.6 GB) and incorporates it into a Flywheel instance specified by the supplied API-Key. A Data Use Agreement (DUA) is required to download this dataset.\n\nReference:\n* https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/data ", "_____no_output_____" ], [ "# Data Use Aggreement\nBefore downloading this data, or any data, from kaggle, you must agree to the rules of this competition: \n\n* https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/rules", "_____no_output_____" ] ], [ [ "%reload_ext autoreload\n%autoreload 2 \n%matplotlib inline", "_____no_output_____" ] ], [ [ "# Requirements:\n- **Python** (Preferably >= 3.6): \n\n- Have admin permissions to create Flywheel Groups and Projects.", "_____no_output_____" ], [ "# Install and import dependencies", "_____no_output_____" ] ], [ [ "!pip install pandas pydicom flywheel-sdk tqdm kaggle jupyter ipywidgets", "_____no_output_____" ], [ "import json\nimport logging\nimport os\nimport re\nimport time\nimport zipfile\nfrom getpass import getpass\nfrom pathlib import Path\n\nimport flywheel\nimport pandas as pd\nimport pydicom\nfrom tqdm.notebook import tqdm", "_____no_output_____" ], [ "# Instantiate a logger\nlogging.basicConfig(level=logging.INFO, format='%(asctime)s %(levelname)s %(message)s')\nlog = logging.getLogger('root')", "_____no_output_____" ] ], [ [ "# Download kaggle dataset\n\nThis requires that you have stored your Kaggle credentials in ~/.kaggle/kaggle.json. These can be acquired by creating a kaggle account at kaggle.com and using \"Create New API Token\" on the user account page. \n\nThis dataset is currently 3.7 GB and may change in the future. Depending on the bandwidth of your internet connection, this may take some time to download.", "_____no_output_____" ] ], [ [ "!kaggle competitions download -c rsna-pneumonia-detection-challenge", "_____no_output_____" ] ], [ [ "# Initialize Constants\nInitialize path to dowload directory, default session label, and default acquisition label.", "_____no_output_____" ] ], [ [ "ROOT_KAGGLE_DATA = '/path/to/repository/rsna-pneumonia-detection-challenge'\nDEFAULT_SESSION_LABEL = 'NA'\nDEFAULT_ACQ_LABEL = 'Chest XR'", "_____no_output_____" ] ], [ [ "# Flywheel API Key and Client\nGet an API_KEY. More on this in the Flywheel SDK doc [here](https://flywheel-io.gitlab.io/product/backend/sdk/branches/master/python/getting_started.html#api-key).", "_____no_output_____" ] ], [ [ "API_KEY = getpass('Enter API_KEY here: ')", "_____no_output_____" ] ], [ [ "Instantiate the Flywheel API client", "_____no_output_____" ] ], [ [ "fw_client = flywheel.Client(API_KEY if 'API_KEY' in locals() else os.environ.get('FW_KEY'))", "_____no_output_____" ] ], [ [ "Show Flywheel logging information", "_____no_output_____" ] ], [ [ "log.info('You are now logged in as %s to %s', fw_client.get_current_user()['email'], fw_client.get_config()['site']['api_url'])", "_____no_output_____" ] ], [ [ "# Read the csv\nThe CSV file consists of the patient id, whether the pnemonia was diagnosed (Target 0/1), and the rectangular region of the image it was found in (x,y,width,height).\n\n```\npatientId,x,y,width,height,Target\n0004cfab-14fd-4e49-80ba-63a80b6bddd6,,,,,0\n00436515-870c-4b36-a041-de91049b9ab4,264.0,152.0,213.0,379.0,1\n```", "_____no_output_____" ] ], [ [ "df = pd.read_csv(Path(ROOT_KAGGLE_DATA) / 'stage_2_train_labels.csv')", "_____no_output_____" ] ], [ [ "# Container helpers\nImport container helper functions to find existing or create new containers.", "_____no_output_____" ] ], [ [ "from container_helpers import (\n find_or_create_group, \n find_or_create_project, \n find_or_create_subject, \n find_or_create_session, \n find_or_create_acquisition,\n upload_file_to_acquisition\n)", "_____no_output_____" ] ], [ [ "# Create the project", "_____no_output_____" ] ], [ [ "# Initialize the group\npublic_data_group = find_or_create_group(fw_client, 'public_data', 'public_data')\n# Initialize the project\nproject_label = 'kaggle-rsna-pneumonia-detection-challenge'\nreadme = 'https://www.kaggle.com/c/rsna-pneumonia-detection-challenge/data'\nchestxray_project = find_or_create_project(project_label, public_data_group)\nif chestxray_project:\n chestxray_project.update(description=readme)", "_____no_output_____" ] ], [ [ "# Iterate through dataframe and upload\nIterate through the training data csv to create the container hierarchy for this project:\n\n1. find or create each subject encountered\n\n a. Encode presence/absence of pneumonia (Target=0/1) and the rectangular region it was found in (box) into a dictionary.\n2. find or create each session (with `DEFAULT_SESSION_LABEL`) encountered\n3. find or create each acquisition (with 'SeriesDescription' or `DEFAULT_ACQ_LABEL`) and add enclosed files.\n\n a. Incorporate presence/absence of pneumonia (Target) and--if found--the rectangular region it was found in (box) into the metadata of the acquisition file.\n", "_____no_output_____" ] ], [ [ "for i, row in tqdm(df.iterrows(), total=len(df)):\n log.info('Processing Subject %s.', row['patientId'])\n # (1) Find or create subject\n subject = find_or_create_subject(row['patientId'], chestxray_project)\n # (1a) Encode pneumonia status and rectangular region of positive status in dictionary.\n if row['Target']:\n row_dict = {\n 'box': {\n 'x': row['x'], \n 'y': row['y'], \n 'width': row['width'], \n 'height': row['height']\n }, \n 'Target': row['Target']\n }\n else:\n row_dict = {'Target': row['Target']}\n if subject:\n log.info('Processing Session %s.', DEFAULT_SESSION_LABEL)\n # (2) Find or create session \n session = find_or_create_session(DEFAULT_SESSION_LABEL, subject)\n if session:\n filepath = str(Path(ROOT_KAGGLE_DATA) / 'stage_2_train_images' / f\"{row['patientId']}.dcm\")\n dcm = pydicom.read_file(filepath, stop_before_pixels=True, force=True)\n # Pack dicoms into zip file\n with zipfile.ZipFile(f'/tmp/{row[\"patientId\"]}.zip', 'w') as myzip:\n myzip.write(filepath)\n\n acq_label = dcm.get('SeriesDescription', DEFAULT_ACQ_LABEL)\n log.info('Processing Acquisition %s.', acq_label)\n # (3) Find or create acquisition\n acq = find_or_create_acquisition(acq_label, session)\n log.info(\n 'Uploading file, %s, to acquisition, %s',\n f'/tmp/{row[\"patientId\"]}.zip',\n acq.label\n )\n kwarg_dict = {\"type\": \"dicom\", \"modality\": \"X-ray\"}\n kwarg_dict[\"info\"] = row_dict\n # Upload file to acquisition and\n # (3a) incorporate Target and box into file metadata\n upload_file_to_acquisition(acq, f'/tmp/{row[\"patientId\"]}.zip', **kwarg_dict)\n # remove temporary zipped dicom file\n os.remove(f'/tmp/{row[\"patientId\"]}.zip')", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "#import libraries\nimport requests\nfrom datetime import datetime as dt\nfrom datetime import timedelta\nimport pandas as pd\n", "_____no_output_____" ], [ "# get events from n days ago\niso = 887 #Pike I changed this to Yemen\nlimit = 400\n\napi_url = 'https://api.acleddata.com/acled/read?terms=accept&iso={}'.format(iso)\nprint (api_url, type(api_url)) #creates request according to ACLED format specifications - p. 13", "https://api.acleddata.com/acled/read?terms=accept&iso=887 <class 'str'>\n" ], [ "response = requests.get(api_url)\ndata = response.json()\ndata.keys()\ndata['count'] ", "_____no_output_____" ] ], [ [ "### From the documentation we know this is the max return --- How can we get all the results?", "_____no_output_____" ] ], [ [ "# Let's mkae a function that updates our search to get the new pages\n\ndef ping_acled(api_url): \n '''\n Takes one parameter search term for API\n '''\n \n response = requests.get(api_url)\n data = response.json()\n return data\n ", "_____no_output_____" ], [ "\nresults = [] # empty data strcture to store results\nnum_results = 500 # condition to continue adding pages\ncount = 0 # tracker of results\npage = 1 #Per the documentation each page will give us more results\n\nwhile num_results == 500: #if less 500 or 0 we know we have all the results\n print (\"starting \", page, \" \", num_results) #just to see our progress\n api_url = 'https://api.acleddata.com/acled/read?terms=accept&iso={}&page={}'.format(iso,page) #the search\n data = ping_acled(api_url) #call the previous function \n results.append(data['data']) #store in our results\n count += data['count'] #Track number of results\n num_results = data['count'] #update our condition\n page += 1 #update our page variable\n print (\"Total Results \", count) #Track our progress\n #Pike this is a bit intimidating I know Yemen is a mess but damn\n #I was worried that the 58000 entries would break the code Does that ever happen\n \n\n", "starting 1 500\nTotal Results 500\nstarting 2 500\nTotal Results 1000\nstarting 3 500\nTotal Results 1500\nstarting 4 500\nTotal Results 2000\nstarting 5 500\nTotal Results 2500\nstarting 6 500\nTotal Results 3000\nstarting 7 500\nTotal Results 3500\nstarting 8 500\nTotal Results 4000\nstarting 9 500\nTotal Results 4500\nstarting 10 500\nTotal Results 5000\nstarting 11 500\nTotal Results 5500\nstarting 12 500\nTotal Results 6000\nstarting 13 500\nTotal Results 6500\nstarting 14 500\nTotal Results 7000\nstarting 15 500\nTotal Results 7500\nstarting 16 500\nTotal Results 8000\nstarting 17 500\nTotal Results 8500\nstarting 18 500\nTotal Results 9000\nstarting 19 500\nTotal Results 9500\nstarting 20 500\nTotal Results 10000\nstarting 21 500\nTotal Results 10500\nstarting 22 500\nTotal Results 11000\nstarting 23 500\nTotal Results 11500\nstarting 24 500\nTotal Results 12000\nstarting 25 500\nTotal Results 12500\nstarting 26 500\nTotal Results 13000\nstarting 27 500\nTotal Results 13500\nstarting 28 500\nTotal Results 14000\nstarting 29 500\nTotal Results 14500\nstarting 30 500\nTotal Results 15000\nstarting 31 500\nTotal Results 15500\nstarting 32 500\nTotal Results 16000\nstarting 33 500\nTotal Results 16500\nstarting 34 500\nTotal Results 17000\nstarting 35 500\nTotal Results 17500\nstarting 36 500\nTotal Results 18000\nstarting 37 500\nTotal Results 18500\nstarting 38 500\nTotal Results 19000\nstarting 39 500\nTotal Results 19500\nstarting 40 500\nTotal Results 20000\nstarting 41 500\nTotal Results 20500\nstarting 42 500\nTotal Results 21000\nstarting 43 500\nTotal Results 21500\nstarting 44 500\nTotal Results 22000\nstarting 45 500\nTotal Results 22500\nstarting 46 500\nTotal Results 23000\nstarting 47 500\nTotal Results 23500\nstarting 48 500\nTotal Results 24000\nstarting 49 500\nTotal Results 24500\nstarting 50 500\nTotal Results 25000\nstarting 51 500\nTotal Results 25500\nstarting 52 500\nTotal Results 26000\nstarting 53 500\nTotal Results 26500\nstarting 54 500\nTotal Results 27000\nstarting 55 500\nTotal Results 27500\nstarting 56 500\nTotal Results 28000\nstarting 57 500\nTotal Results 28500\nstarting 58 500\nTotal Results 29000\nstarting 59 500\nTotal Results 29500\nstarting 60 500\nTotal Results 30000\nstarting 61 500\nTotal Results 30500\nstarting 62 500\nTotal Results 31000\nstarting 63 500\nTotal Results 31500\nstarting 64 500\nTotal Results 32000\nstarting 65 500\nTotal Results 32500\nstarting 66 500\nTotal Results 33000\nstarting 67 500\nTotal Results 33500\nstarting 68 500\nTotal Results 34000\nstarting 69 500\nTotal Results 34500\nstarting 70 500\nTotal Results 35000\nstarting 71 500\nTotal Results 35500\nstarting 72 500\nTotal Results 36000\nstarting 73 500\nTotal Results 36500\nstarting 74 500\nTotal Results 37000\nstarting 75 500\nTotal Results 37500\nstarting 76 500\nTotal Results 38000\nstarting 77 500\nTotal Results 38500\nstarting 78 500\nTotal Results 39000\nstarting 79 500\nTotal Results 39500\nstarting 80 500\nTotal Results 40000\nstarting 81 500\nTotal Results 40500\nstarting 82 500\nTotal Results 41000\nstarting 83 500\nTotal Results 41500\nstarting 84 500\nTotal Results 42000\nstarting 85 500\nTotal Results 42500\nstarting 86 500\nTotal Results 43000\nstarting 87 500\nTotal Results 43500\nstarting 88 500\nTotal Results 44000\nstarting 89 500\nTotal Results 44500\nstarting 90 500\nTotal Results 45000\nstarting 91 500\nTotal Results 45500\nstarting 92 500\nTotal Results 46000\nstarting 93 500\nTotal Results 46500\nstarting 94 500\nTotal Results 47000\nstarting 95 500\nTotal Results 47500\nstarting 96 500\nTotal Results 48000\nstarting 97 500\nTotal Results 48500\nstarting 98 500\nTotal Results 49000\nstarting 99 500\nTotal Results 49500\nstarting 100 500\nTotal Results 50000\nstarting 101 500\nTotal Results 50500\nstarting 102 500\nTotal Results 51000\nstarting 103 500\nTotal Results 51500\nstarting 104 500\nTotal Results 52000\nstarting 105 500\nTotal Results 52500\nstarting 106 500\nTotal Results 53000\nstarting 107 500\nTotal Results 53500\nstarting 108 500\nTotal Results 54000\nstarting 109 500\nTotal Results 54500\nstarting 110 500\nTotal Results 55000\nstarting 111 500\nTotal Results 55500\nstarting 112 500\nTotal Results 56000\nstarting 113 500\nTotal Results 56500\nstarting 114 500\nTotal Results 57000\nstarting 115 500\nTotal Results 57500\nstarting 116 500\nTotal Results 57820\n" ], [ "#Now I want to put them together into one giant result\nsuper_list = []\nfor res in results: \n super_list += res\n print (len(super_list))\n #This is a giant result", "500\n1000\n1500\n2000\n2500\n3000\n3500\n4000\n4500\n5000\n5500\n6000\n6500\n7000\n7500\n8000\n8500\n9000\n9500\n10000\n10500\n11000\n11500\n12000\n12500\n13000\n13500\n14000\n14500\n15000\n15500\n16000\n16500\n17000\n17500\n18000\n18500\n19000\n19500\n20000\n20500\n21000\n21500\n22000\n22500\n23000\n23500\n24000\n24500\n25000\n25500\n26000\n26500\n27000\n27500\n28000\n28500\n29000\n29500\n30000\n30500\n31000\n31500\n32000\n32500\n33000\n33500\n34000\n34500\n35000\n35500\n36000\n36500\n37000\n37500\n38000\n38500\n39000\n39500\n40000\n40500\n41000\n41500\n42000\n42500\n43000\n43500\n44000\n44500\n45000\n45500\n46000\n46500\n47000\n47500\n48000\n48500\n49000\n49500\n50000\n50500\n51000\n51500\n52000\n52500\n53000\n53500\n54000\n54500\n55000\n55500\n56000\n56500\n57000\n57500\n57820\n" ], [ "#Pike changed the name to reflect the proper country\nyemen_res = pd.DataFrame(super_list)\nyemen_res.head()", "_____no_output_____" ] ], [ [ "### Do the right thing, take some time to look at the codebook and see what these columns are", "_____no_output_____" ] ], [ [ "yemen_res.columns\n", "_____no_output_____" ] ], [ [ "### Homework --- Make a map of some ACLED Data (absolutely use the code from the Global Terrorism Database exercise)", "_____no_output_____" ] ], [ [ "from bokeh.tile_providers import get_provider, Vendors \nfrom pyproj import Transformer\ntile_provider =get_provider('STAMEN_TERRAIN')\nimport math\nfrom bokeh.plotting import figure, output_notebook, show\n#Pike got help from my coworker Brandon Mohr on this\n#Will be borrowing this technique for other mapping projects \n#as I like this better than the GTD method", "_____no_output_____" ], [ "yemen_map = yemen_res[[\"latitude\", 'longitude', 'data_id']]", "_____no_output_____" ], [ "yemen_map['latitude'] = yemen_map['latitude'].astype(float)\nyemen_map['longitude'] = yemen_map['longitude'].astype(float)\nyemen_map['data_id'] = yemen_map['data_id'].astype(float)\n#Pike the conversion fix you described in class", "<ipython-input-33-268010bd8d27>:1: SettingWithCopyWarning: \nA value is trying to be set on a copy of a slice from a DataFrame.\nTry using .loc[row_indexer,col_indexer] = value instead\n\nSee the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n yemen_map['latitude'] = yemen_map['latitude'].astype(float)\n<ipython-input-33-268010bd8d27>:2: SettingWithCopyWarning: \nA value is trying to be set on a copy of a slice from a DataFrame.\nTry using .loc[row_indexer,col_indexer] = value instead\n\nSee the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n yemen_map['longitude'] = yemen_map['longitude'].astype(float)\n<ipython-input-33-268010bd8d27>:3: SettingWithCopyWarning: \nA value is trying to be set on a copy of a slice from a DataFrame.\nTry using .loc[row_indexer,col_indexer] = value instead\n\nSee the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n yemen_map['data_id'] = yemen_map['data_id'].astype(float)\n" ], [ "transformer = Transformer.from_crs('epsg:4326','epsg:3857')", "_____no_output_____" ], [ "map_dict = {}\nnan_count = {}\n\nfor idx, row in yemen_map.iterrows():\n if row['data_id'] in map_dict.keys():\n if math.isnan(row[\"latitude\"]):\n if row['data_id'] in nan_count.keys():\n nan_count[row['data_id']] += 1\n else:\n nan_count[row['data_id']] = 1\n \n else:\n point = transformer.transform(row[\"latitude\"],row[\"longitude\"])\n map_dict[row['data_id']].append([point[0],point[1]])\n \n else:\n if math.isnan(row[\"latitude\"]):\n nan_count[row['data_id']] = 1\n else:\n point = transformer.transform(row[\"latitude\"],row[\"longitude\"])\n map_dict[row['data_id']] =[[point[0],point[1]]]\n#Pike this portion of the code is essentially unchanged save for data id for gname \nnan_count", "_____no_output_____" ] ], [ [ "for idx, row in yemen_map.iterrows():\n if row['data_id'] in map_dict.keys():\n if math.isnan(row[\"latitude\"]):\n if row['data_id'] in nan_count.keys():\n nan_count[row['data_id']] += 1\n else:\n nan_count[row['data_id']] = 1\n else:\n point - transformer.transform(row[\"latitude\"],row{\"longitude\"])\n map_dict[row['data_id']].append([point[0],point[1]])", "_____no_output_____" ] ], [ [ "pts = [(19.07, 41.68), (12.01, 54.87)]\nbbox = []\nfor pt in transformer.itransform(pts):\n bbox.append(pt)\n \n#Pike included the Red Sea Bab al Mandab strait and island of Socotra for completeness sake", "_____no_output_____" ], [ "NPA_x = []\nNPA_y = []\nfor k, v in map_dict.items():\n for pt in v:\n NPA_x.append(pt[0])\n NPA_y.append(pt[1])", "_____no_output_____" ], [ "p = figure(x_range=(bbox[0][0], bbox[1][0]),y_range=(bbox[0][1], bbox[1][1]), x_axis_type=\"mercator\", y_axis_type=\"mercator\")\np.add_tile(tile_provider)\np.circle(x = NPA_x, y = NPA_y, color= \"firebrick\")\n\nshow(p)\n\n#Pike as my old ops chief would say Yemen is a show\n", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "from lenslikelihood.power_spectra import *\nmass_function_model = 'shethTormen'\nnormalization = 'As'\npivot_string = '01'\npivot = 0.1\n\nstructure_formation_interp_As = load_interpolated_mapping(mass_function_model, pivot_string)\n\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport matplotlib.cm as cm\nimport os\n\nplt.rcParams['axes.linewidth'] = 2.5\nplt.rcParams['xtick.major.width'] = 2.5\nplt.rcParams['xtick.major.size'] = 8\nplt.rcParams['xtick.minor.size'] = 5\nplt.rcParams['ytick.major.width'] = 2.5\nplt.rcParams['ytick.major.size'] = 8\nplt.rcParams['ytick.minor.size'] = 4\nplt.rcParams['ytick.labelsize'] = 15\nplt.rcParams['xtick.labelsize'] = 15", "_____no_output_____" ], [ "from lenslikelihood.measurements import *\nfrom lenslikelihood.sampling import InterpolatedLikelihood\nimport dill as pickle\nfrom trikde.pdfs import DensitySamples, IndepdendentLikelihoods, MultivariateNormalPriorHyperCube, CustomPriorHyperCube\n\nnbins = 20\nparam_names = ['LOS_normalization', 'beta', 'log10c0', 'delta_power_law_index', 'sigma_sub']\nparam_ranges = [all_param_ranges_version2[name] for name in param_names]\nload_from_pickle = True \nsave_to_pickle = False \n\nfilename_extension = '_joint_logprior'\nbase_path = './../lenslikelihood/precomputed_likelihoods/'\nlikelihoods = []\nfor lens in all_lens_names:\n fname = base_path + lens + filename_extension\n print('loading joint likelihoods for lens '+lens+' ...')\n f = open(fname, 'rb')\n single_lens_likelihood = pickle.load(f)\n f.close()\n likelihoods.append(single_lens_likelihood)\n \nlikelihood_noprior = IndepdendentLikelihoods(likelihoods)", "loading joint likelihoods for lens HE0435 ...\nloading joint likelihoods for lens WGD2038 ...\nloading joint likelihoods for lens B1422 ...\nloading joint likelihoods for lens WFI2033 ...\nloading joint likelihoods for lens PSJ1606 ...\nloading joint likelihoods for lens WFI2026 ...\nloading joint likelihoods for lens RXJ0911 ...\nloading joint likelihoods for lens MG0414 ...\nloading joint likelihoods for lens PG1115 ...\nloading joint likelihoods for lens RXJ1131 ...\nloading joint likelihoods for lens WGDJ0405 ...\n" ] ], [ [ "## Priors on the subhalo and field halo mass functions\n\nA reasonable assumption to impose on the inference is that the number of subhalos varies proportionally with the number of field halos, since subhalos are accreted from the field. We can enforce this by choosing an expected amplitude for the subhalo mass function in $\\Lambda$CDM, and then coupling variations to $\\Sigma_{\\rm{sub}}$ around this value to $\\delta_{\\rm{LOS}}$. ", "_____no_output_____" ] ], [ [ "def couple_mass_functions(samples, sigma_sub_theory=0.025, coupling_strength=0.2):\n \n delta_los_samples = samples[:, 0]\n sigma_sub_samples = samples[:, -1]\n delta_sigma_sub = sigma_sub_samples/sigma_sub_theory\n chi2 = (delta_sigma_sub - delta_los_samples)**2/coupling_strength**2 \n return chi2\n\nextrapolate_likelihood = True\nsigma_sub_theory = 0.05\nkwargs_prior = {'sigma_sub_theory': sigma_sub_theory}\nprior_on_mass_functions = CustomPriorHyperCube(couple_mass_functions, param_names, param_ranges, nbins, kwargs_prior)\n\nlikelihood = IndepdendentLikelihoods(likelihoods + [prior_on_mass_functions])\ninterpolated_lens_likelihood = InterpolatedLikelihood(likelihood, param_names, param_ranges, extrapolate=extrapolate_likelihood)", "_____no_output_____" ] ], [ [ "### Plot the likelihood\n\nFirst we show the likelihood as inferred from the lenses with no additional modeling assumptions", "_____no_output_____" ] ], [ [ "from trikde.triangleplot import TrianglePlot\nfig = plt.figure()\ncmap = 'jet'\ntriangle_plot = TrianglePlot([likelihood_noprior])\ntriangle_plot.set_cmap(cmap, marginal_col='k')\ntriangle_plot.truth_color = 'k'\ntruths = {'sigma_sub': 1.05, 'LOS_normalization': 1., 'beta': 0.85, 'log10c0': np.log10(18.5), 'delta_power_law_index': 0.}\naxes = triangle_plot.make_triplot(filled_contours=False, show_intervals=False, contour_alpha=1.,\n contour_colors=['k', 'k'],\n show_contours=True, contour_levels=[0.32], truths=truths)\n\nbeta = r'$\\beta$'\nbeta_ticks = [-0.2, 3, 6, 9, 12, 15]\nc0 = r'$\\log_{10} c_8$'\nc0_ticks = [0., 1.0, 2.0, 3.0, 4.0]\ndelta_power_law_index = r'$\\Delta \\alpha$'\ndpli_ticks = [-0.6, -0.3, 0., 0.3, 0.6, 0.9]\nsigma_sub = r'$\\Sigma_{\\rm{sub}} \\ \\left[\\rm{kpc^{-2}}\\right]$'\nsigma_sub_ticks = [0., 0.025, 0.05, 0.075, 0.1]\ndelta_LOS = r'$\\delta_{\\rm{LOS}}$'\ndlos_ticks = [0.0, 0.5, 1., 1.5, 2., 2.5]\nticksize = 14\nlabelsize = 18\nrotation = 40\n\naxes[5].set_ylabel(beta, fontsize=labelsize)\naxes[5].set_yticks(beta_ticks)\naxes[5].set_yticklabels(beta_ticks, fontsize=ticksize)\n\naxes[10].set_ylabel(c0, fontsize=labelsize)\naxes[10].set_yticks(c0_ticks)\naxes[10].set_yticklabels(c0_ticks, fontsize=ticksize)\n\naxes[15].set_ylabel(delta_power_law_index, fontsize=labelsize)\naxes[15].set_yticks(dpli_ticks)\naxes[15].set_yticklabels(dpli_ticks, fontsize=ticksize)\n\naxes[20].set_ylabel(sigma_sub, fontsize=labelsize)\naxes[20].set_yticks(sigma_sub_ticks)\naxes[20].set_yticklabels(sigma_sub_ticks, fontsize=ticksize)\n\naxes[20].set_xlabel(delta_LOS, fontsize=labelsize)\naxes[20].set_xticks(dlos_ticks)\naxes[20].set_xticklabels(dlos_ticks, fontsize=ticksize, rotation=rotation)\n\naxes[21].set_xlabel(beta, fontsize=labelsize)\naxes[21].set_xticks(beta_ticks)\naxes[21].set_xticklabels(beta_ticks, fontsize=ticksize, rotation=rotation)\n\naxes[22].set_xlabel(c0, fontsize=labelsize)\naxes[22].set_xticks(c0_ticks)\naxes[22].set_xticklabels(c0_ticks, fontsize=ticksize, rotation=rotation)\n\n\naxes[23].set_xlabel(delta_power_law_index, fontsize=labelsize)\naxes[23].set_xticks(dpli_ticks)\naxes[23].set_xticklabels(dpli_ticks, fontsize=ticksize, rotation=rotation)\n\naxes[24].set_xlabel(sigma_sub, fontsize=labelsize)\naxes[24].set_xticks(sigma_sub_ticks)\naxes[24].set_xticklabels(sigma_sub_ticks, fontsize=ticksize, rotation=rotation)\n\nfrom mpl_toolkits.axes_grid1 import make_axes_locatable\nfrom mpl_toolkits.axes_grid1.inset_locator import inset_axes\n\nax_idx = 9\naxins1 = inset_axes(axes[ax_idx],\n width=\"300%\", # width = 50% of parent_bbox width\n height=\"15%\", # height : 5%\n loc='upper right')\nempty = np.zeros((20, 20))\nempty[0,0] = 1\n\nim1 = axes[ax_idx].imshow(empty, interpolation='None', cmap=cmap)\ncb = fig.colorbar(im1, cax=axins1, orientation=\"horizontal\", ticks=[0, 0.25, 0.5, 0.75, 1])\naxes[ax_idx].set_visible(False)\ncb.set_label('probability', fontsize=15)\n#plt.savefig('./figures/lensing_likelihood.pdf')\n", "/Users/danielgilman/.local/lib/python3.6/site-packages/matplotlib/contour.py:1173: UserWarning: No contour levels were found within the data range.\n warnings.warn(\"No contour levels were found\"\n" ] ], [ [ "### Likelihood with a prior\n\nNow we show the likelihood after adding the prior coupling $\\Sigma_{\\rm{sub}}$ to $\\delta_{LOS}$, assuming $\\Sigma_{\\rm{sub}} = 0.05 \\rm{kpc^{-1}}$ in $\\Lambda$CDM, corresponding to doubly efficient tidal disruption of halos between in the Milky Way relative to massive ellipticals", "_____no_output_____" ] ], [ [ "fig = plt.figure()\ntriangle_plot = TrianglePlot([likelihood])\ntriangle_plot.set_cmap(cmap, marginal_col='k')\ntriangle_plot.truth_color = 'k'\ntruths= {'sigma_sub': 1.05, 'LOS_normalization': 1., 'beta': 0.85, 'log10c0': np.log10(18.5), 'delta_power_law_index': 0.}\naxes = triangle_plot.make_triplot(filled_contours=False, show_intervals=False, show_contours=True,\n contour_levels=[0.32], contour_colors=['k', 'k'],\n display_params=['LOS_normalization', 'beta', 'log10c0', 'delta_power_law_index'],\n truths=truths)\n\naxes[4].set_ylabel(beta, fontsize=labelsize)\naxes[4].set_yticks(beta_ticks)\naxes[4].set_yticklabels(beta_ticks, fontsize=ticksize)\n\naxes[8].set_ylabel(c0, fontsize=labelsize)\naxes[8].set_yticks(c0_ticks)\naxes[8].set_yticklabels(c0_ticks, fontsize=ticksize)\n\naxes[12].set_ylabel(delta_power_law_index, fontsize=labelsize)\naxes[12].set_yticks(dpli_ticks)\naxes[12].set_yticklabels(dpli_ticks, fontsize=ticksize)\n\naxes[12].set_xlabel(delta_LOS, fontsize=labelsize)\naxes[12].set_xticks(dlos_ticks)\naxes[12].set_xticklabels(dlos_ticks, fontsize=ticksize, rotation=rotation)\n\naxes[13].set_xlabel(beta, fontsize=labelsize)\naxes[13].set_xticks(beta_ticks)\naxes[13].set_xticklabels(beta_ticks, fontsize=ticksize, rotation=rotation)\n\naxes[14].set_xlabel(c0, fontsize=labelsize)\naxes[14].set_xticks(c0_ticks)\naxes[14].set_xticklabels(c0_ticks, fontsize=ticksize, rotation=rotation)\n\naxes[15].set_xlabel(delta_power_law_index, fontsize=labelsize)\naxes[15].set_xticks(dpli_ticks)\naxes[15].set_xticklabels(dpli_ticks, fontsize=ticksize, rotation=rotation)\n\naxes[2].annotate(r'$\\Sigma_{\\rm{sub(predicted)}} = 0.05 \\rm{kpc^{-2}}$', fontsize=22,\n xy=(0.26, 0.1), xycoords='axes fraction')\nax_idx = 7\naxins1 = inset_axes(axes[ax_idx],\n width=\"200%\", # width = 50% of parent_bbox width\n height=\"10%\", # height : 5%\n loc='upper right')\nempty = np.zeros((20, 20))\nempty[0,0] = 1\n\nim1 = axes[ax_idx].imshow(empty, interpolation='None', cmap=cmap)\ncb = fig.colorbar(im1, cax=axins1, orientation=\"horizontal\", ticks=[0, 0.25, 0.5, 0.75, 1])\naxes[ax_idx].set_visible(False)\ncb.set_label('probability', fontsize=15)\n\n#plt.savefig('./figures/lensing_likelihood_w.pdf')", "_____no_output_____" ] ], [ [ "## Systematic modeling errors\n\nWe allow for systematic errors in the model by changing the internal mapping between the parameters describing the mass function and concentration-mass relation", "_____no_output_____" ] ], [ [ "error_type = 'INTERPOLATED_GRID'\n\nif error_type == 'INTERPOLATED_GRID':\n \n f = open('./systematic_error_interpolations/systematic_error_interpolation_lowfit_'+mass_function_model+'_pivot'+pivot_string+'_3D', 'rb')\n systematic_interp_lowfit = pickle.load(f)\n f.close()\n\n f = open('./systematic_error_interpolations/systematic_error_interpolation_highfit_'+mass_function_model+'_pivot'+pivot_string+'_3D', 'rb')\n systematic_interp_highfit = pickle.load(f)\n f.close()\n \nelif error_type == 'RELATIVE':\n delta_delta_los = 0.1\n delta_beta = 0.2\n delta_c8 = 0.2\n delta_delta_alpha = 0.05", "_____no_output_____" ] ], [ [ "## Final setup", "_____no_output_____" ] ], [ [ "delta_los_range = [0., 2.5]\nbeta_range = [-0.2, 15.]\nlog10c0_range = [0., 4.]\ndelta_alpha_range = [-0.6, 0.9]\nsigma_sub_range = [0., 0.125]\nparam_ranges_lensing = [delta_los_range, beta_range, log10c0_range, delta_alpha_range, sigma_sub_range]\nn_draw = 50000\nextrapolate_ranges = [[0., 2.5], \n [-0.2, 15.],\n [0., 4.0], \n delta_alpha_range,\n sigma_sub_range]\n\nparam_ranges_pk = [[0.6645, 1.2645], [-0.1, 0.1], [-0.01, 0.01]]\narun_ticks = [-0.10, -0.05, 0.00, 0.05, 0.10]\nbrun_ticks = [-0.010, -0.005, 0.000, 0.005, 0.01]\nns_ticks = [0.7645, 0.9645, 1.1645]", "_____no_output_____" ] ], [ [ "## Compute the likelihood of the power spectrum parameters\n\nWe can compute the likelihood the parameters describing $P\\left(k\\right)$, adding systematic models errors by hand", "_____no_output_____" ] ], [ [ "if error_type == 'INTERPOLATED_GRID':\n \n samples_no_sys, like_no_sys = sample_power_spectra_with_systematic_interp(n_draw, param_ranges_pk, param_ranges_lensing, structure_formation_interp_As, interpolated_lens_likelihood,\n systematic_interp_highfit, extrapolate=extrapolate_likelihood, extrapolate_ranges=extrapolate_ranges, log10c8_sys=False, delta_los_sys=False,\n delta_alpha_sys=False, beta_sys=False, three_D=True)\n \n samples_sys1, like_sys1 = sample_power_spectra_with_systematic_interp(n_draw, param_ranges_pk, param_ranges_lensing, structure_formation_interp_As, interpolated_lens_likelihood,\n systematic_interp_lowfit, extrapolate=extrapolate_likelihood, extrapolate_ranges=extrapolate_ranges, three_D=True)\n\n samples_sys2, like_sys2 = sample_power_spectra_with_systematic_interp(n_draw, param_ranges_pk, param_ranges_lensing, structure_formation_interp_As, interpolated_lens_likelihood,\n systematic_interp_highfit, extrapolate=extrapolate_likelihood, extrapolate_ranges=extrapolate_ranges, three_D=True)\n\n\n samples_sys_noamp_1, like_sys_noamp_1 = sample_power_spectra_with_systematic_interp(n_draw, param_ranges_pk, param_ranges_lensing, structure_formation_interp_As, interpolated_lens_likelihood,\n systematic_interp_lowfit, extrapolate=extrapolate_likelihood, extrapolate_ranges=extrapolate_ranges, log10c8_sys=False, delta_los_sys=False, three_D=True)\n\n samples_sys_noamp_2, like_sys_noamp_2 = sample_power_spectra_with_systematic_interp(n_draw, param_ranges_pk, param_ranges_lensing, structure_formation_interp_As, interpolated_lens_likelihood,\n systematic_interp_highfit, extrapolate=extrapolate_likelihood, extrapolate_ranges=extrapolate_ranges, log10c8_sys=False, delta_los_sys=False, three_D=True)\n\n samples_sys_noslope, like_sys_noslope = sample_power_spectra_with_systematic_interp(n_draw, param_ranges_pk, param_ranges_lensing, structure_formation_interp_As, interpolated_lens_likelihood,\n systematic_interp_lowfit, extrapolate=extrapolate_likelihood, extrapolate_ranges=extrapolate_ranges, delta_alpha_sys=False, beta_sys=False, three_D=True)\n \nelif error_type == 'RELATIVE':\n \n samples_sys1, like_sys1 = sample_power_spectra(n_draw, param_ranges_pk, param_ranges_lensing, structure_formation_interp_As, interpolated_lens_likelihood,\n delta_c8, delta_beta, delta_delta_los, delta_delta_alpha, extrapolate=extrapolate_likelihood, extrapolate_ranges=extrapolate_ranges)\n\n samples_sys2, like_sys2 = sample_power_spectra(n_draw, param_ranges_pk, param_ranges_lensing, structure_formation_interp_As, interpolated_lens_likelihood,\n -delta_c8, -delta_beta, -delta_delta_los, -delta_delta_alpha, extrapolate=extrapolate_likelihood, extrapolate_ranges=extrapolate_ranges)\n\n samples_no_sys, like_no_sys = sample_power_spectra(n_draw, param_ranges_pk, param_ranges_lensing, structure_formation_interp_As, interpolated_lens_likelihood,\n 0., 0., 0., 0., extrapolate=extrapolate_likelihood, extrapolate_ranges=extrapolate_ranges)\n\n samples_sys_noamp_1, like_sys_noamp_1 = sample_power_spectra(n_draw, param_ranges_pk, param_ranges_lensing, structure_formation_interp_As, interpolated_lens_likelihood,\n 0., delta_beta, 0., delta_delta_alpha, extrapolate=extrapolate_likelihood, extrapolate_ranges=extrapolate_ranges)\n\n samples_sys_noamp_2, like_sys_noamp_2 = sample_power_spectra(n_draw, param_ranges_pk, param_ranges_lensing, structure_formation_interp_As, interpolated_lens_likelihood,\n 0., -delta_beta, 0., delta_delta_alpha, extrapolate=extrapolate_likelihood, extrapolate_ranges=extrapolate_ranges)\n\n samples_sys_noslope, like_sys_noslope = sample_power_spectra(n_draw, param_ranges_pk, param_ranges_lensing, structure_formation_interp_As, interpolated_lens_likelihood,\n -delta_c8, 0., 0., 0., extrapolate=extrapolate_likelihood, extrapolate_ranges=extrapolate_ranges)\n \n samples_sys_noslope_2, like_sys_noslope_2 = sample_power_spectra(n_draw, param_ranges_pk, param_ranges_lensing, structure_formation_interp_As, interpolated_lens_likelihood,\n delta_c8, 0., 0., 0., extrapolate=extrapolate_likelihood, extrapolate_ranges=extrapolate_ranges)\n \n \n ", "_____no_output_____" ] ], [ [ "## Plot the likelihood of the parameters describing the power spectrum", "_____no_output_____" ] ], [ [ "nbins = 20\nparam_names_pk = [r'$n_s$', r'$a_{\\rm{run}}$', r'$b_{\\rm{run}}$'] \n\nsamples_marginalized = np.vstack((np.vstack((np.vstack((np.vstack((np.vstack((samples_no_sys, samples_sys1)), samples_sys2)), samples_sys_noamp_1)), samples_sys_noamp_2)), samples_sys_noslope))\nlikelihood_marginalized = np.append(np.append(np.append(np.append(np.append(like_no_sys, like_sys1), like_sys2), like_sys_noamp_1), like_sys_noamp_2), like_sys_noslope)\n# samples_marginalized = samples_no_sys\n# likelihood_marginalized = like_no_sys\ndensity_marginalized = DensitySamples(samples_marginalized, param_names_pk, likelihood_marginalized, \n param_ranges_pk, nbins=nbins, use_kde=False, bandwidth_scale=1.)\npk_likelihood_marginalized = IndepdendentLikelihoods([density_marginalized])\n\ntriplot = TrianglePlot([pk_likelihood_marginalized])\ncmap = 'jet'\ntriplot.set_cmap(cmap, marginal_col='k')\ntriplot.truth_color = 'k'\ntruths= {r'$n_s$': 0.9645, r'$a_{\\rm{run}}$': 0., r'$b_{\\rm{run}}$': 0.}\naxes = triplot.make_triplot(filled_contours=False, show_intervals=False, show_contours=True,\n contour_levels=[0.32], contour_colors=['k', 'k'])\n\naxes[3].set_yticks(arun_ticks)\naxes[3].set_yticklabels(arun_ticks, fontsize=ticksize)\n\naxes[6].set_yticks(brun_ticks)\naxes[6].set_yticklabels(brun_ticks, fontsize=ticksize)\n\naxes[6].set_xticks(ns_ticks)\naxes[6].set_xticklabels(ns_ticks, fontsize=ticksize)\n\naxes[7].set_xticks(arun_ticks)\naxes[7].set_xticklabels(arun_ticks, fontsize=ticksize)\n\naxes[8].set_xticks(brun_ticks)\naxes[8].set_xticklabels(brun_ticks, fontsize=ticksize)\n\nax_idx = 1\naxins1 = inset_axes(axes[ax_idx],\n width=\"200%\", # width = 50% of parent_bbox width\n height=\"10%\", # height : 5%\n loc=6)\nempty = np.zeros((20, 20))\nempty[0,0] = 1\n\nim1 = axes[ax_idx].imshow(empty, interpolation='None', cmap=cmap)\ncb = fig.colorbar(im1, cax=axins1, orientation=\"horizontal\", ticks=[0, 0.25, 0.5, 0.75, 1])\naxes[ax_idx].set_visible(False)\ncb.set_label('probability', fontsize=15)\nplt.savefig('./figures/qP_likelihood_'+mass_function_model+'_pivot'+pivot_string+'.pdf')\n\nimport pickle\nf = open('./interpolated_pq_likelihoods/Pk_likelihood_'+mass_function_model+'_pivot'+pivot_string, 'wb')\npk_likelihood_marginalized_interp = InterpolatedLikelihood(pk_likelihood_marginalized, param_names_pk, param_ranges_pk)\npickle.dump(pk_likelihood_marginalized_interp, f)\n", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Content:\n1. [Definitions](#1.-Definitions)\n2. [The root finding problem](#2.-The-root-finding-problem)\n3. [Fixed point iteration](#3.-Fixed-point-iteration)\n>3.1 [The cobweb diagram](#3.1-The-cobweb-diagram) \n>3.2 [Fixed point iteration theorem](#3.2-Fixed-point-iteration-theorem) \n>3.3 [The code](#3.3-The-code)\n4. [Bisection method](#4.-Bisection-method)", "_____no_output_____" ], [ "# 1. Definitions", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "[Weierstrass function](https://en.wikipedia.org/wiki/Weierstrass_function) is a peculiar function. It is continuous on the real number line but not differentiable anywhere.", "_____no_output_____" ] ], [ [ "import numpy as np\nimport matplotlib.pyplot as plt\n\ndef weierstrass(a,b,M,x):\n val = 0.0\n for n in range(0,M):\n val = val + a**n * np.cos(b**n*np.pi*x)\n return val\n\nx = np.linspace(-2,2,1000) # 1000 points between -2 and +2\na=0.5\nb=3.0\n\nN=x.size\ny=np.zeros(N)\n\nM=1\nfor i in range(N):\n y[i]=weierstrass(a,b,M,x[i])\nplt.plot(x, y, 'b-', label='M=1')\n\nplt.title('Weierstrass function, M=1')\nplt.legend()\nplt.show()", "_____no_output_____" ], [ "M=3\nfor i in range(N):\n y[i]=weierstrass(a,b,M,x[i])\nplt.plot(x, y, 'b-', label='M=3')\n\nplt.title('Weierstrass function, M=3')\nplt.legend()\nplt.show()", "_____no_output_____" ], [ "M=10\nfor i in range(N):\n y[i]=weierstrass(a,b,M,x[i])\nplt.plot(x, y, 'b-', label='M=10')\n\nplt.title('Weierstrass function, M=10')\nplt.legend()\nplt.show()", "_____no_output_____" ] ], [ [ "---\nHomework-16: Find examples for polynomial, rational, trigonometric, exponential and logarithmic functions that are in $C^\\infty[{\\bf R}],~{\\rm where}~{\\bf R}$ is the real number line.\n \n---", "_____no_output_____" ], [ "## 2. The root-finding problem", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "## 3. Fixed point iteration ", "_____no_output_____" ], [ "", "_____no_output_____" ] ], [ [ "import numpy as np\n\ndef f(x):\n val=x-np.sqrt(10.0/x)\n return val\n\ndef g(x):\n val=np.sqrt(10.0/x)\n return val\n\nx=1 # initial guess, x0\n\ndx=x\n\ni=0\nwhile dx > 1e-3:\n dx=np.abs(x-g(x))\n print('Iteration: ',i,' x:',x,' g(x):',g(x),' f(x): ', f(x))\n x=g(x)\n i=i+1\n\nprint('Exact root is x:',np.power(10.0,1.0/3.0))", "Iteration: 0 x: 1 g(x): 3.1622776601683795 f(x): -2.1622776601683795\nIteration: 1 x: 3.1622776601683795 g(x): 1.7782794100389228 f(x): 1.3839982501294568\nIteration: 2 x: 1.7782794100389228 g(x): 2.3713737056616555 f(x): -0.5930942956227327\nIteration: 3 x: 2.3713737056616555 g(x): 2.053525026457146 f(x): 0.31784867920450965\nIteration: 4 x: 2.053525026457146 g(x): 2.20673406908459 f(x): -0.15320904262744417\nIteration: 5 x: 2.20673406908459 g(x): 2.1287516617963727 f(x): 0.07798240728821737\nIteration: 6 x: 2.1287516617963727 g(x): 2.1673921695684175 f(x): -0.03864050777204486\nIteration: 7 x: 2.1673921695684175 g(x): 2.1479850285170086 f(x): 0.01940714105140895\nIteration: 8 x: 2.1479850285170086 g(x): 2.157666779453667 f(x): -0.00968175093665824\nIteration: 9 x: 2.157666779453667 g(x): 2.152820461347157 f(x): 0.004846318106509706\nIteration: 10 x: 2.152820461347157 g(x): 2.155242258210635 f(x): -0.0024217968634778764\nIteration: 11 x: 2.155242258210635 g(x): 2.1540310194228653 f(x): 0.0012112387877696307\nIteration: 12 x: 2.1540310194228653 g(x): 2.154636553703824 f(x): -0.0006055342809587749\nExact root is x: 2.154434690031884\n" ] ], [ [ "Here is another elegant way to print the output ", "_____no_output_____" ] ], [ [ "x=1.0\n\nfor i in range(0,15):\n gx=g(x)\n fx=f(x)\n fstring=(f'''Iteration={i:5d} x={x:10.4f} g(x)={gx:10.4f} f(x)={np.abs(fx):10.4f}''') # using f-string\n print(fstring)\n x=g(x)\n\nout=(f'''Exact root is x={np.power(10.0,1.0/3.0):10.4f}''')\nprint(out)\n\n#other way for formatted print\n#mynumber=3.14\n#print('{:10.8f}'.format(mynumber))", "Iteration= 0 x= 1.0000 g(x)= 3.1623 f(x)= 2.1623\nIteration= 1 x= 3.1623 g(x)= 1.7783 f(x)= 1.3840\nIteration= 2 x= 1.7783 g(x)= 2.3714 f(x)= 0.5931\nIteration= 3 x= 2.3714 g(x)= 2.0535 f(x)= 0.3178\nIteration= 4 x= 2.0535 g(x)= 2.2067 f(x)= 0.1532\nIteration= 5 x= 2.2067 g(x)= 2.1288 f(x)= 0.0780\nIteration= 6 x= 2.1288 g(x)= 2.1674 f(x)= 0.0386\nIteration= 7 x= 2.1674 g(x)= 2.1480 f(x)= 0.0194\nIteration= 8 x= 2.1480 g(x)= 2.1577 f(x)= 0.0097\nIteration= 9 x= 2.1577 g(x)= 2.1528 f(x)= 0.0048\nIteration= 10 x= 2.1528 g(x)= 2.1552 f(x)= 0.0024\nIteration= 11 x= 2.1552 g(x)= 2.1540 f(x)= 0.0012\nIteration= 12 x= 2.1540 g(x)= 2.1546 f(x)= 0.0006\nIteration= 13 x= 2.1546 g(x)= 2.1543 f(x)= 0.0003\nIteration= 14 x= 2.1543 g(x)= 2.1545 f(x)= 0.0002\nExact root is x= 2.1544\n" ] ], [ [ "### 3.1 The cobweb diagram ", "_____no_output_____" ], [ "\n\n", "_____no_output_____" ] ], [ [ "def g_fn(x):\n val=np.sqrt(10.0/x)\n return val\n\nN=15\nx=np.zeros(N,float)\ng=np.zeros(N,float)\n\nx0=1.0 # initial guess\n\nNi=10\nfor i in range(0,Ni):\n x[i]=x0\n g[i]=g_fn(x0)\n x0=g[i] \n \n#print(x,g)", "_____no_output_____" ], [ "import numpy as np\nimport matplotlib.pyplot as plt\n\nfig = plt.figure() # comment if square plot is not needed\nax = fig.add_subplot(111) # comment if square plot is not needed\n\nplt.xlim(0, 4)\nplt.ylim(0, 4)\n\nx_grids = np.linspace(0,4,100) \nN=x_grids.size\ng_grids=np.zeros(N)\n\nfor i in range(N):\n g_grids[i]=g_fn(x_grids[i])\n\nplt.plot(x_grids,x_grids,'k-',label='x')\nplt.plot(x_grids,g_grids,'b-',label='g(x)')\n\nxval=[x[0],x[0]]\ngval=[x[0],g[0]]\nplt.plot(xval,gval)\n\nplt.grid()\n\nfor i in range(0,6):\n\n # horizontal line, same y-value\n xval=[x[i],g[i]]\n gval=[g[i],g[i]]\n plt.plot(xval,gval)\n\n # vertical line, same x-value \n xval=[g[i],x[i+1]]\n gval=[g[i],g[i+1]]\n plt.plot(xval,gval)\n\nax.set_aspect('equal', adjustable='box') # comment if square plot is not needed\n\nplt.title('Cobweb diagram for $x=\\sqrt{10/x}$')\nplt.legend()\nplt.show()", "<ipython-input-6-348e80d82643>:2: RuntimeWarning: divide by zero encountered in double_scalars\n val=np.sqrt(10.0/x)\n" ] ], [ [ "### Let's try another problem: $x - 1/x^2 = 0;~g(x)=1/x^2;~x_0 = 0.1$", "_____no_output_____" ] ], [ [ "import numpy as np\n\ndef g_fn(x):\n val=1.0/x**2\n return val\n\ndef f_fn(x):\n val=x-1.0/x**2\n return val\n\nx=0.1\n\nfor i in range(0,4):\n gx=g_fn(x)\n fx=f_fn(x)\n fstring=(f'''Iteration={i:5d} x={x:10.4f} g(x)={gx:10.4f} f(x)={np.abs(fx):10.4f}''') # using f-string\n print(fstring)\n x=g_fn(x)", "Iteration= 0 x= 0.1000 g(x)= 100.0000 f(x)= 99.9000\nIteration= 1 x= 100.0000 g(x)= 0.0001 f(x)= 99.9999\nIteration= 2 x= 0.0001 g(x)=100000000.0000 f(x)=99999999.9999\nIteration= 3 x=100000000.0000 g(x)= 0.0000 f(x)=100000000.0000\n" ] ], [ [ "Diverges! ", "_____no_output_____" ], [ "### 3.2 Fixed point iteration theorem", "_____no_output_____" ], [ "\n\n\n\n\n", "_____no_output_____" ], [ "### 3.3 The code", "_____no_output_____" ] ], [ [ "import numpy as np\n\n# fn is the g(x) in x = g(x) that we want to solve\n# x is the initial guess, x0\n# xthresh is convergence thershold\n# maxeval - maximum number of evaluation of fn\n# iprint control printing, iprint = 1 for extra output\ndef fixedpoint(fn, x, xthresh, maxeval, iprint):\n \n if iprint == 1:\n print('#iter x g(x) dx')\n\n ieval=0\n \n g=fn(x)\n ieval=ieval+1\n \n dx=np.abs(x-g)\n\n iiter=0 \n while dx > xthresh:\n g=fn(x)\n ieval=ieval+1\n \n dx=np.abs(x-g)\n \n if iprint == 1:\n print('{:5d}{:15.6e}{:15.6e}{:15.6e}'.format(iiter,x, g, dx))\n\n if ieval >= maxeval:\n print('Exiting fixed-point iteration, maximum function evaluations reached')\n break\n \n x=g\n iiter=iiter+1\n\n return x \n print('Exiting fixed-point iteration, convergence reached') ", "_____no_output_____" ], [ "def fn_g(x):\n val=np.sqrt(10.0/x)\n return val", "_____no_output_____" ], [ "x0 = 1.0\nxthresh = 1E-5\nmaxeval = 100 \niprint=1\nx = fixedpoint(fn_g, x0, xthresh, maxeval,iprint)\nprint('The solution is: ',x)", "#iter x g(x) dx\n 0 1.000000e+00 3.162278e+00 2.162278e+00\n 1 3.162278e+00 1.778279e+00 1.383998e+00\n 2 1.778279e+00 2.371374e+00 5.930943e-01\n 3 2.371374e+00 2.053525e+00 3.178487e-01\n 4 2.053525e+00 2.206734e+00 1.532090e-01\n 5 2.206734e+00 2.128752e+00 7.798241e-02\n 6 2.128752e+00 2.167392e+00 3.864051e-02\n 7 2.167392e+00 2.147985e+00 1.940714e-02\n 8 2.147985e+00 2.157667e+00 9.681751e-03\n 9 2.157667e+00 2.152820e+00 4.846318e-03\n 10 2.152820e+00 2.155242e+00 2.421797e-03\n 11 2.155242e+00 2.154031e+00 1.211239e-03\n 12 2.154031e+00 2.154637e+00 6.055343e-04\n 13 2.154637e+00 2.154334e+00 3.027884e-04\n 14 2.154334e+00 2.154485e+00 1.513889e-04\n 15 2.154485e+00 2.154409e+00 7.569577e-05\n 16 2.154409e+00 2.154447e+00 3.784755e-05\n 17 2.154447e+00 2.154428e+00 1.892386e-05\n 18 2.154428e+00 2.154438e+00 9.461909e-06\nThe solution is: 2.15443784400631\n" ] ], [ [ "### Let's try another problem: $\\exp(-x) + x/5 - 1 = 0$", "_____no_output_____" ], [ "Let's look at the graphical solution by plotting the function $f(x)$ and see where it takes the value zero.", "_____no_output_____" ] ], [ [ "import numpy as np\nimport matplotlib.pyplot as plt\n\ndef f(x):\n val=np.exp(-x)+x/5.0-1\n return val\n\nxmin=-5.0\nxmax=10.0\nplt.xlim(xmin, xmax)\n\nplt.ylim(-3, 10)\n\nx = np.linspace(xmin,xmax,100) \nN=x_grids.size\ny=np.zeros(N)\n\nfor i in range(N):\n y[i]=f(x[i])\n\nplt.plot(x,x*0,'k-')\nplt.plot(x,y,'b-')\nplt.grid()\nplt.show()", "_____no_output_____" ] ], [ [ "There are two roots for this equation. One at 0.0 and another near 5.0.", "_____no_output_____" ], [ "There are two ways of rearranging the equation to apply the fixed-point iteration $x_{n+1}=g(x_n)$.\n* Option-1: $g_1(x)=5\\left[ 1- \\exp(-x) \\right]$ \n* Option-2: $g_2(x)=-\\log\\left[ 1 - x/5 \\right]$", "_____no_output_____" ] ], [ [ "def g1(x):\n val=5 * ( 1 - np.exp(-x) )\n return val\n\nx0 = 2 # somewhere in between both the solutions\nmaxeval = 20\nxthresh = 0.0001\niprint=1\nx = fixedpoint(g1, x0, xthresh, maxeval,iprint)\nprint('The solution is: ',x)", "#iter x g(x) dx\n 0 2.000000e+00 4.323324e+00 2.323324e+00\n 1 4.323324e+00 4.933721e+00 6.103976e-01\n 2 4.933721e+00 4.964002e+00 3.028046e-02\n 3 4.964002e+00 4.965075e+00 1.073707e-03\n 4 4.965075e+00 4.965113e+00 3.747867e-05\nThe solution is: 4.965112876983948\n" ], [ "def g2(x):\n val=-np.log(1-x/5.0)\n return val\n\nx0 = 2.0 # somewhere in between both the solutions\nmaxeval = 10\nxthresh = 0.0001\niprint=1\nx = fixedpoint(g2, x0, xthresh, maxeval,iprint)\nprint('The solution is: ',x)", "#iter x g(x) dx\n 0 2.000000e+00 5.108256e-01 1.489174e+00\n 1 5.108256e-01 1.077691e-01 4.030565e-01\n 2 1.077691e-01 2.178950e-02 8.597961e-02\n 3 2.178950e-02 4.367423e-03 1.742207e-02\n 4 4.367423e-03 8.738663e-04 3.493557e-03\n 5 8.738663e-04 1.747885e-04 6.990778e-04\n 6 1.747885e-04 3.495832e-05 1.398302e-04\n 7 3.495832e-05 6.991688e-06 2.796663e-05\nThe solution is: 6.991687951052522e-06\n" ] ], [ [ "---\nHomework-17: $For~the~above~example,~using~the~fixed~point~convergence~relation~explain~why~using~g_1(x)~results~in~the~solution~x^*=4.965~while~g_2(x)~results~in~x^*=0.0.~In~both~cases,~use~x_0=2.0~as~the~initial~guess.$\n \n---", "_____no_output_____" ], [ "## 4. Bisection method", "_____no_output_____" ], [ "\n\n", "_____no_output_____" ] ], [ [ "import numpy as np\n\ndef bisection(fn, a0, b0, xthresh, maxeval, iprint):\n \n if iprint == 1:\n print('#iter a b x dx')\n \n ieval=0\n iiter=1\n \n a=a0\n b=b0\n \n dx = abs(a-b)\n \n while dx > xthresh or iiter < 10:\n \n x = (a+b)/2.0\n \n dx = abs(a-b) \n \n fx = fn(x)\n fb = fn(b)\n \n if (fb < xthresh): # handle an exception\n print('The upper limit seems to be a root. Stopping program.')\n x=b\n break\n \n if iprint == 1:\n print('{:5d}{:15.6e}{:15.6e}{:15.6e}{:15.6e}{:15.6e}{:15.6e}'.format(iiter, a, b, x, dx,fx,fb))\n \n if fx*fb > 0:\n b = x\n else:\n a = x \n \n ieval=ieval+2\n if ieval >= maxeval:\n print('Exiting fixed-point iteration, maximum function evaluations reached')\n break\n \n iiter=iiter+1\n\n print('Exiting fixed-point iteration, convergence reached') \n return x ", "_____no_output_____" ], [ "def fn_f(x):\n val=np.exp(-x)+x/5.0-1\n return val", "_____no_output_____" ], [ "a = -25.0\nb = 30.0\nmaxeval = 100\nxthresh = 0.0001\niprint=1\nx = bisection(fn_f, a, b, xthresh, maxeval,iprint)\nprint('The solution is: ',x)", "#iter a b x dx\n 1 -2.500000e+01 3.000000e+01 2.500000e+00 5.500000e+01 -4.179150e-01 5.000000e+00\n 2 2.500000e+00 3.000000e+01 1.625000e+01 2.750000e+01 2.250000e+00 5.000000e+00\n 3 2.500000e+00 1.625000e+01 9.375000e+00 1.375000e+01 8.750848e-01 2.250000e+00\n 4 2.500000e+00 9.375000e+00 5.937500e+00 6.875000e+00 1.901386e-01 8.750848e-01\n 5 2.500000e+00 5.937500e+00 4.218750e+00 3.437500e+00 -1.415330e-01 1.901386e-01\n 6 4.218750e+00 5.937500e+00 5.078125e+00 1.718750e+00 2.185658e-02 1.901386e-01\n 7 4.218750e+00 5.078125e+00 4.648438e+00 8.593750e-01 -6.073595e-02 2.185658e-02\n 8 4.648438e+00 5.078125e+00 4.863281e+00 4.296875e-01 -1.961866e-02 2.185658e-02\n 9 4.863281e+00 5.078125e+00 4.970703e+00 2.148438e-01 1.078893e-03 2.185658e-02\n 10 4.863281e+00 4.970703e+00 4.916992e+00 1.074219e-01 -9.280444e-03 1.078893e-03\n 11 4.916992e+00 4.970703e+00 4.943848e+00 5.371094e-02 -4.103346e-03 1.078893e-03\n 12 4.943848e+00 4.970703e+00 4.957275e+00 2.685547e-02 -1.512860e-03 1.078893e-03\n 13 4.957275e+00 4.970703e+00 4.963989e+00 1.342773e-02 -2.171413e-04 1.078893e-03\n 14 4.963989e+00 4.970703e+00 4.967346e+00 6.713867e-03 4.308366e-04 1.078893e-03\n 15 4.963989e+00 4.967346e+00 4.965668e+00 3.356934e-03 1.068378e-04 4.308366e-04\n 16 4.963989e+00 4.965668e+00 4.964828e+00 1.678467e-03 -5.515417e-05 1.068378e-04\n 17 4.964828e+00 4.965668e+00 4.965248e+00 8.392334e-04 2.584122e-05 1.068378e-04\nThe upper limit seems to be a root. Stopping program.\nExiting fixed-point iteration, convergence reached\nThe solution is: 4.965248107910156\n" ], [ "def fn_f(x):\n val=(x-1)**2\n return val", "_____no_output_____" ], [ "a = -10\nb = 1\nmaxeval = 100\nxthresh = 0.0001\niprint=1\nx = bisection(fn_f, a, b, xthresh, maxeval,iprint)\nprint('The solution is: ',x)", "#iter a b x dx\nThe upper limit seems to be a root. Stopping program.\nExiting fixed-point iteration, convergence reached\nThe solution is: 1\n" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown", "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown", "markdown" ], [ "code", "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown" ], [ "code", "code", "code", "code", "code" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "# 第二十三讲 微分方程和$e^{At}$\n\n## 微分方程$\\frac{du}{dt} = Au$\n\n现有一阶（First-order）微分方程组：$\\left\\{\\begin{matrix} \\frac{du_1}{dt} & = & -u_1 & + 2u_2\\\\ \\frac{du_2}{dt} & = & u_1 & -2u_2 \\end{matrix}\\right.$，其中初始状态 $u(0) = \\begin{bmatrix}u_1 \\\\ u_2 \\end{bmatrix} = \\begin{bmatrix} 1 \\\\ 0 \\end{bmatrix}$，现在我们需要求解方程的一般形式 $u(t)$。\n\n首先，通过微分方程组可以得到系数矩阵 $A = \\begin{bmatrix} -1 & 2 \\\\ 1 & -2 \\end{bmatrix}$，并且该矩阵的特征值为 $\\left\\{\\begin{matrix} \\lambda_1 & = & 0\\\\ \\lambda_2 & = & -3 \\end{matrix}\\right.$，特征向量为 $\\left\\{\\begin{matrix} x_1 & = & \\begin{bmatrix} 2 \\\\ 1 \\end{bmatrix} \\\\ x_2 & = & \\begin{bmatrix} 1 \\\\ -1 \\end{bmatrix} \\end{matrix}\\right.$。此时，解可以写成 $u(t) = c_1 e^{\\lambda_1 t} x_1 + c_2 e^{\\lambda_2 t}x_2$，这里可以进行检验，取 $u = \\lambda_1 e^{\\lambda_1 t}x_1$，有 $\\frac{du}{dt} = \\lambda_1 e^{\\lambda_1 t}x_1 = A e^{\\lambda_1 t}x_1 = Au$。\n\n接下来，通过 $u(0)$ 解出 $c_1 = \\frac{1}{3}, c_2=\\frac{1}{3}$。于是最终的解形式为 $u(t) = \\frac{1}{3}\\begin{bmatrix}2 \\\\ 1\\end{bmatrix} + \\frac{1}{3}e^{-3t}\\begin{bmatrix}1 \\\\ -1\\end{bmatrix}$，并且方程有一个稳定的状态 $u(\\infty) = \\frac{1}{3}\\begin{bmatrix}2 \\\\ 1\\end{bmatrix}$，因为当 $t \\rightarrow \\infty$ 时，$e^{-3t} \\rightarrow 0$。\n\n这里引入方程趋势和特征向之间的关系：\n* 稳定态（Stability）：方程的值最终会趋向于 $0$，即 $u(t) \\rightarrow 0$，这就要求 $e^{\\lambda t} \\rightarrow 0$，所以当所有的特征值的实数部分（虚数部分只是噪音）均小于 $0$ 时，方程具有稳定性。\n* 收敛态（Steady State）：方程最终收敛于某个值，此时需要有一个特征值为 $0$，而剩余的特征值全小于 $0$。\n* 发散态（Blowup）：如果存在一个特征值的实数部分大于零，那么函数不会收敛。\n\n小技巧：直接判断任意二阶矩阵的特征值是否均小于零。对于二阶矩阵 $A = \\begin{bmatrix} a & b \\\\ c & d \\end{bmatrix}$，特征值具有性质即 $\\lambda_1 + \\lambda_2 = a + d, \\lambda_1 * \\lambda_2 = detA = a * d - b * c$。如果特征向量均小于零，那么 $a + d < 0$ 且 $detA > 0$。\n\n总结：原方程组有两个相互耦合的未知函数，即 $u1,u2$ 相互耦合，而特征值和特征向量的作则就是解耦，也就是对角化。回到原方程 $\\frac{du}{dt} = Au$，如果将 $u$ 表示为特征向量的线性组合 $u=Sv$，那么有 $\\frac{du}{dt} = Au \\Rightarrow S\\frac{dv}{dt} = ASv \\Rightarrow \\frac{dv}{dt} = S^{-1}ASv = \\Lambda v$。此时新方程组线性无关 $\\left\\{\\begin{matrix} \\frac{dv_1}{dt} & = & \\lambda_1v1 \\\\ \\frac{dv_2}{dt} & = & \\lambda_2v_2 \\\\ & \\vdots & \\\\ \\frac{dv_n}{dt} & = & \\lambda_nv_n \\end{matrix}\\right.$，分向量的解形式为 $v(t) = e^{\\lambda t}v(0)$，而原方程的解形式为 $u(t)=Se^{\\Lambda t}S^{-1}u(0)$。", "_____no_output_____" ], [ "## 指数矩阵$e^{At}$\n\n指数矩阵（Exponential Matrix）为指数部分的矩阵，例如 $e^{At}$。上个部分中用到结论 $e^{At} = Se^{\\Lambda t}S^{-1}$，这里给出证明。\n\n根据泰勒级数 $e^x = \\sum ^{\\infty}_{0} \\frac{x^n}{n!}$ 将 $e^{At}$ 展开，得 $$e^{At} = I + At + \\frac{(At)^2}{2} + \\dots + \\frac{(At)^n}{n!}$$\n$$e^{At} = I + At + \\frac{A^2}{2}t^2 + \\dots + \\frac{A^n}{n!}t^n$$\n$$e^{At} = SS^{-1} + S\\Lambda S^{1}t + S\\frac{\\Lambda^2}{2}S^{-1}t^2 + \\dots + S\\frac{\\Lambda^n}{n!}S^{-1}t^n$$\n$$e^{At} = S(I + \\Lambda t + \\frac{\\Lambda^2}{2}t^2 + \\dots + \\frac{\\Lambda^n}{n!}t^n)S^{-1}$$\n$$e^{At} = Se^{\\Lambda t}S^{-1}$$\n\n拓展：\n* 几何级数：$\\frac{1}{1-x} = \\sum^{\\infty}_{0} x^n$\n* 第二个泰勒级数：$ (I - At)^{-1} = I + At + (At)^2 + \\dots (At)^n)$\n\n", "_____no_output_____" ], [ "## 高阶微分方程\n\n对于一个二阶（Second-order）微分方程 ${y}''+b{y}'+ky=0$，可以构造方程组 $\\left\\{\\begin{matrix}{y}'' & = & -b{y}' & -ky\\\\ {y}' & = & {y}' & \\end{matrix}\\right.$，写成矩阵形式有 $\\begin{bmatrix}{y}'' \\\\ {y}' \\end{bmatrix} = \\begin{bmatrix} -b & -k \\\\ 1 & 0 \\end{bmatrix}\\begin{bmatrix}{y}' \\\\ {y} \\end{bmatrix}$。\n\n拓展到五阶 ${y}'''''+b{y}''''+c{y}'''+d{y}''+ey′+fy=0$，矩阵形式有 $\\begin{bmatrix} {y}''''' \\\\ {y}'''' \\\\ {y}''' \\\\ {y}'' \\\\ {y}' \\end{bmatrix} = \\begin{bmatrix} -b & -c & -d & -e & -f \\\\ 1 & 0 & 0 & 0 & 0 \\\\ 0 & 1 & 0 & 0 & 0 \\\\ 0 & 0 & 1 & 0 & 0 \\\\ 0 & 0 & 0 & 1 & 0\\end{bmatrix}\\begin{bmatrix} {y}'''' \\\\ {y}''' \\\\ {y}'' \\\\ {y}' \\\\ y \\end{bmatrix}$。\n", "_____no_output_____" ] ] ]

