4-analyze-hatespeech.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "j0a4mTk9o1Qg"
   },
   "outputs": [],
   "source": [
    "# Copyright 2019 Google Inc.\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."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load the predictions of models, load real labels and count the F1-values"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Version for Hatespeech"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "#from tensorflow import keras\n",
    "#import tensorflow as tf\n",
    "#import tensorflow_hub as hub\n",
    "\n",
    "import os\n",
    "import re\n",
    "import pandas as pd\n",
    "\n",
    "from sklearn.model_selection import train_test_split\n",
    "import pandas as pd\n",
    "\n",
    "from datetime import datetime\n",
    "\n",
    "# predicted logits to distribution\n",
    "import numpy as np\n",
    "\n",
    "# confusion matrix\n",
    "from matplotlib import pyplot as plt\n",
    "import itertools\n",
    "# vs import matplotlib.pyplot as plt\n",
    "\n",
    "from sklearn.metrics import f1_score\n",
    "\n",
    "from sklearn.metrics import confusion_matrix\n",
    "\n",
    "from sklearn.metrics import precision_score, recall_score\n",
    "\n",
    "# avoid printing e-notation, such as 9.9687493e-01. Instead print 0.9968\n",
    "np.set_printoptions(suppress=True,\n",
    "   formatter={'float_kind':'{:0.4f}'.format})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_confusion_matrix(cm, classes, normalize=False, title='Confusion matrix', cmap=plt.cm.Blues):\n",
    "    \"\"\"\n",
    "    This function prints and plots the confusion matrix.\n",
    "    Normalization can be applied by setting `normalize=True`.\n",
    "    (This function is copied from the scikit docs.)\n",
    "    \"\"\"\n",
    "    plt.figure()\n",
    "    plt.imshow(cm, interpolation='nearest', cmap=cmap)\n",
    "    plt.title(title)\n",
    "    plt.colorbar()\n",
    "    tick_marks = np.arange(len(classes))\n",
    "    plt.xticks(tick_marks, classes, rotation=45)\n",
    "    plt.yticks(tick_marks, classes)\n",
    "\n",
    "    if normalize:\n",
    "        cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]\n",
    "    print(cm)\n",
    "    thresh = cm.max() / 2.\n",
    "    for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):\n",
    "        plt.text(j, i, cm[i, j], horizontalalignment=\"center\", color=\"white\" if cm[i, j] > thresh else \"black\")\n",
    "\n",
    "    plt.tight_layout()\n",
    "    plt.ylabel('True label')\n",
    "    plt.xlabel('Predicted label')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset = 'hatespeech'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "/home/max/git/modelcompare/dataset_hatespeech/\n"
     ]
    }
   ],
   "source": [
    "current = os.getcwd()\n",
    "basefolder = current + '/dataset_'+ dataset+'/'\n",
    "datafolder = basefolder + 'data/'  # for example /dataset_businessnews/data/\n",
    "print(basefolder)\n",
    "\n",
    "infolder =  basefolder + 'input/'\n",
    "outfolder = basefolder + 'output/'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### True labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "test = pd.read_csv(basefolder+'input/test.csv'  ,sep='\\t', header = None)\n",
    "test.columns  = ['id','label', 'text']\n",
    "true_labels = test['label'].astype(int)  # they are type object, turn to int"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "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>id</th>\n",
       "      <th>label</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>23437</td>\n",
       "      <td>1</td>\n",
       "      <td>Wizard turned into a hoe after a break up</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>17813</td>\n",
       "      <td>1</td>\n",
       "      <td>RT  MENTION :  Captain Kirk hoe !</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2561</td>\n",
       "      <td>1</td>\n",
       "      <td>MENTION bruh did go out like a bitch thoe lol...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>18228</td>\n",
       "      <td>1</td>\n",
       "      <td>RT  MENTION :  RT  MENTION :  when you tryna g...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>23030</td>\n",
       "      <td>1</td>\n",
       "      <td>Well I wish I could shoot the shit wit u bitch...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      id  label                                               text\n",
       "0  23437      1          Wizard turned into a hoe after a break up\n",
       "1  17813      1                 RT  MENTION :  Captain Kirk hoe ! \n",
       "2   2561      1   MENTION bruh did go out like a bitch thoe lol...\n",
       "3  18228      1  RT  MENTION :  RT  MENTION :  when you tryna g...\n",
       "4  23030      1  Well I wish I could shoot the shit wit u bitch..."
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Counter({1: 2279, 2: 548, 0: 173})"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Counts, Majority\n",
    "\n",
    "# 0 - hate speech\n",
    "# 1 - offensive language\n",
    "# 2 - neither\n",
    "\n",
    "from collections import Counter\n",
    "Counter(true_labels)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.7596666666666667"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "2279 / (2279+548+173)\n",
    "0.7596666666666667\n",
    "\n",
    "# Majority prediction gives accuracy 0.75966"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "f1 majority: 0.7596666666666667\n"
     ]
    }
   ],
   "source": [
    "### count F1 for majority class.\n",
    "# majority, 0.76 is 1\n",
    "\n",
    "length = len(true_labels)\n",
    "#yhat = pd.Series(np.ones(length).astype(int))\n",
    "\n",
    "yhat = pd.Series( [1] * length) # value 1, times length, turned into Series\n",
    "f1_majority = f1_score(true_labels, yhat, average='micro')\n",
    "print('f1 majority:',f1_majority)\n",
    "\n",
    "precision_majority = precision_score(true_labels, yhat, average='micro', labels=[0,1,2])\n",
    "recall_majority = recall_score(true_labels, yhat, average='micro')\n",
    "\n",
    "#names.append('Majority')\n",
    "#scores.append(f1)\n",
    "#traintimes.append(0)\n",
    "#precisions.append()\n",
    "#recalls.append(recall_score(true_labels, yhat, average='micro'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "import shelve\n",
    "def loadResults_tfhub():\n",
    "    import shelve\n",
    "    shelf = shelve.open(outfolder+\"all.shlf\") # the same filename that you used before, please\n",
    "    new = shelf[\"all\"]\n",
    "    shelf.close()\n",
    "    return new"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "def loadResults_flair():\n",
    "    import shelve\n",
    "    shelf = shelve.open(outfolder+\"all_flair.shlf\") # the same filename that you used before, please\n",
    "    new = shelf[\"all_flair\"]\n",
    "    shelf.close()\n",
    "    return new"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "def loadResults_flair(name='all_default'):\n",
    "    import shelve\n",
    "    filename= name+'.shlf'\n",
    "    shelf = shelve.open(outfolder+filename) \n",
    "    new = shelf[name]\n",
    "    shelf.close()\n",
    "    return new"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "def saveResults_flair(all, name='all_default'):\n",
    "    filename= name+'.shlf'\n",
    "    shelf = shelve.open(outfolder+filename)\n",
    "    #shelf = shelve.open(\"all_flair.shlf\")\n",
    "    # serializing\n",
    "    #shelf[\"all_flair\"] = all_flair\n",
    "    shelf[name] = all_flair\n",
    "    shelf.close() # you must close the shelve file!!!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "def loadResults_flair(name='all_default'):\n",
    "    filename= name+'.shlf'\n",
    "    shelf = shelve.open(outfolder+filename) \n",
    "    new = shelf[name]\n",
    "    shelf.close()\n",
    "    return new"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'/home/max/git/newcombined/dataset_hatespeech/output/'"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "outfolder"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### LOAD all sizes and combine"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "    # PROCESS LISTS FOR ONE SIZE OF TRAININ DATA\n",
    "    \n",
    "def unpack(all):\n",
    "        # FLAIR\n",
    "        # unpack values from dict / list structure: names and f1-scores of models.\n",
    "        names = []\n",
    "        scores= []\n",
    "        traintimes=[]\n",
    "        predtimes=[]\n",
    "        precisions = []\n",
    "        recalls = []\n",
    "\n",
    "        for i in range(len(all)):\n",
    "            yhat = all[i]['labels']\n",
    "            all[i]['f1'] = f1_score(true_labels, yhat, average='micro')\n",
    "    \n",
    "            names.append(all[i]['model'])\n",
    "            scores.append(f1_score(true_labels, yhat, average='micro'))\n",
    "            traintimes.append(all[i]['traintime'])\n",
    "            predtimes.append(all[i]['predtime3k'])\n",
    "            precisions.append(precision_score(true_labels, yhat, average='micro', labels=[0,1,2]))\n",
    "            recalls.append(recall_score(true_labels, yhat, average='micro'))            \n",
    "        \n",
    "        return(names, scores, traintimes, predtimes, precisions, recalls)\n",
    "    "
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "    # Add Majority to beginning of list\n",
    "    names.insert(0,'Majority')\n",
    "    scores.insert(0,f1_majority)\n",
    "    traintimes.insert(0,0)\n",
    "    \n",
    "    # instead, add it as a column"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "FINAL_flair_all_trainsz_100\n",
      "FINAL_flair_all_trainsz_200\n",
      "FINAL_flair_all_trainsz_500\n",
      "FINAL_flair_all_trainsz_1000\n",
      "FINAL_flair_all_trainsz_3000\n",
      "FINAL_flair_all_trainsz_7000\n",
      "FINAL_flair_all_trainsz_18783\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import precision_score, recall_score\n",
    "\n",
    "sizes = [100, 200, 500, 1000, 3000, 7000, 18783]\n",
    "\n",
    "\n",
    "#randomized 501 to compare and confirm \n",
    "#sizes = [100, 200, 500, 1000, 1001, 3000, 7000, 18783]\n",
    "\n",
    "#sizes = [100, 200, 500, 1000, 3000, 18783]\n",
    "#sizes = [100]\n",
    "\n",
    "first= True\n",
    "\n",
    "namestart = 'FINAL_flair_all_trainsz_'\n",
    "for size in sizes:\n",
    "    name = namestart+str(size)\n",
    "    print(name)\n",
    "    \n",
    "    #LOAD IT\n",
    "    all = loadResults_flair(name)\n",
    "\n",
    "    names, scores, traintimes, predtimes, precisions, recalls = unpack(all)\n",
    "    \n",
    "\n",
    "\n",
    "    \n",
    "    df_size = pd.DataFrame(list(zip(names, scores)), #, traintimes, precisions, recalls\n",
    "                   #columns =['Name', 'score', 'traintime', 'precision', 'recall']) \n",
    "            #columns =['Model', 'trn_'+str(size)]) \n",
    "             columns =['Model', str(size)]) \n",
    "\n",
    "    if first:\n",
    "        df = df_size\n",
    "        first=False\n",
    "    else:\n",
    "        df = pd.merge(df, df_size, on='Model')\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "metadata": {},
   "outputs": [],
   "source": [
    "# add column with f1 value of majority class as baseline\n",
    "df.insert(1, '0', 0.760)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {},
   "outputs": [],
   "source": [
    "# manually fix name of bert \n",
    "df.replace('bert-base-uncased', 'bert-base-cased', inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "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>Model</th>\n",
       "      <th>0</th>\n",
       "      <th>100</th>\n",
       "      <th>200</th>\n",
       "      <th>500</th>\n",
       "      <th>1000</th>\n",
       "      <th>3000</th>\n",
       "      <th>7000</th>\n",
       "      <th>18783</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>glove</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.763</td>\n",
       "      <td>0.768</td>\n",
       "      <td>0.798</td>\n",
       "      <td>0.862</td>\n",
       "      <td>0.885</td>\n",
       "      <td>0.892</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>fasttext web-crawl</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.761</td>\n",
       "      <td>0.764</td>\n",
       "      <td>0.837</td>\n",
       "      <td>0.902</td>\n",
       "      <td>0.907</td>\n",
       "      <td>0.914</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>fasttext news/wiki</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.767</td>\n",
       "      <td>0.879</td>\n",
       "      <td>0.895</td>\n",
       "      <td>0.903</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>en-twitter</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.765</td>\n",
       "      <td>0.770</td>\n",
       "      <td>0.797</td>\n",
       "      <td>0.810</td>\n",
       "      <td>0.892</td>\n",
       "      <td>0.902</td>\n",
       "      <td>0.907</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>elmo</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.776</td>\n",
       "      <td>0.787</td>\n",
       "      <td>0.792</td>\n",
       "      <td>0.812</td>\n",
       "      <td>0.866</td>\n",
       "      <td>0.902</td>\n",
       "      <td>0.909</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Flair</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.761</td>\n",
       "      <td>0.776</td>\n",
       "      <td>0.787</td>\n",
       "      <td>0.812</td>\n",
       "      <td>0.829</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>bert-base-cased</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.775</td>\n",
       "      <td>0.793</td>\n",
       "      <td>0.816</td>\n",
       "      <td>0.797</td>\n",
       "      <td>0.888</td>\n",
       "      <td>0.907</td>\n",
       "      <td>0.914</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>BytePairEmbedding</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.765</td>\n",
       "      <td>0.766</td>\n",
       "      <td>0.777</td>\n",
       "      <td>0.806</td>\n",
       "      <td>0.867</td>\n",
       "      <td>0.885</td>\n",
       "      <td>0.899</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>gpt-1</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.789</td>\n",
       "      <td>0.791</td>\n",
       "      <td>0.816</td>\n",
       "      <td>0.879</td>\n",
       "      <td>0.892</td>\n",
       "      <td>0.905</td>\n",
       "      <td>0.917</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                Model     0   100   200   500  1000  3000  7000  18783\n",
       "0               glove 0.760 0.760 0.763 0.768 0.798 0.862 0.885  0.892\n",
       "1  fasttext web-crawl 0.760 0.760 0.761 0.764 0.837 0.902 0.907  0.914\n",
       "2  fasttext news/wiki 0.760 0.760 0.760 0.760 0.767 0.879 0.895  0.903\n",
       "3          en-twitter 0.760 0.765 0.770 0.797 0.810 0.892 0.902  0.907\n",
       "4                elmo 0.760 0.776 0.787 0.792 0.812 0.866 0.902  0.909\n",
       "5               Flair 0.760 0.760 0.760 0.761 0.776 0.787 0.812  0.829\n",
       "6     bert-base-cased 0.760 0.775 0.793 0.816 0.797 0.888 0.907  0.914\n",
       "7   BytePairEmbedding 0.760 0.765 0.766 0.777 0.806 0.867 0.885  0.899\n",
       "8               gpt-1 0.760 0.789 0.791 0.816 0.879 0.892 0.905  0.917"
      ]
     },
     "execution_count": 116,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.options.display.float_format = \"{:,.3f}\".format\n",
    "df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Write pandas DF into latex table"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 117,
   "metadata": {},
   "outputs": [],
   "source": [
    "# https://tex.stackexchange.com/questions/394123/export-dataframe-to-table-in-latex\n",
    "\n",
    "with open('mytable_f1_trainsize.tex','w') as tf:\n",
    "    tf.write(df.to_latex())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 118,
   "metadata": {},
   "outputs": [],
   "source": [
    "#df.to_latex()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Present it as a graph"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 119,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Write out before setting Model as inxed? puts them headers on 2 lines\n",
    "df = df.set_index('Model')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "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>0</th>\n",
       "      <th>100</th>\n",
       "      <th>200</th>\n",
       "      <th>500</th>\n",
       "      <th>1000</th>\n",
       "      <th>3000</th>\n",
       "      <th>7000</th>\n",
       "      <th>18783</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Model</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>glove</th>\n",
       "      <td>0.760</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.763</td>\n",
       "      <td>0.768</td>\n",
       "      <td>0.798</td>\n",
       "      <td>0.862</td>\n",
       "      <td>0.885</td>\n",
       "      <td>0.892</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>fasttext web-crawl</th>\n",
       "      <td>0.760</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.761</td>\n",
       "      <td>0.764</td>\n",
       "      <td>0.837</td>\n",
       "      <td>0.902</td>\n",
       "      <td>0.907</td>\n",
