test_model_w_poly_terms.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "UrajMxg_5d2-"
   },
   "source": [
    "# Polynomial Model Testing\n",
    "\n",
    "The purpose of this notbook is to examine the relationship (or lack thereof) between drug's various features by creating interaction terms. This would would hopefully show that there is a strong relationship between aggregate drug markers and their method of action in order to create a more accurate model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {
    "executionInfo": {
     "elapsed": 1718,
     "status": "ok",
     "timestamp": 1603378654499,
     "user": {
      "displayName": "ryan lewis",
      "photoUrl": "",
      "userId": "09680394787948119906"
     },
     "user_tz": 420
    },
    "id": "Z1hdVobyCcLI"
   },
   "outputs": [],
   "source": [
    "import io, requests\n",
    "\n",
    "import datetime\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from time import time\n",
    "from tqdm.notebook import tqdm\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "# import xgboost as xgb\n",
    "\n",
    "from sklearn.preprocessing import power_transform\n",
    "from sklearn.preprocessing import PolynomialFeatures\n",
    "#from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.model_selection import StratifiedKFold\n",
    "from sklearn.svm import NuSVR\n",
    "from sklearn.metrics import log_loss\n",
    "\n",
    "###############################################################################\n",
    "# Undersampling did not show any appreciable improvement in model performance #\n",
    "###############################################################################\n",
    "\n",
    "# from imblearn.under_sampling import TomekLinks # removed 5ish samples and only made a marginal improvement to the logloss values\n",
    "# from imblearn.under_sampling import ClusterCentroids # wasn't useful\n",
    "# from imblearn.under_sampling import NeighbourhoodCleaningRule # made model marginally worse and took significantly longer\n",
    "# from imblearn.under_sampling import InstanceHardnessThreshold #kept getting errors\n",
    "# from imblearn.under_sampling import EditedNearestNeighbours#did not remove any samples\n",
    "# from imblearn.under_sampling import AllKNN#did not remove any samples\n",
    "# from imblearn.under_sampling import CondensedNearestNeighbour#crashed after using all ram\n",
    "# from imblearn.under_sampling import NearMiss# marginally worse after running NearMiss(version=2, n_jobs=2, sampling_strategy=sampling_val)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "7a0LBSJd1vJ7"
   },
   "source": [
    "## Loading the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "executionInfo": {
     "elapsed": 2313,
     "status": "ok",
     "timestamp": 1603378657832,
     "user": {
      "displayName": "ryan lewis",
      "photoUrl": "",
      "userId": "09680394787948119906"
     },
     "user_tz": 420
    },
    "id": "LINf-Doscml1"
   },
   "outputs": [],
   "source": [
    "def dummy_vars(df):\n",
    "    df.loc[:, 'cp_type'] = df.loc[:, 'cp_type'].map({'trt_cp': 0, 'ctl_vehicle': 1})\n",
    "    # trt_cp is a control and has no MOA, ctl_vehicle indicates a real test\n",
    "    df.loc[:, 'cp_dose'] = df.loc[:, 'cp_dose'].map({'D1': 0, 'D2': 1})\n",
    "    # create columns ['cp_type_24', 'cp_type_48', 'cp_type_72'] which are one hot encoded based on cp_time\n",
    "    df['cp_type_24'] = df['cp_time'].astype(str) == '24'\n",
    "    df['cp_type_48'] = df['cp_time'].astype(str) == '48'\n",
    "    df['cp_type_72'] = df['cp_time'].astype(str) == '72'\n",
    "    df.drop(columns=['sig_id','cp_time'], inplace = True)\n",
    "    return df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 104
    },
    "executionInfo": {
     "elapsed": 6428,
     "status": "ok",
     "timestamp": 1603378662286,
     "user": {
      "displayName": "ryan lewis",
      "photoUrl": "",
      "userId": "09680394787948119906"
     },
     "user_tz": 420
    },
    "id": "Gffd7zyYFOF7",
    "outputId": "7207fcc3-2982-4e46-94f7-4379d419396b"
   },
   "outputs": [],
   "source": [
    "def load_files():\n",
    "    # Train Features\n",
    "    train_features = pd.read_csv('train_features.csv')\n",
    "    train_features = dummy_vars(train_features)\n",
    "    print('train_features loaded')\n",
    "\n",
    "    # Train Targets\n",
    "    train_targets = pd.read_csv('train_targets_scored.csv')\n",
    "    train_targets.drop(columns=['sig_id'], inplace = True)\n",
    "    print('train_targets loaded')\n",
    "\n",
    "    # Test Features\n",
    "    test_features = pd.read_csv('test_features.csv')\n",
    "    test_features = dummy_vars(test_features)\n",
    "    #test_features = pd.DataFrame(power_transform(test_features, method='yeo-johnson'))\n",
    "    print('test_features loaded')\n",
    "\n",
    "    # Sample Submission DF (Targets)\n",
    "    ss_lr = pd.read_csv('sample_submission.csv')\n",
    "    ss_lr.drop(columns=['sig_id'], inplace = True)\n",
    "    print('ss_lr loaded')\n",
    "\n",
    "    print('done!')\n",
    "\n",
    "    return train_features, train_targets, test_features, ss_lr"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "fahpRhhW3TJq"
   },
   "source": [
    "## Data Processing/Prediction Pipeline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "executionInfo": {
     "elapsed": 5791,
     "status": "ok",
     "timestamp": 1603378662287,
     "user": {
      "displayName": "ryan lewis",
      "photoUrl": "",
      "userId": "09680394787948119906"
     },
     "user_tz": 420
    },
    "id": "snbrfwC4izZk"
   },
   "outputs": [],
   "source": [
    "def log_loss_metric(y_true, y_pred):\n",
    "    metrics = []\n",
    "    for _target in train_targets.columns:\n",
    "        metrics.append(log_loss(y_true.loc[:, _target], y_pred.loc[:, _target].astype(float), labels = [0,1]))\n",
    "    return np.mean(metrics)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "metadata": {
    "executionInfo": {
     "elapsed": 5457,
     "status": "ok",
     "timestamp": 1603378662288,
     "user": {
      "displayName": "ryan lewis",
      "photoUrl": "",
      "userId": "09680394787948119906"
     },
     "user_tz": 420
    },
    "id": "rQIRw5YvAaIR"
   },
   "outputs": [],
   "source": [
    "def k_fold_fit_pred(model_train_num, column, x_test, target_vals, N_SPLITS, res, ss, coef_df2):\n",
    "    for rand_state in range(model_train_num):\n",
    "        # train the model for each one of the random states \n",
    "        skf = StratifiedKFold(n_splits = N_SPLITS, random_state = rand_state, shuffle = True)\n",
    "\n",