[ "markdown" ]

[ [ "markdown", "markdown", "markdown" ] ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ "Apache-2.0" ]

[ [ [ "Project No.: 3\n\nTime Taken: 3 days\n\nDifficulty: Intermediate.\n\n\nThis is the toughest dataset I've worked with. Learnt a lot. Still a long way to go...\n\nWould love it if you left a comment with advice on where I could have improved, what you liked/disliked about my work, or any thing else. And if you like it, please give it an upvote!", "_____no_output_____" ] ], [ [ "# This Python 3 environment comes with many helpful analytics libraries installed\n# It is defined by the kaggle/python docker image: https://github.com/kaggle/docker-python\n# For example, here's several helpful packages to load in \n\nimport numpy as np # linear algebra\nimport pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n\n# Input data files are available in the \"../input/\" directory.\n# For example, running this (by clicking run or pressing Shift+Enter) will list the files in the input directory\n\nimport os\nprint(os.listdir(\"../input\"))\n\n# Any results you write to the current directory are saved as output.\n\n", "_____no_output_____" ] ], [ [ "# 1. The Problem\n\n**What is the problem?**\n\nTask(T): Predicting the cost of a used car in India.\n\nExperience(E): Data collected from various sources and distributed across various locations in India.\n\nPerformance(P): Mean Absolute Error\n\n**My plan of action:**\n\n* Clean the data (missing values and categorical variables).'.\n* Build the model and check the MAE.\n* Try to improve the model.\n\n\n* Brand matters too! I could select the brand name of the car and treat them as categorical data.\n* Filling the missing values in New_Price might help. I should get all the available values for each brand, get their avg, and fill that brand's missing values. For example, I could get all the available New_Price values for Honda, take their average and use that number for other Honda cars whose New_Price is missing.\n* Try converting Engine, Power and New_Price to numbers.\n* I'll try scaling in the end. Although, I don't think it has much effect on xgboost.", "_____no_output_____" ] ], [ [ "df_full = pd.read_excel(\"../input/Data_Train.xlsx\")\ndf_test = pd.read_excel(\"../input/Data_Test.xlsx\")\ndf_full.head(10)", "_____no_output_____" ], [ "df_full.shape", "_____no_output_____" ] ], [ [ "Mileage contains kmp/kg and kmpl, Engine contains CC, Power contains bhp and New_Price contains Lakh. By removing them I can convert them from 'object' to 'int'/'float'.", "_____no_output_____" ] ], [ [ "df_full.info()\ndf_full.isnull().sum()", "_____no_output_____" ] ], [ [ "# 2. Data Preparation\n\n", "_____no_output_____" ], [ "Let's first modify the 'Name' of the car and extract just the brand name.", "_____no_output_____" ] ], [ [ "df_full['Name'] = df_full.Name.str.split().str.get(0)\ndf_test['Name'] = df_test.Name.str.split().str.get(0)\ndf_full.head()", "_____no_output_____" ], [ "df_full['Name'].value_counts().sum()", "_____no_output_____" ] ], [ [ "df_full.shape = (6019,13). So I guess all rows have been modified.", "_____no_output_____" ], [ "Now I gotta modify 'Mileage', 'Power', 'Engine' and 'New_Price'. But first, I have to deal with missing values.", "_____no_output_____" ], [ "# 2.1 Missing Values", "_____no_output_____" ] ], [ [ "# Get names of columns with missing values\ncols_with_missing = [col for col in df_full.columns\n if df_full[col].isnull().any()]\nprint(\"Columns with missing values:\")\nprint(cols_with_missing)", "_____no_output_____" ], [ "# Let's deal with them one by one.\n\ndf_full['Seats'].fillna(df_full['Seats'].mean(),inplace=True)\ndf_test['Seats'].fillna(df_test['Seats'].mean(),inplace=True)", "_____no_output_____" ] ], [ [ "NOTE: To get more accurate values, we need more data. So I'll combine df_train and df_test data.", "_____no_output_____" ] ], [ [ "data = pd.concat([df_full,df_test], sort=False)", "_____no_output_____" ], [ "import matplotlib.pyplot as plt\n\nplt.figure(figsize=(20,5))\ndata['Mileage'].value_counts().head(100).plot.bar()\nplt.show()", "_____no_output_____" ], [ "df_full['Mileage'] = df_full['Mileage'].fillna('17.0 kmpl')\ndf_test['Mileage'] = df_test['Mileage'].fillna('17.0 kmpl')\n\n#I noticed the 14th entry (and others) have 0.0 kmpl. Let's replace that too.\n\ndf_full['Mileage'] = df_full['Mileage'].replace(\"0.0 kmpl\", \"17.0 kmpl\")\ndf_test['Mileage'] = df_test['Mileage'].replace(\"0.0 kmpl\", \"17.0 kmpl\")", "_____no_output_____" ], [ "plt.figure(figsize=(20,5))\ndata['Engine'].value_counts().head(100).plot.bar()\nplt.show()", "_____no_output_____" ], [ "df_full['Engine'] = df_full['Engine'].fillna('1197 CC')\ndf_test['Engine'] = df_test['Engine'].fillna('1197 CC')", "_____no_output_____" ], [ "plt.figure(figsize=(20,5))\ndata['Power'].value_counts().head(100).plot.bar()\nplt.show()", "_____no_output_____" ], [ "df_full['Power'] = df_full['Power'].fillna('74 bhp')\ndf_test['Power'] = df_test['Power'].fillna('74 bhp')\n\n#I noticed the 76th entry (and others) have null bhp. Let's replace that too. \n#This was creating problems during LabelEncoding.\n\ndf_full['Power'] = df_full['Power'].replace(\"null bhp\", \"74 bhp\")\ndf_test['Power'] = df_test['Power'].replace(\"null bhp\", \"74 bhp\")", "_____no_output_____" ] ], [ [ "Now let's deal with 'New_Price'.", "_____no_output_____" ], [ "**Appoach 1:** Fill the missing values with the value which occurs the most.", "_____no_output_____" ] ], [ [ "plt.figure(figsize=(20,5))\ndata['New_Price'].value_counts().head(100).plot.bar()\nplt.show()", "_____no_output_____" ], [ "# # I'll select 4.78 cuz the others are way too high.\n\n# df_full['New_Price'] = df_full['New_Price'].fillna('4.78 Lakh')\n# df_test['New_Price'] = df_test['New_Price'].fillna('4.78 Lakh')\n\n# # Run the method get_number() defined below first.\n# # Converting it to float.\n\n# df_full['New_Price'] = df_full['New_Price'].apply(get_number).astype('float')\n# df_test['New_Price'] = df_test['New_Price'].apply(get_number).astype('float')", "_____no_output_____" ] ], [ [ "**Approach 2:** Group by Brand names and get the mean of the available values for 'New_Price'. Use these to fill the missing values for the respective brands.\n\nFirst of all I'll have to convert it into numeric data (or else mean() won't work). For that, I'll have to first deal with missing values. So! Here's what we're gonna do:\n\nFirst fill it with 0.0 Lakh, convert the column into float, group and mean, then finally replace all 0.0 values with their respective values. Capiche?\n\n**NOTE: TURNS OUT THIS WAS A COMPLETE AND UTTER WASTE OF MY TIME AND EFFORT. </3**", "_____no_output_____" ] ], [ [ "# Method to extract 'float' from 'object' \n\nimport re\n\ndef get_number(name):\n title_search = re.search('([\\d+\\.+\\d]+\\W)', name)\n \n if title_search:\n return title_search.group(1)\n return \"\"", "_____no_output_____" ] ], [ [ "I got the code for the above step from [here](https://www.kaggle.com/funxexcel/titanic-basic-solution-with-logistic-regression) and modified it.", "_____no_output_____" ] ], [ [ "data['New_Price'] = data['New_Price'].fillna('0.0 Lakh') # dealt with missing values.\n\ndata['New_Price'] = data['New_Price'].apply(get_number).astype('float') #converted to float\n\ntotal = data['New_Price'].groupby(data['Name'])\nprint(total.mean().round(2))", "_____no_output_____" ] ], [ [ "We got avg 'New_Price' values for more than half the brands. There are still 6 Brands whose values are not given. For that, another plan!\n\nFirst of all, deal with the brands that we have values for. After that, use a bar chart to get the value of the most occurring 'New_Price' value. Use that to fill the rest of them.", "_____no_output_____" ] ], [ [ "df_full['New_Price'] = df_full['New_Price'].fillna('0.0 Lakh') # dealt with missing values.\ndf_full['New_Price'] = df_full['New_Price'].apply(get_number).astype('float') #converted to float", "_____no_output_____" ], [ "df_full.loc[df_full['Name']==\"Audi\", 'New_Price'] = df_full.loc[df_full['Name']==\"Audi\", 'New_Price'].replace(0.0,5.02)\ndf_full.loc[df_full['Name']==\"BMW\", 'New_Price'] = df_full.loc[df_full['Name']==\"BMW\", 'New_Price'].replace(0.0,11.14)\ndf_full.loc[df_full['Name']==\"Bentley\", 'New_Price'] = df_full.loc[df_full['Name']==\"Bentley\", 'New_Price'].replace(0.0,1.88)\ndf_full.loc[df_full['Name']==\"Datsun\", 'New_Price'] = df_full.loc[df_full['Name']==\"Datsun\", 'New_Price'].replace(0.0,3.14)\ndf_full.loc[df_full['Name']==\"Fiat\", 'New_Price'] = df_full.loc[df_full['Name']==\"Fiat\", 'New_Price'].replace(0.0,0.95)\n\ndf_full.loc[df_full['Name']==\"Ford\", 'New_Price'] = df_full.loc[df_full['Name']==\"Ford\", 'New_Price'].replace(0.0,1.16)\ndf_full.loc[df_full['Name']==\"Honda\", 'New_Price'] = df_full.loc[df_full['Name']==\"Honda\", 'New_Price'].replace(0.0,1.30)\ndf_full.loc[df_full['Name']==\"Hyundai\", 'New_Price'] = df_full.loc[df_full['Name']==\"Hyundai\", 'New_Price'].replace(0.0,1.03)\ndf_full.loc[df_full['Name']==\"Isuzu\", 'New_Price'] = df_full.loc[df_full['Name']==\"Isuzu\", 'New_Price'].replace(0.0,16.84)\ndf_full.loc[df_full['Name']==\"ISUZU\", 'New_Price'] = df_full.loc[df_full['Name']==\"ISUZU\", 'New_Price'].replace(0.0,16.84)\n\ndf_full.loc[df_full['Name']==\"Jaguar\", 'New_Price'] = df_full.loc[df_full['Name']==\"Jaguar\", 'New_Price'].replace(0.0,8.52)\ndf_full.loc[df_full['Name']==\"Jeep\", 'New_Price'] = df_full.loc[df_full['Name']==\"Jeep\", 'New_Price'].replace(0.0,22.75)\ndf_full.loc[df_full['Name']==\"Land\", 'New_Price'] = df_full.loc[df_full['Name']==\"Land\", 'New_Price'].replace(0.0,4.39)\ndf_full.loc[df_full['Name']==\"Mahindra\", 'New_Price'] = df_full.loc[df_full['Name']==\"Mahindra\", 'New_Price'].replace(0.0,1.20)\ndf_full.loc[df_full['Name']==\"Maruti\", 'New_Price'] = df_full.loc[df_full['Name']==\"Maruti\", 'New_Price'].replace(0.0,1.29)\n\ndf_full.loc[df_full['Name']==\"Mercedes-Benz\", 'New_Price'] = df_full.loc[df_full['Name']==\"Mercedes-Benz\", 'New_Price'].replace(0.0,7.97)\ndf_full.loc[df_full['Name']==\"Mini\", 'New_Price'] = df_full.loc[df_full['Name']==\"Mini\", 'New_Price'].replace(0.0,25.06)\ndf_full.loc[df_full['Name']==\"Mitsubishi\", 'New_Price'] = df_full.loc[df_full['Name']==\"Mitsubishi\", 'New_Price'].replace(0.0,12.03)\ndf_full.loc[df_full['Name']==\"Nissan\", 'New_Price'] = df_full.loc[df_full['Name']==\"Nissan\", 'New_Price'].replace(0.0,1.89)\ndf_full.loc[df_full['Name']==\"Porsche\", 'New_Price'] = df_full.loc[df_full['Name']==\"Porsche\", 'New_Price'].replace(0.0,0.07)\n\ndf_full.loc[df_full['Name']==\"Renault\", 'New_Price'] = df_full.loc[df_full['Name']==\"Renault\", 'New_Price'].replace(0.0,1.49)\ndf_full.loc[df_full['Name']==\"Skoda\", 'New_Price'] = df_full.loc[df_full['Name']==\"Skoda\", 'New_Price'].replace(0.0,3.63)\ndf_full.loc[df_full['Name']==\"Tata\", 'New_Price'] = df_full.loc[df_full['Name']==\"Tata\", 'New_Price'].replace(0.0,2.00)\ndf_full.loc[df_full['Name']==\"Toyota\", 'New_Price'] = df_full.loc[df_full['Name']==\"Toyota\", 'New_Price'].replace(0.0,4.38)\ndf_full.loc[df_full['Name']==\"Volksvagen\", 'New_Price'] = df_full.loc[df_full['Name']==\"Volksvagen\", 'New_Price'].replace(0.0,1.53)\ndf_full.loc[df_full['Name']==\"Volvo\", 'New_Price'] = df_full.loc[df_full['Name']==\"Volvo\", 'New_Price'].replace(0.0,4.62)", "_____no_output_____" ], [ "df_test['New_Price'] = df_test['New_Price'].fillna('0.0 Lakh') # dealt with missing values.\ndf_test['New_Price'] = df_test['New_Price'].apply(get_number).astype('float') #converted to float", "_____no_output_____" ], [ "# Modify df_test too...\n\ndf_test.loc[df_full['Name']==\"Audi\", 'New_Price'] = df_test.loc[df_test['Name']==\"Audi\", 'New_Price'].replace(0.0,5.02)\ndf_test.loc[df_full['Name']==\"BMW\", 'New_Price'] = df_test.loc[df_test['Name']==\"BMW\", 'New_Price'].replace(0.0,11.14)\ndf_test.loc[df_full['Name']==\"Bentley\", 'New_Price'] = df_test.loc[df_test['Name']==\"Bentley\", 'New_Price'].replace(0.0,1.88)\ndf_test.loc[df_full['Name']==\"Datsun\", 'New_Price'] = df_test.loc[df_test['Name']==\"Datsun\", 'New_Price'].replace(0.0,3.14)\ndf_test.loc[df_full['Name']==\"Fiat\", 'New_Price'] = df_test.loc[df_test['Name']==\"Fiat\", 'New_Price'].replace(0.0,0.95)\n\ndf_test.loc[df_full['Name']==\"Ford\", 'New_Price'] = df_test.loc[df_test['Name']==\"Ford\", 'New_Price'].replace(0.0,1.16)\ndf_test.loc[df_full['Name']==\"Honda\", 'New_Price'] = df_test.loc[df_test['Name']==\"Honda\", 'New_Price'].replace(0.0,1.30)\ndf_test.loc[df_full['Name']==\"Hyundai\", 'New_Price'] = df_test.loc[df_test['Name']==\"Hyundai\", 'New_Price'].replace(0.0,1.03)\ndf_test.loc[df_full['Name']==\"Isuzu\", 'New_Price'] = df_test.loc[df_test['Name']==\"Isuzu\", 'New_Price'].replace(0.0,16.84)\ndf_test.loc[df_full['Name']==\"ISUZU\", 'New_Price'] = df_test.loc[df_test['Name']==\"ISUZU\", 'New_Price'].replace(0.0,16.84)\n\ndf_test.loc[df_full['Name']==\"Jaguar\", 'New_Price'] = df_test.loc[df_test['Name']==\"Jaguar\", 'New_Price'].replace(0.0,8.52)\ndf_test.loc[df_full['Name']==\"Jeep\", 'New_Price'] = df_test.loc[df_test['Name']==\"Jeep\", 'New_Price'].replace(0.0,22.75)\ndf_test.loc[df_full['Name']==\"Land\", 'New_Price'] = df_test.loc[df_test['Name']==\"Land\", 'New_Price'].replace(0.0,4.39)\ndf_test.loc[df_full['Name']==\"Mahindra\", 'New_Price'] = df_test.loc[df_test['Name']==\"Mahindra\", 'New_Price'].replace(0.0,1.20)\ndf_test.loc[df_full['Name']==\"Maruti\", 'New_Price'] = df_test.loc[df_test['Name']==\"Maruti\", 'New_Price'].replace(0.0,1.29)\n\ndf_test.loc[df_full['Name']==\"Mercedes-Benz\", 'New_Price'] = df_test.loc[df_test['Name']==\"Mercedes-Benz\", 'New_Price'].replace(0.0,7.97)\ndf_test.loc[df_full['Name']==\"Mini\", 'New_Price'] = df_test.loc[df_test['Name']==\"Mini\", 'New_Price'].replace(0.0,25.06)\ndf_test.loc[df_full['Name']==\"Mitsubishi\", 'New_Price'] = df_test.loc[df_test['Name']==\"Mitsubishi\", 'New_Price'].replace(0.0,12.03)\ndf_test.loc[df_full['Name']==\"Nissan\", 'New_Price'] = df_test.loc[df_test['Name']==\"Nissan\", 'New_Price'].replace(0.0,1.89)\ndf_test.loc[df_full['Name']==\"Porsche\", 'New_Price'] = df_test.loc[df_test['Name']==\"Porsche\", 'New_Price'].replace(0.0,0.07)\n\ndf_test.loc[df_full['Name']==\"Renault\", 'New_Price'] = df_test.loc[df_test['Name']==\"Renault\", 'New_Price'].replace(0.0,1.49)\ndf_test.loc[df_full['Name']==\"Skoda\", 'New_Price'] = df_test.loc[df_test['Name']==\"Skoda\", 'New_Price'].replace(0.0,3.63)\ndf_test.loc[df_full['Name']==\"Tata\", 'New_Price'] = df_test.loc[df_test['Name']==\"Tata\", 'New_Price'].replace(0.0,2.00)\ndf_test.loc[df_full['Name']==\"Toyota\", 'New_Price'] = df_test.loc[df_test['Name']==\"Toyota\", 'New_Price'].replace(0.0,4.38)\ndf_test.loc[df_full['Name']==\"Volksvagen\", 'New_Price'] = df_test.loc[df_test['Name']==\"Volksvagen\", 'New_Price'].replace(0.0,1.53)\ndf_test.loc[df_full['Name']==\"Volvo\", 'New_Price'] = df_test.loc[df_test['Name']==\"Volvo\", 'New_Price'].replace(0.0,4.62)", "_____no_output_____" ] ], [ [ "I must have filled most of the missing values. Now let's use a bar chart to get the most occurring values and fill the rest of them.", "_____no_output_____" ] ], [ [ "plt.figure(figsize=(20,5))\ndf_full['New_Price'].value_counts().head(100).plot.bar()\nplt.show()\n\nplt.figure(figsize=(20,5))\ndf_test['New_Price'].value_counts().head(100).plot.bar()\nplt.show()", "_____no_output_____" ], [ "df_full.loc[df_full['Name']==\"Ambassador\", 'New_Price'] = df_full.loc[df_full['Name']==\"Ambassador\", 'New_Price'].replace(0.0,1.29)\ndf_full.loc[df_full['Name']==\"Chevrolet\", 'New_Price'] = df_full.loc[df_full['Name']==\"Chevrolet\", 'New_Price'].replace(0.0,1.29)\ndf_full.loc[df_full['Name']==\"Force\", 'New_Price'] = df_full.loc[df_full['Name']==\"Force\", 'New_Price'].replace(0.0,1.29)\ndf_full.loc[df_full['Name']==\"Lamborghini\", 'New_Price'] = df_full.loc[df_full['Name']==\"Lamborghini\", 'New_Price'].replace(0.0,1.29)\ndf_full.loc[df_full['Name']==\"OpelCorsa\", 'New_Price'] = df_full.loc[df_full['Name']==\"OpelCorsa\", 'New_Price'].replace(0.0,1.29)\n\ndf_test.loc[df_full['Name']==\"Ambassador\", 'New_Price'] = df_test.loc[df_test['Name']==\"Ambassador\", 'New_Price'].replace(0.0,1.29)\ndf_test.loc[df_full['Name']==\"Chevrolet\", 'New_Price'] = df_test.loc[df_test['Name']==\"Chevrolet\", 'New_Price'].replace(0.0,1.29)\ndf_test.loc[df_full['Name']==\"Force\", 'New_Price'] = df_test.loc[df_test['Name']==\"Force\", 'New_Price'].replace(0.0,1.29)\ndf_test.loc[df_full['Name']==\"Lamborghini\", 'New_Price'] = df_test.loc[df_test['Name']==\"Lamborghini\", 'New_Price'].replace(0.0,1.29)\ndf_test.loc[df_full['Name']==\"OpelCorsa\", 'New_Price'] = df_test.loc[df_test['Name']==\"OpelCorsa\", 'New_Price'].replace(0.0,1.29)", "_____no_output_____" ], [ "df_full.isnull().sum()", "_____no_output_____" ], [ "df_full.head(10)", "_____no_output_____" ], [ "df_full.info()", "_____no_output_____" ] ], [ [ "Now let's convert 'Mileage', 'Engine' and 'Power' into numbers.", "_____no_output_____" ] ], [ [ "#Using the above defined method get_number()\n\ndf_full['Mileage'] = df_full['Mileage'].apply(get_number).astype('float')\ndf_full['Engine'] = df_full['Engine'].apply(get_number).astype('int')\ndf_full['Power'] = df_full['Power'].apply(get_number).astype('float')\n\ndf_test['Mileage'] = df_test['Mileage'].apply(get_number).astype('float')\ndf_test['Engine'] = df_test['Engine'].apply(get_number).astype('int')\ndf_test['Power'] = df_test['Power'].apply(get_number).astype('float')\n\ndf_full.info()", "_____no_output_____" ], [ "help(re) # This baby was realy helpful!", "_____no_output_____" ], [ "df_test.info()", "_____no_output_____" ], [ "df_full.head()", "_____no_output_____" ] ], [ [ "Looks good!!", "_____no_output_____" ], [ "# 2.2 Categorical Variables", "_____no_output_____" ] ], [ [ "from sklearn.model_selection import train_test_split\n\ny = df_full.Price\nX = df_full.drop(['Price'],axis=1)\n# df_test = df_test.drop('New_Price',axis=1)\n\nX_train, X_valid, y_train, y_valid = train_test_split(X,y,train_size=0.82,test_size=0.18,random_state=0)", "_____no_output_____" ], [ "from sklearn.preprocessing import LabelEncoder\n\nlabel_encoder = LabelEncoder()\n\n# X_train[object_cols] = label_encoder.fit_transform(X_train[object_cols])\n# X_valid[object_cols] = label_encoder.transform(X_valid[object_cols])\n# df_test[object_cols] = label_encoder.fit_transform(df_test[object_cols])\n\n# ValueError: bad input shape (4815, 5)\n# That's why I did it manually.\n\nX_train['Name'] = label_encoder.fit_transform(X_train['Name'])\nX_valid['Name'] = label_encoder.transform(X_valid['Name'])\ndf_test['Name'] = label_encoder.fit_transform(df_test['Name'])\n\nX_train['Location'] = label_encoder.fit_transform(X_train['Location'])\nX_valid['Location'] = label_encoder.transform(X_valid['Location'])\ndf_test['Location'] = label_encoder.fit_transform(df_test['Location'])\n\nX_train['Fuel_Type'] = label_encoder.fit_transform(X_train['Fuel_Type'])\nX_valid['Fuel_Type'] = label_encoder.transform(X_valid['Fuel_Type'])\ndf_test['Fuel_Type'] = label_encoder.fit_transform(df_test['Fuel_Type'])\n\nX_train['Transmission'] = label_encoder.fit_transform(X_train['Transmission'])\nX_valid['Transmission'] = label_encoder.transform(X_valid['Transmission'])\ndf_test['Transmission'] = label_encoder.fit_transform(df_test['Transmission'])\n\nX_train['Owner_Type'] = label_encoder.fit_transform(X_train['Owner_Type'])\nX_valid['Owner_Type'] = label_encoder.transform(X_valid['Owner_Type'])\ndf_test['Owner_Type'] = label_encoder.fit_transform(df_test['Owner_Type'])\n", "_____no_output_____" ], [ "X_train.head()", "_____no_output_____" ], [ "X_train.info()", "_____no_output_____" ] ], [ [ "Ah finally!! After 3 days!", "_____no_output_____" ], [ "Quickly tried scaling too. Not a cool move.", "_____no_output_____" ] ], [ [ "# # Let's try scaling too.\n\n# from sklearn.preprocessing import StandardScaler\n# scaler = StandardScaler().fit(X_train)\n# rescaled_X_train = scaler.transform(X_train)\n\n# scaler = StandardScaler().fit(X_valid)\n# rescaled_X_valid = scaler.transform(X_valid)\n\n# scaler = StandardScaler().fit(df_test)\n# rescaled_df_test = scaler.transform(df_test)\n\n# from xgboost import XGBRegressor\n# from sklearn.metrics import mean_absolute_error,mean_squared_error,mean_squared_log_error\n\n# my_model = XGBRegressor(n_estimators=1000, learning_rate=0.05)\n# my_model.fit(rescaled_X_train, y_train, \n# early_stopping_rounds=5, \n# eval_set=[(rescaled_X_valid, y_valid)], \n# verbose=False)\n\n# predictions = my_model.predict(rescaled_X_valid)\n# print(\"MAE: \" + str(mean_absolute_error(predictions, y_valid)))\n# print(\"MSE: \" + str(mean_squared_error(predictions, y_valid)))\n# print(\"MSLE: \" + str(mean_squared_log_error(predictions, y_valid)))\n\n# # MAE: 2.115451765105513\n# # MSE: 17.56415019000094\n# # MSLE: 0.058881434868999126", "_____no_output_____" ] ], [ [ "# 3. Model\n\nI will use XGBRegressor to build the model and MAE to check the performance. I will also check out mean_squared_error and mean_squared_log_error.", "_____no_output_____" ] ], [ [ "from xgboost import XGBRegressor\nfrom sklearn.metrics import mean_absolute_error,mean_squared_error,mean_squared_log_error\n\nmy_model = XGBRegressor(n_estimators=1000, learning_rate=0.05)\nmy_model.fit(X_train, y_train, \n early_stopping_rounds=5, \n eval_set=[(X_valid, y_valid)], \n verbose=False)\n\npredictions = my_model.predict(X_valid)\nprint(\"MAE: \" + str(mean_absolute_error(predictions, y_valid)))\nprint(\"MSE: \" + str(mean_squared_error(predictions, y_valid)))\nprint(\"MSLE: \" + str(mean_squared_log_error(predictions, y_valid)))", "_____no_output_____" ] ], [ [ "# 4. Predictions", "_____no_output_____" ] ], [ [ "preds_test = my_model.predict(df_test)\nprint(preds_test)\n\n# The Price is in the format xx.xx So let's round off and submit.\n\npreds_test = preds_test.round(2)\nprint(preds_test)", "_____no_output_____" ], [ "output = pd.DataFrame({'Price': preds_test})\noutput.to_excel('submission.xlsx', index=False)", "_____no_output_____" ] ], [ [ "# Notes\n\n* Treating 'Mileage' and the others as categorical variables was a mistake. Eg.: Mileage went up from 23.6 to around 338! Converting it to numbers fixed it.\n\n* LabelEncoder won't work if there are missing values.\n\n* ValueError: y contains previously unseen label 'Bentley'. Fixed it by increasing training_size in train_test_split.\n\n* Scaling all the columns made the model worse (as expected).\n\n* With 'New_Price' (33.36L) -\n\nMAE: 1.841521016220765\n\nMSE: 14.468386600963221\n\nMSLE: 0.05295155300850892\n\n* With 'New_Price' (4.78L) -\n\nMAE: 1.9925125514537205\n\nMSE: 15.974590365346188\n\nMSLE: 0.0599331113483451\n\n\n* Without 'New_Price' - \n\nMAE: 1.7999142406259514\n\nMSE: 12.915820113678437\n\nMSLE: 0.05128357937155652\n\n* After manually modifying 'New_Price'\n\nMAE: 1.8252445468636458 Higher! Ugh!\n\nMSE: 13.293730579850678\n\nMSLE: 0.048714052000441106 This is less though...\n", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import pandas as pd\nfrom bs4 import BeautifulSoup as soup\nfrom urllib.request import urlopen\nimport numpy as np\nimport math", "_____no_output_____" ], [ "links = ['https://openpsychometrics.org/tests/characters/stats/FF/1/','https://openpsychometrics.org/tests/characters/stats/SL/1/','https://openpsychometrics.org/tests/characters/stats/MCU/1/','https://openpsychometrics.org/tests/characters/stats/B99/1/','https://openpsychometrics.org/tests/characters/stats/PC/1/','https://openpsychometrics.org/tests/characters/stats/GOT/5/','https://openpsychometrics.org/tests/characters/stats/GOT/2/','https://openpsychometrics.org/tests/characters/stats/F/5/','https://openpsychometrics.org/tests/characters/stats/VD/2/','https://openpsychometrics.org/tests/characters/stats/FF/2/']", "_____no_output_____" ], [ "col_headers = []\ndata = urlopen(links[0])\ndata_html = data.read()\ndata.close()\ndata_soup = soup(data_html, 'html.parser')\nheaders = data_soup.findAll('th')\ncol_headers = [ch.text for ch in headers]\nrows_data = data_soup.findAll('tr')", "_____no_output_____" ], [ "col_headers = col_headers[:2]\ncol_headers = ['Trait', 'rating']", "_____no_output_____" ], [ "table_rows = []\nfor link in links:\n data = urlopen(link)\n data_html = data.read()\n data.close()\n data_soup = soup(data_html, 'html.parser')\n rows_data = data_soup.findAll('tr')\n for row in rows_data:\n current_row = []\n row_data = row.findAll('td')\n for data in row_data:\n current_row.append(data.text)\n table_rows.append(current_row[:2])\n ", "_____no_output_____" ], [ "df = pd.DataFrame(table_rows, columns=col_headers)", "_____no_output_____" ], [ "df.rating = df.rating.apply(pd.to_numeric, errors='coerce')", "_____no_output_____" ], [ "df = df.dropna()", "_____no_output_____" ], [ "df = df.reset_index(drop=True)\ndf", "_____no_output_____" ], [ "df = df[df['rating']>=85]\ndf = df.reset_index(drop = True)\ndf", "_____no_output_____" ], [ "df['freq']=df.groupby(by='Trait')['Trait'].transform('count')\ndf = df.sort_values(by =['freq','rating'], ascending = (0,0))\ndf.drop_duplicates(subset =\"Trait\",keep ='first', inplace = True)\n\ndf = df.reset_index(drop=True)\ndf", "_____no_output_____" ], [ "want = df.iloc[:10,:]\nwant", "_____no_output_____" ], [ "", "_____no_output_____" ], [ "", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "!git clone https://github.com/huggingface/transformers.git", "fatal: destination path 'transformers' already exists and is not an empty directory.\r\n" ], [ "%cd transformers", "/home/ociftci/Untitled Folder/transformers\n" ], [ "!pwd", "/home/ociftci/Untitled Folder/transformers\r\n" ], [ "!git reset --hard 52f44dd", "HEAD is now at 52f44dd6 change TokenClassificationTask class methods to static methods (#7902)\r\n" ], [ "!cp ./examples/token-classification/run_ner.py ../", "_____no_output_____" ], [ "%cd ..", "/home/ociftci/Untitled Folder\n" ], [ "#!wget https://raw.githubusercontent.com/huggingface/transformers/master/examples/token-classification/run_ner.py\n!wget https://raw.githubusercontent.com/huggingface/transformers/master/examples/token-classification/utils_ner.py\n!wget https://raw.githubusercontent.com/huggingface/transformers/master/examples/token-classification/tasks.py\n\n!git clone https://github.com/huggingface/transformers\n%cd transformers\n!pip install .\n!pip install -r ./examples/requirements.txt\n%cd ..", "--2020-11-13 11:54:01-- https://raw.githubusercontent.com/huggingface/transformers/master/examples/token-classification/utils_ner.py\nResolving raw.githubusercontent.com (raw.githubusercontent.com)... 151.101.240.133\nConnecting to raw.githubusercontent.com (raw.githubusercontent.com)|151.101.240.133|:443... connected.\nHTTP request sent, awaiting response... 200 OK\nLength: 15657 (15K) [text/plain]\nSaving to: ‘utils_ner.py.3’\n\nutils_ner.py.3 100%[===================>] 15,29K --.-KB/s in 0,001s \n\n2020-11-13 11:54:01 (20,1 MB/s) - ‘utils_ner.py.3’ saved [15657/15657]\n\n--2020-11-13 11:54:01-- https://raw.githubusercontent.com/huggingface/transformers/master/examples/token-classification/tasks.py\nResolving raw.githubusercontent.com (raw.githubusercontent.com)... 151.101.240.133\nConnecting to raw.githubusercontent.com (raw.githubusercontent.com)|151.101.240.133|:443... connected.\nHTTP request sent, awaiting response... 200 OK\nLength: 5509 (5,4K) [text/plain]\nSaving to: ‘tasks.py.3’\n\ntasks.py.3 100%[===================>] 5,38K --.-KB/s in 0s \n\n2020-11-13 11:54:01 (33,0 MB/s) - ‘tasks.py.3’ saved [5509/5509]\n\nfatal: destination path 'transformers' already exists and is not an empty directory.\n/home/ociftci/Untitled Folder/transformers\nProcessing /home/ociftci/Untitled Folder/transformers\n Installing build dependencies ... \u001b[?25ldone\n\u001b[?25h Getting requirements to build wheel ... \u001b[?25ldone\n\u001b[?25h Preparing wheel metadata ... \u001b[?25ldone\n\u001b[?25hRequirement already satisfied: numpy in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from transformers==3.4.0) (1.19.4)\nRequirement already satisfied: sentencepiece==0.1.91 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from transformers==3.4.0) (0.1.91)\nRequirement already satisfied: tokenizers==0.9.2 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from transformers==3.4.0) (0.9.2)\nRequirement already satisfied: regex!=2019.12.17 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from transformers==3.4.0) (2020.11.11)\nRequirement already satisfied: protobuf in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from transformers==3.4.0) (3.13.0)\nRequirement already satisfied: requests in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from transformers==3.4.0) (2.25.0)\nRequirement already satisfied: tqdm>=4.27 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from transformers==3.4.0) (4.51.0)\nRequirement already satisfied: dataclasses; python_version < \"3.7\" in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from transformers==3.4.0) (0.7)\nRequirement already satisfied: filelock in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from transformers==3.4.0) (3.0.12)\nRequirement already satisfied: packaging in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from transformers==3.4.0) (20.4)\nRequirement already satisfied: sacremoses in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from transformers==3.4.0) (0.0.43)\nRequirement already satisfied: six>=1.9 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from protobuf->transformers==3.4.0) (1.15.0)\nRequirement already satisfied: setuptools in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from protobuf->transformers==3.4.0) (50.3.1.post20201107)\nRequirement already satisfied: urllib3<1.27,>=1.21.1 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from requests->transformers==3.4.0) (1.25.11)\nRequirement already satisfied: idna<3,>=2.5 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from requests->transformers==3.4.0) (2.10)\nRequirement already satisfied: certifi>=2017.4.17 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from requests->transformers==3.4.0) (2020.11.8)\nRequirement already satisfied: chardet<4,>=3.0.2 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from requests->transformers==3.4.0) (3.0.4)\nRequirement already satisfied: pyparsing>=2.0.2 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from packaging->transformers==3.4.0) (2.4.7)\nRequirement already satisfied: joblib in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from sacremoses->transformers==3.4.0) (0.17.0)\nRequirement already satisfied: click in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from sacremoses->transformers==3.4.0) (7.1.2)\nBuilding wheels for collected packages: transformers\n Building wheel for transformers (PEP 517) ... \u001b[?25ldone\n\u001b[?25h Created wheel for transformers: filename=transformers-3.4.0-py3-none-any.whl size=1293574 sha256=e865e6763cefa9f1f286d2f79378bd8aefee0eda20f71b6df0737f1a232ed0d9\n Stored in directory: /tmp/pip-ephem-wheel-cache-3imfab10/wheels/a0/07/9c/9e03b1038c15284ad824ae7fe624bb042b18d888f95ac4b12a\nSuccessfully built transformers\nInstalling collected packages: transformers\n Attempting uninstall: transformers\n Found existing installation: transformers 3.4.0\n Uninstalling transformers-3.4.0:\n Successfully uninstalled transformers-3.4.0\nSuccessfully installed transformers-3.4.0\nRequirement already satisfied: tensorboard in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from -r ./examples/requirements.txt (line 1)) (2.4.0)\nRequirement already satisfied: scikit-learn in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from -r ./examples/requirements.txt (line 2)) (0.23.2)\nRequirement already satisfied: seqeval in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from -r ./examples/requirements.txt (line 3)) (1.2.2)\nRequirement already satisfied: psutil in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from -r ./examples/requirements.txt (line 4)) (5.7.3)\nRequirement already satisfied: sacrebleu in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from -r ./examples/requirements.txt (line 5)) (1.4.14)\nRequirement already satisfied: rouge-score in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from -r ./examples/requirements.txt (line 6)) (0.0.4)\nRequirement already satisfied: tensorflow_datasets in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from -r ./examples/requirements.txt (line 7)) (4.1.0)\nRequirement already satisfied: pytorch-lightning==1.0.4 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from -r ./examples/requirements.txt (line 8)) (1.0.4)\nRequirement already satisfied: matplotlib in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from -r ./examples/requirements.txt (line 9)) (3.3.3)\nRequirement already satisfied: git-python==1.0.3 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from -r ./examples/requirements.txt (line 10)) (1.0.3)\nRequirement already satisfied: faiss-cpu in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from -r ./examples/requirements.txt (line 11)) (1.6.4.post2)\nRequirement already satisfied: streamlit in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from -r ./examples/requirements.txt (line 12)) (0.71.0)\nRequirement already satisfied: elasticsearch in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from -r ./examples/requirements.txt (line 13)) (7.10.0)\nRequirement already satisfied: nltk in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from -r ./examples/requirements.txt (line 14)) (3.5)\nRequirement already satisfied: pandas in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from -r ./examples/requirements.txt (line 15)) (1.1.4)\nRequirement already satisfied: datasets in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from -r ./examples/requirements.txt (line 16)) (1.1.2)\nRequirement already satisfied: fire in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from -r ./examples/requirements.txt (line 17)) (0.3.1)\nRequirement already satisfied: pytest in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from -r ./examples/requirements.txt (line 18)) (6.1.2)\nRequirement already satisfied: conllu in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from -r ./examples/requirements.txt (line 19)) (4.2.1)\nRequirement already satisfied: sentencepiece!=0.1.92 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from -r ./examples/requirements.txt (line 20)) (0.1.91)\nRequirement already satisfied: werkzeug>=0.11.15 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorboard->-r ./examples/requirements.txt (line 1)) (1.0.1)\nRequirement already satisfied: google-auth<2,>=1.6.3 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorboard->-r ./examples/requirements.txt (line 1)) (1.23.0)\nRequirement already satisfied: setuptools>=41.0.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorboard->-r ./examples/requirements.txt (line 1)) (50.3.1.post20201107)\nRequirement already satisfied: protobuf>=3.6.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorboard->-r ./examples/requirements.txt (line 1)) (3.13.0)\nRequirement already satisfied: six>=1.10.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorboard->-r ./examples/requirements.txt (line 1)) (1.15.0)\nRequirement already satisfied: google-auth-oauthlib<0.5,>=0.4.1 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorboard->-r ./examples/requirements.txt (line 1)) (0.4.2)\nRequirement already satisfied: absl-py>=0.4 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorboard->-r ./examples/requirements.txt (line 1)) (0.11.0)\nRequirement already satisfied: markdown>=2.6.8 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorboard->-r ./examples/requirements.txt (line 1)) (3.3.3)\nRequirement already satisfied: tensorboard-plugin-wit>=1.6.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorboard->-r ./examples/requirements.txt (line 1)) (1.7.0)\nRequirement already satisfied: wheel>=0.26; python_version >= \"3\" in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorboard->-r ./examples/requirements.txt (line 1)) (0.35.1)\nRequirement already satisfied: numpy>=1.12.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorboard->-r ./examples/requirements.txt (line 1)) (1.19.4)\nRequirement already satisfied: grpcio>=1.24.3 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorboard->-r ./examples/requirements.txt (line 1)) (1.33.2)\nRequirement already satisfied: requests<3,>=2.21.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorboard->-r ./examples/requirements.txt (line 1)) (2.25.0)\nRequirement already satisfied: joblib>=0.11 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from scikit-learn->-r ./examples/requirements.txt (line 2)) (0.17.0)\nRequirement already satisfied: threadpoolctl>=2.0.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from scikit-learn->-r ./examples/requirements.txt (line 2)) (2.1.0)\nRequirement already satisfied: scipy>=0.19.1 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from scikit-learn->-r ./examples/requirements.txt (line 2)) (1.5.4)\nRequirement already satisfied: portalocker in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from sacrebleu->-r ./examples/requirements.txt (line 5)) (2.0.0)\nRequirement already satisfied: termcolor in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorflow_datasets->-r ./examples/requirements.txt (line 7)) (1.1.0)\nRequirement already satisfied: typing-extensions; python_version < \"3.8\" in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorflow_datasets->-r ./examples/requirements.txt (line 7)) (3.7.4.3)\nRequirement already satisfied: future in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorflow_datasets->-r ./examples/requirements.txt (line 7)) (0.18.2)\nRequirement already satisfied: tensorflow-metadata in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorflow_datasets->-r ./examples/requirements.txt (line 7)) (0.25.0)\nRequirement already satisfied: dataclasses; python_version < \"3.7\" in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorflow_datasets->-r ./examples/requirements.txt (line 7)) (0.7)\nRequirement already satisfied: attrs>=18.1.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorflow_datasets->-r ./examples/requirements.txt (line 7)) (20.3.0)\nRequirement already satisfied: importlib-resources; python_version < \"3.9\" in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorflow_datasets->-r ./examples/requirements.txt (line 7)) (3.3.0)\nRequirement already satisfied: dill in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorflow_datasets->-r ./examples/requirements.txt (line 7)) (0.3.3)\nRequirement already satisfied: tqdm in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorflow_datasets->-r ./examples/requirements.txt (line 7)) (4.51.0)\nRequirement already satisfied: promise in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorflow_datasets->-r ./examples/requirements.txt (line 7)) (2.3)\nRequirement already satisfied: PyYAML>=5.1 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from pytorch-lightning==1.0.4->-r ./examples/requirements.txt (line 8)) (5.3.1)\nRequirement already satisfied: torch>=1.3 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from pytorch-lightning==1.0.4->-r ./examples/requirements.txt (line 8)) (1.7.0+cu101)\nRequirement already satisfied: fsspec>=0.8.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from pytorch-lightning==1.0.4->-r ./examples/requirements.txt (line 8)) (0.8.4)\nRequirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.3 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from matplotlib->-r ./examples/requirements.txt (line 9)) (2.4.7)\nRequirement already satisfied: pillow>=6.2.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from matplotlib->-r ./examples/requirements.txt (line 9)) (8.0.1)\nRequirement already satisfied: cycler>=0.10 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from matplotlib->-r ./examples/requirements.txt (line 9)) (0.10.0)\nRequirement already satisfied: kiwisolver>=1.0.1 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from matplotlib->-r ./examples/requirements.txt (line 9)) (1.3.1)\nRequirement already satisfied: python-dateutil>=2.1 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from matplotlib->-r ./examples/requirements.txt (line 9)) (2.8.1)\nRequirement already satisfied: gitpython in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from git-python==1.0.3->-r ./examples/requirements.txt (line 10)) (3.1.11)\nRequirement already satisfied: packaging in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from streamlit->-r ./examples/requirements.txt (line 12)) (20.4)\nRequirement already satisfied: blinker in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from streamlit->-r ./examples/requirements.txt (line 12)) (1.4)\nRequirement already satisfied: toml in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from streamlit->-r ./examples/requirements.txt (line 12)) (0.10.2)\nRequirement already satisfied: watchdog in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from streamlit->-r ./examples/requirements.txt (line 12)) (0.10.3)\nRequirement already satisfied: enum-compat in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from streamlit->-r ./examples/requirements.txt (line 12)) (0.0.3)\nRequirement already satisfied: pyarrow in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from streamlit->-r ./examples/requirements.txt (line 12)) (2.0.0)\nRequirement already satisfied: botocore>=1.13.44 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from streamlit->-r ./examples/requirements.txt (line 12)) (1.19.16)\nRequirement already satisfied: altair>=3.2.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from streamlit->-r ./examples/requirements.txt (line 12)) (4.1.0)\nRequirement already satisfied: cachetools>=4.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from streamlit->-r ./examples/requirements.txt (line 12)) (4.1.1)\nRequirement already satisfied: validators in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from streamlit->-r ./examples/requirements.txt (line 12)) (0.18.1)\nRequirement already satisfied: pydeck>=0.1.dev5 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from streamlit->-r ./examples/requirements.txt (line 12)) (0.5.0)\nRequirement already satisfied: base58 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from streamlit->-r ./examples/requirements.txt (line 12)) (2.0.1)\nRequirement already satisfied: boto3 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from streamlit->-r ./examples/requirements.txt (line 12)) (1.16.16)\nRequirement already satisfied: tornado>=5.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from streamlit->-r ./examples/requirements.txt (line 12)) (6.1)\nRequirement already satisfied: tzlocal in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from streamlit->-r ./examples/requirements.txt (line 12)) (2.1)\nRequirement already satisfied: astor in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from streamlit->-r ./examples/requirements.txt (line 12)) (0.8.1)\nRequirement already satisfied: click>=7.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from streamlit->-r ./examples/requirements.txt (line 12)) (7.1.2)\nRequirement already satisfied: urllib3<2,>=1.21.1 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from elasticsearch->-r ./examples/requirements.txt (line 13)) (1.25.11)\nRequirement already satisfied: certifi in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from elasticsearch->-r ./examples/requirements.txt (line 13)) (2020.11.8)\nRequirement already satisfied: regex in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from nltk->-r ./examples/requirements.txt (line 14)) (2020.11.11)\nRequirement already satisfied: pytz>=2017.2 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from pandas->-r ./examples/requirements.txt (line 15)) (2020.4)\nRequirement already satisfied: xxhash in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from datasets->-r ./examples/requirements.txt (line 16)) (2.0.0)\nRequirement already satisfied: multiprocess in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from datasets->-r ./examples/requirements.txt (line 16)) (0.70.11.1)\nRequirement already satisfied: filelock in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from datasets->-r ./examples/requirements.txt (line 16)) (3.0.12)\nRequirement already satisfied: importlib-metadata>=0.12; python_version < \"3.8\" in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from pytest->-r ./examples/requirements.txt (line 18)) (2.0.0)\nRequirement already satisfied: pluggy<1.0,>=0.12 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from pytest->-r ./examples/requirements.txt (line 18)) (0.13.1)\nRequirement already satisfied: iniconfig in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from pytest->-r ./examples/requirements.txt (line 18)) (1.1.1)\nRequirement already satisfied: py>=1.8.2 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from pytest->-r ./examples/requirements.txt (line 18)) (1.9.0)\nRequirement already satisfied: pyasn1-modules>=0.2.1 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from google-auth<2,>=1.6.3->tensorboard->-r ./examples/requirements.txt (line 1)) (0.2.8)\nRequirement already satisfied: rsa<5,>=3.1.4; python_version >= \"3.5\" in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from google-auth<2,>=1.6.3->tensorboard->-r ./examples/requirements.txt (line 1)) (4.6)\nRequirement already satisfied: requests-oauthlib>=0.7.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from google-auth-oauthlib<0.5,>=0.4.1->tensorboard->-r ./examples/requirements.txt (line 1)) (1.3.0)\nRequirement already satisfied: chardet<4,>=3.0.2 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from requests<3,>=2.21.0->tensorboard->-r ./examples/requirements.txt (line 1)) (3.0.4)\nRequirement already satisfied: idna<3,>=2.5 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from requests<3,>=2.21.0->tensorboard->-r ./examples/requirements.txt (line 1)) (2.10)\nRequirement already satisfied: googleapis-common-protos<2,>=1.52.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from tensorflow-metadata->tensorflow_datasets->-r ./examples/requirements.txt (line 7)) (1.52.0)\nRequirement already satisfied: zipp>=0.4; python_version < \"3.8\" in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from importlib-resources; python_version < \"3.9\"->tensorflow_datasets->-r ./examples/requirements.txt (line 7)) (3.4.0)\nRequirement already satisfied: gitdb<5,>=4.0.1 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from gitpython->git-python==1.0.3->-r ./examples/requirements.txt (line 10)) (4.0.5)\nRequirement already satisfied: pathtools>=0.1.1 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from watchdog->streamlit->-r ./examples/requirements.txt (line 12)) (0.1.2)\nRequirement already satisfied: jmespath<1.0.0,>=0.7.1 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from botocore>=1.13.44->streamlit->-r ./examples/requirements.txt (line 12)) (0.10.0)\nRequirement already satisfied: jinja2 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from altair>=3.2.0->streamlit->-r ./examples/requirements.txt (line 12)) (2.11.2)\nRequirement already satisfied: toolz in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from altair>=3.2.0->streamlit->-r ./examples/requirements.txt (line 12)) (0.11.1)\nRequirement already satisfied: jsonschema in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from altair>=3.2.0->streamlit->-r ./examples/requirements.txt (line 12)) (3.2.0)\nRequirement already satisfied: entrypoints in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from altair>=3.2.0->streamlit->-r ./examples/requirements.txt (line 12)) (0.3)\nRequirement already satisfied: decorator>=3.4.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from validators->streamlit->-r ./examples/requirements.txt (line 12)) (4.4.2)\nRequirement already satisfied: traitlets>=4.3.2 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (4.3.3)\nRequirement already satisfied: ipykernel>=5.1.2; python_version >= \"3.4\" in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (5.3.4)\nRequirement already satisfied: ipywidgets>=7.0.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (7.5.1)\nRequirement already satisfied: s3transfer<0.4.0,>=0.3.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from boto3->streamlit->-r ./examples/requirements.txt (line 12)) (0.3.3)\nRequirement already satisfied: pyasn1<0.5.0,>=0.4.6 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from pyasn1-modules>=0.2.1->google-auth<2,>=1.6.3->tensorboard->-r ./examples/requirements.txt (line 1)) (0.4.8)\nRequirement already satisfied: oauthlib>=3.0.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from requests-oauthlib>=0.7.0->google-auth-oauthlib<0.5,>=0.4.1->tensorboard->-r ./examples/requirements.txt (line 1)) (3.1.0)\nRequirement already satisfied: smmap<4,>=3.0.1 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from gitdb<5,>=4.0.1->gitpython->git-python==1.0.3->-r ./examples/requirements.txt (line 10)) (3.0.4)\nRequirement already satisfied: MarkupSafe>=0.23 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from jinja2->altair>=3.2.0->streamlit->-r ./examples/requirements.txt (line 12)) (1.1.1)\nRequirement already satisfied: pyrsistent>=0.14.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from jsonschema->altair>=3.2.0->streamlit->-r ./examples/requirements.txt (line 12)) (0.17.3)\nRequirement already satisfied: ipython-genutils in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from traitlets>=4.3.2->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (0.2.0)\nRequirement already satisfied: ipython>=5.0.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from ipykernel>=5.1.2; python_version >= \"3.4\"->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (5.8.0)\nRequirement already satisfied: jupyter-client in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from ipykernel>=5.1.2; python_version >= \"3.4\"->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (6.1.7)\nRequirement already satisfied: nbformat>=4.2.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from ipywidgets>=7.0.0->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (5.0.8)\nRequirement already satisfied: widgetsnbextension~=3.5.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from ipywidgets>=7.0.0->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (3.5.1)\nRequirement already satisfied: pygments in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from ipython>=5.0.0->ipykernel>=5.1.2; python_version >= \"3.4\"->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (2.7.2)\nRequirement already satisfied: simplegeneric>0.8 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from ipython>=5.0.0->ipykernel>=5.1.2; python_version >= \"3.4\"->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (0.8.1)\nRequirement already satisfied: pickleshare in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from ipython>=5.0.0->ipykernel>=5.1.2; python_version >= \"3.4\"->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (0.7.5)\nRequirement already satisfied: prompt-toolkit<2.0.0,>=1.0.4 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from ipython>=5.0.0->ipykernel>=5.1.2; python_version >= \"3.4\"->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (1.0.15)\nRequirement already satisfied: pexpect; sys_platform != \"win32\" in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from ipython>=5.0.0->ipykernel>=5.1.2; python_version >= \"3.4\"->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (4.8.0)\nRequirement already satisfied: jupyter-core>=4.6.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from jupyter-client->ipykernel>=5.1.2; python_version >= \"3.4\"->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (4.6.3)\nRequirement already satisfied: pyzmq>=13 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from jupyter-client->ipykernel>=5.1.2; python_version >= \"3.4\"->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (19.0.2)\nRequirement already satisfied: notebook>=4.4.1 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from widgetsnbextension~=3.5.0->ipywidgets>=7.0.0->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (6.1.5)\nRequirement already satisfied: wcwidth in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from prompt-toolkit<2.0.0,>=1.0.4->ipython>=5.0.0->ipykernel>=5.1.2; python_version >= \"3.4\"->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (0.2.5)\nRequirement already satisfied: ptyprocess>=0.5 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from pexpect; sys_platform != \"win32\"->ipython>=5.0.0->ipykernel>=5.1.2; python_version >= \"3.4\"->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (0.6.0)\nRequirement already satisfied: Send2Trash in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from notebook>=4.4.1->widgetsnbextension~=3.5.0->ipywidgets>=7.0.0->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (1.5.0)\nRequirement already satisfied: nbconvert in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from notebook>=4.4.1->widgetsnbextension~=3.5.0->ipywidgets>=7.0.0->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (6.0.7)\nRequirement already satisfied: terminado>=0.8.3 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from notebook>=4.4.1->widgetsnbextension~=3.5.0->ipywidgets>=7.0.0->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (0.9.1)\nRequirement already satisfied: argon2-cffi in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from notebook>=4.4.1->widgetsnbextension~=3.5.0->ipywidgets>=7.0.0->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (20.1.0)\nRequirement already satisfied: prometheus-client in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from notebook>=4.4.1->widgetsnbextension~=3.5.0->ipywidgets>=7.0.0->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (0.8.0)\nRequirement already satisfied: mistune<2,>=0.8.1 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from nbconvert->notebook>=4.4.1->widgetsnbextension~=3.5.0->ipywidgets>=7.0.0->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (0.8.4)\nRequirement already satisfied: jupyterlab-pygments in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from nbconvert->notebook>=4.4.1->widgetsnbextension~=3.5.0->ipywidgets>=7.0.0->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (0.1.2)\nRequirement already satisfied: pandocfilters>=1.4.1 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from nbconvert->notebook>=4.4.1->widgetsnbextension~=3.5.0->ipywidgets>=7.0.0->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (1.4.2)\nRequirement already satisfied: testpath in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from nbconvert->notebook>=4.4.1->widgetsnbextension~=3.5.0->ipywidgets>=7.0.0->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (0.4.4)\nRequirement already satisfied: defusedxml in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from nbconvert->notebook>=4.4.1->widgetsnbextension~=3.5.0->ipywidgets>=7.0.0->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (0.6.0)\nRequirement already satisfied: nbclient<0.6.0,>=0.5.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from nbconvert->notebook>=4.4.1->widgetsnbextension~=3.5.0->ipywidgets>=7.0.0->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (0.5.1)\nRequirement already satisfied: bleach in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from nbconvert->notebook>=4.4.1->widgetsnbextension~=3.5.0->ipywidgets>=7.0.0->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (3.2.1)\nRequirement already satisfied: cffi>=1.0.0 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from argon2-cffi->notebook>=4.4.1->widgetsnbextension~=3.5.0->ipywidgets>=7.0.0->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (1.11.5)\nRequirement already satisfied: async-generator in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from nbclient<0.6.0,>=0.5.0->nbconvert->notebook>=4.4.1->widgetsnbextension~=3.5.0->ipywidgets>=7.0.0->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (1.10)\nRequirement already satisfied: nest-asyncio in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from nbclient<0.6.0,>=0.5.0->nbconvert->notebook>=4.4.1->widgetsnbextension~=3.5.0->ipywidgets>=7.0.0->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (1.4.2)\nRequirement already satisfied: webencodings in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from bleach->nbconvert->notebook>=4.4.1->widgetsnbextension~=3.5.0->ipywidgets>=7.0.0->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (0.5.1)\nRequirement already satisfied: pycparser in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from cffi>=1.0.0->argon2-cffi->notebook>=4.4.1->widgetsnbextension~=3.5.0->ipywidgets>=7.0.0->pydeck>=0.1.dev5->streamlit->-r ./examples/requirements.txt (line 12)) (2.20)\n/home/ociftci/Untitled Folder\n" ], [ "!pip install pyarrow --upgrade", "Requirement already up-to-date: pyarrow in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (2.0.0)\r\nRequirement already satisfied, skipping upgrade: numpy>=1.14 in /home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages (from pyarrow) (1.19.4)\r\n" ], [ "import transformers", "_____no_output_____" ], [ "!mkdir data\n# aynı dizinde data klasöründe train,test,dev tsv dosyaları yükleniyor.\n# eğer text dosyaları varsa hazır, direk run_ner.py çalıştırılabilir.\n\nblankLineIndicator = \"BlankLineIndicator\"\nblank = \"\"\nfirstColumnIndex = 0\nsecondColumnIndex = 1", "mkdir: cannot create directory ‘data’: File exists\r\n" ], [ "!cp -r ./NCBI-disease/ ./data/", "cp: cannot stat './NCBI-disease/': No such file or directory\r\n" ], [ "!mv ./data/NCBI-disease/devel.tsv ./data/NCBI-disease/dev.tsv", "mv: cannot stat './data/NCBI-disease/devel.tsv': No such file or directory\r\n" ], [ "!mv ./dev.tsv ./NCBI-disease/", "mv: cannot stat './dev.tsv': No such file or directory\r\n" ], [ "!pwd", "/home/ociftci/Untitled Folder\r\n" ], [ "#!unzip NERdata.zip -d data", "_____no_output_____" ], [ "#!ls data/", "_____no_output_____" ], [ "train_dev_tsv = []\n\nwith open('./data/NCBI-disease/train_dev.tsv', 'r') as f:\n train_dev_pd = f.readlines()\n\n\nfor row in train_dev_pd:\n row = row.split('\\n')[0].split('\\t')\n #Token Sütun İsmi\n if row:\n pass\n if row[firstColumnIndex] == '':\n train_dev_tsv.append(blank)\n else: #Token Sütun İsmi\n #print(row)\n train_dev_tsv.append(row[firstColumnIndex] + \" \" + row[secondColumnIndex])\n", "_____no_output_____" ], [ "train_dev_tsv[0]", "_____no_output_____" ], [ "test_tsv = []\n\nwith open('./data/NCBI-disease/test.tsv', 'r') as f:\n test_pd = f.readlines()\n\n\nfor row in test_pd:\n row = row.split('\\n')[0].split('\\t')\n #Token Sütun İsmi\n if row:\n pass\n if row[firstColumnIndex] == '':\n test_tsv.append(blank)\n else: #Token Sütun İsmi\n #print(row)\n test_tsv.append(row[firstColumnIndex] + \" \" + row[secondColumnIndex])\n", "_____no_output_____" ], [ "dev_tsv = []\n\nwith open('./data/NCBI-disease/dev.tsv', 'r') as f:\n test_pd = f.readlines()\n\n\nfor row in test_pd:\n row = row.split('\\n')[0].split('\\t')\n #Token Sütun İsmi\n if row:\n pass\n if row[firstColumnIndex] == '':\n dev_tsv.append(blank)\n else: #Token Sütun İsmi\n #print(row)\n dev_tsv.append(row[firstColumnIndex] + \" \" + row[secondColumnIndex])\n", "_____no_output_____" ], [ "print(len(train_dev_pd))\nprint(len(train_dev_tsv))", "166017\n166017\n" ], [ "train_dev_tsv[12288].split()[0]", "_____no_output_____" ], [ "train_dev_pd[0], train_dev_tsv[0]", "_____no_output_____" ], [ "l = []\nfor item in train_dev_tsv:\n try:\n item = item.split()[1]\n if item != 'B' and item != 'I' and item != 'O':\n print(item)\n l.append(item)\n except:\n pass\n\nl = set(l)\nprint(l)\n\nwith open('labels.txt', 'w') as f:\n for item in list(l):\n f.write(item + '\\n')\n", "{'B', 'I', 'O'}\n" ], [ "#!cut -f2 BC2GM/train.tsv | sort | uniq", "_____no_output_____" ], [ "dev_tsv[0]", "_____no_output_____" ], [ "def create_txt(file_name, lines):\n file = open(file_name, 'w') \n for line in lines:\n file.write(line + \"\\n\") \n file.close()\n\n#create_txt(\"./data/train.txt\",train_tsv)\ncreate_txt(\"data/train.txt\", train_dev_tsv)\ncreate_txt(\"data/test.txt\", test_tsv)\ncreate_txt(\"data/dev.txt\", dev_tsv)\n# txt file'a çeviriyoruz", "_____no_output_____" ], [ "# !cat data/NCBI-disease/train.tsv | tr \"\\t\" \" \" | head -10", "_____no_output_____" ], [ "#labels.txt -> unique varlık ismi sınıflarının olduğu text dosyası", "_____no_output_____" ], [ "OUTPUT_DIR = \"electra-ner\"", "_____no_output_____" ], [ "!cd data", "_____no_output_____" ], [ "!ls ./data", "dev.txt NCBI-disease test.txt train.txt\r\n" ], [ "!python3 run_ner.py --data_dir ./data/ \\\n--labels ./labels.txt \\\n--model_name_or_path enelpi/med-electra-small-discriminator \\\n--output_dir $OUTPUT_DIR \\\n--max_seq_length 128 \\\n--num_train_epochs 3 \\\n--per_device_train_batch_size 16 \\\n--overwrite_output_dir \\\n--save_steps 10000 \\\n--seed 41 \\\n--do_train \\\n--do_eval \\\n--do_predict", "11/13/2020 22:18:13 - WARNING - __main__ - Process rank: -1, device: cuda:0, n_gpu: 2, distributed training: False, 16-bits training: False\n11/13/2020 22:18:13 - INFO - __main__ - Training/evaluation parameters TrainingArguments(output_dir='electra-ner', overwrite_output_dir=True, do_train=True, do_eval=True, do_predict=True, evaluate_during_training=False, evaluation_strategy=<EvaluationStrategy.NO: 'no'>, prediction_loss_only=False, per_device_train_batch_size=16, per_device_eval_batch_size=8, per_gpu_train_batch_size=None, per_gpu_eval_batch_size=None, gradient_accumulation_steps=1, eval_accumulation_steps=None, learning_rate=5e-05, weight_decay=0.0, adam_beta1=0.9, adam_beta2=0.999, adam_epsilon=1e-08, max_grad_norm=1.0, num_train_epochs=3.0, max_steps=-1, warmup_steps=0, logging_dir='runs/Nov13_22-18-13_inzva-GPU', logging_first_step=False, logging_steps=500, save_steps=10000, save_total_limit=None, no_cuda=False, seed=41, fp16=False, fp16_opt_level='O1', local_rank=-1, tpu_num_cores=None, tpu_metrics_debug=False, debug=False, dataloader_drop_last=False, eval_steps=500, dataloader_num_workers=0, past_index=-1, run_name='electra-ner', disable_tqdm=False, remove_unused_columns=True, label_names=None, load_best_model_at_end=False, metric_for_best_model=None, greater_is_better=None)\nSome weights of the model checkpoint at enelpi/med-electra-small-discriminator were not used when initializing ElectraForTokenClassification: ['discriminator_predictions.dense.weight', 'discriminator_predictions.dense.bias', 'discriminator_predictions.dense_prediction.weight', 'discriminator_predictions.dense_prediction.bias']\n- This IS expected if you are initializing ElectraForTokenClassification from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPretraining model).\n- This IS NOT expected if you are initializing ElectraForTokenClassification from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\nSome weights of ElectraForTokenClassification were not initialized from the model checkpoint at enelpi/med-electra-small-discriminator and are newly initialized: ['classifier.weight', 'classifier.bias']\nYou should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.