       "      <td>0.914</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>fasttext news/wiki</th>\n",
       "      <td>0.760</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.767</td>\n",
       "      <td>0.879</td>\n",
       "      <td>0.895</td>\n",
       "      <td>0.903</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>en-twitter</th>\n",
       "      <td>0.760</td>\n",
       "      <td>0.765</td>\n",
       "      <td>0.770</td>\n",
       "      <td>0.797</td>\n",
       "      <td>0.810</td>\n",
       "      <td>0.892</td>\n",
       "      <td>0.902</td>\n",
       "      <td>0.907</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>elmo</th>\n",
       "      <td>0.760</td>\n",
       "      <td>0.776</td>\n",
       "      <td>0.787</td>\n",
       "      <td>0.792</td>\n",
       "      <td>0.812</td>\n",
       "      <td>0.866</td>\n",
       "      <td>0.902</td>\n",
       "      <td>0.909</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Flair</th>\n",
       "      <td>0.760</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.760</td>\n",
       "      <td>0.761</td>\n",
       "      <td>0.776</td>\n",
       "      <td>0.787</td>\n",
       "      <td>0.812</td>\n",
       "      <td>0.829</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>bert-base-cased</th>\n",
       "      <td>0.760</td>\n",
       "      <td>0.775</td>\n",
       "      <td>0.793</td>\n",
       "      <td>0.816</td>\n",
       "      <td>0.797</td>\n",
       "      <td>0.888</td>\n",
       "      <td>0.907</td>\n",
       "      <td>0.914</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>BytePairEmbedding</th>\n",
       "      <td>0.760</td>\n",
       "      <td>0.765</td>\n",
       "      <td>0.766</td>\n",
       "      <td>0.777</td>\n",
       "      <td>0.806</td>\n",
       "      <td>0.867</td>\n",
       "      <td>0.885</td>\n",
       "      <td>0.899</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>gpt-1</th>\n",
       "      <td>0.760</td>\n",
       "      <td>0.789</td>\n",
       "      <td>0.791</td>\n",
       "      <td>0.816</td>\n",
       "      <td>0.879</td>\n",
       "      <td>0.892</td>\n",
       "      <td>0.905</td>\n",
       "      <td>0.917</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                       0   100   200   500  1000  3000  7000  18783\n",
       "Model                                                              \n",
       "glove              0.760 0.760 0.763 0.768 0.798 0.862 0.885  0.892\n",
       "fasttext web-crawl 0.760 0.760 0.761 0.764 0.837 0.902 0.907  0.914\n",
       "fasttext news/wiki 0.760 0.760 0.760 0.760 0.767 0.879 0.895  0.903\n",
       "en-twitter         0.760 0.765 0.770 0.797 0.810 0.892 0.902  0.907\n",
       "elmo               0.760 0.776 0.787 0.792 0.812 0.866 0.902  0.909\n",
       "Flair              0.760 0.760 0.760 0.761 0.776 0.787 0.812  0.829\n",
       "bert-base-cased    0.760 0.775 0.793 0.816 0.797 0.888 0.907  0.914\n",
       "BytePairEmbedding  0.760 0.765 0.766 0.777 0.806 0.867 0.885  0.899\n",
       "gpt-1              0.760 0.789 0.791 0.816 0.879 0.892 0.905  0.917"
      ]
     },
     "execution_count": 120,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "df.loc[0]"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "data = [0, 100,200,500,1000,3000,7000,18000]\n",
    "#data = [0, 100,200,500,1000, 3000,18000]\n",
    "\n",
    "plt.figure(figsize=(10,6))\n",
    "plt.plot(data, df.loc[0][1:], color=\"blue\")\n",
    "plt.plot(data, df.loc[1][1:], color=\"orange\")\n",
    "\n",
    "#plt.axis('equal')\n",
    "plt.title('F1 per training data')\n",
    "plt.xlabel('Training data')\n",
    "plt.ylabel('F1')\n",
    "\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f2dbfa42080>"
      ]
     },
     "execution_count": 121,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df.T.plot(figsize=(10,6))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f2dbf9d6c50>"
      ]
     },
     "execution_count": 122,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# colormap='jet'\n",
    "#df.T.plot(figsize=(10,6), title=\"F1 per train samples: Hatespeech data\", marker='.', markersize=7)\n",
    "df.T.plot(figsize=(10,6), marker='.', markersize=7)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Single, original"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "name = 'FINAL_flair_all_trainsz_100'\n",
    "all = loadResults_flair(name)\n",
    "\n",
    "df_new = unpack(all)\n",
    "#df_new"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "names"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "df = pd.DataFrame(list(zip(names, scores)), \n",
    "               columns =['Name', 'score']) \n",
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.7596666666666667\n"
     ]
    }
   ],
   "source": [
    "### count F1 for majority class.\n",
    "# majority, 0.76 is 1\n",
    "\n",
    "length = len(true_labels)\n",
    "#yhat = pd.Series(np.ones(length).astype(int))\n",
    "\n",
    "yhat = pd.Series( [1] * length) # value 1, times length, turned into Series\n",
    "f1 = f1_score(true_labels, yhat, average='micro')\n",
    "print(f1)\n",
    "\n",
    "names.append('Majority')\n",
    "scores.append(f1)\n",
    "traintimes.append(0)\n",
    "precisions.append(precision_score(true_labels, yhat, average='micro', labels=[0,1,2]))\n",
    "recalls.append(recall_score(true_labels, yhat, average='micro'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 133,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['glove',\n",
       " 'fasttext web-crawl',\n",
       " 'fasttext news/wiki',\n",
       " 'en-twitter',\n",
       " 'elmo',\n",
       " 'Flair',\n",
       " 'bert-base-uncased',\n",
       " 'BytePairEmbedding',\n",
       " 'gpt-1',\n",
       " 'Majority',\n",
       " 'Majority']"
      ]
     },
     "execution_count": 133,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "names"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0.892,\n",
       " 0.9136666666666666,\n",
       " 0.9026666666666666,\n",
       " 0.9066666666666666,\n",
       " 0.9086666666666666,\n",
       " 0.8293333333333334,\n",
       " 0.9136666666666666,\n",
       " 0.8986666666666666,\n",
       " 0.9173333333333333]"
      ]
     },
     "execution_count": 79,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scores"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Traintimes, prediction times"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['glove',\n",
       " 'fasttext web-crawl',\n",
       " 'fasttext news/wiki',\n",
       " 'en-twitter',\n",
       " 'elmo',\n",
       " 'Flair',\n",
       " 'bert-base-cased',\n",
       " 'BytePairEmbedding',\n",
       " 'gpt-1']"
      ]
     },
     "execution_count": 82,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "names"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[4.842005729675293,\n",
       " 5.090792655944824,\n",
       " 4.955539703369141,\n",
       " 5.2229108810424805,\n",
       " 130.29307055473328,\n",
       " 93.0849277973175,\n",
       " 68.86565017700195,\n",
       " 8.170584917068481,\n",
       " 65.25658059120178]"
      ]
     },
     "execution_count": 83,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predtimes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "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>Model</th>\n",
       "      <th>Traintime</th>\n",
       "      <th>Prediction time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>glove</td>\n",
       "      <td>71.011</td>\n",
       "      <td>1.614</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>fasttext web-crawl</td>\n",
       "      <td>208.219</td>\n",
       "      <td>1.697</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>fasttext news/wiki</td>\n",
       "      <td>209.539</td>\n",
       "      <td>1.652</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>en-twitter</td>\n",
       "      <td>129.847</td>\n",
       "      <td>1.741</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>elmo</td>\n",
       "      <td>77.256</td>\n",
       "      <td>43.431</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Flair</td>\n",
       "      <td>56.918</td>\n",
       "      <td>31.028</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>bert-base-cased</td>\n",
       "      <td>259.950</td>\n",
       "      <td>22.955</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>BytePairEmbedding</td>\n",
       "      <td>61.296</td>\n",
       "      <td>2.724</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>gpt-1</td>\n",
       "      <td>198.082</td>\n",
       "      <td>21.752</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                Model  Traintime  Prediction time\n",
       "0               glove     71.011            1.614\n",
       "1  fasttext web-crawl    208.219            1.697\n",
       "2  fasttext news/wiki    209.539            1.652\n",
       "3          en-twitter    129.847            1.741\n",
       "4                elmo     77.256           43.431\n",
       "5               Flair     56.918           31.028\n",
       "6     bert-base-cased    259.950           22.955\n",
       "7   BytePairEmbedding     61.296            2.724\n",
       "8               gpt-1    198.082           21.752"
      ]
     },
     "execution_count": 86,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# dictionary of lists  \n",
    "dict = {'Model': names, 'Traintime': traintimes, 'Prediction time': predtimes}  \n",
    "    \n",
    "df_times = pd.DataFrame(dict) \n",
    "df_times.set_index('Model') \n",
    "\n",
    "\n",
    "#df_times['Prediction time'] = df_times['Prediction time'] / 3000\n",
    "df_times['Prediction time'] = df_times['Prediction time'] / 3 # per 1000 sample\n",
    "df_times\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### FLAIR"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "9"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# in folder dataset_hatespeech/output/    # can see name of resultfile there\n",
    "#name = 'all_flair_512LSTM_2ep'\n",
    "name = 'all_flair_512LSTM_15ep_8model'\n",
    "name = 'FINAL_flair_all_trainsz_100'\n",
    "\n",
    "#name = 'FINAL_flair_all_trainsz_200'\n",
    "#name = 'FINAL_flair_all_trainsz_500'\n",
    "\n",
    "name = 'FINAL_flair_all_trainsz_1000'\n",
    "\n",
    "name = 'FINAL_flair_all_trainsz_18783'\n",
    "\n",
    "name = 'FINAL_flair_all_trainsz_100'\n",
    "\n",
    "all_flair = loadResults_flair(name)\n",
    "all = all_flair\n",
    "len(all_flair)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import precision_score, recall_score\n",
    "# FLAIR\n",
    "# unpack values from dict / list structure: names and f1-scores of models.\n",
    "names = []\n",
    "scores= []\n",
    "traintimes=[]\n",
    "precisions = []\n",
    "recalls = []\n",
    "\n",
    "for i in range(len(all)):\n",
    "    yhat = all[i]['labels']\n",
    "    all[i]['f1'] = f1_score(true_labels, yhat, average='micro')\n",
    "    #all[i]['precision'] = precision_score(true_labels, yhat, average='micro', labels=[0,1,2])\n",
    "    #all[i]['recall'] = recall_score(true_labels, yhat, average='micro')\n",
    "    \n",
    "    names.append(all[i]['model'])\n",
    "    scores.append(f1_score(true_labels, yhat, average='micro'))\n",
    "    traintimes.append(all[i]['traintime'])\n",
    "    precisions.append(precision_score(true_labels, yhat, average='micro', labels=[0,1,2]))\n",
    "    recalls.append(recall_score(true_labels, yhat, average='micro'))\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### TF-hlub"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [],
   "source": [
    "all_tfhub = loadResults_tfhub()\n",
    "len(all_tfhub)\n",
    "all = all_tfhub\n",
    "\n",
    "# TF-HUB\n",
    "# unpack values from dict / list structure: names and f1-scores of models.\n",
    "names = []\n",
    "scores= []\n",
    "#traintimes=[]\n",
    "\n",
    "for i in range(len(all)):\n",
    "    yhat = all[i]['labels']\n",
    "    all[i]['f1'] = f1_score(true_labels, yhat, average='micro')\n",
    "    \n",
    "    names.append(all[i]['model'])\n",
    "    scores.append(f1_score(true_labels, yhat, average='micro'))\n",
    "    #traintimes.append(all[i]['traintime'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Common"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['glove',\n",
       " 'fasttext web-crawl',\n",
       " 'fasttext news/wiki',\n",
       " 'en-twitter',\n",
       " 'elmo',\n",
       " 'Flair',\n",
       " 'bert-base-uncased',\n",
       " 'BytePairEmbedding',\n",
       " 'gpt-1']"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "names"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['glove',\n",
       " 'fasttext web-crawl',\n",
       " 'fasttext news/wiki',\n",
       " 'en-twitter',\n",
       " 'elmo',\n",
       " 'Flair',\n",
       " 'bert-base-uncased',\n",
       " 'BytePairEmbedding',\n",
       " 'gpt-1']"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "names"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0.7596666666666667,\n",
       " 0.7596666666666667,\n",
       " 0.7596666666666667,\n",
       " 0.7646666666666667,\n",
       " 0.7763333333333333,\n",
       " 0.7596666666666667,\n",
       " 0.7746666666666666,\n",
       " 0.765,\n",
       " 0.7886666666666666]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scores"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[0.7976666666666666,\n",
       " 0.8366666666666667,\n",
       " 0.7673333333333334,\n",
       " 0.8096666666666666,\n",
       " 0.8116666666666666,\n",
       " 0.7756666666666665,\n",
       " 0.797,\n",
       " 0.8063333333333333,\n",
       " 0.8786666666666667]"
      ]
     },
     "execution_count": 75,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scores"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[420.35936999320984,\n",
       " 512.5102701187134,\n",
       " 490.92159509658813,\n",
       " 505.1277804374695,\n",
       " 557.0145864486694,\n",
       " 521.807178735733,\n",
       " 1232.1364481449127,\n",
       " 638.7153763771057,\n",
       " 1123.3866958618164]"
      ]
     },
     "execution_count": 120,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "traintimes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Bar\n",
    "index = np.arange(len(scores))\n",
    "values = scores\n",
    "labels = names\n",
    "\n",
    "plt.figure(figsize=(12,5))\n",
    "plt.ylim(0.6, 0.90) # is set, but not effecting bar\n",
    "plt.bar(index, values)\n",
    "plt.xlabel('Model', fontsize=10)\n",
    "plt.ylabel('F1', fontsize=10)\n",
    "plt.xticks(index, labels, fontsize=10, rotation=10)\n",
    "plt.title('Test F1 with 18.7K training data')\n",
    "\n",
    "plt.show()\n",
    "\n",
    "#from importlib import reload  \n",
    "#reload(plt)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 119,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Bar\n",
    "index = np.arange(len(scores))\n",
    "values = scores\n",
    "labels = names\n",
    "\n",
    "plt.figure(figsize=(12,5))\n",
    "plt.ylim(0.80, 0.96) # is set, but not effecting bar\n",
    "plt.bar(index, values)\n",
    "plt.xlabel('Model', fontsize=10)\n",
    "plt.ylabel('F1', fontsize=10)\n",
    "plt.xticks(index, labels, fontsize=10, rotation=10)\n",
    "plt.title('F1 with 18.7K training data')\n",
    "\n",
    "plt.show()\n",
    "\n",
    "#from importlib import reload  \n",
    "#reload(plt)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "del labels[3:6]\n",
    "del values[3:6]\n",
    "del labels[3]\n",
    "del values[3]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Bar\n",
    "index = np.arange(len(scores))\n",
    "values = scores\n",
    "labels = names\n",
    "\n",
    "plt.figure(figsize=(12,5))\n",
    "plt.ylim(0.82, 0.96) # is set, but not effecting bar\n",
    "plt.bar(index, values)\n",
    "plt.xlabel('Model', fontsize=10)\n",
    "plt.ylabel('Test-F1', fontsize=10)\n",
    "plt.xticks(index, labels, fontsize=10, rotation=10)\n",
    "plt.title('Test-F1')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [],
   "source": [
    "#del labels[3:5]\n",
    "#del values[3:5]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Turn into dataframe"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "# input list containing dictionaries: all[0]['labels'] where labels are on 2nd level\n",
    "# output: raise the labels-lists into 1st level and turn into DataFrame\n",
    "\n",
    "labels = []\n",
    "for i in range(len(names)): #(len(all))\n",
    "    labels.append(all[i]['labels'])\n",
    "\n",
    "labels = pd.DataFrame(labels).T\n",
    "labels.columns= names\n",
    "#lab\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Correlation between predicted labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "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>glove</th>\n",
       "      <th>fasttext web-crawl</th>\n",
       "      <th>fasttext news/wiki</th>\n",
       "      <th>en-twitter</th>\n",
       "      <th>elmo</th>\n",
       "      <th>Flair</th>\n",
       "      <th>bert-base-uncased</th>\n",
       "      <th>BytePairEmbedding</th>\n",
       "      <th>gpt-1</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>glove</th>\n",
       "      <td>1.000</td>\n",
       "      <td>0.831</td>\n",
       "      <td>0.814</td>\n",
       "      <td>0.817</td>\n",
       "      <td>0.786</td>\n",