    "        for (train, val) in skf.split(target_vals, target_vals):\n",
    "        # for each of the train/validate splits...\n",
    "            \n",
    "            #print(X[train].shape, end = '\\r')\n",
    "            #print(target_vals[train].shape)\n",
    "            \n",
    "            #x_train = X[train, column].reshape(-1, 1).astype(float)\n",
    "            x_train = X_df.values[train].astype(float)\n",
    "            y_train = target_vals[train].astype(float)\n",
    "\n",
    "            x_train = scaler.fit_transform(x_train)\n",
    "\n",
    "            #x_val = X[val, column].astype(float)#.reshape(-1, 1).astype(float)\n",
    "            x_val = X_df.values[val].astype(float)\n",
    "            y_val = target_vals[val].astype(float)\n",
    "\n",
    "            x_val = scaler.transform(x_val)\n",
    "\n",
    "            model.fit(x_train, y_train)\n",
    "            temp_df = pd.DataFrame([list(model.coef_)], columns = list(X_df.columns))\n",
    "            #print(temp_df.head())\n",
    "            coef_df = pd.concat([coef_df2,temp_df])\n",
    "            #print(coef_df.head())\n",
    "            \n",
    "            try:\n",
    "                ss_pred_vals = model.predict_proba(x_test[:, column].reshape(-1,1))[:,1]\n",
    "            except: # For the tree based models\n",
    "                #ss_pred_vals = model.predict(x_test[:, column].reshape(-1,1))\n",
    "                ss_pred_vals = model.predict(x_test)\n",
    "            ss.loc[:, train_targets.columns[column]] += (\n",
    "                ss_pred_vals / (N_SPLITS * model_train_num))\n",
    "\n",
    "            try:\n",
    "                res_pred_vals = model.predict_proba(x_val)[:,1]\n",
    "            except: # For the tree based models\n",
    "                res_pred_vals = model.predict(x_val)\n",
    "            res.loc[val, train_targets.columns[column]] += (\n",
    "                res_pred_vals / model_train_num)\n",
    "\n",
    "    return res, ss, coef_df2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "metadata": {
    "executionInfo": {
     "elapsed": 632,
     "status": "ok",
     "timestamp": 1603378797207,
     "user": {
      "displayName": "ryan lewis",
      "photoUrl": "",
      "userId": "09680394787948119906"
     },
     "user_tz": 420
    },
    "id": "mjpia2xNk77Y"
   },
   "outputs": [],
   "source": [
    "# cols = [c for c in ss_lr.columns.values]\n",
    "def run_score(train_targets, X_df.values, x_test_df.values, res, ss, logloss_vals, coef_df2):\n",
    "# train_targets is the training targets\n",
    "# X is the train features\n",
    "# x_test is the holdout set of features\n",
    "# ss_ is blank since it's the empty sample submission predictions\n",
    "\n",
    "    N_SPLITS = 5 # number of k-fold splits\n",
    "    model_train_num = 3 # number of random states the model is trained on for each target\n",
    "\n",
    "    for column in tqdm(range(train_targets.shape[1])):\n",
    "    # create a model for each of the targets\n",
    "        start_time = time()\n",
    "        target_vals = train_targets.values[:, column]\n",
    "\n",
    "        if target_vals.sum() >= 5:\n",
    "        # split if there's enough positive target values\n",
    "            res, ss, coef_df2 = k_fold_fit_pred(model_train_num, column, x_test_df.values, target_vals, N_SPLITS, res, ss, coef_df2)\n",
    "            \n",
    "        score = log_loss(\n",
    "            train_targets.loc[:, train_targets.columns[column]],\n",
    "            res.loc[:, train_targets.columns[column]])\n",
    "        feat_time = str(datetime.timedelta(seconds = time() - start_time))[2:7]\n",
    "        #print(f'[{feat_time}] Target {column+1}/{train_targets.shape[1]} Log Loss:', score)\n",
    "        target_counts.append(np.sum(target_vals))\n",
    "        logloss_vals.append(score)\n",
    "        #print(coef_df.head())\n",
    "\n",
    "    print(f'Log Loss: {log_loss_metric(train_targets, res)}')\n",
    "\n",
    "    # these lines automatically set the target values equal to zero if it cp_type = 1 AKA control group  \n",
    "    res.loc[train_features['cp_type'] == 1, train_targets.columns] = 0\n",
    "    ss.loc[test_features['cp_type'] == 1, train_targets.columns] = 0\n",
    "    print(f'After manually eliminating control groups: {log_loss_metric(train_targets, res)}\\n\\n')\n",
    "    \n",
    "    return res, ss, logloss_vals, coef_df2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train_features loaded\n",
      "train_targets loaded\n",
      "test_features loaded\n",
      "ss_lr loaded\n",
      "done!\n"
     ]
    }
   ],
   "source": [
    "train_features, train_targets, test_features, ss_lr = load_files()\n",
    "\n",
    "g_tags = [g for g in list(train_features.columns) if g[0]+g[1] == 'g-']\n",
    "c_tags = [c for c in list(train_features.columns) if c[0]+c[1] == 'c-']\n",
    "d_tags = [d for d in list(train_features.columns) if d not in c_tags]\n",
    "d_tags = [d for d in d_tags if d not in g_tags]\n",
    "\n",
    "# train_features[c_tags] = scaler.fit_transform(train_features[c_tags])\n",
    "# test_features[c_tags] = scaler.fit_transform(test_features[c_tags])\n",
    "\n",
    "# train_features[g_tags] = scaler.fit_transform(train_features[g_tags])\n",
    "# test_features[g_tags] = scaler.fit_transform(test_features[g_tags])\n",
    "\n",
    "#X = train_features.values[:,:]\n",
    "poly = PolynomialFeatures(interaction_only=True)\n",
    "X = np.concatenate([train_features.loc[:,d_tags].values,\n",
    "               poly.fit_transform(train_features.loc[:,c_tags]),\n",
    "               train_features.loc[:,g_tags]],\n",
    "              axis=1)\n",
    "#x_test = test_features.values[:,:]\n",
    "x_test = np.concatenate([test_features.loc[:,d_tags].values,\n",
    "               poly.fit_transform(test_features.loc[:,c_tags]),\n",
    "               test_features.loc[:,g_tags].values],\n",
    "              axis=1)\n",
    "\n",
    "#This is for the sample_submission.csv df\n",
    "ss_lr.loc[:, train_targets.columns] = 0\n",
    "\n",
    "#These are the target values which are being predicted and tested\n",
    "res_lr = train_targets.copy()\n",
    "res_lr.loc[:, train_targets.columns] = 0\n",
    "\n",
    "cols = ss_lr.columns.values"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [],
   "source": [
    "feat_labels = d_tags + poly.get_feature_names(input_features=c_tags) + g_tags\n",
    "coef_df = pd.DataFrame(columns = feat_labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Running and Evaluating the Models\n",
    "\n",
    "The features fall into three distinct groups:\n",