\n11/13/2020 22:18:18 - INFO - filelock - Lock 140716208148096 acquired on ./data/cached_train_ElectraTokenizer_128.lock\n11/13/2020 22:18:18 - INFO - utils_ner - Loading features from cached file ./data/cached_train_ElectraTokenizer_128\n11/13/2020 22:18:19 - INFO - filelock - Lock 140716208148096 released on ./data/cached_train_ElectraTokenizer_128.lock\n11/13/2020 22:18:19 - INFO - filelock - Lock 140716208148040 acquired on ./data/cached_dev_ElectraTokenizer_128.lock\n11/13/2020 22:18:19 - INFO - utils_ner - Loading features from cached file ./data/cached_dev_ElectraTokenizer_128\n11/13/2020 22:18:19 - INFO - filelock - Lock 140716208148040 released on ./data/cached_dev_ElectraTokenizer_128.lock\n 0%| | 0/597 [00:00<?, ?it/s]Traceback (most recent call last):\n File \"run_ner.py\", line 311, in <module>\n main()\n File \"run_ner.py\", line 247, in main\n model_path=model_args.model_name_or_path if os.path.isdir(model_args.model_name_or_path) else None\n File \"/home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages/transformers/trainer.py\", line 775, in train\n tr_loss += self.training_step(model, inputs)\n File \"/home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages/transformers/trainer.py\", line 1112, in training_step\n loss = self.compute_loss(model, inputs)\n File \"/home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages/transformers/trainer.py\", line 1136, in compute_loss\n outputs = model(**inputs)\n File \"/home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages/torch/nn/modules/module.py\", line 727, in _call_impl\n result = self.forward(*input, **kwargs)\n File \"/home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages/torch/nn/parallel/data_parallel.py\", line 157, in forward\n inputs, kwargs = self.scatter(inputs, kwargs, self.device_ids)\n File \"/home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages/torch/nn/parallel/data_parallel.py\", line 168, in scatter\n return scatter_kwargs(inputs, kwargs, device_ids, dim=self.dim)\n File \"/home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages/torch/nn/parallel/scatter_gather.py\", line 37, in scatter_kwargs\n kwargs = scatter(kwargs, target_gpus, dim) if kwargs else []\n File \"/home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages/torch/nn/parallel/scatter_gather.py\", line 28, in scatter\n res = scatter_map(inputs)\n File \"/home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages/torch/nn/parallel/scatter_gather.py\", line 19, in scatter_map\n return list(map(type(obj), zip(*map(scatter_map, obj.items()))))\n File \"/home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages/torch/nn/parallel/scatter_gather.py\", line 15, in scatter_map\n return list(zip(*map(scatter_map, obj)))\n File \"/home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages/torch/nn/parallel/scatter_gather.py\", line 13, in scatter_map\n return Scatter.apply(target_gpus, None, dim, obj)\n File \"/home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages/torch/nn/parallel/_functions.py\", line 92, in forward\n outputs = comm.scatter(input, target_gpus, chunk_sizes, ctx.dim, streams)\n File \"/home/ociftci/anaconda3/envs/ner_env/lib/python3.6/site-packages/torch/nn/parallel/comm.py\", line 186, in scatter\n return tuple(torch._C._scatter(tensor, devices, chunk_sizes, dim, streams))\nRuntimeError: CUDA error: out of memory\n 0%| | 0/597 [00:00<?, ?it/s]\n" ], [ "import torch\nprint(torch.__version__)\nprint(torch.cuda.is_available())", "1.7.0+cu101\nTrue\n" ], [ "!lshw -c video", "WARNING: you should run this program as super-user.\n *-display \n description: VGA compatible controller\n product: GV102\n vendor: NVIDIA Corporation\n physical id: 0\n bus info: pci@0000:01:00.0\n version: a1\n width: 64 bits\n clock: 33MHz\n capabilities: vga_controller bus_master cap_list rom\n configuration: driver=nvidia latency=0\n resources: irq:136 memory:78000000-78ffffff memory:60000000-6fffffff memory:76000000-77ffffff ioport:5000(size=128) memory:c0000-dffff\n *-display\n description: VGA compatible controller\n product: GV102\n vendor: NVIDIA Corporation\n physical id: 0\n bus info: pci@0000:05:00.0\n version: a1\n width: 64 bits\n clock: 33MHz\n capabilities: vga_controller bus_master cap_list rom\n configuration: driver=nvidia latency=0\n resources: irq:137 memory:74000000-74ffffff memory:40000000-4fffffff memory:72000000-73ffffff ioport:3000(size=128) memory:75000000-7507ffff\nWARNING: output may be incomplete or inaccurate, you should run this program as super-user.\n" ], [ "!nvcc --version", "nvcc: NVIDIA (R) Cuda compiler driver\r\nCopyright (c) 2005-2018 NVIDIA Corporation\r\nBuilt on Sat_Aug_25_21:08:01_CDT_2018\r\nCuda compilation tools, release 10.0, V10.0.130\r\n" ], [ "!modinfo nvidia", "filename: /lib/modules/5.3.0-46-generic/updates/dkms/nvidia.ko\r\nalias: char-major-195-*\r\nversion: 435.21\r\nsupported: external\r\nlicense: NVIDIA\r\nsrcversion: 2D4143B613DF8762A7AAE09\r\nalias: pci:v000010DEd00000E00sv*sd*bc04sc80i00*\r\nalias: pci:v000010DEd*sv*sd*bc03sc02i00*\r\nalias: pci:v000010DEd*sv*sd*bc03sc00i00*\r\ndepends: ipmi_msghandler\r\nretpoline: Y\r\nname: nvidia\r\nvermagic: 5.3.0-46-generic SMP mod_unload \r\nparm: NvSwitchRegDwords:NvSwitch regkey (charp)\r\nparm: NVreg_Mobile:int\r\nparm: NVreg_ResmanDebugLevel:int\r\nparm: NVreg_RmLogonRC:int\r\nparm: NVreg_ModifyDeviceFiles:int\r\nparm: NVreg_DeviceFileUID:int\r\nparm: NVreg_DeviceFileGID:int\r\nparm: NVreg_DeviceFileMode:int\r\nparm: NVreg_InitializeSystemMemoryAllocations:int\r\nparm: NVreg_UsePageAttributeTable:int\r\nparm: NVreg_MapRegistersEarly:int\r\nparm: NVreg_RegisterForACPIEvents:int\r\nparm: NVreg_EnablePCIeGen3:int\r\nparm: NVreg_EnableMSI:int\r\nparm: NVreg_TCEBypassMode:int\r\nparm: NVreg_EnableStreamMemOPs:int\r\nparm: NVreg_EnableBacklightHandler:int\r\nparm: NVreg_RestrictProfilingToAdminUsers:int\r\nparm: NVreg_PreserveVideoMemoryAllocations:int\r\nparm: NVreg_DynamicPowerManagement:int\r\nparm: NVreg_EnableUserNUMAManagement:int\r\nparm: NVreg_MemoryPoolSize:int\r\nparm: NVreg_KMallocHeapMaxSize:int\r\nparm: NVreg_VMallocHeapMaxSize:int\r\nparm: NVreg_IgnoreMMIOCheck:int\r\nparm: NVreg_NvLinkDisable:int\r\nparm: NVreg_RegisterPCIDriver:int\r\nparm: NVreg_RegistryDwords:charp\r\nparm: NVreg_RegistryDwordsPerDevice:charp\r\nparm: NVreg_RmMsg:charp\r\nparm: NVreg_GpuBlacklist:charp\r\nparm: NVreg_TemporaryFilePath:charp\r\nparm: NVreg_AssignGpus:charp\r\n" ], [ "# 11/06/2020 20:45:35 - INFO - __main__ - eval_accuracy_score = 0.9825284728742295\n# 11/06/2020 20:45:35 - INFO - __main__ - eval_precision = 0.8032166508987701\n# 11/06/2020 20:45:35 - INFO - __main__ - eval_recall = 0.884375\n# 11/06/2020 20:45:35 - INFO - __main__ - eval_f1 = 0.8418443232523549", "_____no_output_____" ], [ "!nvidia-smi", "Thu Nov 12 22:15:55 2020 \r\n+-----------------------------------------------------------------------------+\r\n| NVIDIA-SMI 435.21 Driver Version: 435.21 CUDA Version: 10.1 |\r\n|-------------------------------+----------------------+----------------------+\r\n| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |\r\n| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |\r\n|===============================+======================+======================|\r\n| 0 GeForce RTX 208... Off | 00000000:01:00.0 On | N/A |\r\n| 37% 63C P0 71W / 260W | 195MiB / 11018MiB | 0% Default |\r\n+-------------------------------+----------------------+----------------------+\r\n| 1 GeForce RTX 208... Off | 00000000:05:00.0 Off | N/A |\r\n| 66% 79C P2 256W / 260W | 10910MiB / 11019MiB | 99% Default |\r\n+-------------------------------+----------------------+----------------------+\r\n \r\n+-----------------------------------------------------------------------------+\r\n| Processes: GPU Memory |\r\n| GPU PID Type Process name Usage |\r\n|=============================================================================|\r\n| 0 1280 G /usr/lib/xorg/Xorg 65MiB |\r\n| 0 1428 G /usr/bin/gnome-shell 99MiB |\r\n| 0 1845 G /opt/teamviewer/tv_bin/TeamViewer 17MiB |\r\n| 1 28158 C python 10899MiB |\r\n+-----------------------------------------------------------------------------+\r\n" ], [ "Seq length 256 distilbert-base-uncased", "_____no_output_____" ], [ "torch.version.cuda", "_____no_output_____" ], [ "!export CUDA_VISIBLE_DEVICES=0", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Artificial Intelligence Nanodegree\n\n## Convolutional Neural Networks\n\n---\n\nIn this notebook, we train a CNN on augmented images from the CIFAR-10 database.\n\n### 1. Load CIFAR-10 Database", "_____no_output_____" ] ], [ [ "import keras\nfrom keras.datasets import cifar10\n\n# load the pre-shuffled train and test data\n(x_train, y_train), (x_test, y_test) = cifar10.load_data()", "Using TensorFlow backend.\n" ] ], [ [ "### 2. Visualize the First 24 Training Images", "_____no_output_____" ] ], [ [ "import numpy as np\nimport matplotlib.pyplot as plt\n%matplotlib inline\n\nfig = plt.figure(figsize=(20,5))\nfor i in range(36):\n ax = fig.add_subplot(3, 12, i + 1, xticks=[], yticks=[])\n ax.imshow(np.squeeze(x_train[i]))", "_____no_output_____" ] ], [ [ "### 3. Rescale the Images by Dividing Every Pixel in Every Image by 255", "_____no_output_____" ] ], [ [ "# rescale [0,255] --> [0,1]\nx_train = x_train.astype('float32')/255\nx_test = x_test.astype('float32')/255 ", "_____no_output_____" ] ], [ [ "### 4. Break Dataset into Training, Testing, and Validation Sets", "_____no_output_____" ] ], [ [ "from keras.utils import np_utils\n\n# break training set into training and validation sets\n(x_train, x_valid) = x_train[5000:], x_train[:5000]\n(y_train, y_valid) = y_train[5000:], y_train[:5000]\n\n# one-hot encode the labels\nnum_classes = len(np.unique(y_train))\ny_train = keras.utils.to_categorical(y_train, num_classes)\ny_test = keras.utils.to_categorical(y_test, num_classes)\ny_valid = keras.utils.to_categorical(y_valid, num_classes)\n\n# print shape of training set\nprint('x_train shape:', x_train.shape)\n\n# print number of training, validation, and test images\nprint(x_train.shape[0], 'train samples')\nprint(x_test.shape[0], 'test samples')\nprint(x_valid.shape[0], 'validation samples')", "x_train shape: (45000, 32, 32, 3)\n45000 train samples\n10000 test samples\n5000 validation samples\n" ] ], [ [ "### 5. Create and Configure Augmented Image Generator", "_____no_output_____" ] ], [ [ "from keras.preprocessing.image import ImageDataGenerator\n\ndatagen_train = ImageDataGenerator(width_shift_range=.1, \n height_shift_range=.1, horizontal_flip=True)\n\ndatagen_train.fit(x_train)", "_____no_output_____" ] ], [ [ "### 6. Visualize Original and Augmented Images", "_____no_output_____" ] ], [ [ "fig = plt.figure(figsize=(20,2))\nfig.suptitle('Actual images', fontsize=20)\nfor i in range(10):\n ax = fig.add_subplot(1, 10, i+1, xticks=[], yticks=[])\n ax.imshow(np.squeeze(x_train[i]))\n \nfig = plt.figure(figsize=(20,2))\nfig.suptitle('Augmented images')\nfor x_batch in datagen_train.flow(x_train[:10]):\n for i in range(10):\n ax = fig.add_subplot(1, 10, i+1, xticks=[], yticks=[])\n ax.imshow(x_batch[i])\n break;\n", "_____no_output_____" ] ], [ [ "### 7. Define the Model Architecture ", "_____no_output_____" ] ], [ [ "from keras.models import Sequential\nfrom keras.layers import Conv2D, MaxPooling2D, Flatten, Dense, Dropout\n\nmodel = Sequential()\nmodel.add(Conv2D(filters=16, kernel_size=2, padding='same', activation='relu', \n input_shape=(32, 32, 3)))\nmodel.add(MaxPooling2D(pool_size=2))\nmodel.add(Conv2D(filters=32, kernel_size=2, padding='same', activation='relu'))\nmodel.add(MaxPooling2D(pool_size=2))\nmodel.add(Conv2D(filters=64, kernel_size=2, padding='same', activation='relu'))\nmodel.add(MaxPooling2D(pool_size=2))\nmodel.add(Dropout(0.3))\nmodel.add(Flatten())\nmodel.add(Dense(500, activation='relu'))\nmodel.add(Dropout(0.4))\nmodel.add(Dense(10, activation='softmax'))\n\nmodel.summary()", "_________________________________________________________________\nLayer (type) Output Shape Param # \n=================================================================\nconv2d_1 (Conv2D) (None, 32, 32, 16) 208 \n_________________________________________________________________\nmax_pooling2d_1 (MaxPooling2 (None, 16, 16, 16) 0 \n_________________________________________________________________\nconv2d_2 (Conv2D) (None, 16, 16, 32) 2080 \n_________________________________________________________________\nmax_pooling2d_2 (MaxPooling2 (None, 8, 8, 32) 0 \n_________________________________________________________________\nconv2d_3 (Conv2D) (None, 8, 8, 64) 8256 \n_________________________________________________________________\nmax_pooling2d_3 (MaxPooling2 (None, 4, 4, 64) 0 \n_________________________________________________________________\ndropout_1 (Dropout) (None, 4, 4, 64) 0 \n_________________________________________________________________\nflatten_1 (Flatten) (None, 1024) 0 \n_________________________________________________________________\ndense_1 (Dense) (None, 500) 512500 \n_________________________________________________________________\ndropout_2 (Dropout) (None, 500) 0 \n_________________________________________________________________\ndense_2 (Dense) (None, 10) 5010 \n=================================================================\nTotal params: 528,054\nTrainable params: 528,054\nNon-trainable params: 0\n_________________________________________________________________\n" ] ], [ [ "### 8. Compile the Model ", "_____no_output_____" ] ], [ [ "# compile the model\nmodel.compile(loss='categorical_crossentropy', optimizer='rmsprop', \n metrics=['accuracy'])", "_____no_output_____" ] ], [ [ "### 9. Train the Model ", "_____no_output_____" ] ], [ [ "from keras.callbacks import ModelCheckpoint \n\nbatch_size = 32\nepochs = 100\n\n# train the model\ncheckpointer = ModelCheckpoint(filepath='aug_model.weights.best.mine.hdf5', verbose=1, \n save_best_only=True)\nmodel.fit_generator(datagen_train.flow(x_train, y_train, batch_size=batch_size),\n steps_per_epoch=x_train.shape[0] // batch_size,\n epochs=epochs, verbose=2, callbacks=[checkpointer],\n validation_data=(x_valid, y_valid),\n validation_steps=x_valid.shape[0] // batch_size)", "Epoch 1/100\n - 64s - loss: 1.6690 - acc: 0.3944 - val_loss: 1.5590 - val_acc: 0.4352\n\nEpoch 00001: val_loss improved from inf to 1.55902, saving model to aug_model.weights.best.hdf5\nEpoch 2/100\n - 54s - loss: 1.3915 - acc: 0.4976 - val_loss: 1.1726 - val_acc: 0.5744\n\nEpoch 00002: val_loss improved from 1.55902 to 1.17256, saving model to aug_model.weights.best.hdf5\nEpoch 3/100\n - 52s - loss: 1.2954 - acc: 0.5354 - val_loss: 1.2066 - val_acc: 0.5770\n\nEpoch 00003: val_loss did not improve from 1.17256\nEpoch 4/100\n - 55s - loss: 1.2305 - acc: 0.5648 - val_loss: 1.1527 - val_acc: 0.5934\n\nEpoch 00004: val_loss improved from 1.17256 to 1.15274, saving model to aug_model.weights.best.hdf5\nEpoch 5/100\n - 57s - loss: 1.2038 - acc: 0.5731 - val_loss: 1.1211 - val_acc: 0.6062\n\nEpoch 00005: val_loss improved from 1.15274 to 1.12108, saving model to aug_model.weights.best.hdf5\nEpoch 6/100\n - 57s - loss: 1.1766 - acc: 0.5859 - val_loss: 1.1099 - val_acc: 0.6174\n\nEpoch 00006: val_loss improved from 1.12108 to 1.10988, saving model to aug_model.weights.best.hdf5\nEpoch 7/100\n - 65s - loss: 1.1752 - acc: 0.5914 - val_loss: 0.9883 - val_acc: 0.6580\n\nEpoch 00007: val_loss improved from 1.10988 to 0.98825, saving model to aug_model.weights.best.hdf5\nEpoch 8/100\n - 67s - loss: 1.1651 - acc: 0.5979 - val_loss: 1.0768 - val_acc: 0.6416\n\nEpoch 00008: val_loss did not improve from 0.98825\nEpoch 9/100\n - 66s - loss: 1.1677 - acc: 0.5957 - val_loss: 1.2447 - val_acc: 0.6168\n\nEpoch 00009: val_loss did not improve from 0.98825\nEpoch 10/100\n - 80s - loss: 1.1747 - acc: 0.5972 - val_loss: 1.0569 - val_acc: 0.6564\n\nEpoch 00010: val_loss did not improve from 0.98825\nEpoch 11/100\n - 53s - loss: 1.1842 - acc: 0.5943 - val_loss: 0.9703 - val_acc: 0.6726\n\nEpoch 00011: val_loss improved from 0.98825 to 0.97035, saving model to aug_model.weights.best.hdf5\nEpoch 12/100\n - 71s - loss: 1.2000 - acc: 0.5913 - val_loss: 1.0361 - val_acc: 0.6450\n\nEpoch 00012: val_loss did not improve from 0.97035\nEpoch 13/100\n - 83s - loss: 1.2158 - acc: 0.5863 - val_loss: 1.0435 - val_acc: 0.6390\n\nEpoch 00013: val_loss did not improve from 0.97035\nEpoch 14/100\n - 55s - loss: 1.2252 - acc: 0.5841 - val_loss: 1.1879 - val_acc: 0.6378\n\nEpoch 00014: val_loss did not improve from 0.97035\nEpoch 15/100\n - 56s - loss: 1.2427 - acc: 0.5770 - val_loss: 0.9678 - val_acc: 0.6768\n\nEpoch 00015: val_loss improved from 0.97035 to 0.96785, saving model to aug_model.weights.best.hdf5\nEpoch 16/100\n - 63s - loss: 1.2540 - acc: 0.5767 - val_loss: 1.0361 - val_acc: 0.6394\n\nEpoch 00016: val_loss did not improve from 0.96785\nEpoch 17/100\n - 55s - loss: 1.2676 - acc: 0.5708 - val_loss: 0.9741 - val_acc: 0.6658\n\nEpoch 00017: val_loss did not improve from 0.96785\nEpoch 18/100\n - 59s - loss: 1.2829 - acc: 0.5672 - val_loss: 1.2069 - val_acc: 0.6016\n\nEpoch 00018: val_loss did not improve from 0.96785\nEpoch 19/100\n - 55s - loss: 1.3015 - acc: 0.5610 - val_loss: 1.1444 - val_acc: 0.6146\n\nEpoch 00019: val_loss did not improve from 0.96785\nEpoch 20/100\n - 52s - loss: 1.3232 - acc: 0.5571 - val_loss: 1.2112 - val_acc: 0.5716\n\nEpoch 00020: val_loss did not improve from 0.96785\nEpoch 21/100\n - 55s - loss: 1.3394 - acc: 0.5498 - val_loss: 1.2940 - val_acc: 0.5604\n\nEpoch 00021: val_loss did not improve from 0.96785\nEpoch 22/100\n - 52s - loss: 1.3635 - acc: 0.5419 - val_loss: 1.2010 - val_acc: 0.6032\n\nEpoch 00022: val_loss did not improve from 0.96785\nEpoch 23/100\n - 52s - loss: 1.3709 - acc: 0.5383 - val_loss: 1.1548 - val_acc: 0.6052\n\nEpoch 00023: val_loss did not improve from 0.96785\nEpoch 24/100\n - 52s - loss: 1.3855 - acc: 0.5360 - val_loss: 1.1610 - val_acc: 0.6080\n\nEpoch 00024: val_loss did not improve from 0.96785\nEpoch 25/100\n - 54s - loss: 1.4122 - acc: 0.5219 - val_loss: 1.1643 - val_acc: 0.6190\n\nEpoch 00025: val_loss did not improve from 0.96785\nEpoch 26/100\n - 53s - loss: 1.4311 - acc: 0.5181 - val_loss: 1.3308 - val_acc: 0.5356\n\nEpoch 00026: val_loss did not improve from 0.96785\nEpoch 27/100\n - 53s - loss: 1.4460 - acc: 0.5124 - val_loss: 1.2807 - val_acc: 0.5570\n\nEpoch 00027: val_loss did not improve from 0.96785\nEpoch 28/100\n - 60s - loss: 1.4613 - acc: 0.5112 - val_loss: 1.2505 - val_acc: 0.5740\n\nEpoch 00028: val_loss did not improve from 0.96785\nEpoch 29/100\n - 80s - loss: 1.4706 - acc: 0.5032 - val_loss: 1.5598 - val_acc: 0.4462\n\nEpoch 00029: val_loss did not improve from 0.96785\nEpoch 30/100\n - 79s - loss: 1.5003 - acc: 0.4970 - val_loss: 1.1695 - val_acc: 0.5842\n\nEpoch 00030: val_loss did not improve from 0.96785\nEpoch 31/100\n - 60s - loss: 1.5121 - acc: 0.4939 - val_loss: 1.2706 - val_acc: 0.5850\n\nEpoch 00031: val_loss did not improve from 0.96785\nEpoch 32/100\n - 71s - loss: 1.5207 - acc: 0.4872 - val_loss: 1.1488 - val_acc: 0.6032\n\nEpoch 00032: val_loss did not improve from 0.96785\nEpoch 33/100\n - 85s - loss: 1.5382 - acc: 0.4826 - val_loss: 1.3100 - val_acc: 0.5440\n\nEpoch 00033: val_loss did not improve from 0.96785\nEpoch 34/100\n - 78s - loss: 1.5507 - acc: 0.4763 - val_loss: 1.1904 - val_acc: 0.5924\n\nEpoch 00034: val_loss did not improve from 0.96785\nEpoch 35/100\n - 74s - loss: 1.5715 - acc: 0.4695 - val_loss: 1.4230 - val_acc: 0.5040\n\nEpoch 00035: val_loss did not improve from 0.96785\nEpoch 36/100\n - 71s - loss: 1.5849 - acc: 0.4636 - val_loss: 1.3997 - val_acc: 0.5286\n\nEpoch 00036: val_loss did not improve from 0.96785\nEpoch 37/100\n - 71s - loss: 1.6072 - acc: 0.4556 - val_loss: 1.2331 - val_acc: 0.5720\n\nEpoch 00037: val_loss did not improve from 0.96785\nEpoch 38/100\n - 76s - loss: 1.6077 - acc: 0.4545 - val_loss: 1.3933 - val_acc: 0.5222\n\nEpoch 00038: val_loss did not improve from 0.96785\nEpoch 39/100\n - 66s - loss: 1.6159 - acc: 0.4519 - val_loss: 1.4958 - val_acc: 0.4560\n\nEpoch 00039: val_loss did not improve from 0.96785\nEpoch 40/100\n - 80s - loss: 1.6509 - acc: 0.4409 - val_loss: 1.4325 - val_acc: 0.4758\n\nEpoch 00040: val_loss did not improve from 0.96785\nEpoch 41/100\n - 77s - loss: 1.6628 - acc: 0.4331 - val_loss: 1.3907 - val_acc: 0.5086\n\nEpoch 00041: val_loss did not improve from 0.96785\nEpoch 42/100\n - 91s - loss: 1.6651 - acc: 0.4336 - val_loss: 1.3285 - val_acc: 0.5428\n\nEpoch 00042: val_loss did not improve from 0.96785\nEpoch 43/100\n - 81s - loss: 1.6686 - acc: 0.4311 - val_loss: 1.3000 - val_acc: 0.5452\n\nEpoch 00043: val_loss did not improve from 0.96785\nEpoch 44/100\n - 51s - loss: 1.6724 - acc: 0.4278 - val_loss: 1.3552 - val_acc: 0.5390\n\nEpoch 00044: val_loss did not improve from 0.96785\nEpoch 45/100\n - 40s - loss: 1.6865 - acc: 0.4245 - val_loss: 1.3496 - val_acc: 0.5116\n\nEpoch 00045: val_loss did not improve from 0.96785\nEpoch 46/100\n - 40s - loss: 1.6878 - acc: 0.4194 - val_loss: 1.3678 - val_acc: 0.5272\n\nEpoch 00046: val_loss did not improve from 0.96785\nEpoch 47/100\n - 40s - loss: 1.6947 - acc: 0.4211 - val_loss: 1.3246 - val_acc: 0.5384\n\nEpoch 00047: val_loss did not improve from 0.96785\nEpoch 48/100\n - 41s - loss: 1.6874 - acc: 0.4209 - val_loss: 1.3495 - val_acc: 0.5438\n\nEpoch 00048: val_loss did not improve from 0.96785\nEpoch 49/100\n - 40s - loss: 1.6892 - acc: 0.4225 - val_loss: 1.3779 - val_acc: 0.5306\n\nEpoch 00049: val_loss did not improve from 0.96785\nEpoch 50/100\n - 41s - loss: 1.6878 - acc: 0.4168 - val_loss: 1.3686 - val_acc: 0.5306\n\nEpoch 00050: val_loss did not improve from 0.96785\nEpoch 51/100\n - 43s - loss: 1.6930 - acc: 0.4172 - val_loss: 1.2627 - val_acc: 0.5682\n\nEpoch 00051: val_loss did not improve from 0.96785\nEpoch 52/100\n - 43s - loss: 1.6996 - acc: 0.4125 - val_loss: 1.7034 - val_acc: 0.5202\n\nEpoch 00052: val_loss did not improve from 0.96785\nEpoch 53/100\n - 41s - loss: 1.6946 - acc: 0.4156 - val_loss: 1.3737 - val_acc: 0.5146\n\nEpoch 00053: val_loss did not improve from 0.96785\nEpoch 54/100\n - 41s - loss: 1.6955 - acc: 0.4132 - val_loss: 1.3678 - val_acc: 0.5176\n\nEpoch 00054: val_loss did not improve from 0.96785\nEpoch 55/100\n - 46s - loss: 1.7002 - acc: 0.4104 - val_loss: 1.3815 - val_acc: 0.5266\n\nEpoch 00055: val_loss did not improve from 0.96785\nEpoch 56/100\n - 44s - loss: 1.7032 - acc: 0.4108 - val_loss: 1.5053 - val_acc: 0.4550\n\nEpoch 00056: val_loss did not improve from 0.96785\nEpoch 57/100\n - 41s - loss: 1.7119 - acc: 0.4040 - val_loss: 1.4515 - val_acc: 0.4832\n" ] ], [ [ "### 10. Load the Model with the Best Validation Accuracy", "_____no_output_____" ] ], [ [ "# load the weights that yielded the best validation accuracy\nmodel.load_weights('aug_model.weights.best.mine.hdf5')", "_____no_output_____" ] ], [ [ "### 11. Calculate Classification Accuracy on Test Set", "_____no_output_____" ] ], [ [ "# evaluate and print test accuracy\nscore = model.evaluate(x_test, y_test, verbose=0)\nprint('\\n', 'Test accuracy:', score[1])", "\n Test accuracy: 0.6665\n" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import os, json, sys, time, random\nimport numpy as np\nimport torch\nfrom easydict import EasyDict\nfrom math import floor\nfrom easydict import EasyDict\n\nfrom steves_utils.vanilla_train_eval_test_jig import Vanilla_Train_Eval_Test_Jig\n\nfrom steves_utils.torch_utils import get_dataset_metrics, independent_accuracy_assesment\nfrom steves_models.configurable_vanilla import Configurable_Vanilla\nfrom steves_utils.torch_sequential_builder import build_sequential\nfrom steves_utils.lazy_map import Lazy_Map\nfrom steves_utils.sequence_aggregator import Sequence_Aggregator\n\nfrom steves_utils.stratified_dataset.traditional_accessor import Traditional_Accessor_Factory\n\nfrom steves_utils.cnn_do_report import (\n get_loss_curve,\n get_results_table,\n get_parameters_table,\n get_domain_accuracies,\n)\n\nfrom steves_utils.torch_utils import (\n confusion_by_domain_over_dataloader,\n independent_accuracy_assesment\n)\n\nfrom steves_utils.utils_v2 import (\n per_domain_accuracy_from_confusion,\n get_datasets_base_path\n)\n\n# from steves_utils.ptn_do_report import TBD", "_____no_output_____" ], [ "required_parameters = {\n \"experiment_name\",\n \"lr\",\n \"device\",\n \"dataset_seed\",\n \"seed\",\n \"labels\",\n \"domains_target\",\n \"domains_source\",\n \"num_examples_per_domain_per_label_source\",\n \"num_examples_per_domain_per_label_target\",\n \"batch_size\",\n \"n_epoch\",\n \"patience\",\n \"criteria_for_best\",\n \"normalize_source\",\n \"normalize_target\",\n \"x_net\",\n \"NUM_LOGS_PER_EPOCH\",\n \"BEST_MODEL_PATH\",\n \"pickle_name_source\",\n \"pickle_name_target\",\n \"torch_default_dtype\",\n}", "_____no_output_____" ], [ "from steves_utils.ORACLE.utils_v2 import (\n ALL_SERIAL_NUMBERS,\n ALL_DISTANCES_FEET_NARROWED,\n)\n\nstandalone_parameters = {}\nstandalone_parameters[\"experiment_name\"] = \"MANUAL CORES CNN\"\nstandalone_parameters[\"lr\"] = 0.0001\nstandalone_parameters[\"device\"] = \"cuda\"\n\nstandalone_parameters[\"dataset_seed\"] = 1337\nstandalone_parameters[\"seed\"] = 1337\nstandalone_parameters[\"labels\"] = ALL_SERIAL_NUMBERS\n\nstandalone_parameters[\"domains_source\"] = [8,32,50]\nstandalone_parameters[\"domains_target\"] = [14,20,26,38,44,]\n\nstandalone_parameters[\"num_examples_per_domain_per_label_source\"]=-1\nstandalone_parameters[\"num_examples_per_domain_per_label_target\"]=-1\n\nstandalone_parameters[\"pickle_name_source\"] = \"oracle.Run1_framed_2000Examples_stratified_ds.2022A.pkl\"\nstandalone_parameters[\"pickle_name_target\"] = \"oracle.Run2_framed_2000Examples_stratified_ds.2022A.pkl\"\n\nstandalone_parameters[\"torch_default_dtype\"] = \"torch.float32\" \n\nstandalone_parameters[\"batch_size\"]=128\n\nstandalone_parameters[\"n_epoch\"] = 3\n\nstandalone_parameters[\"patience\"] = 10\n\nstandalone_parameters[\"criteria_for_best\"] = \"target_accuracy\"\nstandalone_parameters[\"normalize_source\"] = False\nstandalone_parameters[\"normalize_target\"] = False\n\nstandalone_parameters[\"x_net\"] = [\n {\"class\": \"nnReshape\", \"kargs\": {\"shape\":[-1, 1, 2, 256]}},\n {\"class\": \"Conv2d\", \"kargs\": { \"in_channels\":1, \"out_channels\":256, \"kernel_size\":(1,7), \"bias\":False, \"padding\":(0,3), },},\n {\"class\": \"ReLU\", \"kargs\": {\"inplace\": True}},\n {\"class\": \"BatchNorm2d\", \"kargs\": {\"num_features\":256}},\n\n {\"class\": \"Conv2d\", \"kargs\": { \"in_channels\":256, \"out_channels\":80, \"kernel_size\":(2,7), \"bias\":True, \"padding\":(0,3), },},\n {\"class\": \"ReLU\", \"kargs\": {\"inplace\": True}},\n {\"class\": \"BatchNorm2d\", \"kargs\": {\"num_features\":80}},\n {\"class\": \"Flatten\", \"kargs\": {}},\n\n {\"class\": \"Linear\", \"kargs\": {\"in_features\": 80*256, \"out_features\": 256}}, # 80 units per IQ pair\n {\"class\": \"ReLU\", \"kargs\": {\"inplace\": True}},\n {\"class\": \"BatchNorm1d\", \"kargs\": {\"num_features\":256}},\n\n {\"class\": \"Linear\", \"kargs\": {\"in_features\": 256, \"out_features\": len(standalone_parameters[\"labels\"])}},\n]\n\nstandalone_parameters[\"NUM_LOGS_PER_EPOCH\"] = 10\nstandalone_parameters[\"BEST_MODEL_PATH\"] = \"./best_model.pth\"", "_____no_output_____" ], [ "# Parameters\nparameters = {\n \"experiment_name\": \"cnn_1:oracle.run1.framed\",\n \"labels\": [\n \"3123D52\",\n \"3123D65\",\n \"3123D79\",\n \"3123D80\",\n \"3123D54\",\n \"3123D70\",\n \"3123D7B\",\n \"3123D89\",\n \"3123D58\",\n \"3123D76\",\n \"3123D7D\",\n \"3123EFE\",\n \"3123D64\",\n \"3123D78\",\n \"3123D7E\",\n \"3124E4A\",\n ],\n \"domains_source\": [8, 32, 50],\n \"domains_target\": [14, 20, 26, 38, 44],\n \"pickle_name_source\": \"oracle.Run1_framed_2000Examples_stratified_ds.2022A.pkl\",\n \"pickle_name_target\": \"oracle.Run1_framed_2000Examples_stratified_ds.2022A.pkl\",\n \"device\": \"cuda\",\n \"lr\": 0.0001,\n \"batch_size\": 128,\n \"normalize_source\": False,\n \"normalize_target\": False,\n \"num_examples_per_domain_per_label_source\": -1,\n \"num_examples_per_domain_per_label_target\": -1,\n \"torch_default_dtype\": \"torch.float32\",\n \"n_epoch\": 50,\n \"patience\": 3,\n \"criteria_for_best\": \"target_accuracy\",\n \"x_net\": [\n {\"class\": \"nnReshape\", \"kargs\": {\"shape\": [-1, 1, 2, 256]}},\n {\n \"class\": \"Conv2d\",\n \"kargs\": {\n \"in_channels\": 1,\n \"out_channels\": 256,\n \"kernel_size\": [1, 7],\n \"bias\": False,\n \"padding\": [0, 3],\n },\n },\n {\"class\": \"ReLU\", \"kargs\": {\"inplace\": True}},\n {\"class\": \"BatchNorm2d\", \"kargs\": {\"num_features\": 256}},\n {\n \"class\": \"Conv2d\",\n \"kargs\": {\n \"in_channels\": 256,\n \"out_channels\": 80,\n \"kernel_size\": [2, 7],\n \"bias\": True,\n \"padding\": [0, 3],\n },\n },\n {\"class\": \"ReLU\", \"kargs\": {\"inplace\": True}},\n {\"class\": \"BatchNorm2d\", \"kargs\": {\"num_features\": 80}},\n {\"class\": \"Flatten\", \"kargs\": {}},\n {\"class\": \"Linear\", \"kargs\": {\"in_features\": 20480, \"out_features\": 256}},\n {\"class\": \"ReLU\", \"kargs\": {\"inplace\": True}},\n {\"class\": \"BatchNorm1d\", \"kargs\": {\"num_features\": 256}},\n {\"class\": \"Linear\", \"kargs\": {\"in_features\": 256, \"out_features\": 16}},\n ],\n \"NUM_LOGS_PER_EPOCH\": 10,\n \"BEST_MODEL_PATH\": \"./best_model.pth\",\n \"dataset_seed\": 1337,\n \"seed\": 1337,\n}\n", "_____no_output_____" ], [ "# Set this to True if you want to run this template directly\nSTANDALONE = False\nif STANDALONE:\n print(\"parameters not injected, running with standalone_parameters\")\n parameters = standalone_parameters\n\nif not 'parameters' in locals() and not 'parameters' in globals():\n raise Exception(\"Parameter injection failed\")\n\n#Use an easy dict for all the parameters\np = EasyDict(parameters)\n\nsupplied_keys = set(p.keys())\n\nif supplied_keys != required_parameters:\n print(\"Parameters are incorrect\")\n if len(supplied_keys - required_parameters)>0: print(\"Shouldn't have:\", str(supplied_keys - required_parameters))\n if len(required_parameters - supplied_keys)>0: print(\"Need to have:\", str(required_parameters - supplied_keys))\n raise RuntimeError(\"Parameters are incorrect\")\n\n", "_____no_output_____" ], [ "###################################\n# Set the RNGs and make it all deterministic\n###################################\nnp.random.seed(p.seed)\nrandom.seed(p.seed)\ntorch.manual_seed(p.seed)\n\ntorch.use_deterministic_algorithms(True) ", "_____no_output_____" ], [ "torch.set_default_dtype(eval(p.torch_default_dtype))", "_____no_output_____" ], [ "###################################\n# Build the network(s)\n# Note: It's critical to do this AFTER setting the RNG\n###################################\nx_net = build_sequential(p.x_net)", "_____no_output_____" ], [ "start_time_secs = time.time()", "_____no_output_____" ], [ "def wrap_in_dataloader(p, ds):\n return torch.utils.data.DataLoader(\n ds,\n batch_size=p.batch_size,\n shuffle=True,\n num_workers=1,\n persistent_workers=True,\n prefetch_factor=50,\n pin_memory=True\n )\n\ntaf_source = Traditional_Accessor_Factory(\n labels=p.labels,\n domains=p.domains_source,\n num_examples_per_domain_per_label=p.num_examples_per_domain_per_label_source,\n pickle_path=os.path.join(get_datasets_base_path(), p.pickle_name_source),\n seed=p.dataset_seed\n)\ntrain_original_source, val_original_source, test_original_source = \\\n taf_source.get_train(), taf_source.get_val(), taf_source.get_test()\n\n\ntaf_target = Traditional_Accessor_Factory(\n labels=p.labels,\n domains=p.domains_target,\n num_examples_per_domain_per_label=p.num_examples_per_domain_per_label_source,\n pickle_path=os.path.join(get_datasets_base_path(), p.pickle_name_target),\n seed=p.dataset_seed\n)\ntrain_original_target, val_original_target, test_original_target = \\\n taf_target.get_train(), taf_target.get_val(), taf_target.get_test()\n\n\n# For CNN We only use X and Y. And we only train on the source.\n# Properly form the data using a transform lambda and Lazy_Map. Finally wrap them in a dataloader\n\ntransform_lambda = lambda ex: ex[:2] # Strip the tuple to just (x,y)\n\n\ntrain_processed_source = wrap_in_dataloader(\n p,\n Lazy_Map(train_original_source, transform_lambda)\n)\nval_processed_source = wrap_in_dataloader(\n p,\n Lazy_Map(val_original_source, transform_lambda)\n)\ntest_processed_source = wrap_in_dataloader(\n p,\n Lazy_Map(test_original_source, transform_lambda)\n)\n\ntrain_processed_target = wrap_in_dataloader(\n p,\n Lazy_Map(train_original_target, transform_lambda)\n)\nval_processed_target = wrap_in_dataloader(\n p,\n Lazy_Map(val_original_target, transform_lambda)\n)\ntest_processed_target = wrap_in_dataloader(\n p,\n Lazy_Map(test_original_target, transform_lambda)\n)\n\n\n\ndatasets = EasyDict({\n \"source\": {\n \"original\": {\"train\":train_original_source, \"val\":val_original_source, \"test\":test_original_source},\n \"processed\": {\"train\":train_processed_source, \"val\":val_processed_source, \"test\":test_processed_source}\n },\n \"target\": {\n \"original\": {\"train\":train_original_target, \"val\":val_original_target, \"test\":test_original_target},\n \"processed\": {\"train\":train_processed_target, \"val\":val_processed_target, \"test\":test_processed_target}\n },\n})", "_____no_output_____" ], [ "ep = next(iter(test_processed_target))\nep[0].dtype", "_____no_output_____" ], [ "model = Configurable_Vanilla(\n x_net=x_net,\n label_loss_object=torch.nn.NLLLoss(),\n learning_rate=p.lr\n)", "_____no_output_____" ], [ "jig = Vanilla_Train_Eval_Test_Jig(\n model=model,\n path_to_best_model=p.BEST_MODEL_PATH,\n device=p.device,\n label_loss_object=torch.nn.NLLLoss(),\n)\n\njig.train(\n train_iterable=datasets.source.processed.train,\n source_val_iterable=datasets.source.processed.val,\n target_val_iterable=datasets.target.processed.val,\n patience=p.patience,\n num_epochs=p.n_epoch,\n num_logs_per_epoch=p.NUM_LOGS_PER_EPOCH,\n criteria_for_best=p.criteria_for_best\n)", "epoch: 1, [batch: 1 / 525], examples_per_second: 814.0396, train_label_loss: 2.7947, \n" ], [ "total_experiment_time_secs = time.time() - start_time_secs", "_____no_output_____" ], [ "source_test_label_accuracy, source_test_label_loss = jig.test(datasets.source.processed.test)\ntarget_test_label_accuracy, target_test_label_loss = jig.test(datasets.target.processed.test)\n\nsource_val_label_accuracy, source_val_label_loss = jig.test(datasets.source.processed.val)\ntarget_val_label_accuracy, target_val_label_loss = jig.test(datasets.target.processed.val)\n\nhistory = jig.get_history()\n\ntotal_epochs_trained = len(history[\"epoch_indices\"])\n\nval_dl = wrap_in_dataloader(p, Sequence_Aggregator((datasets.source.original.val, datasets.target.original.val)))\n\nconfusion = confusion_by_domain_over_dataloader(model, p.device, val_dl, forward_uses_domain=False)\nper_domain_accuracy = per_domain_accuracy_from_confusion(confusion)\n\n# Add a key to per_domain_accuracy for if it was a source domain\nfor domain, accuracy in per_domain_accuracy.items():\n per_domain_accuracy[domain] = {\n \"accuracy\": accuracy,\n \"source?\": domain in p.domains_source\n }\n\n# Do an independent accuracy assesment JUST TO BE SURE!\n# _source_test_label_accuracy = independent_accuracy_assesment(model, datasets.source.processed.test, p.device)\n# _target_test_label_accuracy = independent_accuracy_assesment(model, datasets.target.processed.test, p.device)\n# _source_val_label_accuracy = independent_accuracy_assesment(model, datasets.source.processed.val, p.device)\n# _target_val_label_accuracy = independent_accuracy_assesment(model, datasets.target.processed.val, p.device)\n\n# assert(_source_test_label_accuracy == source_test_label_accuracy)\n# assert(_target_test_label_accuracy == target_test_label_accuracy)\n# assert(_source_val_label_accuracy == source_val_label_accuracy)\n# assert(_target_val_label_accuracy == target_val_label_accuracy)\n\n###################################\n# Write out the results\n###################################\n\nexperiment = {\n \"experiment_name\": p.experiment_name,\n \"parameters\": p,\n \"results\": {\n \"source_test_label_accuracy\": source_test_label_accuracy,\n \"source_test_label_loss\": source_test_label_loss,\n \"target_test_label_accuracy\": target_test_label_accuracy,\n \"target_test_label_loss\": target_test_label_loss,\n \"source_val_label_accuracy\": source_val_label_accuracy,\n \"source_val_label_loss\": source_val_label_loss,\n \"target_val_label_accuracy\": target_val_label_accuracy,\n \"target_val_label_loss\": target_val_label_loss,\n \"total_epochs_trained\": total_epochs_trained,\n \"total_experiment_time_secs\": total_experiment_time_secs,\n \"confusion\": confusion,\n \"per_domain_accuracy\": per_domain_accuracy,\n },\n \"history\": history,\n \"dataset_metrics\": get_dataset_metrics(datasets, \"cnn\"),\n}", "_____no_output_____" ], [ "get_loss_curve(experiment)", "_____no_output_____" ], [ "get_results_table(experiment)", "_____no_output_____" ], [ "get_domain_accuracies(experiment)", "_____no_output_____" ], [ "print(\"Source Test Label Accuracy:\", experiment[\"results\"][\"source_test_label_accuracy\"], \"Target Test Label Accuracy:\", experiment[\"results\"][\"target_test_label_accuracy\"])\nprint(\"Source Val Label Accuracy:\", experiment[\"results\"][\"source_val_label_accuracy\"], \"Target Val Label Accuracy:\", experiment[\"results\"][\"target_val_label_accuracy\"])", "Source Test Label Accuracy: 0.9120833333333334 Target Test Label Accuracy: 0.083875\nSource Val Label Accuracy: 0.9145138888888888 Target Val Label Accuracy: 0.08370833333333333\n" ], [ "json.dumps(experiment)", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "import pandas as pd", "_____no_output_____" ], [ "# 1. Create a DataFrame for the 201908-citibike-tripdata data. \ncitibike = pd.DataFrame(pd.read_csv('201908-citibike-tripdata.csv'))\ncitibike.head()", "_____no_output_____" ], [ "# 2. Check the datatypes of your columns. \ncitibike.dtypes", "_____no_output_____" ], [ "# 3. Convert the 'tripduration' column to datetime datatype.\ncitibike['tripduration'] = pd.to_datetime(citibike['tripduration'], unit='m')", "_____no_output_____" ], [ "# 4. Check the datatypes of your columns. \ncitibike.dtypes", "_____no_output_____" ], [ "citibike.head()", "_____no_output_____" ], [ "# 5. Export the Dataframe as a new CSV file without the index.\ncitibike.to_csv('citibike_modified.csv', index = False)", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Introducing Scikit-Learn", "_____no_output_____" ], [ "There are several Python libraries which provide solid implementations of a range of machine learning algorithms.\nOne of the best known is [Scikit-Learn](http://scikit-learn.org), a package that provides efficient versions of a large number of common algorithms.\nScikit-Learn is characterized by a clean, uniform, and streamlined API, as well as by very useful and complete [online documentation](https://scikit-learn.org/stable/documentation.html).\nA benefit of this uniformity is that once you understand the basic use and syntax of Scikit-Learn for one type of model, switching to a new model or algorithm is very straightforward.\n\nThis section provides an overview of the Scikit-Learn API.\n\nWe will start by covering *data representation* in Scikit-Learn, followed by covering the *Estimator* API, and finally go through a couple examples.", "_____no_output_____" ], [ "## Data Representation in Scikit-Learn", "_____no_output_____" ], [ "Machine learning is about creating models from data: for that reason, we'll start by discussing how data can be represented in order to be understood by the computer.\nThe best way to think about data within Scikit-Learn is in terms of tables of data.", "_____no_output_____" ], [ "### Data as table\n\nA basic table is a two-dimensional grid of data, in which the rows represent individual elements of the dataset, and the columns represent quantities related to each of these elements.\nFor example, consider the [Iris dataset](https://en.wikipedia.org/wiki/Iris_flower_data_set), famously analyzed by Ronald Fisher in 1936.\nWe can download this dataset in the form of a Pandas ``DataFrame`` using the [seaborn](http://seaborn.pydata.org/) library:", "_____no_output_____" ] ], [ [ "import pandas as pd\nimport numpy as np\nfrom IPython.display import Pretty as disp\nhint = 'https://raw.githubusercontent.com/soltaniehha/Business-Analytics/master/docs/hints/' # path to hints on GitHub\nimport matplotlib.pyplot as plt\nimport seaborn as sns\n%matplotlib inline\nsns.set(rc={'figure.figsize':(10,8)}) # Figure size", "/usr/local/lib/python3.6/dist-packages/statsmodels/tools/_testing.py:19: FutureWarning: pandas.util.testing is deprecated. Use the functions in the public API at pandas.testing instead.\n import pandas.util.testing as tm\n" ], [ "iris = sns.load_dataset('iris')\niris.head()", "_____no_output_____" ] ], [ [ "<img src=\"https://github.com/soltaniehha/Business-Analytics/blob/master/figs/11-01-Petal-sepal.jpg?raw=true\" width=\"300\" align=\"center\"/>", "_____no_output_____" ] ], [ [ "iris.species.unique()", "_____no_output_____" ] ], [ [ "Here each row of the data refers to a single observed flower, and the number of rows is the total number of flowers in the dataset.\nIn general, we will refer to the rows of the matrix as *samples*, and the number of rows as ``n_samples``.\n\nLikewise, each column of the data refers to a particular quantitative piece of information that describes each sample.\nIn general, we will refer to the columns of the matrix as *features*, and the number of columns as ``n_features``.", "_____no_output_____" ], [ "#### Features matrix\n\nThis table layout makes clear that the information can be thought of as a two-dimensional numerical array or matrix, which we will call the *features matrix*.\nBy convention, this features matrix is often stored in a variable named ``X``.\nThe features matrix is assumed to be two-dimensional, with shape ``[n_samples, n_features]``, and is most often contained in a NumPy array or a Pandas ``DataFrame``.\n\nThe samples (i.e., rows) always refer to the individual objects described by the dataset.\nFor example, the sample might be a flower, a person, a document, an image, a sound file, a video, an astronomical object, or anything else you can describe with a set of quantitative measurements.\n\nThe features (i.e., columns) always refer to the distinct observations that describe each sample in a quantitative manner.\nFeatures are generally real-valued, but may be Boolean or discrete-valued in some cases.", "_____no_output_____" ], [ "#### Target array\n\nIn addition to the feature matrix ``X``, we also generally work with a *label* or *target* array, which by convention we will usually call ``y``.\nThe target array is usually one dimensional, with length ``n_samples``, and is generally contained in a NumPy array or Pandas ``Series``.\nThe target array may have continuous numerical values, or discrete classes/labels.\n\nOften one point of confusion is how the target array differs from the other features columns. The distinguishing feature of the target array is that it is usually the quantity we want to *predict from the data*: in statistical terms, it is the dependent variable.\nFor example, in the preceding data we may wish to construct a model that can predict the species of flower based on the other measurements; in this case, the ``species`` column would be considered the target array.\n\nWith this target array in mind, we can use Seaborn to conveniently visualize the data:", "_____no_output_____" ] ], [ [ "sns.pairplot(iris, hue='species', height=2.5);", "_____no_output_____" ] ], [ [ "For use in Scikit-Learn, we will extract the features matrix and target array from the ``DataFrame``, which we can do using some of the Pandas ``DataFrame`` operations we've learned:", "_____no_output_____" ] ], [ [ "X_iris = iris.drop('species', axis=1)\nX_iris.shape", "_____no_output_____" ], [ "y_iris = iris['species']\ny_iris.shape", "_____no_output_____" ] ], [ [ "To summarize, the expected layout of features and target values is visualized in the following diagram:", "_____no_output_____" ], [ "<img src=\"https://github.com/soltaniehha/Business-Analytics/blob/master/figs/11-01-samples-features.png?raw=true\" width=\"700\" align=\"center\"/>", "_____no_output_____" ], [ "With this data properly formatted, we can move on to consider the *estimator* API of Scikit-Learn:", "_____no_output_____" ], [ "## Scikit-Learn's Estimator API", "_____no_output_____" ], [ "The Scikit-Learn API is designed with the following guiding principles in mind, as outlined in the [Scikit-Learn API paper(2013)](http://arxiv.org/abs/1309.0238):\n\n- *Consistency*: All objects share a common interface drawn from a limited set of methods, with consistent documentation.\n\n- *Inspection*: All specified parameter values are exposed as public attributes.\n\n- *Limited object hierarchy*: Only algorithms are represented by Python classes; datasets are represented\n in standard formats (NumPy arrays, Pandas ``DataFrame``s) and parameter\n names use standard Python strings.\n\n- *Composition*: Many machine learning tasks can be expressed as sequences of more fundamental algorithms,\n and Scikit-Learn makes use of this wherever possible.\n\n- *Sensible defaults*: When models require user-specified parameters, the library defines an appropriate default value.\n\nIn practice, these principles make Scikit-Learn very easy to use, once the basic principles are understood.\nEvery machine learning algorithm in Scikit-Learn is implemented via the Estimator API, which provides a consistent interface for a wide range of machine learning applications.", "_____no_output_____" ], [ "### Basics of the API\n\nMost commonly, the steps in using the Scikit-Learn estimator API are as follows\n(we will step through a couple of detailed examples in the sections that follow).\n\n1. Choose a class of model by importing the appropriate estimator class from Scikit-Learn.\n2. Choose model hyperparameters by instantiating this class with desired values.\n3. Arrange data into a features matrix and target vector following the discussion above.\n4. Fit the model to your data by calling the ``fit()`` method of the model instance.\n5. Apply the Model to new data:\n - For supervised learning, often we predict labels for unknown data using the ``predict()`` method.\n - For unsupervised learning, we often transform or infer properties of the data using the ``transform()`` or ``predict()`` method.\n\nWe will now step through simple examples of applying supervised learning methods.", "_____no_output_____" ], [ "### Supervised learning example: Simple linear regression\n\nAs an example of this process, let's consider a simple linear regression—that is, the common case of fitting a line to $(x, y)$ data.\nWe will use the following columns from `iris` for our regression example: `petal_width` & `petal_length`", "_____no_output_____" ] ], [ [ "X = iris[['petal_width']]\ny = iris[['petal_length']]\nplt.scatter(X, y);", "_____no_output_____" ] ], [ [ "With this data in place, we can use the recipe outlined earlier. Let's walk through the process: ", "_____no_output_____" ], [ "#### 1. Choose a class of model\n\nIn Scikit-Learn, every class of model is represented by a Python class.\nSo, for example, if we would like to compute a simple linear regression model, we can import the linear regression class:", "_____no_output_____" ] ], [ [ "from sklearn.linear_model import LinearRegression", "_____no_output_____" ] ], [ [ "Note that other more general linear regression models exist as well; you can read more about them in the [``sklearn.linear_model`` module documentation](http://Scikit-Learn.org/stable/modules/linear_model.html).", "_____no_output_____" ], [ "#### 2. Choose model hyperparameters\n\nAn important point is that *a class of model is not the same as an instance of a model*.\n\nOnce we have decided on our model class, there are still some options open to us.\nDepending on the model class we are working with, we might need to answer one or more questions like the following:\n\n- Would we like to fit for the offset (i.e., *y*-intercept)?\n- Would we like the model to be normalized?\n- Would we like to preprocess our features to add model flexibility?\n- What degree of regularization would we like to use in our model?\n- How many model components would we like to use?\n\nThese are examples of the important choices that must be made *once the model class is selected*.\nThese choices are often represented as *hyperparameters*, or parameters that must be set before the model is fit to data.\nIn Scikit-Learn, hyperparameters are chosen by passing values at model instantiation.\n\nFor our linear regression example, we can instantiate the ``LinearRegression`` class and specify that we would like to fit the intercept using the ``fit_intercept`` hyperparameter:", "_____no_output_____" ] ], [ [ "model = LinearRegression(fit_intercept=True)\nmodel", "_____no_output_____" ] ], [ [ "Keep in mind that when the model is instantiated, the only action is the storing of these hyperparameter values.\nIn particular, we have not yet applied the model to any data: the Scikit-Learn API makes very clear the distinction between *choice of model* and *application of model to data*.", "_____no_output_____" ], [ "#### 3. Arrange data into a features matrix and target vector\n\nPreviously we detailed the Scikit-Learn data representation, which requires a two-dimensional features matrix and a one-dimensional target array.\nHere our target variable ``y`` is already in the correct form (a length-``n_samples`` array). Our features matrix is also in the right shape since we only have 1 feature it is a matrix of size ``[n_samples, n_features]``.\nLet's check the shapes:", "_____no_output_____" ] ], [ [ "print(X.shape)\nprint(y.shape)", "(150, 1)\n(150, 1)\n" ] ], [ [ "#### 4. Fit the model to your data\n\nNow it is time to apply our model to data.\nThis can be done with the ``fit()`` method of the model:", "_____no_output_____" ] ], [ [ "model.fit(X, y)", "_____no_output_____" ] ], [ [ "This ``fit()`` command causes a number of model-dependent internal computations to take place, and the results of these computations are stored in model-specific attributes that the user can explore.\nIn Scikit-Learn, by convention all model parameters that were learned during the ``fit()`` process have trailing underscores; for example in this linear model, we have the following:", "_____no_output_____" ] ], [ [ "model.coef_", "_____no_output_____" ], [ "model.intercept_", "_____no_output_____" ] ], [ [ "These two parameters represent the slope and intercept of the simple linear fit to the data.\nComparing to the data definition, we see that they are very close to the input slope of 2.2 and intercept of 1.\n\nOne question that frequently comes up regards the uncertainty in such internal model parameters.\nIn general, Scikit-Learn does not provide tools to draw conclusions from internal model parameters themselves: interpreting model parameters is much more a *statistical modeling* question than a *machine learning* question.\nMachine learning rather focuses on what the model *predicts*.\nIf you would like to dive into the meaning of fit parameters within the model, other tools are available, including the [Statsmodels Python package](http://statsmodels.sourceforge.net/).", "_____no_output_____" ], [ "#### 5. Predict labels for unknown data\n\nOnce the model is trained, the main task of supervised machine learning is to evaluate it based on what it says about new data that was not part of the training set.\nIn Scikit-Learn, this can be done using the ``predict()`` method.\nFor the sake of this example, our \"new data\" will be a grid of `x` values, and we will ask what `y` values the model predicts:", "_____no_output_____" ] ], [ [ "xfit = np.linspace(0, 2.5)\nxfit = pd.DataFrame(xfit)\nxfit.shape", "_____no_output_____" ] ], [ [ "We have coerced these *x* values into a ``[n_samples, n_features]`` features matrix, after which we can feed it to the model:", "_____no_output_____" ] ], [ [ "yfit = model.predict(xfit)", "_____no_output_____" ] ], [ [ "Finally, let's visualize the results by plotting first the raw data, and then this model fit:", "_____no_output_____" ] ], [ [ "plt.scatter(X, y)\nplt.plot(xfit, yfit, c='gray')\nplt.xlabel('petal_width')\nplt.ylabel('petal_length');", "_____no_output_____" ] ], [ [ "Typically the efficacy of the model is evaluated by comparing its results to some known baseline, as we will see in the next example", "_____no_output_____" ], [ "### Supervised learning example: Iris classification\n\nLet's take a look at another example of this process, using the Iris dataset we discussed earlier.\nOur question will be this: given a model trained on a portion of the Iris data, how well can we predict the remaining labels?\n\nFor this task, we will use an extremely simple generative model known as Gaussian naive Bayes, which proceeds by assuming each class is drawn from an axis-aligned Gaussian distribution.\nBecause it is so fast and has no hyperparameters to choose, Gaussian naive Bayes is often a good model to use as a baseline classification, before exploring whether improvements can be found through more sophisticated models.\n\nWe would like to evaluate the model on data it has not seen before, and so we will split the data into a *training set* and a *testing set*.\nThis could be done by hand, but it is more convenient to use the ``train_test_split`` utility function:", "_____no_output_____" ] ], [ [ "from sklearn.model_selection import train_test_split\n\nXtrain, Xtest, ytrain, ytest = train_test_split(X_iris, y_iris, test_size=0.3, random_state=833)", "_____no_output_____" ] ], [ [ "With the data arranged, we can follow our recipe to predict the labels:", "_____no_output_____" ] ], [ [ "from sklearn.naive_bayes import GaussianNB # 1. choose model class\nmodel = GaussianNB() # 2. instantiate model\nmodel.fit(Xtrain, ytrain) # 3. fit model to data\ny_model = model.predict(Xtest) # 4. predict on new data", "_____no_output_____" ] ], [ [ "Finally, we can use the ``accuracy_score`` utility to see the fraction of predicted labels that match their true value:", "_____no_output_____" ] ], [ [ "from sklearn.metrics import accuracy_score\naccuracy_score(ytest, y_model)", "_____no_output_____" ] ], [ [ "With an accuracy topping 93%, we see that even this very naive classification algorithm is effective for this particular dataset!\n\nTo learn more about Gaussian naive Bayes check out [this YouTube video](https://www.youtube.com/watch?v=r1in0YNetG8).", "_____no_output_____" ], [ "# Your Turn\nFor this exercise we are going to make some predictions using Telco customer churn data. Our goal is to make a simple model that can predict whether a customer will churn or not based on the historical data. We will follow the steps above.\n\nBut first, let's load the data:", "_____no_output_____" ] ], [ [ "df = pd.read_csv('https://raw.githubusercontent.com/soltaniehha/Business-Analytics/master/data/Telco-Customer-Churn.csv')\ndf.head(3)", "_____no_output_____" ] ], [ [ "Column `TotalCharges` has 11 rows with an empty string (\" \"). Replace these values by `0` as they represent new customers that haven't received a bill yet. Once you replaced the values, convert that column to a `float32` data type.", "_____no_output_____" ] ], [ [ "# Your answer goes here\n", "_____no_output_____" ], [ "# SOLUTION: Uncomment and execute the cell below to get help\n#disp(hint + '09-02-TotalCharges')", "_____no_output_____" ] ], [ [ "If we check the `df.info()` now we should see that `TotalCharges` is now a `float32`:", "_____no_output_____" ] ], [ [ "df.info()", "<class 'pandas.core.frame.DataFrame'>\nRangeIndex: 7043 entries, 0 to 7042\nData columns (total 21 columns):\n # Column Non-Null Count Dtype \n--- ------ -------------- ----- \n 0 customerID 7043 non-null object \n 1 gender 7043 non-null object \n 2 SeniorCitizen 7043 non-null int64 \n 3 Partner 7043 non-null object \n 4 Dependents 7043 non-null object \n 5 tenure 7043 non-null int64 \n 6 PhoneService 7043 non-null object \n 7 MultipleLines 7043 non-null object \n 8 InternetService 7043 non-null object \n 9 OnlineSecurity 7043 non-null object \n 10 OnlineBackup 7043 non-null object \n 11 DeviceProtection 7043 non-null object \n 12 TechSupport 7043 non-null object \n 13 StreamingTV 7043 non-null object \n 14 StreamingMovies 7043 non-null object \n 15 Contract 7043 non-null object \n 16 PaperlessBilling 7043 non-null object \n 17 PaymentMethod 7043 non-null object \n 18 MonthlyCharges 7043 non-null float64\n 19 TotalCharges 7043 non-null float32\n 20 Churn 7043 non-null object \ndtypes: float32(1), float64(1), int64(2), object(17)\nmemory usage: 1.1+ MB\n" ] ], [ [ "Before we get to splitting our dataset into train/test we would have to make sure all of our values are numerical as most of ML algorithms work with numerical values only. \n\n### How to Convert Categorical Data to Numerical Data?\nIn order to do this we have to convert all of the categorical variables to numerical values. This can be done with a process called one-hot encoding. Take the \"Churn\" column as an example. We have two unique values: \"Yes\"/\"No\". One-hot encoding will create two variables, one called `Churn_Yes` and the other one `Churn_No`. We will go from\n\n| Churn |\n|--|\n|Yes|\n|No|\n|Yes|\n|...|\n\nto\n\n|Churn_Yes | Churn_No |\n|--|--|\n|1|0|\n|0|1|\n|1|0|\n|...|...|\n\nAs you can see, having two variables is redundant since they mirror each other. So, in any one-hot encoding scenario, we need `n-1` variables for a categorical variable that had `n` categories.\n\nBelow, we will use a *pandas* function called `get_dummies()`. If we want *pandas* to automcatically drop one of the extra variables for us we use the `drop_first=True` argument. ", "_____no_output_____" ] ], [ [ "churn_df = df.drop('customerID', axis=1) # dropping customerID as it doesn't have any predictive power\ndf_dummified = pd.get_dummies(churn_df, drop_first=True) # One-hot encoding\ndf_dummified.rename(columns={'Churn_Yes': 'Churn'}, inplace=True) # renaming Churn_Yes to Churn\ndf_dummified.head()", "_____no_output_____" ] ], [ [ "Using `df_dummified` dataframe, create two dataframes for features matrix and target vector and call them `X_df` and `y_df` respectively:", "_____no_output_____" ] ], [ [ "# Your answer goes here\n# X_df\n", "_____no_output_____" ], [ "# Your answer goes here\n# y_df\n", "_____no_output_____" ], [ "# SOLUTION: Uncomment and execute the cell below to get help\n#disp(hint + '09-02-X_df-y_df')", "_____no_output_____" ] ], [ [ "Check their shape:", "_____no_output_____" ] ], [ [ "# Your answer goes here\n", "X_df: (7043, 30)\ny_df: (7043,)\n" ], [ "# SOLUTION: Uncomment and execute the cell below to get help\n#disp(hint + '09-02-shape')", "_____no_output_____" ] ], [ [ "From `X_df` and `y_df` create the train/test splits. \n* Set 30% of the data to test and the remainder to train. \n* Use `random_state=833`, so you get the same result as in the notebook. \n* Name the resulting objects: `Xtrain`, `Xtest`, `ytrain`, `ytest`.", "_____no_output_____" ] ], [ [ "# Your answer goes here\n", "_____no_output_____" ], [ "# SOLUTION: Uncomment and execute the cell below to get help\n#disp(hint + '09-02-split')", "_____no_output_____" ] ], [ [ "From `sklearn.naive_bayes` import `GaussianNB`:", "_____no_output_____" ] ], [ [ "# Your answer goes here\n", "_____no_output_____" ], [ "# SOLUTION: Uncomment and execute the cell below to get help\n#disp(hint + '09-02-import')", "_____no_output_____" ] ], [ [ "Instantiate a `GaussianNB` model and call it `model`:", "_____no_output_____" ] ], [ [ "# Your answer goes here\n", "_____no_output_____" ], [ "# SOLUTION: Uncomment and execute the cell below to get help\n#disp(hint + '09-02-model')", "_____no_output_____" ] ], [ [ "Fit model to data:", "_____no_output_____" ] ], [ [ "# Your answer goes here\n", "_____no_output_____" ], [ "# SOLUTION: Uncomment and execute the cell below to get help\n#disp(hint + '09-02-fit')", "_____no_output_____" ] ], [ [ "Now let's make some predictions. Use the `Xtest` feature dataframe to predict whether these customers are churning or not. Call the outcome (predictions) `y_model`:", "_____no_output_____" ] ], [ [ "# Your answer goes here\n", "_____no_output_____" ], [ "# SOLUTION: Uncomment and execute the cell below to get help\n#disp(hint + '09-02-predict')", "_____no_output_____" ] ], [ [ "Accuracy is not the most reliable metric when it comes to evaluating classification algorithms, but it's one of the most simple ones. Let's calculate it below:", "_____no_output_____" ] ], [ [ "# Your answer goes here\n", "Our model is 0.65 accurate!\n" ], [ "# SOLUTION: Uncomment and execute the cell below to get help\n#disp(hint + '09-02-accuracy')", "_____no_output_____" ] ], [ [ "### Is this model any good? \n\nIf we had said no one will churn, what the accuracy would have been?\n\nWe would have identified all the ones who didn't churn correctly, but missed all the ones who did actually churn. Let's calculate the accuracy for this simplistic model below:\n\nhint: all you need to work with is `ytest`", "_____no_output_____" ] ], [ [ "# Your answer goes here\n", "_____no_output_____" ], [ "# SOLUTION: Uncomment and execute the cell below to get help\n#disp(hint + '09-02-accuracy-base')", "_____no_output_____" ] ], [ [ "So, after all our ML model is not that great. But the scope of this exercise is not to fine-tune this model, but understand the pipeline and how to read the outcome. Let's continue with some simple questions.", "_____no_output_____" ], [ "How many people did we have in the test dataset? \n\nSave it to a variable called `n_test`:", "_____no_output_____" ] ], [ [ "# Your answer goes here\n", "_____no_output_____" ], [ "# SOLUTION: Uncomment and execute the cell below to get help\n#disp(hint + '09-02-test-size')", "_____no_output_____" ] ], [ [ "How many of these customers churned? \n\nSave it to a variable and call it `P`:", "_____no_output_____" ] ], [ [ "# Your answer goes here\n", "_____no_output_____" ], [ "# SOLUTION: Uncomment and execute the cell below to get help\n#disp(hint + '09-02-churned')", "_____no_output_____" ] ], [ [ "How many true positives did the model generate? (True positive = correctly identified in the \"positive\" class. Take \"Yes\" as the positive class).\n\nSave it to a variable and call it `TP`:", "_____no_output_____" ] ], [ [ "# Your answer goes here\n", "_____no_output_____" ], [ "# SOLUTION: Uncomment and execute the cell below to get help\n#disp(hint + '09-02-TP')", "_____no_output_____" ] ], [ [ "What is the true positive rate (or sensitivity)?\n\n**Sensitivity** (also called the **true positive rate (TPR)**, the recall, or probability of detection in some fields) measures the proportion of actual positives that are correctly identified as such (e.g., the percentage of sick people who are correctly identified as having the condition). ~Wikipedia\n\n$TPR=\\frac{TP}{P}$, where $TP$ is ture positives and $P$ is count of all positives.", "_____no_output_____" ] ], [ [ "# Your answer goes here\n", "_____no_output_____" ], [ "# SOLUTION: Uncomment and execute the cell below to get help\n#disp(hint + '09-02-sensitivity')", "_____no_output_____" ] ], [ [ "How many of the customers in the test set didn't churn?\n\nSave it to `N`:", "_____no_output_____" ] ], [ [ "# Your answer goes here\n", "_____no_output_____" ], [ "# SOLUTION: Uncomment and execute the cell below to get help\n#disp(hint + '09-02-N')", "_____no_output_____" ] ], [ [ "How many true negatives did the model generate? (True negative = correctly rejected). \nSave a to a variable and call it `TN`:", "_____no_output_____" ] ], [ [ "# Your answer goes here\n", "_____no_output_____" ], [ "# SOLUTION: Uncomment and execute the cell below to get help\n#disp(hint + '09-02-TN')", "_____no_output_____" ] ], [ [ "What is the true negative rate (or specificity)?\n\n**Specificity** (also called the **true negative rate**) measures the proportion of actual negatives that are correctly identified as such (e.g., the percentage of healthy people who are correctly identified as not having the condition). ~Wikipedia\n\n$TNR=\\frac{TN}{N}$, where $TN$ is ture negatives and $N$ is count of all negatives.", "_____no_output_____" ] ], [ [ "# Your answer goes here\n", "_____no_output_____" ], [ "# SOLUTION: Uncomment and execute the cell below to get help\n#disp(hint + '09-02-specificity')", "_____no_output_____" ] ], [ [ "You may check out [this Wikipedia page](https://en.wikipedia.org/wiki/Sensitivity_and_specificity) for more info on sensitivity and specificity.", "_____no_output_____" ], [ "<img src=\"https://github.com/soltaniehha/Business-Analytics/blob/master/figs/09-02-sensitivity.png?raw=true\" width=\"400\" align=\"center\"/>", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ] ]