       "      <td>0.554</td>\n",
       "      <td>0.786</td>\n",
       "      <td>0.782</td>\n",
       "      <td>0.775</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>fasttext web-crawl</th>\n",
       "      <td>0.831</td>\n",
       "      <td>1.000</td>\n",
       "      <td>0.886</td>\n",
       "      <td>0.897</td>\n",
       "      <td>0.844</td>\n",
       "      <td>0.558</td>\n",
       "      <td>0.856</td>\n",
       "      <td>0.810</td>\n",
       "      <td>0.845</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>fasttext news/wiki</th>\n",
       "      <td>0.814</td>\n",
       "      <td>0.886</td>\n",
       "      <td>1.000</td>\n",
       "      <td>0.867</td>\n",
       "      <td>0.798</td>\n",
       "      <td>0.578</td>\n",
       "      <td>0.838</td>\n",
       "      <td>0.776</td>\n",
       "      <td>0.805</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>en-twitter</th>\n",
       "      <td>0.817</td>\n",
       "      <td>0.897</td>\n",
       "      <td>0.867</td>\n",
       "      <td>1.000</td>\n",
       "      <td>0.835</td>\n",
       "      <td>0.585</td>\n",
       "      <td>0.849</td>\n",
       "      <td>0.793</td>\n",
       "      <td>0.819</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>elmo</th>\n",
       "      <td>0.786</td>\n",
       "      <td>0.844</td>\n",
       "      <td>0.798</td>\n",
       "      <td>0.835</td>\n",
       "      <td>1.000</td>\n",
       "      <td>0.547</td>\n",
       "      <td>0.844</td>\n",
       "      <td>0.808</td>\n",
       "      <td>0.825</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Flair</th>\n",
       "      <td>0.554</td>\n",
       "      <td>0.558</td>\n",
       "      <td>0.578</td>\n",
       "      <td>0.585</td>\n",
       "      <td>0.547</td>\n",
       "      <td>1.000</td>\n",
       "      <td>0.578</td>\n",
       "      <td>0.525</td>\n",
       "      <td>0.516</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>bert-base-uncased</th>\n",
       "      <td>0.786</td>\n",
       "      <td>0.856</td>\n",
       "      <td>0.838</td>\n",
       "      <td>0.849</td>\n",
       "      <td>0.844</td>\n",
       "      <td>0.578</td>\n",
       "      <td>1.000</td>\n",
       "      <td>0.813</td>\n",
       "      <td>0.810</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>BytePairEmbedding</th>\n",
       "      <td>0.782</td>\n",
       "      <td>0.810</td>\n",
       "      <td>0.776</td>\n",
       "      <td>0.793</td>\n",
       "      <td>0.808</td>\n",
       "      <td>0.525</td>\n",
       "      <td>0.813</td>\n",
       "      <td>1.000</td>\n",
       "      <td>0.780</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>gpt-1</th>\n",
       "      <td>0.775</td>\n",
       "      <td>0.845</td>\n",
       "      <td>0.805</td>\n",
       "      <td>0.819</td>\n",
       "      <td>0.825</td>\n",
       "      <td>0.516</td>\n",
       "      <td>0.810</td>\n",
       "      <td>0.780</td>\n",
       "      <td>1.000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                    glove  fasttext web-crawl  fasttext news/wiki  en-twitter  \\\n",
       "glove               1.000               0.831               0.814       0.817   \n",
       "fasttext web-crawl  0.831               1.000               0.886       0.897   \n",
       "fasttext news/wiki  0.814               0.886               1.000       0.867   \n",
       "en-twitter          0.817               0.897               0.867       1.000   \n",
       "elmo                0.786               0.844               0.798       0.835   \n",
       "Flair               0.554               0.558               0.578       0.585   \n",
       "bert-base-uncased   0.786               0.856               0.838       0.849   \n",
       "BytePairEmbedding   0.782               0.810               0.776       0.793   \n",
       "gpt-1               0.775               0.845               0.805       0.819   \n",
       "\n",
       "                    elmo  Flair  bert-base-uncased  BytePairEmbedding  gpt-1  \n",
       "glove              0.786  0.554              0.786              0.782  0.775  \n",
       "fasttext web-crawl 0.844  0.558              0.856              0.810  0.845  \n",
       "fasttext news/wiki 0.798  0.578              0.838              0.776  0.805  \n",
       "en-twitter         0.835  0.585              0.849              0.793  0.819  \n",
       "elmo               1.000  0.547              0.844              0.808  0.825  \n",
       "Flair              0.547  1.000              0.578              0.525  0.516  \n",
       "bert-base-uncased  0.844  0.578              1.000              0.813  0.810  \n",
       "BytePairEmbedding  0.808  0.525              0.813              1.000  0.780  \n",
       "gpt-1              0.825  0.516              0.810              0.780  1.000  "
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "labels.corr()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7fa50513f588>"
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x720 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "pick = labels\n",
    "import seaborn as sns\n",
    "\n",
    "#f, ax = plt.subplots(figsize=(10, 8))\n",
    "f, ax = plt.subplots(figsize=(12, 10))\n",
    "corr = pick.corr()\n",
    "sns.heatmap(corr, mask=np.zeros_like(corr, dtype=np.bool), cmap=sns.diverging_palette(220, 10, as_cmap=True),\n",
    "            square=True, ax=ax)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Flair is very low cor with others - remove it to use the color shades better"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "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>glove</th>\n",
       "      <th>fasttext web-crawl</th>\n",
       "      <th>fasttext news/wiki</th>\n",
       "      <th>en-twitter</th>\n",
       "      <th>elmo</th>\n",
       "      <th>bert-base-uncased</th>\n",
       "      <th>BytePairEmbedding</th>\n",
       "      <th>gpt-1</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</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",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2995</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2996</th>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2997</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2998</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2999</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3000 rows × 8 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      glove  fasttext web-crawl  fasttext news/wiki  en-twitter  elmo  \\\n",
       "0         1                   1                   1           1     1   \n",
       "1         1                   1                   1           1     1   \n",
       "2         1                   1                   1           1     1   \n",
       "3         1                   1                   1           1     1   \n",
       "4         1                   1                   1           1     1   \n",
       "...     ...                 ...                 ...         ...   ...   \n",
       "2995      1                   1                   1           1     1   \n",
       "2996      2                   2                   2           2     2   \n",
       "2997      1                   1                   1           1     1   \n",
       "2998      1                   1                   2           1     1   \n",
       "2999      1                   1                   1           1     1   \n",
       "\n",
       "      bert-base-uncased  BytePairEmbedding  gpt-1  \n",
       "0                     1                  1      1  \n",
       "1                     1                  1      1  \n",
       "2                     1                  1      1  \n",
       "3                     1                  1      1  \n",
       "4                     1                  1      1  \n",
       "...                 ...                ...    ...  \n",
       "2995                  1                  1      1  \n",
       "2996                  2                  2      2  \n",
       "2997                  1                  1      1  \n",
       "2998                  1                  1      1  \n",
       "2999                  1                  1      1  \n",
       "\n",
       "[3000 rows x 8 columns]"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "del labels['Flair']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "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>glove</th>\n",
       "      <th>fasttext web-crawl</th>\n",
       "      <th>fasttext news/wiki</th>\n",
       "      <th>en-twitter</th>\n",
       "      <th>elmo</th>\n",
       "      <th>bert-base-uncased</th>\n",
       "      <th>BytePairEmbedding</th>\n",
       "      <th>gpt-1</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>glove</th>\n",
       "      <td>1.000</td>\n",
       "      <td>0.831</td>\n",
       "      <td>0.814</td>\n",
       "      <td>0.817</td>\n",
       "      <td>0.786</td>\n",
       "      <td>0.786</td>\n",
       "      <td>0.782</td>\n",
       "      <td>0.775</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>fasttext web-crawl</th>\n",
       "      <td>0.831</td>\n",
       "      <td>1.000</td>\n",
       "      <td>0.886</td>\n",
       "      <td>0.897</td>\n",
       "      <td>0.844</td>\n",
       "      <td>0.856</td>\n",
       "      <td>0.810</td>\n",
       "      <td>0.845</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>fasttext news/wiki</th>\n",
       "      <td>0.814</td>\n",
       "      <td>0.886</td>\n",
       "      <td>1.000</td>\n",
       "      <td>0.867</td>\n",
       "      <td>0.798</td>\n",
       "      <td>0.838</td>\n",
       "      <td>0.776</td>\n",
       "      <td>0.805</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>en-twitter</th>\n",
       "      <td>0.817</td>\n",
       "      <td>0.897</td>\n",
       "      <td>0.867</td>\n",
       "      <td>1.000</td>\n",
       "      <td>0.835</td>\n",
       "      <td>0.849</td>\n",
       "      <td>0.793</td>\n",
       "      <td>0.819</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>elmo</th>\n",
       "      <td>0.786</td>\n",
       "      <td>0.844</td>\n",
       "      <td>0.798</td>\n",
       "      <td>0.835</td>\n",
       "      <td>1.000</td>\n",
       "      <td>0.844</td>\n",
       "      <td>0.808</td>\n",
       "      <td>0.825</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>bert-base-uncased</th>\n",
       "      <td>0.786</td>\n",
       "      <td>0.856</td>\n",
       "      <td>0.838</td>\n",
       "      <td>0.849</td>\n",
       "      <td>0.844</td>\n",
       "      <td>1.000</td>\n",
       "      <td>0.813</td>\n",
       "      <td>0.810</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>BytePairEmbedding</th>\n",
       "      <td>0.782</td>\n",
       "      <td>0.810</td>\n",
       "      <td>0.776</td>\n",
       "      <td>0.793</td>\n",
       "      <td>0.808</td>\n",
       "      <td>0.813</td>\n",
       "      <td>1.000</td>\n",
       "      <td>0.780</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>gpt-1</th>\n",
       "      <td>0.775</td>\n",
       "      <td>0.845</td>\n",
       "      <td>0.805</td>\n",
       "      <td>0.819</td>\n",
       "      <td>0.825</td>\n",
       "      <td>0.810</td>\n",
       "      <td>0.780</td>\n",
       "      <td>1.000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                    glove  fasttext web-crawl  fasttext news/wiki  en-twitter  \\\n",
       "glove               1.000               0.831               0.814       0.817   \n",
       "fasttext web-crawl  0.831               1.000               0.886       0.897   \n",
       "fasttext news/wiki  0.814               0.886               1.000       0.867   \n",
       "en-twitter          0.817               0.897               0.867       1.000   \n",
       "elmo                0.786               0.844               0.798       0.835   \n",
       "bert-base-uncased   0.786               0.856               0.838       0.849   \n",
       "BytePairEmbedding   0.782               0.810               0.776       0.793   \n",
       "gpt-1               0.775               0.845               0.805       0.819   \n",
       "\n",
       "                    elmo  bert-base-uncased  BytePairEmbedding  gpt-1  \n",
       "glove              0.786              0.786              0.782  0.775  \n",
       "fasttext web-crawl 0.844              0.856              0.810  0.845  \n",
       "fasttext news/wiki 0.798              0.838              0.776  0.805  \n",
       "en-twitter         0.835              0.849              0.793  0.819  \n",
       "elmo               1.000              0.844              0.808  0.825  \n",
       "bert-base-uncased  0.844              1.000              0.813  0.810  \n",
       "BytePairEmbedding  0.808              0.813              1.000  0.780  \n",
       "gpt-1              0.825              0.810              0.780  1.000  "
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "labels.corr()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7fa504ee0e48>"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "pick = labels\n",
    "import seaborn as sns\n",
    "\n",
    "f, ax = plt.subplots(figsize=(10, 8))\n",
    "#f, ax = plt.subplots(figsize=(12, 10))\n",
    "corr = pick.corr()\n",
    "sns.heatmap(corr, mask=np.zeros_like(corr, dtype=np.bool), cmap=sns.diverging_palette(220, 10, as_cmap=True),\n",
    "            square=True, ax=ax)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "# last 3 columns\n",
    "pick = labels.iloc[:, 3:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "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>nnlm-128_tsize_3000</th>\n",
       "      <th>DAN_tsize_3000</th>\n",
       "      <th>univ3_tsize_3000</th>\n",
       "      <th>nnlm-128_tsize_18783_epoch30</th>\n",
       "      <th>DAN_tsize_18783_epoch30</th>\n",
       "      <th>univ3_tsize_18783_epoch30</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>[4.233589e-05, 0.9999193, 3.840332e-05]</td>\n",
       "      <td>[1.6480626e-05, 0.9998909, 9.264603e-05]</td>\n",
       "      <td>[6.398087e-05, 0.9998853, 5.0707247e-05]</td>\n",
       "      <td>[0.00013908364, 0.9998254, 3.547703e-05]</td>\n",
       "      <td>[2.5715732e-05, 0.9998716, 0.000102616505]</td>\n",
       "      <td>[7.91436e-05, 0.9998338, 8.701929e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>[0.0013433183, 0.99797875, 0.0006780409]</td>\n",
       "      <td>[7.580178e-05, 0.96989506, 0.030029025]</td>\n",
       "      <td>[0.0001076711, 0.999806, 8.637823e-05]</td>\n",
       "      <td>[0.002030536, 0.99750996, 0.00045953243]</td>\n",
       "      <td>[2.547616e-05, 0.9989743, 0.0010002507]</td>\n",
       "      <td>[7.870038e-05, 0.9997701, 0.000151296]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>[0.0011234806, 0.99871135, 0.00016510795]</td>\n",
       "      <td>[2.4509067e-05, 0.99997044, 5.0195786e-06]</td>\n",
       "      <td>[4.7642898e-05, 0.9999232, 2.913791e-05]</td>\n",
       "      <td>[0.0006725391, 0.99926144, 6.598066e-05]</td>\n",
       "      <td>[1.7972814e-05, 0.999974, 8.038951e-06]</td>\n",
       "      <td>[6.0237904e-05, 0.9999143, 2.548163e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>[0.0004791329, 0.9995173, 3.5879834e-06]</td>\n",
       "      <td>[1.3438928e-05, 0.99997985, 6.728323e-06]</td>\n",
       "      <td>[3.8377642e-05, 0.99993515, 2.6411251e-05]</td>\n",
       "      <td>[0.0003392862, 0.9996592, 1.5918495e-06]</td>\n",
       "      <td>[1.8328115e-05, 0.9999672, 1.4433712e-05]</td>\n",
       "      <td>[4.4898494e-05, 0.9999287, 2.633073e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>[3.1970467e-05, 0.99996805, 7.1044792e-09]</td>\n",
       "      <td>[1.818452e-05, 0.999979, 2.731076e-06]</td>\n",
       "      <td>[6.852479e-05, 0.99991035, 2.1157512e-05]</td>\n",
       "      <td>[4.3006254e-05, 0.99995697, 8.022627e-09]</td>\n",
       "      <td>[1.7900435e-05, 0.9999765, 5.5927703e-06]</td>\n",
       "      <td>[7.507247e-05, 0.9999107, 1.421926e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>[0.081907175, 0.7386889, 0.17940396]</td>\n",
       "      <td>[0.0016064331, 0.005709461, 0.9926842]</td>\n",
       "      <td>[0.0004184056, 0.9995516, 3.00833e-05]</td>\n",
       "      <td>[0.08427183, 0.8487675, 0.06696065]</td>\n",
       "      <td>[0.0026555285, 0.06760905, 0.9297355]</td>\n",
       "      <td>[0.00016169962, 0.99982136, 1.6898892e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>[0.0006111432, 0.9598467, 0.03954213]</td>\n",
       "      <td>[1.0904048e-05, 0.999979, 1.0073294e-05]</td>\n",
       "      <td>[4.641684e-05, 0.99991405, 3.959844e-05]</td>\n",
       "      <td>[0.001172376, 0.9711367, 0.027690923]</td>\n",
       "      <td>[2.252558e-05, 0.99995816, 1.9321673e-05]</td>\n",
       "      <td>[5.2135645e-05, 0.99991536, 3.253769e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>[8.144646e-07, 0.9998646, 0.00013461642]</td>\n",
       "      <td>[0.00010322058, 0.98762506, 0.012271626]</td>\n",
       "      <td>[4.8737867e-05, 0.9998951, 5.6094e-05]</td>\n",
       "      <td>[2.1608248e-06, 0.99978226, 0.00021563252]</td>\n",
       "      <td>[0.00030092022, 0.9585573, 0.04114188]</td>\n",