    "- Chemical Markers (Denoted as c-__)\n",
    "- Genetic Markers (Denoted as g-__)\n",
    "- Experimental Variables (Denoted as cp-__)\n",
    "\n",
    "Based on the varying number of features in each of these categories, chemical marker feature interaction terms were chosen as the first group to explore. This is because there were 100 of these features and the feature space went from 900 terms to ~5500. Looking at the genetic markers would have expanded the feature space significantly beyond that and would have required ~60GB of RAM which makes the model impractical to analyze in a reasonable amount of time while leveraging GPUs."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b2d739f9ff5742f995a048f5df0d1a78",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(HTML(value=''), FloatProgress(value=0.0, max=206.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[W] [10:16:56.339160] QWL-QN: max iterations reached\n",
      "[W] [11:57:16.705341] QWL-QN: max iterations reached\n",
      "[W] [11:57:40.893372] QWL-QN: max iterations reached\n",
      "[W] [11:58:22.589945] QWL-QN: max iterations reached\n",
      "[W] [11:58:39.623664] QWL-QN: max iterations reached\n",
      "[W] [11:59:10.205158] QWL-QN: max iterations reached\n",
      "[W] [11:59:27.273932] QWL-QN: max iterations reached\n",
      "[W] [11:59:44.365539] QWL-QN: max iterations reached\n",
      "[W] [12:00:24.288939] QWL-QN: max iterations reached\n",
      "[W] [12:10:12.110509] QWL-QN: max iterations reached\n",
      "[W] [12:10:45.550960] QWL-QN: max iterations reached\n",
      "[W] [12:11:02.447477] QWL-QN: max iterations reached\n",
      "[W] [12:11:30.649704] QWL-QN: max iterations reached\n",
      "[W] [12:11:47.772635] QWL-QN: max iterations reached\n",
      "[W] [12:12:04.915928] QWL-QN: max iterations reached\n",
      "[W] [12:12:21.861254] QWL-QN: max iterations reached\n",
      "[W] [13:52:47.169839] QWL-QN: max iterations reached\n",
      "[W] [13:53:35.720645] QWL-QN: max iterations reached\n",
      "[W] [13:54:21.895133] QWL-QN: max iterations reached\n",
      "\n",
      "Log Loss: 0.07183212526967282\n",
      "After manually eliminating control groups: 0.07155615905436347\n",
      "\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from cuml.linear_model import LogisticRegression\n",
    "#from sklearn.linear_model import LogisticRegression\n",
    "\n",
    "feat_labels = d_tags + poly.get_feature_names(input_features=c_tags) + g_tags\n",
    "coef_df = pd.DataFrame(columns = feat_labels)\n",
    "\n",
    "# column_names = [\"a\", \"b\", \"c\"]\n",
    "# coef_df = pd.DataFrame(columns = column_names)\n",
    "rowval = 0\n",
    "for i in [256]: # left for further hyperparameter tuning\n",
    "    from cuml.experimental.preprocessing import StandardScaler\n",
    "    scaler = StandardScaler()\n",
    "    #model = LogisticRegression(C = 128, max_iter = 100000, penalty='elasticnet', l1_ratio=1)\n",
    "    model = LogisticRegression(C = 128, penalty='elasticnet', l1_ratio=1)\n",
    "    target_counts = []\n",
    "    logistic_regr_loss_vals = []\n",
    "    ss_lr.loc[:, train_targets.columns] = 0\n",
    "    res_lr.loc[:, train_targets.columns] = 0\n",
    "    res_lr, ss_lr, logistic_regr_loss_vals, coef_df = run_score(train_targets, X_df.values, x_test_df.values, res_lr, ss_lr, logistic_regr_loss_vals, coef_df)\n",
    "    \n",
    "plt.scatter(target_counts,logistic_regr_loss_vals)\n",
    "plt.xlabel('target_counts')\n",
    "plt.ylabel('log loss values')\n",
    "plt.title('cuML Logistic Regression')\n",
    "plt.show()\n",
    "res_lr.to_pickle('pickles/res_cuml_lr.p')\n",
    "ss_lr.to_pickle('pickles/ss_cuml_lr.p')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Analyzing the Results\n",
    "- The weights from each of the models (306 targets, 10 per target) were saved into the following dataframe to get an intuition of which features could be correllated positively or negatively with the targets.\n",
    "- For each of the features the standard deviation and the mean was taken.\n",
    "- The features which had a low correllation value (mean close to zero) and relatively little deviation from that correllation value were eliminated.\n",
    "- Using the reduced list of features the model was run again.\n",
    "- The model with the reduced feature space had a roughly 40% improvement in the loss score.\n",
    "## Thoughts\n",
    "Typically a larger feature space would allow a model to perform better as it gives more opportunity for a given algorithm to find paterns in the noise. Since a training penalty was used it makes sense that \"pruning\" the features which showed a low correllation with the target improved the score by preventing the model from overfitting on the noise in the training set. This process would make sense to repeat as long as the model scores keep improving. After that it would be interesting to see if applying an l2 penalty or no penalty could further improve the model's performance."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cp_type</th>\n",
       "      <th>cp_dose</th>\n",
       "      <th>cp_type_24</th>\n",
       "      <th>cp_type_48</th>\n",
       "      <th>cp_type_72</th>\n",
       "      <th>1</th>\n",
       "      <th>c-0</th>\n",
       "      <th>c-1</th>\n",
       "      <th>c-2</th>\n",
       "      <th>c-3</th>\n",
       "      <th>...</th>\n",
       "      <th>g-762</th>\n",
       "      <th>g-763</th>\n",
       "      <th>g-764</th>\n",
       "      <th>g-765</th>\n",
       "      <th>g-766</th>\n",
       "      <th>g-767</th>\n",
       "      <th>g-768</th>\n",
       "      <th>g-769</th>\n",
       "      <th>g-770</th>\n",
       "      <th>g-771</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>[-0.05545241691160695]</td>\n",
       "      <td>[2.1199877096858333]</td>\n",
       "      <td>[3.5271482203135216]</td>\n",
       "      <td>[3.653198715312279]</td>\n",
       "      <td>[0.24798512687284416]</td>\n",
       "      <td>[0.99958100092857]</td>\n",
       "      <td>[0.3584368250253535]</td>\n",
       "      <td>[-0.33353422802050986]</td>\n",
       "      <td>[-0.38113551132676826]</td>\n",
       "      <td>[0.7091074024443236]</td>\n",
       "      <td>...</td>\n",
       "      <td>[0.8749090264333937]</td>\n",
       "      <td>[3.9134005806488834]</td>\n",
       "      <td>[-0.013360612907235822]</td>\n",
       "      <td>[1.998728818097307]</td>\n",
       "      <td>[1.1963255070951602]</td>\n",
       "      <td>[1.4745708244240772]</td>\n",
       "      <td>[0.9484212982815089]</td>\n",
       "      <td>[1.2338218095092064]</td>\n",