[ "ADSL" ]

[ "ADSL" ]

[ "ADSL" ]

[ [ [ "# Import Dependencies\nimport numpy as np\nimport pandas as pd\nfrom pathlib import Path\n\n# Processing Libraries\nfrom sklearn.preprocessing import StandardScaler, LabelEncoder\n\n# Models\nfrom sklearn.linear_model import LogisticRegression\nfrom sklearn.ensemble import RandomForestClassifier", "_____no_output_____" ] ], [ [ "### Predictions\nI believe the Logistic Regression model will perform better than the Random Forest Classifier as the data is not randomized.\n\nHow about after scaling the data?\nI expect the Logistic Regression to perform better on the scaled data. In this particular case, the Logistic Regression is the appropriate regression analysis to conduct as the dependent variable are binary.\n\nConclusion\nAs expected, the Logistic Regression performed better than the Random Forest Classifier. ", "_____no_output_____" ] ], [ [ "# Retrieving the datasets\ntrain_df = pd.read_csv(\"Resources/2019loans.csv\")\ntest_df = pd.read_csv(\"Resources/2020Q1loans.csv\")", "_____no_output_____" ], [ "# Processing the Data\nX_train_dummies = pd.get_dummies(train_df)\nprint(X_train_dummies.columns)\nX_train_dummies", "Index(['Unnamed: 0', 'index', 'loan_amnt', 'int_rate', 'installment',\n 'annual_inc', 'dti', 'delinq_2yrs', 'inq_last_6mths', 'open_acc',\n 'pub_rec', 'revol_bal', 'total_acc', 'out_prncp', 'out_prncp_inv',\n 'total_pymnt', 'total_pymnt_inv', 'total_rec_prncp', 'total_rec_int',\n 'total_rec_late_fee', 'recoveries', 'collection_recovery_fee',\n 'last_pymnt_amnt', 'collections_12_mths_ex_med', 'policy_code',\n 'acc_now_delinq', 'tot_coll_amt', 'tot_cur_bal', 'open_acc_6m',\n 'open_act_il', 'open_il_12m', 'open_il_24m', 'mths_since_rcnt_il',\n 'total_bal_il', 'il_util', 'open_rv_12m', 'open_rv_24m', 'max_bal_bc',\n 'all_util', 'total_rev_hi_lim', 'inq_fi', 'total_cu_tl', 'inq_last_12m',\n 'acc_open_past_24mths', 'avg_cur_bal', 'bc_open_to_buy', 'bc_util',\n 'chargeoff_within_12_mths', 'delinq_amnt', 'mo_sin_old_il_acct',\n 'mo_sin_old_rev_tl_op', 'mo_sin_rcnt_rev_tl_op', 'mo_sin_rcnt_tl',\n 'mort_acc', 'mths_since_recent_bc', 'mths_since_recent_inq',\n 'num_accts_ever_120_pd', 'num_actv_bc_tl', 'num_actv_rev_tl',\n 'num_bc_sats', 'num_bc_tl', 'num_il_tl', 'num_op_rev_tl',\n 'num_rev_accts', 'num_rev_tl_bal_gt_0', 'num_sats', 'num_tl_120dpd_2m',\n 'num_tl_30dpd', 'num_tl_90g_dpd_24m', 'num_tl_op_past_12m',\n 'pct_tl_nvr_dlq', 'percent_bc_gt_75', 'pub_rec_bankruptcies',\n 'tax_liens', 'tot_hi_cred_lim', 'total_bal_ex_mort', 'total_bc_limit',\n 'total_il_high_credit_limit', 'home_ownership_ANY',\n 'home_ownership_MORTGAGE', 'home_ownership_OWN', 'home_ownership_RENT',\n 'verification_status_Not Verified',\n 'verification_status_Source Verified', 'verification_status_Verified',\n 'loan_status_high_risk', 'loan_status_low_risk', 'pymnt_plan_n',\n 'initial_list_status_f', 'initial_list_status_w',\n 'application_type_Individual', 'application_type_Joint App',\n 'hardship_flag_N', 'hardship_flag_Y', 'debt_settlement_flag_N',\n 'debt_settlement_flag_Y'],\n dtype='object')\n" ], [ "train_df", "_____no_output_____" ], [ "test_df", "_____no_output_____" ], [ "# Convert categorical data to numeric and separate target feature for training data\nX_train = train_df.drop([\"loan_status\"], axis=1)\nX_train = pd.get_dummies(X_train)\ny_train = LabelEncoder().fit_transform(train_df[\"loan_status\"])\nprint(X_train.columns)\ny_train", "Index(['Unnamed: 0', 'index', 'loan_amnt', 'int_rate', 'installment',\n 'annual_inc', 'dti', 'delinq_2yrs', 'inq_last_6mths', 'open_acc',\n 'pub_rec', 'revol_bal', 'total_acc', 'out_prncp', 'out_prncp_inv',\n 'total_pymnt', 'total_pymnt_inv', 'total_rec_prncp', 'total_rec_int',\n 'total_rec_late_fee', 'recoveries', 'collection_recovery_fee',\n 'last_pymnt_amnt', 'collections_12_mths_ex_med', 'policy_code',\n 'acc_now_delinq', 'tot_coll_amt', 'tot_cur_bal', 'open_acc_6m',\n 'open_act_il', 'open_il_12m', 'open_il_24m', 'mths_since_rcnt_il',\n 'total_bal_il', 'il_util', 'open_rv_12m', 'open_rv_24m', 'max_bal_bc',\n 'all_util', 'total_rev_hi_lim', 'inq_fi', 'total_cu_tl', 'inq_last_12m',\n 'acc_open_past_24mths', 'avg_cur_bal', 'bc_open_to_buy', 'bc_util',\n 'chargeoff_within_12_mths', 'delinq_amnt', 'mo_sin_old_il_acct',\n 'mo_sin_old_rev_tl_op', 'mo_sin_rcnt_rev_tl_op', 'mo_sin_rcnt_tl',\n 'mort_acc', 'mths_since_recent_bc', 'mths_since_recent_inq',\n 'num_accts_ever_120_pd', 'num_actv_bc_tl', 'num_actv_rev_tl',\n 'num_bc_sats', 'num_bc_tl', 'num_il_tl', 'num_op_rev_tl',\n 'num_rev_accts', 'num_rev_tl_bal_gt_0', 'num_sats', 'num_tl_120dpd_2m',\n 'num_tl_30dpd', 'num_tl_90g_dpd_24m', 'num_tl_op_past_12m',\n 'pct_tl_nvr_dlq', 'percent_bc_gt_75', 'pub_rec_bankruptcies',\n 'tax_liens', 'tot_hi_cred_lim', 'total_bal_ex_mort', 'total_bc_limit',\n 'total_il_high_credit_limit', 'home_ownership_ANY',\n 'home_ownership_MORTGAGE', 'home_ownership_OWN', 'home_ownership_RENT',\n 'verification_status_Not Verified',\n 'verification_status_Source Verified', 'verification_status_Verified',\n 'pymnt_plan_n', 'initial_list_status_f', 'initial_list_status_w',\n 'application_type_Individual', 'application_type_Joint App',\n 'hardship_flag_N', 'hardship_flag_Y', 'debt_settlement_flag_N',\n 'debt_settlement_flag_Y'],\n dtype='object')\n" ], [ "# Convert categorical data to numeric and separate target feature for testing data\nX_test = test_df.drop([\"loan_status\"], axis=1)\nX_test = pd.get_dummies(X_test)\ny_test = LabelEncoder().fit_transform(test_df[\"loan_status\"])\nprint(X_test.columns)\ny_test", "Index(['Unnamed: 0', 'index', 'loan_amnt', 'int_rate', 'installment',\n 'annual_inc', 'dti', 'delinq_2yrs', 'inq_last_6mths', 'open_acc',\n 'pub_rec', 'revol_bal', 'total_acc', 'out_prncp', 'out_prncp_inv',\n 'total_pymnt', 'total_pymnt_inv', 'total_rec_prncp', 'total_rec_int',\n 'total_rec_late_fee', 'recoveries', 'collection_recovery_fee',\n 'last_pymnt_amnt', 'collections_12_mths_ex_med', 'policy_code',\n 'acc_now_delinq', 'tot_coll_amt', 'tot_cur_bal', 'open_acc_6m',\n 'open_act_il', 'open_il_12m', 'open_il_24m', 'mths_since_rcnt_il',\n 'total_bal_il', 'il_util', 'open_rv_12m', 'open_rv_24m', 'max_bal_bc',\n 'all_util', 'total_rev_hi_lim', 'inq_fi', 'total_cu_tl', 'inq_last_12m',\n 'acc_open_past_24mths', 'avg_cur_bal', 'bc_open_to_buy', 'bc_util',\n 'chargeoff_within_12_mths', 'delinq_amnt', 'mo_sin_old_il_acct',\n 'mo_sin_old_rev_tl_op', 'mo_sin_rcnt_rev_tl_op', 'mo_sin_rcnt_tl',\n 'mort_acc', 'mths_since_recent_bc', 'mths_since_recent_inq',\n 'num_accts_ever_120_pd', 'num_actv_bc_tl', 'num_actv_rev_tl',\n 'num_bc_sats', 'num_bc_tl', 'num_il_tl', 'num_op_rev_tl',\n 'num_rev_accts', 'num_rev_tl_bal_gt_0', 'num_sats', 'num_tl_120dpd_2m',\n 'num_tl_30dpd', 'num_tl_90g_dpd_24m', 'num_tl_op_past_12m',\n 'pct_tl_nvr_dlq', 'percent_bc_gt_75', 'pub_rec_bankruptcies',\n 'tax_liens', 'tot_hi_cred_lim', 'total_bal_ex_mort', 'total_bc_limit',\n 'total_il_high_credit_limit', 'home_ownership_ANY',\n 'home_ownership_MORTGAGE', 'home_ownership_OWN', 'home_ownership_RENT',\n 'verification_status_Not Verified',\n 'verification_status_Source Verified', 'verification_status_Verified',\n 'pymnt_plan_n', 'initial_list_status_f', 'initial_list_status_w',\n 'application_type_Individual', 'application_type_Joint App',\n 'hardship_flag_N', 'hardship_flag_Y', 'debt_settlement_flag_N'],\n dtype='object')\n" ], [ "# add missing dummy variables to testing set\nX_test[\"debt_settlement_flag_Y\"] = X_test.apply(lambda row: round(abs(row[\"debt_settlement_flag_N\"] - 1),0), axis=1)\nX_test = X_test.convert_dtypes()\nX_test", "_____no_output_____" ], [ "# Train the Logistic Regression model on the unscaled data and print the model score\nclassifier = LogisticRegression(max_iter=10000)\nclassifier.fit(X_train, y_train)\ntrain_score = classifier.score(X_train, y_train)\ntest_score = classifier.score(X_test, y_test)\nprint(f\"Train Score: {train_score:.3f}\")\nprint(f\"Test Score: {test_score:.3f}\")", "Train Score: 0.703\nTest Score: 0.565\n" ], [ "# Train a Random Forest Classifier model and print the model score\nrfc = RandomForestClassifier(random_state=42)\nrfc.fit(X_train, y_train)\ntrain_score = rfc.score(X_train, y_train)\ntest_score = rfc.score(X_test, y_test)\nprint(f\"Train Score: {train_score:.3f}\")\nprint(f\"Test Score: {test_score:.3f}\")", "Train Score: 1.000\nTest Score: 0.631\n" ], [ "# Scale the data\nscaler = StandardScaler().fit(X_train)\nX_train_scaled = scaler.transform(X_train)\nX_test_scaled = scaler.transform(X_test)", "_____no_output_____" ], [ "# Train the Logistic Regression model on the scaled data and print the model score\nclassifier = LogisticRegression(max_iter=10000)\nclassifier.fit(X_train_scaled, y_train)\ntrain_score = classifier.score(X_train_scaled, y_train)\ntest_score = classifier.score(X_test_scaled, y_test)\nprint(f\"Train Score: {train_score:.3f}\")\nprint(f\"Test Score: {test_score:.3f}\")", "Train Score: 0.713\nTest Score: 0.720\n" ], [ "# Train a Random Forest Classifier model on the scaled data and print the model score\nrfc = RandomForestClassifier(random_state=42)\nrfc.fit(X_train_scaled, y_train)\ntrain_score = rfc.score(X_train_scaled, y_train)\ntest_score = rfc.score(X_test_scaled, y_test)\nprint(f\"Train Score: {train_score:.3f}\")\nprint(f\"Test Score: {test_score:.3f}\")", "Train Score: 1.000\nTest Score: 0.630\n" ] ] ]

[ "code", "markdown", "code" ]