       "      <td>[6.164825e-05, 0.9998791, 5.9188635e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>[0.00034172254, 0.9969459, 0.0027123827]</td>\n",
       "      <td>[1.3682003e-05, 0.99988484, 0.0001014915]</td>\n",
       "      <td>[0.0001048133, 0.9995121, 0.00038309305]</td>\n",
       "      <td>[0.00057175703, 0.9981173, 0.0013109304]</td>\n",
       "      <td>[1.6481239e-05, 0.99988806, 9.542799e-05]</td>\n",
       "      <td>[6.790512e-05, 0.99990237, 2.9676981e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>[4.506633e-05, 0.9999548, 1.6939906e-07]</td>\n",
       "      <td>[6.86458e-06, 0.9999902, 2.8723064e-06]</td>\n",
       "      <td>[4.578529e-05, 0.99992275, 3.144341e-05]</td>\n",
       "      <td>[7.477624e-05, 0.99992514, 1.1235904e-07]</td>\n",
       "      <td>[5.6351914e-06, 0.9999902, 4.1621734e-06]</td>\n",
       "      <td>[6.1978855e-05, 0.9999176, 2.0371614e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>[0.028411375, 6.653153e-05, 0.97152215]</td>\n",
       "      <td>[0.0018834813, 0.0010752678, 0.9970413]</td>\n",
       "      <td>[0.00011976459, 0.00021957459, 0.9996606]</td>\n",
       "      <td>[0.021710116, 2.5961535e-05, 0.9782639]</td>\n",
       "      <td>[0.00081079284, 0.0007934927, 0.99839574]</td>\n",
       "      <td>[7.8857935e-05, 0.0003978529, 0.9995233]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>[0.00021463603, 0.99944204, 0.00034326394]</td>\n",
       "      <td>[6.1479936e-05, 0.9999318, 6.7185747e-06]</td>\n",
       "      <td>[5.3447086e-05, 0.9999268, 1.981259e-05]</td>\n",
       "      <td>[0.00060418097, 0.99912363, 0.0002722138]</td>\n",
       "      <td>[6.452492e-05, 0.99992394, 1.1570731e-05]</td>\n",
       "      <td>[7.210481e-05, 0.9999161, 1.1803859e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>[0.9867077, 0.009154943, 0.004137294]</td>\n",
       "      <td>[0.17751005, 0.43451717, 0.3879728]</td>\n",
       "      <td>[0.00017330452, 0.00029272275, 0.999534]</td>\n",
       "      <td>[0.9535876, 0.0369543, 0.009458121]</td>\n",
       "      <td>[0.06544465, 0.4085732, 0.52598214]</td>\n",
       "      <td>[0.00020762562, 0.00060981687, 0.9991825]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>[0.005461895, 0.99341416, 0.0011239109]</td>\n",
       "      <td>[1.4571097e-05, 0.9999753, 1.0091025e-05]</td>\n",
       "      <td>[4.281666e-05, 0.9999161, 4.1079733e-05]</td>\n",
       "      <td>[0.010873717, 0.98802453, 0.0011016902]</td>\n",
       "      <td>[1.33834255e-05, 0.99996865, 1.796761e-05]</td>\n",
       "      <td>[4.5374585e-05, 0.9999213, 3.3305845e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>[7.315173e-06, 0.9999902, 2.536194e-06]</td>\n",
       "      <td>[5.9142135e-06, 0.9999912, 2.8827149e-06]</td>\n",
       "      <td>[4.5550813e-05, 0.9999248, 2.9640507e-05]</td>\n",
       "      <td>[1.2437995e-05, 0.99998605, 1.4981332e-06]</td>\n",
       "      <td>[6.814804e-06, 0.99998903, 4.2227966e-06]</td>\n",
       "      <td>[5.612093e-05, 0.99992216, 2.1637394e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>[4.2200754e-05, 0.14187208, 0.8580857]</td>\n",
       "      <td>[0.0013588266, 0.31469786, 0.6839433]</td>\n",
       "      <td>[0.00026939638, 0.0002832935, 0.9994473]</td>\n",
       "      <td>[0.00022437192, 0.40065616, 0.5991194]</td>\n",
       "      <td>[0.0018520205, 0.23281698, 0.76533103]</td>\n",
       "      <td>[0.00013584517, 0.00024865704, 0.99961543]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>[0.19498013, 0.28203487, 0.522985]</td>\n",
       "      <td>[6.760271e-05, 0.0001779354, 0.9997545]</td>\n",
       "      <td>[0.00012323644, 0.00015144021, 0.9997253]</td>\n",
       "      <td>[0.42409116, 0.3661773, 0.2097315]</td>\n",
       "      <td>[0.00018075421, 0.00020491573, 0.99961436]</td>\n",
       "      <td>[8.948143e-05, 0.00019811906, 0.9997124]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>[0.000438633, 0.082147196, 0.9174142]</td>\n",
       "      <td>[0.0026416993, 0.7575073, 0.239851]</td>\n",
       "      <td>[8.45979e-05, 0.00014179268, 0.99977356]</td>\n",
       "      <td>[0.00028815673, 0.02189521, 0.97781664]</td>\n",
       "      <td>[0.0020194512, 0.35163933, 0.6463412]</td>\n",
       "      <td>[0.00022410582, 0.013034229, 0.9867417]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>[0.0076374356, 0.99044526, 0.0019172331]</td>\n",
       "      <td>[5.1096198e-05, 0.9999249, 2.3965078e-05]</td>\n",
       "      <td>[3.943014e-05, 0.9999386, 2.1962676e-05]</td>\n",
       "      <td>[0.0049450295, 0.99437827, 0.0006767206]</td>\n",
       "      <td>[0.0011020844, 0.9986578, 0.00024008311]</td>\n",
       "      <td>[5.753277e-05, 0.99992466, 1.777614e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>[0.0009561412, 0.99867696, 0.00036692788]</td>\n",
       "      <td>[0.3191397, 0.12964316, 0.55121714]</td>\n",
       "      <td>[9.3881274e-05, 0.000116055446, 0.99979013]</td>\n",
       "      <td>[0.0021360728, 0.9973984, 0.00046561955]</td>\n",
       "      <td>[0.06276215, 0.010787978, 0.9264499]</td>\n",
       "      <td>[5.8154856e-05, 0.00014961355, 0.9997923]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>[0.018522294, 0.98142964, 4.812113e-05]</td>\n",
       "      <td>[1.42467015e-05, 0.99998283, 2.856866e-06]</td>\n",
       "      <td>[5.2919135e-05, 0.999915, 3.2032738e-05]</td>\n",
       "      <td>[0.012432584, 0.9875252, 4.218152e-05]</td>\n",
       "      <td>[6.9983995e-05, 0.9999213, 8.655985e-06]</td>\n",
       "      <td>[7.0063914e-05, 0.9999087, 2.1275844e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>[3.3813408e-06, 0.99999654, 1.1062413e-07]</td>\n",
       "      <td>[4.903143e-05, 0.99991894, 3.2119373e-05]</td>\n",
       "      <td>[3.9518014e-05, 0.99992836, 3.202855e-05]</td>\n",
       "      <td>[3.4952702e-06, 0.99999654, 5.4307318e-08]</td>\n",
       "      <td>[3.8394963e-05, 0.9999201, 4.1535608e-05]</td>\n",
       "      <td>[4.8331603e-05, 0.99993205, 1.9714593e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>[0.022528995, 0.07266271, 0.9048083]</td>\n",
       "      <td>[0.000281977, 0.63859683, 0.36112115]</td>\n",
       "      <td>[7.190047e-05, 0.00026368184, 0.9996644]</td>\n",
       "      <td>[0.009834621, 0.031162515, 0.95900285]</td>\n",
       "      <td>[3.8877777e-05, 0.99868363, 0.0012774161]</td>\n",
       "      <td>[5.3785814e-05, 0.00030585722, 0.99964035]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>[3.0262268e-05, 0.9998877, 8.194801e-05]</td>\n",
       "      <td>[1.4272013e-05, 0.9999516, 3.404179e-05]</td>\n",
       "      <td>[3.8888775e-05, 0.9999335, 2.7662583e-05]</td>\n",
       "      <td>[2.863268e-05, 0.9999547, 1.6737697e-05]</td>\n",
       "      <td>[9.463277e-06, 0.9999757, 1.4949457e-05]</td>\n",
       "      <td>[5.3389886e-05, 0.99991834, 2.8288976e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>[0.00017858457, 0.99981993, 1.5306515e-06]</td>\n",
       "      <td>[8.602782e-06, 0.9999882, 3.274369e-06]</td>\n",
       "      <td>[3.7480313e-05, 0.9999404, 2.2209082e-05]</td>\n",
       "      <td>[0.00046800202, 0.99953043, 1.5787145e-06]</td>\n",
       "      <td>[7.325024e-06, 0.9999875, 5.193338e-06]</td>\n",
       "      <td>[4.9465325e-05, 0.9999325, 1.801631e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>[0.00028489454, 0.99971455, 4.936044e-07]</td>\n",
       "      <td>[0.001231889, 0.9987533, 1.4753598e-05]</td>\n",
       "      <td>[5.077563e-05, 0.99992466, 2.4521478e-05]</td>\n",
       "      <td>[0.0009937334, 0.9990043, 2.0188118e-06]</td>\n",
       "      <td>[0.0066787857, 0.99322104, 0.00010020443]</td>\n",
       "      <td>[5.763131e-05, 0.999928, 1.4367954e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>[0.0001413786, 0.9998586, 4.243305e-08]</td>\n",
       "      <td>[1.1435212e-05, 0.9999863, 2.2247548e-06]</td>\n",
       "      <td>[4.3633758e-05, 0.9999362, 2.0162879e-05]</td>\n",
       "      <td>[0.00067893823, 0.999321, 1.2991575e-07]</td>\n",
       "      <td>[1.5814856e-05, 0.99997985, 4.2840243e-06]</td>\n",
       "      <td>[7.19162e-05, 0.9999155, 1.2658613e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>[0.0021451085, 0.99774146, 0.00011343301]</td>\n",
       "      <td>[9.384714e-05, 0.99988794, 1.8204153e-05]</td>\n",
       "      <td>[3.7041853e-05, 0.99993753, 2.5337724e-05]</td>\n",
       "      <td>[0.002937752, 0.9970304, 3.1891446e-05]</td>\n",
       "      <td>[0.000570712, 0.9993119, 0.000117387346]</td>\n",
       "      <td>[5.7950838e-05, 0.9999305, 1.1548351e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>[0.00015609733, 0.086633295, 0.9132106]</td>\n",
       "      <td>[6.858496e-05, 0.0013311019, 0.9986003]</td>\n",
       "      <td>[5.630375e-05, 0.00017668531, 0.999767]</td>\n",
       "      <td>[0.00021700047, 0.12237597, 0.8774071]</td>\n",
       "      <td>[0.00014342695, 0.0009917042, 0.9988649]</td>\n",
       "      <td>[7.705059e-05, 0.0055413716, 0.9943816]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>[6.387743e-05, 0.99993515, 1.0078197e-06]</td>\n",
       "      <td>[8.210377e-06, 0.99998844, 3.3694355e-06]</td>\n",
       "      <td>[4.0054543e-05, 0.9999366, 2.3396988e-05]</td>\n",
       "      <td>[0.00041335513, 0.99958247, 4.15565e-06]</td>\n",
       "      <td>[1.3541771e-05, 0.9999796, 6.7871615e-06]</td>\n",
       "      <td>[5.4944056e-05, 0.99992704, 1.7942342e-05]</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",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2970</th>\n",
       "      <td>[0.00017322828, 0.9997682, 5.8635833e-05]</td>\n",
       "      <td>[7.1989675e-06, 0.9999864, 6.41879e-06]</td>\n",
       "      <td>[4.035385e-05, 0.9999207, 3.8841205e-05]</td>\n",
       "      <td>[0.00013102862, 0.999859, 1.002201e-05]</td>\n",
       "      <td>[8.479956e-06, 0.9999844, 7.094274e-06]</td>\n",
       "      <td>[5.0472172e-05, 0.9999002, 4.9328337e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2971</th>\n",
       "      <td>[0.0038565872, 0.99614257, 8.881378e-07]</td>\n",
       "      <td>[4.5307224e-05, 0.9999262, 2.8450857e-05]</td>\n",
       "      <td>[6.0800274e-05, 0.99990463, 3.452863e-05]</td>\n",
       "      <td>[0.0063016093, 0.9936958, 2.5687439e-06]</td>\n",
       "      <td>[2.0126467e-05, 0.9999597, 2.0086696e-05]</td>\n",
       "      <td>[6.4939864e-05, 0.9999056, 2.9467843e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2972</th>\n",
       "      <td>[8.80227e-05, 0.9999119, 7.357802e-08]</td>\n",
       "      <td>[1.3555074e-05, 0.99998295, 3.4149105e-06]</td>\n",
       "      <td>[3.9216964e-05, 0.9999268, 3.4026827e-05]</td>\n",
       "      <td>[8.079995e-05, 0.9999192, 5.4082644e-08]</td>\n",
       "      <td>[1.0396791e-05, 0.9999838, 5.8043006e-06]</td>\n",
       "      <td>[4.5115954e-05, 0.9999186, 3.6302445e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2973</th>\n",
       "      <td>[0.0015256882, 0.99838865, 8.565819e-05]</td>\n",
       "      <td>[1.9083656e-05, 0.99996376, 1.7134922e-05]</td>\n",
       "      <td>[4.0429615e-05, 0.99991965, 3.993368e-05]</td>\n",
       "      <td>[0.011751247, 0.9881732, 7.558385e-05]</td>\n",
       "      <td>[1.2453335e-05, 0.9999789, 8.749407e-06]</td>\n",
       "      <td>[4.8256217e-05, 0.99992466, 2.707365e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2974</th>\n",
       "      <td>[0.003918652, 0.9948362, 0.0012451549]</td>\n",
       "      <td>[9.222197e-05, 0.9998895, 1.8230361e-05]</td>\n",
       "      <td>[4.7365535e-05, 0.9999229, 2.9792693e-05]</td>\n",
       "      <td>[0.00568506, 0.9927377, 0.0015771401]</td>\n",
       "      <td>[0.0001500005, 0.9998109, 3.910941e-05]</td>\n",
       "      <td>[5.503107e-05, 0.9999263, 1.8559715e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2975</th>\n",
       "      <td>[0.0030204197, 0.996965, 1.4611041e-05]</td>\n",
       "      <td>[6.7992733e-06, 0.9999887, 4.5344436e-06]</td>\n",
       "      <td>[4.1351144e-05, 0.9999343, 2.4335111e-05]</td>\n",
       "      <td>[0.0013939402, 0.9986022, 3.8363874e-06]</td>\n",
       "      <td>[6.4183373e-06, 0.9999894, 4.162622e-06]</td>\n",
       "      <td>[5.2535448e-05, 0.9999291, 1.8341178e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2976</th>\n",
       "      <td>[4.9134807e-05, 0.0010570005, 0.9988939]</td>\n",
       "      <td>[6.8653746e-05, 0.00039237682, 0.999539]</td>\n",
       "      <td>[0.00010582302, 0.00013120257, 0.99976295]</td>\n",
       "      <td>[6.0102193e-05, 0.00019435708, 0.99974555]</td>\n",
       "      <td>[9.494994e-05, 9.341852e-05, 0.9998117]</td>\n",
       "      <td>[6.442183e-05, 0.00026567586, 0.9996699]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2977</th>\n",
       "      <td>[0.0022013718, 0.9977958, 2.8232332e-06]</td>\n",
       "      <td>[2.5639873e-05, 0.9999691, 5.26665e-06]</td>\n",
       "      <td>[5.2532116e-05, 0.9999238, 2.36427e-05]</td>\n",
       "      <td>[0.0014636873, 0.99853516, 1.1816319e-06]</td>\n",
       "      <td>[1.3079727e-05, 0.99998057, 6.3716625e-06]</td>\n",
       "      <td>[8.573858e-05, 0.99989474, 1.9587023e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2978</th>\n",
       "      <td>[0.03501048, 0.8525938, 0.11239565]</td>\n",
       "      <td>[0.009120187, 0.58312917, 0.40775067]</td>\n",
       "      <td>[0.011614903, 0.017614027, 0.97077113]</td>\n",
       "      <td>[0.16959986, 0.6882856, 0.14211453]</td>\n",
       "      <td>[0.0028511407, 0.9726916, 0.024457194]</td>\n",
       "      <td>[0.048414215, 0.036332503, 0.91525334]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2979</th>\n",
       "      <td>[0.0022350894, 0.9975278, 0.00023716157]</td>\n",
       "      <td>[0.022991601, 0.97697073, 3.7619695e-05]</td>\n",
       "      <td>[9.840087e-05, 0.99988043, 2.105654e-05]</td>\n",
       "      <td>[0.0017004324, 0.998131, 0.00016857816]</td>\n",
       "      <td>[0.001993633, 0.99798137, 2.5018555e-05]</td>\n",
       "      <td>[0.00042199215, 0.9995654, 1.2613746e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2980</th>\n",
       "      <td>[0.00010207494, 0.9990933, 0.0008046568]</td>\n",
       "      <td>[1.3390532e-05, 0.99994266, 4.3964817e-05]</td>\n",
       "      <td>[4.5865465e-05, 0.99991846, 3.5620935e-05]</td>\n",
       "      <td>[0.00012538987, 0.9989166, 0.0009579694]</td>\n",
       "      <td>[1.2080088e-05, 0.9999558, 3.2193548e-05]</td>\n",
       "      <td>[4.9621558e-05, 0.9999237, 2.6679521e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2981</th>\n",
       "      <td>[0.0009147265, 0.99754816, 0.0015371647]</td>\n",
       "      <td>[5.166359e-06, 0.99998033, 1.4554392e-05]</td>\n",
       "      <td>[4.0075138e-05, 0.9999299, 3.0068322e-05]</td>\n",
       "      <td>[0.009910916, 0.98633945, 0.0037496267]</td>\n",
       "      <td>[7.797504e-06, 0.9999833, 8.987446e-06]</td>\n",
       "      <td>[4.5276945e-05, 0.999912, 4.2622516e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2982</th>\n",
       "      <td>[0.010657696, 0.98857003, 0.0007722146]</td>\n",
       "      <td>[2.5701902e-05, 0.9999689, 5.4119646e-06]</td>\n",
       "      <td>[5.9823455e-05, 0.99991965, 2.0463423e-05]</td>\n",
       "      <td>[0.010100571, 0.98953, 0.00036947953]</td>\n",
       "      <td>[2.1385e-05, 0.9999696, 9.035759e-06]</td>\n",
       "      <td>[0.000110848836, 0.9998785, 1.0575883e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2983</th>\n",
       "      <td>[7.1778713e-06, 0.99999285, 1.9216408e-08]</td>\n",
       "      <td>[2.0386318e-05, 0.99997544, 4.1761705e-06]</td>\n",
       "      <td>[3.7208443e-05, 0.9999368, 2.5983507e-05]</td>\n",
       "      <td>[4.693483e-06, 0.99999535, 5.278013e-09]</td>\n",
       "      <td>[1.6061853e-05, 0.99997807, 5.8738624e-06]</td>\n",
       "      <td>[4.7603855e-05, 0.9999274, 2.5006837e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2984</th>\n",
       "      <td>[5.3931566e-05, 0.9999461, 1.0684114e-08]</td>\n",
       "      <td>[1.5818103e-05, 0.99998105, 3.1040581e-06]</td>\n",
       "      <td>[4.847375e-05, 0.99992764, 2.3856917e-05]</td>\n",
       "      <td>[0.00018049637, 0.9998196, 3.956059e-08]</td>\n",
       "      <td>[1.3465946e-05, 0.99998164, 4.930999e-06]</td>\n",
       "      <td>[5.0970968e-05, 0.9999248, 2.4255407e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2985</th>\n",
       "      <td>[2.847215e-05, 0.9999696, 1.8788573e-06]</td>\n",
       "      <td>[1.36609415e-05, 0.9999789, 7.3400865e-06]</td>\n",
       "      <td>[4.3154858e-05, 0.99993134, 2.5539239e-05]</td>\n",
       "      <td>[6.223519e-05, 0.99993634, 1.4155079e-06]</td>\n",
       "      <td>[2.096273e-05, 0.9999702, 8.876558e-06]</td>\n",
       "      <td>[6.0911327e-05, 0.99992037, 1.8680157e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2986</th>\n",
       "      <td>[0.06812218, 0.9316607, 0.00021711104]</td>\n",
       "      <td>[1.711237e-05, 0.99997807, 4.78472e-06]</td>\n",
       "      <td>[4.8654856e-05, 0.9999299, 2.1502186e-05]</td>\n",
       "      <td>[0.15090811, 0.84902453, 6.727478e-05]</td>\n",
       "      <td>[1.6015809e-05, 0.9999778, 6.1886635e-06]</td>\n",
       "      <td>[5.6788038e-05, 0.99992776, 1.5480824e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2987</th>\n",
       "      <td>[8.5589374e-05, 0.9999144, 6.913387e-09]</td>\n",
       "      <td>[0.0009913222, 0.99899787, 1.0895573e-05]</td>\n",