       "      <td>[0.9618653384274806]</td>\n",
       "      <td>[1.593447621983805]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>[0.4508431468627332]</td>\n",
       "      <td>[2.873485045489368]</td>\n",
       "      <td>[2.9652882798202773]</td>\n",
       "      <td>[3.2701409307817215]</td>\n",
       "      <td>[3.8364220715811967]</td>\n",
       "      <td>[0.9990469869683756]</td>\n",
       "      <td>[-0.24276896443051477]</td>\n",
       "      <td>[-0.17584555663391227]</td>\n",
       "      <td>[-0.09340903850804319]</td>\n",
       "      <td>[-0.15055492453984126]</td>\n",
       "      <td>...</td>\n",
       "      <td>[1.1234876098069022]</td>\n",
       "      <td>[0.9783480523572948]</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[0.6425866206498884]</td>\n",
       "      <td>[-0.11338716529274412]</td>\n",
       "      <td>[0.5954042392528222]</td>\n",
       "      <td>[-0.0367879514866143]</td>\n",
       "      <td>[1.4282135378204226]</td>\n",
       "      <td>[-0.21471668822199522]</td>\n",
       "      <td>[0.4299493304933841]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[0.5588296295353339]</td>\n",
       "      <td>[1.2114260479070598]</td>\n",
       "      <td>[2.0293362075683783]</td>\n",
       "      <td>[0.9491503528663043]</td>\n",
       "      <td>[0.9996118950505938]</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[-0.6937736262063956]</td>\n",
       "      <td>[-0.04667038184078219]</td>\n",
       "      <td>[-0.5925380500991164]</td>\n",
       "      <td>...</td>\n",
       "      <td>[5.788151692534951]</td>\n",
       "      <td>[2.7926051300193055]</td>\n",
       "      <td>[0.06368329672739806]</td>\n",
       "      <td>[1.5946004453286233]</td>\n",
       "      <td>[4.913552866965104]</td>\n",
       "      <td>[0.19162081412667947]</td>\n",
       "      <td>[-0.09844020216902022]</td>\n",
       "      <td>[0.6618449783924507]</td>\n",
       "      <td>[-0.27918635370739786]</td>\n",
       "      <td>[0.4656207367885598]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>[0.0426137180178707]</td>\n",
       "      <td>[0.5504140823396886]</td>\n",
       "      <td>[3.7113987675694897]</td>\n",
       "      <td>[9.48382957453369]</td>\n",
       "      <td>[4.47779354037929]</td>\n",
       "      <td>[0.999483124750076]</td>\n",
       "      <td>[0.06604431351820879]</td>\n",
       "      <td>[-0.24232866670645634]</td>\n",
       "      <td>[0.09737170104846313]</td>\n",
       "      <td>[0.04854733305685826]</td>\n",
       "      <td>...</td>\n",
       "      <td>[-1.731097853854435]</td>\n",
       "      <td>[3.816077104419831]</td>\n",
       "      <td>[0.015288095598431715]</td>\n",
       "      <td>[1.3130385967948641]</td>\n",
       "      <td>[2.302586763906089]</td>\n",
       "      <td>[2.7182835191329793]</td>\n",
       "      <td>[-0.327208491625899]</td>\n",
       "      <td>[-0.6626519876377063]</td>\n",
       "      <td>[-0.5035431886749584]</td>\n",
       "      <td>[0.5250355946095726]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>[0.24410124395251065]</td>\n",
       "      <td>[-0.2176690097060736]</td>\n",
       "      <td>[2.9799582306740913]</td>\n",
       "      <td>[3.191714050714197]</td>\n",
       "      <td>[3.6493622766874414]</td>\n",
       "      <td>[0.9998932923168317]</td>\n",
       "      <td>[0.06610327255740722]</td>\n",
       "      <td>[0.15397887868935686]</td>\n",
       "      <td>[0.17268113054416961]</td>\n",
       "      <td>[-0.03283548422204124]</td>\n",
       "      <td>...</td>\n",
       "      <td>[0.7443678210568891]</td>\n",
       "      <td>[3.866481668597004]</td>\n",
       "      <td>[-0.07609707814174005]</td>\n",
       "      <td>[2.3216025935674516]</td>\n",
       "      <td>[1.1900319403529864]</td>\n",
       "      <td>[-2.739423235704266]</td>\n",
       "      <td>[0.07328868753990539]</td>\n",
       "      <td>[1.1671496413835754]</td>\n",
       "      <td>[2.011707225467895]</td>\n",
       "      <td>[1.8059200670074873]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>[0.6457402035564366]</td>\n",
       "      <td>[2.274219563995644]</td>\n",
       "      <td>[3.6887390328126792]</td>\n",
       "      <td>[10.746700448891072]</td>\n",
       "      <td>[9.98068432057992]</td>\n",
       "      <td>[0.9993569100743362]</td>\n",
       "      <td>[-0.06586094751585886]</td>\n",
       "      <td>[-0.05392791746758637]</td>\n",
       "      <td>[-0.15868257648870882]</td>\n",
       "      <td>[-0.10562081354962681]</td>\n",
       "      <td>...</td>\n",
       "      <td>[-4.621885630292439]</td>\n",
       "      <td>[1.5421440505467052]</td>\n",
       "      <td>[-0.5972204494116028]</td>\n",
       "      <td>[5.83811125819721]</td>\n",
       "      <td>[-3.0982747509699933]</td>\n",
       "      <td>[2.123410697435634]</td>\n",
       "      <td>[-0.0790811773891357]</td>\n",
       "      <td>[2.1045458492678057]</td>\n",
       "      <td>[1.2877144972670604]</td>\n",
       "      <td>[0.09800260303370109]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>[-0.09633325250316424]</td>\n",
       "      <td>[-1.0159244612729097]</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[2.0674422304904128]</td>\n",
       "      <td>[3.083527291780644]</td>\n",
       "      <td>[0.9993164002537506]</td>\n",
       "      <td>[-2.6130878767565955]</td>\n",
       "      <td>[2.1781666724041022]</td>\n",
       "      <td>[-2.766773929863105]</td>\n",
       "      <td>[-2.277124291085381]</td>\n",
       "      <td>...</td>\n",
       "      <td>[1.1344063508288114]</td>\n",
       "      <td>[1.843797389500503]</td>\n",
       "      <td>[-0.10084825653756632]</td>\n",
       "      <td>[0.5115710785428866]</td>\n",
       "      <td>[2.1552417036740192]</td>\n",
       "      <td>[0.7134899749472898]</td>\n",
       "      <td>[0.13239595500284837]</td>\n",
       "      <td>[0.3863704167349505]</td>\n",
       "      <td>[-0.6482491808008887]</td>\n",
       "      <td>[0.6463436299323346]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[-4.399376163281369]</td>\n",
       "      <td>[4.2008719566817545]</td>\n",
       "      <td>[6.149729247185884]</td>\n",
       "      <td>[5.601577053711939]</td>\n",
       "      <td>[0.9991082626097362]</td>\n",
       "      <td>[0.0182075850767299]</td>\n",
       "      <td>[0.22619862323736076]</td>\n",
       "      <td>[0.22508599472841737]</td>\n",
       "      <td>[0.052571926797061345]</td>\n",
       "      <td>...</td>\n",
       "      <td>[0.8217197428907099]</td>\n",
       "      <td>[2.5435900514602476]</td>\n",