[ [ "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "given = \"\"\"\nGrey\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tRIGHT\tRIGHT\t2\nGrey\tLEFT\tLEFT\t2\nBLACK\tRIGHT\tRIGHT\t2\nGrey\tLEFT\tLEFT\t2\nBLACK\tRIGHT\tRIGHT\t2\nGrey\tEMPTY\tEMPTY\t4\nGrey\tLEFT\tLEFT\t3\nBLACK\tTOP\tTOP\t1\nBLACK\tEMPTY\tEMPTY\t5\nGrey\tTOP\tTOP\t3\nGrey\tRIGHT\tRIGHT\t5\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tEMPTY\tEMPTY\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nBLACK\tTOP\tTOP\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tTOP\tTOP\t5\nBLACK\tRIGHT\tRIGHT\t3\nGrey\tLEFT\tLEFT\t3\nBLACK\tLEFT\tLEFT\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nBLACK\tTOP\tTOP\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tLEFT\tLEFT\t3\nGrey\tBOTTOM\tBOTTOM\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tEMPTY\tEMPTY\t4\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nBLACK\tTOP\tTOP\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tLEFT\tLEFT\t3\nBLACK\tBOTTOM\tBOTTOM\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nBLACK\tTOP\tTOP\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tLEFT\tLEFT\t3\nGrey\tLEFT\tLEFT\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tEMPTY\tEMPTY\t4\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nBLACK\tTOP\tTOP\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tLEFT\tLEFT\t3\nBLACK\tLEFT\tLEFT\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nBLACK\tTOP\tTOP\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tRIGHT\tRIGHT\t5\nGrey\tLEFT\tLEFT\t5\nGrey\tLEFT\tLEFT\t2\nBLACK\tRIGHT\tRIGHT\t5\nGrey\tRIGHT\tRIGHT\t5\nBLACK\tRIGHT\tRIGHT\t5\nGrey\tRIGHT\tRIGHT\t5\nGrey\tLEFT\tLEFT\t4\nBLACK\tTOP\tTOP\t2\nBLACK\tRIGHT\tRIGHT\t5\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tLEFT\tLEFT\t3\nGrey\tTOP\tTOP\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tLEFT\tLEFT\t3\nGrey\tEMPTY\tEMPTY\t1\nGrey\tEMPTY\tEMPTY\t2\nMAP CIRCLE\t\t\t\nGrey\tLEFT\tLEFT\t3\nGrey\tEMPTY\tEMPTY\t5\nBLACK\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tLEFT\tLEFT\t3\nBLACK\tEMPTY\tEMPTY\t1\nBLACK\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nBLACK\tTOP\tTOP\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tLEFT\tLEFT\t3\nGrey\tRIGHT\tRIGHT\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tEMPTY\tEMPTY\t4\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tLEFT\tLEFT\t3\nBLACK\tRIGHT\tRIGHT\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tLEFT\tLEFT\t3\nGrey\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tRIGHT\tRIGHT\t2\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tEMPTY\tEMPTY\t4\nGrey\tLEFT\tLEFT\t3\nBLACK\tTOP\tTOP\t1\nBLACK\tEMPTY\tEMPTY\t5\nGrey\tTOP\tTOP\t3\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tTOP\tTOP\t5\nGrey\tLEFT\tLEFT\t5\nBLACK\tTOP\tTOP\t4\nGrey\tLEFT\tLEFT\t2\nBLACK\tLEFT\tLEFT\t2\nBLACK\tBOTTOM\tBOTTOM\t5\nBLACK\tTOP\tTOP\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tTOP\tTOP\t4\nGrey\tTOP\tTOP\t2\nBLACK\tTOP\tTOP\t5\nBLACK\tTOP\tTOP\t4\nGrey\tLEFT\tLEFT\t2\nGrey\tBOTTOM\tBOTTOM\t3\nGrey\tRIGHT\tRIGHT\t2\nGrey\tTOP\tTOP\t5\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tRIGHT\tRIGHT\t4\nGrey\tLEFT\tLEFT\t2\nGrey\tLEFT\tLEFT\t3\nBLACK\tLEFT\tLEFT\t2\nBLACK\tBOTTOM\tBOTTOM\t5\nBLACK\tRIGHT\tRIGHT\t4\nGrey\tLEFT\tLEFT\t2\nGrey\tLEFT\tLEFT\t3\nBLACK\tLEFT\tLEFT\t2\nBLACK\tBOTTOM\tBOTTOM\t5\nBLACK\tLEFT\tLEFT\t2\nBLACK\tBOTTOM\tBOTTOM\t5\nGrey\tTOP\tTOP\t5\nGrey\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t5\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tLEFT\tLEFT\t2\nGrey\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tRIGHT\tRIGHT\t2\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tEMPTY\tEMPTY\t4\nGrey\tLEFT\tLEFT\t3\nMAP CIRCLE\t\t\t\nBLACK\tEMPTY\tEMPTY\t5\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tEMPTY\tEMPTY\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tRIGHT\tRIGHT\t3\nGrey\tRIGHT\tRIGHT\t3\nBLACK\tLEFT\tLEFT\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tBOTTOM\tBOTTOM\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tEMPTY\tEMPTY\t4\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tBOTTOM\tBOTTOM\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tTOP\tTOP\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tLEFT\tLEFT\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tEMPTY\tEMPTY\t4\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tLEFT\tLEFT\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tTOP\tTOP\t4\nGrey\tTOP\tTOP\t2\nBLACK\tTOP\tTOP\t5\nGrey\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tLEFT\tLEFT\t5\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tEMPTY\tEMPTY\t5\nBLACK\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tEMPTY\tEMPTY\t1\nGrey\tEMPTY\tEMPTY\t2\nMAP CIRCLE\t\t\t\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tTOP\tTOP\t4\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tEMPTY\tEMPTY\t1\nBLACK\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tTOP\tTOP\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tRIGHT\tRIGHT\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tEMPTY\tEMPTY\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tTOP\tTOP\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tTOP\tTOP\t4\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tRIGHT\tRIGHT\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tLEFT\tLEFT\t4\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tRIGHT\tRIGHT\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tTOP\tTOP\t3\nGrey\tLEFT\tLEFT\t2\nBLACK\tLEFT\tLEFT\t4\nGrey\tLEFT\tLEFT\t2\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tRIGHT\tRIGHT\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tEMPTY\tEMPTY\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tLEFT\tLEFT\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tRIGHT\tRIGHT\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tEMPTY\tEMPTY\t4\nGrey\tLEFT\tLEFT\t2\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tLEFT\tLEFT\t2\nBLACK\tLEFT\tLEFT\t4\nGrey\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tRIGHT\tRIGHT\t2\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tEMPTY\tEMPTY\t4\nGrey\tLEFT\tLEFT\t3\nBLACK\tTOP\tTOP\t1\nBLACK\tEMPTY\tEMPTY\t5\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tLEFT\tLEFT\t2\nBLACK\tEMPTY\tEMPTY\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tRIGHT\tRIGHT\t3\nGrey\tRIGHT\tRIGHT\t3\nBLACK\tLEFT\tLEFT\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tTOP\tTOP\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tLEFT\tLEFT\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tEMPTY\tEMPTY\t4\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tTOP\tTOP\t4\nGrey\tEMPTY\tEMPTY\t3\nGrey\tLEFT\tLEFT\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tBOTTOM\tBOTTOM\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tLEFT\tLEFT\t5\nGrey\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t2\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tLEFT\tLEFT\t5\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tTOP\tTOP\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tEMPTY\tEMPTY\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tEMPTY\tEMPTY\t5\nBLACK\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tEMPTY\tEMPTY\t1\nBLACK\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tLEFT\tLEFT\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tEMPTY\tEMPTY\t1\nBLACK\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tLEFT\tLEFT\t2\nBLACK\tLEFT\tLEFT\t4\nGrey\tLEFT\tLEFT\t2\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tEMPTY\tEMPTY\t5\nBLACK\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tLEFT\tLEFT\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tEMPTY\tEMPTY\t5\nBLACK\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tLEFT\tLEFT\t2\nBLACK\tLEFT\tLEFT\t4\nGrey\tLEFT\tLEFT\t2\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tEMPTY\tEMPTY\t5\nBLACK\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nBLACK\tLEFT\tLEFT\t3\nGrey\tEMPTY\tEMPTY\t5\nBLACK\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tTOP\tTOP\t5\nGrey\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nMAP CIRCLE\t\t\t\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tEMPTY\tEMPTY\t4\nGrey\tLEFT\tLEFT\t3\nBLACK\tTOP\tTOP\t1\nBLACK\tEMPTY\tEMPTY\t5\nGrey\tRIGHT\tRIGHT\t4\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tEMPTY\tEMPTY\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tTOP\tTOP\t4\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tRIGHT\tRIGHT\t3\nGrey\tRIGHT\tRIGHT\t3\nBLACK\tLEFT\tLEFT\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tTOP\tTOP\t4\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tBOTTOM\tBOTTOM\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tTOP\tTOP\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tBOTTOM\tBOTTOM\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tEMPTY\tEMPTY\t4\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tTOP\tTOP\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tLEFT\tLEFT\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tEMPTY\tEMPTY\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tLEFT\tLEFT\t4\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tBOTTOM\tBOTTOM\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tLEFT\tLEFT\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tTOP\tTOP\t4\nGrey\tTOP\tTOP\t2\nBLACK\tTOP\tTOP\t5\nGrey\tRIGHT\tRIGHT\t2\nGrey\tTOP\tTOP\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tLEFT\tLEFT\t5\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tEMPTY\tEMPTY\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tEMPTY\tEMPTY\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tTOP\tTOP\t3\nGrey\tLEFT\tLEFT\t2\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tLEFT\tLEFT\t2\nBLACK\tLEFT\tLEFT\t4\nGrey\tEMPTY\tEMPTY\t1\nGrey\tEMPTY\tEMPTY\t2\nGrey\tEMPTY\tEMPTY\t1\nGrey\tEMPTY\tEMPTY\t2\nGrey\tEMPTY\tEMPTY\t1\nGrey\tEMPTY\tEMPTY\t2\nGrey\tEMPTY\tEMPTY\t1\nGrey\tEMPTY\tEMPTY\t2\nGrey\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tRIGHT\tRIGHT\t2\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tEMPTY\tEMPTY\t4\nGrey\tLEFT\tLEFT\t3\nBLACK\tTOP\tTOP\t1\nBLACK\tEMPTY\tEMPTY\t5\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tEMPTY\tEMPTY\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tRIGHT\tRIGHT\t3\nGrey\tRIGHT\tRIGHT\t3\nBLACK\tLEFT\tLEFT\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tBOTTOM\tBOTTOM\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tEMPTY\tEMPTY\t4\nGrey\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tBOTTOM\tBOTTOM\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tLEFT\tLEFT\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tEMPTY\tEMPTY\t4\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tLEFT\tLEFT\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tTOP\tTOP\t4\nGrey\tTOP\tTOP\t2\nBLACK\tTOP\tTOP\t5\nGrey\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tLEFT\tLEFT\t5\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tLEFT\tLEFT\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tLEFT\tLEFT\t5\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nBLACK\tRIGHT\tRIGHT\t4\nGrey\tLEFT\tLEFT\t2\nGrey\tLEFT\tLEFT\t5\nBLACK\tTOP\tTOP\t4\nGrey\tLEFT\tLEFT\t2\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nBLACK\tRIGHT\tRIGHT\t4\nGrey\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tLEFT\tLEFT\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nMAP CIRCLE\t\t\t\nBLACK\tTOP\tTOP\t2\nBLACK\tLEFT\tLEFT\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nGrey\tTOP\tTOP\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tLEFT\tLEFT\t2\nBLACK\tLEFT\tLEFT\t4\nGrey\tLEFT\tLEFT\t2\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tRIGHT\tRIGHT\t2\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tEMPTY\tEMPTY\t4\nGrey\tLEFT\tLEFT\t3\nBLACK\tTOP\tTOP\t1\nBLACK\tEMPTY\tEMPTY\t5\nGrey\tRIGHT\tRIGHT\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tEMPTY\tEMPTY\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tRIGHT\tRIGHT\t3\nGrey\tRIGHT\tRIGHT\t3\nBLACK\tLEFT\tLEFT\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tBOTTOM\tBOTTOM\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tEMPTY\tEMPTY\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tBOTTOM\tBOTTOM\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tLEFT\tLEFT\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tEMPTY\tEMPTY\t4\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tLEFT\tLEFT\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tLEFT\tLEFT\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tEMPTY\tEMPTY\t4\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nGrey\tLEFT\tLEFT\t2\nBLACK\tBOTTOM\tBOTTOM\t4\nBLACK\tLEFT\tLEFT\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tRIGHT\tRIGHT\t4\nGrey\tLEFT\tLEFT\t2\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tRIGHT\tRIGHT\t2\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tEMPTY\tEMPTY\t4\nGrey\tLEFT\tLEFT\t3\nBLACK\tTOP\tTOP\t1\nBLACK\tEMPTY\tEMPTY\t5\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tEMPTY\tEMPTY\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tRIGHT\tRIGHT\t3\nGrey\tRIGHT\tRIGHT\t3\nBLACK\tLEFT\tLEFT\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tBOTTOM\tBOTTOM\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tEMPTY\tEMPTY\t4\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tBOTTOM\tBOTTOM\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tBOTTOM\tBOTTOM\t1\nGrey\tEMPTY\tEMPTY\t2\nMAP CIRCLE\t\t\t\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tLEFT\tLEFT\t2\nBLACK\tBOTTOM\tBOTTOM\t4\nBLACK\tBOTTOM\tBOTTOM\t1\nGrey\tEMPTY\tEMPTY\t2\nBLACK\tBOTTOM\tBOTTOM\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tLEFT\tLEFT\t2\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tRIGHT\tRIGHT\t2\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tEMPTY\tEMPTY\t5\nGrey\tLEFT\tLEFT\t3\nBLACK\tTOP\tTOP\t1\nBLACK\tEMPTY\tEMPTY\t5\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tEMPTY\tEMPTY\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nBLACK\tRIGHT\tRIGHT\t3\nGrey\tRIGHT\tRIGHT\t3\nBLACK\tLEFT\tLEFT\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nBLACK\tBOTTOM\tBOTTOM\t4\nBLACK\tBOTTOM\tBOTTOM\t2\nBLACK\tRIGHT\tRIGHT\t3\nGrey\tRIGHT\tRIGHT\t3\nBLACK\tLEFT\tLEFT\t3\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tLEFT\tLEFT\t2\nBLACK\tBOTTOM\tBOTTOM\t4\nGrey\tEMPTY\tEMPTY\t4\nGrey\tLEFT\tLEFT\t3\nBLACK\tTOP\tTOP\t1\nBLACK\tEMPTY\tEMPTY\t5\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tRIGHT\tRIGHT\t2\nBLACK\tLEFT\tLEFT\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tEMPTY\tEMPTY\t4\nGrey\tLEFT\tLEFT\t3\nBLACK\tTOP\tTOP\t1\nBLACK\tEMPTY\tEMPTY\t5\nGrey\tTOP\tTOP\t3\nGrey\tLEFT\tLEFT\t2\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tRIGHT\tRIGHT\t2\nBLACK\tBOTTOM\tBOTTOM\t2\nGrey\tTOP\tTOP\t3\nGrey\tEMPTY\tEMPTY\t3\nGrey\tBOTTOM\tBOTTOM\t5\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tRIGHT\tRIGHT\t2\nBLACK\tBOTTOM\tBOTTOM\t2\nGrey\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tRIGHT\tRIGHT\t2\nBLACK\tBOTTOM\tBOTTOM\t2\nGrey\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\nGrey\tRIGHT\tRIGHT\t2\nGrey\tRIGHT\tRIGHT\t2\nGrey\tBOTTOM\tBOTTOM\t4\nGrey\tBOTTOM\tBOTTOM\t2\n\"\"\"\n\ntranscript = \"\"\"\nI AM SAM. I AM SAM. SAM I AM.\n\nTHAT SAM-I-AM! THAT SAM-I-AM! I DO NOT LIKE THAT SAM-I-AM!\n\nDO WOULD YOU LIKE GREEN EGGS AND HAM?\n\nI DO NOT LIKE THEM,SAM-I-AM.\nI DO NOT LIKE GREEN EGGS AND HAM.\n\nWOULD YOU LIKE THEM HERE OR THERE?\n\nI WOULD NOT LIKE THEM HERE OR THERE.\nI WOULD NOT LIKE THEM ANYWHERE.\nI DO NOT LIKE GREEN EGGS AND HAM.\nI DO NOT LIKE THEM, SAM-I-AM.\n\nWOULD YOU LIKE THEM IN A HOUSE?\nWOULD YOU LIKE THEN WITH A MOUSE?\n\nI DO NOT LIKE THEM IN A HOUSE.\nI DO NOT LIKE THEM WITH A MOUSE.\nI DO NOT LIKE THEM HERE OR THERE.\nI DO NOT LIKE THEM ANYWHERE.\nI DO NOT LIKE GREEN EGGS AND HAM.\nI DO NOT LIKE THEM, SAM-I-AM.\n\nWOULD YOU EAT THEM IN A BOX?\nWOULD YOU EAT THEM WITH A FOX?\n\nNOT IN A BOX. NOT WITH A FOX.\nNOT IN A HOUSE. NOT WITH A MOUSE.\nI WOULD NOT EAT THEM HERE OR THERE.\nI WOULD NOT EAT THEM ANYWHERE.\nI WOULD NOT EAT GREEN EGGS AND HAM.\nI DO NOT LIKE THEM, SAM-I-AM.\n\nWOULD YOU? COULD YOU? IN A CAR?\nEAT THEM! EAT THEM! HERE THEY ARE.\n\nI WOULD NOT, COULD NOT, IN A CAR.\n\nYOU MAY LIKE THEM. YOU WILL SEE.\nYOU MAY LIKE THEM IN A TREE!\n\nI WOULD NOT, COULD NOT IN A TREE.\nNOT IN A CAR! YOU LET ME BE.\nI DO NOT LIKE THEM IN A BOX.\nI DO NOT LIKE THEM WITH A FOX.\nI DO NOT LIKE THEM IN A HOUSE.\nI DO NOT LIKE THEM WITH A MOUSE.\nI DO NOT LIKE THEM HERE OR THERE.\nI DO NOT LIKE THEM ANYWHERE.\nI DO NOT LIKE GREEN EGGS AND HAM.\nI DO NOT LIKE THEM, SAM-I-AM.\n\nA TRAIN! A TRAIN! A TRAIN! A TRAIN!\nCOULD YOU, WOULD YOU ON A TRAIN?\n\nNOT ON TRAIN! NOT IN A TREE!\nNOT IN A CAR! SAM! LET ME BE!\nI WOULD NOT, COULD NOT, IN A BOX.\nI WOULD NOT, COULD NOT, WITH A FOX.\nI WILL NOT EAT THEM IN A HOUSE.\nI WILL NOT EAT THEM HERE OR THERE.\nI WILL NOT EAT THEM ANYWHERE.\nI DO NOT EAT GREEM EGGS AND HAM.\nI DO NOT LIKE THEM, SAM-I-AM.\n\nSAY! IN THE DARK? HERE IN THE DARK!\nWOULD YOU, COULD YOU, IN THE DARK?\n\nI WOULD NOT, COULD NOT, IN THE DARK.\n\nWOULD YOU COULD YOU IN THE RAIN?\n\nI WOULD NOT, COULD NOT IN THE RAIN.\nNOT IN THE DARK. NOT ON A TRAIN.\nNOT IN A CAR. NOT IN A TREE.\nI DO NOT LIKE THEM, SAM, YOU SEE.\nNOT IN A HOUSE. NOT IN A BOX.\nNOT WITH A MOUSE. NOT WITH A FOX.\nI WILL NOT EAT THEM HERE OR THERE.\nI DO NOT LIKE THEM ANYWHERE!\n\nYOU DO NOT LIKE GREEN EGGS AND HAM?\n\nI DO NOT LIKE THEM, SAM-I-AM.\n\nCOULD YOU, WOULD YOU, WITH A GOAT?\n\nI WOULD NOT, COULD NOT WITH A GOAT!\n\nWOULD YOU, COULD YOU, ON A BOAT?\n\nI COULD NOT, WOULD NOT, ON A BOAT.\nI WILL NOT, WILL NOT, WITH A GOAT.\nI WILL NOT EAT THEM IN THE RAIN.\nNOT IN THE DARK! NOT IN A TREE!\nNOT IN A CAR! YOU LET ME BE!\nI DO NOT LIKE THEM IN A BOX.\nI DO NOT LIKE THEM WITH A FOX.\nI WILL NOT EAT THEM IN A HOUSE.\nI DO NOT LIKE THEM WITH A MOUSE.\nI DO NOT LIKE THEM HERE OR THERE.\nI DO NOT LIKE THEM ANYWHERE!\nI DO NOT LIKE GREEN EGGS AND HAM!\nI DO NOT LIKE THEM, SAM-I-AM.\n\nYOU DO NOT LIKE THEM. SO YOU SAY.\nTRY THEM! TRY THEM! AND YOU MAY.\nTRY THEM AND YOU MAY, I SAY.\n\nsAM! IF YOU LET ME BE,\nI WILL TRY THEM. YOU WILL SEE.\n\n(... and he tries them ...)\n\nSAY! I LIKE GREEN EGGS AND HAM!\nI DO! I LIKE THEM, SAM-I-AM!\nAND I WOULD EAT THEM IN A BOAT.\nAND I WOULD EAT THEM WITH A GOAT...\nAND I WILL EAT THEM, IN THE RAIN.\nAND IN THE DARK. AND ON A TRAIN.\nAND IN A CAR. AND IN A TREE.\nTHEY ARE SO GOOD, SO GOOD, YOU SEE!\nSO I WILL EAT THEM IN A BOX.\nAND I WILL EAT THEM WITH A FOX.\nAND I WILL EAT THEM IN A HOUSE.\nAND I WILL EAT THEM WITH A MOUSE.\nAND I WILL EAT THEM HERE AND THERE.\nSAY! I WILL EAT THEM ANYWHERE!\nI DO SO LIKE GREEN EGGS AND HAM!\nTHANK YOU! THANK YOU, SAM I AM.\n\"\"\"\n\nVALUES = {\n 'GREY': 0,\n 'BLACK': 0,\n \n 'EMPTY': 0,\n 'TOP': 1,\n 'RIGHT': 2,\n 'BOTTOM': 3,\n 'LEFT': 4,\n \n '1': 0,\n '2': 1,\n '3': 2,\n '4': 3,\n '5': 4,\n}\n", "_____no_output_____" ], [ "def parse(txt):\n chunk = []\n result = [chunk]\n for line in txt.strip('\\n').split('\\n'):\n if 'MAP CIRCLE' in line:\n print('chunk length', len(chunk))\n chunk = []\n result.append(chunk)\n continue\n color, _, position, fill = line.upper().split('\\t')\n chunk.append((color, position, fill))\n #chunk.append((VALUES[color], VALUES[position], VALUES[fill]))\n return result", "_____no_output_____" ], [ "import collections\n\nparsed = parse(given)\n", "chunk length 79\nchunk length 90\nchunk length 69\nchunk length 161\nchunk length 190\nchunk length 121\n" ], [ "ngram_qs = [\n collections.deque([], maxlen=i) for i in range(1, 10)\n]\n\ncounter = collections.Counter()\n\nfor chunk in parsed:\n for ngram_q in ngram_qs:\n ngram_q.clear()\n for piece in chunk:\n for ngram_q in ngram_qs:\n ngram_q.append(''.join(map(str, piece)))\n if len(ngram_q) == ngram_q.maxlen:\n # print(ngram_q.maxlen, '\\n', ngram_q)\n counter[','.join(ngram_q)] += 1\n\nfor s, freq in sorted(counter.items(), key=lambda x: x[1], reverse=True):\n print(str(freq).rjust(8), s)\n if freq == 1:\n break", " 84 GREYEMPTY3\n 83 GREYBOTTOM2\n 60 BLACKLEFT2\n 59 GREYEMPTY2\n 44 GREYTOP3\n 41 BLACKBOTTOM3\n 37 GREYBOTTOM5\n 35 GREYLEFT2\n 35 GREYTOP3,GREYEMPTY3\n 34 GREYBOTTOM5,GREYBOTTOM2\n 34 GREYEMPTY3,GREYBOTTOM5\n 33 BLACKLEFT2,GREYTOP3\n 33 GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 32 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 31 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 30 GREYEMPTY2,BLACKBOTTOM3\n 27 BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 26 BLACKBOTTOM4\n 25 GREYLEFT3\n 25 GREYRIGHT4\n 25 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 24 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 23 BLACKLEFT2,GREYRIGHT4\n 22 BLACKTOP4\n 19 BLACKBOTTOM3,GREYEMPTY3\n 17 GREYRIGHT2\n 17 BLACKTOP4,GREYBOTTOM2\n 17 BLACKEMPTY4\n 17 GREYEMPTY2,BLACKEMPTY4\n 15 GREYBOTTOM4\n 15 GREYBOTTOM4,GREYBOTTOM2\n 15 GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3\n 14 BLACKBOTTOM3,BLACKLEFT2\n 14 BLACKLEFT4\n 13 GREYBOTTOM2,GREYRIGHT2\n 13 GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2\n 13 GREYEMPTY3,BLACKBOTTOM4\n 12 GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2\n 12 BLACKBOTTOM4,GREYBOTTOM2\n 12 GREYRIGHT4,GREYEMPTY3\n 11 BLACKEMPTY5\n 11 BLACKLEFT3\n 11 BLACKEMPTY2\n 11 BLACKEMPTY2,BLACKBOTTOM3\n 11 GREYLEFT2,BLACKBOTTOM4\n 10 GREYEMPTY4\n 10 GREYEMPTY4,GREYLEFT3\n 10 BLACKTOP1\n 10 GREYLEFT3,BLACKTOP1\n 10 BLACKTOP1,BLACKEMPTY5\n 10 GREYLEFT3,BLACKTOP1,BLACKEMPTY5\n 10 BLACKLEFT3,BLACKLEFT2\n 10 BLACKLEFT2,GREYRIGHT4,GREYEMPTY3\n 10 GREYEMPTY3,BLACKTOP4\n 10 GREYEMPTY3,BLACKBOTTOM4,GREYBOTTOM2\n 9 GREYEMPTY4,GREYLEFT3,BLACKTOP1\n 9 GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5\n 9 BLACKEMPTY5,GREYTOP3\n 9 BLACKRIGHT3\n 9 BLACKEMPTY4,BLACKLEFT2\n 9 GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2\n 9 BLACKBOTTOM1\n 9 BLACKBOTTOM1,GREYEMPTY2\n 9 BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3\n 9 GREYEMPTY1\n 9 GREYEMPTY1,GREYEMPTY2\n 9 GREYRIGHT2,BLACKLEFT2\n 9 GREYRIGHT2,BLACKLEFT2,GREYTOP3\n 9 GREYEMPTY3,BLACKTOP4,GREYBOTTOM2\n 8 BLACKTOP1,BLACKEMPTY5,GREYTOP3\n 8 GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYTOP3\n 8 BLACKEMPTY3\n 8 BLACKEMPTY3,BLACKLEFT2\n 8 GREYBOTTOM1\n 8 GREYBOTTOM1,GREYEMPTY2\n 8 BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4\n 8 GREYLEFT1\n 8 GREYLEFT1,GREYEMPTY2\n 8 GREYEMPTY5\n 8 GREYBOTTOM2,GREYRIGHT2,BLACKLEFT2\n 8 GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2,BLACKLEFT2\n 8 GREYBOTTOM2,GREYRIGHT2,BLACKLEFT2,GREYTOP3\n 8 GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2,BLACKLEFT2,GREYTOP3\n 8 GREYBOTTOM2,GREYEMPTY4\n 8 GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4\n 8 GREYBOTTOM2,GREYEMPTY4,GREYLEFT3\n 8 GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3\n 8 GREYRIGHT2,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 8 GREYRIGHT2,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 8 GREYRIGHT2,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 8 GREYRIGHT3\n 8 BLACKRIGHT3,GREYRIGHT3\n 8 GREYRIGHT3,BLACKLEFT3\n 8 BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3\n 8 GREYRIGHT3,BLACKLEFT3,BLACKLEFT2\n 8 BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2\n 8 GREYRIGHT4,GREYEMPTY3,BLACKTOP4\n 8 BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKTOP4\n 8 GREYRIGHT4,GREYEMPTY3,BLACKTOP4,GREYBOTTOM2\n 8 BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKTOP4,GREYBOTTOM2\n 8 GREYEMPTY3,BLACKLEFT4\n 7 GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYTOP3\n 7 GREYBOTTOM2,BLACKEMPTY3\n 7 GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2\n 7 GREYRIGHT4,BLACKTOP4\n 7 BLACKLEFT2,GREYRIGHT4,BLACKTOP4\n 7 GREYRIGHT4,BLACKTOP4,GREYBOTTOM2\n 7 BLACKLEFT2,GREYRIGHT4,BLACKTOP4,GREYBOTTOM2\n 7 GREYBOTTOM2,GREYLEFT3\n 7 GREYLEFT1,GREYEMPTY2,BLACKEMPTY4\n 7 GREYTOP1\n 7 GREYTOP1,GREYEMPTY2\n 7 GREYTOP1,GREYEMPTY2,BLACKBOTTOM3\n 7 GREYEMPTY5,BLACKEMPTY2\n 7 GREYEMPTY5,BLACKEMPTY2,BLACKBOTTOM3\n 7 GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1\n 7 GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1\n 7 GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5\n 7 GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5\n 7 GREYBOTTOM2,GREYRIGHT2,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 7 GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 7 GREYBOTTOM2,GREYRIGHT2,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 7 GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 7 GREYBOTTOM2,GREYRIGHT2,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 7 GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 7 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4\n 7 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4\n 7 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4\n 7 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3\n 7 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3\n 7 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3\n 7 GREYBOTTOM2,BLACKRIGHT3\n 7 GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3\n 7 GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3\n 7 GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2\n 6 GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3\n 6 GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2\n 6 GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4\n 6 BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2\n 6 GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4\n 6 GREYRIGHT2,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4\n 6 GREYBOTTOM2,GREYRIGHT2,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4\n 6 GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4\n 6 GREYRIGHT2,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3\n 6 GREYBOTTOM2,GREYRIGHT2,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3\n 6 BLACKEMPTY5,GREYTOP3,GREYEMPTY3\n 6 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 6 GREYBOTTOM2,GREYBOTTOM1\n 6 GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2\n 6 BLACKBOTTOM3,BLACKLEFT2,GREYTOP3\n 6 GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3\n 6 BLACKEMPTY4,GREYEMPTY3\n 6 GREYEMPTY2,BLACKEMPTY4,GREYEMPTY3\n 6 GREYLEFT2,BLACKLEFT4\n 6 BLACKLEFT4,GREYEMPTY3\n 6 GREYEMPTY3,BLACKLEFT4,GREYEMPTY3\n 6 BLACKLEFT4,GREYEMPTY3,BLACKBOTTOM4\n 6 GREYEMPTY3,BLACKLEFT4,GREYEMPTY3,BLACKBOTTOM4\n 6 BLACKLEFT4,GREYEMPTY3,BLACKBOTTOM4,GREYBOTTOM2\n 6 GREYEMPTY3,BLACKLEFT4,GREYEMPTY3,BLACKBOTTOM4,GREYBOTTOM2\n 6 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1\n 6 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1\n 6 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1\n 6 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5\n 6 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5\n 6 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5\n 5 GREYRIGHT5\n 5 GREYTOP5\n 5 BLACKLEFT3,BLACKLEFT2,GREYRIGHT4\n 5 BLACKTOP4,GREYBOTTOM2,GREYLEFT3\n 5 GREYRIGHT4,BLACKTOP4,GREYBOTTOM2,GREYLEFT3\n 5 BLACKLEFT2,GREYRIGHT4,BLACKTOP4,GREYBOTTOM2,GREYLEFT3\n 5 BLACKEMPTY4,BLACKLEFT2,GREYRIGHT4\n 5 GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYRIGHT4\n 5 BLACKLEFT1\n 5 BLACKLEFT1,GREYEMPTY2\n 5 BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3\n 5 GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYTOP3\n 5 GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYTOP3\n 5 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 5 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3\n 5 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2\n 5 BLACKEMPTY3,BLACKLEFT2,GREYTOP3\n 5 BLACKEMPTY3,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 5 BLACKLEFT3,BLACKLEFT2,GREYTOP3\n 5 GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYTOP3\n 5 BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYTOP3\n 5 GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4\n 5 GREYBOTTOM2,BLACKBOTTOM1\n 5 GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2\n 5 GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3\n 5 GREYTOP1,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3\n 5 BLACKLEFT5\n 5 BLACKLEFT5,GREYEMPTY2\n 5 BLACKLEFT5,GREYEMPTY2,BLACKBOTTOM3\n 5 GREYRIGHT2,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1\n 5 BLACKTOP1,BLACKEMPTY5,GREYTOP3,GREYEMPTY3\n 5 GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYTOP3,GREYEMPTY3\n 5 BLACKTOP1,BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 5 GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 5 BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT4\n 5 BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT4,GREYEMPTY3\n 5 BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT4,GREYEMPTY3,BLACKBOTTOM4\n 5 BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT4,GREYEMPTY3,BLACKBOTTOM4,GREYBOTTOM2\n 4 GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2\n 4 GREYLEFT1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2\n 4 BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2\n 4 GREYLEFT5\n 4 BLACKEMPTY1\n 4 BLACKEMPTY1,BLACKEMPTY2\n 4 BLACKEMPTY1,BLACKEMPTY2,BLACKBOTTOM3\n 4 GREYRIGHT1\n 4 GREYRIGHT1,GREYEMPTY2\n 4 GREYRIGHT1,GREYEMPTY2,BLACKEMPTY4\n 4 BLACKBOTTOM5\n 4 BLACKLEFT2,BLACKBOTTOM5\n 4 GREYTOP4\n 4 GREYBOTTOM2,GREYTOP4\n 4 GREYTOP2\n 4 GREYTOP4,GREYTOP2\n 4 GREYBOTTOM2,GREYTOP4,GREYTOP2\n 4 BLACKTOP5\n 4 GREYTOP2,BLACKTOP5\n 4 GREYTOP4,GREYTOP2,BLACKTOP5\n 4 GREYBOTTOM2,GREYTOP4,GREYTOP2,BLACKTOP5\n 4 BLACKRIGHT4\n 4 BLACKRIGHT4,GREYLEFT2\n 4 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3\n 4 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3\n 4 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2\n 4 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2\n 4 GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYTOP3\n 4 GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYTOP3\n 4 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYTOP3\n 4 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYTOP3\n 4 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYTOP3\n 4 GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 4 GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 4 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 4 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 4 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 4 BLACKEMPTY3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 4 BLACKEMPTY3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 4 GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3\n 4 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3\n 4 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3\n 4 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3\n 4 BLACKEMPTY3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3\n 4 GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3\n 4 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3\n 4 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3\n 4 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3\n 4 BLACKEMPTY3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3\n 4 GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3\n 4 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3\n 4 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3\n 4 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3\n 4 BLACKEMPTY3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3\n 4 GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2\n 4 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2\n 4 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2\n 4 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2\n 4 GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYTOP3\n 4 BLACKLEFT3,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 4 GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 4 BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 4 GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 4 GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1\n 4 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1\n 4 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1\n 4 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1\n 4 GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2\n 4 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2\n 4 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2\n 4 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2\n 4 BLACKEMPTY4,BLACKLEFT2,GREYTOP3\n 4 GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3\n 4 BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 4 GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 4 BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 4 GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 4 BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 4 GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 4 GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2\n 4 GREYBOTTOM2,GREYLEFT1\n 4 GREYBOTTOM2,GREYLEFT1,GREYEMPTY2\n 4 GREYBOTTOM2,GREYLEFT1,GREYEMPTY2,BLACKEMPTY4\n 4 BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 4 GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 4 BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 4 GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 4 BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 4 GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 4 BLACKLEFT5,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3\n 4 BLACKEMPTY2,BLACKBOTTOM3,GREYEMPTY3\n 4 GREYEMPTY3,GREYEMPTY1\n 4 BLACKBOTTOM3,GREYEMPTY3,GREYEMPTY1\n 4 GREYEMPTY3,GREYEMPTY1,GREYEMPTY2\n 4 BLACKBOTTOM3,GREYEMPTY3,GREYEMPTY1,GREYEMPTY2\n 4 BLACKTOP3\n 4 GREYEMPTY2,BLACKTOP3\n 4 BLACKLEFT4,GREYLEFT2\n 4 GREYLEFT2,BLACKLEFT4,GREYLEFT2\n 4 BLACKLEFT4,GREYLEFT2,BLACKBOTTOM4\n 4 GREYLEFT2,BLACKLEFT4,GREYLEFT2,BLACKBOTTOM4\n 4 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYTOP3\n 4 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYTOP3\n 4 GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYTOP3,GREYEMPTY3\n 4 GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 4 BLACKTOP1,BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 4 GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 4 GREYTOP3,GREYLEFT2\n 4 BLACKBOTTOM2\n 3 BLACKEMPTY3,BLACKLEFT2,GREYRIGHT4\n 3 GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYRIGHT4\n 3 BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4\n 3 BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4,BLACKTOP4\n 3 BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4,BLACKTOP4,GREYBOTTOM2\n 3 BLACKRIGHT5\n 3 BLACKBOTTOM3,GREYLEFT3\n 3 BLACKRIGHT1\n 3 BLACKRIGHT1,GREYEMPTY2\n 3 BLACKTOP4,GREYLEFT2\n 3 GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 3 GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 3 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 3 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 3 GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 3 GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 3 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 3 GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3\n 3 GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3\n 3 GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3\n 3 GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYTOP3\n 3 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYTOP3\n 3 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYTOP3\n 3 GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 3 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 3 BLACKLEFT3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 3 GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 3 BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 3 GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 3 GREYBOTTOM5,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 3 BLACKLEFT3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 3 GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 3 BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 3 GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 3 BLACKLEFT3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1\n 3 GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1\n 3 BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1\n 3 BLACKLEFT3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2\n 3 GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2\n 3 GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4\n 3 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4\n 3 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4\n 3 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4\n 3 BLACKLEFT3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4\n 3 GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2\n 3 GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1\n 3 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1\n 3 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1\n 3 GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2\n 3 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2\n 3 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2\n 3 GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3\n 3 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3\n 3 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3\n 3 GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2\n 3 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2\n 3 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2\n 3 BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3\n 3 BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKTOP4\n 3 BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKTOP4,GREYBOTTOM2\n 3 BLACKTOP4,GREYBOTTOM2,GREYLEFT1\n 3 GREYEMPTY3,BLACKTOP4,GREYBOTTOM2,GREYLEFT1\n 3 GREYRIGHT4,GREYEMPTY3,BLACKTOP4,GREYBOTTOM2,GREYLEFT1\n 3 BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKTOP4,GREYBOTTOM2,GREYLEFT1\n 3 BLACKTOP4,GREYBOTTOM2,GREYLEFT1,GREYEMPTY2\n 3 GREYEMPTY3,BLACKTOP4,GREYBOTTOM2,GREYLEFT1,GREYEMPTY2\n 3 GREYRIGHT4,GREYEMPTY3,BLACKTOP4,GREYBOTTOM2,GREYLEFT1,GREYEMPTY2\n 3 BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKTOP4,GREYBOTTOM2,GREYLEFT1,GREYEMPTY2\n 3 BLACKTOP4,GREYBOTTOM2,GREYLEFT1,GREYEMPTY2,BLACKEMPTY4\n 3 GREYEMPTY3,BLACKTOP4,GREYBOTTOM2,GREYLEFT1,GREYEMPTY2,BLACKEMPTY4\n 3 GREYRIGHT4,GREYEMPTY3,BLACKTOP4,GREYBOTTOM2,GREYLEFT1,GREYEMPTY2,BLACKEMPTY4\n 3 BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKTOP4,GREYBOTTOM2,GREYLEFT1,GREYEMPTY2,BLACKEMPTY4\n 3 GREYLEFT2,GREYTOP1\n 3 GREYLEFT2,GREYTOP1,GREYEMPTY2\n 3 GREYLEFT2,GREYTOP1,GREYEMPTY2,BLACKBOTTOM3\n 3 GREYLEFT2,GREYTOP1,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3\n 3 GREYEMPTY3,BLACKLEFT5\n 3 BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT5\n 3 GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT5\n 3 GREYTOP1,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT5\n 3 GREYEMPTY3,BLACKLEFT5,GREYEMPTY2\n 3 BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT5,GREYEMPTY2\n 3 GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT5,GREYEMPTY2\n 3 GREYTOP1,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT5,GREYEMPTY2\n 3 GREYEMPTY3,BLACKLEFT5,GREYEMPTY2,BLACKBOTTOM3\n 3 BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT5,GREYEMPTY2,BLACKBOTTOM3\n 3 GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT5,GREYEMPTY2,BLACKBOTTOM3\n 3 GREYTOP1,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT5,GREYEMPTY2,BLACKBOTTOM3\n 3 GREYEMPTY3,BLACKLEFT5,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3\n 3 BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT5,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3\n 3 GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT5,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3\n 3 GREYTOP1,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT5,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3\n 3 GREYEMPTY5,BLACKEMPTY2,BLACKBOTTOM3,GREYEMPTY3\n 3 GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYTOP3,GREYEMPTY3\n 3 GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYTOP3,GREYEMPTY3\n 3 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYTOP3,GREYEMPTY3\n 3 GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 3 GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 3 GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYRIGHT4\n 3 BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYRIGHT4\n 3 GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYRIGHT4\n 3 BLACKLEFT3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3\n 3 GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3\n 3 BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3\n 3 GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3\n 3 GREYLEFT1,GREYEMPTY2,BLACKEMPTY4,GREYEMPTY3\n 3 GREYEMPTY3,GREYLEFT1\n 3 GREYEMPTY3,GREYLEFT1,GREYEMPTY2\n 3 BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3\n 3 GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,GREYEMPTY1\n 3 GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,GREYEMPTY1,GREYEMPTY2\n 3 BLACKBOTTOM3,GREYLEFT2\n 3 BLACKBOTTOM3,GREYLEFT2,BLACKLEFT4\n 3 BLACKBOTTOM3,GREYLEFT2,BLACKLEFT4,GREYLEFT2\n 3 BLACKBOTTOM3,GREYLEFT2,BLACKLEFT4,GREYLEFT2,BLACKBOTTOM4\n 3 GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT4\n 3 GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT4,GREYEMPTY3\n 3 GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT4,GREYEMPTY3,BLACKBOTTOM4\n 3 GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT4,GREYEMPTY3,BLACKBOTTOM4,GREYBOTTOM2\n 3 GREYEMPTY2,GREYEMPTY1\n 3 GREYEMPTY1,GREYEMPTY2,GREYEMPTY1\n 3 GREYEMPTY2,GREYEMPTY1,GREYEMPTY2\n 3 GREYEMPTY1,GREYEMPTY2,GREYEMPTY1,GREYEMPTY2\n 3 GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 3 BLACKTOP1,BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3\n 3 GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3\n 3 BLACKTOP1,BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2\n 3 GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2\n 3 BLACKTOP1,BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYTOP3\n 3 BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3\n 3 BLACKBOTTOM4,GREYBOTTOM4\n 3 GREYLEFT2,BLACKBOTTOM4,GREYBOTTOM4\n 3 BLACKBOTTOM4,GREYBOTTOM4,GREYBOTTOM2\n 3 GREYLEFT2,BLACKBOTTOM4,GREYBOTTOM4,GREYBOTTOM2\n 3 BLACKBOTTOM4,GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2\n 3 GREYLEFT2,BLACKBOTTOM4,GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2\n 3 BLACKBOTTOM4,GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2,BLACKLEFT2\n 3 GREYLEFT2,BLACKBOTTOM4,GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2,BLACKLEFT2\n 3 BLACKBOTTOM4,GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2,BLACKLEFT2,GREYTOP3\n 3 GREYLEFT2,BLACKBOTTOM4,GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2,BLACKLEFT2,GREYTOP3\n 3 BLACKBOTTOM4,GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 3 GREYLEFT2,BLACKBOTTOM4,GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 3 BLACKBOTTOM4,GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 3 GREYLEFT2,BLACKBOTTOM4,GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 3 BLACKBOTTOM4,GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 3 GREYRIGHT4,GREYEMPTY3,BLACKBOTTOM4\n 3 GREYRIGHT4,GREYEMPTY3,BLACKBOTTOM4,GREYBOTTOM2\n 3 BLACKLEFT2,GREYTOP3,GREYLEFT2\n 3 GREYTOP3,GREYLEFT2,BLACKBOTTOM4\n 3 BLACKLEFT2,GREYTOP3,GREYLEFT2,BLACKBOTTOM4\n 3 GREYRIGHT2,BLACKBOTTOM2\n 3 GREYBOTTOM2,GREYRIGHT2,BLACKBOTTOM2\n 3 GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2,BLACKBOTTOM2\n 2 BLACKRIGHT2\n 2 GREYLEFT2,BLACKRIGHT2\n 2 GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYRIGHT4\n 2 GREYBOTTOM2,GREYTOP5\n 2 GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYRIGHT4\n 2 BLACKEMPTY4,BLACKLEFT2,GREYRIGHT4,BLACKTOP4\n 2 GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYRIGHT4,BLACKTOP4\n 2 BLACKEMPTY4,BLACKLEFT2,GREYRIGHT4,BLACKTOP4,GREYBOTTOM2\n 2 GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYRIGHT4,BLACKTOP4,GREYBOTTOM2\n 2 BLACKEMPTY4,BLACKLEFT2,GREYRIGHT4,BLACKTOP4,GREYBOTTOM2,GREYLEFT3\n 2 GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYRIGHT4,BLACKTOP4,GREYBOTTOM2,GREYLEFT3\n 2 GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4,BLACKTOP4\n 2 GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4,BLACKTOP4,GREYBOTTOM2\n 2 BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4,BLACKTOP4,GREYBOTTOM2,GREYLEFT3\n 2 GREYLEFT1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYRIGHT4\n 2 BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4\n 2 GREYLEFT5,GREYLEFT2\n 2 BLACKRIGHT5,GREYRIGHT5\n 2 BLACKTOP2\n 2 GREYEMPTY2,BLACKBOTTOM3,GREYLEFT3\n 2 BLACKEMPTY2,BLACKBOTTOM3,BLACKLEFT2\n 2 BLACKEMPTY1,BLACKEMPTY2,BLACKBOTTOM3,BLACKLEFT2\n 2 BLACKEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4\n 2 BLACKEMPTY1,BLACKEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4\n 2 GREYRIGHT4,BLACKBOTTOM4\n 2 BLACKLEFT2,GREYRIGHT4,BLACKBOTTOM4\n 2 GREYRIGHT4,BLACKBOTTOM4,GREYBOTTOM2\n 2 BLACKLEFT2,GREYRIGHT4,BLACKBOTTOM4,GREYBOTTOM2\n 2 BLACKBOTTOM4,GREYBOTTOM2,GREYLEFT3\n 2 GREYRIGHT4,BLACKBOTTOM4,GREYBOTTOM2,GREYLEFT3\n 2 BLACKLEFT2,GREYRIGHT4,BLACKBOTTOM4,GREYBOTTOM2,GREYLEFT3\n 2 GREYTOP3,GREYBOTTOM2\n 2 GREYLEFT5,BLACKTOP4\n 2 GREYLEFT5,BLACKTOP4,GREYLEFT2\n 2 GREYLEFT2,BLACKLEFT2\n 2 BLACKTOP4,GREYLEFT2,BLACKLEFT2\n 2 GREYLEFT5,BLACKTOP4,GREYLEFT2,BLACKLEFT2\n 2 GREYLEFT2,GREYLEFT3\n 2 BLACKRIGHT4,GREYLEFT2,GREYLEFT3\n 2 GREYLEFT3,BLACKLEFT2\n 2 GREYLEFT2,GREYLEFT3,BLACKLEFT2\n 2 BLACKRIGHT4,GREYLEFT2,GREYLEFT3,BLACKLEFT2\n 2 GREYLEFT3,BLACKLEFT2,BLACKBOTTOM5\n 2 GREYLEFT2,GREYLEFT3,BLACKLEFT2,BLACKBOTTOM5\n 2 BLACKRIGHT4,GREYLEFT2,GREYLEFT3,BLACKLEFT2,BLACKBOTTOM5\n 2 GREYBOTTOM5,GREYLEFT2\n 2 GREYEMPTY3,GREYBOTTOM5,GREYLEFT2\n 2 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYLEFT2\n 2 GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2\n 2 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2\n 2 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2\n 2 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2\n 2 GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3\n 2 GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3\n 2 GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3\n 2 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3\n 2 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3\n 2 GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 2 GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 2 GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 2 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 2 GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 2 GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 2 GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 2 GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 2 GREYBOTTOM2,GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 2 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1\n 2 BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1\n 2 GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1\n 2 GREYBOTTOM1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1\n 2 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2\n 2 BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2\n 2 GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2\n 2 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3\n 2 BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3\n 2 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2\n 2 GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4\n 2 GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3\n 2 BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3\n 2 GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3\n 2 GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKTOP4\n 2 BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKTOP4\n 2 GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKTOP4\n 2 GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKTOP4,GREYBOTTOM2\n 2 BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKTOP4,GREYBOTTOM2\n 2 GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKTOP4,GREYBOTTOM2\n 2 GREYBOTTOM2,GREYLEFT1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2\n 2 GREYLEFT1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3\n 2 GREYBOTTOM2,GREYLEFT1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3\n 2 GREYLEFT1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 2 GREYBOTTOM2,GREYLEFT1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 2 GREYLEFT1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 2 GREYBOTTOM2,GREYLEFT1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 2 GREYLEFT1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 2 GREYBOTTOM2,GREYLEFT1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 2 GREYBOTTOM2,BLACKLEFT1\n 2 GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1\n 2 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1\n 2 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1\n 2 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1\n 2 BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1\n 2 GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1\n 2 GREYLEFT1,GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1\n 2 GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2\n 2 GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2\n 2 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2\n 2 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2\n 2 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2\n 2 BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2\n 2 GREYEMPTY2,BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2\n 2 GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3\n 2 GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3\n 2 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3\n 2 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3\n 2 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3\n 2 BLACKEMPTY4,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3\n 2 GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2\n 2 GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2\n 2 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2\n 2 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2\n 2 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2\n 2 BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3\n 2 GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3\n 2 GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3\n 2 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3\n 2 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3\n 2 BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 2 GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 2 GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 2 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 2 BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 2 GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 2 GREYBOTTOM5,GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 2 BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 2 GREYBOTTOM2,BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 2 GREYBOTTOM5,GREYBOTTOM2,GREYTOP4\n 2 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYTOP4\n 2 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYTOP4\n 2 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYTOP4\n 2 BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYTOP4\n 2 GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYTOP4\n 2 BLACKLEFT1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYTOP4\n 2 GREYBOTTOM5,GREYBOTTOM2,GREYTOP4,GREYTOP2\n 2 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYTOP4,GREYTOP2\n 2 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYTOP4,GREYTOP2\n 2 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYTOP4,GREYTOP2\n 2 BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYTOP4,GREYTOP2\n 2 GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYTOP4,GREYTOP2\n 2 GREYBOTTOM5,GREYBOTTOM2,GREYTOP4,GREYTOP2,BLACKTOP5\n 2 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYTOP4,GREYTOP2,BLACKTOP5\n 2 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYTOP4,GREYTOP2,BLACKTOP5\n 2 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYTOP4,GREYTOP2,BLACKTOP5\n 2 BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYTOP4,GREYTOP2,BLACKTOP5\n 2 BLACKTOP5,GREYLEFT2\n 2 GREYTOP2,BLACKTOP5,GREYLEFT2\n 2 GREYTOP4,GREYTOP2,BLACKTOP5,GREYLEFT2\n 2 GREYBOTTOM2,GREYTOP4,GREYTOP2,BLACKTOP5,GREYLEFT2\n 2 GREYBOTTOM5,GREYBOTTOM2,GREYTOP4,GREYTOP2,BLACKTOP5,GREYLEFT2\n 2 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYTOP4,GREYTOP2,BLACKTOP5,GREYLEFT2\n 2 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYTOP4,GREYTOP2,BLACKTOP5,GREYLEFT2\n 2 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYTOP4,GREYTOP2,BLACKTOP5,GREYLEFT2\n 2 BLACKTOP5,GREYLEFT2,GREYTOP1\n 2 GREYTOP2,BLACKTOP5,GREYLEFT2,GREYTOP1\n 2 GREYTOP4,GREYTOP2,BLACKTOP5,GREYLEFT2,GREYTOP1\n 2 GREYBOTTOM2,GREYTOP4,GREYTOP2,BLACKTOP5,GREYLEFT2,GREYTOP1\n 2 GREYBOTTOM5,GREYBOTTOM2,GREYTOP4,GREYTOP2,BLACKTOP5,GREYLEFT2,GREYTOP1\n 2 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYTOP4,GREYTOP2,BLACKTOP5,GREYLEFT2,GREYTOP1\n 2 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYTOP4,GREYTOP2,BLACKTOP5,GREYLEFT2,GREYTOP1\n 2 BLACKTOP5,GREYLEFT2,GREYTOP1,GREYEMPTY2\n 2 GREYTOP2,BLACKTOP5,GREYLEFT2,GREYTOP1,GREYEMPTY2\n 2 GREYTOP4,GREYTOP2,BLACKTOP5,GREYLEFT2,GREYTOP1,GREYEMPTY2\n 2 GREYBOTTOM2,GREYTOP4,GREYTOP2,BLACKTOP5,GREYLEFT2,GREYTOP1,GREYEMPTY2\n 2 GREYBOTTOM5,GREYBOTTOM2,GREYTOP4,GREYTOP2,BLACKTOP5,GREYLEFT2,GREYTOP1,GREYEMPTY2\n 2 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYTOP4,GREYTOP2,BLACKTOP5,GREYLEFT2,GREYTOP1,GREYEMPTY2\n 2 BLACKTOP5,GREYLEFT2,GREYTOP1,GREYEMPTY2,BLACKBOTTOM3\n 2 GREYTOP2,BLACKTOP5,GREYLEFT2,GREYTOP1,GREYEMPTY2,BLACKBOTTOM3\n 2 GREYTOP4,GREYTOP2,BLACKTOP5,GREYLEFT2,GREYTOP1,GREYEMPTY2,BLACKBOTTOM3\n 2 GREYBOTTOM2,GREYTOP4,GREYTOP2,BLACKTOP5,GREYLEFT2,GREYTOP1,GREYEMPTY2,BLACKBOTTOM3\n 2 GREYBOTTOM5,GREYBOTTOM2,GREYTOP4,GREYTOP2,BLACKTOP5,GREYLEFT2,GREYTOP1,GREYEMPTY2,BLACKBOTTOM3\n 2 BLACKTOP5,GREYLEFT2,GREYTOP1,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3\n 2 GREYTOP2,BLACKTOP5,GREYLEFT2,GREYTOP1,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3\n 2 GREYTOP4,GREYTOP2,BLACKTOP5,GREYLEFT2,GREYTOP1,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3\n 2 GREYBOTTOM2,GREYTOP4,GREYTOP2,BLACKTOP5,GREYLEFT2,GREYTOP1,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3\n 2 GREYLEFT2,GREYTOP1,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT5\n 2 BLACKTOP5,GREYLEFT2,GREYTOP1,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT5\n 2 GREYTOP2,BLACKTOP5,GREYLEFT2,GREYTOP1,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT5\n 2 GREYTOP4,GREYTOP2,BLACKTOP5,GREYLEFT2,GREYTOP1,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT5\n 2 GREYLEFT2,GREYTOP1,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT5,GREYEMPTY2\n 2 BLACKTOP5,GREYLEFT2,GREYTOP1,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT5,GREYEMPTY2\n 2 GREYTOP2,BLACKTOP5,GREYLEFT2,GREYTOP1,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT5,GREYEMPTY2\n 2 GREYLEFT2,GREYTOP1,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT5,GREYEMPTY2,BLACKBOTTOM3\n 2 BLACKTOP5,GREYLEFT2,GREYTOP1,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT5,GREYEMPTY2,BLACKBOTTOM3\n 2 GREYLEFT2,GREYTOP1,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT5,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3\n 2 GREYEMPTY3,GREYEMPTY5\n 2 GREYEMPTY3,GREYEMPTY5,BLACKEMPTY2\n 2 GREYEMPTY3,GREYEMPTY5,BLACKEMPTY2,BLACKBOTTOM3\n 2 GREYEMPTY3,GREYEMPTY5,BLACKEMPTY2,BLACKBOTTOM3,GREYEMPTY3\n 2 GREYBOTTOM2,BLACKEMPTY1\n 2 GREYBOTTOM2,BLACKEMPTY1,BLACKEMPTY2\n 2 GREYBOTTOM2,BLACKEMPTY1,BLACKEMPTY2,BLACKBOTTOM3\n 2 GREYBOTTOM2,GREYRIGHT1\n 2 GREYBOTTOM2,GREYRIGHT1,GREYEMPTY2\n 2 GREYBOTTOM2,GREYRIGHT1,GREYEMPTY2,BLACKEMPTY4\n 2 GREYRIGHT1,GREYEMPTY2,BLACKEMPTY4,GREYEMPTY3\n 2 BLACKTOP4,GREYEMPTY3\n 2 GREYBOTTOM2,BLACKRIGHT1\n 2 GREYBOTTOM2,BLACKRIGHT1,GREYEMPTY2\n 2 BLACKRIGHT1,GREYEMPTY2,BLACKTOP3\n 2 GREYBOTTOM2,BLACKRIGHT1,GREYEMPTY2,BLACKTOP3\n 2 BLACKTOP3,GREYEMPTY3\n 2 GREYEMPTY2,BLACKTOP3,GREYEMPTY3\n 2 BLACKBOTTOM4,GREYEMPTY3\n 2 GREYEMPTY3,BLACKBOTTOM4,GREYEMPTY3\n 2 BLACKBOTTOM4,GREYEMPTY3,BLACKLEFT4\n 2 GREYEMPTY3,BLACKBOTTOM4,GREYEMPTY3,BLACKLEFT4\n 2 BLACKLEFT4,GREYBOTTOM2\n 2 GREYEMPTY3,BLACKLEFT4,GREYBOTTOM2\n 2 BLACKBOTTOM4,GREYEMPTY3,BLACKLEFT4,GREYBOTTOM2\n 2 GREYEMPTY3,BLACKBOTTOM4,GREYEMPTY3,BLACKLEFT4,GREYBOTTOM2\n 2 BLACKTOP3,GREYLEFT2\n 2 GREYEMPTY2,BLACKTOP3,GREYLEFT2\n 2 BLACKEMPTY4,GREYLEFT2\n 2 GREYEMPTY2,BLACKEMPTY4,GREYLEFT2\n 2 BLACKBOTTOM4,GREYLEFT2\n 2 GREYLEFT2,BLACKBOTTOM4,GREYLEFT2\n 2 BLACKBOTTOM4,GREYLEFT2,BLACKLEFT4\n 2 GREYLEFT2,BLACKBOTTOM4,GREYLEFT2,BLACKLEFT4\n 2 BLACKLEFT3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKTOP4\n 2 GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKTOP4\n 2 BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKTOP4\n 2 GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKTOP4\n 2 BLACKLEFT3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKTOP4,GREYBOTTOM2\n 2 GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKTOP4,GREYBOTTOM2\n 2 BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKTOP4,GREYBOTTOM2\n 2 GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKTOP4,GREYBOTTOM2\n 2 GREYBOTTOM2,GREYLEFT1,GREYEMPTY2,BLACKEMPTY4,GREYEMPTY3\n 2 BLACKTOP4,GREYBOTTOM2,GREYLEFT1,GREYEMPTY2,BLACKEMPTY4,GREYEMPTY3\n 2 GREYEMPTY3,BLACKTOP4,GREYBOTTOM2,GREYLEFT1,GREYEMPTY2,BLACKEMPTY4,GREYEMPTY3\n 2 GREYRIGHT4,GREYEMPTY3,BLACKTOP4,GREYBOTTOM2,GREYLEFT1,GREYEMPTY2,BLACKEMPTY4,GREYEMPTY3\n 2 BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKTOP4,GREYBOTTOM2,GREYLEFT1,GREYEMPTY2,BLACKEMPTY4,GREYEMPTY3\n 2 GREYEMPTY3,BLACKBOTTOM1\n 2 GREYEMPTY3,BLACKBOTTOM1,GREYEMPTY2\n 2 GREYEMPTY3,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3\n 2 GREYEMPTY3,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3\n 2 GREYBOTTOM2,BLACKLEFT5\n 2 GREYBOTTOM2,BLACKLEFT5,GREYEMPTY2\n 2 GREYBOTTOM2,BLACKLEFT5,GREYEMPTY2,BLACKBOTTOM3\n 2 GREYEMPTY1,GREYEMPTY2,BLACKTOP3\n 2 GREYEMPTY3,GREYEMPTY1,GREYEMPTY2,BLACKTOP3\n 2 BLACKBOTTOM3,GREYEMPTY3,GREYEMPTY1,GREYEMPTY2,BLACKTOP3\n 2 GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,GREYEMPTY1,GREYEMPTY2,BLACKTOP3\n 2 BLACKEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT4\n 2 BLACKEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT4,GREYEMPTY3\n 2 BLACKEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT4,GREYEMPTY3,BLACKBOTTOM4\n 2 BLACKEMPTY2,BLACKBOTTOM3,GREYEMPTY3,BLACKLEFT4,GREYEMPTY3,BLACKBOTTOM4,GREYBOTTOM2\n 2 BLACKEMPTY2,BLACKBOTTOM3,GREYLEFT2\n 2 BLACKEMPTY2,BLACKBOTTOM3,GREYLEFT2,BLACKLEFT4\n 2 BLACKEMPTY2,BLACKBOTTOM3,GREYLEFT2,BLACKLEFT4,GREYLEFT2\n 2 BLACKEMPTY2,BLACKBOTTOM3,GREYLEFT2,BLACKLEFT4,GREYLEFT2,BLACKBOTTOM4\n 2 BLACKBOTTOM4,GREYEMPTY5\n 2 GREYLEFT2,BLACKBOTTOM4,GREYEMPTY5\n 2 BLACKLEFT4,GREYLEFT2,BLACKBOTTOM4,GREYEMPTY5\n 2 GREYLEFT2,BLACKLEFT4,GREYLEFT2,BLACKBOTTOM4,GREYEMPTY5\n 2 BLACKBOTTOM3,GREYLEFT2,BLACKLEFT4,GREYLEFT2,BLACKBOTTOM4,GREYEMPTY5\n 2 BLACKEMPTY2,BLACKBOTTOM3,GREYLEFT2,BLACKLEFT4,GREYLEFT2,BLACKBOTTOM4,GREYEMPTY5\n 2 BLACKBOTTOM4,GREYEMPTY5,BLACKEMPTY2\n 2 GREYLEFT2,BLACKBOTTOM4,GREYEMPTY5,BLACKEMPTY2\n 2 BLACKLEFT4,GREYLEFT2,BLACKBOTTOM4,GREYEMPTY5,BLACKEMPTY2\n 2 GREYLEFT2,BLACKLEFT4,GREYLEFT2,BLACKBOTTOM4,GREYEMPTY5,BLACKEMPTY2\n 2 BLACKBOTTOM3,GREYLEFT2,BLACKLEFT4,GREYLEFT2,BLACKBOTTOM4,GREYEMPTY5,BLACKEMPTY2\n 2 BLACKEMPTY2,BLACKBOTTOM3,GREYLEFT2,BLACKLEFT4,GREYLEFT2,BLACKBOTTOM4,GREYEMPTY5,BLACKEMPTY2\n 2 BLACKBOTTOM4,GREYEMPTY5,BLACKEMPTY2,BLACKBOTTOM3\n 2 GREYLEFT2,BLACKBOTTOM4,GREYEMPTY5,BLACKEMPTY2,BLACKBOTTOM3\n 2 BLACKLEFT4,GREYLEFT2,BLACKBOTTOM4,GREYEMPTY5,BLACKEMPTY2,BLACKBOTTOM3\n 2 GREYLEFT2,BLACKLEFT4,GREYLEFT2,BLACKBOTTOM4,GREYEMPTY5,BLACKEMPTY2,BLACKBOTTOM3\n 2 BLACKBOTTOM3,GREYLEFT2,BLACKLEFT4,GREYLEFT2,BLACKBOTTOM4,GREYEMPTY5,BLACKEMPTY2,BLACKBOTTOM3\n 2 BLACKEMPTY2,BLACKBOTTOM3,GREYLEFT2,BLACKLEFT4,GREYLEFT2,BLACKBOTTOM4,GREYEMPTY5,BLACKEMPTY2,BLACKBOTTOM3\n 2 GREYBOTTOM2,GREYEMPTY5\n 2 BLACKEMPTY5,GREYRIGHT4\n 2 BLACKTOP1,BLACKEMPTY5,GREYRIGHT4\n 2 GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYRIGHT4\n 2 GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYRIGHT4\n 2 GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYRIGHT4\n 2 GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYRIGHT4\n 2 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYRIGHT4\n 2 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYRIGHT4\n 2 BLACKEMPTY5,GREYRIGHT4,GREYEMPTY3\n 2 BLACKTOP1,BLACKEMPTY5,GREYRIGHT4,GREYEMPTY3\n 2 GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYRIGHT4,GREYEMPTY3\n 2 GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYRIGHT4,GREYEMPTY3\n 2 GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYRIGHT4,GREYEMPTY3\n 2 GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYRIGHT4,GREYEMPTY3\n 2 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYRIGHT4,GREYEMPTY3\n 2 BLACKEMPTY3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3\n 2 GREYBOTTOM2,BLACKEMPTY3,BLACKLEFT2,GREYRIGHT4,GREYEMPTY3\n 2 GREYEMPTY2,GREYEMPTY1,GREYEMPTY2,GREYEMPTY1\n 2 GREYEMPTY1,GREYEMPTY2,GREYEMPTY1,GREYEMPTY2,GREYEMPTY1\n 2 GREYEMPTY2,GREYEMPTY1,GREYEMPTY2,GREYEMPTY1,GREYEMPTY2\n 2 GREYEMPTY1,GREYEMPTY2,GREYEMPTY1,GREYEMPTY2,GREYEMPTY1,GREYEMPTY2\n 2 GREYBOTTOM2,GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 2 GREYEMPTY4,GREYLEFT3,BLACKTOP1,BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKEMPTY3\n 2 GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3\n 2 GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3\n 2 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3\n 2 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3\n 2 BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 2 GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 2 GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 2 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3\n 2 BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 2 GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 2 GREYBOTTOM5,GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5\n 2 BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 2 GREYBOTTOM2,BLACKBOTTOM1,GREYEMPTY2,BLACKBOTTOM3,BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2\n 2 GREYTOP3,BLACKRIGHT4\n 2 BLACKLEFT2,GREYTOP3,BLACKRIGHT4\n 2 GREYTOP3,BLACKRIGHT4,GREYLEFT2\n 2 BLACKLEFT2,GREYTOP3,BLACKRIGHT4,GREYLEFT2\n 2 BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKBOTTOM4\n 2 BLACKLEFT2,GREYRIGHT4,GREYEMPTY3,BLACKBOTTOM4,GREYBOTTOM2\n 2 BLACKBOTTOM4,GREYBOTTOM2,BLACKRIGHT3\n 2 GREYEMPTY3,BLACKBOTTOM4,GREYBOTTOM2,BLACKRIGHT3\n 2 BLACKBOTTOM4,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3\n 2 GREYEMPTY3,BLACKBOTTOM4,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3\n 2 BLACKBOTTOM4,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3\n 2 GREYEMPTY3,BLACKBOTTOM4,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3\n 2 BLACKBOTTOM4,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2\n 2 GREYEMPTY3,BLACKBOTTOM4,GREYBOTTOM2,BLACKRIGHT3,GREYRIGHT3,BLACKLEFT3,BLACKLEFT2\n 2 BLACKBOTTOM3,GREYEMPTY3,GREYLEFT1\n 2 GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,GREYLEFT1\n 2 BLACKBOTTOM3,GREYEMPTY3,GREYLEFT1,GREYEMPTY2\n 2 GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,GREYLEFT1,GREYEMPTY2\n 2 GREYEMPTY3,GREYLEFT1,GREYEMPTY2,BLACKEMPTY4\n 2 BLACKBOTTOM3,GREYEMPTY3,GREYLEFT1,GREYEMPTY2,BLACKEMPTY4\n 2 GREYEMPTY2,BLACKBOTTOM3,GREYEMPTY3,GREYLEFT1,GREYEMPTY2,BLACKEMPTY4\n 2 GREYRIGHT4,GREYLEFT2\n 2 BLACKLEFT2,GREYRIGHT4,GREYLEFT2\n 2 GREYRIGHT4,GREYLEFT2,BLACKBOTTOM4\n 2 BLACKLEFT2,GREYRIGHT4,GREYLEFT2,BLACKBOTTOM4\n 2 GREYTOP3,GREYEMPTY3,BLACKBOTTOM4\n 2 BLACKLEFT2,GREYTOP3,GREYEMPTY3,BLACKBOTTOM4\n 2 GREYBOTTOM2,GREYBOTTOM4\n 2 GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM4\n 2 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM4\n 2 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM4\n 2 GREYBOTTOM2,GREYBOTTOM4,GREYBOTTOM2\n 2 GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM4,GREYBOTTOM2\n 2 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM4,GREYBOTTOM2\n 2 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM4,GREYBOTTOM2\n 2 GREYBOTTOM2,GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2\n 2 GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2\n 2 GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2\n 2 GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2,GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2\n 2 BLACKBOTTOM2,GREYBOTTOM4\n 2 GREYRIGHT2,BLACKBOTTOM2,GREYBOTTOM4\n 2 GREYBOTTOM2,GREYRIGHT2,BLACKBOTTOM2,GREYBOTTOM4\n 2 GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2,BLACKBOTTOM2,GREYBOTTOM4\n 2 BLACKBOTTOM2,GREYBOTTOM4,GREYBOTTOM2\n 2 GREYRIGHT2,BLACKBOTTOM2,GREYBOTTOM4,GREYBOTTOM2\n 2 GREYBOTTOM2,GREYRIGHT2,BLACKBOTTOM2,GREYBOTTOM4,GREYBOTTOM2\n 2 GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2,BLACKBOTTOM2,GREYBOTTOM4,GREYBOTTOM2\n 2 BLACKBOTTOM2,GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2\n 2 GREYRIGHT2,BLACKBOTTOM2,GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2\n 2 GREYBOTTOM2,GREYRIGHT2,BLACKBOTTOM2,GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2\n 2 GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2,BLACKBOTTOM2,GREYBOTTOM4,GREYBOTTOM2,GREYRIGHT2\n 1 GREYRIGHT2,GREYLEFT2\n" ], [ "N = 8\n\n# ⇦ ⇨ ⇧ ⇩\n# ⬅ ➡ ⬆ ⬇\n\nVIZ = {\n 'BLACKTOP': '\\t⬆',\n 'BLACKRIGHT': '\\t➡',\n 'BLACKBOTTOM': '\\t⬇',\n 'BLACKLEFT': '\\t⬅',\n 'BLACKEMPTY': '\\t⬤',\n 'GREYTOP': '\\t⇧',\n 'GREYRIGHT': '\\t⇨',\n 'GREYBOTTOM': '\\t⇩',\n 'GREYLEFT': '\\t⇦',\n 'GREYEMPTY': '\\t◯',\n \n # '1': '▕',\n # '2': '▕▎',\n # '3': '▕▍',\n # '4': '▕▋',\n # '5': '▕▉',\n}\n\ndef print_lines(n_start, lines):\n for offset, line in enumerate(lines):\n n = n_start + offset\n if n % N == 0:\n nth = '*'\n else:\n nth = ' '\n #if len(line) > 2:\n # for k, v in VIZ.items():\n # line = line.replace(k, v)\n print(str(n).ljust(3), nth, line)\n", "_____no_output_____" ], [ "for chunk in parsed:\n buffer.clear()\n print('-' * 20)\n for i, piece in enumerate(chunk):\n if len(buffer) == buffer.maxlen:\n word = ','.join(buffer)\n if word in SEQUENCES:\n print_lines(i - buffer.maxlen, ['%s = %s' % (word, SEQUENCES[word])] * 5)\n buffer.clear()\n buffer.append(''.join(map(str, piece))) # Don't lose this piece.\n continue\n print_lines(i - buffer.maxlen, [buffer[0]])\n buffer.append(''.join(map(str, piece)))\n", "--------------------\n0 * GREYBOTTOM4\n1 GREYBOTTOM2\n2 GREYRIGHT2\n3 GREYLEFT2\n4 BLACKRIGHT2\n5 GREYLEFT2\n6 BLACKRIGHT2\n7 GREYEMPTY4\n8 * GREYLEFT3\n9 BLACKTOP1\n10 BLACKEMPTY5\n11 GREYTOP3\n12 GREYRIGHT5\n13 GREYBOTTOM5\n14 GREYBOTTOM2\n15 BLACKEMPTY3\n16 * BLACKLEFT2\n17 GREYRIGHT4\n18 BLACKTOP4\n19 GREYBOTTOM2\n20 GREYTOP5\n21 BLACKRIGHT3\n22 GREYLEFT3\n23 BLACKLEFT3\n24 * BLACKLEFT2\n25 GREYRIGHT4\n26 BLACKTOP4\n27 GREYBOTTOM2\n28 GREYLEFT3\n29 GREYBOTTOM1\n30 GREYEMPTY2\n31 BLACKEMPTY4\n32 * BLACKLEFT2\n33 GREYRIGHT4\n34 BLACKTOP4\n35 GREYBOTTOM2\n36 GREYLEFT3\n37 BLACKBOTTOM1\n38 GREYEMPTY2\n39 BLACKBOTTOM3\n40 * BLACKLEFT2\n41 GREYRIGHT4\n42 BLACKTOP4\n43 GREYBOTTOM2\n44 GREYLEFT3\n45 GREYLEFT1\n46 GREYEMPTY2\n47 BLACKEMPTY4\n48 * BLACKLEFT2\n49 GREYRIGHT4\n50 BLACKTOP4\n51 GREYBOTTOM2\n52 GREYLEFT3\n53 BLACKLEFT1\n54 GREYEMPTY2\n55 BLACKBOTTOM3\n56 * BLACKLEFT2\n57 GREYRIGHT4\n58 BLACKTOP4\n59 GREYBOTTOM2\n60 GREYRIGHT5\n61 GREYLEFT5\n62 GREYLEFT2\n63 BLACKRIGHT5\n64 * GREYRIGHT5\n65 BLACKRIGHT5\n66 GREYRIGHT5\n67 GREYLEFT4\n68 BLACKTOP2\n69 BLACKRIGHT5\n70 GREYEMPTY2\n71 BLACKBOTTOM3\n72 * GREYLEFT3\n73 GREYTOP1\n--------------------\n0 * GREYLEFT3\n1 GREYEMPTY5\n2 BLACKEMPTY2\n3 BLACKBOTTOM3\n4 GREYLEFT3\n5 BLACKEMPTY1\n6 BLACKEMPTY2\n7 BLACKBOTTOM3\n8 * BLACKLEFT2\n9 GREYRIGHT4\n10 BLACKTOP4\n11 GREYBOTTOM2\n12 GREYLEFT3\n13 GREYRIGHT1\n14 GREYEMPTY2\n15 BLACKEMPTY4\n16 * BLACKLEFT2\n17 GREYRIGHT4\n18 BLACKBOTTOM4\n19 GREYBOTTOM2\n20 GREYLEFT3\n21 BLACKRIGHT1\n22 GREYEMPTY2\n23 BLACKBOTTOM3\n24 * BLACKLEFT2\n25 GREYRIGHT4\n26 BLACKBOTTOM4\n27 GREYBOTTOM2\n28 GREYLEFT3\n29 GREYBOTTOM4\n30 GREYBOTTOM2\n31 GREYRIGHT2\n32 * BLACKLEFT2\n33 GREYTOP3\n34 GREYBOTTOM2\n35 GREYBOTTOM5\n36 GREYBOTTOM2\n37 GREYEMPTY4\n38 GREYLEFT3\n39 BLACKTOP1\n40 * BLACKEMPTY5\n41 GREYTOP3\n42 GREYBOTTOM2\n43 GREYTOP5\n44 GREYLEFT5\n45 BLACKTOP4\n46 GREYLEFT2\n47 BLACKLEFT2\n48 * BLACKBOTTOM5\n49 BLACKTOP4\n50 GREYBOTTOM2\n51 GREYTOP4\n52 GREYTOP2\n53 BLACKTOP5\n54 BLACKTOP4\n55 GREYLEFT2\n56 * GREYBOTTOM3\n57 GREYRIGHT2\n58 GREYTOP5\n59 GREYBOTTOM2\n60 BLACKRIGHT4\n61 GREYLEFT2\n62 GREYLEFT3\n63 BLACKLEFT2\n64 * BLACKBOTTOM5\n65 BLACKRIGHT4\n66 GREYLEFT2\n67 GREYLEFT3\n68 BLACKLEFT2\n69 BLACKBOTTOM5\n70 BLACKLEFT2\n71 BLACKBOTTOM5\n72 * GREYTOP5\n73 GREYLEFT2\n74 GREYRIGHT5\n75 BLACKLEFT2\n76 GREYTOP3\n77 GREYEMPTY3\n78 GREYBOTTOM5\n79 GREYLEFT2\n80 * GREYBOTTOM4\n81 GREYBOTTOM2\n82 GREYRIGHT2\n83 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n84 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n85 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n86 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n87 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n--------------------\n0 * BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n1 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n2 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n3 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n4 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n5 BLACKEMPTY3\n6 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n7 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n8 * BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n9 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n10 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n11 BLACKRIGHT3\n12 GREYRIGHT3\n13 BLACKLEFT3\n14 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n15 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n16 * BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n17 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n18 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n19 GREYBOTTOM1\n20 GREYEMPTY2\n21 BLACKEMPTY4\n22 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n23 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n24 * BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n25 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n26 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n27 BLACKBOTTOM1\n28 GREYEMPTY2\n29 BLACKBOTTOM3\n30 BLACKLEFT2\n31 GREYRIGHT4\n32 * GREYEMPTY3\n33 BLACKTOP4\n34 GREYBOTTOM2\n35 GREYLEFT1\n36 GREYEMPTY2\n37 BLACKEMPTY4\n38 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n39 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n40 * BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n41 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n42 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n43 BLACKLEFT1\n44 GREYEMPTY2\n45 BLACKBOTTOM3\n46 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n47 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n48 * BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n49 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n50 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n51 GREYTOP4\n52 GREYTOP2\n53 BLACKTOP5\n54 GREYLEFT2\n55 GREYTOP1\n56 * GREYEMPTY2\n57 BLACKBOTTOM3\n58 GREYEMPTY3\n59 BLACKLEFT5\n60 GREYEMPTY2\n61 BLACKBOTTOM3\n62 GREYEMPTY3\n63 GREYEMPTY5\n--------------------\n0 * BLACKLEFT2\n1 GREYRIGHT4\n2 GREYEMPTY3\n3 BLACKTOP4\n4 GREYBOTTOM2\n5 BLACKEMPTY1\n6 BLACKEMPTY2\n7 BLACKBOTTOM3\n8 * BLACKLEFT2\n9 GREYRIGHT4\n10 GREYEMPTY3\n11 BLACKTOP4\n12 GREYBOTTOM2\n13 GREYRIGHT1\n14 GREYEMPTY2\n15 BLACKEMPTY4\n16 * GREYEMPTY3\n17 BLACKTOP4\n18 GREYEMPTY3\n19 BLACKTOP4\n20 GREYBOTTOM2\n21 BLACKRIGHT1\n22 GREYEMPTY2\n23 BLACKTOP3\n24 * GREYEMPTY3\n25 BLACKBOTTOM4\n26 GREYEMPTY3\n27 BLACKLEFT4\n28 GREYBOTTOM2\n29 BLACKRIGHT1\n30 GREYEMPTY2\n31 BLACKTOP3\n32 * GREYLEFT2\n33 BLACKLEFT4\n34 GREYLEFT2\n35 BLACKBOTTOM4\n36 GREYRIGHT1\n37 GREYEMPTY2\n38 BLACKEMPTY4\n39 GREYEMPTY3\n40 * BLACKLEFT4\n41 GREYEMPTY3\n42 BLACKBOTTOM4\n43 GREYBOTTOM2\n44 GREYRIGHT1\n45 GREYEMPTY2\n46 BLACKEMPTY4\n47 GREYLEFT2\n48 * BLACKBOTTOM4\n49 GREYLEFT2\n50 BLACKLEFT4\n51 GREYBOTTOM4\n52 GREYBOTTOM2\n53 GREYRIGHT2\n54 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n55 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n56 * BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n57 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n58 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n59 GREYEMPTY4\n60 GREYLEFT3\n61 BLACKTOP1\n62 BLACKEMPTY5\n63 GREYTOP3\n64 * GREYEMPTY3\n65 GREYBOTTOM5\n66 GREYLEFT2\n67 BLACKEMPTY3\n68 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n69 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n70 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n71 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n72 * BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n73 BLACKRIGHT3\n74 GREYRIGHT3\n75 BLACKLEFT3\n76 BLACKLEFT2\n77 GREYRIGHT4\n78 GREYEMPTY3\n79 BLACKTOP4\n80 * GREYBOTTOM2\n81 GREYLEFT1\n82 GREYEMPTY2\n83 BLACKEMPTY4\n84 GREYEMPTY3\n85 GREYBOTTOM1\n86 GREYEMPTY2\n87 BLACKTOP4\n88 * GREYEMPTY3\n89 GREYLEFT1\n90 GREYEMPTY2\n91 BLACKBOTTOM3\n92 GREYEMPTY3\n93 BLACKBOTTOM1\n94 GREYEMPTY2\n95 BLACKBOTTOM3\n96 * GREYEMPTY3\n97 GREYLEFT5\n98 GREYLEFT2\n99 GREYRIGHT2\n100 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n101 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n102 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n103 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n104 * BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n105 BLACKLEFT5\n106 GREYEMPTY2\n107 BLACKBOTTOM3\n108 GREYEMPTY3\n109 GREYTOP1\n110 GREYEMPTY2\n111 BLACKBOTTOM3\n112 * GREYEMPTY3\n113 GREYEMPTY1\n114 GREYEMPTY2\n115 BLACKTOP3\n116 GREYEMPTY3\n117 GREYEMPTY5\n118 BLACKEMPTY2\n119 BLACKBOTTOM3\n120 * GREYEMPTY3\n121 BLACKEMPTY1\n122 BLACKEMPTY2\n123 BLACKBOTTOM3\n124 GREYEMPTY3\n125 BLACKLEFT4\n126 GREYEMPTY3\n127 BLACKBOTTOM4\n128 * GREYBOTTOM2\n129 BLACKEMPTY1\n130 BLACKEMPTY2\n131 BLACKBOTTOM3\n132 GREYLEFT2\n133 BLACKLEFT4\n134 GREYLEFT2\n135 BLACKBOTTOM4\n136 * GREYEMPTY5\n137 BLACKEMPTY2\n138 BLACKBOTTOM3\n139 GREYEMPTY3\n140 BLACKLEFT4\n141 GREYEMPTY3\n142 BLACKBOTTOM4\n143 GREYBOTTOM2\n144 * GREYEMPTY5\n145 BLACKEMPTY2\n146 BLACKBOTTOM3\n147 GREYLEFT2\n148 BLACKLEFT4\n149 GREYLEFT2\n150 BLACKBOTTOM4\n151 GREYEMPTY5\n152 * BLACKEMPTY2\n153 BLACKBOTTOM3\n154 BLACKLEFT3\n155 GREYEMPTY5\n--------------------\n0 * BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n1 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n2 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n3 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n4 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n5 GREYEMPTY4\n6 GREYLEFT3\n7 BLACKTOP1\n8 * BLACKEMPTY5\n9 GREYRIGHT4\n10 GREYEMPTY3\n11 GREYBOTTOM5\n12 GREYBOTTOM2\n13 BLACKEMPTY3\n14 BLACKLEFT2\n15 GREYRIGHT4\n16 * GREYEMPTY3\n17 BLACKTOP4\n18 GREYBOTTOM2\n19 BLACKRIGHT3\n20 GREYRIGHT3\n21 BLACKLEFT3\n22 BLACKLEFT2\n23 GREYRIGHT4\n24 * GREYEMPTY3\n25 BLACKTOP4\n26 GREYBOTTOM2\n27 BLACKBOTTOM1\n28 GREYEMPTY2\n29 BLACKBOTTOM3\n30 BLACKLEFT2\n31 GREYRIGHT4\n32 * GREYEMPTY3\n33 BLACKTOP4\n34 GREYBOTTOM2\n35 GREYBOTTOM1\n36 GREYEMPTY2\n37 BLACKEMPTY4\n38 BLACKLEFT2\n39 GREYRIGHT4\n40 * GREYEMPTY3\n41 BLACKTOP4\n42 GREYBOTTOM2\n43 GREYLEFT1\n44 GREYEMPTY2\n45 BLACKEMPTY4\n46 GREYEMPTY3\n47 BLACKBOTTOM4\n48 * GREYEMPTY3\n49 BLACKLEFT4\n50 GREYBOTTOM2\n51 BLACKBOTTOM1\n52 GREYEMPTY2\n53 BLACKBOTTOM3\n54 GREYEMPTY3\n55 BLACKLEFT4\n56 * GREYEMPTY3\n57 BLACKBOTTOM4\n58 GREYBOTTOM2\n59 GREYTOP4\n60 GREYTOP2\n61 BLACKTOP5\n62 GREYRIGHT2\n63 GREYTOP1\n64 * GREYEMPTY2\n65 BLACKBOTTOM3\n66 GREYEMPTY3\n67 BLACKLEFT5\n68 GREYEMPTY2\n69 BLACKBOTTOM3\n70 GREYEMPTY3\n71 GREYEMPTY1\n72 * GREYEMPTY2\n73 BLACKBOTTOM3\n74 GREYEMPTY3\n75 GREYEMPTY1\n76 GREYEMPTY2\n77 BLACKTOP3\n78 GREYLEFT2\n79 BLACKBOTTOM4\n80 * GREYLEFT2\n81 BLACKLEFT4\n82 GREYEMPTY1\n83 GREYEMPTY2\n84 GREYEMPTY1\n85 GREYEMPTY2\n86 GREYEMPTY1\n87 GREYEMPTY2\n88 * GREYEMPTY1\n89 GREYEMPTY2\n90 GREYBOTTOM4\n91 GREYBOTTOM2\n92 GREYRIGHT2\n93 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n94 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n95 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n96 * BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n97 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n98 GREYEMPTY4\n99 GREYLEFT3\n100 BLACKTOP1\n101 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n102 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n103 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n104 * BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n105 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n106 BLACKEMPTY3\n107 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n108 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n109 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n110 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n111 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n112 * BLACKRIGHT3\n113 GREYRIGHT3\n114 BLACKLEFT3\n115 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n116 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n117 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n118 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n119 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n120 * GREYBOTTOM1\n121 GREYEMPTY2\n122 BLACKEMPTY4\n123 GREYLEFT2\n124 GREYTOP3\n125 GREYEMPTY3\n126 GREYBOTTOM5\n127 GREYBOTTOM2\n128 * BLACKBOTTOM1\n129 GREYEMPTY2\n130 BLACKBOTTOM3\n131 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n132 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n133 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n134 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n135 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n136 * GREYLEFT1\n137 GREYEMPTY2\n138 BLACKEMPTY4\n139 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n140 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n141 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n142 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n143 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n144 * BLACKLEFT1\n145 GREYEMPTY2\n146 BLACKBOTTOM3\n147 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n148 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n149 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n150 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n151 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n152 * GREYTOP4\n153 GREYTOP2\n154 BLACKTOP5\n155 GREYLEFT2\n156 GREYTOP1\n157 GREYEMPTY2\n158 BLACKBOTTOM3\n159 GREYEMPTY3\n160 * BLACKLEFT5\n161 GREYEMPTY2\n162 BLACKBOTTOM3\n163 GREYEMPTY3\n164 BLACKLEFT4\n165 GREYEMPTY3\n166 BLACKBOTTOM4\n167 GREYBOTTOM2\n168 * BLACKLEFT5\n169 GREYEMPTY2\n170 BLACKBOTTOM3\n171 BLACKLEFT2\n172 GREYTOP3\n173 BLACKRIGHT4\n174 GREYLEFT2\n175 GREYLEFT5\n176 * BLACKTOP4\n177 GREYLEFT2\n178 BLACKLEFT2\n179 GREYTOP3\n180 BLACKRIGHT4\n181 GREYLEFT2\n182 GREYTOP1\n183 GREYEMPTY2\n184 * BLACKBOTTOM3\n--------------------\n0 * BLACKTOP2\n1 BLACKLEFT3\n2 BLACKLEFT2\n3 GREYRIGHT4\n4 BLACKLEFT2\n5 GREYRIGHT4\n6 GREYTOP1\n7 GREYEMPTY2\n8 * BLACKBOTTOM3\n9 GREYLEFT2\n10 BLACKLEFT4\n11 GREYLEFT2\n12 BLACKBOTTOM4\n13 GREYBOTTOM4\n14 GREYBOTTOM2\n15 GREYRIGHT2\n16 * BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n17 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n18 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n19 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n20 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n21 GREYEMPTY4\n22 GREYLEFT3\n23 BLACKTOP1\n24 * BLACKEMPTY5\n25 GREYRIGHT4\n26 GREYEMPTY3\n27 BLACKBOTTOM4\n28 GREYBOTTOM2\n29 BLACKEMPTY3\n30 BLACKLEFT2\n31 GREYRIGHT4\n32 * GREYEMPTY3\n33 BLACKBOTTOM4\n34 GREYBOTTOM2\n35 BLACKRIGHT3\n36 GREYRIGHT3\n37 BLACKLEFT3\n38 BLACKLEFT2\n39 GREYRIGHT4\n40 * GREYEMPTY3\n41 BLACKBOTTOM4\n42 GREYBOTTOM2\n43 GREYBOTTOM1\n44 GREYEMPTY2\n45 BLACKEMPTY4\n46 GREYEMPTY3\n47 BLACKBOTTOM1\n48 * GREYEMPTY2\n49 BLACKBOTTOM3\n50 GREYEMPTY3\n51 GREYLEFT1\n52 GREYEMPTY2\n53 BLACKEMPTY4\n54 GREYEMPTY3\n55 BLACKLEFT1\n56 * GREYEMPTY2\n57 BLACKBOTTOM3\n58 GREYEMPTY3\n59 GREYLEFT1\n60 GREYEMPTY2\n61 BLACKEMPTY4\n62 BLACKLEFT2\n63 GREYRIGHT4\n64 * GREYLEFT2\n65 BLACKBOTTOM4\n66 BLACKLEFT1\n67 GREYEMPTY2\n68 BLACKBOTTOM3\n69 BLACKLEFT2\n70 GREYRIGHT4\n71 GREYLEFT2\n72 * BLACKBOTTOM4\n73 GREYBOTTOM4\n74 GREYBOTTOM2\n75 GREYRIGHT2\n76 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n77 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n78 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n79 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n80 * BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n81 GREYEMPTY4\n82 GREYLEFT3\n83 BLACKTOP1\n84 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n85 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n86 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n87 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n88 * BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n89 BLACKEMPTY3\n90 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n91 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n92 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n93 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n94 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n95 BLACKRIGHT3\n96 * GREYRIGHT3\n97 BLACKLEFT3\n98 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n99 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n100 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n101 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n102 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n103 GREYBOTTOM1\n104 * GREYEMPTY2\n105 BLACKEMPTY4\n106 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n107 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n108 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n109 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n110 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n111 BLACKBOTTOM1\n112 * GREYEMPTY2\n113 BLACKBOTTOM3\n114 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n115 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n116 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n117 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n118 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n--------------------\n0 * BLACKLEFT2\n1 GREYTOP3\n2 GREYLEFT2\n3 BLACKBOTTOM4\n4 BLACKBOTTOM1\n5 GREYEMPTY2\n6 BLACKBOTTOM3\n7 BLACKLEFT2\n8 * GREYTOP3\n9 GREYLEFT2\n10 BLACKBOTTOM4\n11 GREYBOTTOM4\n12 GREYBOTTOM2\n13 GREYRIGHT2\n14 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n15 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n16 * BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n17 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n18 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n19 GREYEMPTY5\n20 GREYLEFT3\n21 BLACKTOP1\n22 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n23 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n24 * BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n25 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n26 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n27 BLACKEMPTY3\n28 BLACKLEFT2,GREYTOP3,GREYEMPTY3,BLACKBOTTOM4,GREYBOTTOM2 = B\n29 BLACKLEFT2,GREYTOP3,GREYEMPTY3,BLACKBOTTOM4,GREYBOTTOM2 = B\n30 BLACKLEFT2,GREYTOP3,GREYEMPTY3,BLACKBOTTOM4,GREYBOTTOM2 = B\n31 BLACKLEFT2,GREYTOP3,GREYEMPTY3,BLACKBOTTOM4,GREYBOTTOM2 = B\n32 * BLACKLEFT2,GREYTOP3,GREYEMPTY3,BLACKBOTTOM4,GREYBOTTOM2 = B\n33 BLACKRIGHT3\n34 GREYRIGHT3\n35 BLACKLEFT3\n36 BLACKLEFT2\n37 GREYTOP3\n38 GREYEMPTY3\n39 BLACKBOTTOM4\n40 * BLACKBOTTOM2\n41 BLACKRIGHT3\n42 GREYRIGHT3\n43 BLACKLEFT3\n44 BLACKLEFT2\n45 GREYTOP3\n46 GREYLEFT2\n47 BLACKBOTTOM4\n48 * GREYEMPTY4\n49 GREYLEFT3\n50 BLACKTOP1\n51 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n52 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n53 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n54 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n55 BLACKEMPTY5,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = C\n56 * GREYBOTTOM4\n57 GREYBOTTOM2\n58 GREYRIGHT2\n59 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n60 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n61 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n62 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n63 BLACKLEFT2,GREYTOP3,GREYEMPTY3,GREYBOTTOM5,GREYBOTTOM2 = PIECE\n64 * GREYEMPTY4\n65 GREYLEFT3\n66 BLACKTOP1\n67 BLACKEMPTY5\n68 GREYTOP3\n69 GREYLEFT2\n70 GREYBOTTOM5\n71 GREYBOTTOM4\n72 * GREYBOTTOM2\n73 GREYRIGHT2\n74 BLACKBOTTOM2\n75 GREYTOP3\n76 GREYEMPTY3\n77 GREYBOTTOM5\n78 GREYBOTTOM2\n79 GREYBOTTOM4\n80 * GREYBOTTOM2\n81 GREYRIGHT2\n82 BLACKBOTTOM2\n83 GREYBOTTOM4\n84 GREYBOTTOM2\n85 GREYRIGHT2\n86 BLACKBOTTOM2\n87 GREYBOTTOM4\n" ], [ "\"\"\"\n1BOTTOMBLACK\t\n1BOTTOMGrey\t\n1EMPTYBLACK\t\n1EMPTYGrey\t\n1LEFTBLACK\t\n1LEFTGrey\t\n1RIGHTBLACK\t\n1RIGHTGrey\t\n1TOPBLACK\tEGGS\n1TOPGrey\t\n2BOTTOMBLACK\tTHAT\n2BOTTOMGrey\tI\n2EMPTYBLACK\t\n2EMPTYGrey\t\n2LEFTBLACK\t\n2LEFTGrey\tYOU\n2RIGHTBLACK\tTHANK\n2RIGHTGrey\tSAM\n2TOPBLACK\t\n2TOPGrey\t\n3BOTTOMBLACK\t\n3BOTTOMGrey\t\n3EMPTYBLACK\t\n3EMPTYGrey\t\n3LEFTBLACK\t\n3LEFTGrey\tAND\n3RIGHTBLACK\t\n3RIGHTGrey\t\n3TOPBLACK\t\n3TOPGrey\t\n4BOTTOMBLACK\t\n4BOTTOMGrey\tAM\n4EMPTYBLACK\t\n4EMPTYGrey\tHAM\n4LEFTBLACK\t\n4LEFTGrey\t\n4RIGHTBLACK\t\n4RIGHTGrey\t\n4TOPBLACK\t\n4TOPGrey\t\n5BOTTOMBLACK\t\n5BOTTOMGrey\t\n5EMPTYBLACK\t\n5EMPTYGrey\t\n5LEFTBLACK\t\n5LEFTGrey\t\n5RIGHTBLACK\t\n5RIGHTGrey\t\n5TOPBLACK\t\n5TOPGrey\t\n\"\"\"\n\nWORDS = {\n 'BLACKTOP1': 'EGGS',\n 'GREYLEFT3': 'AND',\n 'GREYEMPTY4': 'HAM',\n 'GREYEMPTY2': 'A',\n 'GREYEMPTY1': 'TRAIN',\n 'BLACKLEFT4': 'COULD',\n 'GREYLEFT2': 'YOU',\n 'BLACKBOTTOM4': 'WOULD',\n 'BLACKTOP3': 'ON',\n 'GREYEMPTY3': 'NOT',\n 'BLACKBOTTOM3': 'IN',\n 'BLACKLEFT5': 'TREE',\n 'GREYTOP1': 'CAR',\n 'GREYRIGHT2': 'SAM',\n 'BLACKTOP5': 'LET',\n 'GREYTOP2': 'ME',\n 'GREYTOP4': 'BE',\n 'GREYBOTTOM2': 'I',\n 'BLACKBOTTOM1': 'BOX',\n 'BLACKEMPTY4': 'IN',\n 'GREYLEFT1': 'FOX',\n 'BLACKTOP4': 'WILL',\n 'GREYRIGHT4': 'EAT',\n 'BLACKLEFT2': 'THEM',\n 'GREYBOTTOM1': 'HOUSE',\n 'BLACKLEFT3': 'HERE',\n 'GREYRIGHT3': 'OR',\n 'BLACKRIGHT3': 'THERE',\n 'BLACKEMPTY3': ''\n}", "_____no_output_____" ], [ "for chunk in reversed(parsed):\n for i, piece in enumerate(reversed(chunk)):\n word = ''.join(map(str, piece))\n if word in WORDS:\n print(WORDS[word])\n else:\n print(word)\n", "I\nGREYBOTTOM4\nSAM\nSAM\nI\nGREYBOTTOM4\nBLACKBOTTOM2\nSAM\nI\nGREYBOTTOM4\nBLACKBOTTOM2\nSAM\nI\nGREYBOTTOM4\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nBLACKBOTTOM2\nSAM\nI\nGREYBOTTOM4\nGREYBOTTOM5\nYOU\nGREYTOP3\nBLACKEMPTY5\nEGGS\nAND\nHAM\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nSAM\nI\nGREYBOTTOM4\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nBLACKEMPTY5\nEGGS\nAND\nHAM\nWOULD\nYOU\nGREYTOP3\nTHEM\nHERE\nOR\nTHERE\nBLACKBOTTOM2\nWOULD\nNOT\nGREYTOP3\nTHEM\nHERE\nOR\nTHERE\nI\nWOULD\nNOT\nGREYTOP3\nTHEM\nBLACKEMPTY3\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nBLACKEMPTY5\nEGGS\nAND\nGREYEMPTY5\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nSAM\nI\nGREYBOTTOM4\nWOULD\nYOU\nGREYTOP3\nTHEM\nIN\nA\nBOX\nWOULD\nYOU\nGREYTOP3\nTHEM\nA\nHOUSE\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nIN\nA\nBOX\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nIN\nA\nHOUSE\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nHERE\nOR\nTHERE\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nBLACKEMPTY3\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nBLACKEMPTY5\nEGGS\nAND\nHAM\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nSAM\nI\nGREYBOTTOM4\nWOULD\nYOU\nEAT\nTHEM\nIN\nA\nBLACKLEFT1\nWOULD\nYOU\nEAT\nTHEM\nIN\nA\nFOX\nNOT\nIN\nA\nBLACKLEFT1\nNOT\nIN\nA\nFOX\nNOT\nIN\nA\nBOX\nNOT\nIN\nA\nHOUSE\nI\nWOULD\nNOT\nEAT\nTHEM\nHERE\nOR\nTHERE\nI\nWOULD\nNOT\nEAT\nTHEM\nBLACKEMPTY3\nI\nWOULD\nNOT\nEAT\nBLACKEMPTY5\nEGGS\nAND\nHAM\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nSAM\nI\nGREYBOTTOM4\nWOULD\nYOU\nCOULD\nYOU\nIN\nA\nCAR\nEAT\nTHEM\nEAT\nTHEM\nHERE\nBLACKTOP2\nI\nWOULD\nNOT\nCOULD\nNOT\nIN\nA\nCAR\nYOU\nBLACKRIGHT4\nGREYTOP3\nTHEM\nYOU\nWILL\nGREYLEFT5\nYOU\nBLACKRIGHT4\nGREYTOP3\nTHEM\nIN\nA\nTREE\nI\nWOULD\nNOT\nCOULD\nNOT\nIN\nA\nTREE\nNOT\nIN\nA\nCAR\nYOU\nLET\nME\nBE\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nIN\nA\nBLACKLEFT1\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nIN\nA\nFOX\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nIN\nA\nBOX\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nYOU\nIN\nA\nHOUSE\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nHERE\nOR\nTHERE\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nBLACKEMPTY3\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nBLACKEMPTY5\nEGGS\nAND\nHAM\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nSAM\nI\nGREYBOTTOM4\nA\nTRAIN\nA\nTRAIN\nA\nTRAIN\nA\nTRAIN\nCOULD\nYOU\nWOULD\nYOU\nON\nA\nTRAIN\nNOT\nIN\nA\nTRAIN\nNOT\nIN\nA\nTREE\nNOT\nIN\nA\nCAR\nSAM\nLET\nME\nBE\nI\nWOULD\nNOT\nCOULD\nNOT\nIN\nA\nBOX\nI\nCOULD\nNOT\nWOULD\nNOT\nIN\nA\nFOX\nI\nWILL\nNOT\nEAT\nTHEM\nIN\nA\nHOUSE\nI\nWILL\nNOT\nEAT\nTHEM\nIN\nA\nBOX\nI\nWILL\nNOT\nEAT\nTHEM\nHERE\nOR\nTHERE\nI\nWILL\nNOT\nEAT\nTHEM\nBLACKEMPTY3\nI\nGREYBOTTOM5\nNOT\nEAT\nBLACKEMPTY5\nEGGS\nAND\nHAM\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nI\nGREYBOTTOM4\nGREYTOP5\nIN\nBLACKEMPTY2\nGREYEMPTY5\nHERE\nIN\nBLACKEMPTY2\nGREYEMPTY5\nWOULD\nYOU\nCOULD\nYOU\nIN\nBLACKEMPTY2\nGREYEMPTY5\nI\nWOULD\nNOT\nCOULD\nNOT\nIN\nBLACKEMPTY2\nGREYEMPTY5\nWOULD\nYOU\nCOULD\nYOU\nIN\nBLACKEMPTY2\nBLACKEMPTY1\nI\nWOULD\nNOT\nCOULD\nNOT\nIN\nBLACKEMPTY2\nBLACKEMPTY1\nNOT\nIN\nBLACKEMPTY2\nGREYEMPTY5\nNOT\nON\nA\nTRAIN\nNOT\nIN\nA\nCAR\nNOT\nIN\nA\nTREE\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nSAM\nYOU\nGREYLEFT5\nNOT\nIN\nA\nBOX\nNOT\nIN\nA\nFOX\nNOT\nWILL\nA\nHOUSE\nNOT\nIN\nA\nFOX\nI\nWILL\nNOT\nEAT\nTHEM\nHERE\nOR\nTHERE\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nBLACKEMPTY3\nYOU\nGREYBOTTOM5\nNOT\nGREYTOP3\nBLACKEMPTY5\nEGGS\nAND\nHAM\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nSAM\nI\nGREYBOTTOM4\nCOULD\nYOU\nWOULD\nYOU\nIN\nA\nGREYRIGHT1\nI\nWOULD\nNOT\nCOULD\nNOT\nIN\nA\nGREYRIGHT1\nWOULD\nYOU\nCOULD\nYOU\nON\nA\nBLACKRIGHT1\nI\nCOULD\nNOT\nWOULD\nNOT\nON\nA\nBLACKRIGHT1\nI\nWILL\nNOT\nWILL\nNOT\nIN\nA\nGREYRIGHT1\nI\nWILL\nNOT\nEAT\nTHEM\nIN\nBLACKEMPTY2\nBLACKEMPTY1\nI\nWILL\nNOT\nEAT\nTHEM\nA\nTRAIN\nNOT\nIN\nBLACKEMPTY2\nGREYEMPTY5\nNOT\nIN\nA\nTREE\nNOT\nIN\nA\nCAR\nYOU\nLET\nME\nBE\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nIN\nA\nBLACKLEFT1\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nIN\nA\nFOX\nI\nWILL\nNOT\nEAT\nTHEM\nIN\nA\nBOX\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nIN\nA\nHOUSE\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nHERE\nOR\nTHERE\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nBLACKEMPTY3\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nBLACKEMPTY5\nAND\nHAM\nI\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nSAM\nI\nGREYBOTTOM4\nYOU\nGREYBOTTOM5\nNOT\nGREYTOP3\nTHEM\nGREYRIGHT5\nYOU\nGREYTOP5\nBLACKBOTTOM5\nTHEM\nBLACKBOTTOM5\nTHEM\nAND\nYOU\nBLACKRIGHT4\nBLACKBOTTOM5\nTHEM\nAND\nYOU\nBLACKRIGHT4\nI\nGREYTOP5\nSAM\nGREYBOTTOM3\nYOU\nWILL\nLET\nME\nBE\nI\nWILL\nBLACKBOTTOM5\nTHEM\nYOU\nWILL\nGREYLEFT5\nGREYTOP5\nI\nGREYTOP3\nBLACKEMPTY5\nEGGS\nAND\nHAM\nI\nGREYBOTTOM5\nI\nGREYTOP3\nTHEM\nSAM\nI\nGREYBOTTOM4\nAND\nI\nWOULD\nEAT\nTHEM\nIN\nA\nBLACKRIGHT1\nAND\nI\nWOULD\nEAT\nTHEM\nIN\nA\nGREYRIGHT1\nAND\nI\nWILL\nEAT\nTHEM\nIN\nBLACKEMPTY2\nBLACKEMPTY1\nAND\nIN\nBLACKEMPTY2\nGREYEMPTY5\nAND\nA\nTRAIN\nAND\nIN\nA\nCAR\nAND\nIN\nA\nBLACKRIGHT5\nBLACKTOP2\nGREYLEFT4\nGREYRIGHT5\nBLACKRIGHT5\nGREYRIGHT5\nBLACKRIGHT5\nYOU\nGREYLEFT5\nGREYRIGHT5\nI\nWILL\nEAT\nTHEM\nIN\nA\nBLACKLEFT1\nAND\nI\nWILL\nEAT\nTHEM\nIN\nA\nFOX\nAND\nI\nWILL\nEAT\nTHEM\nIN\nA\nBOX\nAND\nI\nWILL\nEAT\nTHEM\nIN\nA\nHOUSE\nAND\nI\nWILL\nEAT\nTHEM\nHERE\nAND\nTHERE\nGREYTOP5\nI\nWILL\nEAT\nTHEM\nBLACKEMPTY3\nI\nGREYBOTTOM5\nGREYRIGHT5\nGREYTOP3\nBLACKEMPTY5\nEGGS\nAND\nHAM\nBLACKRIGHT2\nYOU\nBLACKRIGHT2\nYOU\nSAM\nI\nGREYBOTTOM4\n" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Demonstration of integrating POI Points to OSM road network\n1. Use anyway you like to get the sample [POI data](https://assets.onemap.sg/shp/supermarkets.zip) consisting of supermarkets from [OneMap SG](https://www.onemap.sg/).\n2. Use [OSMnx](https://osmnx.readthedocs.io/en/stable/index.html) to download the pedestrian network from [OpenStreetMap](https://openstreetmap.org), we use a bounding box of Toa Payoh for the demo.\n3. Save the network as `.shp` and read in as two `GeoDataFrame`s: junction as `nodes` and road segment as `edges`.\n4. Integrate POIs into the network using the `connect_poi` function.", "_____no_output_____" ] ], [ [ "import os\nimport wget\nimport osmnx as ox\nimport geopandas as gpd\nfrom toolbox import connect_poi", "_____no_output_____" ] ], [ [ "## 1. Prepare POIs", "_____no_output_____" ] ], [ [ "# get POI data\nurl = \"https://assets.onemap.sg/shp/supermarkets.zip\"\nPATH = 'data/supermarkets.zip'\n\nif os.path.exists(PATH):\n print('File existed.')\nelse:\n PATH = wget.download(url, PATH)\n print('File downloaded.')", "File downloaded.\n" ], [ "# load and subset the POI based on a bounding box\nbbox = (103.8427, 1.3308, 103.8601, 1.3416) # set bbox of Toa Payoh\npois = gpd.read_file('supermarkets', vfs='zip://{}'.format(PATH), crs='epsg:3857')\npois = pois.to_crs(epsg=4326)\npois['lon'] = pois['geometry'].apply(lambda p: p.x)\npois['lat'] = pois['geometry'].apply(lambda p: p.y)\npois = pois[(pois['lon'] >= bbox[0]) & (pois['lon'] <= bbox[2]) &\n (pois['lat'] >= bbox[1]) & (pois['lat'] <= bbox[3])]\npois['key'] = pois.index # set a primary key column\npois.head(3)", "/Users/Yuwen/anaconda/lib/python3.6/site-packages/fiona/env.py:398: FionaDeprecationWarning: The vfs keyword argument is deprecated. Instead, pass a URL that uses a zip or tar (for example) scheme.\n return f(*args, **kwargs)\n" ] ], [ [ "[NOTE] For use in pandana, you may want to ensure the key column for the input is numeric-only to avoid processing errors. Preferably use unique integers (int or str) only, and be aware not to intersect with the node key, which is 'osmid' if you use OSM data, in the nodes gdf.", "_____no_output_____" ], [ "## 2. Prepare network", "_____no_output_____" ] ], [ [ "# get road network and save as .shp\nG = ox.graph_from_bbox(bbox[3], bbox[1], bbox[2], bbox[0], network_type='walk')\nox.save_graph_shapefile(G, filepath='data/sample/', encoding='utf-8')", "_____no_output_____" ], [ "# load as GeoDataFrame\nnodes = gpd.read_file('data/sample/nodes.shp')\nedges = gpd.read_file('data/sample/edges.shp')", "_____no_output_____" ] ], [ [ "## 3. Integrate POIs and network", "_____no_output_____" ] ], [ [ "connect_poi?", "_____no_output_____" ], [ "# it's a one-liner, but is still at beta at the moment\nnew_nodes, new_edges = connect_poi(pois, nodes, edges, key_col='key', path=None)", "Building rtree...\nUpdating external nodes...\nProjecting POIs to the network...\nUpdating internal nodes...\nUpdating internal edges...\nUpdating external links...\nRemove faulty projections: 0/7 (0.00%)\nNOTE: duplication in node coordinates keys\nNodes count: 845\nNode coordinates key count: 843\nMissing 'from' nodes: 0\nMissing 'to' nodes: 0\n" ] ], [ [ "## 4. Check output\n1. First is an example of how edges will be broken into segments when there is a POI to be linked onto it. This process accommodates for multiple POIs. E.g., for 2 POIs projecting onto a same edge (but not overlapping nor on either vertices of the edge), the edge will be replaced with 3 segments.\n2. Then a figure illustrating how the new network looks like after the update.\n\n[NOTE] Be noted that the aggregated length of segments will not equal exactly to the length of the original edge for some reasons that are not handled at the moment.", "_____no_output_____" ] ], [ [ "# original edge\nedges[edges['from'] == 3370311549][['from', 'to', 'length']]", "_____no_output_____" ], [ "# new edges replacing the original (953, 954) and connecting the poi (964)\nnew_edges[(new_edges['from'] == 3370311549) |\n (new_edges['from'] == 9990000005) |\n (new_edges['to'] == 9990000005) ][['from', 'to', 'length']]", "_____no_output_____" ], [ "# output\npoi_links = new_edges[new_edges['highway'] == 'projected_footway']\nax = edges.plot(linewidth=0.8, figsize=(18,10), label='Original Road Edges')\npoi_links.plot(color='indianred', linewidth=2, ax=ax, label='New Connection Edges')\npois.plot(color='indianred', marker='.', markersize=200, ax=ax, label='POI')\nax.legend(loc=2, fontsize=18)\nax.set_title('The integrated network of supermarkets and road network at Toa Payoh', fontsize=22);", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Pincipal component analysis", "_____no_output_____" ] ], [ [ "import numpy as np\nimport matplotlib.pyplot as plt", "_____no_output_____" ] ], [ [ "this is reducing data to a lower dimensional space\n\n$q^tq = 1$\n\n$q^Tx_i =$ distance of datapoint after projecting it on line of vector q\n\n$(q^Tx_i)q =$ stretches vector to the datapoint\n- this is the prediction of the data\n- using the vector q which is constant for all data points\n- and the value which stretches the vector to as close as possible to the real data point \n\nWe optimize q using minimize squared error\n\n$q = \\arg \\min_q \\sum_{i=1}^n||x_i - qq^Tx_i||^2$\n\n$= \\arg \\min_q \\sum_{i=1}^n x_i^Tx_i - q^T\\underbrace{(\\sum_{i=1}^nx_ix_i^T)}_{XX^T}q$\n\n$q =\\arg \\max_q q^T(XX^T)q$\n\nq = 1st eigenvector of $XX^T$\n\n$\\lambda = q^T(XX^T)q$ 1st eigenvalue", "_____no_output_____" ] ], [ [ "x = np.linspace(0, 5, 201)\n\ndef f(x):\n return np.exp(-x)*\n\ny = f(x)\n\nfig, ax = plt.subplots(1, figsize=(8,4))\n\nax.plot(x,y)\n\nax.grid(True)\nplt.show()\n", "/usr/local/lib/python2.7/dist-packages/ipykernel_launcher.py:4: RuntimeWarning: divide by zero encountered in log\n after removing the cwd from sys.path.\n" ] ] ]