       "      <td>[8.760096e-05, 0.9998946, 1.7733038e-05]</td>\n",
       "      <td>[5.171855e-05, 0.99994826, 5.354327e-09]</td>\n",
       "      <td>[0.0001350763, 0.99985313, 1.1741929e-05]</td>\n",
       "      <td>[0.00010627753, 0.999884, 9.6272715e-06]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2988</th>\n",
       "      <td>[0.013489461, 0.9865102, 3.423163e-07]</td>\n",
       "      <td>[0.0017347258, 0.998251, 1.4286028e-05]</td>\n",
       "      <td>[0.00014023812, 0.9998367, 2.2949485e-05]</td>\n",
       "      <td>[0.05583445, 0.944164, 1.5643257e-06]</td>\n",
       "      <td>[0.015329741, 0.98452073, 0.00014957698]</td>\n",
       "      <td>[9.4243944e-05, 0.999892, 1.3700564e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2989</th>\n",
       "      <td>[0.5913317, 0.06488435, 0.34378394]</td>\n",
       "      <td>[0.07855934, 0.034699924, 0.88674074]</td>\n",
       "      <td>[0.026767017, 0.97229403, 0.0009390159]</td>\n",
       "      <td>[0.7011099, 0.03436372, 0.26452637]</td>\n",
       "      <td>[0.19836606, 0.009145228, 0.7924887]</td>\n",
       "      <td>[0.0008914261, 0.9989819, 0.00012670354]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2990</th>\n",
       "      <td>[0.37693667, 0.6213987, 0.0016646808]</td>\n",
       "      <td>[0.0005321994, 0.99932873, 0.00013918553]</td>\n",
       "      <td>[0.0012563246, 0.9986883, 5.536021e-05]</td>\n",
       "      <td>[0.42243278, 0.57580805, 0.0017591851]</td>\n",
       "      <td>[0.00037372633, 0.9994367, 0.00018960005]</td>\n",
       "      <td>[0.0006359286, 0.9993395, 2.4561226e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2991</th>\n",
       "      <td>[0.0994842, 0.9005138, 1.9788483e-06]</td>\n",
       "      <td>[0.07102145, 0.92889076, 8.781737e-05]</td>\n",
       "      <td>[0.009643516, 0.99015206, 0.00020440473]</td>\n",
       "      <td>[0.12673624, 0.87326056, 3.2519938e-06]</td>\n",
       "      <td>[0.020055577, 0.9798353, 0.000109053195]</td>\n",
       "      <td>[0.026576467, 0.9733123, 0.00011121639]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2992</th>\n",
       "      <td>[0.00030034542, 0.64738744, 0.35231218]</td>\n",
       "      <td>[0.0004756091, 0.3433362, 0.65618825]</td>\n",
       "      <td>[0.00011879748, 0.0012255969, 0.99865556]</td>\n",
       "      <td>[0.0004849217, 0.82130843, 0.17820662]</td>\n",
       "      <td>[0.00086266483, 0.16405062, 0.8350867]</td>\n",
       "      <td>[0.00011355382, 0.0038210351, 0.99606544]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2993</th>\n",
       "      <td>[0.15911534, 0.73032546, 0.110559225]</td>\n",
       "      <td>[0.0039536227, 0.98666734, 0.0093789585]</td>\n",
       "      <td>[0.00012895101, 0.9998222, 4.8844817e-05]</td>\n",
       "      <td>[0.43858352, 0.35041612, 0.21100032]</td>\n",
       "      <td>[0.011102119, 0.98604786, 0.002850076]</td>\n",
       "      <td>[0.0002352526, 0.999734, 3.0689174e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2994</th>\n",
       "      <td>[0.0063568847, 0.009023929, 0.9846192]</td>\n",
       "      <td>[0.00021208546, 0.00028024692, 0.99950767]</td>\n",
       "      <td>[0.00017807873, 0.00017199268, 0.9996499]</td>\n",
       "      <td>[0.007363762, 0.03469444, 0.9579418]</td>\n",
       "      <td>[0.00038859833, 0.0002748932, 0.99933654]</td>\n",
       "      <td>[0.000146678, 0.00027509205, 0.9995782]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2995</th>\n",
       "      <td>[2.7447456e-07, 0.99999976, 4.6237902e-09]</td>\n",
       "      <td>[7.92075e-06, 0.9999862, 5.9381587e-06]</td>\n",
       "      <td>[4.268119e-05, 0.9999305, 2.687175e-05]</td>\n",
       "      <td>[5.5074986e-07, 0.9999994, 3.1386886e-09]</td>\n",
       "      <td>[1.1006394e-05, 0.9999789, 1.0096202e-05]</td>\n",
       "      <td>[5.4615495e-05, 0.99992585, 1.9504161e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2996</th>\n",
       "      <td>[0.00035443844, 0.17763066, 0.8220149]</td>\n",
       "      <td>[9.221037e-05, 0.012253262, 0.9876545]</td>\n",
       "      <td>[6.109948e-05, 0.00012058155, 0.9998184]</td>\n",
       "      <td>[0.00054795184, 0.3214301, 0.67802197]</td>\n",
       "      <td>[9.5943695e-05, 0.002728676, 0.99717534]</td>\n",
       "      <td>[3.364683e-05, 0.0001740074, 0.9997924]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2997</th>\n",
       "      <td>[2.6139876e-06, 0.9999968, 6.370505e-07]</td>\n",
       "      <td>[0.0004047803, 0.96675104, 0.032844197]</td>\n",
       "      <td>[0.0005413972, 0.9994062, 5.2410072e-05]</td>\n",
       "      <td>[1.0340856e-05, 0.9999889, 7.5271475e-07]</td>\n",
       "      <td>[0.00015814147, 0.99613684, 0.0037049148]</td>\n",
       "      <td>[0.00016778223, 0.99978524, 4.694916e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2998</th>\n",
       "      <td>[0.07386664, 0.8809504, 0.045182902]</td>\n",
       "      <td>[0.00016258945, 0.9986877, 0.0011496462]</td>\n",
       "      <td>[4.5637506e-05, 0.99992454, 2.9840941e-05]</td>\n",
       "      <td>[0.02169743, 0.95809954, 0.020203082]</td>\n",
       "      <td>[0.00021270107, 0.9985361, 0.0012511375]</td>\n",
       "      <td>[6.2617466e-05, 0.99991655, 2.0800484e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2999</th>\n",
       "      <td>[1.2920809e-05, 0.9999871, 3.234183e-09]</td>\n",
       "      <td>[0.00026157586, 0.9997327, 5.7703e-06]</td>\n",
       "      <td>[3.9747934e-05, 0.9999399, 2.0363928e-05]</td>\n",
       "      <td>[3.4449095e-05, 0.99996555, 3.870269e-09]</td>\n",
       "      <td>[6.391212e-05, 0.9999279, 8.209448e-06]</td>\n",
       "      <td>[6.275527e-05, 0.9999261, 1.1155761e-05]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3000 rows × 6 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                             nnlm-128_tsize_3000  \\\n",
       "0        [4.233589e-05, 0.9999193, 3.840332e-05]   \n",
       "1       [0.0013433183, 0.99797875, 0.0006780409]   \n",
       "2      [0.0011234806, 0.99871135, 0.00016510795]   \n",
       "3       [0.0004791329, 0.9995173, 3.5879834e-06]   \n",
       "4     [3.1970467e-05, 0.99996805, 7.1044792e-09]   \n",
       "5           [0.081907175, 0.7386889, 0.17940396]   \n",
       "6          [0.0006111432, 0.9598467, 0.03954213]   \n",
       "7       [8.144646e-07, 0.9998646, 0.00013461642]   \n",
       "8       [0.00034172254, 0.9969459, 0.0027123827]   \n",
       "9       [4.506633e-05, 0.9999548, 1.6939906e-07]   \n",
       "10       [0.028411375, 6.653153e-05, 0.97152215]   \n",
       "11    [0.00021463603, 0.99944204, 0.00034326394]   \n",
       "12         [0.9867077, 0.009154943, 0.004137294]   \n",
       "13       [0.005461895, 0.99341416, 0.0011239109]   \n",
       "14       [7.315173e-06, 0.9999902, 2.536194e-06]   \n",
       "15        [4.2200754e-05, 0.14187208, 0.8580857]   \n",
       "16            [0.19498013, 0.28203487, 0.522985]   \n",
       "17         [0.000438633, 0.082147196, 0.9174142]   \n",
       "18      [0.0076374356, 0.99044526, 0.0019172331]   \n",
       "19     [0.0009561412, 0.99867696, 0.00036692788]   \n",
       "20       [0.018522294, 0.98142964, 4.812113e-05]   \n",
       "21    [3.3813408e-06, 0.99999654, 1.1062413e-07]   \n",
       "22          [0.022528995, 0.07266271, 0.9048083]   \n",
       "23      [3.0262268e-05, 0.9998877, 8.194801e-05]   \n",
       "24    [0.00017858457, 0.99981993, 1.5306515e-06]   \n",
       "25     [0.00028489454, 0.99971455, 4.936044e-07]   \n",
       "26       [0.0001413786, 0.9998586, 4.243305e-08]   \n",
       "27     [0.0021451085, 0.99774146, 0.00011343301]   \n",
       "28       [0.00015609733, 0.086633295, 0.9132106]   \n",
       "29     [6.387743e-05, 0.99993515, 1.0078197e-06]   \n",
       "...                                          ...   \n",
       "2970   [0.00017322828, 0.9997682, 5.8635833e-05]   \n",
       "2971    [0.0038565872, 0.99614257, 8.881378e-07]   \n",
       "2972      [8.80227e-05, 0.9999119, 7.357802e-08]   \n",
       "2973    [0.0015256882, 0.99838865, 8.565819e-05]   \n",
       "2974      [0.003918652, 0.9948362, 0.0012451549]   \n",
       "2975     [0.0030204197, 0.996965, 1.4611041e-05]   \n",
       "2976    [4.9134807e-05, 0.0010570005, 0.9988939]   \n",
       "2977    [0.0022013718, 0.9977958, 2.8232332e-06]   \n",
       "2978         [0.03501048, 0.8525938, 0.11239565]   \n",
       "2979    [0.0022350894, 0.9975278, 0.00023716157]   \n",
       "2980    [0.00010207494, 0.9990933, 0.0008046568]   \n",
       "2981    [0.0009147265, 0.99754816, 0.0015371647]   \n",
       "2982     [0.010657696, 0.98857003, 0.0007722146]   \n",
       "2983  [7.1778713e-06, 0.99999285, 1.9216408e-08]   \n",
       "2984   [5.3931566e-05, 0.9999461, 1.0684114e-08]   \n",
       "2985    [2.847215e-05, 0.9999696, 1.8788573e-06]   \n",
       "2986      [0.06812218, 0.9316607, 0.00021711104]   \n",
       "2987    [8.5589374e-05, 0.9999144, 6.913387e-09]   \n",
       "2988      [0.013489461, 0.9865102, 3.423163e-07]   \n",
       "2989         [0.5913317, 0.06488435, 0.34378394]   \n",
       "2990       [0.37693667, 0.6213987, 0.0016646808]   \n",
       "2991       [0.0994842, 0.9005138, 1.9788483e-06]   \n",
       "2992     [0.00030034542, 0.64738744, 0.35231218]   \n",
       "2993       [0.15911534, 0.73032546, 0.110559225]   \n",
       "2994      [0.0063568847, 0.009023929, 0.9846192]   \n",
       "2995  [2.7447456e-07, 0.99999976, 4.6237902e-09]   \n",
       "2996      [0.00035443844, 0.17763066, 0.8220149]   \n",
       "2997    [2.6139876e-06, 0.9999968, 6.370505e-07]   \n",
       "2998        [0.07386664, 0.8809504, 0.045182902]   \n",
       "2999    [1.2920809e-05, 0.9999871, 3.234183e-09]   \n",
       "\n",
       "                                  DAN_tsize_3000  \\\n",
       "0       [1.6480626e-05, 0.9998909, 9.264603e-05]   \n",
       "1        [7.580178e-05, 0.96989506, 0.030029025]   \n",
       "2     [2.4509067e-05, 0.99997044, 5.0195786e-06]   \n",
       "3      [1.3438928e-05, 0.99997985, 6.728323e-06]   \n",
       "4         [1.818452e-05, 0.999979, 2.731076e-06]   \n",
       "5         [0.0016064331, 0.005709461, 0.9926842]   \n",
       "6       [1.0904048e-05, 0.999979, 1.0073294e-05]   \n",
       "7       [0.00010322058, 0.98762506, 0.012271626]   \n",
       "8      [1.3682003e-05, 0.99988484, 0.0001014915]   \n",
       "9        [6.86458e-06, 0.9999902, 2.8723064e-06]   \n",
       "10       [0.0018834813, 0.0010752678, 0.9970413]   \n",
       "11     [6.1479936e-05, 0.9999318, 6.7185747e-06]   \n",
       "12           [0.17751005, 0.43451717, 0.3879728]   \n",
       "13     [1.4571097e-05, 0.9999753, 1.0091025e-05]   \n",
       "14     [5.9142135e-06, 0.9999912, 2.8827149e-06]   \n",
       "15         [0.0013588266, 0.31469786, 0.6839433]   \n",
       "16       [6.760271e-05, 0.0001779354, 0.9997545]   \n",
       "17           [0.0026416993, 0.7575073, 0.239851]   \n",
       "18     [5.1096198e-05, 0.9999249, 2.3965078e-05]   \n",
       "19           [0.3191397, 0.12964316, 0.55121714]   \n",
       "20    [1.42467015e-05, 0.99998283, 2.856866e-06]   \n",
       "21     [4.903143e-05, 0.99991894, 3.2119373e-05]   \n",
       "22         [0.000281977, 0.63859683, 0.36112115]   \n",
       "23      [1.4272013e-05, 0.9999516, 3.404179e-05]   \n",
       "24       [8.602782e-06, 0.9999882, 3.274369e-06]   \n",
       "25       [0.001231889, 0.9987533, 1.4753598e-05]   \n",
       "26     [1.1435212e-05, 0.9999863, 2.2247548e-06]   \n",
       "27     [9.384714e-05, 0.99988794, 1.8204153e-05]   \n",
       "28       [6.858496e-05, 0.0013311019, 0.9986003]   \n",
       "29     [8.210377e-06, 0.99998844, 3.3694355e-06]   \n",
       "...                                          ...   \n",
       "2970     [7.1989675e-06, 0.9999864, 6.41879e-06]   \n",
       "2971   [4.5307224e-05, 0.9999262, 2.8450857e-05]   \n",
       "2972  [1.3555074e-05, 0.99998295, 3.4149105e-06]   \n",
       "2973  [1.9083656e-05, 0.99996376, 1.7134922e-05]   \n",
       "2974    [9.222197e-05, 0.9998895, 1.8230361e-05]   \n",
       "2975   [6.7992733e-06, 0.9999887, 4.5344436e-06]   \n",
       "2976    [6.8653746e-05, 0.00039237682, 0.999539]   \n",
       "2977     [2.5639873e-05, 0.9999691, 5.26665e-06]   \n",
       "2978       [0.009120187, 0.58312917, 0.40775067]   \n",
       "2979    [0.022991601, 0.97697073, 3.7619695e-05]   \n",
       "2980  [1.3390532e-05, 0.99994266, 4.3964817e-05]   \n",
       "2981   [5.166359e-06, 0.99998033, 1.4554392e-05]   \n",
       "2982   [2.5701902e-05, 0.9999689, 5.4119646e-06]   \n",
       "2983  [2.0386318e-05, 0.99997544, 4.1761705e-06]   \n",
       "2984  [1.5818103e-05, 0.99998105, 3.1040581e-06]   \n",
       "2985  [1.36609415e-05, 0.9999789, 7.3400865e-06]   \n",
       "2986     [1.711237e-05, 0.99997807, 4.78472e-06]   \n",
       "2987   [0.0009913222, 0.99899787, 1.0895573e-05]   \n",
       "2988     [0.0017347258, 0.998251, 1.4286028e-05]   \n",
       "2989       [0.07855934, 0.034699924, 0.88674074]   \n",
       "2990   [0.0005321994, 0.99932873, 0.00013918553]   \n",
       "2991      [0.07102145, 0.92889076, 8.781737e-05]   \n",
       "2992       [0.0004756091, 0.3433362, 0.65618825]   \n",
       "2993    [0.0039536227, 0.98666734, 0.0093789585]   \n",
       "2994  [0.00021208546, 0.00028024692, 0.99950767]   \n",
       "2995     [7.92075e-06, 0.9999862, 5.9381587e-06]   \n",
       "2996      [9.221037e-05, 0.012253262, 0.9876545]   \n",
       "2997     [0.0004047803, 0.96675104, 0.032844197]   \n",
       "2998    [0.00016258945, 0.9986877, 0.0011496462]   \n",
       "2999      [0.00026157586, 0.9997327, 5.7703e-06]   \n",
       "\n",
       "                                 univ3_tsize_3000  \\\n",
       "0        [6.398087e-05, 0.9998853, 5.0707247e-05]   \n",
       "1          [0.0001076711, 0.999806, 8.637823e-05]   \n",
       "2        [4.7642898e-05, 0.9999232, 2.913791e-05]   \n",
       "3      [3.8377642e-05, 0.99993515, 2.6411251e-05]   \n",
       "4       [6.852479e-05, 0.99991035, 2.1157512e-05]   \n",
       "5          [0.0004184056, 0.9995516, 3.00833e-05]   \n",
       "6        [4.641684e-05, 0.99991405, 3.959844e-05]   \n",
       "7          [4.8737867e-05, 0.9998951, 5.6094e-05]   \n",
       "8        [0.0001048133, 0.9995121, 0.00038309305]   \n",
       "9        [4.578529e-05, 0.99992275, 3.144341e-05]   \n",
       "10      [0.00011976459, 0.00021957459, 0.9996606]   \n",
       "11       [5.3447086e-05, 0.9999268, 1.981259e-05]   \n",
       "12       [0.00017330452, 0.00029272275, 0.999534]   \n",
       "13       [4.281666e-05, 0.9999161, 4.1079733e-05]   \n",
       "14      [4.5550813e-05, 0.9999248, 2.9640507e-05]   \n",
       "15       [0.00026939638, 0.0002832935, 0.9994473]   \n",
       "16      [0.00012323644, 0.00015144021, 0.9997253]   \n",
       "17       [8.45979e-05, 0.00014179268, 0.99977356]   \n",
       "18       [3.943014e-05, 0.9999386, 2.1962676e-05]   \n",
       "19    [9.3881274e-05, 0.000116055446, 0.99979013]   \n",
       "20       [5.2919135e-05, 0.999915, 3.2032738e-05]   \n",
       "21      [3.9518014e-05, 0.99992836, 3.202855e-05]   \n",
       "22       [7.190047e-05, 0.00026368184, 0.9996644]   \n",
       "23      [3.8888775e-05, 0.9999335, 2.7662583e-05]   \n",
       "24      [3.7480313e-05, 0.9999404, 2.2209082e-05]   \n",
       "25      [5.077563e-05, 0.99992466, 2.4521478e-05]   \n",
       "26      [4.3633758e-05, 0.9999362, 2.0162879e-05]   \n",
       "27     [3.7041853e-05, 0.99993753, 2.5337724e-05]   \n",
       "28        [5.630375e-05, 0.00017668531, 0.999767]   \n",
       "29      [4.0054543e-05, 0.9999366, 2.3396988e-05]   \n",
       "...                                           ...   \n",
       "2970     [4.035385e-05, 0.9999207, 3.8841205e-05]   \n",
       "2971    [6.0800274e-05, 0.99990463, 3.452863e-05]   \n",
       "2972    [3.9216964e-05, 0.9999268, 3.4026827e-05]   \n",
       "2973    [4.0429615e-05, 0.99991965, 3.993368e-05]   \n",
       "2974    [4.7365535e-05, 0.9999229, 2.9792693e-05]   \n",
       "2975    [4.1351144e-05, 0.9999343, 2.4335111e-05]   \n",
       "2976   [0.00010582302, 0.00013120257, 0.99976295]   \n",
       "2977      [5.2532116e-05, 0.9999238, 2.36427e-05]   \n",
       "2978       [0.011614903, 0.017614027, 0.97077113]   \n",
       "2979     [9.840087e-05, 0.99988043, 2.105654e-05]   \n",
       "2980   [4.5865465e-05, 0.99991846, 3.5620935e-05]   \n",
       "2981    [4.0075138e-05, 0.9999299, 3.0068322e-05]   \n",
       "2982   [5.9823455e-05, 0.99991965, 2.0463423e-05]   \n",
       "2983    [3.7208443e-05, 0.9999368, 2.5983507e-05]   \n",
       "2984    [4.847375e-05, 0.99992764, 2.3856917e-05]   \n",
       "2985   [4.3154858e-05, 0.99993134, 2.5539239e-05]   \n",
       "2986    [4.8654856e-05, 0.9999299, 2.1502186e-05]   \n",
       "2987     [8.760096e-05, 0.9998946, 1.7733038e-05]   \n",
       "2988    [0.00014023812, 0.9998367, 2.2949485e-05]   \n",
       "2989      [0.026767017, 0.97229403, 0.0009390159]   \n",
       "2990      [0.0012563246, 0.9986883, 5.536021e-05]   \n",
       "2991     [0.009643516, 0.99015206, 0.00020440473]   \n",
       "2992    [0.00011879748, 0.0012255969, 0.99865556]   \n",
       "2993    [0.00012895101, 0.9998222, 4.8844817e-05]   \n",
       "2994    [0.00017807873, 0.00017199268, 0.9996499]   \n",