       "      <td>[-0.07574055307376949]</td>\n",
       "      <td>[1.510107783564854]</td>\n",
       "      <td>[2.212027335090691]</td>\n",
       "      <td>[1.8527335953800863]</td>\n",
       "      <td>[-0.09520844547898685]</td>\n",
       "      <td>[0.015883639267998108]</td>\n",
       "      <td>[0.0908405165418373]</td>\n",
       "      <td>[0.4244739953961869]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>[-0.24625160788576905]</td>\n",
       "      <td>[0.9462396892893684]</td>\n",
       "      <td>[0.6890876083519271]</td>\n",
       "      <td>[1.808495087128961]</td>\n",
       "      <td>[1.0718481833115392]</td>\n",
       "      <td>[0.9995233638360629]</td>\n",
       "      <td>[1.5160315273437244]</td>\n",
       "      <td>[1.051277074764439]</td>\n",
       "      <td>[1.1874534074270884]</td>\n",
       "      <td>[0.9437537396251652]</td>\n",
       "      <td>...</td>\n",
       "      <td>[0.9649616429039903]</td>\n",
       "      <td>[2.709557624743327]</td>\n",
       "      <td>[-0.03092638516197354]</td>\n",
       "      <td>[3.702824636079579]</td>\n",
       "      <td>[0.0681840002002953]</td>\n",
       "      <td>[1.038348131104793]</td>\n",
       "      <td>[0.0]</td>\n",
       "      <td>[5.39155902323272]</td>\n",
       "      <td>[0.6943104635201732]</td>\n",
       "      <td>[1.7526399545926277]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>[0.44366848070881226]</td>\n",
       "      <td>[-0.33402716638056346]</td>\n",
       "      <td>[0.8576992152801075]</td>\n",
       "      <td>[1.169437054441907]</td>\n",
       "      <td>[1.121871152532458]</td>\n",
       "      <td>[0.9987961461192264]</td>\n",
       "      <td>[-0.7304594274112479]</td>\n",
       "      <td>[-0.5612566026451028]</td>\n",
       "      <td>[-0.5702166753003511]</td>\n",
       "      <td>[-0.5695919341003639]</td>\n",
       "      <td>...</td>\n",
       "      <td>[-7.785759929948195]</td>\n",
       "      <td>[1.5042033486114101]</td>\n",
       "      <td>[-0.07397717952066662]</td>\n",
       "      <td>[1.7196716524206275]</td>\n",
       "      <td>[6.173270138577494]</td>\n",
       "      <td>[0.6095348657910896]</td>\n",
       "      <td>[0.3671648751341954]</td>\n",
       "      <td>[0.508575311397899]</td>\n",
       "      <td>[2.217889953633818]</td>\n",
       "      <td>[0.4647836914218902]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3060 rows × 5828 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                   cp_type                 cp_dose            cp_type_24  \\\n",
       "0   [-0.05545241691160695]    [2.1199877096858333]  [3.5271482203135216]   \n",
       "0     [0.4508431468627332]     [2.873485045489368]  [2.9652882798202773]   \n",
       "0                    [0.0]    [0.5588296295353339]  [1.2114260479070598]   \n",
       "0     [0.0426137180178707]    [0.5504140823396886]  [3.7113987675694897]   \n",
       "0    [0.24410124395251065]   [-0.2176690097060736]  [2.9799582306740913]   \n",
       "..                     ...                     ...                   ...   \n",
       "0     [0.6457402035564366]     [2.274219563995644]  [3.6887390328126792]   \n",
       "0   [-0.09633325250316424]   [-1.0159244612729097]                 [0.0]   \n",
       "0                    [0.0]    [-4.399376163281369]  [4.2008719566817545]   \n",
       "0   [-0.24625160788576905]    [0.9462396892893684]  [0.6890876083519271]   \n",
       "0    [0.44366848070881226]  [-0.33402716638056346]  [0.8576992152801075]   \n",
       "\n",
       "              cp_type_48             cp_type_72                     1  \\\n",
       "0    [3.653198715312279]  [0.24798512687284416]    [0.99958100092857]   \n",
       "0   [3.2701409307817215]   [3.8364220715811967]  [0.9990469869683756]   \n",
       "0   [2.0293362075683783]   [0.9491503528663043]  [0.9996118950505938]   \n",
       "0     [9.48382957453369]     [4.47779354037929]   [0.999483124750076]   \n",
       "0    [3.191714050714197]   [3.6493622766874414]  [0.9998932923168317]   \n",
       "..                   ...                    ...                   ...   \n",
       "0   [10.746700448891072]     [9.98068432057992]  [0.9993569100743362]   \n",
       "0   [2.0674422304904128]    [3.083527291780644]  [0.9993164002537506]   \n",
       "0    [6.149729247185884]    [5.601577053711939]  [0.9991082626097362]   \n",
       "0    [1.808495087128961]   [1.0718481833115392]  [0.9995233638360629]   \n",
       "0    [1.169437054441907]    [1.121871152532458]  [0.9987961461192264]   \n",
       "\n",
       "                       c-0                     c-1                     c-2  \\\n",
       "0     [0.3584368250253535]  [-0.33353422802050986]  [-0.38113551132676826]   \n",
       "0   [-0.24276896443051477]  [-0.17584555663391227]  [-0.09340903850804319]   \n",
       "0                    [0.0]   [-0.6937736262063956]  [-0.04667038184078219]   \n",
       "0    [0.06604431351820879]  [-0.24232866670645634]   [0.09737170104846313]   \n",
       "0    [0.06610327255740722]   [0.15397887868935686]   [0.17268113054416961]   \n",
       "..                     ...                     ...                     ...   \n",
       "0   [-0.06586094751585886]  [-0.05392791746758637]  [-0.15868257648870882]   \n",
       "0    [-2.6130878767565955]    [2.1781666724041022]    [-2.766773929863105]   \n",
       "0     [0.0182075850767299]   [0.22619862323736076]   [0.22508599472841737]   \n",
       "0     [1.5160315273437244]     [1.051277074764439]    [1.1874534074270884]   \n",
       "0    [-0.7304594274112479]   [-0.5612566026451028]   [-0.5702166753003511]   \n",
       "\n",
       "                       c-3  ...                 g-762                 g-763  \\\n",
       "0     [0.7091074024443236]  ...  [0.8749090264333937]  [3.9134005806488834]   \n",
       "0   [-0.15055492453984126]  ...  [1.1234876098069022]  [0.9783480523572948]   \n",
       "0    [-0.5925380500991164]  ...   [5.788151692534951]  [2.7926051300193055]   \n",
       "0    [0.04854733305685826]  ...  [-1.731097853854435]   [3.816077104419831]   \n",
       "0   [-0.03283548422204124]  ...  [0.7443678210568891]   [3.866481668597004]   \n",
       "..                     ...  ...                   ...                   ...   \n",
       "0   [-0.10562081354962681]  ...  [-4.621885630292439]  [1.5421440505467052]   \n",
       "0     [-2.277124291085381]  ...  [1.1344063508288114]   [1.843797389500503]   \n",
       "0   [0.052571926797061345]  ...  [0.8217197428907099]  [2.5435900514602476]   \n",
       "0     [0.9437537396251652]  ...  [0.9649616429039903]   [2.709557624743327]   \n",