[ "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "BSD-2-Clause" ]

[ "BSD-2-Clause" ]

[ "BSD-2-Clause" ]

[ [ [ "import numpy as np\nimport pandas as pd\nimport scipy\nimport scipy.stats\nimport verdict", "_____no_output_____" ], [ "import matplotlib.pyplot as plt", "_____no_output_____" ] ], [ [ "# Parameters", "_____no_output_____" ] ], [ [ "sample_number = 0", "_____no_output_____" ] ], [ [ "# Load Mock Data", "_____no_output_____" ] ], [ [ "# Load using my preferred wrapper for hdf5 files.\ndata = verdict.Dict.from_hdf5( './data/synthetic_data/sample{}/observers_file.h5'.format( sample_number ) )", "_____no_output_____" ] ], [ [ "# Assign Errors\n\nOur goal is to create semi-realistic errors.\nWe won't put too much effort into this, so we'll roughly fit a gamma function to the distribution of errors from COS-Halos data, and sample from the distribution.\nWe do provide a bound on 1.5 the maximum observed error, however, to prevent getting anything too crazy.", "_____no_output_____" ], [ "## Derive Errors from Data", "_____no_output_____" ] ], [ [ "df = pd.read_fwf(\n './data/cos_halos_data.txt',\n skiprows = 31,\n names = [ 'ID', 'z', 'Ion', 'Vel', 'e_Vel', 'b', 'e_b', 'logN', 'elogN', 'l_logNA', ',logNA', 'e_logNA' ],\n)", "_____no_output_____" ], [ "# Find ions used\nions = list( set( df['Ion'] ) )\nions.remove( np.nan )\nions.append( 'H I' )", "_____no_output_____" ], [ "# Get out errors from data, mildly processed\nelogNs = {}\nfor ion in ions:\n \n if ion != 'H I':\n # Locate\n ion_df = df[df['Ion'] == ion]\n elogN = []\n for e in ion_df['elogN'].values:\n try:\n elogN.append( float( e ) )\n except ValueError:\n continue\n else:\n # Manual H I input from COS-Halos hydrogen paper\n elogN = [ 0.03, 0.05, 0.05, 0.03, 0.09, 0.03, 0.15, 0.15, 0.08, 0.06, 0.03, 0.06, 0.03, 0.04, 0.03, 0.13, 0.06, 0.04, 0.05, 0.06, 0.02, 0.04, 0.06, 0.04, 0.04, 0.03, 0.07, 0.06, 0.10, 0.12, 0.06, 0.02, 0.03, 0.09, 0.04, 0.03, 0.04, ]\n\n elogNs[ion] = np.array( elogN )", "_____no_output_____" ], [ "# Stats of the errors\n\nerrs = {}\nfor ion in ions:\n \n elogN = elogNs[ion]\n\n # Extract\n errs[ion] = {\n 'mean': np.mean( elogN ),\n 'std': np.std( elogN ),\n }\n \n \n # Create a gamma fn, *very* roughly fitted\n beta = errs[ion]['mean'] / errs[ion]['std']\n alpha = errs[ion]['mean'] * beta\n dist = scipy.stats.gamma( a=alpha, scale=1/beta )\n \n # Plot and store\n fig = plt.figure()\n ax = plt.gca()\n ax.hist(\n elogN,\n bins = np.linspace( 0, elogN.max(), 32 )\n )\n ax.plot(\n np.linspace( 0, elogN.max(), 128 ),\n dist.pdf( np.linspace( 0, elogN.max(), 128 ) ),\n )\n ax.annotate(\n s = ion,\n xy = ( 1, 1 ),\n xytext = ( -5, -5 ),\n va = 'top',\n ha = 'right',\n xycoords = 'axes fraction',\n textcoords = 'offset points',\n fontsize = 20,\n )\n errs[ion]['dist'] = dist", "_____no_output_____" ] ], [ [ "## Apply to Data", "_____no_output_____" ] ], [ [ "data_out = {}\nfor ion in data.keys():\n \n dist = errs[ion]['dist']\n \n ion_errs = []\n modified_columns = []\n for i, column in enumerate( data[ion] ):\n \n ion_err = np.inf\n while ( ion_err > elogNs[ion].max()*1.5 ) or ( ion_err < elogNs[ion].min() / 1.5 ):\n ion_err = dist.rvs()\n \n # Assume error is conservative, apply\n modified_column = column * 10.**np.random.uniform( -ion_err, ion_err )\n \n # Round and store\n ion_errs.append( np.round( ion_err, decimals=3 ) )\n modified_columns.append( np.round( np.log10( modified_column ), decimals=3 ) )\n \n data_out[ion] = {\n 'logN': np.array( modified_columns ),\n 'elogN': np.array( ion_errs ),\n }\ndata_out = verdict.Dict( data_out )", "<ipython-input-21-1a3bbd98ad63>:19: RuntimeWarning: divide by zero encountered in log10\n modified_columns.append( np.round( np.log10( modified_column ), decimals=3 ) )\n" ] ], [ [ "# Save Output", "_____no_output_____" ] ], [ [ "data_out.to_hdf5( './data/sample{}.h5'.format( sample_number ) )", "_____no_output_____" ] ], [ [ "# Reload for Checking", "_____no_output_____" ] ], [ [ "data_out = verdict.Dict.from_hdf5( './data/synthetic_data_samples/sample{}.h5'.format( sample_number ) )", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code", "code", "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "from google.colab import drive\ndrive.mount('/content/drive')", "Mounted at /content/drive\n" ], [ "# from google.colab import drive\n# drive.mount('/content/drive')\n!pwd", "/content\n" ], [ "\npath = '/content/drive/MyDrive/Research/AAAI/cifar_new/k_001/sixth_run1_'", "_____no_output_____" ], [ "import torch.nn as nn\nimport torch.nn.functional as F\nimport pandas as pd\nimport numpy as np\nimport matplotlib.pyplot as plt\nimport torch\nimport torchvision\nimport torchvision.transforms as transforms\nfrom torch.utils.data import Dataset, DataLoader\nfrom torchvision import transforms, utils\nfrom matplotlib import pyplot as plt\nimport copy\n\n# Ignore warnings\nimport warnings\nwarnings.filterwarnings(\"ignore\")\nn_seed = 0\nk = 0.001", "_____no_output_____" ], [ "torch.backends.cudnn.deterministic = True\ntorch.backends.cudnn.benchmark= False", "_____no_output_____" ], [ "transform = transforms.Compose(\n [transforms.ToTensor(),\n transforms.Normalize((0.5, 0.5, 0.5), (0.5, 0.5, 0.5))])\n\ntrainset = torchvision.datasets.CIFAR10(root='./data', train=True, download=True, transform=transform)\n\n\ntestset = torchvision.datasets.CIFAR10(root='./data', train=False, download=True, transform=transform)\n", "Downloading https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz to ./data/cifar-10-python.tar.gz\n" ], [ "trainloader = torch.utils.data.DataLoader(trainset, batch_size=10, shuffle=False)\ntestloader = torch.utils.data.DataLoader(testset, batch_size=10, shuffle=False)\n\n\nclasses = ('plane', 'car', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck')\n\nforeground_classes = {'plane', 'car', 'bird'}\n#foreground_classes = {'bird', 'cat', 'deer'}\nbackground_classes = {'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck'}\n#background_classes = {'plane', 'car', 'dog', 'frog', 'horse','ship', 'truck'}\n\nfg1,fg2,fg3 = 0,1,2", "_____no_output_____" ], [ "dataiter = iter(trainloader)\nbackground_data=[]\nbackground_label=[]\nforeground_data=[]\nforeground_label=[]\nbatch_size=10\n\nfor i in range(5000):\n images, labels = dataiter.next()\n for j in range(batch_size):\n if(classes[labels[j]] in background_classes):\n img = images[j].tolist()\n background_data.append(img)\n background_label.append(labels[j])\n else:\n img = images[j].tolist()\n foreground_data.append(img)\n foreground_label.append(labels[j])\n \nforeground_data = torch.tensor(foreground_data)\nforeground_label = torch.tensor(foreground_label)\nbackground_data = torch.tensor(background_data)\nbackground_label = torch.tensor(background_label)", "_____no_output_____" ], [ "def create_mosaic_img(bg_idx,fg_idx,fg): \n \"\"\"\n bg_idx : list of indexes of background_data[] to be used as background images in mosaic\n fg_idx : index of image to be used as foreground image from foreground data\n fg : at what position/index foreground image has to be stored out of 0-8\n \"\"\"\n image_list=[]\n j=0\n for i in range(9):\n if i != fg:\n image_list.append(background_data[bg_idx[j]])#.type(\"torch.DoubleTensor\"))\n j+=1\n else: \n image_list.append(foreground_data[fg_idx])#.type(\"torch.DoubleTensor\"))\n label = foreground_label[fg_idx]- fg1 # minus fg1 because our fore ground classes are fg1,fg2,fg3 but we have to store it as 0,1,2\n #image_list = np.concatenate(image_list ,axis=0)\n image_list = torch.stack(image_list) \n return image_list,label", "_____no_output_____" ], [ "desired_num = 30000\nmosaic_list_of_images =[] # list of mosaic images, each mosaic image is saved as list of 9 images\nfore_idx =[] # list of indexes at which foreground image is present in a mosaic image i.e from 0 to 9 \nmosaic_label=[] # label of mosaic image = foreground class present in that mosaic\nfor i in range(desired_num):\n np.random.seed(i)\n bg_idx = np.random.randint(0,35000,8)\n fg_idx = np.random.randint(0,15000)\n fg = np.random.randint(0,9)\n fore_idx.append(fg)\n image_list,label = create_mosaic_img(bg_idx,fg_idx,fg)\n mosaic_list_of_images.append(image_list)\n mosaic_label.append(label)", "_____no_output_____" ], [ "plt.imshow(torch.transpose(mosaic_list_of_images[0][1],dim0= 0,dim1 = 2))", "Clipping input data to the valid range for imshow with RGB data ([0..1] for floats or [0..255] for integers).\n" ], [ "class MosaicDataset(Dataset):\n \"\"\"MosaicDataset dataset.\"\"\"\n\n def __init__(self, mosaic_list_of_images, mosaic_label, fore_idx):\n \"\"\"\n Args:\n csv_file (string): Path to the csv file with annotations.\n root_dir (string): Directory with all the images.\n transform (callable, optional): Optional transform to be applied\n on a sample.\n \"\"\"\n self.mosaic = mosaic_list_of_images\n self.label = mosaic_label\n self.fore_idx = fore_idx\n\n def __len__(self):\n return len(self.label)\n\n def __getitem__(self, idx):\n return self.mosaic[idx] , self.label[idx], self.fore_idx[idx]\n\nbatch = 250\nmsd = MosaicDataset(mosaic_list_of_images, mosaic_label , fore_idx)\ntrain_loader = DataLoader( msd,batch_size= batch ,shuffle=True)", "_____no_output_____" ], [ "class Focus(nn.Module):\n def __init__(self):\n super(Focus, self).__init__()\n self.conv1 = nn.Conv2d(in_channels=3, out_channels=32, kernel_size=3, padding=0,bias=False)\n self.conv2 = nn.Conv2d(in_channels=32, out_channels=64, kernel_size=3, padding=0,bias=False)\n self.conv3 = nn.Conv2d(in_channels=64, out_channels=64, kernel_size=3, padding=0,bias=False)\n self.conv4 = nn.Conv2d(in_channels=64, out_channels=128, kernel_size=3, padding=0,bias=False)\n self.conv5 = nn.Conv2d(in_channels=128, out_channels=256, kernel_size=3, padding=0,bias=False)\n self.conv6 = nn.Conv2d(in_channels=256, out_channels=256, kernel_size=3, padding=1,bias=False)\n self.pool = nn.MaxPool2d(kernel_size=2, stride=2)\n self.batch_norm1 = nn.BatchNorm2d(32,track_running_stats=False)\n self.batch_norm2 = nn.BatchNorm2d(64,track_running_stats=False)\n self.batch_norm3 = nn.BatchNorm2d(256,track_running_stats=False)\n self.dropout1 = nn.Dropout2d(p=0.05)\n self.dropout2 = nn.Dropout2d(p=0.1)\n self.fc1 = nn.Linear(256,64,bias=False)\n self.fc2 = nn.Linear(64, 32,bias=False)\n self.fc3 = nn.Linear(32, 10,bias=False)\n self.fc4 = nn.Linear(10, 1,bias=False)\n\n torch.nn.init.xavier_normal_(self.conv1.weight)\n torch.nn.init.xavier_normal_(self.conv2.weight)\n torch.nn.init.xavier_normal_(self.conv3.weight)\n torch.nn.init.xavier_normal_(self.conv4.weight)\n torch.nn.init.xavier_normal_(self.conv5.weight)\n torch.nn.init.xavier_normal_(self.conv6.weight)\n\n torch.nn.init.xavier_normal_(self.fc1.weight)\n torch.nn.init.xavier_normal_(self.fc2.weight)\n torch.nn.init.xavier_normal_(self.fc3.weight)\n torch.nn.init.xavier_normal_(self.fc4.weight)\n\n def forward(self,z): #y is avg image #z batch of list of 9 images\n y = torch.zeros([batch,256, 3,3], dtype=torch.float64)\n x = torch.zeros([batch,9],dtype=torch.float64)\n ftr = torch.zeros([batch,9,256,3,3])\n y = y.to(\"cuda\")\n x = x.to(\"cuda\")\n ftr = ftr.to(\"cuda\")\n \n for i in range(9):\n out,ftrs = self.helper(z[:,i])\n #print(out.shape)\n x[:,i] = out\n ftr[:,i] = ftrs\n log_x = F.log_softmax(x,dim=1) # log_alpha\n x = F.softmax(x,dim=1) \n\n for i in range(9): \n x1 = x[:,i] \n y = y + torch.mul(x1[:,None,None,None],ftr[:,i])\n\n return x, y, log_x, #alpha, log_alpha, avg_data\n \n def helper(self, x):\n #x1 = x\n #x1 =x\n x = self.conv1(x)\n x = F.relu(self.batch_norm1(x))\n\n x = (F.relu(self.conv2(x)))\n x = self.pool(x)\n \n x = self.conv3(x)\n x = F.relu(self.batch_norm2(x))\n\n x = (F.relu(self.conv4(x)))\n x = self.pool(x)\n x = self.dropout1(x)\n\n x = self.conv5(x)\n \n x = F.relu(self.batch_norm3(x))\n\n x = self.conv6(x)\n x1 = F.tanh(x)\n x = F.relu(x)\n x = self.pool(x)\n\n x = x.view(x.size(0), -1)\n\n x = self.dropout2(x)\n x = F.relu(self.fc1(x))\n x = F.relu(self.fc2(x))\n x = self.dropout2(x)\n x = F.relu(self.fc3(x))\n x = self.fc4(x)\n x = x[:,0] \n # print(x.shape)\n return x,x1", "_____no_output_____" ], [ "torch.manual_seed(n_seed)\nfocus_net = Focus().double()\nfocus_net = focus_net.to(\"cuda\")", "_____no_output_____" ], [ "class Classification(nn.Module):\n def __init__(self):\n super(Classification, self).__init__()\n self.conv1 = nn.Conv2d(in_channels=256, out_channels=128, kernel_size=3, padding=1)\n self.conv2 = nn.Conv2d(in_channels=128, out_channels=128, kernel_size=3, padding=1)\n self.conv3 = nn.Conv2d(in_channels=128, out_channels=256, kernel_size=3, padding=1)\n self.conv4 = nn.Conv2d(in_channels=256, out_channels=256, kernel_size=3, padding=1)\n self.conv5 = nn.Conv2d(in_channels=256, out_channels=512, kernel_size=3, padding=1)\n self.conv6 = nn.Conv2d(in_channels=512, out_channels=512, kernel_size=3, padding=1)\n self.pool = nn.MaxPool2d(kernel_size=2, stride=2,padding=1)\n self.batch_norm1 = nn.BatchNorm2d(128,track_running_stats=False)\n self.batch_norm2 = nn.BatchNorm2d(256,track_running_stats=False)\n self.batch_norm3 = nn.BatchNorm2d(512,track_running_stats=False)\n self.dropout1 = nn.Dropout2d(p=0.05)\n self.dropout2 = nn.Dropout2d(p=0.1)\n self.global_average_pooling = nn.AvgPool2d(kernel_size=2)\n self.fc1 = nn.Linear(512,128)\n # self.fc2 = nn.Linear(128, 64)\n # self.fc3 = nn.Linear(64, 10)\n self.fc2 = nn.Linear(128, 3)\n\n torch.nn.init.xavier_normal_(self.conv1.weight)\n torch.nn.init.xavier_normal_(self.conv2.weight)\n torch.nn.init.xavier_normal_(self.conv3.weight)\n torch.nn.init.xavier_normal_(self.conv4.weight)\n torch.nn.init.xavier_normal_(self.conv5.weight)\n torch.nn.init.xavier_normal_(self.conv6.weight)\n\n torch.nn.init.zeros_(self.conv1.bias)\n torch.nn.init.zeros_(self.conv2.bias)\n torch.nn.init.zeros_(self.conv3.bias)\n torch.nn.init.zeros_(self.conv4.bias)\n torch.nn.init.zeros_(self.conv5.bias)\n torch.nn.init.zeros_(self.conv6.bias)\n\n\n torch.nn.init.xavier_normal_(self.fc1.weight)\n torch.nn.init.xavier_normal_(self.fc2.weight)\n torch.nn.init.zeros_(self.fc1.bias)\n torch.nn.init.zeros_(self.fc2.bias)\n\n\n def forward(self, x):\n x = self.conv1(x)\n x = F.relu(self.batch_norm1(x))\n\n x = (F.relu(self.conv2(x)))\n x = self.pool(x)\n \n x = self.conv3(x)\n x = F.relu(self.batch_norm2(x))\n\n x = (F.relu(self.conv4(x)))\n x = self.pool(x)\n x = self.dropout1(x)\n\n x = self.conv5(x)\n x = F.relu(self.batch_norm3(x))\n\n x = (F.relu(self.conv6(x)))\n x = self.pool(x)\n #print(x.shape)\n x = self.global_average_pooling(x)\n x = x.squeeze()\n #x = x.view(x.size(0), -1)\n #print(x.shape)\n x = self.dropout2(x)\n x = F.relu(self.fc1(x))\n #x = F.relu(self.fc2(x))\n #x = self.dropout2(x)\n #x = F.relu(self.fc3(x))\n x = self.fc2(x)\n return x", "_____no_output_____" ], [ "torch.manual_seed(n_seed)\nclassify = Classification().double()\nclassify = classify.to(\"cuda\")", "_____no_output_____" ], [ "test_images =[] #list of mosaic images, each mosaic image is saved as laist of 9 images\nfore_idx_test =[] #list of indexes at which foreground image is present in a mosaic image \ntest_label=[] # label of mosaic image = foreground class present in that mosaic\nfor i in range(10000):\n np.random.seed(i+30000)\n bg_idx = np.random.randint(0,35000,8)\n fg_idx = np.random.randint(0,15000)\n fg = np.random.randint(0,9)\n fore_idx_test.append(fg)\n image_list,label = create_mosaic_img(bg_idx,fg_idx,fg)\n test_images.append(image_list)\n test_label.append(label)", "_____no_output_____" ], [ "test_data = MosaicDataset(test_images,test_label,fore_idx_test)\ntest_loader = DataLoader( test_data,batch_size= batch ,shuffle=False)", "_____no_output_____" ], [ "criterion = nn.CrossEntropyLoss()\ndef my_cross_entropy(x, y,alpha,log_alpha,k):\n # log_prob = -1.0 * F.log_softmax(x, 1)\n # loss = log_prob.gather(1, y.unsqueeze(1))\n # loss = loss.mean()\n loss = criterion(x,y)\n \n #alpha = torch.clamp(alpha,min=1e-10) \n \n b = -1.0* alpha * log_alpha\n b = torch.mean(torch.sum(b,dim=1))\n closs = loss\n entropy = b \n loss = (1-k)*loss + ((k)*b)\n return loss,closs,entropy", "_____no_output_____" ], [ "import torch.optim as optim\n# criterion_classify = nn.CrossEntropyLoss()\noptimizer_focus = optim.Adam(focus_net.parameters(), lr=0.001)#, momentum=0.9)\noptimizer_classify = optim.Adam(classify.parameters(), lr=0.001)#, momentum=0.9)", "_____no_output_____" ], [ "col1=[]\ncol2=[]\ncol3=[]\ncol4=[]\ncol5=[]\ncol6=[]\ncol7=[]\ncol8=[]\ncol9=[]\ncol10=[]\ncol11=[]\ncol12=[]\ncol13=[]\ncol14 = [] # train average sparsity\ncol15 = [] # test average sparsity\n\n", "_____no_output_____" ], [ "correct = 0\ntotal = 0\ncount = 0\nflag = 1\nfocus_true_pred_true =0\nfocus_false_pred_true =0\nfocus_true_pred_false =0\nfocus_false_pred_false =0\n\nargmax_more_than_half = 0\nargmax_less_than_half =0\nsparse_val = 0 \nfocus_net.eval()\nclassify.eval()\nwith torch.no_grad():\n for data in train_loader:\n inputs, labels , fore_idx = data\n inputs = inputs.double()\n inputs, labels , fore_idx = inputs.to(\"cuda\"),labels.to(\"cuda\"), fore_idx.to(\"cuda\")\n alphas, avg_images,_ = focus_net(inputs)\n # print(inputs.shape, alphas.shape, avg_images.shape)\n outputs = classify(avg_images)\n\n _, predicted = torch.max(outputs.data, 1)\n \n sparse_val += torch.sum(torch.sum(alphas>0.01,dim=1)).item()\n\n\n for j in range(labels.size(0)):\n count += 1\n focus = torch.argmax(alphas[j])\n if alphas[j][focus] >= 0.5 :\n argmax_more_than_half += 1\n else:\n argmax_less_than_half += 1\n\n if(focus == fore_idx[j] and predicted[j] == labels[j]):\n focus_true_pred_true += 1\n elif(focus != fore_idx[j] and predicted[j] == labels[j]):\n focus_false_pred_true += 1\n elif(focus == fore_idx[j] and predicted[j] != labels[j]):\n focus_true_pred_false += 1\n elif(focus != fore_idx[j] and predicted[j] != labels[j]):\n focus_false_pred_false += 1\n\n total += labels.size(0)\n correct += (predicted == labels).sum().item()\n\nprint('Accuracy of the network on the 30000 train images: %f %%' % ( 100 * correct / total))\nprint(\"total correct\", correct)\nprint(\"total train set images\", total)\n\nprint(\"focus_true_pred_true %d =============> FTPT : %f %%\" % (focus_true_pred_true , (100 * focus_true_pred_true / total) ) )\nprint(\"focus_false_pred_true %d =============> FFPT : %f %%\" % (focus_false_pred_true, (100 * focus_false_pred_true / total) ) )\nprint(\"focus_true_pred_false %d =============> FTPF : %f %%\" %( focus_true_pred_false , ( 100 * focus_true_pred_false / total) ) )\nprint(\"focus_false_pred_false %d =============> FFPF : %f %%\" % (focus_false_pred_false, ( 100 * focus_false_pred_false / total) ) )\nprint(\"Sparsity_Value %d =============> AVG Sparsity : %f \" % (sparse_val,(sparse_val)/total))\nprint(\"argmax_more_than_half ==================> \",argmax_more_than_half)\nprint(\"argmax_less_than_half ==================> \",argmax_less_than_half)\nprint(count)\n\nprint(\"=\"*100)\n\ncol1.append(0)\ncol2.append(argmax_more_than_half)\ncol3.append(argmax_less_than_half)\ncol4.append(focus_true_pred_true)\ncol5.append(focus_false_pred_true)\ncol6.append(focus_true_pred_false)\ncol7.append(focus_false_pred_false)\ncol14.append(sparse_val)", "Accuracy of the network on the 30000 train images: 33.813333 %\ntotal correct 10144\ntotal train set images 30000\nfocus_true_pred_true 405 =============> FTPT : 1.350000 %\nfocus_false_pred_true 9739 =============> FFPT : 32.463333 %\nfocus_true_pred_false 1624 =============> FTPF : 5.413333 %\nfocus_false_pred_false 18232 =============> FFPF : 60.773333 %\nSparsity_Value 270000 =============> AVG Sparsity : 9.000000 \nargmax_more_than_half ==================> 0\nargmax_less_than_half ==================> 30000\n30000\n====================================================================================================\n" ], [ "correct = 0\ntotal = 0\ncount = 0\nflag = 1\nfocus_true_pred_true =0\nfocus_false_pred_true =0\nfocus_true_pred_false =0\nfocus_false_pred_false =0\n\nargmax_more_than_half = 0\nargmax_less_than_half =0\nsparse_val = 0\nfocus_net.eval()\nclassify.eval()\nwith torch.no_grad():\n for data in test_loader:\n inputs, labels , fore_idx = data\n inputs = inputs.double()\n inputs, labels , fore_idx = inputs.to(\"cuda\"),labels.to(\"cuda\"), fore_idx.to(\"cuda\")\n alphas, avg_images,_ = focus_net(inputs)\n outputs = classify(avg_images)\n\n _, predicted = torch.max(outputs.data, 1)\n \n sparse_val += torch.sum(torch.sum(alphas>0.01,dim=1)).item()\n\n for j in range(labels.size(0)):\n focus = torch.argmax(alphas[j])\n if alphas[j][focus] >= 0.5 :\n argmax_more_than_half += 1\n else:\n argmax_less_than_half += 1\n\n if(focus == fore_idx[j] and predicted[j] == labels[j]):\n focus_true_pred_true += 1\n elif(focus != fore_idx[j] and predicted[j] == labels[j]):\n focus_false_pred_true += 1\n elif(focus == fore_idx[j] and predicted[j] != labels[j]):\n focus_true_pred_false += 1\n elif(focus != fore_idx[j] and predicted[j] != labels[j]):\n focus_false_pred_false += 1\n\n total += labels.size(0)\n correct += (predicted == labels).sum().item()\n\nprint('Accuracy of the network on the 10000 test images: %f %%' % (100 * correct / total))\nprint(\"total correct\", correct)\nprint(\"total test set images\", total)\n\nprint(\"focus_true_pred_true %d =============> FTPT : %f %%\" % (focus_true_pred_true , (100 * focus_true_pred_true / total) ) )\nprint(\"focus_false_pred_true %d =============> FFPT : %f %%\" % (focus_false_pred_true, (100 * focus_false_pred_true / total) ) )\nprint(\"focus_true_pred_false %d =============> FTPF : %f %%\" %( focus_true_pred_false , ( 100 * focus_true_pred_false / total) ) )\nprint(\"focus_false_pred_false %d =============> FFPF : %f %%\" % (focus_false_pred_false, ( 100 * focus_false_pred_false / total) ) )\nprint(\"Sparsity_Value %d =============> AVG Sparsity : %f \" % (sparse_val,(sparse_val)/total))\nprint(\"argmax_more_than_half ==================> \",argmax_more_than_half)\nprint(\"argmax_less_than_half ==================> \",argmax_less_than_half)\ncol8.append(argmax_more_than_half)\ncol9.append(argmax_less_than_half)\ncol10.append(focus_true_pred_true)\ncol11.append(focus_false_pred_true)\ncol12.append(focus_true_pred_false)\ncol13.append(focus_false_pred_false)\ncol15.append(sparse_val)", "Accuracy of the network on the 10000 test images: 33.550000 %\ntotal correct 3355\ntotal test set images 10000\nfocus_true_pred_true 138 =============> FTPT : 1.380000 %\nfocus_false_pred_true 3217 =============> FFPT : 32.170000 %\nfocus_true_pred_false 514 =============> FTPF : 5.140000 %\nfocus_false_pred_false 6131 =============> FFPF : 61.310000 %\nSparsity_Value 90000 =============> AVG Sparsity : 9.000000 \nargmax_more_than_half ==================> 0\nargmax_less_than_half ==================> 10000\n" ], [ "nos_epochs = 100\nfocus_true_pred_true =0\nfocus_false_pred_true =0\nfocus_true_pred_false =0\nfocus_false_pred_false =0\n\nargmax_more_than_half = 0\nargmax_less_than_half =0\nfocus_net.train()\nclassify.train()\ntr_loss = []\n\n\nfor epoch in range(nos_epochs): # loop over the dataset multiple times\n focus_net.train()\n classify.train()\n focus_true_pred_true =0\n focus_false_pred_true =0\n focus_true_pred_false =0\n focus_false_pred_false =0\n \n argmax_more_than_half = 0\n argmax_less_than_half =0\n sparse_val = 0 \n \n running_loss = 0.0\n epoch_loss = []\n cnt=0\n\n iteration = desired_num // batch\n \n #training data set\n \n for i, data in enumerate(train_loader):\n inputs , labels , fore_idx = data\n inputs = inputs.double()\n inputs, labels = inputs.to(\"cuda\"), labels.to(\"cuda\")\n # zero the parameter gradients\n optimizer_focus.zero_grad()\n optimizer_classify.zero_grad()\n alphas, avg_images,log_alphas = focus_net(inputs)\n outputs = classify(avg_images)\n \n # outputs, alphas, avg_images = classify(inputs)\n\n _, predicted = torch.max(outputs.data, 1)\n# print(outputs)\n# print(outputs.shape,labels.shape , torch.argmax(outputs, dim=1))\n\n loss,_,_ = my_cross_entropy(outputs, labels,alphas,log_alphas,k) \n loss.backward()\n optimizer_focus.step()\n optimizer_classify.step()\n\n running_loss += loss.item()\n mini = 60\n if cnt % mini == mini-1: # print every 40 mini-batches\n print('[%d, %5d] loss: %.3f' %(epoch + 1, cnt + 1, running_loss / mini))\n epoch_loss.append(running_loss/mini)\n running_loss = 0.0\n cnt=cnt+1\n \n \n if epoch % 1 == 0:\n sparse_val += torch.sum(torch.sum(alphas>0.01,dim=1)).item()\n for j in range (batch):\n focus = torch.argmax(alphas[j])\n\n if(alphas[j][focus] >= 0.5):\n argmax_more_than_half +=1\n else:\n argmax_less_than_half +=1\n\n if(focus == fore_idx[j] and predicted[j] == labels[j]):\n focus_true_pred_true += 1\n\n elif(focus != fore_idx[j] and predicted[j] == labels[j]):\n focus_false_pred_true +=1\n\n elif(focus == fore_idx[j] and predicted[j] != labels[j]):\n focus_true_pred_false +=1\n\n elif(focus != fore_idx[j] and predicted[j] != labels[j]):\n focus_false_pred_false +=1\n\n\n tr_loss.append(np.mean(epoch_loss))\n if epoch % 1 == 0:\n col1.append(epoch+1)\n col2.append(argmax_more_than_half)\n col3.append(argmax_less_than_half)\n col4.append(focus_true_pred_true)\n col5.append(focus_false_pred_true)\n col6.append(focus_true_pred_false)\n col7.append(focus_false_pred_false)\n col14.append(sparse_val)\n \n #************************************************************************\n #testing data set \n focus_net.eval()\n classify.eval() \n with torch.no_grad():\n focus_true_pred_true =0\n focus_false_pred_true =0\n focus_true_pred_false =0\n focus_false_pred_false =0\n\n argmax_more_than_half = 0\n argmax_less_than_half =0\n sparse_val = 0\n for data in test_loader:\n inputs, labels , fore_idx = data\n inputs = inputs.double()\n inputs, labels = inputs.to(\"cuda\"), labels.to(\"cuda\")\n alphas, avg_images,log_alphas = focus_net(inputs)\n outputs = classify(avg_images)\n #outputs, alphas, avg_images = classify(inputs)\n _, predicted = torch.max(outputs.data, 1)\n sparse_val += torch.sum(torch.sum(alphas>0.01,dim=1)).item()\n for j in range (batch):\n focus = torch.argmax(alphas[j])\n\n if(alphas[j][focus] >= 0.5):\n argmax_more_than_half +=1\n else:\n argmax_less_than_half +=1\n\n if(focus == fore_idx[j] and predicted[j] == labels[j]):\n focus_true_pred_true += 1\n\n elif(focus != fore_idx[j] and predicted[j] == labels[j]):\n focus_false_pred_true +=1\n\n elif(focus == fore_idx[j] and predicted[j] != labels[j]):\n focus_true_pred_false +=1\n\n elif(focus != fore_idx[j] and predicted[j] != labels[j]):\n focus_false_pred_false +=1\n \n col8.append(argmax_more_than_half)\n col9.append(argmax_less_than_half)\n col10.append(focus_true_pred_true)\n col11.append(focus_false_pred_true)\n col12.append(focus_true_pred_false)\n col13.append(focus_false_pred_false)\n col15.append(sparse_val)\n if(np.mean(epoch_loss) <= 0.05):\n break;\nprint('Finished Training') ", "[1, 60] loss: 1.221\n[1, 120] loss: 1.097\n[2, 60] loss: 1.030\n[2, 120] loss: 0.953\n[3, 60] loss: 0.857\n[3, 120] loss: 0.782\n[4, 60] loss: 0.725\n[4, 120] loss: 0.689\n[5, 60] loss: 0.663\n[5, 120] loss: 0.640\n[6, 60] loss: 0.607\n[6, 120] loss: 0.591\n[7, 60] loss: 0.591\n[7, 120] loss: 0.587\n[8, 60] loss: 0.566\n[8, 120] loss: 0.549\n[9, 60] loss: 0.539\n[9, 120] loss: 0.565\n[10, 60] loss: 0.546\n[10, 120] loss: 0.572\n[11, 60] loss: 0.551\n[11, 120] loss: 0.555\n[12, 60] loss: 0.599\n[12, 120] loss: 0.547\n[13, 60] loss: 0.491\n[13, 120] loss: 0.480\n[14, 60] loss: 0.462\n[14, 120] loss: 0.465\n[15, 60] loss: 0.424\n[15, 120] loss: 0.407\n[16, 60] loss: 0.352\n[16, 120] loss: 0.338\n[17, 60] loss: 0.282\n[17, 120] loss: 0.289\n[18, 60] loss: 0.325\n[18, 120] loss: 0.257\n[19, 60] loss: 0.225\n[19, 120] loss: 0.220\n[20, 60] loss: 0.172\n[20, 120] loss: 0.166\n[21, 60] loss: 0.133\n[21, 120] loss: 0.152\n[22, 60] loss: 0.110\n[22, 120] loss: 0.122\n[23, 60] loss: 0.085\n[23, 120] loss: 0.101\n[24, 60] loss: 0.081\n[24, 120] loss: 0.083\n[25, 60] loss: 0.073\n[25, 120] loss: 0.072\n[26, 60] loss: 0.054\n[26, 120] loss: 0.076\n[27, 60] loss: 0.047\n[27, 120] loss: 0.057\n[28, 60] loss: 0.045\n[28, 120] loss: 0.056\n[29, 60] loss: 0.042\n[29, 120] loss: 0.048\nFinished Training\n" ], [ "torch.save(focus_net.state_dict(),path+\"weights_focus_0.pt\") \ntorch.save(classify.state_dict(),path+\"weights_classify_0.pt\")", "_____no_output_____" ], [ "columns = [\"epochs\", \"argmax > 0.5\" ,\"argmax < 0.5\", \"focus_true_pred_true\", \"focus_false_pred_true\", \"focus_true_pred_false\", \"focus_false_pred_false\" ,\"sparse_val\"]", "_____no_output_____" ], [ "df_train = pd.DataFrame()\ndf_test = pd.DataFrame()", "_____no_output_____" ], [ "len(col1),col9", "_____no_output_____" ], [ "plt.plot(np.arange(1,epoch+2),tr_loss)\nplt.xlabel(\"epochs\", fontsize=14, fontweight = 'bold')\nplt.ylabel(\"Loss\", fontsize=14, fontweight = 'bold')\nplt.title(\"Train Loss\")\nplt.grid()\nplt.show()", "_____no_output_____" ], [ "np.save(\"train_loss.npy\",{\"training_loss\":tr_loss})", "_____no_output_____" ], [ "df_train[columns[0]] = col1\ndf_train[columns[1]] = col2\ndf_train[columns[2]] = col3\ndf_train[columns[3]] = col4\ndf_train[columns[4]] = col5\ndf_train[columns[5]] = col6\ndf_train[columns[6]] = col7\ndf_train[columns[7]] = col14\n\ndf_test[columns[0]] = col1\ndf_test[columns[1]] = col8\ndf_test[columns[2]] = col9\ndf_test[columns[3]] = col10\ndf_test[columns[4]] = col11\ndf_test[columns[5]] = col12\ndf_test[columns[6]] = col13\ndf_test[columns[7]] = col15", "_____no_output_____" ], [ "df_train", "_____no_output_____" ], [ "df_train.to_csv(path+\"_train.csv\",index=False)", "_____no_output_____" ], [ "# plt.figure(12,12)\nplt.plot(col1,col2, label='argmax > 0.5')\nplt.plot(col1,col3, label='argmax < 0.5')\n\nplt.legend(loc='center left', bbox_to_anchor=(1, 0.5))\nplt.xlabel(\"epochs\")\nplt.ylabel(\"training data\")\nplt.title(\"On Training set\")\nplt.show()\n\n", "_____no_output_____" ], [ "plt.figure(figsize=(6,5))\nplt.plot(col1,np.array(col4)/300, label =\"FTPT \")\nplt.plot(col1,np.array(col5)/300, label =\"FFPT \")\nplt.plot(col1,np.array(col6)/300, label =\"FTPF \")\nplt.plot(col1,np.array(col7)/300, label =\"FFPF \")\nplt.title(\"On Training set\")\n#plt.legend(loc='center left', bbox_to_anchor=(1, 0.5))\nplt.xlabel(\"epochs\", fontsize=14, fontweight = 'bold')\nplt.ylabel(\"percentage train data\", fontsize=14, fontweight = 'bold')\n# plt.xlabel(\"epochs\")\n# plt.ylabel(\"training data\")\nplt.legend()\n\nplt.savefig(path + \"_train.png\",bbox_inches=\"tight\")\nplt.savefig(path + \"_train.pdf\",bbox_inches=\"tight\")\nplt.grid()\nplt.show()", "_____no_output_____" ], [ "plt.figure(figsize=(6,5))\nplt.plot(col1,np.array(col14)/30000, label =\"sparsity_val\")\n\nplt.title(\"On Training set\")\n#plt.legend(loc='center left', bbox_to_anchor=(1, 0.5))\nplt.xlabel(\"epochs\", fontsize=14, fontweight = 'bold')\nplt.ylabel(\"average sparsity value\", fontsize=14, fontweight = 'bold')\n# plt.xlabel(\"epochs\")\n# plt.ylabel(\"sparsity_value\")\n\nplt.savefig(path + \"sparsity_train.png\",bbox_inches=\"tight\")\nplt.savefig(path + \"sparsity_train.pdf\",bbox_inches=\"tight\")\nplt.grid()\nplt.show()", "_____no_output_____" ], [ "df_test", "_____no_output_____" ], [ "df_test.to_csv(path+\"_test.csv\")", "_____no_output_____" ], [ "# plt.figure(12,12)\nplt.plot(col1,col8, label='argmax > 0.5')\nplt.plot(col1,col9, label='argmax < 0.5')\n\nplt.legend(loc='center left', bbox_to_anchor=(1, 0.5))\nplt.xlabel(\"epochs\")\nplt.ylabel(\"Testing data\")\nplt.title(\"On Testing set\")\nplt.show()\n\n", "_____no_output_____" ], [ "plt.figure(figsize=(6,5))\nplt.plot(col1,np.array(col10)/100, label =\"FTPT \")\nplt.plot(col1,np.array(col11)/100, label =\"FFPT \")\nplt.plot(col1,np.array(col12)/100, label =\"FTPF \")\nplt.plot(col1,np.array(col13)/100, label =\"FFPF \")\nplt.title(\"On Testing set\")\n\n#plt.legend(loc='center left', bbox_to_anchor=(1, 0.5))\nplt.xlabel(\"epochs\", fontsize=14, fontweight = 'bold')\nplt.ylabel(\"percentage test data\", fontsize=14, fontweight = 'bold')\n# plt.xlabel(\"epochs\")\n# plt.ylabel(\"training data\")\nplt.legend()\n\n#plt.legend(loc='center left', bbox_to_anchor=(1, 0.5))\nplt.xlabel(\"epochs\")\n# plt.ylabel(\"Testing data\")\nplt.savefig(path + \"_test.png\",bbox_inches=\"tight\")\nplt.savefig(path + \"_test.pdf\",bbox_inches=\"tight\")\nplt.grid()\nplt.show()", "_____no_output_____" ], [ "plt.figure(figsize=(6,5))\nplt.plot(col1,np.array(col15)/10000, label =\"sparsity_val\")\n\nplt.title(\"On Testing set\")\n#plt.legend(loc='center left', bbox_to_anchor=(1, 0.5))\nplt.xlabel(\"epochs\", fontsize=14, fontweight = 'bold')\nplt.ylabel(\"average sparsity value\", fontsize=14, fontweight = 'bold')\nplt.grid()\n\nplt.savefig(path + \"sparsity_test.png\",bbox_inches=\"tight\")\nplt.savefig(path + \"sparsity_test.pdf\",bbox_inches=\"tight\")\nplt.show()", "_____no_output_____" ], [ "correct = 0\ntotal = 0\ncount = 0\nflag = 1\nfocus_true_pred_true =0\nfocus_false_pred_true =0\nfocus_true_pred_false =0\nfocus_false_pred_false =0\n\nargmax_more_than_half = 0\nargmax_less_than_half =0\nsparse_val = 0\nfocus_net.eval()\nclassify.eval()\nwith torch.no_grad():\n for data in train_loader:\n inputs, labels , fore_idx = data\n inputs = inputs.double()\n inputs, labels , fore_idx = inputs.to(\"cuda\"),labels.to(\"cuda\"), fore_idx.to(\"cuda\")\n alphas, avg_images,_ = focus_net(inputs)\n outputs = classify(avg_images)\n\n _, predicted = torch.max(outputs.data, 1)\n sparse_val += torch.sum(torch.sum(alphas>0.01,dim=1)).item()\n for j in range(labels.size(0)):\n count += 1\n focus = torch.argmax(alphas[j])\n if alphas[j][focus] >= 0.5 :\n argmax_more_than_half += 1\n else:\n argmax_less_than_half += 1\n\n if(focus == fore_idx[j] and predicted[j] == labels[j]):\n focus_true_pred_true += 1\n elif(focus != fore_idx[j] and predicted[j] == labels[j]):\n focus_false_pred_true += 1\n elif(focus == fore_idx[j] and predicted[j] != labels[j]):\n focus_true_pred_false += 1\n elif(focus != fore_idx[j] and predicted[j] != labels[j]):\n focus_false_pred_false += 1\n\n total += labels.size(0)\n correct += (predicted == labels).sum().item()\n\nprint('Accuracy of the network on the 30000 train images: %f %%' % ( 100 * correct / total))\nprint(\"total correct\", correct)\nprint(\"total train set images\", total)\n\nprint(\"focus_true_pred_true %d =============> FTPT : %f %%\" % (focus_true_pred_true , (100 * focus_true_pred_true / total) ) )\nprint(\"focus_false_pred_true %d =============> FFPT : %f %%\" % (focus_false_pred_true, (100 * focus_false_pred_true / total) ) )\nprint(\"focus_true_pred_false %d =============> FTPF : %f %%\" %( focus_true_pred_false , ( 100 * focus_true_pred_false / total) ) )\nprint(\"focus_false_pred_false %d =============> FFPF : %f %%\" % (focus_false_pred_false, ( 100 * focus_false_pred_false / total) ) )\nprint(\"Sparsity_Value %d =============> AVG Sparsity : %f \" % (sparse_val,(sparse_val)/total))\nprint(\"argmax_more_than_half ==================> \",argmax_more_than_half)\nprint(\"argmax_less_than_half ==================> \",argmax_less_than_half)\nprint(count)\n\nprint(\"=\"*100)", "Accuracy of the network on the 30000 train images: 99.466667 %\ntotal correct 29840\ntotal train set images 30000\nfocus_true_pred_true 23409 =============> FTPT : 78.030000 %\nfocus_false_pred_true 6431 =============> FFPT : 21.436667 %\nfocus_true_pred_false 51 =============> FTPF : 0.170000 %\nfocus_false_pred_false 109 =============> FFPF : 0.363333 %\nSparsity_Value 118107 =============> AVG Sparsity : 3.936900 \nargmax_more_than_half ==================> 21426\nargmax_less_than_half ==================> 8574\n30000\n====================================================================================================\n" ], [ "correct = 0\ntotal = 0\ncount = 0\nflag = 1\nfocus_true_pred_true =0\nfocus_false_pred_true =0\nfocus_true_pred_false =0\nfocus_false_pred_false =0\n\nargmax_more_than_half = 0\nargmax_less_than_half =0\nsparse_val = 0 \nfocus_net.eval()\nclassify.eval()\nwith torch.no_grad():\n for data in test_loader:\n inputs, labels , fore_idx = data\n inputs = inputs.double()\n inputs, labels , fore_idx = inputs.to(\"cuda\"),labels.to(\"cuda\"), fore_idx.to(\"cuda\")\n alphas, avg_images , _ = focus_net(inputs)\n outputs = classify(avg_images)\n\n _, predicted = torch.max(outputs.data, 1)\n sparse_val += torch.sum(torch.sum(alphas>0.01,dim=1)).item()\n for j in range(labels.size(0)):\n focus = torch.argmax(alphas[j])\n if alphas[j][focus] >= 0.5 :\n argmax_more_than_half += 1\n else:\n argmax_less_than_half += 1\n\n if(focus == fore_idx[j] and predicted[j] == labels[j]):\n focus_true_pred_true += 1\n elif(focus != fore_idx[j] and predicted[j] == labels[j]):\n focus_false_pred_true += 1\n elif(focus == fore_idx[j] and predicted[j] != labels[j]):\n focus_true_pred_false += 1\n elif(focus != fore_idx[j] and predicted[j] != labels[j]):\n focus_false_pred_false += 1\n\n total += labels.size(0)\n correct += (predicted == labels).sum().item()\n\nprint('Accuracy of the network on the 10000 test images: %f %%' % (\n 100 * correct / total))\nprint(\"total correct\", correct)\nprint(\"total test set images\", total)\n\nprint(\"focus_true_pred_true %d =============> FTPT : %f %%\" % (focus_true_pred_true , (100 * focus_true_pred_true / total) ) )\nprint(\"focus_false_pred_true %d =============> FFPT : %f %%\" % (focus_false_pred_true, (100 * focus_false_pred_true / total) ) )\nprint(\"focus_true_pred_false %d =============> FTPF : %f %%\" %( focus_true_pred_false , ( 100 * focus_true_pred_false / total) ) )\nprint(\"focus_false_pred_false %d =============> FFPF : %f %%\" % (focus_false_pred_false, ( 100 * focus_false_pred_false / total) ) )\nprint(\"Sparsity_Value %d =============> AVG Sparsity : %f \" % (sparse_val,(sparse_val)/total))\n\nprint(\"argmax_more_than_half ==================> \",argmax_more_than_half)\nprint(\"argmax_less_than_half ==================> \",argmax_less_than_half)", "Accuracy of the network on the 10000 test images: 94.650000 %\ntotal correct 9465\ntotal test set images 10000\nfocus_true_pred_true 7454 =============> FTPT : 74.540000 %\nfocus_false_pred_true 2011 =============> FFPT : 20.110000 %\nfocus_true_pred_false 215 =============> FTPF : 2.150000 %\nfocus_false_pred_false 320 =============> FFPF : 3.200000 %\nSparsity_Value 40924 =============> AVG Sparsity : 4.092400 \nargmax_more_than_half ==================> 7072\nargmax_less_than_half ==================> 2928\n" ], [ "correct = 0\ntotal = 0\nfocus_net.eval()\nclassify.eval()\nwith torch.no_grad():\n for data in train_loader:\n inputs, labels , fore_idx = data\n inputs = inputs.double()\n inputs, labels = inputs.to(\"cuda\"), labels.to(\"cuda\")\n alphas, avg_images,_ = focus_net(inputs)\n outputs = classify(avg_images)\n\n _, predicted = torch.max(outputs.data, 1)\n\n total += labels.size(0)\n correct += (predicted == labels).sum().item()\n\nprint('Accuracy of the network on the 30000 train images: %f %%' % ( 100 * correct / total))\nprint(\"total correct\", correct)\nprint(\"total train set images\", total)", "Accuracy of the network on the 30000 train images: 99.490000 %\ntotal correct 29847\ntotal train set images 30000\n" ], [ "correct = 0\ntotal = 0\nfocus_net.eval()\nclassify.eval()\nwith torch.no_grad():\n for data in test_loader:\n inputs, labels , fore_idx = data\n inputs = inputs.double()\n inputs, labels = inputs.to(\"cuda\"), labels.to(\"cuda\")\n alphas, avg_images,_ = focus_net(inputs)\n outputs = classify(avg_images)\n\n _, predicted = torch.max(outputs.data, 1)\n\n total += labels.size(0)\n correct += (predicted == labels).sum().item()\n\nprint('Accuracy of the network on the 10000 test images: %f %%' % ( 100 * correct / total))\nprint(\"total correct\", correct)\nprint(\"total test set images\", total)", "Accuracy of the network on the 10000 test images: 94.650000 %\ntotal correct 9465\ntotal test set images 10000\n" ], [ "", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# These are all the modules we'll be using later. Make sure you can import them\n# before proceeding further.\nfrom __future__ import print_function\nimport numpy as np\nimport tensorflow as tf\nfrom six.moves import cPickle as pickle\nfrom six.moves import range", "_____no_output_____" ], [ "# The folder when dumped big 3D array has been stored from previous excercise\ndata_root = 'D:\\\\1_Workspaces\\\\UNDER_VCS\\\\github\\\\1_ML_NN\\\\python_with_math\\\\data'\n#a big 3D array to a big file. \npickle_file = 'notMNIST.pickle'\n\nwith open(data_root + '\\\\' + pickle_file, 'rb') as f:\n save = pickle.load(f)\n train_dataset = save['train_dataset']\n train_labels = save['train_labels']\n valid_dataset = save['valid_dataset']\n valid_labels = save['valid_labels']\n test_dataset = save['test_dataset']\n test_labels = save['test_labels']\n del save # hint to help gc free up memory\n print('Training set', train_dataset.shape, train_labels.shape)\n print('Validation set', valid_dataset.shape, valid_labels.shape)\n print('Test set', test_dataset.shape, test_labels.shape)", "Training set (200000, 28, 28) (200000,)\nValidation set (10000, 28, 28) (10000,)\nTest set (10000, 28, 28) (10000,)\n" ], [ "image_size = 28\nnum_labels = 10\n\ndef reformat(dataset, labels):\n dataset = dataset.reshape((-1, image_size * image_size)).astype(np.float32)\n # Map 0 to [1.0, 0.0, 0.0 ...], 1 to [0.0, 1.0, 0.0 ...]\n labels = (np.arange(num_labels) == labels[:,None]).astype(np.float32)\n return dataset, labels\ntrain_dataset, train_labels = reformat(train_dataset, train_labels)\nvalid_dataset, valid_labels = reformat(valid_dataset, valid_labels)\ntest_dataset, test_labels = reformat(test_dataset, test_labels)\nprint('Training set', train_dataset.shape, train_labels.shape)\nprint('Validation set', valid_dataset.shape, valid_labels.shape)\nprint('Test set', test_dataset.shape, test_labels.shape)", "Training set (200000, 784) (200000, 10)\nValidation set (10000, 784) (10000, 10)\nTest set (10000, 784) (10000, 10)\n" ], [ "# It loads all the data into TensorFlow and build the computation graph corresponding to our training:\n\n# With gradient descent training, even this much data is prohibitive.\n# Subset the training data for faster turnaround.\ntrain_subset = 10000\n\ngraph = tf.Graph()\nwith graph.as_default():\n\n # Input data.\n # Load the training, validation and test data into constants that are\n # attached to the graph.\n tf_train_dataset = tf.constant(train_dataset[:train_subset, :])\n tf_train_labels = tf.constant(train_labels[:train_subset])\n tf_valid_dataset = tf.constant(valid_dataset)\n tf_test_dataset = tf.constant(test_dataset)\n \n # Variables.\n # These are the parameters that we are going to be training. The weight\n # matrix will be initialized using random values following a (truncated)\n # normal distribution. The biases get initialized to zero.\n weights = tf.Variable(\n tf.truncated_normal([image_size * image_size, num_labels]))\n biases = tf.Variable(tf.zeros([num_labels]))\n \n # Training computation.\n # We multiply the inputs with the weight matrix, and add biases. We compute\n # the softmax and cross-entropy (it's one operation in TensorFlow, because\n # it's very common, and it can be optimized). We take the average of this\n # cross-entropy across all training examples: that's our loss.\n logits = tf.matmul(tf_train_dataset, weights) + biases\n loss = tf.reduce_mean(\n tf.nn.softmax_cross_entropy_with_logits(labels=tf_train_labels, logits=logits))\n \n # Optimizer.\n # We are going to find the minimum of this loss using gradient descent.\n optimizer = tf.train.GradientDescentOptimizer(0.5).minimize(loss)\n \n # Predictions for the training, validation, and test data.\n # These are not part of training, but merely here so that we can report\n # accuracy figures as we train.\n train_prediction = tf.nn.softmax(logits)\n valid_prediction = tf.nn.softmax(\n tf.matmul(tf_valid_dataset, weights) + biases)\n test_prediction = tf.nn.softmax(tf.matmul(tf_test_dataset, weights) + biases)", "_____no_output_____" ], [ "num_steps = 10000 #why 801?\n\ndef accuracy(predictions, labels):\n return (100.0 * np.sum(np.argmax(predictions, 1) == np.argmax(labels, 1))\n / predictions.shape[0])\n\n#it performs the train\nwith tf.Session(graph=graph) as session:\n # This is a one-time operation which ensures the parameters get initialized as\n # we described in the graph: random weights for the matrix, zeros for the\n # biases. \n tf.global_variables_initializer().run()\n print('Tensorflow graph initialized')\n for step in range(num_steps):\n # Run the computations. We tell .run() that we want to run the optimizer,\n # and get the loss value and the training predictions returned as numpy\n # arrays.\n _, l, predictions = session.run([optimizer, loss, train_prediction])\n if (step % 100 == 0):\n print('Loss at step %d: %f' % (step, l))\n print('Training accuracy: %.1f%%' % accuracy(\n predictions, train_labels[:train_subset, :]))\n # Calling .eval() on valid_prediction is basically like calling run(), but\n # just to get that one numpy array. Note that it recomputes all its graph\n # dependencies.\n print('Validation accuracy: %.1f%%' % accuracy(\n valid_prediction.eval(), valid_labels))\n print('Test accuracy: %.1f%%' % accuracy(test_prediction.eval(), test_labels))\n#TODO plot graph from accuracy data", "Tensorflow graph initialized\nLoss at step 0: 19.005764\nTraining accuracy: 8.6%\nValidation accuracy: 9.5%\nLoss at step 100: 2.306310\nTraining accuracy: 71.9%\nValidation accuracy: 70.7%\nLoss at step 200: 1.862624\nTraining accuracy: 75.1%\nValidation accuracy: 72.8%\nLoss at step 300: 1.619365\nTraining accuracy: 76.3%\nValidation accuracy: 73.7%\nLoss at step 400: 1.452627\nTraining accuracy: 77.3%\nValidation accuracy: 74.1%\nLoss at step 500: 1.328036\nTraining accuracy: 77.8%\nValidation accuracy: 74.4%\nLoss at step 600: 1.230201\nTraining accuracy: 78.4%\nValidation accuracy: 74.8%\nLoss at step 700: 1.151069\nTraining accuracy: 79.0%\nValidation accuracy: 74.9%\nLoss at step 800: 1.085528\nTraining accuracy: 79.5%\nValidation accuracy: 75.0%\nLoss at step 900: 1.030110\nTraining accuracy: 79.9%\nValidation accuracy: 75.1%\nLoss at step 1000: 0.982457\nTraining accuracy: 80.3%\nValidation accuracy: 75.3%\nLoss at step 1100: 0.940917\nTraining accuracy: 80.7%\nValidation accuracy: 75.3%\nLoss at step 1200: 0.904299\nTraining accuracy: 81.0%\nValidation accuracy: 75.5%\nLoss at step 1300: 0.871726\nTraining accuracy: 81.6%\nValidation accuracy: 75.5%\nLoss at step 1400: 0.842531\nTraining accuracy: 81.8%\nValidation accuracy: 75.6%\nLoss at step 1500: 0.816180\nTraining accuracy: 82.2%\nValidation accuracy: 75.7%\nLoss at step 1600: 0.792249\nTraining accuracy: 82.4%\nValidation accuracy: 75.9%\nLoss at step 1700: 0.770396\nTraining accuracy: 82.6%\nValidation accuracy: 76.0%\nLoss at step 1800: 0.750347\nTraining accuracy: 82.8%\nValidation accuracy: 76.2%\nLoss at step 1900: 0.731872\nTraining accuracy: 83.0%\nValidation accuracy: 76.2%\nLoss at step 2000: 0.714789\nTraining accuracy: 83.3%\nValidation accuracy: 76.2%\nLoss at step 2100: 0.698939\nTraining accuracy: 83.5%\nValidation accuracy: 76.3%\nLoss at step 2200: 0.684192\nTraining accuracy: 83.7%\nValidation accuracy: 76.4%\nLoss at step 2300: 0.670434\nTraining accuracy: 84.0%\nValidation accuracy: 76.5%\nLoss at step 2400: 0.657569\nTraining accuracy: 84.1%\nValidation accuracy: 76.4%\nLoss at step 2500: 0.645511\nTraining accuracy: 84.3%\nValidation accuracy: 76.4%\nLoss at step 2600: 0.634187\nTraining accuracy: 84.5%\nValidation accuracy: 76.5%\nLoss at step 2700: 0.623530\nTraining accuracy: 84.7%\nValidation accuracy: 76.4%\nLoss at step 2800: 0.613483\nTraining accuracy: 84.9%\nValidation accuracy: 76.4%\nLoss at step 2900: 0.603993\nTraining accuracy: 85.0%\nValidation accuracy: 76.5%\nLoss at step 3000: 0.595014\nTraining accuracy: 85.2%\nValidation accuracy: 76.5%\nLoss at step 3100: 0.586504\nTraining accuracy: 85.4%\nValidation accuracy: 76.5%\nLoss at step 3200: 0.578426\nTraining accuracy: 85.6%\nValidation accuracy: 76.6%\nLoss at step 3300: 0.570746\nTraining accuracy: 85.8%\nValidation accuracy: 76.6%\nLoss at step 3400: 0.563433\nTraining accuracy: 85.8%\nValidation accuracy: 76.6%\nLoss at step 3500: 0.556461\nTraining accuracy: 86.0%\nValidation accuracy: 76.7%\nLoss at step 3600: 0.549805\nTraining accuracy: 86.1%\nValidation accuracy: 76.7%\nLoss at step 3700: 0.543443\nTraining accuracy: 86.2%\nValidation accuracy: 76.6%\nLoss at step 3800: 0.537355\nTraining accuracy: 86.4%\nValidation accuracy: 76.7%\nLoss at step 3900: 0.531522\nTraining accuracy: 86.5%\nValidation accuracy: 76.7%\nLoss at step 4000: 0.525927\nTraining accuracy: 86.5%\nValidation accuracy: 76.7%\nLoss at step 4100: 0.520556\nTraining accuracy: 86.7%\nValidation accuracy: 76.8%\nLoss at step 4200: 0.515394\nTraining accuracy: 86.8%\nValidation accuracy: 76.8%\nLoss at step 4300: 0.510427\nTraining accuracy: 86.9%\nValidation accuracy: 76.8%\nLoss at step 4400: 0.505645\nTraining accuracy: 87.0%\nValidation accuracy: 76.9%\nLoss at step 4500: 0.501034\nTraining accuracy: 87.1%\nValidation accuracy: 76.9%\nLoss at step 4600: 0.496588\nTraining accuracy: 87.3%\nValidation accuracy: 77.0%\nLoss at step 4700: 0.492295\nTraining accuracy: 87.5%\nValidation accuracy: 77.0%\nLoss at step 4800: 0.488147\nTraining accuracy: 87.6%\nValidation accuracy: 77.0%\nLoss at step 4900: 0.484136\nTraining accuracy: 87.7%\nValidation accuracy: 77.0%\nLoss at step 5000: 0.480255\nTraining accuracy: 87.8%\nValidation accuracy: 77.0%\nLoss at step 5100: 0.476498\nTraining accuracy: 87.9%\nValidation accuracy: 76.9%\nLoss at step 5200: 0.472857\nTraining accuracy: 88.0%\nValidation accuracy: 77.0%\nLoss at step 5300: 0.469328\nTraining accuracy: 88.0%\nValidation accuracy: 76.9%\nLoss at step 5400: 0.465905\nTraining accuracy: 88.1%\nValidation accuracy: 77.0%\nLoss at step 5500: 0.462581\nTraining accuracy: 88.2%\nValidation accuracy: 77.0%\nLoss at step 5600: 0.459354\nTraining accuracy: 88.2%\nValidation accuracy: 77.0%\nLoss at step 5700: 0.456219\nTraining accuracy: 88.2%\nValidation accuracy: 77.0%\nLoss at step 5800: 0.453170\nTraining accuracy: 88.3%\nValidation accuracy: 77.0%\nLoss at step 5900: 0.450206\nTraining accuracy: 88.3%\nValidation accuracy: 77.1%\nLoss at step 6000: 0.447320\nTraining accuracy: 88.4%\nValidation accuracy: 77.1%\nLoss at step 6100: 0.444511\nTraining accuracy: 88.4%\nValidation accuracy: 77.2%\nLoss at step 6200: 0.441775\nTraining accuracy: 88.5%\nValidation accuracy: 77.2%\nLoss at step 6300: 0.439109\nTraining accuracy: 88.5%\nValidation accuracy: 77.2%\nLoss at step 6400: 0.436509\nTraining accuracy: 88.6%\nValidation accuracy: 77.2%\nLoss at step 6500: 0.433974\nTraining accuracy: 88.6%\nValidation accuracy: 77.2%\nLoss at step 6600: 0.431500\nTraining accuracy: 88.7%\nValidation accuracy: 77.2%\nLoss at step 6700: 0.429086\nTraining accuracy: 88.8%\nValidation accuracy: 77.2%\nLoss at step 6800: 0.426728\nTraining accuracy: 88.8%\nValidation accuracy: 77.2%\nLoss at step 6900: 0.424425\nTraining accuracy: 88.9%\nValidation accuracy: 77.2%\nLoss at step 7000: 0.422174\nTraining accuracy: 89.0%\nValidation accuracy: 77.2%\nLoss at step 7100: 0.419974\nTraining accuracy: 89.1%\nValidation accuracy: 77.1%\nLoss at step 7200: 0.417823\nTraining accuracy: 89.1%\nValidation accuracy: 77.2%\nLoss at step 7300: 0.415718\nTraining accuracy: 89.2%\nValidation accuracy: 77.2%\nLoss at step 7400: 0.413659\nTraining accuracy: 89.3%\nValidation accuracy: 77.2%\nLoss at step 7500: 0.411643\nTraining accuracy: 89.3%\nValidation accuracy: 77.2%\nLoss at step 7600: 0.409670\nTraining accuracy: 89.4%\nValidation accuracy: 77.2%\nLoss at step 7700: 0.407737\nTraining accuracy: 89.5%\nValidation accuracy: 77.2%\nLoss at step 7800: 0.405843\nTraining accuracy: 89.5%\nValidation accuracy: 77.2%\nLoss at step 7900: 0.403987\nTraining accuracy: 89.6%\nValidation accuracy: 77.2%\nLoss at step 8000: 0.402168\nTraining accuracy: 89.7%\nValidation accuracy: 77.2%\nLoss at step 8100: 0.400384\nTraining accuracy: 89.7%\nValidation accuracy: 77.2%\nLoss at step 8200: 0.398635\nTraining accuracy: 89.8%\nValidation accuracy: 77.2%\nLoss at step 8300: 0.396918\nTraining accuracy: 89.8%\nValidation accuracy: 77.1%\nLoss at step 8400: 0.395234\nTraining accuracy: 89.8%\nValidation accuracy: 77.1%\nLoss at step 8500: 0.393582\nTraining accuracy: 89.8%\nValidation accuracy: 77.2%\nLoss at step 8600: 0.391959\nTraining accuracy: 89.9%\nValidation accuracy: 77.2%\nLoss at step 8700: 0.390365\nTraining accuracy: 89.9%\nValidation accuracy: 77.2%\nLoss at step 8800: 0.388800\nTraining accuracy: 90.0%\nValidation accuracy: 77.2%\nLoss at step 8900: 0.387263\nTraining accuracy: 90.0%\nValidation accuracy: 77.2%\nLoss at step 9000: 0.385752\nTraining accuracy: 90.1%\nValidation accuracy: 77.3%\nLoss at step 9100: 0.384267\nTraining accuracy: 90.1%\nValidation accuracy: 77.2%\nLoss at step 9200: 0.382808\nTraining accuracy: 90.2%\nValidation accuracy: 77.2%\nLoss at step 9300: 0.381373\nTraining accuracy: 90.2%\nValidation accuracy: 77.2%\nLoss at step 9400: 0.379962\nTraining accuracy: 90.3%\nValidation accuracy: 77.2%\nLoss at step 9500: 0.378574\nTraining accuracy: 90.3%\nValidation accuracy: 77.2%\nLoss at step 9600: 0.377209\nTraining accuracy: 90.3%\nValidation accuracy: 77.2%\nLoss at step 9700: 0.375866\nTraining accuracy: 90.3%\nValidation accuracy: 77.2%\nLoss at step 9800: 0.374544\nTraining accuracy: 90.4%\nValidation accuracy: 77.2%\nLoss at step 9900: 0.373243\nTraining accuracy: 90.4%\nValidation accuracy: 77.2%\nTest accuracy: 84.2%\n" ], [ "# Let's now switch to stochastic gradient descent training instead, which is much faster.\n\nbatch_size = 128\n\ngraph = tf.Graph()\nwith graph.as_default():\n\n # Input data. For the training data, we use a placeholder that will be fed\n # at run time with a training minibatch.\n tf_train_dataset = tf.placeholder(tf.float32,\n shape=(batch_size, image_size * image_size))\n tf_train_labels = tf.placeholder(tf.float32, shape=(batch_size, num_labels))\n tf_valid_dataset = tf.constant(valid_dataset)\n tf_test_dataset = tf.constant(test_dataset)\n \n # Variables.\n weights = tf.Variable(\n tf.truncated_normal([image_size * image_size, num_labels]))\n biases = tf.Variable(tf.zeros([num_labels]))\n \n # Training computation.\n logits = tf.matmul(tf_train_dataset, weights) + biases\n loss = tf.reduce_mean(\n tf.nn.softmax_cross_entropy_with_logits(labels=tf_train_labels, logits=logits))\n \n # Optimizer.\n optimizer = tf.train.GradientDescentOptimizer(0.5).minimize(loss)\n \n # Predictions for the training, validation, and test data.\n train_prediction = tf.nn.softmax(logits)\n valid_prediction = tf.nn.softmax(\n tf.matmul(tf_valid_dataset, weights) + biases)\n test_prediction = tf.nn.softmax(tf.matmul(tf_test_dataset, weights) + biases)", "_____no_output_____" ], [ "num_steps = 10000\n\nwith tf.Session(graph=graph) as session:\n tf.global_variables_initializer().run()\n print(\"Initialized\")\n for step in range(num_steps):\n # Pick an offset within the training data, which has been randomized.\n # Note: we could use better randomization across epochs.\n offset = (step * batch_size) % (train_labels.shape[0] - batch_size)\n # Generate a minibatch.\n batch_data = train_dataset[offset:(offset + batch_size), :]\n batch_labels = train_labels[offset:(offset + batch_size), :]\n # Prepare a dictionary telling the session where to feed the minibatch.\n # The key of the dictionary is the placeholder node of the graph to be fed,\n # and the value is the numpy array to feed to it.\n feed_dict = {tf_train_dataset : batch_data, tf_train_labels : batch_labels}\n _, l, predictions = session.run(\n [optimizer, loss, train_prediction], feed_dict=feed_dict)\n if (step % 500 == 0):\n print(\"Minibatch loss at step %d: %f\" % (step, l))\n print(\"Minibatch accuracy: %.1f%%\" % accuracy(predictions, batch_labels))\n print(\"Validation accuracy: %.1f%%\" % accuracy(\n valid_prediction.eval(), valid_labels))\n print(\"Test accuracy: %.1f%%\" % accuracy(test_prediction.eval(), test_labels))\n\n#TODO measure time", "Initialized\nMinibatch loss at step 0: 17.899033\nMinibatch accuracy: 7.8%\nValidation accuracy: 10.0%\nMinibatch loss at step 500: 1.193777\nMinibatch accuracy: 79.7%\nValidation accuracy: 75.0%\nMinibatch loss at step 1000: 1.287950\nMinibatch accuracy: 82.0%\nValidation accuracy: 76.5%\nMinibatch loss at step 1500: 0.656950\nMinibatch accuracy: 81.2%\nValidation accuracy: 77.4%\nMinibatch loss at step 2000: 0.924349\nMinibatch accuracy: 78.9%\nValidation accuracy: 77.3%\nMinibatch loss at step 2500: 1.161976\nMinibatch accuracy: 74.2%\nValidation accuracy: 78.2%\nMinibatch loss at step 3000: 0.851813\nMinibatch accuracy: 80.5%\nValidation accuracy: 78.8%\nMinibatch loss at step 3500: 0.921817\nMinibatch accuracy: 81.2%\nValidation accuracy: 78.6%\nMinibatch loss at step 4000: 0.835667\nMinibatch accuracy: 80.5%\nValidation accuracy: 79.0%\nMinibatch loss at step 4500: 0.808454\nMinibatch accuracy: 79.7%\nValidation accuracy: 79.4%\nMinibatch loss at step 5000: 0.658319\nMinibatch accuracy: 81.2%\nValidation accuracy: 79.5%\nMinibatch loss at step 5500: 0.857714\nMinibatch accuracy: 74.2%\nValidation accuracy: 79.5%\nMinibatch loss at step 6000: 0.938224\nMinibatch accuracy: 76.6%\nValidation accuracy: 79.7%\nMinibatch loss at step 6500: 0.562490\nMinibatch accuracy: 81.2%\nValidation accuracy: 80.0%\nMinibatch loss at step 7000: 0.773726\nMinibatch accuracy: 76.6%\nValidation accuracy: 80.2%\nMinibatch loss at step 7500: 0.964890\nMinibatch accuracy: 78.9%\nValidation accuracy: 80.0%\nMinibatch loss at step 8000: 1.113154\nMinibatch accuracy: 72.7%\nValidation accuracy: 80.1%\nMinibatch loss at step 8500: 0.644020\nMinibatch accuracy: 83.6%\nValidation accuracy: 80.6%\nMinibatch loss at step 9000: 0.761561\nMinibatch accuracy: 82.0%\nValidation accuracy: 80.0%\nMinibatch loss at step 9500: 0.601618\nMinibatch accuracy: 85.2%\nValidation accuracy: 80.9%\nTest accuracy: 86.6%\n" ], [ "# Turn the logistic regression example with SGD into a 1-hidden layer neural network with rectified linear units nn.relu() \n# and 1024 hidden nodes. This model should improve your validation / test accuracy.\n\n#Do TODOs\n", "_____no_output_____" ] ] ]