       "2995      [4.268119e-05, 0.9999305, 2.687175e-05]   \n",
       "2996     [6.109948e-05, 0.00012058155, 0.9998184]   \n",
       "2997     [0.0005413972, 0.9994062, 5.2410072e-05]   \n",
       "2998   [4.5637506e-05, 0.99992454, 2.9840941e-05]   \n",
       "2999    [3.9747934e-05, 0.9999399, 2.0363928e-05]   \n",
       "\n",
       "                    nnlm-128_tsize_18783_epoch30  \\\n",
       "0       [0.00013908364, 0.9998254, 3.547703e-05]   \n",
       "1       [0.002030536, 0.99750996, 0.00045953243]   \n",
       "2       [0.0006725391, 0.99926144, 6.598066e-05]   \n",
       "3       [0.0003392862, 0.9996592, 1.5918495e-06]   \n",
       "4      [4.3006254e-05, 0.99995697, 8.022627e-09]   \n",
       "5            [0.08427183, 0.8487675, 0.06696065]   \n",
       "6          [0.001172376, 0.9711367, 0.027690923]   \n",
       "7     [2.1608248e-06, 0.99978226, 0.00021563252]   \n",
       "8       [0.00057175703, 0.9981173, 0.0013109304]   \n",
       "9      [7.477624e-05, 0.99992514, 1.1235904e-07]   \n",
       "10       [0.021710116, 2.5961535e-05, 0.9782639]   \n",
       "11     [0.00060418097, 0.99912363, 0.0002722138]   \n",
       "12           [0.9535876, 0.0369543, 0.009458121]   \n",
       "13       [0.010873717, 0.98802453, 0.0011016902]   \n",
       "14    [1.2437995e-05, 0.99998605, 1.4981332e-06]   \n",
       "15        [0.00022437192, 0.40065616, 0.5991194]   \n",
       "16            [0.42409116, 0.3661773, 0.2097315]   \n",
       "17       [0.00028815673, 0.02189521, 0.97781664]   \n",
       "18      [0.0049450295, 0.99437827, 0.0006767206]   \n",
       "19      [0.0021360728, 0.9973984, 0.00046561955]   \n",
       "20        [0.012432584, 0.9875252, 4.218152e-05]   \n",
       "21    [3.4952702e-06, 0.99999654, 5.4307318e-08]   \n",
       "22        [0.009834621, 0.031162515, 0.95900285]   \n",
       "23      [2.863268e-05, 0.9999547, 1.6737697e-05]   \n",
       "24    [0.00046800202, 0.99953043, 1.5787145e-06]   \n",
       "25      [0.0009937334, 0.9990043, 2.0188118e-06]   \n",
       "26      [0.00067893823, 0.999321, 1.2991575e-07]   \n",
       "27       [0.002937752, 0.9970304, 3.1891446e-05]   \n",
       "28        [0.00021700047, 0.12237597, 0.8774071]   \n",
       "29      [0.00041335513, 0.99958247, 4.15565e-06]   \n",
       "...                                          ...   \n",
       "2970     [0.00013102862, 0.999859, 1.002201e-05]   \n",
       "2971    [0.0063016093, 0.9936958, 2.5687439e-06]   \n",
       "2972    [8.079995e-05, 0.9999192, 5.4082644e-08]   \n",
       "2973      [0.011751247, 0.9881732, 7.558385e-05]   \n",
       "2974       [0.00568506, 0.9927377, 0.0015771401]   \n",
       "2975    [0.0013939402, 0.9986022, 3.8363874e-06]   \n",
       "2976  [6.0102193e-05, 0.00019435708, 0.99974555]   \n",
       "2977   [0.0014636873, 0.99853516, 1.1816319e-06]   \n",
       "2978         [0.16959986, 0.6882856, 0.14211453]   \n",
       "2979     [0.0017004324, 0.998131, 0.00016857816]   \n",
       "2980    [0.00012538987, 0.9989166, 0.0009579694]   \n",
       "2981     [0.009910916, 0.98633945, 0.0037496267]   \n",
       "2982       [0.010100571, 0.98953, 0.00036947953]   \n",
       "2983    [4.693483e-06, 0.99999535, 5.278013e-09]   \n",
       "2984    [0.00018049637, 0.9998196, 3.956059e-08]   \n",
       "2985   [6.223519e-05, 0.99993634, 1.4155079e-06]   \n",
       "2986      [0.15090811, 0.84902453, 6.727478e-05]   \n",
       "2987    [5.171855e-05, 0.99994826, 5.354327e-09]   \n",
       "2988       [0.05583445, 0.944164, 1.5643257e-06]   \n",
       "2989         [0.7011099, 0.03436372, 0.26452637]   \n",
       "2990      [0.42243278, 0.57580805, 0.0017591851]   \n",
       "2991     [0.12673624, 0.87326056, 3.2519938e-06]   \n",
       "2992      [0.0004849217, 0.82130843, 0.17820662]   \n",
       "2993        [0.43858352, 0.35041612, 0.21100032]   \n",
       "2994        [0.007363762, 0.03469444, 0.9579418]   \n",
       "2995   [5.5074986e-07, 0.9999994, 3.1386886e-09]   \n",
       "2996      [0.00054795184, 0.3214301, 0.67802197]   \n",
       "2997   [1.0340856e-05, 0.9999889, 7.5271475e-07]   \n",
       "2998       [0.02169743, 0.95809954, 0.020203082]   \n",
       "2999   [3.4449095e-05, 0.99996555, 3.870269e-09]   \n",
       "\n",
       "                         DAN_tsize_18783_epoch30  \\\n",
       "0     [2.5715732e-05, 0.9998716, 0.000102616505]   \n",
       "1        [2.547616e-05, 0.9989743, 0.0010002507]   \n",
       "2        [1.7972814e-05, 0.999974, 8.038951e-06]   \n",
       "3      [1.8328115e-05, 0.9999672, 1.4433712e-05]   \n",
       "4      [1.7900435e-05, 0.9999765, 5.5927703e-06]   \n",
       "5          [0.0026555285, 0.06760905, 0.9297355]   \n",
       "6      [2.252558e-05, 0.99995816, 1.9321673e-05]   \n",
       "7         [0.00030092022, 0.9585573, 0.04114188]   \n",
       "8      [1.6481239e-05, 0.99988806, 9.542799e-05]   \n",
       "9      [5.6351914e-06, 0.9999902, 4.1621734e-06]   \n",
       "10     [0.00081079284, 0.0007934927, 0.99839574]   \n",
       "11     [6.452492e-05, 0.99992394, 1.1570731e-05]   \n",
       "12           [0.06544465, 0.4085732, 0.52598214]   \n",
       "13    [1.33834255e-05, 0.99996865, 1.796761e-05]   \n",
       "14     [6.814804e-06, 0.99998903, 4.2227966e-06]   \n",
       "15        [0.0018520205, 0.23281698, 0.76533103]   \n",
       "16    [0.00018075421, 0.00020491573, 0.99961436]   \n",
       "17         [0.0020194512, 0.35163933, 0.6463412]   \n",
       "18      [0.0011020844, 0.9986578, 0.00024008311]   \n",
       "19          [0.06276215, 0.010787978, 0.9264499]   \n",
       "20      [6.9983995e-05, 0.9999213, 8.655985e-06]   \n",
       "21     [3.8394963e-05, 0.9999201, 4.1535608e-05]   \n",
       "22     [3.8877777e-05, 0.99868363, 0.0012774161]   \n",
       "23      [9.463277e-06, 0.9999757, 1.4949457e-05]   \n",
       "24       [7.325024e-06, 0.9999875, 5.193338e-06]   \n",
       "25     [0.0066787857, 0.99322104, 0.00010020443]   \n",
       "26    [1.5814856e-05, 0.99997985, 4.2840243e-06]   \n",
       "27      [0.000570712, 0.9993119, 0.000117387346]   \n",
       "28      [0.00014342695, 0.0009917042, 0.9988649]   \n",
       "29     [1.3541771e-05, 0.9999796, 6.7871615e-06]   \n",
       "...                                          ...   \n",
       "2970     [8.479956e-06, 0.9999844, 7.094274e-06]   \n",
       "2971   [2.0126467e-05, 0.9999597, 2.0086696e-05]   \n",
       "2972   [1.0396791e-05, 0.9999838, 5.8043006e-06]   \n",
       "2973    [1.2453335e-05, 0.9999789, 8.749407e-06]   \n",
       "2974     [0.0001500005, 0.9998109, 3.910941e-05]   \n",
       "2975    [6.4183373e-06, 0.9999894, 4.162622e-06]   \n",
       "2976     [9.494994e-05, 9.341852e-05, 0.9998117]   \n",
       "2977  [1.3079727e-05, 0.99998057, 6.3716625e-06]   \n",
       "2978      [0.0028511407, 0.9726916, 0.024457194]   \n",
       "2979    [0.001993633, 0.99798137, 2.5018555e-05]   \n",
       "2980   [1.2080088e-05, 0.9999558, 3.2193548e-05]   \n",
       "2981     [7.797504e-06, 0.9999833, 8.987446e-06]   \n",
       "2982       [2.1385e-05, 0.9999696, 9.035759e-06]   \n",
       "2983  [1.6061853e-05, 0.99997807, 5.8738624e-06]   \n",
       "2984   [1.3465946e-05, 0.99998164, 4.930999e-06]   \n",
       "2985     [2.096273e-05, 0.9999702, 8.876558e-06]   \n",
       "2986   [1.6015809e-05, 0.9999778, 6.1886635e-06]   \n",
       "2987   [0.0001350763, 0.99985313, 1.1741929e-05]   \n",
       "2988    [0.015329741, 0.98452073, 0.00014957698]   \n",
       "2989        [0.19836606, 0.009145228, 0.7924887]   \n",
       "2990   [0.00037372633, 0.9994367, 0.00018960005]   \n",
       "2991    [0.020055577, 0.9798353, 0.000109053195]   \n",
       "2992      [0.00086266483, 0.16405062, 0.8350867]   \n",
       "2993      [0.011102119, 0.98604786, 0.002850076]   \n",
       "2994   [0.00038859833, 0.0002748932, 0.99933654]   \n",
       "2995   [1.1006394e-05, 0.9999789, 1.0096202e-05]   \n",
       "2996    [9.5943695e-05, 0.002728676, 0.99717534]   \n",
       "2997   [0.00015814147, 0.99613684, 0.0037049148]   \n",
       "2998    [0.00021270107, 0.9985361, 0.0012511375]   \n",
       "2999     [6.391212e-05, 0.9999279, 8.209448e-06]   \n",
       "\n",
       "                       univ3_tsize_18783_epoch30  \n",
       "0         [7.91436e-05, 0.9998338, 8.701929e-05]  \n",
       "1         [7.870038e-05, 0.9997701, 0.000151296]  \n",
       "2       [6.0237904e-05, 0.9999143, 2.548163e-05]  \n",
       "3       [4.4898494e-05, 0.9999287, 2.633073e-05]  \n",
       "4        [7.507247e-05, 0.9999107, 1.421926e-05]  \n",
       "5     [0.00016169962, 0.99982136, 1.6898892e-05]  \n",
       "6      [5.2135645e-05, 0.99991536, 3.253769e-05]  \n",
       "7       [6.164825e-05, 0.9998791, 5.9188635e-05]  \n",
       "8      [6.790512e-05, 0.99990237, 2.9676981e-05]  \n",
       "9      [6.1978855e-05, 0.9999176, 2.0371614e-05]  \n",
       "10      [7.8857935e-05, 0.0003978529, 0.9995233]  \n",
       "11      [7.210481e-05, 0.9999161, 1.1803859e-05]  \n",
       "12     [0.00020762562, 0.00060981687, 0.9991825]  \n",
       "13     [4.5374585e-05, 0.9999213, 3.3305845e-05]  \n",
       "14     [5.612093e-05, 0.99992216, 2.1637394e-05]  \n",
       "15    [0.00013584517, 0.00024865704, 0.99961543]  \n",
       "16      [8.948143e-05, 0.00019811906, 0.9997124]  \n",
       "17       [0.00022410582, 0.013034229, 0.9867417]  \n",
       "18      [5.753277e-05, 0.99992466, 1.777614e-05]  \n",
       "19     [5.8154856e-05, 0.00014961355, 0.9997923]  \n",
       "20     [7.0063914e-05, 0.9999087, 2.1275844e-05]  \n",
       "21    [4.8331603e-05, 0.99993205, 1.9714593e-05]  \n",
       "22    [5.3785814e-05, 0.00030585722, 0.99964035]  \n",
       "23    [5.3389886e-05, 0.99991834, 2.8288976e-05]  \n",
       "24      [4.9465325e-05, 0.9999325, 1.801631e-05]  \n",
       "25       [5.763131e-05, 0.999928, 1.4367954e-05]  \n",
       "26       [7.19162e-05, 0.9999155, 1.2658613e-05]  \n",
       "27     [5.7950838e-05, 0.9999305, 1.1548351e-05]  \n",
       "28       [7.705059e-05, 0.0055413716, 0.9943816]  \n",
       "29    [5.4944056e-05, 0.99992704, 1.7942342e-05]  \n",
       "...                                          ...  \n",
       "2970   [5.0472172e-05, 0.9999002, 4.9328337e-05]  \n",
       "2971   [6.4939864e-05, 0.9999056, 2.9467843e-05]  \n",
       "2972   [4.5115954e-05, 0.9999186, 3.6302445e-05]  \n",
       "2973   [4.8256217e-05, 0.99992466, 2.707365e-05]  \n",
       "2974    [5.503107e-05, 0.9999263, 1.8559715e-05]  \n",
       "2975   [5.2535448e-05, 0.9999291, 1.8341178e-05]  \n",
       "2976    [6.442183e-05, 0.00026567586, 0.9996699]  \n",
       "2977   [8.573858e-05, 0.99989474, 1.9587023e-05]  \n",
       "2978      [0.048414215, 0.036332503, 0.91525334]  \n",
       "2979   [0.00042199215, 0.9995654, 1.2613746e-05]  \n",
       "2980   [4.9621558e-05, 0.9999237, 2.6679521e-05]  \n",
       "2981    [4.5276945e-05, 0.999912, 4.2622516e-05]  \n",
       "2982  [0.000110848836, 0.9998785, 1.0575883e-05]  \n",
       "2983   [4.7603855e-05, 0.9999274, 2.5006837e-05]  \n",
       "2984   [5.0970968e-05, 0.9999248, 2.4255407e-05]  \n",
       "2985  [6.0911327e-05, 0.99992037, 1.8680157e-05]  \n",
       "2986  [5.6788038e-05, 0.99992776, 1.5480824e-05]  \n",
       "2987    [0.00010627753, 0.999884, 9.6272715e-06]  \n",
       "2988    [9.4243944e-05, 0.999892, 1.3700564e-05]  \n",
       "2989    [0.0008914261, 0.9989819, 0.00012670354]  \n",
       "2990    [0.0006359286, 0.9993395, 2.4561226e-05]  \n",
       "2991     [0.026576467, 0.9733123, 0.00011121639]  \n",
       "2992   [0.00011355382, 0.0038210351, 0.99606544]  \n",
       "2993     [0.0002352526, 0.999734, 3.0689174e-05]  \n",
       "2994     [0.000146678, 0.00027509205, 0.9995782]  \n",
       "2995  [5.4615495e-05, 0.99992585, 1.9504161e-05]  \n",
       "2996     [3.364683e-05, 0.0001740074, 0.9997924]  \n",
       "2997   [0.00016778223, 0.99978524, 4.694916e-05]  \n",
       "2998  [6.2617466e-05, 0.99991655, 2.0800484e-05]  \n",
       "2999    [6.275527e-05, 0.9999261, 1.1155761e-05]  \n",
       "\n",
       "[3000 rows x 6 columns]"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "probs = []\n",
    "for i in range(len(names)): #(len(all))\n",
    "    probs.append(all[i]['probs'])\n",
    "\n",
    "probs = pd.DataFrame(probs).T\n",
    "probs.columns= names\n",
    "probs\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "# last 3 columns\n",
    "probs = probs.iloc[:, 3:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "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>nnlm-128_tsize_18783_epoch30</th>\n",
       "      <th>DAN_tsize_18783_epoch30</th>\n",
       "      <th>univ3_tsize_18783_epoch30</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>[0.00013908364, 0.9998254, 3.547703e-05]</td>\n",
       "      <td>[2.5715732e-05, 0.9998716, 0.000102616505]</td>\n",
       "      <td>[7.91436e-05, 0.9998338, 8.701929e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>[0.002030536, 0.99750996, 0.00045953243]</td>\n",
       "      <td>[2.547616e-05, 0.9989743, 0.0010002507]</td>\n",
       "      <td>[7.870038e-05, 0.9997701, 0.000151296]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>[0.0006725391, 0.99926144, 6.598066e-05]</td>\n",
       "      <td>[1.7972814e-05, 0.999974, 8.038951e-06]</td>\n",
       "      <td>[6.0237904e-05, 0.9999143, 2.548163e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>[0.0003392862, 0.9996592, 1.5918495e-06]</td>\n",
       "      <td>[1.8328115e-05, 0.9999672, 1.4433712e-05]</td>\n",
       "      <td>[4.4898494e-05, 0.9999287, 2.633073e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>[4.3006254e-05, 0.99995697, 8.022627e-09]</td>\n",
       "      <td>[1.7900435e-05, 0.9999765, 5.5927703e-06]</td>\n",
       "      <td>[7.507247e-05, 0.9999107, 1.421926e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>[0.08427183, 0.8487675, 0.06696065]</td>\n",
       "      <td>[0.0026555285, 0.06760905, 0.9297355]</td>\n",
       "      <td>[0.00016169962, 0.99982136, 1.6898892e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>[0.001172376, 0.9711367, 0.027690923]</td>\n",
       "      <td>[2.252558e-05, 0.99995816, 1.9321673e-05]</td>\n",
       "      <td>[5.2135645e-05, 0.99991536, 3.253769e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>[2.1608248e-06, 0.99978226, 0.00021563252]</td>\n",
       "      <td>[0.00030092022, 0.9585573, 0.04114188]</td>\n",
       "      <td>[6.164825e-05, 0.9998791, 5.9188635e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>[0.00057175703, 0.9981173, 0.0013109304]</td>\n",
       "      <td>[1.6481239e-05, 0.99988806, 9.542799e-05]</td>\n",
       "      <td>[6.790512e-05, 0.99990237, 2.9676981e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>[7.477624e-05, 0.99992514, 1.1235904e-07]</td>\n",
       "      <td>[5.6351914e-06, 0.9999902, 4.1621734e-06]</td>\n",
       "      <td>[6.1978855e-05, 0.9999176, 2.0371614e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>[0.021710116, 2.5961535e-05, 0.9782639]</td>\n",
       "      <td>[0.00081079284, 0.0007934927, 0.99839574]</td>\n",
       "      <td>[7.8857935e-05, 0.0003978529, 0.9995233]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>[0.00060418097, 0.99912363, 0.0002722138]</td>\n",
       "      <td>[6.452492e-05, 0.99992394, 1.1570731e-05]</td>\n",
       "      <td>[7.210481e-05, 0.9999161, 1.1803859e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>[0.9535876, 0.0369543, 0.009458121]</td>\n",
       "      <td>[0.06544465, 0.4085732, 0.52598214]</td>\n",
       "      <td>[0.00020762562, 0.00060981687, 0.9991825]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>[0.010873717, 0.98802453, 0.0011016902]</td>\n",
       "      <td>[1.33834255e-05, 0.99996865, 1.796761e-05]</td>\n",
       "      <td>[4.5374585e-05, 0.9999213, 3.3305845e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>[1.2437995e-05, 0.99998605, 1.4981332e-06]</td>\n",
       "      <td>[6.814804e-06, 0.99998903, 4.2227966e-06]</td>\n",
       "      <td>[5.612093e-05, 0.99992216, 2.1637394e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>[0.00022437192, 0.40065616, 0.5991194]</td>\n",
       "      <td>[0.0018520205, 0.23281698, 0.76533103]</td>\n",
       "      <td>[0.00013584517, 0.00024865704, 0.99961543]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>[0.42409116, 0.3661773, 0.2097315]</td>\n",
       "      <td>[0.00018075421, 0.00020491573, 0.99961436]</td>\n",
       "      <td>[8.948143e-05, 0.00019811906, 0.9997124]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>[0.00028815673, 0.02189521, 0.97781664]</td>\n",
       "      <td>[0.0020194512, 0.35163933, 0.6463412]</td>\n",
       "      <td>[0.00022410582, 0.013034229, 0.9867417]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>[0.0049450295, 0.99437827, 0.0006767206]</td>\n",
       "      <td>[0.0011020844, 0.9986578, 0.00024008311]</td>\n",
       "      <td>[5.753277e-05, 0.99992466, 1.777614e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>[0.0021360728, 0.9973984, 0.00046561955]</td>\n",
       "      <td>[0.06276215, 0.010787978, 0.9264499]</td>\n",
       "      <td>[5.8154856e-05, 0.00014961355, 0.9997923]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>[0.012432584, 0.9875252, 4.218152e-05]</td>\n",
       "      <td>[6.9983995e-05, 0.9999213, 8.655985e-06]</td>\n",