       "0    [-0.5695919341003639]  ...  [-7.785759929948195]  [1.5042033486114101]   \n",
       "\n",
       "                      g-764                 g-765                   g-766  \\\n",
       "0   [-0.013360612907235822]   [1.998728818097307]    [1.1963255070951602]   \n",
       "0                     [0.0]  [0.6425866206498884]  [-0.11338716529274412]   \n",
       "0     [0.06368329672739806]  [1.5946004453286233]     [4.913552866965104]   \n",
       "0    [0.015288095598431715]  [1.3130385967948641]     [2.302586763906089]   \n",
       "0    [-0.07609707814174005]  [2.3216025935674516]    [1.1900319403529864]   \n",
       "..                      ...                   ...                     ...   \n",
       "0     [-0.5972204494116028]    [5.83811125819721]   [-3.0982747509699933]   \n",
       "0    [-0.10084825653756632]  [0.5115710785428866]    [2.1552417036740192]   \n",
       "0    [-0.07574055307376949]   [1.510107783564854]     [2.212027335090691]   \n",
       "0    [-0.03092638516197354]   [3.702824636079579]    [0.0681840002002953]   \n",
       "0    [-0.07397717952066662]  [1.7196716524206275]     [6.173270138577494]   \n",
       "\n",
       "                    g-767                   g-768                   g-769  \\\n",
       "0    [1.4745708244240772]    [0.9484212982815089]    [1.2338218095092064]   \n",
       "0    [0.5954042392528222]   [-0.0367879514866143]    [1.4282135378204226]   \n",
       "0   [0.19162081412667947]  [-0.09844020216902022]    [0.6618449783924507]   \n",
       "0    [2.7182835191329793]    [-0.327208491625899]   [-0.6626519876377063]   \n",
       "0    [-2.739423235704266]   [0.07328868753990539]    [1.1671496413835754]   \n",
       "..                    ...                     ...                     ...   \n",
       "0     [2.123410697435634]   [-0.0790811773891357]    [2.1045458492678057]   \n",
       "0    [0.7134899749472898]   [0.13239595500284837]    [0.3863704167349505]   \n",
       "0    [1.8527335953800863]  [-0.09520844547898685]  [0.015883639267998108]   \n",
       "0     [1.038348131104793]                   [0.0]      [5.39155902323272]   \n",
       "0    [0.6095348657910896]    [0.3671648751341954]     [0.508575311397899]   \n",
       "\n",
       "                     g-770                  g-771  \n",
       "0     [0.9618653384274806]    [1.593447621983805]  \n",
       "0   [-0.21471668822199522]   [0.4299493304933841]  \n",
       "0   [-0.27918635370739786]   [0.4656207367885598]  \n",
       "0    [-0.5035431886749584]   [0.5250355946095726]  \n",
       "0      [2.011707225467895]   [1.8059200670074873]  \n",
       "..                     ...                    ...  \n",
       "0     [1.2877144972670604]  [0.09800260303370109]  \n",
       "0    [-0.6482491808008887]   [0.6463436299323346]  \n",
       "0     [0.0908405165418373]   [0.4244739953961869]  \n",
       "0     [0.6943104635201732]   [1.7526399545926277]  \n",
       "0      [2.217889953633818]   [0.4647836914218902]  \n",
       "\n",
       "[3060 rows x 5828 columns]"
      ]
     },
     "execution_count": 77,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "coef_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "saved!\n",
      "done calculating the mean!\n",
      "done calculating std_dev!\n"
     ]
    },
    {
     "ename": "NameError",
     "evalue": "name 'std' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-83-a3868c2435b5>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0mcoef_df_std\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcoef_df\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mstd\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      6\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'done calculating std_dev!'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 7\u001b[0;31m \u001b[0mcoef_df_stats\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mpd\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mDataFrame\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mcoef_df_avg\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mcoef_df\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mstd\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      8\u001b[0m \u001b[0mcoef_df_stats\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mhead\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mNameError\u001b[0m: name 'std' is not defined"
     ]
    }
   ],
   "source": [
    "coef_df.to_csv('coef_df.csv')\n",
    "print('saved!')\n",
    "coef_df_avg = coef_df.mean()\n",
    "print('done calculating the mean!')\n",
    "coef_df_std = coef_df.std()\n",
    "print('done calculating std_dev!')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>cp_type</th>\n",
       "      <th>cp_dose</th>\n",
       "      <th>cp_type_24</th>\n",
       "      <th>cp_type_48</th>\n",
       "      <th>cp_type_72</th>\n",
       "      <th>1</th>\n",
       "      <th>c-0</th>\n",
       "      <th>c-1</th>\n",
       "      <th>c-2</th>\n",
       "      <th>c-3</th>\n",
       "      <th>...</th>\n",
       "      <th>g-762</th>\n",
       "      <th>g-763</th>\n",
       "      <th>g-764</th>\n",
       "      <th>g-765</th>\n",
       "      <th>g-766</th>\n",
       "      <th>g-767</th>\n",
       "      <th>g-768</th>\n",
       "      <th>g-769</th>\n",
       "      <th>g-770</th>\n",
       "      <th>g-771</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>avg</th>\n",
       "      <td>-1.576901</td>\n",
       "      <td>-0.294892</td>\n",
       "      <td>1.995278</td>\n",
       "      <td>4.422627</td>\n",
       "      <td>4.742534</td>\n",
       "      <td>0.941453</td>\n",
       "      <td>0.005768</td>\n",
       "      <td>-0.224598</td>\n",
       "      <td>-0.060883</td>\n",
       "      <td>0.253388</td>\n",
       "      <td>...</td>\n",
       "      <td>0.872295</td>\n",
       "      <td>2.324504</td>\n",
       "      <td>-0.168635</td>\n",
       "      <td>3.244462</td>\n",
       "      <td>3.910462</td>\n",
       "      <td>2.014526</td>\n",
       "      <td>0.521484</td>\n",
       "      <td>1.769784</td>\n",
       "      <td>0.721522</td>\n",
       "      <td>1.715793</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std_dev</th>\n",
       "      <td>7.104119</td>\n",
       "      <td>1.579887</td>\n",
       "      <td>4.180392</td>\n",
       "      <td>4.694272</td>\n",
       "      <td>5.205373</td>\n",
       "      <td>0.225002</td>\n",
       "      <td>1.571074</td>\n",
       "      <td>1.763629</td>\n",
       "      <td>1.363721</td>\n",
       "      <td>1.538444</td>\n",
       "      <td>...</td>\n",