[ "code" ]

[ [ "code", "code", "code", "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "## Fitting a diagonal covariance Gaussian mixture model to text data\n\nIn a previous assignment, we explored k-means clustering for a high-dimensional Wikipedia dataset. We can also model this data with a mixture of Gaussians, though with increasing dimension we run into two important issues associated with using a full covariance matrix for each component.\n * Computational cost becomes prohibitive in high dimensions: score calculations have complexity cubic in the number of dimensions M if the Gaussian has a full covariance matrix.\n * A model with many parameters require more data: observe that a full covariance matrix for an M-dimensional Gaussian will have M(M+1)/2 parameters to fit. With the number of parameters growing roughly as the square of the dimension, it may quickly become impossible to find a sufficient amount of data to make good inferences.\n\nBoth of these issues are avoided if we require the covariance matrix of each component to be diagonal, as then it has only M parameters to fit and the score computation decomposes into M univariate score calculations. Recall from the lecture that the M-step for the full covariance is:\n\n\\begin{align*}\n\\hat{\\Sigma}_k &= \\frac{1}{N_k^{soft}} \\sum_{i=1}^N r_{ik} (x_i-\\hat{\\mu}_k)(x_i - \\hat{\\mu}_k)^T\n\\end{align*}\n\nNote that this is a square matrix with M rows and M columns, and the above equation implies that the (v, w) element is computed by\n\n\\begin{align*}\n\\hat{\\Sigma}_{k, v, w} &= \\frac{1}{N_k^{soft}} \\sum_{i=1}^N r_{ik} (x_{iv}-\\hat{\\mu}_{kv})(x_{iw} - \\hat{\\mu}_{kw})\n\\end{align*}\n\nWhen we assume that this is a diagonal matrix, then non-diagonal elements are assumed to be zero and we only need to compute each of the M elements along the diagonal independently using the following equation. \n\n\\begin{align*}\n\\hat{\\sigma}^2_{k, v} &= \\hat{\\Sigma}_{k, v, v} \\\\\n&= \\frac{1}{N_k^{soft}} \\sum_{i=1}^N r_{ik} (x_{iv}-\\hat{\\mu}_{kv})^2\n\\end{align*}\n\nIn this section, we will use an EM implementation to fit a Gaussian mixture model with **diagonal** covariances to a subset of the Wikipedia dataset. The implementation uses the above equation to compute each variance term. \n\nWe'll begin by importing the dataset and coming up with a useful representation for each article. After running our algorithm on the data, we will explore the output to see whether we can give a meaningful interpretation to the fitted parameters in our model.", "_____no_output_____" ], [ "**Note to Amazon EC2 users**: To conserve memory, make sure to stop all the other notebooks before running this notebook.", "_____no_output_____" ], [ "## Import necessary packages", "_____no_output_____" ] ], [ [ "from __future__ import print_function # to conform python 2.x print to python 3.x\nimport turicreate", "_____no_output_____" ] ], [ [ "We also have a Python file containing implementations for several functions that will be used during the course of this assignment.", "_____no_output_____" ] ], [ [ "from em_utilities import *", "_____no_output_____" ] ], [ [ "## Load Wikipedia data and extract TF-IDF features", "_____no_output_____" ], [ "Load Wikipedia data and transform each of the first 5000 document into a TF-IDF representation.", "_____no_output_____" ] ], [ [ "wiki = turicreate.SFrame('people_wiki.sframe/').head(5000)\nwiki['tf_idf'] = turicreate.text_analytics.tf_idf(wiki['text'])", "_____no_output_____" ] ], [ [ "Using a utility we provide, we will create a sparse matrix representation of the documents. This is the same utility function you used during the previous assignment on k-means with text data.", "_____no_output_____" ] ], [ [ "wiki = wiki.add_row_number()\ntf_idf, map_word_to_index = sframe_to_scipy(wiki, 'tf_idf')\nmap_index_to_word = dict([[map_word_to_index[i], i] for i in map_word_to_index.keys()])", "_____no_output_____" ] ], [ [ "As in the previous assignment, we will normalize each document's TF-IDF vector to be a unit vector. ", "_____no_output_____" ] ], [ [ "%%time\ntf_idf = normalize(tf_idf)", "CPU times: user 4.56 ms, sys: 733 µs, total: 5.29 ms\nWall time: 5.8 ms\n" ] ], [ [ "We can check that the length (Euclidean norm) of each row is now 1.0, as expected.", "_____no_output_____" ] ], [ [ "for i in range(5):\n doc = tf_idf[i]\n print(np.linalg.norm(doc.todense()))", "1.0\n1.0\n0.9999999999999997\n1.0000000000000007\n1.0\n" ] ], [ [ "## EM in high dimensions\n\nEM for high-dimensional data requires some special treatment:\n * E step and M step must be vectorized as much as possible, as explicit loops are dreadfully slow in Python.\n * All operations must be cast in terms of sparse matrix operations, to take advantage of computational savings enabled by sparsity of data.\n * Initially, some words may be entirely absent from a cluster, causing the M step to produce zero mean and variance for those words. This means any data point with one of those words will have 0 probability of being assigned to that cluster since the cluster allows for no variability (0 variance) around that count being 0 (0 mean). Since there is a small chance for those words to later appear in the cluster, we instead assign a small positive variance (~1e-10). Doing so also prevents numerical overflow.\n \nWe provide the complete implementation for you in the file `em_utilities.py`. For those who are interested, you can read through the code to see how the sparse matrix implementation differs from the previous assignment. \n\nYou are expected to answer some quiz questions using the results of clustering.", "_____no_output_____" ], [ "**Initializing mean parameters using k-means**\n\nRecall from the lectures that EM for Gaussian mixtures is very sensitive to the choice of initial means. With a bad initial set of means, EM may produce clusters that span a large area and are mostly overlapping. To eliminate such bad outcomes, we first produce a suitable set of initial means by using the cluster centers from running k-means. That is, we first run k-means and then take the final set of means from the converged solution as the initial means in our EM algorithm.", "_____no_output_____" ] ], [ [ "%%time \n\nfrom sklearn.cluster import KMeans\n\nnp.random.seed(5)\nnum_clusters = 25\n\n# Use scikit-learn's k-means to simplify workflow\n#kmeans_model = KMeans(n_clusters=num_clusters, n_init=5, max_iter=400, random_state=1, n_jobs=-1) # uncomment to use parallelism -- may break on your installation\nkmeans_model = KMeans(n_clusters=num_clusters, n_init=5, max_iter=400, random_state=1, n_jobs=1)\nkmeans_model.fit(tf_idf)\ncentroids, cluster_assignment = kmeans_model.cluster_centers_, kmeans_model.labels_\n\nmeans = [centroid for centroid in centroids]", "/home/offbeat/Environments/turicreate/lib/python3.7/site-packages/sklearn/cluster/_kmeans.py:974: FutureWarning: 'n_jobs' was deprecated in version 0.23 and will be removed in 0.25.\n \" removed in 0.25.\", FutureWarning)\n" ] ], [ [ "**Initializing cluster weights**\n\nWe will initialize each cluster weight to be the proportion of documents assigned to that cluster by k-means above.", "_____no_output_____" ] ], [ [ "%%time \n\nnum_docs = tf_idf.shape[0]\nweights = []\nfor i in range(num_clusters):\n # Compute the number of data points assigned to cluster i:\n num_assigned = np.sum(cluster_assignment == i) # YOUR CODE HERE\n w = float(num_assigned) / num_docs\n weights.append(w)", "CPU times: user 1.51 ms, sys: 345 µs, total: 1.86 ms\nWall time: 1.22 ms\n" ], [ "np.sum(cluster_assignment == 1)", "_____no_output_____" ] ], [ [ "**Initializing covariances**\n\nTo initialize our covariance parameters, we compute $\\hat{\\sigma}_{k, j}^2 = \\sum_{i=1}^{N}(x_{i,j} - \\hat{\\mu}_{k, j})^2$ for each feature $j$. For features with really tiny variances, we assign 1e-8 instead to prevent numerical instability. We do this computation in a vectorized fashion in the following code block.", "_____no_output_____" ] ], [ [ "covs = []\nfor i in range(num_clusters):\n member_rows = tf_idf[cluster_assignment==i]\n cov = (member_rows.multiply(member_rows) - 2*member_rows.dot(diag(means[i]))).sum(axis=0).A1 / member_rows.shape[0] \\\n + means[i]**2\n cov[cov < 1e-8] = 1e-8\n covs.append(cov)", "_____no_output_____" ] ], [ [ "**Running EM**\n\nNow that we have initialized all of our parameters, run EM.", "_____no_output_____" ] ], [ [ "out = EM_for_high_dimension(tf_idf, means, covs, weights, cov_smoothing=1e-10)", "_____no_output_____" ], [ "out['loglik']", "_____no_output_____" ] ], [ [ "## Interpret clustering results", "_____no_output_____" ], [ "In contrast to k-means, EM is able to explicitly model clusters of varying sizes and proportions. The relative magnitude of variances in the word dimensions tell us much about the nature of the clusters.\n\nWrite yourself a cluster visualizer as follows. Examining each cluster's mean vector, list the 5 words with the largest mean values (5 most common words in the cluster). For each word, also include the associated variance parameter (diagonal element of the covariance matrix). \n\nA sample output may be:\n```\n==========================================================\nCluster 0: Largest mean parameters in cluster \n\nWord Mean Variance \nfootball 1.08e-01 8.64e-03\nseason 5.80e-02 2.93e-03\nclub 4.48e-02 1.99e-03\nleague 3.94e-02 1.08e-03\nplayed 3.83e-02 8.45e-04\n...\n```", "_____no_output_____" ] ], [ [ "# Fill in the blanks\ndef visualize_EM_clusters(tf_idf, means, covs, map_index_to_word):\n print('')\n print('==========================================================')\n \n num_clusters = len(means)\n for c in range(num_clusters):\n print('Cluster {0:d}: Largest mean parameters in cluster '.format(c))\n print('\\n{0: <12}{1: <12}{2: <12}'.format('Word', 'Mean', 'Variance'))\n \n # The k'th element of sorted_word_ids should be the index of the word \n # that has the k'th-largest value in the cluster mean. Hint: Use np.argsort().\n sorted_word_ids = np.argsort(means[c])[::-1] # YOUR CODE HERE\n\n for i in sorted_word_ids[:5]:\n print('{0: <12}{1:<10.2e}{2:10.2e}'.format(map_index_to_word[i], \n means[c][i],\n covs[c][i]))\n print('\\n==========================================================')", "_____no_output_____" ], [ "'''By EM'''\nvisualize_EM_clusters(tf_idf, out['means'], out['covs'], map_index_to_word)", "\n==========================================================\nCluster 0: Largest mean parameters in cluster \n\nWord Mean Variance \nminister 7.57e-02 7.42e-03\nelection 5.89e-02 3.21e-03\nparty 5.89e-02 2.61e-03\nliberal 2.93e-02 4.55e-03\nelected 2.91e-02 8.95e-04\n\n==========================================================\nCluster 1: Largest mean parameters in cluster \n\nWord Mean Variance \nfilm 1.76e-01 6.07e-03\nfilms 5.50e-02 2.97e-03\nfestival 4.66e-02 3.60e-03\nfeature 3.69e-02 1.81e-03\ndirected 3.39e-02 2.22e-03\n\n==========================================================\nCluster 2: Largest mean parameters in cluster \n\nWord Mean Variance \nart 1.26e-01 6.83e-03\nmuseum 5.62e-02 7.27e-03\ngallery 3.65e-02 3.40e-03\nartist 3.61e-02 1.44e-03\ndesign 3.20e-02 4.59e-03\n\n==========================================================\nCluster 3: Largest mean parameters in cluster \n\nWord Mean Variance \nbasketball 1.86e-01 7.78e-03\nnba 1.01e-01 1.22e-02\npoints 6.25e-02 5.92e-03\ncoach 5.57e-02 5.91e-03\nteam 4.68e-02 1.30e-03\n\n==========================================================\nCluster 4: Largest mean parameters in cluster \n\nWord Mean Variance \nhockey 2.45e-01 1.64e-02\nnhl 1.56e-01 1.64e-02\nice 6.40e-02 2.97e-03\nseason 5.05e-02 2.52e-03\nleague 4.31e-02 1.53e-03\n\n==========================================================\nCluster 5: Largest mean parameters in cluster \n\nWord Mean Variance \nrepublican 7.93e-02 5.20e-03\nsenate 5.41e-02 6.28e-03\nhouse 4.64e-02 2.41e-03\ndistrict 4.60e-02 2.37e-03\ndemocratic 4.46e-02 3.02e-03\n\n==========================================================\nCluster 6: Largest mean parameters in cluster \n\nWord Mean Variance \nshe 1.60e-01 4.65e-03\nher 1.00e-01 3.14e-03\nmiss 2.22e-02 7.76e-03\nwomen 1.43e-02 1.36e-03\nwomens 1.21e-02 1.46e-03\n\n==========================================================\nCluster 7: Largest mean parameters in cluster \n\nWord Mean Variance \nchampionships7.78e-02 5.17e-03\nm 4.70e-02 7.58e-03\nolympics 4.69e-02 2.59e-03\nmedal 4.28e-02 2.44e-03\nshe 4.18e-02 5.99e-03\n\n==========================================================\nCluster 8: Largest mean parameters in cluster \n\nWord Mean Variance \nbook 1.45e-02 9.38e-04\npublished 1.23e-02 6.16e-04\nthat 1.10e-02 1.73e-04\nnovel 1.07e-02 1.43e-03\nhe 1.04e-02 6.05e-05\n\n==========================================================\nCluster 9: Largest mean parameters in cluster \n\nWord Mean Variance \nshe 1.37e-01 4.25e-03\nher 8.99e-02 2.74e-03\nactress 7.65e-02 4.29e-03\nfilm 5.98e-02 3.44e-03\ndrama 5.03e-02 6.40e-03\n\n==========================================================\nCluster 10: Largest mean parameters in cluster \n\nWord Mean Variance \nsoccer 1.15e-01 2.86e-02\nchess 4.52e-02 1.66e-02\nteam 4.13e-02 2.15e-03\ncoach 3.09e-02 4.45e-03\nleague 3.07e-02 2.01e-03\n\n==========================================================\nCluster 11: Largest mean parameters in cluster \n\nWord Mean Variance \npresident 2.52e-02 1.29e-03\nchairman 2.44e-02 1.97e-03\ncommittee 2.34e-02 2.38e-03\nserved 2.24e-02 6.99e-04\nexecutive 2.15e-02 1.23e-03\n\n==========================================================\nCluster 12: Largest mean parameters in cluster \n\nWord Mean Variance \nmusic 7.26e-02 3.48e-03\njazz 6.07e-02 1.14e-02\nhong 3.78e-02 9.92e-03\nkong 3.50e-02 8.64e-03\nchinese 3.12e-02 5.33e-03\n\n==========================================================\nCluster 13: Largest mean parameters in cluster \n\nWord Mean Variance \nuniversity 3.47e-02 8.89e-04\nhistory 3.38e-02 2.81e-03\nphilosophy 2.86e-02 5.35e-03\nprofessor 2.74e-02 1.08e-03\nstudies 2.41e-02 1.95e-03\n\n==========================================================\nCluster 14: Largest mean parameters in cluster \n\nWord Mean Variance \ntheatre 4.93e-02 6.17e-03\nactor 3.56e-02 2.91e-03\ntelevision 3.21e-02 1.67e-03\nfilm 2.93e-02 1.16e-03\ncomedy 2.86e-02 3.91e-03\n\n==========================================================\nCluster 15: Largest mean parameters in cluster \n\nWord Mean Variance \nalbum 6.76e-02 4.78e-03\nband 5.35e-02 4.21e-03\nmusic 4.18e-02 1.96e-03\nreleased 3.13e-02 1.11e-03\nsong 2.50e-02 1.81e-03\n\n==========================================================\nCluster 16: Largest mean parameters in cluster \n\nWord Mean Variance \ntour 1.14e-01 1.92e-02\npga 1.08e-01 2.65e-02\nracing 8.45e-02 8.26e-03\nchampionship6.27e-02 4.54e-03\nformula 6.06e-02 1.31e-02\n\n==========================================================\nCluster 17: Largest mean parameters in cluster \n\nWord Mean Variance \nnews 5.76e-02 8.06e-03\nradio 5.18e-02 4.62e-03\nshow 3.75e-02 2.56e-03\nbbc 3.63e-02 7.41e-03\nchef 3.27e-02 1.18e-02\n\n==========================================================\nCluster 18: Largest mean parameters in cluster \n\nWord Mean Variance \nfootball 1.11e-01 5.60e-03\nyards 7.37e-02 1.72e-02\nnfl 6.98e-02 9.15e-03\ncoach 6.74e-02 7.85e-03\nquarterback 4.02e-02 7.16e-03\n\n==========================================================\nCluster 19: Largest mean parameters in cluster \n\nWord Mean Variance \nleague 5.21e-02 3.13e-03\nclub 5.04e-02 2.64e-03\nseason 4.77e-02 2.30e-03\nrugby 4.35e-02 8.18e-03\ncup 4.22e-02 2.46e-03\n\n==========================================================\nCluster 20: Largest mean parameters in cluster \n\nWord Mean Variance \norchestra 1.31e-01 1.06e-02\nmusic 1.23e-01 6.15e-03\nsymphony 8.70e-02 1.08e-02\nconductor 8.16e-02 1.01e-02\nphilharmonic4.96e-02 3.27e-03\n\n==========================================================\nCluster 21: Largest mean parameters in cluster \n\nWord Mean Variance \nlaw 9.52e-02 8.35e-03\ncourt 6.84e-02 5.24e-03\njudge 4.59e-02 4.44e-03\nattorney 3.74e-02 4.30e-03\ndistrict 3.72e-02 4.20e-03\n\n==========================================================\nCluster 22: Largest mean parameters in cluster \n\nWord Mean Variance \nfootball 1.21e-01 6.14e-03\nafl 9.58e-02 1.31e-02\naustralian 7.91e-02 1.58e-03\nclub 5.93e-02 1.76e-03\nseason 5.58e-02 1.83e-03\n\n==========================================================\nCluster 23: Largest mean parameters in cluster \n\nWord Mean Variance \nresearch 5.70e-02 2.68e-03\nscience 3.50e-02 2.95e-03\nuniversity 3.34e-02 7.14e-04\nprofessor 3.20e-02 1.26e-03\nphysics 2.61e-02 5.43e-03\n\n==========================================================\nCluster 24: Largest mean parameters in cluster \n\nWord Mean Variance \nbaseball 1.16e-01 5.57e-03\nleague 1.03e-01 3.63e-03\nmajor 5.09e-02 1.19e-03\ngames 4.66e-02 1.93e-03\nsox 4.55e-02 6.28e-03\n\n==========================================================\n" ] ], [ [ "**Quiz Question**. Select all the topics that have a cluster in the model created above. [multiple choice]\n\n- Baseball\n- Basketball\n- Soccer/Football\n- Music\n- Politics\n- Law\n- Finance", "_____no_output_____" ], [ "## Comparing to random initialization", "_____no_output_____" ], [ "Create variables for randomly initializing the EM algorithm. Complete the following code block.", "_____no_output_____" ] ], [ [ "np.random.seed(5) # See the note below to see why we set seed=5.\nnum_clusters = len(means)\nnum_docs, num_words = tf_idf.shape\n\nrandom_means = []\nrandom_covs = []\nrandom_weights = []\n\nfor k in range(num_clusters):\n \n # Create a numpy array of length num_words with random normally distributed values.\n # Use the standard univariate normal distribution (mean 0, variance 1).\n # YOUR CODE HERE\n mean = np.random.normal(0, 1, size=num_words)\n \n # Create a numpy array of length num_words with random values uniformly distributed between 1 and 5.\n # YOUR CODE HERE\n cov = np.random.uniform(1, 5, size=(num_words))\n\n # Initially give each cluster equal weight.\n # YOUR CODE HERE\n weight = 1\n \n random_means.append(mean)\n random_covs.append(cov)\n random_weights.append(weight)", "_____no_output_____" ] ], [ [ "**Quiz Question**: Try fitting EM with the random initial parameters you created above. (Use `cov_smoothing=1e-5`.) Store the result to `out_random_init`. What is the final loglikelihood that the algorithm converges to?", "_____no_output_____" ] ], [ [ "out_random_init = EM_for_high_dimension(tf_idf, random_means, random_covs, random_weights, cov_smoothing=1e-5)", "_____no_output_____" ], [ "print(\"{:e}\".format(out_random_init['loglik'][-1]))", "2.362457e+09\n" ] ], [ [ "**Quiz Question:** Is the final loglikelihood larger or smaller than the final loglikelihood we obtained above when initializing EM with the results from running k-means?", "_____no_output_____" ] ], [ [ "out['loglik']", "_____no_output_____" ] ], [ [ "**Quiz Question**: For the above model, `out_random_init`, use the `visualize_EM_clusters` method you created above. Are the clusters more or less interpretable than the ones found after initializing using k-means?", "_____no_output_____" ] ], [ [ "# YOUR CODE HERE. Use visualize_EM_clusters, which will require you to pass in tf_idf and map_index_to_word.\nvisualize_EM_clusters(tf_idf, out_random_init['means'], out_random_init['covs'], map_index_to_word)", "\n==========================================================\nCluster 0: Largest mean parameters in cluster \n\nWord Mean Variance \nshe 2.24e-02 3.54e-03\naward 1.53e-02 1.18e-03\nmusic 1.45e-02 1.29e-03\nuniversity 1.43e-02 6.32e-04\nlaw 1.27e-02 2.59e-03\n\n==========================================================\nCluster 1: Largest mean parameters in cluster \n\nWord Mean Variance \nshe 1.99e-02 2.80e-03\nleague 1.91e-02 2.14e-03\nseason 1.59e-02 1.17e-03\nfootball 1.53e-02 2.15e-03\nhe 1.36e-02 1.13e-04\n\n==========================================================\nCluster 2: Largest mean parameters in cluster \n\nWord Mean Variance \nshe 5.65e-02 6.39e-03\nher 5.21e-02 5.62e-03\nmusic 1.13e-02 9.50e-04\nde 1.08e-02 1.92e-03\nopera 1.03e-02 3.26e-03\n\n==========================================================\nCluster 3: Largest mean parameters in cluster \n\nWord Mean Variance \nfilm 2.98e-02 5.79e-03\nhe 1.38e-02 1.07e-04\nhockey 1.33e-02 5.33e-03\nshe 1.31e-02 1.69e-03\nher 1.24e-02 1.20e-03\n\n==========================================================\nCluster 4: Largest mean parameters in cluster \n\nWord Mean Variance \nshe 1.96e-02 3.21e-03\nfilm 1.38e-02 1.72e-03\nhe 1.36e-02 1.14e-04\nher 1.35e-02 1.09e-03\nuniversity 1.31e-02 5.36e-04\n\n==========================================================\nCluster 5: Largest mean parameters in cluster \n\nWord Mean Variance \nshe 2.03e-02 2.65e-03\nher 1.37e-02 1.09e-03\nhe 1.31e-02 1.21e-04\nlaw 9.93e-03 1.96e-03\ncourt 9.54e-03 1.47e-03\n\n==========================================================\nCluster 6: Largest mean parameters in cluster \n\nWord Mean Variance \nshe 2.70e-02 3.62e-03\nher 1.44e-02 1.18e-03\nhe 1.18e-02 1.06e-04\nserved 1.11e-02 3.79e-04\nstate 1.03e-02 5.08e-04\n\n==========================================================\nCluster 7: Largest mean parameters in cluster \n\nWord Mean Variance \nshe 3.41e-02 4.33e-03\nher 2.17e-02 1.89e-03\nmusic 1.73e-02 2.16e-03\nalbum 1.52e-02 2.38e-03\nmarathon 1.25e-02 5.52e-03\n\n==========================================================\nCluster 8: Largest mean parameters in cluster \n\nWord Mean Variance \nleague 1.87e-02 2.01e-03\nshe 1.73e-02 2.67e-03\nhe 1.46e-02 1.17e-04\nseason 1.38e-02 8.80e-04\nplayed 1.35e-02 6.32e-04\n\n==========================================================\nCluster 9: Largest mean parameters in cluster \n\nWord Mean Variance \nshe 3.38e-02 4.81e-03\nher 1.52e-02 1.23e-03\nteam 1.46e-02 8.86e-04\nplayed 1.45e-02 7.99e-04\ncup 1.39e-02 1.44e-03\n\n==========================================================\nCluster 10: Largest mean parameters in cluster \n\nWord Mean Variance \nmusic 1.54e-02 1.51e-03\nyork 1.49e-02 8.13e-04\nhe 1.31e-02 1.01e-04\nuniversity 1.23e-02 5.72e-04\nshe 1.22e-02 2.06e-03\n\n==========================================================\nCluster 11: Largest mean parameters in cluster \n\nWord Mean Variance \nshe 2.07e-02 2.65e-03\nher 1.61e-02 1.67e-03\nfilm 1.55e-02 2.19e-03\nmusic 1.36e-02 1.71e-03\nuniversity 1.36e-02 3.43e-04\n\n==========================================================\nCluster 12: Largest mean parameters in cluster \n\nWord Mean Variance \nshe 2.06e-02 2.98e-03\ntour 1.46e-02 3.64e-03\nhe 1.42e-02 1.22e-04\nher 1.26e-02 1.11e-03\nmusic 1.25e-02 1.40e-03\n\n==========================================================\nCluster 13: Largest mean parameters in cluster \n\nWord Mean Variance \nshe 2.12e-02 2.25e-03\nher 1.65e-02 1.35e-03\nart 1.60e-02 2.76e-03\nnixon 1.50e-02 9.78e-03\nmusic 1.44e-02 1.41e-03\n\n==========================================================\nCluster 14: Largest mean parameters in cluster \n\nWord Mean Variance \nmusic 1.85e-02 2.31e-03\nfilm 1.69e-02 2.75e-03\nhe 1.31e-02 9.45e-05\nresearch 1.08e-02 1.06e-03\nuniversity 1.04e-02 2.96e-04\n\n==========================================================\nCluster 15: Largest mean parameters in cluster \n\nWord Mean Variance \nshe 2.14e-02 3.05e-03\nhe 1.35e-02 1.11e-04\nleague 1.32e-02 1.34e-03\nher 1.30e-02 1.32e-03\nseason 1.22e-02 1.01e-03\n\n==========================================================\nCluster 16: Largest mean parameters in cluster \n\nWord Mean Variance \nshe 2.12e-02 2.73e-03\nher 1.65e-02 1.29e-03\nfilm 1.36e-02 1.26e-03\nhe 1.17e-02 1.01e-04\nshow 1.14e-02 9.72e-04\n\n==========================================================\nCluster 17: Largest mean parameters in cluster \n\nWord Mean Variance \nfilm 1.66e-02 3.18e-03\nmusic 1.28e-02 1.23e-03\nhe 1.18e-02 9.96e-05\nshe 1.11e-02 1.28e-03\nher 1.05e-02 1.17e-03\n\n==========================================================\nCluster 18: Largest mean parameters in cluster \n\nWord Mean Variance \nshe 4.03e-02 5.17e-03\nher 2.28e-02 1.75e-03\nband 1.64e-02 2.43e-03\nmusic 1.24e-02 9.29e-04\nuniversity 1.19e-02 3.84e-04\n\n==========================================================\nCluster 19: Largest mean parameters in cluster \n\nWord Mean Variance \nleague 1.51e-02 1.26e-03\nseason 1.43e-02 1.04e-03\nhe 1.33e-02 9.20e-05\nshe 1.33e-02 1.71e-03\nclub 1.11e-02 8.05e-04\n\n==========================================================\nCluster 20: Largest mean parameters in cluster \n\nWord Mean Variance \nshe 2.63e-02 3.91e-03\nfilm 1.82e-02 1.93e-03\nher 1.45e-02 1.43e-03\nhe 1.26e-02 1.05e-04\nlaw 1.24e-02 2.35e-03\n\n==========================================================\nCluster 21: Largest mean parameters in cluster \n\nWord Mean Variance \nshe 1.86e-02 2.57e-03\nhe 1.39e-02 1.30e-04\nleague 1.37e-02 1.27e-03\nseason 1.22e-02 1.09e-03\nher 1.21e-02 1.09e-03\n\n==========================================================\nCluster 22: Largest mean parameters in cluster \n\nWord Mean Variance \nshe 1.50e-02 2.21e-03\nmusic 1.49e-02 1.63e-03\nhe 1.29e-02 1.08e-04\nparty 1.12e-02 8.99e-04\nher 1.00e-02 1.08e-03\n\n==========================================================\nCluster 23: Largest mean parameters in cluster \n\nWord Mean Variance \nshe 2.27e-02 2.67e-03\nher 1.79e-02 1.67e-03\nmusic 1.62e-02 1.32e-03\nband 1.20e-02 1.10e-03\nyork 1.18e-02 6.68e-04\n\n==========================================================\nCluster 24: Largest mean parameters in cluster \n\nWord Mean Variance \nshe 2.62e-02 3.89e-03\nher 1.77e-02 1.67e-03\nalbum 1.50e-02 2.47e-03\nhe 1.28e-02 1.26e-04\nsoccer 1.16e-02 4.19e-03\n\n==========================================================\n" ] ], [ [ "**Note**: Random initialization may sometimes produce a superior fit than k-means initialization. We do not claim that random initialization is always worse. However, this section does illustrate that random initialization often produces much worse clustering than k-means counterpart. This is the reason why we provide the particular random seed (`np.random.seed(5)`).", "_____no_output_____" ], [ "## Takeaway\n\nIn this assignment we were able to apply the EM algorithm to a mixture of Gaussians model of text data. This was made possible by modifying the model to assume a diagonal covariance for each cluster, and by modifying the implementation to use a sparse matrix representation. In the second part you explored the role of k-means initialization on the convergence of the model as well as the interpretability of the clusters.", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Example notebook for the functions contained in cry_file_readwrite.py", "_____no_output_____" ], [ "### Crystal_input class", "_____no_output_____" ] ], [ [ "from crystal_functions.file_readwrite import Crystal_input", "_____no_output_____" ] ], [ [ "#### Create a crystal input object from blocks", "_____no_output_____" ] ], [ [ "geom_block = ['MGO BULK - GEOMETRY TEST\\n',\n 'CRYSTAL\\n',\n '0 0 0\\n',\n '225\\n',\n '4.217\\n',\n '2\\n',\n '12 0. 0. 0.\\n',\n '8 0.5 0.5 0.5\\n',\n 'END\\n']\nbs_block = ['BASISSET\\n','POB-DZVP\\n']\nfunc_block = ['DFT\\n', 'B3LYP\\n', 'XXLGRID\\n', 'ENDDFT\\n']\nscf_block = [['TOLINTEG\\n', '7 7 7 7 14\\n'],\n ['SHRINK\\n', '12 24\\n'],\n ['MAXCYCLE\\n', '200\\n'],\n ['FMIXING\\n', '70\\n'],\n 'DIIS\\n',\n 'ENDSCF\\n']", "_____no_output_____" ], [ "mgo_input = Crystal_input().from_blocks(geom_block,bs_block,func_block,scf_block)\nmgo_input.geom_block", "_____no_output_____" ] ], [ [ "#### Create a crystal input object from an existing input file", "_____no_output_____" ] ], [ [ "mgo_input = Crystal_input().from_file('data/mgo.d12')\nmgo_input.geom_block", "_____no_output_____" ] ], [ [ "### Crystal_output class", "_____no_output_____" ] ], [ [ "from crystal_functions.file_readwrite import Crystal_output", "_____no_output_____" ] ], [ [ "#### 3D system", "_____no_output_____" ] ], [ [ "mgo_output = Crystal_output().read_cry_output('data/mgo_optgeom.out')\nmgo_output", "_____no_output_____" ] ], [ [ "#### Functions and properties", "_____no_output_____" ] ], [ [ "#Final energy\nprint(\"Final energy = %s eV \\n\" % mgo_output.get_final_energy())\n\n#Fermi energy\nprint(\"Fermi energy = %s eV \\n\" % mgo_output.get_fermi_energy())\n\n#Primitive lattice\nprint(\"Primitive lattice \\n %s \\n\" % mgo_output.get_primitive_lattice())\n\n#Reciprocal lattice\nprint(\"Reciprocal lattice \\n %s \\n\" % mgo_output.get_reciprocal_lattice())\n\n#Band gap\nprint(\"Band gap = %s eV \\n\" % mgo_output.get_band_gap())\n\n#Last geometry\nprint(\"Last geometry = \\n %s \\n\" % mgo_output.get_last_geom())\n\n#Symmetry operators\nprint(\"Symmetry operators = \\n %s \\n\" % mgo_output.get_symm_ops())\n\n#Forces\nprint(\"Forces on cell = \\n %s \\n Forces on atoms = \\n %s \\n Gradient = \\n %s \\n\" % (mgo_output.get_forces(grad=True)[0],\n mgo_output.get_forces(grad=True)[1],\n mgo_output.grad))\n#Scf convergence\nprint(\"Total energy = \\n %s \\n Delta energy = \\n %s \\n\" % (mgo_output.get_scf_convergence()[0],\n mgo_output.get_scf_convergence()[1]))", "Final energy = -7495.341792877063 eV \n\nFermi energy = -4.13671240282 eV \n\nPrimitive lattice \n [[0. 2.1085 2.1085]\n [2.1085 0. 2.1085]\n [2.1085 2.1085 0. ]] \n\nReciprocal lattice \n [[-1.48996571 1.48996571 1.48996571]\n [ 1.48996571 -1.48996571 1.48996571]\n [ 1.48996571 1.48996571 -1.48996571]] \n\nBand gap = 7.1237 eV \n\n" ] ], [ [ "#### 0D system", "_____no_output_____" ] ], [ [ "co_output = Crystal_output().read_cry_output('data/co.out')", "_____no_output_____" ], [ "#Final energy\nprint(\"Final energy = %s eV \\n\" % co_output.get_final_energy())\n\n#Fermi energy\nprint(\"Fermi energy = %s eV \\n\" % co_output.get_fermi_energy())\n\n#Primitive lattice\nprint(\"Primitive lattice \\n %s \\n\" % co_output.get_primitive_lattice())\n\n#Reciprocal lattice\nprint(\"Reciprocal lattice \\n %s \\n\" % co_output.get_reciprocal_lattice())\n\n#Band gap\nprint(\"Band gap = %s eV \\n\" % co_output.get_band_gap())\n\n#Last geometry\nprint(\"Last geometry = \\n %s \\n\" % co_output.get_last_geom())\n\n#Symmetry operators\nprint(\"Symmetry operators = \\n %s \\n\" % co_output.get_symm_ops())\n\n#Forces\nprint(\"Forces on cell = \\n %s \\n Forces on atoms = \\n %s \\n Gradient = \\n %s \\n\" % (co_output.get_forces(grad=True)[0],\n co_output.get_forces(grad=True)[1],\n co_output.grad))\n#Scf convergence\nprint(\"Total energy = \\n %s \\n Delta energy = \\n %s \\n\" % (co_output.get_scf_convergence()[0],\n co_output.get_scf_convergence()[1]))", "Final energy = -3037.8807073392627 eV \n\nFermi energy = -7.247937617924001 eV \n\nWARNING: no lattice vectors found in the output file. lattice = []\nPrimitive lattice \n None \n\nReciprocal lattice \n None \n\nBand gap = 10.9012 eV \n\nLast geometry = \n [[[12.0, 0.0, 0.0], [0.0, 12.0, 0.0], [0.0, 0.0, 12.0]], [6, 8], [[-0.06372620712817, -0.03982887945511, -0.03186310356409], [0.8637262071282, 0.5398288794551, 0.4318631035641]]] \n\nSymmetry operators = \n None \n\nForces on cell = \n [] \n Forces on atoms = \n [[-0.48652346 -0.30407721 -0.24326191]\n [ 0.48652346 0.30407721 0.24326191]] \n Gradient = \n ['0.881299', '0.033031', '0.014444', '0.005748', '0.000292', '0.000266'] \n\nTotal energy = \n [-3050.4186324 -3037.81431056 -3037.86470591 -3037.87933672\n -3037.88070722 -3037.88070734 -3037.88070734] \n Delta energy = \n [-3.04767680e+03 1.25988782e+01 -5.03410900e-02 -1.46397332e-02\n -1.37145456e-03 -1.17281134e-07 -7.48313500e-10] \n\n" ] ], [ [ "### Properties_output class", "_____no_output_____" ] ], [ [ "from crystal_functions.file_readwrite import Properties_output", "_____no_output_____" ] ], [ [ "#### Bands", "_____no_output_____" ] ], [ [ "mgo_bands = Properties_output()\nmgo_bands.read_cry_bands('../examples/data/mgo_BAND_dat.BAND')\nmgo_bands", "_____no_output_____" ] ], [ [ "#### Doss", "_____no_output_____" ] ], [ [ "mgo_doss = Properties_output()\nmgo_doss = mgo_doss.read_cry_doss('data/mgo_DOSS_dat.DOSS')\nmgo_doss", "_____no_output_____" ] ], [ [ "### write_cry_input function", "_____no_output_____" ] ], [ [ "from crystal_functions.file_readwrite import Crystal_input\nfrom crystal_functions.file_readwrite import write_crystal_input", "_____no_output_____" ] ], [ [ "#### Use existing input file", "_____no_output_____" ] ], [ [ "#Read the original input\nmgo_original_input = Crystal_input().from_file('data/mgo.d12')\n\n#Write the input\nnew_input_name = 'data/mgo_from_file.d12'\nwrite_crystal_input(new_input_name,crystal_input=mgo_original_input)", "_____no_output_____" ] ], [ [ "#### Build from blocks", "_____no_output_____" ] ], [ [ "#Define the blocks\ngeom_block = ['MGO BULK - GEOMETRY TEST\\n',\n 'CRYSTAL\\n',\n '0 0 0\\n',\n '225\\n',\n '4.217\\n',\n '2\\n',\n '12 0. 0. 0.\\n',\n '8 0.5 0.5 0.5\\n',\n 'END\\n']\nbs_block = ['12 4\\n',\n '0 0 8 2.0 1.0\\n',\n ' 68370.0 0.0002226\\n',\n ' 9661.0 0.001901\\n',\n ' 2041.0 0.011042\\n',\n ' 529.6 0.05005\\n',\n ' 159.17 0.1690\\n',\n ' 54.71 0.36695\\n',\n ' 21.236 0.4008\\n',\n ' 8.791 0.1487\\n',\n '0 1 5 8.0 1.0\\n',\n ' 143.7 -0.00671 0.00807\\n',\n ' 31.27 -0.07927 0.06401\\n',\n ' 9.661 -0.08088 0.2092\\n',\n ' 3.726 0.2947 0.3460\\n',\n ' 1.598 0.5714 0.3731\\n',\n '0 1 1 2.0 1.0\\n',\n ' 0.688 1.0 1.0\\n',\n '0 1 1 0.0 1.0\\n',\n ' 0.28 1.0 1.0\\n',\n '8 4\\n',\n '0 0 8 2.0 1.0\\n',\n ' 8020.0 0.00108\\n',\n ' 1338.0 0.00804\\n',\n ' 255.4 0.05324\\n',\n ' 69.22 0.1681\\n',\n ' 23.90 0.3581\\n',\n ' 9.264 0.3855\\n',\n ' 3.851 0.1468\\n',\n ' 1.212 0.0728\\n',\n '0 1 4 6.0 1.0\\n',\n ' 49.43 -0.00883 0.00958\\n',\n ' 10.47 -0.0915 0.0696\\n',\n ' 3.235 -0.0402 0.2065\\n',\n ' 1.217 0.379 0.347\\n',\n '0 1 1 0.0 1.0\\n',\n ' 0.4764 1.0 1.0\\n',\n '0 1 1 0.0 1.0\\n',\n ' 0.1802 1.0 1.0\\n',\n '99 0\\n',\n 'ENDBS\\n']\nfunc_block = ['DFT\\n', 'B3LYP\\n', 'XXLGRID\\n', 'ENDDFT\\n']\nscf_block = [['TOLINTEG\\n', '7 7 7 7 14\\n'],\n ['SHRINK\\n', '12 24\\n'],\n ['MAXCYCLE\\n', '200\\n'],\n ['FMIXING\\n', '70\\n'],\n 'DIIS\\n',\n 'ENDSCF\\n']\n\n#Write the input object\nmgo_from_blocks = Crystal_input().from_blocks(geom_block,bs_block,func_block,scf_block)\nnew_input_name = 'data/mgo_from_blocks.d12'\nwrite_crystal_input(new_input_name,mgo_from_blocks)", "_____no_output_____" ], [ "#Write input for external object (ASE or pymatgen)\n#Start from original inout \nmgo_original_input = Crystal_input().from_file('data/mgo.d12')\n\n#Generate the external object\nfrom pymatgen.core import Structure, Lattice \nfrom pymatgen.symmetry.analyzer import SpacegroupAnalyzer\n\nsubstrate = Structure.from_spacegroup(\"Fm-3m\", Lattice.cubic(4.217), [\"Mg\",\"O\"], [[0, 0, 0],[0.5,0.5,0.5]])\nsubstrate_primitive = SpacegroupAnalyzer(substrate).get_primitive_standard_structure() \n\n#Write the input\nnew_input_name = 'data/mgo_external_obj.d12'\nwrite_crystal_input(new_input_name,crystal_input=mgo_original_input,external_obj=substrate_primitive)", "_____no_output_____" ] ], [ [ "### write_cry_properties function", "_____no_output_____" ], [ "#### bands from k points coordinates", "_____no_output_____" ] ], [ [ "from crystal_functions.file_readwrite import Properties_input\nfrom crystal_functions.file_readwrite import write_properties_input\n\n#Create the bands input object\nbands_input = Properties_input()\n#Add the newk block to the input object\nbands_input.make_newk_block(12,24)\n\n#Prepare the band_block\nk_path = [[0,0,0],[0.5,0,0],[0.5,0.5,0.5],[0.25,0,0.5]]\nn_kpoints = 200\nfirst_band = 1\nlast_band = 26\nbands_input.make_bands_block(k_path,n_kpoints,first_band,last_band)\n\n#Write the input\nwrite_properties_input('data/bands_input_1.d3',bands_input)", "_____no_output_____" ], [ "bands_input.property_block", "_____no_output_____" ] ], [ [ "#### bands from pymatgen HighSymmKpath object", "_____no_output_____" ] ], [ [ "from pymatgen.symmetry.bandstructure import HighSymmKpath\nfrom pymatgen.symmetry.analyzer import SpacegroupAnalyzer\n\nfrom crystal_functions.file_readwrite import write_properties_input\nfrom crystal_functions.file_readwrite import Crystal_output\nfrom crystal_functions.file_readwrite import Properties_input\nfrom crystal_functions.convert import cry_out2pmg\n\n\n#Create the bands input object\nbands_input = Properties_input()\n#Add the newk block to the input object\nbands_input.make_newk_block(12,24)\n\n#Read the structure\nmgo = Crystal_output().read_cry_output('data/mgo.out')\nmgo = cry_out2pmg(mgo)\nmgo_prim = SpacegroupAnalyzer(mgo).get_primitive_standard_structure(international_monoclinic=False)\n\n#Obtain the k path object\nk_path = HighSymmKpath(mgo_prim)\n\nn_kpoints = 200\nfirst_band = 1\nlast_band = 26\nbands_input.make_bands_block(k_path,n_kpoints,first_band,last_band)\n\n#Write the input\nwrite_properties_input('data/bands_input_2.d3',bands_input)", "_____no_output_____" ], [ "bands_input.property_block", "_____no_output_____" ] ], [ [ "#### doss", "_____no_output_____" ] ], [ [ "from crystal_functions.file_readwrite import write_properties_input\nfrom crystal_functions.file_readwrite import Properties_input\n\n#Create the doss input object\ndoss_input = Properties_input()\n#Add the newk block to the input object\ndoss_input.make_newk_block(12,24)\n\n#Prepare the doss_block\n\ndoss_input.make_doss_block(n_points=200,e_range=[-5,15],plotting_option=2,poly=12,print_option=1)\n\n#Write the input\nwrite_properties_input('data/doss_input.d3',doss_input)", "_____no_output_____" ], [ "!cat data/doss_input.d3", "NEWK\r\n12 24\r\n1 0\r\nDOSS\r\n0 200 -1 -1 2 12 1\r\n-0.18374651800348382 0.5512395540104514\r\nEND\r\n" ] ], [ [ "#### pdoss (atoms)", "_____no_output_____" ] ], [ [ "from crystal_functions.file_readwrite import write_properties_input\nfrom crystal_functions.file_readwrite import Properties_input\n\n#Create the doss input object\npdoss_input = Properties_input()\n#Add the newk block to the input object\npdoss_input.make_newk_block(12,24)\n\n#Prepare the pdoss_block\nprojections = [[1],[2]]\npdoss_input.make_pdoss_block(projections,proj_type='atom',n_points=200,band_range=[1,26],\n plotting_option=2,poly=12,print_option=1)\n\n#Write the input\nwrite_properties_input('data/pdoss_input.d3',pdoss_input)", "_____no_output_____" ], [ "!cat data/pdoss_input.d3", "NEWK\r\n12 24\r\n1 0\r\nPDOS\r\n2 200 1 26 2 12 1\r\n-1 1\r\n-1 2\r\nEND\r\n" ] ], [ [ "#### pdoss (ao)", "_____no_output_____" ] ], [ [ "from crystal_functions.file_readwrite import write_properties_input\nfrom crystal_functions.file_readwrite import Properties_input\n\n#Create the doss input object\npdoss_input = Properties_input()\n#Add the newk block to the input object\npdoss_input.make_newk_block(12,24)\n\n#Prepare the pdoss_block\nprojections = [[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13],\n [14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26]]\npdoss_input.make_pdoss_block(projections,proj_type='ao',n_points=200,band_range=[1,26],\n plotting_option=2,poly=12,print_option=1)\n\n#Write the input\nwrite_properties_input('data/pdoss_input_ao.d3',pdoss_input)", "_____no_output_____" ], [ "!cat data/pdoss_input_ao.d3", "NEWK\r\n12 24\r\n1 0\r\nPDOS\r\n2 200 1 26 2 12 1\r\n13 1 2 3 4 5 6 7 8 9 10 11 12 13\r\n13 14 15 16 17 18 19 20 21 22 23 24 25 26\r\nEND\r\n" ] ], [ [ "### write_cry_gui function", "_____no_output_____" ], [ "#### from pymatgen structure", "_____no_output_____" ], [ "#### bulk (3D)", "_____no_output_____" ] ], [ [ "from crystal_functions.file_readwrite import Crystal_gui, write_crystal_gui\nfrom crystal_functions.convert import cry_pmg2gui\n\nfrom pymatgen.core import Structure, Lattice\n\n#Generate bulk structure\nsubstrate = Structure.from_spacegroup(\"Fm-3m\", Lattice.cubic(4.217), [\"Mg\",'O'], [[0, 0, 0],[0.5,0.5,0.5]])\n\n#The dimensionality parameter is optional. The function will recognise it is a pymatgen bulk object\n#Create the gui object by converting the pymatgen structure\nmgo_gui = cry_pmg2gui(substrate,dimensionality=2)\n\nwrite_crystal_gui('data/mgo_write_gui.gui',mgo_gui)", "_____no_output_____" ], [ "! cat data/mgo_write_gui.gui", "2 1 1\r\n4.217 0.0 0.0\r\n0.0 4.217 0.0\r\n0.0 0.0 500.0\r\n32\r\n1.0 0.0 0.0\r\n0.0 1.0 0.0\r\n0.0 0.0 1.0\r\n0.0 0.0 0.0\r\n0.0 -1.0 0.0\r\n1.0 0.0 0.0\r\n0.0 0.0 1.0\r\n0.0 0.0 0.0\r\n-1.0 0.0 0.0\r\n0.0 -1.0 0.0\r\n0.0 0.0 1.0\r\n0.0 0.0 0.0\r\n0.0 1.0 0.0\r\n-1.0 0.0 0.0\r\n0.0 0.0 1.0\r\n0.0 0.0 0.0\r\n-1.0 0.0 0.0\r\n0.0 1.0 0.0\r\n0.0 0.0 1.0\r\n0.0 0.0 0.0\r\n0.0 1.0 0.0\r\n1.0 0.0 0.0\r\n0.0 0.0 1.0\r\n0.0 0.0 0.0\r\n1.0 0.0 0.0\r\n0.0 -1.0 0.0\r\n0.0 0.0 1.0\r\n0.0 0.0 0.0\r\n0.0 -1.0 0.0\r\n-1.0 0.0 0.0\r\n0.0 0.0 1.0\r\n0.0 0.0 0.0\r\n-1.0 0.0 0.0\r\n0.0 -1.0 0.0\r\n0.0 0.0 -1.0\r\n0.0 2.1085 0.0\r\n0.0 1.0 0.0\r\n-1.0 0.0 0.0\r\n0.0 0.0 -1.0\r\n0.0 2.1085 0.0\r\n1.0 0.0 0.0\r\n0.0 1.0 0.0\r\n0.0 0.0 -1.0\r\n0.0 2.1085 0.0\r\n0.0 -1.0 0.0\r\n1.0 0.0 0.0\r\n0.0 0.0 -1.0\r\n0.0 2.1085 0.0\r\n1.0 0.0 0.0\r\n0.0 -1.0 0.0\r\n0.0 0.0 -1.0\r\n0.0 2.1085 0.0\r\n0.0 -1.0 0.0\r\n-1.0 0.0 0.0\r\n0.0 0.0 -1.0\r\n0.0 2.1085 0.0\r\n-1.0 0.0 0.0\r\n0.0 1.0 0.0\r\n0.0 0.0 -1.0\r\n0.0 2.1085 0.0\r\n0.0 1.0 0.0\r\n1.0 0.0 0.0\r\n0.0 0.0 -1.0\r\n0.0 2.1085 0.0\r\n-1.0 0.0 0.0\r\n0.0 -1.0 0.0\r\n0.0 0.0 -1.0\r\n2.1085 0.0 0.0\r\n0.0 1.0 0.0\r\n-1.0 0.0 0.0\r\r\n0.0 0.0 -1.0\r\n2.1085 0.0 0.0\r\n1.0 0.0 0.0\r\n0.0 1.0 0.0\r\n0.0 0.0 -1.0\r\n2.1085 0.0 0.0\r\n0.0 -1.0 0.0\r\n1.0 0.0 0.0\r\n0.0 0.0 -1.0\r\n2.1085 0.0 0.0\r\n1.0 0.0 0.0\r\n0.0 -1.0 0.0\r\n0.0 0.0 -1.0\r\n2.1085 0.0 0.0\r\n0.0 -1.0 0.0\r\n-1.0 0.0 0.0\r\n0.0 0.0 -1.0\r\n2.1085 0.0 0.0\r\n-1.0 0.0 0.0\r\n0.0 1.0 0.0\r\n0.0 0.0 -1.0\r\n2.1085 0.0 0.0\r\n0.0 1.0 0.0\r\n1.0 0.0 0.0\r\n0.0 0.0 -1.0\r\n2.1085 0.0 0.0\r\n1.0 0.0 0.0\r\n0.0 1.0 0.0\r\n0.0 0.0 1.0\r\n2.1085 2.1085 0.0\r\n0.0 -1.0 0.0\r\n1.0 0.0 0.0\r\n0.0 0.0 1.0\r\n2.1085 2.1085 0.0\r\n-1.0 0.0 0.0\r\n0.0 -1.0 0.0\r\n0.0 0.0 1.0\r\n2.1085 2.1085 0.0\r\n0.0 1.0 0.0\r\n-1.0 0.0 0.0\r\n0.0 0.0 1.0\r\n2.1085 2.1085 0.0\r\n-1.0 0.0 0.0\r\n0.0 1.0 0.0\r\n0.0 0.0 1.0\r\n2.1085 2.1085 0.0\r\n0.0 1.0 0.0\r\n1.0 0.0 0.0\r\n0.0 0.0 1.0\r\n2.1085 2.1085 0.0\r\n1.0 0.0 0.0\r\n0.0 -1.0 0.0\r\n0.0 0.0 1.0\r\n2.1085 2.1085 0.0\r\n0.0 -1.0 0.0\r\n-1.0 0.0 0.0\r\n0.0 0.0 1.0\r\n2.1085 2.1085 0.0\r\n8\r\n12 0.0 0.0 -1.05425\r\n12 2.1085 2.1085 -1.05425\r\n12 2.1085 0.0 1.05425\r\n12 0.0 2.1085 1.05425\r\n8 2.1085 2.1085 1.05425\r\n8 0.0 0.0 1.05425\r\n8 0.0 2.1085 -1.05425\r\n8 2.1085 0.0 -1.05425\r\n225 32" ] ], [ [ "#### slab (2D)", "_____no_output_____" ] ], [ [ "import sys\nsys.path.insert(1,'../crystal_functions/')\nfrom file_readwrite import *\nfrom convert import *", "_____no_output_____" ], [ "#from crystal_functions.file_readwrite import write_crystal_gui\n\nfrom pymatgen.core import Structure, Lattice\nfrom pymatgen.core.surface import SlabGenerator\n\n#Generate bulk structure\nsubstrate = Structure.from_spacegroup(\"Fm-3m\", Lattice.cubic(4.217), [\"Mg\",'O'], [[0, 0, 0],[0.5,0.5,0.5]])\n#Generate slab\nsubstrate = SlabGenerator(substrate, (1,0,0), 5., 10., center_slab=True).get_slab()\n\n#The dimensionality parameter is optional. \nmgo_slab_gui = cry_pmg2gui(substrate, dimensionality=2)\n\nwrite_crystal_gui('data/mgo_100_write_gui.gui',mgo_slab_gui)", "_____no_output_____" ], [ "! cat data/mgo_100_write_gui.gui", "2 1 1\r\n2.9818692962636706 0.0 1.8258683445725374e-16\r\n-1.8258683445725374e-16 2.9818692962636706 1.8258683445725374e-16\r\n0.0 0.0 500.0\r\n8\r\n1.0 -0.0 0.0\r\n0.0 1.0 0.0\r\n0.0 0.0 1.0\r\n0.0 0.0 0.0\r\n-0.0 -1.0 0.0\r\n1.0 0.0 0.0\r\n0.0 -0.0 1.0\r\n0.0 0.0 0.0\r\n-1.0 0.0 0.0\r\n0.0 -1.0 0.0\r\n-0.0 -0.0 1.0\r\n0.0 0.0 0.0\r\n0.0 1.0 0.0\r\n-1.0 -0.0 0.0\r\n-0.0 0.0 1.0\r\n0.0 0.0 0.0\r\n-1.0 -0.0 0.0\r\n0.0 1.0 0.0\r\n-0.0 -0.0 1.0\r\n0.0 0.0 0.0\r\n-0.0 1.0 0.0\r\n1.0 0.0 0.0\r\n0.0 0.0 1.0\r\n0.0 0.0 0.0\r\n1.0 0.0 0.0\r\n0.0 -1.0 0.0\r\n0.0 -0.0 1.0\r\n0.0 0.0 0.0\r\n0.0 -1.0 0.0\r\n-1.0 -0.0 0.0\r\n-0.0 -0.0 1.0\r\n0.0 0.0 0.0\r\n8\r\n12 0.0 0.0 3.16275\r\n12 1.4909346481318353 1.4909346481318353 1.054250000000001\r\n12 0.0 0.0 -1.0542500000000008\r\n12 1.4909346481318353 1.4909346481318353 -3.16275\r\n8 0.0 0.0 1.0542500000000008\r\n8 1.4909346481318353 1.4909346481318353 3.1627500000000004\r\n8 0.0 0.0 -3.16275\r\n8 1.4909346481318353 1.4909346481318353 -1.0542500000000006\r\n129 8" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown", "markdown", "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Text Data Explanation Benchmarking: Emotion Multiclass Classification", "_____no_output_____" ], [ "This notebook demonstrates how to use the benchmark utility to benchmark the performance of an explainer for text data. In this demo, we showcase explanation performance for partition explainer on an Emotion Multiclass Classification model. The metrics used to evaluate are \"keep positive\" and \"keep negative\". The masker used is Text Masker.\n\nThe new benchmark utility uses the new API with MaskedModel as wrapper around user-imported model and evaluates masked values of inputs.", "_____no_output_____" ] ], [ [ "import copy\nimport pandas as pd\nimport numpy as np\nimport matplotlib.pyplot as plt\nfrom transformers import AutoTokenizer, AutoModelForSequenceClassification\nimport shap.benchmark as benchmark\nimport shap\nimport scipy as sp\nimport nlp\nimport torch\n\npd.set_option('display.max_columns', None)\npd.set_option('display.max_rows', None)\npd.set_option('max_colwidth', None)", "_____no_output_____" ] ], [ [ "### Load Data and Model", "_____no_output_____" ] ], [ [ "train, test = nlp.load_dataset(\"emotion\", split = [\"train\", \"test\"])\n\ndata={'text':train['text'],\n 'emotion':train['label']}\n \ndata = pd.DataFrame(data)", "Using custom data configuration default\n" ], [ "tokenizer = AutoTokenizer.from_pretrained(\"nateraw/bert-base-uncased-emotion\",use_fast=True)\nmodel = AutoModelForSequenceClassification.from_pretrained(\"nateraw/bert-base-uncased-emotion\")", "_____no_output_____" ] ], [ [ "### Class Label Mapping", "_____no_output_____" ] ], [ [ "# set mapping between label and id\nid2label = model.config.id2label\nlabel2id = model.config.label2id\nlabels = sorted(label2id, key=label2id.get)", "_____no_output_____" ] ], [ [ "### Define Score Function", "_____no_output_____" ] ], [ [ "def f(x):\n tv = torch.tensor([tokenizer.encode(v, padding='max_length', max_length=128,truncation=True) for v in x])\n attention_mask = (tv!=0).type(torch.int64)\n outputs = model(tv,attention_mask=attention_mask)[0].detach().numpy()\n scores = (np.exp(outputs).T / np.exp(outputs).sum(-1)).T\n val = sp.special.logit(scores) \n return val", "_____no_output_____" ] ], [ [ "### Create Explainer Object", "_____no_output_____" ] ], [ [ "explainer = shap.Explainer(f,tokenizer,output_names=labels)", "explainers.Partition is still in an alpha state, so use with caution...\n" ] ], [ [ "### Run SHAP Explanation", "_____no_output_____" ] ], [ [ "shap_values = explainer(data['text'][0:20])", "_____no_output_____" ] ], [ [ "### Define Metrics (Sort Order & Perturbation Method)", "_____no_output_____" ] ], [ [ "sort_order = 'positive'\nperturbation = 'keep'", "_____no_output_____" ] ], [ [ "### Benchmark Explainer", "_____no_output_____" ] ], [ [ "sequential_perturbation = benchmark.perturbation.SequentialPerturbation(explainer.model, explainer.masker, sort_order, perturbation)\nxs, ys, auc = sequential_perturbation.model_score(shap_values, data['text'][0:20])\nsequential_perturbation.plot(xs, ys, auc)", "_____no_output_____" ], [ "sort_order = 'negative'\nperturbation = 'keep'", "_____no_output_____" ], [ "sequential_perturbation = benchmark.perturbation.SequentialPerturbation(explainer.model, explainer.masker, sort_order, perturbation)\nxs, ys, auc = sequential_perturbation.model_score(shap_values, data['text'][0:20])\nsequential_perturbation.plot(xs, ys, auc)", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code", "code" ] ]