       "      <td>[7.0063914e-05, 0.9999087, 2.1275844e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>[3.4952702e-06, 0.99999654, 5.4307318e-08]</td>\n",
       "      <td>[3.8394963e-05, 0.9999201, 4.1535608e-05]</td>\n",
       "      <td>[4.8331603e-05, 0.99993205, 1.9714593e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>[0.009834621, 0.031162515, 0.95900285]</td>\n",
       "      <td>[3.8877777e-05, 0.99868363, 0.0012774161]</td>\n",
       "      <td>[5.3785814e-05, 0.00030585722, 0.99964035]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>[2.863268e-05, 0.9999547, 1.6737697e-05]</td>\n",
       "      <td>[9.463277e-06, 0.9999757, 1.4949457e-05]</td>\n",
       "      <td>[5.3389886e-05, 0.99991834, 2.8288976e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>[0.00046800202, 0.99953043, 1.5787145e-06]</td>\n",
       "      <td>[7.325024e-06, 0.9999875, 5.193338e-06]</td>\n",
       "      <td>[4.9465325e-05, 0.9999325, 1.801631e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>[0.0009937334, 0.9990043, 2.0188118e-06]</td>\n",
       "      <td>[0.0066787857, 0.99322104, 0.00010020443]</td>\n",
       "      <td>[5.763131e-05, 0.999928, 1.4367954e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>[0.00067893823, 0.999321, 1.2991575e-07]</td>\n",
       "      <td>[1.5814856e-05, 0.99997985, 4.2840243e-06]</td>\n",
       "      <td>[7.19162e-05, 0.9999155, 1.2658613e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>[0.002937752, 0.9970304, 3.1891446e-05]</td>\n",
       "      <td>[0.000570712, 0.9993119, 0.000117387346]</td>\n",
       "      <td>[5.7950838e-05, 0.9999305, 1.1548351e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>[0.00021700047, 0.12237597, 0.8774071]</td>\n",
       "      <td>[0.00014342695, 0.0009917042, 0.9988649]</td>\n",
       "      <td>[7.705059e-05, 0.0055413716, 0.9943816]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>[0.00041335513, 0.99958247, 4.15565e-06]</td>\n",
       "      <td>[1.3541771e-05, 0.9999796, 6.7871615e-06]</td>\n",
       "      <td>[5.4944056e-05, 0.99992704, 1.7942342e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2970</th>\n",
       "      <td>[0.00013102862, 0.999859, 1.002201e-05]</td>\n",
       "      <td>[8.479956e-06, 0.9999844, 7.094274e-06]</td>\n",
       "      <td>[5.0472172e-05, 0.9999002, 4.9328337e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2971</th>\n",
       "      <td>[0.0063016093, 0.9936958, 2.5687439e-06]</td>\n",
       "      <td>[2.0126467e-05, 0.9999597, 2.0086696e-05]</td>\n",
       "      <td>[6.4939864e-05, 0.9999056, 2.9467843e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2972</th>\n",
       "      <td>[8.079995e-05, 0.9999192, 5.4082644e-08]</td>\n",
       "      <td>[1.0396791e-05, 0.9999838, 5.8043006e-06]</td>\n",
       "      <td>[4.5115954e-05, 0.9999186, 3.6302445e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2973</th>\n",
       "      <td>[0.011751247, 0.9881732, 7.558385e-05]</td>\n",
       "      <td>[1.2453335e-05, 0.9999789, 8.749407e-06]</td>\n",
       "      <td>[4.8256217e-05, 0.99992466, 2.707365e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2974</th>\n",
       "      <td>[0.00568506, 0.9927377, 0.0015771401]</td>\n",
       "      <td>[0.0001500005, 0.9998109, 3.910941e-05]</td>\n",
       "      <td>[5.503107e-05, 0.9999263, 1.8559715e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2975</th>\n",
       "      <td>[0.0013939402, 0.9986022, 3.8363874e-06]</td>\n",
       "      <td>[6.4183373e-06, 0.9999894, 4.162622e-06]</td>\n",
       "      <td>[5.2535448e-05, 0.9999291, 1.8341178e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2976</th>\n",
       "      <td>[6.0102193e-05, 0.00019435708, 0.99974555]</td>\n",
       "      <td>[9.494994e-05, 9.341852e-05, 0.9998117]</td>\n",
       "      <td>[6.442183e-05, 0.00026567586, 0.9996699]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2977</th>\n",
       "      <td>[0.0014636873, 0.99853516, 1.1816319e-06]</td>\n",
       "      <td>[1.3079727e-05, 0.99998057, 6.3716625e-06]</td>\n",
       "      <td>[8.573858e-05, 0.99989474, 1.9587023e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2978</th>\n",
       "      <td>[0.16959986, 0.6882856, 0.14211453]</td>\n",
       "      <td>[0.0028511407, 0.9726916, 0.024457194]</td>\n",
       "      <td>[0.048414215, 0.036332503, 0.91525334]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2979</th>\n",
       "      <td>[0.0017004324, 0.998131, 0.00016857816]</td>\n",
       "      <td>[0.001993633, 0.99798137, 2.5018555e-05]</td>\n",
       "      <td>[0.00042199215, 0.9995654, 1.2613746e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2980</th>\n",
       "      <td>[0.00012538987, 0.9989166, 0.0009579694]</td>\n",
       "      <td>[1.2080088e-05, 0.9999558, 3.2193548e-05]</td>\n",
       "      <td>[4.9621558e-05, 0.9999237, 2.6679521e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2981</th>\n",
       "      <td>[0.009910916, 0.98633945, 0.0037496267]</td>\n",
       "      <td>[7.797504e-06, 0.9999833, 8.987446e-06]</td>\n",
       "      <td>[4.5276945e-05, 0.999912, 4.2622516e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2982</th>\n",
       "      <td>[0.010100571, 0.98953, 0.00036947953]</td>\n",
       "      <td>[2.1385e-05, 0.9999696, 9.035759e-06]</td>\n",
       "      <td>[0.000110848836, 0.9998785, 1.0575883e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2983</th>\n",
       "      <td>[4.693483e-06, 0.99999535, 5.278013e-09]</td>\n",
       "      <td>[1.6061853e-05, 0.99997807, 5.8738624e-06]</td>\n",
       "      <td>[4.7603855e-05, 0.9999274, 2.5006837e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2984</th>\n",
       "      <td>[0.00018049637, 0.9998196, 3.956059e-08]</td>\n",
       "      <td>[1.3465946e-05, 0.99998164, 4.930999e-06]</td>\n",
       "      <td>[5.0970968e-05, 0.9999248, 2.4255407e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2985</th>\n",
       "      <td>[6.223519e-05, 0.99993634, 1.4155079e-06]</td>\n",
       "      <td>[2.096273e-05, 0.9999702, 8.876558e-06]</td>\n",
       "      <td>[6.0911327e-05, 0.99992037, 1.8680157e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2986</th>\n",
       "      <td>[0.15090811, 0.84902453, 6.727478e-05]</td>\n",
       "      <td>[1.6015809e-05, 0.9999778, 6.1886635e-06]</td>\n",
       "      <td>[5.6788038e-05, 0.99992776, 1.5480824e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2987</th>\n",
       "      <td>[5.171855e-05, 0.99994826, 5.354327e-09]</td>\n",
       "      <td>[0.0001350763, 0.99985313, 1.1741929e-05]</td>\n",
       "      <td>[0.00010627753, 0.999884, 9.6272715e-06]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2988</th>\n",
       "      <td>[0.05583445, 0.944164, 1.5643257e-06]</td>\n",
       "      <td>[0.015329741, 0.98452073, 0.00014957698]</td>\n",
       "      <td>[9.4243944e-05, 0.999892, 1.3700564e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2989</th>\n",
       "      <td>[0.7011099, 0.03436372, 0.26452637]</td>\n",
       "      <td>[0.19836606, 0.009145228, 0.7924887]</td>\n",
       "      <td>[0.0008914261, 0.9989819, 0.00012670354]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2990</th>\n",
       "      <td>[0.42243278, 0.57580805, 0.0017591851]</td>\n",
       "      <td>[0.00037372633, 0.9994367, 0.00018960005]</td>\n",
       "      <td>[0.0006359286, 0.9993395, 2.4561226e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2991</th>\n",
       "      <td>[0.12673624, 0.87326056, 3.2519938e-06]</td>\n",
       "      <td>[0.020055577, 0.9798353, 0.000109053195]</td>\n",
       "      <td>[0.026576467, 0.9733123, 0.00011121639]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2992</th>\n",
       "      <td>[0.0004849217, 0.82130843, 0.17820662]</td>\n",
       "      <td>[0.00086266483, 0.16405062, 0.8350867]</td>\n",
       "      <td>[0.00011355382, 0.0038210351, 0.99606544]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2993</th>\n",
       "      <td>[0.43858352, 0.35041612, 0.21100032]</td>\n",
       "      <td>[0.011102119, 0.98604786, 0.002850076]</td>\n",
       "      <td>[0.0002352526, 0.999734, 3.0689174e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2994</th>\n",
       "      <td>[0.007363762, 0.03469444, 0.9579418]</td>\n",
       "      <td>[0.00038859833, 0.0002748932, 0.99933654]</td>\n",
       "      <td>[0.000146678, 0.00027509205, 0.9995782]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2995</th>\n",
       "      <td>[5.5074986e-07, 0.9999994, 3.1386886e-09]</td>\n",
       "      <td>[1.1006394e-05, 0.9999789, 1.0096202e-05]</td>\n",
       "      <td>[5.4615495e-05, 0.99992585, 1.9504161e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2996</th>\n",
       "      <td>[0.00054795184, 0.3214301, 0.67802197]</td>\n",
       "      <td>[9.5943695e-05, 0.002728676, 0.99717534]</td>\n",
       "      <td>[3.364683e-05, 0.0001740074, 0.9997924]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2997</th>\n",
       "      <td>[1.0340856e-05, 0.9999889, 7.5271475e-07]</td>\n",
       "      <td>[0.00015814147, 0.99613684, 0.0037049148]</td>\n",
       "      <td>[0.00016778223, 0.99978524, 4.694916e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2998</th>\n",
       "      <td>[0.02169743, 0.95809954, 0.020203082]</td>\n",
       "      <td>[0.00021270107, 0.9985361, 0.0012511375]</td>\n",
       "      <td>[6.2617466e-05, 0.99991655, 2.0800484e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2999</th>\n",
       "      <td>[3.4449095e-05, 0.99996555, 3.870269e-09]</td>\n",
       "      <td>[6.391212e-05, 0.9999279, 8.209448e-06]</td>\n",
       "      <td>[6.275527e-05, 0.9999261, 1.1155761e-05]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3000 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                    nnlm-128_tsize_18783_epoch30  \\\n",
       "0       [0.00013908364, 0.9998254, 3.547703e-05]   \n",
       "1       [0.002030536, 0.99750996, 0.00045953243]   \n",
       "2       [0.0006725391, 0.99926144, 6.598066e-05]   \n",
       "3       [0.0003392862, 0.9996592, 1.5918495e-06]   \n",
       "4      [4.3006254e-05, 0.99995697, 8.022627e-09]   \n",
       "5            [0.08427183, 0.8487675, 0.06696065]   \n",
       "6          [0.001172376, 0.9711367, 0.027690923]   \n",
       "7     [2.1608248e-06, 0.99978226, 0.00021563252]   \n",
       "8       [0.00057175703, 0.9981173, 0.0013109304]   \n",
       "9      [7.477624e-05, 0.99992514, 1.1235904e-07]   \n",
       "10       [0.021710116, 2.5961535e-05, 0.9782639]   \n",
       "11     [0.00060418097, 0.99912363, 0.0002722138]   \n",
       "12           [0.9535876, 0.0369543, 0.009458121]   \n",
       "13       [0.010873717, 0.98802453, 0.0011016902]   \n",
       "14    [1.2437995e-05, 0.99998605, 1.4981332e-06]   \n",
       "15        [0.00022437192, 0.40065616, 0.5991194]   \n",
       "16            [0.42409116, 0.3661773, 0.2097315]   \n",
       "17       [0.00028815673, 0.02189521, 0.97781664]   \n",
       "18      [0.0049450295, 0.99437827, 0.0006767206]   \n",
       "19      [0.0021360728, 0.9973984, 0.00046561955]   \n",
       "20        [0.012432584, 0.9875252, 4.218152e-05]   \n",
       "21    [3.4952702e-06, 0.99999654, 5.4307318e-08]   \n",
       "22        [0.009834621, 0.031162515, 0.95900285]   \n",
       "23      [2.863268e-05, 0.9999547, 1.6737697e-05]   \n",
       "24    [0.00046800202, 0.99953043, 1.5787145e-06]   \n",
       "25      [0.0009937334, 0.9990043, 2.0188118e-06]   \n",
       "26      [0.00067893823, 0.999321, 1.2991575e-07]   \n",
       "27       [0.002937752, 0.9970304, 3.1891446e-05]   \n",
       "28        [0.00021700047, 0.12237597, 0.8774071]   \n",
       "29      [0.00041335513, 0.99958247, 4.15565e-06]   \n",
       "...                                          ...   \n",
       "2970     [0.00013102862, 0.999859, 1.002201e-05]   \n",
       "2971    [0.0063016093, 0.9936958, 2.5687439e-06]   \n",
       "2972    [8.079995e-05, 0.9999192, 5.4082644e-08]   \n",
       "2973      [0.011751247, 0.9881732, 7.558385e-05]   \n",
       "2974       [0.00568506, 0.9927377, 0.0015771401]   \n",
       "2975    [0.0013939402, 0.9986022, 3.8363874e-06]   \n",
       "2976  [6.0102193e-05, 0.00019435708, 0.99974555]   \n",
       "2977   [0.0014636873, 0.99853516, 1.1816319e-06]   \n",
       "2978         [0.16959986, 0.6882856, 0.14211453]   \n",
       "2979     [0.0017004324, 0.998131, 0.00016857816]   \n",
       "2980    [0.00012538987, 0.9989166, 0.0009579694]   \n",
       "2981     [0.009910916, 0.98633945, 0.0037496267]   \n",
       "2982       [0.010100571, 0.98953, 0.00036947953]   \n",
       "2983    [4.693483e-06, 0.99999535, 5.278013e-09]   \n",
       "2984    [0.00018049637, 0.9998196, 3.956059e-08]   \n",
       "2985   [6.223519e-05, 0.99993634, 1.4155079e-06]   \n",
       "2986      [0.15090811, 0.84902453, 6.727478e-05]   \n",
       "2987    [5.171855e-05, 0.99994826, 5.354327e-09]   \n",
       "2988       [0.05583445, 0.944164, 1.5643257e-06]   \n",
       "2989         [0.7011099, 0.03436372, 0.26452637]   \n",
       "2990      [0.42243278, 0.57580805, 0.0017591851]   \n",
       "2991     [0.12673624, 0.87326056, 3.2519938e-06]   \n",
       "2992      [0.0004849217, 0.82130843, 0.17820662]   \n",
       "2993        [0.43858352, 0.35041612, 0.21100032]   \n",
       "2994        [0.007363762, 0.03469444, 0.9579418]   \n",
       "2995   [5.5074986e-07, 0.9999994, 3.1386886e-09]   \n",
       "2996      [0.00054795184, 0.3214301, 0.67802197]   \n",
       "2997   [1.0340856e-05, 0.9999889, 7.5271475e-07]   \n",
       "2998       [0.02169743, 0.95809954, 0.020203082]   \n",
       "2999   [3.4449095e-05, 0.99996555, 3.870269e-09]   \n",
       "\n",
       "                         DAN_tsize_18783_epoch30  \\\n",
       "0     [2.5715732e-05, 0.9998716, 0.000102616505]   \n",
       "1        [2.547616e-05, 0.9989743, 0.0010002507]   \n",
       "2        [1.7972814e-05, 0.999974, 8.038951e-06]   \n",
       "3      [1.8328115e-05, 0.9999672, 1.4433712e-05]   \n",
       "4      [1.7900435e-05, 0.9999765, 5.5927703e-06]   \n",
       "5          [0.0026555285, 0.06760905, 0.9297355]   \n",
       "6      [2.252558e-05, 0.99995816, 1.9321673e-05]   \n",
       "7         [0.00030092022, 0.9585573, 0.04114188]   \n",
       "8      [1.6481239e-05, 0.99988806, 9.542799e-05]   \n",
       "9      [5.6351914e-06, 0.9999902, 4.1621734e-06]   \n",
       "10     [0.00081079284, 0.0007934927, 0.99839574]   \n",
       "11     [6.452492e-05, 0.99992394, 1.1570731e-05]   \n",
       "12           [0.06544465, 0.4085732, 0.52598214]   \n",
       "13    [1.33834255e-05, 0.99996865, 1.796761e-05]   \n",
       "14     [6.814804e-06, 0.99998903, 4.2227966e-06]   \n",
       "15        [0.0018520205, 0.23281698, 0.76533103]   \n",
       "16    [0.00018075421, 0.00020491573, 0.99961436]   \n",
       "17         [0.0020194512, 0.35163933, 0.6463412]   \n",
       "18      [0.0011020844, 0.9986578, 0.00024008311]   \n",
       "19          [0.06276215, 0.010787978, 0.9264499]   \n",
       "20      [6.9983995e-05, 0.9999213, 8.655985e-06]   \n",
       "21     [3.8394963e-05, 0.9999201, 4.1535608e-05]   \n",
       "22     [3.8877777e-05, 0.99868363, 0.0012774161]   \n",
       "23      [9.463277e-06, 0.9999757, 1.4949457e-05]   \n",
       "24       [7.325024e-06, 0.9999875, 5.193338e-06]   \n",
       "25     [0.0066787857, 0.99322104, 0.00010020443]   \n",
       "26    [1.5814856e-05, 0.99997985, 4.2840243e-06]   \n",
       "27      [0.000570712, 0.9993119, 0.000117387346]   \n",
       "28      [0.00014342695, 0.0009917042, 0.9988649]   \n",
       "29     [1.3541771e-05, 0.9999796, 6.7871615e-06]   \n",
       "...                                          ...   \n",
       "2970     [8.479956e-06, 0.9999844, 7.094274e-06]   \n",
       "2971   [2.0126467e-05, 0.9999597, 2.0086696e-05]   \n",
       "2972   [1.0396791e-05, 0.9999838, 5.8043006e-06]   \n",
       "2973    [1.2453335e-05, 0.9999789, 8.749407e-06]   \n",
       "2974     [0.0001500005, 0.9998109, 3.910941e-05]   \n",
       "2975    [6.4183373e-06, 0.9999894, 4.162622e-06]   \n",
       "2976     [9.494994e-05, 9.341852e-05, 0.9998117]   \n",
       "2977  [1.3079727e-05, 0.99998057, 6.3716625e-06]   \n",
       "2978      [0.0028511407, 0.9726916, 0.024457194]   \n",
       "2979    [0.001993633, 0.99798137, 2.5018555e-05]   \n",
       "2980   [1.2080088e-05, 0.9999558, 3.2193548e-05]   \n",
       "2981     [7.797504e-06, 0.9999833, 8.987446e-06]   \n",
       "2982       [2.1385e-05, 0.9999696, 9.035759e-06]   \n",
       "2983  [1.6061853e-05, 0.99997807, 5.8738624e-06]   \n",
       "2984   [1.3465946e-05, 0.99998164, 4.930999e-06]   \n",
       "2985     [2.096273e-05, 0.9999702, 8.876558e-06]   \n",
       "2986   [1.6015809e-05, 0.9999778, 6.1886635e-06]   \n",
       "2987   [0.0001350763, 0.99985313, 1.1741929e-05]   \n",
       "2988    [0.015329741, 0.98452073, 0.00014957698]   \n",
       "2989        [0.19836606, 0.009145228, 0.7924887]   \n",
       "2990   [0.00037372633, 0.9994367, 0.00018960005]   \n",
       "2991    [0.020055577, 0.9798353, 0.000109053195]   \n",
       "2992      [0.00086266483, 0.16405062, 0.8350867]   \n",
       "2993      [0.011102119, 0.98604786, 0.002850076]   \n",
       "2994   [0.00038859833, 0.0002748932, 0.99933654]   \n",
       "2995   [1.1006394e-05, 0.9999789, 1.0096202e-05]   \n",
       "2996    [9.5943695e-05, 0.002728676, 0.99717534]   \n",