       "      <td>4.020408</td>\n",
       "      <td>2.016622</td>\n",
       "      <td>0.518873</td>\n",
       "      <td>2.689194</td>\n",
       "      <td>4.967190</td>\n",
       "      <td>3.708776</td>\n",
       "      <td>1.124953</td>\n",
       "      <td>3.852678</td>\n",
       "      <td>1.791939</td>\n",
       "      <td>1.363272</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2 rows × 5828 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "          cp_type   cp_dose  cp_type_24  cp_type_48  cp_type_72         1  \\\n",
       "avg     -1.576901 -0.294892    1.995278    4.422627    4.742534  0.941453   \n",
       "std_dev  7.104119  1.579887    4.180392    4.694272    5.205373  0.225002   \n",
       "\n",
       "              c-0       c-1       c-2       c-3  ...     g-762     g-763  \\\n",
       "avg      0.005768 -0.224598 -0.060883  0.253388  ...  0.872295  2.324504   \n",
       "std_dev  1.571074  1.763629  1.363721  1.538444  ...  4.020408  2.016622   \n",
       "\n",
       "            g-764     g-765     g-766     g-767     g-768     g-769     g-770  \\\n",
       "avg     -0.168635  3.244462  3.910462  2.014526  0.521484  1.769784  0.721522   \n",
       "std_dev  0.518873  2.689194  4.967190  3.708776  1.124953  3.852678  1.791939   \n",
       "\n",
       "            g-771  \n",
       "avg      1.715793  \n",
       "std_dev  1.363272  \n",
       "\n",
       "[2 rows x 5828 columns]"
      ]
     },
     "execution_count": 85,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "coef_df_stats = pd.DataFrame([coef_df_avg,coef_df_std], index = ['avg','std_dev'])\n",
    "coef_df_stats.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [],
   "source": [
    "abs_stats = coef_df_stats.abs()\n",
    "#abs_stats.values = abs(abs_stats.values)\n",
    "abs_stats = abs_stats.sort_values(by = 'avg', axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2, 5828)"
      ]
     },
     "execution_count": 92,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "abs_stats.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "3088"
      ]
     },
     "execution_count": 112,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "drop_avg = [feat for feat in abs_stats.columns if abs_stats[feat]['avg'] < 0.125]\n",
    "drop_std_dev = [feat for feat in abs_stats.columns if abs_stats[feat]['std_dev'] < 0.125]\n",
    "drop_list = [feat for feat in drop_avg if feat in drop_std_dev]\n",
    "len(drop_list)\n",
    "#abs_stats[abs_stats[['avg']] > 0.1].shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((23814, 5828), (3982, 5828))"
      ]
     },
     "execution_count": 107,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X.shape, x_test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_df = pd.DataFrame(X, columns=feat_labels)\n",
    "x_test_df = pd.DataFrame(x_test, columns=feat_labels)\n",
    "\n",
    "X_df.drop(columns=drop_list, inplace = True)\n",
    "x_test_df.drop(columns=drop_list, inplace = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d8e18e5c32f34b59a4d2273879ec37dc",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(HTML(value=''), FloatProgress(value=0.0, max=206.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[W] [13:26:57.377369] QWL-QN: max iterations reached\n",
      "[W] [13:27:30.516795] QWL-QN: max iterations reached\n",
      "[W] [13:27:52.706437] QWL-QN: max iterations reached\n",
      "[W] [13:39:19.877055] QWL-QN: max iterations reached\n",
      "[W] [13:39:32.201073] QWL-QN: max iterations reached\n",
      "[W] [13:39:44.804299] QWL-QN: max iterations reached\n",
      "[W] [13:39:57.105934] QWL-QN: max iterations reached\n",
      "[W] [13:40:09.436435] QWL-QN: max iterations reached\n",
      "[W] [13:40:41.504399] QWL-QN: max iterations reached\n",
      "[W] [13:41:13.937157] QWL-QN: max iterations reached\n",
      "\n",
      "Log Loss: 0.04436834251935356\n",
      "After manually eliminating control groups: 0.04422945050685287\n",
      "\n",
      "\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from cuml.linear_model import LogisticRegression\n",
    "#from sklearn.linear_model import LogisticRegression\n",
    "\n",
    "feat_labels = d_tags + poly.get_feature_names(input_features=c_tags) + g_tags\n",
    "coef_df2 = pd.DataFrame(columns = list(X_df.columns))\n",
    "\n",
    "# column_names = [\"a\", \"b\", \"c\"]\n",
    "# coef_df = pd.DataFrame(columns = column_names)\n",
    "rowval = 0\n",
    "for i in [256]: # left for further hyperparameter tuning\n",
    "    from cuml.experimental.preprocessing import StandardScaler\n",
    "    scaler = StandardScaler()\n",
    "    #model = LogisticRegression(C = 128, max_iter = 100000, penalty='elasticnet', l1_ratio=1)\n",
    "    model = LogisticRegression(C = 128, penalty='elasticnet', l1_ratio=1)\n",
    "    target_counts = []\n",
    "    logistic_regr_loss_vals = []\n",
    "    ss_lr.loc[:, train_targets.columns] = 0\n",
    "    res_lr.loc[:, train_targets.columns] = 0\n",
    "    res_lr, ss_lr, logistic_regr_loss_vals, coef_df2 = run_score(train_targets, X_df.values, x_test_df.values, res_lr, ss_lr, logistic_regr_loss_vals, coef_df2)\n",
    "    \n",
    "plt.scatter(target_counts,logistic_regr_loss_vals)\n",
    "plt.xlabel('target_counts')\n",
    "plt.ylabel('log loss values')\n",
    "plt.title('cuML Logistic Regression')\n",
    "plt.show()\n",
    "res_lr.to_pickle('pickles/res_cuml_lr.p')\n",
    "ss_lr.to_pickle('pickles/ss_cuml_lr.p')"
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "RAPIDS_logistic_rergression_w_undersampling.ipynb",
   "provenance": [
    {
     "file_id": "1CP5eADhskrXuoXJreMbYrVeEAx_clw4r",
     "timestamp": 1602973236795
    }
   ]
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.6"
  },
  "widgets": {
   "application/vnd.jupyter.widget-state+json": {
    "0254a811c551456ea72cef2362a8d2da": {
     "model_module": "@jupyter-widgets/controls",
     "model_name": "ProgressStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "ProgressStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "bar_color": null,
      "description_width": ""
     }
    },
    "0b03083b658f4e478de3b57791f42da8": {
     "model_module": "@jupyter-widgets/base",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "1abb228a8d164b2a892055bc77aa56fd": {
     "model_module": "@jupyter-widgets/controls",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_4baa80741801474695d594eb4f4bb215",
      "placeholder": "​",
      "style": "IPY_MODEL_1e29afe4e9fe43fb8d854cff56fde3af",
      "value": " 206/206 [1:12:39&lt;00:00, 21.16s/it]"
     }
    },
    "1e29afe4e9fe43fb8d854cff56fde3af": {