[ "CC0-1.0" ]

[ "CC0-1.0" ]

[ "CC0-1.0" ]

[ [ [ "# Mapas Autorganizados (SOM)", "_____no_output_____" ], [ "<img src=\"../img/som_1.jpg\" width=\"500\">", "_____no_output_____" ], [ "Se toma un conjunto de datos multidimensional los cuales se pueden mostrar como un mapa bidimensional. El objetivo del **SOM** es reducir las columnas y quedarnos solo con las que nos aporten más info. O lo que es lo mismo, reducir lo máximo la dimensión de la entrada de datos.", "_____no_output_____" ], [ "## Ejemplo", "_____no_output_____" ], [ "<img src=\"../img/som_2.jpg\" heigh=\"700\">", "_____no_output_____" ], [ "En el mapa **SOM** central nos indica la prosperidad o pobreza de los paises agrupados en colores, usando como inputs varios datos de éstos.", "_____no_output_____" ], [ "## Cómo aprenden los SOM:", "_____no_output_____" ], [ "Abajo podemos ver 3 columnas de datos de entrada ($X_{1},X_{2},X_{3}$) convertido en un **SOM** en 2-D.", "_____no_output_____" ], [ "<img src=\"../img/som_3.jpg\" width=\"400\">", "_____no_output_____" ], [ "Abajo tenemos la misma red neuronal pero puesto los nodos de salida en una sola columna. Cada Nodo le llegan 3 sinapsis-enlace de los nodos de entrada, cada nodo tiene un peso asociado para cada enlace $W_{i,j}$.\n\nEn las redes neurnales artificiales los pesos se usaban para multiplicar la entrada del peso de ese nodo para luego sumarlo, agruparlo o lo que fuese en la neurona final y aplicarle una función de activación, en cambio en los mapas autorganizados **SOM** no hay función de activación, los pesos son una característica del propio nodo y por lo tanto dichos pesos no se colocan en las sinápsis sino en el propio nodo (ahora los nodos tienen una serie de coordenadas dados por los pesos).", "_____no_output_____" ], [ "<img src=\"../img/som_4.jpg\" width=\"700\">", "_____no_output_____" ], [ "El objetivo es descubrir el nodo que más cercano está a cada uno de los nodos de entrada, para ello calculamos su distancia (por ejmplo la Euclidea). En el caso de arriba observamos que el nodo3 es el más cercano.", "_____no_output_____" ], [ "<img src=\"../img/som_6.jpg\" width=\"500\">", "_____no_output_____" ], [ "<img src=\"../img/som_5.jpg\" width=\"500\">", "_____no_output_____" ], [ "Si pasamos los nodos a un plano bidimensional y si nos imaginamos que el nodo verde (figura **1**) es el nodo más cercano a las variables input $X_j$, lo que pasará después es que el **SOM** actualizará los pesos para que cada uno de ellos distribuya a las observaciones.\n\nSi nos fijamos en el dibujo inferior, lo que vemos en términos visuales descrito arriba es que el punto más cercano \"estirará\" del mapa hasta él para que se acerque todavía más en las siguientes iteraciones. En definitiva, para cada uno de los nodos, va a intentar cambiar las coordenadas(pesos) de dicho nodos para que cada vez se parezcan más a las observaciones.\n\nA continuación lo que se hará será tomar del punto en cuestión(nodo verde en **1**) un radio que hará que los nodos que se encuentren dentro de esta zona amarilla se actualizarán sus coordenadas-pesos, siendo los nodos más cercanos al verde los que más se actualicen y viceversa.\n\nLuego se procederá igual, por ejemplo con el nodo azúl que se parezca más a las variables de entrada (ver **2**), con su propio radio de actuación y hará lo mismo. En el momento que un nodo entra en conflicto con el azúl o el verde, obviamente será atraído más por el nodo más cercano (nos fijamos en la figura **3**). Si nos fijamos en la figura **4** vemos que para un nodo intermedio entre el azúl y verde, al estar en medio tenría un color intermedio, el turquesa. Y para los que están cerca del verde tendrían un colo casi verde del todo, con muy poco de azúl.\n\nObteniendo así finalmente un mapa de colores.", "_____no_output_____" ], [ "<img src=\"../img/som_7.jpg\" width=\"400\">", "_____no_output_____" ], [ "<img src=\"../img/som_8.jpg\" width=\"400\">", "_____no_output_____" ], [ "Si tenemos más unidades de mayor coincidencia con las variables de entrada pasará lo mismo obteniendo el mapa de color.\n\nEn cada iteración (epoch) los radios de acción amarillos se irán haciendo más pequeños. EL proceso se vuelve más y más preciso.", "_____no_output_____" ], [ "## Importante recordar:\n- Los SOMs retienen la topología del conjunto de entrada.\n- Los SOMs revelan correlaciones que no se identifican fácilmente.\n- Los SOMs clasifican datos sin supervisión alguna.\n- No hay vector de variable target -> no hay propagación hacia atrás.\n- No hay conexiones laterales entre los nodos.", "_____no_output_____" ], [ "[Ejemplo de SOM](http://ai-junkie.com/ann/som/som1.html) con programa descargable *.exe*", "_____no_output_____" ], [ "## Cómo leer un mapa autorganizado SOM avanzado:\n\nEjemplo de SOM de patrones en el voto de 535 senadores nortaméricanos\n<img src=\"../img/som_9.jpg\" width=\"600\">", "_____no_output_____" ], [ "# Pasos para entrenar un SOM:\n\n- <span style='color:#288c17'> <b>PASO 1:</span> Empezamos con un dataset compuesto de *n_features* variables independientes.\n\n- <span style='color:#288c17'> <b>PASO 2:</span> Preparamos una parrilla compuesta de nodos, cada uno con un vector de pesos de *n_features* elementos.\n\n- <span style='color:#288c17'> <b>PASO 3:</span> Aleatoriamente inicializamos valores del vector de pesos a números pequeños cercanos a $0$ (pero no $0$).\n\n- <span style='color:#288c17'> <b>PASO 4:</span> Seleccionar una observación aleatoria del dataser.\n\n- <span style='color:#288c17'> <b>PASO 5:</span> Calcular la distancia Euclídea desde dicho puntos a las diferentes neuronas de la red.\n\n- <span style='color:#288c17'> <b>PASO 6:</span> Seleccionar la neurona con la menor distancia al punto. Dicha neurona es el nodo ganador.\n\n- <span style='color:#288c17'> <b>PASO 7:</span> Actualizar lso epsos del nodo ganador para moverlo más cerca dle punto.\n\n- <span style='color:#288c17'> <b>PASO 8:</span> Utilizar una función Gaussiana al vecindario del punto de medie el nodo ganador y actualizar los pesos de los vecinos para moverlos más cerca del punto. El radio de los vecinos afectados es la desviación típica de la Gaussiana.\n\n- <span style='color:#288c17'> <b>PASO 9:</span> Repetir los pasos <span style='color:#288c17'> <b>1</span> a <span style='color:#288c17'> <b>5</span> y actualizar los pesos después de cada observación (*Reinforcement Learning*) o después de un conjunto de observaciones (*Batch Learning*), hasta que la red neuronal converja en un punto donde los vecindarios no cambien.", "_____no_output_____" ] ] ]

[ "markdown" ]

[ [ "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown", "markdown" ] ]

[ "CC0-1.0" ]

[ "CC0-1.0" ]

[ "CC0-1.0" ]

[ [ [ "**Download** (right-click, save target as ...) this page as a jupyterlab notebook [Lab 30](http://54.243.252.9/engr-1330-webroot/8-Labs/Lab30/Lab30.ipynb)\n\n___", "_____no_output_____" ], [ "# <font color=darkblue>Laboratory 30: Logarithmic and Power-Law Models </font>\n\nLAST NAME, FIRST NAME\n\nR00000000\n\nENGR 1330 Laboratory 30 - In Lab\n\n___\n", "_____no_output_____" ], [ "## Background\n\nThe last couple lessons contain the needed logic, so just apply to the problem below\n\n## Exercise 1\n\nUsing methods in the lesson, fit an appropriate data model to the voltage versus time data below. Use the data model to determine\n\n1. What voltage do you expect after 1.5 seconds?\n2. What voltage do you expect after 15.0 seconds?\n3. How long will it take for the voltage to decline to 1.5 volts?\n4. How long will it take for the voltage to decline to 0.17 volts?\n\n|Time(sec)|Voltage(V)|\n|---:|---:|\n|0|9.8|\n|1|5.9|\n|2|3.9|\n|3|2.1|\n|4|1.0|\n|5|0.8|\n|6|0.6|\n|7|0.4|\n|8|0.3|\n|9|0.2|\n|10|0.1|\n", "_____no_output_____" ] ], [ [ "# Load the necessary packages\n# build the data lists\n# build a dataframe\n# Initialise and fit regression model using `statsmodels`\n# Predict values\n# Plot regression against actual data\n# Is the model \"good\" if not try a different data model (linear,exponential,power law,logarithmic,polynomial)\n\n# with your \"good\" model answer the questions", "_____no_output_____" ] ] ]

[ "markdown", "code" ]

[ [ "markdown", "markdown", "markdown" ], [ "code" ] ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ "BSD-3-Clause" ]

[ [ [ "# Unsupervised outliers detection (event detection)", "_____no_output_____" ] ], [ [ "import drama as drm\nimport numpy as np\nimport matplotlib.pylab as plt\nfrom matplotlib import gridspec\n\nfrom drama.outlier_finder import grid_run_drama\n\nfrom keras.datasets import mnist\n\n%matplotlib inline\n\nn_try = 5", "Using TensorFlow backend.\n" ], [ "# MNIST dataset\n(x_train, y_train), (x_test, y_test) = mnist.load_data()\n\nimage_size = x_train.shape[1]\nx_train = x_train.astype('float32') / 255\nx_test = x_test.astype('float32') / 255\nprint(x_train.shape, y_train.shape, x_test.shape, y_test.shape)", "(60000, 28, 28) (60000,) (10000, 28, 28) (10000,)\n" ], [ "inlier_labels = [2,3,4,6,8]\noutlier_labels = [5,0]\nn_inliers = 500\nn_outliers = 10\n\nX = []\ny = []\nfor i in inlier_labels:\n filt = y_train==i\n ns = np.sum(filt)\n X.extend(x_train[filt][:n_inliers])\n y.extend(n_inliers*[0])\n \nfor i in outlier_labels:\n filt = y_train==i\n ns = np.sum(filt)\n X.extend(x_train[filt][:n_outliers])\n y.extend(n_outliers*[1])\n \nX = np.array(X)\ny = np.array(y)", "_____no_output_____" ], [ "X.shape,y.shape", "_____no_output_____" ], [ "X = np.reshape(X, [-1, image_size*image_size])", "_____no_output_____" ], [ "lof_all = np.zeros((n_try,3))\nifr_all = np.zeros((n_try,3))\ndf = drm.sk_check(X.reshape(-1,784),X.reshape(-1,784),y,[1])\nfor i in range(n_try):\n for j,scr in enumerate(['AUC','MCC','RWS']):\n lof_all[i,j] = df[scr][0]\n ifr_all[i,j] = df[scr][1]", "/home/gf/packages/anaconda3/lib/python3.6/site-packages/sklearn/ensemble/iforest.py:223: FutureWarning: behaviour=\"old\" is deprecated and will be removed in version 0.22. Please use behaviour=\"new\", which makes the decision_function change to match other anomaly detection algorithm API.\n FutureWarning)\n" ], [ "df", "_____no_output_____" ] ], [ [ "# Outlier detection", "_____no_output_____" ] ], [ [ "metrics = ['cityblock', 'L2', 'L4', 'braycurtis', 'canberra', 'chebyshev',\n 'correlation', 'mahalanobis', 'wL2', 'wL4']\ndrt_list = ['DAE1D', 'DVAE1D']\n\nresult = []\nfor i in range(n_try):\n# auc,mcc,rws,conf = grid_run_drama(X,y)\n auc,mcc,rws,conf = grid_run_drama(X,y,\n drt_list=drt_list,\n metrics=metrics,\n n_split=2)\n \n arr = np.stack([auc,mcc,rws],axis=-1)\n result.append(arr)\nresult = np.array(result)\n\ndrts = np.unique(conf[:,1])\nmetrs = np.unique(conf[:,2])\n\nres = result.reshape(n_try,len(drt_list),len(metrics),-1)\n\ndrm.plot_table(np.mean(res,axis=0),drts,metrs)\n\nauc = np.sum((res[:, :, :, 0].T>lof_all[:, 0]) & (res[:, :, :, 0].T>ifr_all[:, 0]),axis=-1).T\nmcc = np.sum((res[:, :, :, 1].T>lof_all[:, 1]) & (res[:, :, :, 1].T>ifr_all[:, 1]),axis=-1).T\nrws = np.sum((res[:, :, :, 2].T>lof_all[:, 2]) & (res[:, :, :, 2].T>ifr_all[:, 2]),axis=-1).T\n\nfig = plt.figure(figsize=(20,10))\nplt.clf()\nax = fig.add_subplot(111)\nax.set_aspect('auto')\nax.imshow(auc, cmap=plt.cm.jet,interpolation='nearest')\n\nwidth, height = auc.shape\n\nfor x in range(width):\n for y in range(height):\n ax.annotate('AUC: {:d}\\n MCC: {:d}\\n RWS: {:d}'.format(auc[x][y],mcc[x][y],rws[x][y]), xy=(y, x), \n horizontalalignment='center',\n verticalalignment='center',fontsize=18);\n\nplt.xticks(range(len(metrs)),metrs,fontsize=15)\nplt.yticks(range(len(drts)), drts,fontsize=15)\n\nplt.title('Number of successes (LOF and i-forest) out of 20 data set',fontsize=25)\nplt.annotate('** Colors depend on AUC.', (0,0), (0, -30), xycoords='axes fraction',\n textcoords='offset points', va='top',fontsize=15)\n\n# plt.savefig('AND_success.jpg',dpi=150,bbox_inches='tight')", "Unsupervised outlier detection mode.\n" ], [ "inlier_labels = [1,2,3,4]\noutlier_labels = [5]\nn_inliers = 1000\nn_outliers = 30\n\nX = []\ny = []\nfor i in inlier_labels:\n filt = y_train==i\n ns = np.sum(filt)\n X.extend(x_train[filt][:n_inliers])\n y.extend(n_inliers*[0])\n \nfor i in outlier_labels:\n filt = y_train==i\n ns = np.sum(filt)\n X.extend(x_train[filt][:n_outliers])\n y.extend(n_outliers*[1])\n \nX = np.array(X)\ny = np.array(y)\nX = np.reshape(X, [-1, image_size*image_size, 1])", "_____no_output_____" ], [ "metrics = ['cityblock', 'L2', 'L4', 'braycurtis', 'canberra', 'chebyshev',\n 'correlation', 'mahalanobis', 'wL2', 'wL4']\ndrt_list = ['CAE1D', 'CVAE1D']\n\nresult = []\nfor i in range(n_try):\n# auc,mcc,rws,conf = grid_run_drama(X,y)\n auc,mcc,rws,conf = grid_run_drama(X,y,\n drt_list=drt_list,\n metrics=metrics,\n n_split=2)\n \n arr = np.stack([auc,mcc,rws],axis=-1)\n result.append(arr)\nresult = np.array(result)\n\ndrts = np.unique(conf[:,1])\nmetrs = np.unique(conf[:,2])\n\nres = result.reshape(n_try,len(drt_list),len(metrics),-1)\n\ndrm.plot_table(np.mean(res,axis=0),drts,metrs)\n\nauc = np.sum((res[:, :, :, 0].T>lof_all[:, 0]) & (res[:, :, :, 0].T>ifr_all[:, 0]),axis=-1).T\nmcc = np.sum((res[:, :, :, 1].T>lof_all[:, 1]) & (res[:, :, :, 1].T>ifr_all[:, 1]),axis=-1).T\nrws = np.sum((res[:, :, :, 2].T>lof_all[:, 2]) & (res[:, :, :, 2].T>ifr_all[:, 2]),axis=-1).T\n\nfig = plt.figure(figsize=(20,10))\nplt.clf()\nax = fig.add_subplot(111)\nax.set_aspect('auto')\nax.imshow(auc, cmap=plt.cm.jet,interpolation='nearest')\n\nwidth, height = auc.shape\n\nfor x in range(width):\n for y in range(height):\n ax.annotate('AUC: {:d}\\n MCC: {:d}\\n RWS: {:d}'.format(auc[x][y],mcc[x][y],rws[x][y]), xy=(y, x), \n horizontalalignment='center',\n verticalalignment='center',fontsize=18);\n\nplt.xticks(range(len(metrs)),metrs,fontsize=15)\nplt.yticks(range(len(drts)), drts,fontsize=15)\n\nplt.title('Number of successes (LOF and i-forest) out of 20 data set',fontsize=25)\nplt.annotate('** Colors depend on AUC.', (0,0), (0, -30), xycoords='axes fraction',\n textcoords='offset points', va='top',fontsize=15)\n\n# plt.savefig('AND_success.jpg',dpi=150,bbox_inches='tight')", "Unsupervised outlier detection mode.\n" ], [ "inlier_labels = [1,2,3,4]\noutlier_labels = [5]\nn_inliers = 1000\nn_outliers = 30\n\nX = []\ny = []\nfor i in inlier_labels:\n filt = y_train==i\n ns = np.sum(filt)\n X.extend(x_train[filt][:n_inliers])\n y.extend(n_inliers*[0])\n \nfor i in outlier_labels:\n filt = y_train==i\n ns = np.sum(filt)\n X.extend(x_train[filt][:n_outliers])\n y.extend(n_outliers*[1])\n \nX = np.array(X)\ny = np.array(y)\nX = np.reshape(X, [-1, image_size, image_size, 1])", "_____no_output_____" ], [ "metrics = ['cityblock', 'L2', 'L4', 'braycurtis', 'canberra', 'chebyshev',\n 'correlation', 'mahalanobis', 'wL2', 'wL4']\ndrt_list = ['CAE2D', 'CVAE2D']\n\nresult = []\nfor i in range(n_try):\n# auc,mcc,rws,conf = grid_run_drama(X,y)\n auc,mcc,rws,conf = grid_run_drama(X,y,\n drt_list=drt_list,\n metrics=metrics,\n n_split=2)\n \n arr = np.stack([auc,mcc,rws],axis=-1)\n result.append(arr)\nresult = np.array(result)\n\ndrts = np.unique(conf[:,1])\nmetrs = np.unique(conf[:,2])\n\nres = result.reshape(n_try,len(drt_list),len(metrics),-1)\n\ndrm.plot_table(np.mean(res,axis=0),drts,metrs)\n\nauc = np.sum((res[:, :, :, 0].T>lof_all[:, 0]) & (res[:, :, :, 0].T>ifr_all[:, 0]),axis=-1).T\nmcc = np.sum((res[:, :, :, 1].T>lof_all[:, 1]) & (res[:, :, :, 1].T>ifr_all[:, 1]),axis=-1).T\nrws = np.sum((res[:, :, :, 2].T>lof_all[:, 2]) & (res[:, :, :, 2].T>ifr_all[:, 2]),axis=-1).T\n\nfig = plt.figure(figsize=(20,10))\nplt.clf()\nax = fig.add_subplot(111)\nax.set_aspect('auto')\nax.imshow(auc, cmap=plt.cm.jet,interpolation='nearest')\n\nwidth, height = auc.shape\n\nfor x in range(width):\n for y in range(height):\n ax.annotate('AUC: {:d}\\n MCC: {:d}\\n RWS: {:d}'.format(auc[x][y],mcc[x][y],rws[x][y]), xy=(y, x), \n horizontalalignment='center',\n verticalalignment='center',fontsize=18);\n\nplt.xticks(range(len(metrs)),metrs,fontsize=15)\nplt.yticks(range(len(drts)), drts,fontsize=15)\n\nplt.title('Number of successes (LOF and i-forest) out of 20 data set',fontsize=25)\nplt.annotate('** Colors depend on AUC.', (0,0), (0, -30), xycoords='axes fraction',\n textcoords='offset points', va='top',fontsize=15)\n\n# plt.savefig('AND_success.jpg',dpi=150,bbox_inches='tight')", "Unsupervised outlier detection mode.\n" ] ] ]

[ "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Ensemble Learning\n### Advantages of CARTs\n- simple to understand, interpret, and use\n- flexibility: ability to describe non-linear dependencies\n- preprocessing: no need to standardize or normalize features\n\n### Limitations of CARTS\n- classification: can only produce orthogonal decision boundaries\n- sensitive to small variations in the training set\n- high variance: unconstrained CARTs may overfit the training set\n\n### Solution: Ensemble learning\n<b>STEPS:</b>\n- train different models on the same dataset\n- let each model make its predictions\n- meta-model: aggregates predictions of individual models\n- final prediction: more robust and less prone to errors\n\n<img src='images/ensemble_learning.png'>\n<img src='images/hard_voting.png'>", "_____no_output_____" ] ], [ [ "# import the packages\nimport numpy as np\nimport pandas as pd\nimport seaborn as sns\nimport matplotlib.pyplot as plt\n\nfrom sklearn.linear_model import LogisticRegression\nfrom sklearn.tree import DecisionTreeClassifier\nfrom sklearn.neighbors import KNeighborsClassifier\nfrom sklearn.ensemble import VotingClassifier\n\nfrom sklearn.model_selection import train_test_split, cross_val_score\nfrom sklearn.metrics import accuracy_score\nfrom sklearn.datasets import load_breast_cancer\n\nsns.set()\n%matplotlib inline", "_____no_output_____" ], [ "# voting classifier\ndata = load_breast_cancer()\nbreast_cancer = pd.DataFrame(data.data, columns=data.feature_names)\nbreast_cancer.head(3)", "_____no_output_____" ], [ "X, y = breast_cancer, data.target\nX_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.3, stratify=y)", "_____no_output_____" ], [ "# instantiate individual classifiers\nlogreg = LogisticRegression()\nknn = KNeighborsClassifier()\ndt = DecisionTreeClassifier()\n\nclassifiers = [\n ('Logistic Regression', logreg),\n ('K Nearest Neighbors', knn),\n ('Classification Tree', dt)\n]\n\n# iterate over the defined list of tuples containing the classifiers\nfor clf_name, clf in classifiers:\n # fit the model \n clf.fit(X_train, y_train)\n \n # predict the labels of the test set\n y_pred = clf.predict(X_test)\n \n # print the accuracy score \n print(f'Accuracy score of {clf_name}: {round(accuracy_score(y_test, y_pred), 3)}')\n", "Accuracy score of Logistic Regression: 0.942\nAccuracy score of K Nearest Neighbors: 0.918\nAccuracy score of Classification Tree: 0.93\n" ], [ "# instantiate a voting classifier\nvoting_clf = VotingClassifier(estimators=classifiers)\n\n# fit the training set and predict the test set\nvoting_clf.fit(X_train, y_train)\ny_pred = voting_clf.predict(X_test)\n\n# show the results\nprint(f'Accuracy score of the Voting Classifier: {round(accuracy_score(y_test, y_pred), 3)}')", "Accuracy score of the Voting Classifier: 0.942\n" ] ] ]

[ "markdown", "code" ]

[ [ "markdown" ], [ "code", "code", "code", "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Further information\n\n## What are the differences between recursion and iteration?\n\nWhen giving instructions to a computer it is possible to use recursion to\ndirectly implement a common mathematical definition. For example consider the\nfollowing sequence:\n\n$$\n \\left\\{\\begin{array}{l}\n a_1 = 1\\\\\n a_{n + 1}= 3a_n, n > 1\n \\end{array}\\right.\n$$\n\nWe can define this in Python as follows:", "_____no_output_____" ] ], [ [ "def generate_sequence(n):\n \"\"\"\n Generate the sequence defined by:\n\n a_1 = 1\n a_n = 3 a_{n - 1}\n\n This is done using recursion.\n \"\"\"\n if n == 1:\n return 1\n return 3 * generate_sequence(n - 1)", "_____no_output_____" ] ], [ [ "The first 6 terms:", "_____no_output_____" ] ], [ [ "[generate_sequence(n) for n in range(1, 7)]", "_____no_output_____" ] ], [ [ "We note that in this case this corresponds to powers of $3$, and indeed we can\nprove that: $a_n = 3 ^ {n - 1}$. We will not carry out the proof here but one\napproach to doing it would be to use proof by induction which is closely related\nto recursive functions.\n\nWe can write a different python function that uses this formulae. This is called\n**iteration**:", "_____no_output_____" ] ], [ [ "def calculate_sequence(n):\n \"\"\"\n Calculate the nth term of the sequence defined by:\n\n a_1 = 1\n a_n = 3 a_{n - 1}\n\n This is done using iteration using the direct formula:\n\n a_n = 3 ^ n\n \"\"\"\n return 3 ** (n - 1)", "_____no_output_____" ], [ "[calculate_sequence(n) for n in range(1, 7)]", "_____no_output_____" ] ], [ [ "We can in fact use a Jupyter [magic\ncommand](https://ipython.readthedocs.io/en/stable/interactive/magics.html) to\ntime the run time of a command. It is clear that recursion is slower.", "_____no_output_____" ] ], [ [ "%timeit [generate_sequence(n) for n in range(1, 25)]", "28.7 µs ± 831 ns per loop (mean ± std. dev. of 7 runs, 10000 loops each)\n" ], [ "%timeit [calculate_sequence(n) for n in range(1, 25)]", "7.63 µs ± 89.5 ns per loop (mean ± std. dev. of 7 runs, 100000 loops each)\n" ] ], [ [ "In practice:\n\n- Using recursion is powerful as it can be used to directly implement recursive\n definitions.\n- Using iteration is more computationally efficient but it is not always\n straightforward to obtain an iterative formula.\n\n\n(what_is_caching)=\n## What is caching\n\nOne of the reasons that recursion is computationally inefficient is that it\nalways has to recalculate previously calculated values.\n\nFor example:\n\n\n$$\n \\left\\{\\begin{array}{l}\n a_1 = 1\\\\\n a_{n + 1}= 3a_n, n > 1\n \\end{array}\\right.\n$$\n\nOne way to overcome this is to use caching which means that when a function is\ncalled for a value it has already computed it remembers the value.\n\nPython has a caching tool available in the functools library:", "_____no_output_____" ] ], [ [ "import functools\n\n\ndef generate_sequence(n):\n \"\"\"\n Generate the sequence defined by:\n\n a_1 = 1\n a_n = 3 a_{n - 1}\n\n This is done using recursion.\n \"\"\"\n if n == 1:\n return 1\n return 3 * generate_sequence(n - 1)\n\n\[email protected]_cache()\ndef cached_generate_sequence(n):\n \"\"\"\n Generate the sequence defined by:\n\n a_1 = 1\n a_n = 3 a_{n - 1}\n\n This is done using recursion but also includes a cache.\n \"\"\"\n if n == 1:\n return 1\n return 3 * cached_generate_sequence(n - 1)", "_____no_output_____" ] ], [ [ "Timing both these approaches confirms a substantial increase in time for the\ncached version.", "_____no_output_____" ] ], [ [ "%timeit [generate_sequence(n) for n in range(1, 25)]", "27.6 µs ± 372 ns per loop (mean ± std. dev. of 7 runs, 10000 loops each)\n" ], [ "%timeit [cached_generate_sequence(n) for n in range(1, 25)]", "1.53 µs ± 7.47 ns per loop (mean ± std. dev. of 7 runs, 1000000 loops each)\n" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ] ]

[ "CC-BY-4.0" ]

[ "CC-BY-4.0" ]

[ "CC-BY-4.0" ]

[ [ [ "# Import conventions we'll be using here. See Part 1\nimport matplotlib\n# matplotlib.use('nbagg')\nimport matplotlib.pyplot as plt\nimport numpy as np", "_____no_output_____" ] ], [ [ "# Limits, Legends, and Layouts\n\nIn this section, we'll focus on what happens around the edges of the axes: Ticks, ticklabels, limits, layouts, and legends.", "_____no_output_____" ], [ "# Limits and autoscaling\n\nBy default, Matplotlib will attempt to determine limits for you that encompasses all the data you have plotted. This is the \"autoscale\" feature. For image plots, the limits are not padded while plots such as scatter plots and bar plots are given some padding.", "_____no_output_____" ] ], [ [ "fig, (ax1, ax2) = plt.subplots(nrows=1, ncols=2, figsize=plt.figaspect(0.5))\n\nax1.plot([-10, -5, 0, 5, 10, 15], [-1.2, 2, 3.5, -0.3, -4, 1])\nax2.scatter([-10, -5, 0, 5, 10, 15], [-1.2, 2, 3.5, -0.3, -4, 1])\n\nplt.show()", "_____no_output_____" ] ], [ [ "### `ax.margins(...)`\n\nIf you'd like to add a bit of \"padding\" to a plot, `ax.margins(<some_small_fraction>)` is a very handy way to do so. Instead of choosing \"even-ish\" numbers as min/max ranges for each axis, `margins` will make Matplotlib calculate the min/max of each axis by taking the range of the data and adding on a fractional amount of padding.\n\nAs an example:", "_____no_output_____" ] ], [ [ "fig, (ax1, ax2) = plt.subplots(1, 2, figsize=plt.figaspect(0.5))\n\nax1.plot([-10, -5, 0, 5, 10, 15], [-1.2, 2, 3.5, -0.3, -4, 1])\nax2.scatter([-10, -5, 0, 5, 10, 15], [-1.2, 2, 3.5, -0.3, -4, 1])\n\nax1.margins(x=0.0, y=0.1) # 10% padding in the y-direction only\nax2.margins(0.05) # 5% padding in all directions\n\nplt.show()", "_____no_output_____" ] ], [ [ "### `ax.axis(...)`\n\nThe `ax.axis(...)` method is a convienent way of controlling the axes limits and enabling/disabling autoscaling.\n\nIf you ever need to get all of the current plot limits, calling `ax.axis()` with no arguments will return the xmin/max/etc:\n\n xmin, xmax, ymin, ymax = ax.axis()\n \nIf you'd like to manually set all of the x/y limits at once, you can use `ax.axis` for this, as well (note that we're calling it with a single argument that's a sequence, not 4 individual arguments):\n\n ax.axis([xmin, xmax, ymin, ymax])\n \nHowever, you'll probably use `axis` mostly with either the `\"tight\"` or `\"equal\"` options. There are other options as well; see the documentation for full details. In a nutshell, though:\n\n * *tight*: Set axes limits to the exact range of the data\n * *equal*: Set axes scales such that one cm/inch in the y-direction is the same as one cm/inch in the x-direction. In Matplotlib terms, this sets the aspect ratio of the plot to 1. That _doesn't_ mean that the axes \"box\" will be square.\n \nAnd as an example:", "_____no_output_____" ] ], [ [ "fig, axes = plt.subplots(nrows=3)\n\nfor ax in axes:\n ax.plot([-10, -5, 0, 5, 10, 15], [-1.2, 2, 3.5, -0.3, -4, 1])\n\naxes[0].set_title('Normal Autoscaling', y=0.7, x=0.8)\n\naxes[1].set_title('ax.axis(\"tight\")', y=0.7, x=0.8)\naxes[1].axis('tight')\n\naxes[2].set_title('ax.axis(\"equal\")', y=0.7, x=0.8)\naxes[2].axis('equal')\n\nplt.show()", "_____no_output_____" ] ], [ [ "### Manually setting only one limit\n\nAnother trick with limits is to specify only half of a limit. When done **after** a plot is made, this has the effect of allowing the user to anchor a limit while letting Matplotlib autoscale the rest of it.", "_____no_output_____" ] ], [ [ "# Good -- setting limits after plotting is done\nfig, (ax1, ax2) = plt.subplots(1, 2, figsize=plt.figaspect(0.5))\nax1.plot([-10, -5, 0, 5, 10, 15], [-1.2, 2, 3.5, -0.3, -4, 1])\nax2.scatter([-10, -5, 0, 5, 10, 15], [-1.2, 2, 3.5, -0.3, -4, 1])\nax1.set_ylim(bottom=-10)\nax2.set_xlim(right=25)\nplt.show()", "_____no_output_____" ], [ "# Bad -- Setting limits before plotting is done\nfig, (ax1, ax2) = plt.subplots(1, 2, figsize=plt.figaspect(0.5))\nax1.set_ylim(bottom=-10)\nax2.set_xlim(right=25)\nax1.plot([-10, -5, 0, 5, 10, 15], [-1.2, 2, 3.5, -0.3, -4, 1])\nax2.scatter([-10, -5, 0, 5, 10, 15], [-1.2, 2, 3.5, -0.3, -4, 1])\nplt.show()", "_____no_output_____" ] ], [ [ "# Legends\n\nAs you've seen in some of the examples so far, the X and Y axis can also be labeled, as well as the subplot itself via the title. \n\nHowever, another thing you can label is the line/point/bar/etc that you plot. You can provide a label to your plot, which allows your legend to automatically build itself. ", "_____no_output_____" ] ], [ [ "fig, ax = plt.subplots()\nax.plot([1, 2, 3, 4], [10, 20, 25, 30], label='Philadelphia')\nax.plot([1, 2, 3, 4], [30, 23, 13, 4], label='Boston')\nax.set(ylabel='Temperature (deg C)', xlabel='Time', title='A tale of two cities')\nax.legend()\nplt.show()", "_____no_output_____" ] ], [ [ "In `classic` mode, legends will go in the upper right corner by default (you can control this with the `loc` kwarg). As of v2.0, by default Matplotlib will choose a location to avoid overlapping plot elements as much as possible. To force this option, you can pass in:\n\n ax.legend(loc=\"best\")\n \nAlso, if you happen to be plotting something that you do not want to appear in the legend, just set the label to \"\\_nolegend\\_\".", "_____no_output_____" ] ], [ [ "fig, ax = plt.subplots(1, 1)\nax.bar([1, 2, 3, 4], [10, 20, 25, 30], label=\"Foobar\", align='center', color='lightblue')\nax.plot([1, 2, 3, 4], [10, 20, 25, 30], label=\"_nolegend_\", marker='o', color='darkred')\nax.legend(loc='best')\nplt.show()", "_____no_output_____" ] ], [ [ "# Exercise 4.1\n\nOnce again, let's use a bit of what we've learned. Try to reproduce the following figure:\n\n<img src=\"images/exercise_4-1.png\">\n\nHint: You'll need to combine `ax.axis(...)` and `ax.margins(...)`. Here's the data and some code to get you started:", "_____no_output_____" ] ], [ [ "import numpy as np\nimport matplotlib.pyplot as plt\n\nt = np.linspace(0, 2 * np.pi, 150)\nx1, y1 = np.cos(t), np.sin(t)\nx2, y2 = 2 * x1, 2 * y1\n\ncolors = ['darkred', 'darkgreen']\n\n# Try to plot the two circles, scale the axes as shown and add a legend\n# Hint: it's easiest to combine `ax.axis(...)` and `ax.margins(...)` to scale the axes", "_____no_output_____" ], [ "%load solutions/4.1-legends_and_scaling.py", "_____no_output_____" ] ], [ [ "# Dealing with the boundaries: Layout, ticks, spines, etc\n\nOne key thing we haven't talked about yet is all of the annotation on the outside of the axes, the borders of the axes, and how to adjust the amount of space around the axes. We won't go over every detail, but this next section should give you a reasonable working knowledge of how to configure what happens around the edges of your axes.\n\n## Ticks, Tick Lines, Tick Labels and Tickers\nThis is a constant source of confusion:\n\n* A Tick is the *location* of a Tick Label.\n* A Tick Line is the line that denotes the location of the tick.\n* A Tick Label is the text that is displayed at that tick.\n* A [`Ticker`](http://matplotlib.org/api/ticker_api.html#module-matplotlib.ticker) automatically determines the ticks for an Axis and formats the tick labels.\n\n[`tick_params()`](https://matplotlib.org/api/axes_api.html#ticks-and-tick-labels) is often used to help configure your tickers.", "_____no_output_____" ] ], [ [ "fig, ax = plt.subplots()\nax.plot([1, 2, 3, 4], [10, 20, 25, 30])\n\n# Manually set ticks and tick labels *on the x-axis* (note ax.xaxis.set, not ax.set!)\nax.xaxis.set(ticks=range(1, 5), ticklabels=[3, 100, -12, \"foo\"]) \n\n# Make the y-ticks a bit longer and go both in and out...\nax.tick_params(axis='y', direction='inout', length=10)\n\nplt.show()", "_____no_output_____" ] ], [ [ "A commonly-asked question is \"How do I plot categories?\"\n\nStarting in version 2.0 of mpl, just like any other data.\n\nFor example:", "_____no_output_____" ] ], [ [ "data = [('apples', 2), ('oranges', 3), ('peaches', 1)]\nfruit, value = zip(*data)\n\nfig, ax = plt.subplots()\nax.bar(fruit, value, align='center', color='gray')\nplt.show()", "_____no_output_____" ] ], [ [ "## Subplot Spacing\nThe spacing between the subplots can be adjusted using [`fig.subplots_adjust()`](http://matplotlib.org/api/pyplot_api.html?#matplotlib.pyplot.subplots_adjust). Play around with the example below to see how the different arguments affect the spacing.", "_____no_output_____" ] ], [ [ "fig, axes = plt.subplots(2, 2, figsize=(9, 9))\nfig.subplots_adjust(wspace=0.5, hspace=0.3,\n left=0.125, right=0.9,\n top=0.9, bottom=0.1)\nplt.show()", "_____no_output_____" ] ], [ [ "A common \"gotcha\" is that the labels are not automatically adjusted to avoid overlapping those of another subplot. Matplotlib does not currently have any sort of robust layout engine, as it is a design decision to minimize the amount of \"magical plotting\". We intend to let users have complete, 100% control over their plots. LaTeX users would be quite familiar with the amount of frustration that can occur with automatic placement of figures in their documents.\n\nThat said, there have been some efforts to develop tools that users can use to help address the most common compaints. The \"[Tight Layout](http://matplotlib.org/users/tight_layout_guide.html)\" feature, when invoked, will attempt to resize margins and subplots so that nothing overlaps.\n\nIf you have multiple subplots, and want to avoid overlapping titles/axis labels/etc, `fig.tight_layout` is a great way to do so:", "_____no_output_____" ] ], [ [ "def example_plot(ax):\n ax.plot([1, 2])\n ax.set_xlabel('x-label', fontsize=16)\n ax.set_ylabel('y-label', fontsize=8)\n ax.set_title('Title', fontsize=24)\n\nfig, ((ax1, ax2), (ax3, ax4)) = plt.subplots(nrows=2, ncols=2)\nexample_plot(ax1)\nexample_plot(ax2)\nexample_plot(ax3)\nexample_plot(ax4)\n\n# Enable fig.tight_layout to compare...\nfig.tight_layout()\n\nplt.show()", "_____no_output_____" ] ], [ [ "## GridSpec\nUnder the hood, Matplotlib utilizes [`GridSpec`](http://matplotlib.org/api/gridspec_api.html) to lay out the subplots. While `plt.subplots()` is fine for simple cases, sometimes you will need more advanced subplot layouts. In such cases, you should use GridSpec directly. GridSpec is outside the scope of this tutorial, but it is handy to know that it exists. [Here](http://matplotlib.org/users/gridspec.html) is a guide on how to use it.", "_____no_output_____" ], [ "## Sharing axes\nThere will be times when you want to have the x axis and/or the y axis of your subplots to be \"shared\". Sharing an axis means that the axis in one or more subplots will be tied together such that any change in one of the axis changes all of the other shared axes. This works very nicely with autoscaling arbitrary datasets that may have overlapping domains. Furthermore, when interacting with the plots (panning and zooming), all of the shared axes will pan and zoom automatically.", "_____no_output_____" ] ], [ [ "fig, (ax1, ax2) = plt.subplots(1, 2, sharex=True, sharey=True)\nax1.plot([1, 2, 3, 4], [1, 2, 3, 4])\nax2.plot([3, 4, 5, 6], [6, 5, 4, 3])\nplt.show()", "_____no_output_____" ] ], [ [ "## \"Twinning\" axes\nSometimes one may want to overlay two plots on the same axes, but the scales may be entirely different. You can simply treat them as separate plots, but then twin them.", "_____no_output_____" ] ], [ [ "fig, ax1 = plt.subplots(1, 1)\nax1.plot([1, 2, 3, 4], [1, 2, 3, 4])\nax2 = ax1.twinx()\nax2.scatter([1, 2, 3, 4], [60, 50, 40, 30])\nax1.set(xlabel='X', ylabel='First scale')\nax2.set(ylabel='Other scale')\nplt.show()", "_____no_output_____" ] ], [ [ "# Axis Spines\nSpines are the axis lines for a plot. Each plot can have four spines: \"top\", \"bottom\", \"left\" and \"right\". By default, they are set so that they frame the plot, but they can be individually positioned and configured via the [`set_position()`](http://matplotlib.org/api/spines_api.html#matplotlib.spines.Spine.set_position) method of the spine. Here are some different configurations.", "_____no_output_____" ] ], [ [ "fig, ax = plt.subplots()\nax.plot([-2, 2, 3, 4], [-10, 20, 25, 5])\nax.spines['top'].set_visible(False)\nax.xaxis.set_ticks_position('bottom') # no ticklines at the top\nax.spines['right'].set_visible(False)\nax.yaxis.set_ticks_position('left') # no ticklines on the right\n\n# \"outward\"\n# Move the two remaining spines \"out\" away from the plot by 10 points\n#ax.spines['bottom'].set_position(('outward', 10))\n#ax.spines['left'].set_position(('outward', 10))\n\n# \"data\"\n# Have the spines stay intersected at (0,0)\n#ax.spines['bottom'].set_position(('data', 0))\n#ax.spines['left'].set_position(('data', 0))\n\n# \"axes\"\n# Have the two remaining spines placed at a fraction of the axes\n#ax.spines['bottom'].set_position(('axes', 0.75))\n#ax.spines['left'].set_position(('axes', 0.3))\n\nplt.show()", "_____no_output_____" ] ], [ [ "# Exercise 4.2\n\nThis one is a bit trickier. Once again, try to reproduce the figure below:\n\n<img src=\"images/exercise_4-2.png\">\n\n\nA few key hints: The two subplots have no vertical space between them (this means that the `hspace` is `0`). Note that the bottom spine is at 0 in data coordinates and the tick lines are missing from the right and top sides.\n\nBecause you're going to be doing a lot of the same things to both subplots, to avoid repitive code you might consider writing a function that takes an `Axes` object and makes the spine changes, etc to it. \n\n", "_____no_output_____" ] ], [ [ "import matplotlib.pyplot as plt\nimport numpy as np\n\n# Try to reproduce the figure shown in images/exercise_4.2.png\n# This one is a bit trickier!\n\n# Here's the data...\ndata = [('dogs', 4, 4), ('frogs', -3, 1), ('cats', 1, 5), ('goldfish', -2, 2)]\nanimals, friendliness, popularity = zip(*data)", "_____no_output_____" ], [ "%load solutions/4.2-spines_ticks_and_subplot_spacing.py", "_____no_output_____" ] ] ]

[ "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code", "markdown", "code" ]

[ [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown", "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code" ], [ "markdown" ], [ "code", "code" ] ]

[ "MIT" ]

[ "MIT" ]

[ "MIT" ]

[ [ [ "# Matplot normal usage and example", "_____no_output_____" ] ], [ [ "import matplotlib.pyplot as plt", "_____no_output_____" ], [ "%matplotlib inline", "_____no_output_____" ], [ "import numpy as np", "_____no_output_____" ], [ "x = np.linspace(0,5, 11)\ny = x ** 2", "_____no_output_____" ], [ "x", "_____no_output_____" ], [ "y", "_____no_output_____" ], [ "plt.plot(x,y,\"black\")\nplt.xlabel(\"x label\")\nplt.ylabel(\"y label\")\nplt.title(\"title\")\nplt.show()", "_____no_output_____" ], [ "plt.subplot(1, 2, 1)\nplt.plot(x,y, \"b\")\n\nplt.subplot(2, 2, 2)\nplt.plot(y, x, \"r\")\n", "_____no_output_____" ], [ "fig = plt.figure()\n\naxes = fig.add_axes([0.1, 0.1, 0.8, 0.8])\naxes.plot(x, y)\naxes.set_xlabel(\"x axis\")\naxes.set_ylabel(\"y axis\")\naxes.set_title(\"title\")", "_____no_output_____" ], [ "fig = plt.figure()\n\naxes1 = fig.add_axes([0.1, 0.1, 0.8, 0.8])\naxes1.plot(x,y,\"r\")\naxes1.set_xlabel(\"x axis\")\naxes1.set_ylabel(\"y axis\")\naxes1.set_title(\"Larger Plot\")\n\naxes2 = fig.add_axes([0.2, 0.4, 0.4, 0.4])\naxes2.plot(y,x,\"b\")\naxes2.set_title(\"Smaller Plot\")", "_____no_output_____" ], [ "fig,axes = plt.subplots(nrows=2, ncols=2)\n\naxes[0][0].plot(x, y, \"r\")\naxes[1][1].plot(y, x, \"r\")\n\nplt.tight_layout()", "_____no_output_____" ] ], [ [ "### Figure Size and DPI", "_____no_output_____" ] ], [ [ "fig, axes = plt.subplots(figsize=(8,2))\nax = fig.add_axes([0,0,1,1])\nax.plot(x,y)", "_____no_output_____" ], [ "fig, axes = plt.subplots(figsize=(8,2), dpi=100)\nax = fig.add_axes([0,0,1,1])\nax.plot(x,y)", "_____no_output_____" ], [ "fig, axes = plt.subplots(nrows=2, ncols =1, figsize=(8,2))\naxes[0].plot(x,y)\naxes[1].plot(y,x)\n\nplt.tight_layout()", "_____no_output_____" ], [ "fig", "_____no_output_____" ], [ "fig.savefig(\"axis plot\", dpi=100)", "_____no_output_____" ], [ "fig = plt.figure()\n\nax = fig.add_axes([0,0,1,1])\nax.plot(x, x**2, label=\"x Squared\")\nax.plot(x, x**3, label=\"x Cube\")\n\nax.legend(loc=0)", "_____no_output_____" ] ], [ [ "### Plot Color and decoration", "_____no_output_____" ] ], [ [ "fig = plt.figure()\n\nax = fig.add_axes([0,0,1,1])\nax.plot(x, y, \"#000000\", lw=3, linestyle=\"--\", marker=\"o\", markersize=20, \n markerfacecolor=\"yellow\", markeredgewidth=3, markeredgecolor=\"green\") # rgb, hex #linewidth is lw\n", "_____no_output_____" ], [ "fig = plt.figure()\n\nax = fig.add_axes([0,0,1,1])\nax.plot(x,y, \"r--\")\nax.set_xlim([0,1])\nax.set_ylim([0,2])", "_____no_output_____" ] ], [ [ "#### Further Reading:\nhttp://www.labri.fr/perso/nrougier/teaching/matplotlib/matplotlib.html\n\nhttps://matplotlib.org/gallery.html", "_____no_output_____" ] ] ]

[ "markdown", "code", "markdown", "code", "markdown", "code", "markdown" ]

[ [ "markdown" ], [ "code", "code", "code", "code", "code", "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code", "code", "code", "code", "code" ], [ "markdown" ], [ "code", "code" ], [ "markdown" ] ]

[ "MIT" ]

[ "MIT" ]