     "model_module": "@jupyter-widgets/controls",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "1f77a084b822458699718400bf5a291b": {
     "model_module": "@jupyter-widgets/controls",
     "model_name": "ProgressStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "ProgressStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "bar_color": null,
      "description_width": ""
     }
    },
    "25286e32af134ccfa3f3770f82a36c16": {
     "model_module": "@jupyter-widgets/base",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "259ba841852240529c05cb6774efcc3f": {
     "model_module": "@jupyter-widgets/controls",
     "model_name": "HBoxModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HBoxModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HBoxView",
      "box_style": "",
      "children": [
       "IPY_MODEL_ba3f897e98e84be5a180ef100aa6d934",
       "IPY_MODEL_6887ebab40d248c2b345cb7417ee2a9a",
       "IPY_MODEL_1abb228a8d164b2a892055bc77aa56fd"
      ],
      "layout": "IPY_MODEL_0b03083b658f4e478de3b57791f42da8"
     }
    },
    "3015131d021f426699d2e82071e50786": {
     "model_module": "@jupyter-widgets/controls",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_317612f56be54ddd97851c95edc4dabd",
      "placeholder": "​",
      "style": "IPY_MODEL_8f12a242d04d4a3dadf321c146a9333d",
      "value": "  0%"
     }
    },
    "317612f56be54ddd97851c95edc4dabd": {
     "model_module": "@jupyter-widgets/base",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "3f3d40d522ef4253835e90175025d687": {
     "model_module": "@jupyter-widgets/base",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "432cbfdc404a4bf5b1b900515cda4e86": {
     "model_module": "@jupyter-widgets/controls",
     "model_name": "HBoxModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HBoxModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HBoxView",
      "box_style": "",
      "children": [
       "IPY_MODEL_3015131d021f426699d2e82071e50786",
       "IPY_MODEL_7e83495b6d7749b189cfa5aea250ae60",
       "IPY_MODEL_6669c81dca1543f79898fa66a3e5a7b5"
      ],
      "layout": "IPY_MODEL_b45bb3a2b52f4880b0ad58e700721365"
     }
    },
    "4baa80741801474695d594eb4f4bb215": {
     "model_module": "@jupyter-widgets/base",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "5c177f172ae8443a8b71a76cdff14220": {
     "model_module": "@jupyter-widgets/controls",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "62b032770c3546498354c2319cc29917": {
     "model_module": "@jupyter-widgets/controls",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "6669c81dca1543f79898fa66a3e5a7b5": {
     "model_module": "@jupyter-widgets/controls",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_3f3d40d522ef4253835e90175025d687",
      "placeholder": "​",
      "style": "IPY_MODEL_62b032770c3546498354c2319cc29917",
      "value": " 0/206 [00:00&lt;?, ?it/s]"
     }
    },
    "6887ebab40d248c2b345cb7417ee2a9a": {
     "model_module": "@jupyter-widgets/controls",
     "model_name": "FloatProgressModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "FloatProgressModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "ProgressView",
      "bar_style": "success",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_25286e32af134ccfa3f3770f82a36c16",
      "max": 206,
      "min": 0,
      "orientation": "horizontal",
      "style": "IPY_MODEL_1f77a084b822458699718400bf5a291b",
      "value": 206
     }
    },
    "7299f41bbc7c493fac12c6500da94599": {
     "model_module": "@jupyter-widgets/base",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "7e83495b6d7749b189cfa5aea250ae60": {
     "model_module": "@jupyter-widgets/controls",
     "model_name": "FloatProgressModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "FloatProgressModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "ProgressView",
      "bar_style": "",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_c12b441719b146b1be880df7d93b0ec7",
      "max": 206,
      "min": 0,
      "orientation": "horizontal",
      "style": "IPY_MODEL_0254a811c551456ea72cef2362a8d2da",
      "value": 0
     }
    },
    "8f12a242d04d4a3dadf321c146a9333d": {
     "model_module": "@jupyter-widgets/controls",
     "model_name": "DescriptionStyleModel",
     "state": {
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "DescriptionStyleModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "StyleView",
      "description_width": ""
     }
    },
    "b45bb3a2b52f4880b0ad58e700721365": {
     "model_module": "@jupyter-widgets/base",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    },
    "ba3f897e98e84be5a180ef100aa6d934": {
     "model_module": "@jupyter-widgets/controls",
     "model_name": "HTMLModel",
     "state": {
      "_dom_classes": [],
      "_model_module": "@jupyter-widgets/controls",
      "_model_module_version": "1.5.0",
      "_model_name": "HTMLModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/controls",
      "_view_module_version": "1.5.0",
      "_view_name": "HTMLView",
      "description": "",
      "description_tooltip": null,
      "layout": "IPY_MODEL_7299f41bbc7c493fac12c6500da94599",
      "placeholder": "​",
      "style": "IPY_MODEL_5c177f172ae8443a8b71a76cdff14220",
      "value": "100%"
     }
    },
    "c12b441719b146b1be880df7d93b0ec7": {
     "model_module": "@jupyter-widgets/base",
     "model_name": "LayoutModel",
     "state": {
      "_model_module": "@jupyter-widgets/base",
      "_model_module_version": "1.2.0",
      "_model_name": "LayoutModel",
      "_view_count": null,
      "_view_module": "@jupyter-widgets/base",
      "_view_module_version": "1.2.0",
      "_view_name": "LayoutView",
      "align_content": null,
      "align_items": null,
      "align_self": null,
      "border": null,
      "bottom": null,
      "display": null,
      "flex": null,
      "flex_flow": null,
      "grid_area": null,
      "grid_auto_columns": null,
      "grid_auto_flow": null,
      "grid_auto_rows": null,
      "grid_column": null,
      "grid_gap": null,
      "grid_row": null,
      "grid_template_areas": null,
      "grid_template_columns": null,
      "grid_template_rows": null,
      "height": null,
      "justify_content": null,
      "justify_items": null,
      "left": null,
      "margin": null,
      "max_height": null,
      "max_width": null,
      "min_height": null,
      "min_width": null,
      "object_fit": null,
      "object_position": null,
      "order": null,
      "overflow": null,
      "overflow_x": null,
      "overflow_y": null,
      "padding": null,
      "right": null,
      "top": null,
      "visibility": null,
      "width": null
     }
    }
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}