       "2997   [0.00015814147, 0.99613684, 0.0037049148]   \n",
       "2998    [0.00021270107, 0.9985361, 0.0012511375]   \n",
       "2999     [6.391212e-05, 0.9999279, 8.209448e-06]   \n",
       "\n",
       "                       univ3_tsize_18783_epoch30  \n",
       "0         [7.91436e-05, 0.9998338, 8.701929e-05]  \n",
       "1         [7.870038e-05, 0.9997701, 0.000151296]  \n",
       "2       [6.0237904e-05, 0.9999143, 2.548163e-05]  \n",
       "3       [4.4898494e-05, 0.9999287, 2.633073e-05]  \n",
       "4        [7.507247e-05, 0.9999107, 1.421926e-05]  \n",
       "5     [0.00016169962, 0.99982136, 1.6898892e-05]  \n",
       "6      [5.2135645e-05, 0.99991536, 3.253769e-05]  \n",
       "7       [6.164825e-05, 0.9998791, 5.9188635e-05]  \n",
       "8      [6.790512e-05, 0.99990237, 2.9676981e-05]  \n",
       "9      [6.1978855e-05, 0.9999176, 2.0371614e-05]  \n",
       "10      [7.8857935e-05, 0.0003978529, 0.9995233]  \n",
       "11      [7.210481e-05, 0.9999161, 1.1803859e-05]  \n",
       "12     [0.00020762562, 0.00060981687, 0.9991825]  \n",
       "13     [4.5374585e-05, 0.9999213, 3.3305845e-05]  \n",
       "14     [5.612093e-05, 0.99992216, 2.1637394e-05]  \n",
       "15    [0.00013584517, 0.00024865704, 0.99961543]  \n",
       "16      [8.948143e-05, 0.00019811906, 0.9997124]  \n",
       "17       [0.00022410582, 0.013034229, 0.9867417]  \n",
       "18      [5.753277e-05, 0.99992466, 1.777614e-05]  \n",
       "19     [5.8154856e-05, 0.00014961355, 0.9997923]  \n",
       "20     [7.0063914e-05, 0.9999087, 2.1275844e-05]  \n",
       "21    [4.8331603e-05, 0.99993205, 1.9714593e-05]  \n",
       "22    [5.3785814e-05, 0.00030585722, 0.99964035]  \n",
       "23    [5.3389886e-05, 0.99991834, 2.8288976e-05]  \n",
       "24      [4.9465325e-05, 0.9999325, 1.801631e-05]  \n",
       "25       [5.763131e-05, 0.999928, 1.4367954e-05]  \n",
       "26       [7.19162e-05, 0.9999155, 1.2658613e-05]  \n",
       "27     [5.7950838e-05, 0.9999305, 1.1548351e-05]  \n",
       "28       [7.705059e-05, 0.0055413716, 0.9943816]  \n",
       "29    [5.4944056e-05, 0.99992704, 1.7942342e-05]  \n",
       "...                                          ...  \n",
       "2970   [5.0472172e-05, 0.9999002, 4.9328337e-05]  \n",
       "2971   [6.4939864e-05, 0.9999056, 2.9467843e-05]  \n",
       "2972   [4.5115954e-05, 0.9999186, 3.6302445e-05]  \n",
       "2973   [4.8256217e-05, 0.99992466, 2.707365e-05]  \n",
       "2974    [5.503107e-05, 0.9999263, 1.8559715e-05]  \n",
       "2975   [5.2535448e-05, 0.9999291, 1.8341178e-05]  \n",
       "2976    [6.442183e-05, 0.00026567586, 0.9996699]  \n",
       "2977   [8.573858e-05, 0.99989474, 1.9587023e-05]  \n",
       "2978      [0.048414215, 0.036332503, 0.91525334]  \n",
       "2979   [0.00042199215, 0.9995654, 1.2613746e-05]  \n",
       "2980   [4.9621558e-05, 0.9999237, 2.6679521e-05]  \n",
       "2981    [4.5276945e-05, 0.999912, 4.2622516e-05]  \n",
       "2982  [0.000110848836, 0.9998785, 1.0575883e-05]  \n",
       "2983   [4.7603855e-05, 0.9999274, 2.5006837e-05]  \n",
       "2984   [5.0970968e-05, 0.9999248, 2.4255407e-05]  \n",
       "2985  [6.0911327e-05, 0.99992037, 1.8680157e-05]  \n",
       "2986  [5.6788038e-05, 0.99992776, 1.5480824e-05]  \n",
       "2987    [0.00010627753, 0.999884, 9.6272715e-06]  \n",
       "2988    [9.4243944e-05, 0.999892, 1.3700564e-05]  \n",
       "2989    [0.0008914261, 0.9989819, 0.00012670354]  \n",
       "2990    [0.0006359286, 0.9993395, 2.4561226e-05]  \n",
       "2991     [0.026576467, 0.9733123, 0.00011121639]  \n",
       "2992   [0.00011355382, 0.0038210351, 0.99606544]  \n",
       "2993     [0.0002352526, 0.999734, 3.0689174e-05]  \n",
       "2994     [0.000146678, 0.00027509205, 0.9995782]  \n",
       "2995  [5.4615495e-05, 0.99992585, 1.9504161e-05]  \n",
       "2996     [3.364683e-05, 0.0001740074, 0.9997924]  \n",
       "2997   [0.00016778223, 0.99978524, 4.694916e-05]  \n",
       "2998  [6.2617466e-05, 0.99991655, 2.0800484e-05]  \n",
       "2999    [6.275527e-05, 0.9999261, 1.1155761e-05]  \n",
       "\n",
       "[3000 rows x 3 columns]"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "probs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Cosine distance of 2 models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 129,
   "metadata": {},
   "outputs": [],
   "source": [
    "def euclidean_distance(x, y):   \n",
    "    return np.sqrt(np.sum((x - y) ** 2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 133,
   "metadata": {},
   "outputs": [],
   "source": [
    "def cosine_similarity(x, y):\n",
    "    return np.dot(x, y) / (np.sqrt(np.dot(x, x)) * np.sqrt(np.dot(y, y)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "metadata": {},
   "outputs": [],
   "source": [
    "# or different definition. likely requires aa = a.reshape(1,3)\n",
    "# from sklearn.metrics.pairwise import cosine_similarity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#import scipy.spatial.distance as SSD"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "nnlm-128_tsize_18783_epoch30      [0.00013908364, 0.9998254, 3.547703e-05]\n",
       "DAN_tsize_18783_epoch30         [2.5715732e-05, 0.9998716, 0.000102616505]\n",
       "univ3_tsize_18783_epoch30           [7.91436e-05, 0.9998338, 8.701929e-05]\n",
       "Name: 0, dtype: object"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "probs.iloc[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.0001 0.9998 0.0000] [0.0000 0.9999 0.0001]\n"
     ]
    }
   ],
   "source": [
    "test_a = probs.iloc[0][0]\n",
    "test_b = probs.iloc[0][1]\n",
    "print(test_a, test_b)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.15028952"
      ]
     },
     "execution_count": 134,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cosine_similarity(test_a, test_b) # cosine of 0 is 1."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1.1667262"
      ]
     },
     "execution_count": 136,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "euclidean_distance(test_a, test_b)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#### too near each other, cos sim is 1 differenc eat 3-4th decimal only"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.0843 0.8488 0.0670] [0.0027 0.0676 0.9297]\n"
     ]
    }
   ],
   "source": [
    "test_a = probs.iloc[5][0]\n",
    "test_b = probs.iloc[5][1]\n",
    "print(test_a, test_b)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.15028952"
      ]
     },
     "execution_count": 94,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cosine_similarity(test_a, test_b)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 130,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['nnlm-128_tsize_18783_epoch30', 'DAN_tsize_18783_epoch30',\n",
       "       'univ3_tsize_18783_epoch30'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 130,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# save the names and change easier to assign\n",
    "#savenames = probs.columns\n",
    "savenames"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "probs.columns = ['a','b','c']\n",
    "probs.columns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0    1.000000\n",
       "1    0.999998\n",
       "2    1.000000\n",
       "3    1.000000\n",
       "4    1.000000\n",
       "5    0.150290\n",
       "6    0.999594\n",
       "7    0.999089\n",
       "dtype: float64"
      ]
     },
     "execution_count": 123,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cosines = probs[['a','b']].apply(lambda row: cosine_similarity(row['a'],row['b']), axis=1)\n",
    "cosines.head(8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9155512768672391"
      ]
     },
     "execution_count": 126,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cosines.mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 127,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.917959241147061"
      ]
     },
     "execution_count": 127,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cosines = probs[['a','c']].apply(lambda row: cosine_similarity(row['a'],row['c']), axis=1)\n",
    "cosines.mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 128,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9316679687557334"
      ]
     },
     "execution_count": 128,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cosines = probs[['b','c']].apply(lambda row: cosine_similarity(row['b'],row['c']), axis=1)\n",
    "cosines.mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Correlation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9999981"
      ]
     },
     "execution_count": 75,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cosine_similarity(test_a, test_c)"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "plt.matshow(lab.corr())\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<style  type=\"text/css\" >\n",
       "    #T_53b9b266_5cf6_11e9_8604_b9b39a98459arow0_col0 {\n",
       "            background-color:  #b40426;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_53b9b266_5cf6_11e9_8604_b9b39a98459arow0_col1 {\n",
       "            background-color:  #3b4cc0;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_53b9b266_5cf6_11e9_8604_b9b39a98459arow0_col2 {\n",
       "            background-color:  #3b4cc0;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_53b9b266_5cf6_11e9_8604_b9b39a98459arow1_col0 {\n",
       "            background-color:  #3b4cc0;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_53b9b266_5cf6_11e9_8604_b9b39a98459arow1_col1 {\n",
       "            background-color:  #b40426;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_53b9b266_5cf6_11e9_8604_b9b39a98459arow1_col2 {\n",
       "            background-color:  #82a6fb;\n",
       "            color:  #000000;\n",
       "        }    #T_53b9b266_5cf6_11e9_8604_b9b39a98459arow2_col0 {\n",
       "            background-color:  #4b64d5;\n",
       "            color:  #f1f1f1;\n",
       "        }    #T_53b9b266_5cf6_11e9_8604_b9b39a98459arow2_col1 {\n",
       "            background-color:  #92b4fe;\n",
       "            color:  #000000;\n",
       "        }    #T_53b9b266_5cf6_11e9_8604_b9b39a98459arow2_col2 {\n",
       "            background-color:  #b40426;\n",
       "            color:  #f1f1f1;\n",
       "        }</style><table id=\"T_53b9b266_5cf6_11e9_8604_b9b39a98459a\" ><thead>    <tr>        <th class=\"blank level0\" ></th>        <th class=\"col_heading level0 col0\" >nnlm-128_tsize_18783_epoch30</th>        <th class=\"col_heading level0 col1\" >DAN_tsize_18783_epoch30</th>        <th class=\"col_heading level0 col2\" >univ3_tsize_18783_epoch30</th>    </tr></thead><tbody>\n",
       "                <tr>\n",
       "                        <th id=\"T_53b9b266_5cf6_11e9_8604_b9b39a98459alevel0_row0\" class=\"row_heading level0 row0\" >nnlm-128_tsize_18783_epoch30</th>\n",
       "                        <td id=\"T_53b9b266_5cf6_11e9_8604_b9b39a98459arow0_col0\" class=\"data row0 col0\" >1</td>\n",
       "                        <td id=\"T_53b9b266_5cf6_11e9_8604_b9b39a98459arow0_col1\" class=\"data row0 col1\" >0.67</td>\n",
       "                        <td id=\"T_53b9b266_5cf6_11e9_8604_b9b39a98459arow0_col2\" class=\"data row0 col2\" >0.68</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_53b9b266_5cf6_11e9_8604_b9b39a98459alevel0_row1\" class=\"row_heading level0 row1\" >DAN_tsize_18783_epoch30</th>\n",
       "                        <td id=\"T_53b9b266_5cf6_11e9_8604_b9b39a98459arow1_col0\" class=\"data row1 col0\" >0.67</td>\n",
       "                        <td id=\"T_53b9b266_5cf6_11e9_8604_b9b39a98459arow1_col1\" class=\"data row1 col1\" >1</td>\n",
       "                        <td id=\"T_53b9b266_5cf6_11e9_8604_b9b39a98459arow1_col2\" class=\"data row1 col2\" >0.75</td>\n",
       "            </tr>\n",
       "            <tr>\n",
       "                        <th id=\"T_53b9b266_5cf6_11e9_8604_b9b39a98459alevel0_row2\" class=\"row_heading level0 row2\" >univ3_tsize_18783_epoch30</th>\n",
       "                        <td id=\"T_53b9b266_5cf6_11e9_8604_b9b39a98459arow2_col0\" class=\"data row2 col0\" >0.68</td>\n",
       "                        <td id=\"T_53b9b266_5cf6_11e9_8604_b9b39a98459arow2_col1\" class=\"data row2 col1\" >0.75</td>\n",
       "                        <td id=\"T_53b9b266_5cf6_11e9_8604_b9b39a98459arow2_col2\" class=\"data row2 col2\" >1</td>\n",
       "            </tr>\n",
       "    </tbody></table>"
      ],
      "text/plain": [
       "<pandas.io.formats.style.Styler at 0x7f519a64c080>"
      ]
     },
     "execution_count": 111,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "corr = pick.corr()\n",
    "corr.style.background_gradient(cmap='coolwarm').set_precision(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 106,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_corr(df,size=10):\n",
    "    '''Function plots a graphical correlation matrix for each pair of columns in the dataframe.\n",
    "\n",
    "    Input:\n",
    "        df: pandas DataFrame\n",
    "        size: vertical and horizontal size of the plot'''\n",
    "\n",
    "    corr = df.corr()\n",
    "    fig, ax = plt.subplots(figsize=(size, size))\n",
    "    ax.matshow(corr)\n",
    "    plt.xticks(range(len(corr.columns)), corr.columns);\n",
    "    plt.yticks(range(len(corr.columns)), corr.columns);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_corr(pick)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7fe2459652e8>"
      ]
     },
     "execution_count": 97,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1008x864 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "pick = labels\n",
    "import seaborn as sns\n",
    "\n",
    "f, ax = plt.subplots(figsize=(14, 12))\n",
    "corr = pick.corr()\n",
    "sns.heatmap(corr, mask=np.zeros_like(corr, dtype=np.bool), cmap=sns.diverging_palette(220, 10, as_cmap=True),\n",
    "            square=True, ax=ax)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Compare probabilities"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "metadata": {},
   "outputs": [],
   "source": [
    "def cross_entropy(predictions, targets, epsilon=1e-12):\n",
    "    \"\"\"\n",
    "    Computes cross entropy between targets (encoded as one-hot vectors)\n",
    "    and predictions. \n",
    "    Input: predictions (N, k) ndarray\n",
    "           targets (N, k) ndarray        \n",
    "    Returns: scalar\n",
    "    \"\"\"\n",
    "    predictions = np.clip(predictions, epsilon, 1. - epsilon)\n",
    "    ce = - np.mean(np.log(predictions) * targets) \n",
    "    return ce"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "False\n"
     ]
    }
   ],
   "source": [
    "predictions = np.array([[0.25,0.25,0.25,0.25],\n",
    "                        [0.01,0.01,0.01,0.96]])\n",
    "targets = np.array([[0,0,0,1],\n",
    "                  [0,0,0,1]])\n",
    "ans = 0.71355817782  #Correct answer\n",
    "x = cross_entropy(predictions, targets)\n",
    "print(np.isclose(x,ans))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "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>nnlm-128_tsize_18783_epoch30</th>\n",
       "      <th>DAN_tsize_18783_epoch30</th>\n",
       "      <th>univ3_tsize_18783_epoch30</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>[0.00013908364, 0.9998254, 3.547703e-05]</td>\n",
       "      <td>[2.5715732e-05, 0.9998716, 0.000102616505]</td>\n",
       "      <td>[7.91436e-05, 0.9998338, 8.701929e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>[0.002030536, 0.99750996, 0.00045953243]</td>\n",
       "      <td>[2.547616e-05, 0.9989743, 0.0010002507]</td>\n",
       "      <td>[7.870038e-05, 0.9997701, 0.000151296]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>[0.0006725391, 0.99926144, 6.598066e-05]</td>\n",
       "      <td>[1.7972814e-05, 0.999974, 8.038951e-06]</td>\n",
       "      <td>[6.0237904e-05, 0.9999143, 2.548163e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>[0.0003392862, 0.9996592, 1.5918495e-06]</td>\n",
       "      <td>[1.8328115e-05, 0.9999672, 1.4433712e-05]</td>\n",
       "      <td>[4.4898494e-05, 0.9999287, 2.633073e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>[4.3006254e-05, 0.99995697, 8.022627e-09]</td>\n",
       "      <td>[1.7900435e-05, 0.9999765, 5.5927703e-06]</td>\n",
       "      <td>[7.507247e-05, 0.9999107, 1.421926e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>[0.08427183, 0.8487675, 0.06696065]</td>\n",
       "      <td>[0.0026555285, 0.06760905, 0.9297355]</td>\n",
       "      <td>[0.00016169962, 0.99982136, 1.6898892e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>[0.001172376, 0.9711367, 0.027690923]</td>\n",
       "      <td>[2.252558e-05, 0.99995816, 1.9321673e-05]</td>\n",
       "      <td>[5.2135645e-05, 0.99991536, 3.253769e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>[2.1608248e-06, 0.99978226, 0.00021563252]</td>\n",
       "      <td>[0.00030092022, 0.9585573, 0.04114188]</td>\n",
       "      <td>[6.164825e-05, 0.9998791, 5.9188635e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>[0.00057175703, 0.9981173, 0.0013109304]</td>\n",
       "      <td>[1.6481239e-05, 0.99988806, 9.542799e-05]</td>\n",
       "      <td>[6.790512e-05, 0.99990237, 2.9676981e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>[7.477624e-05, 0.99992514, 1.1235904e-07]</td>\n",
       "      <td>[5.6351914e-06, 0.9999902, 4.1621734e-06]</td>\n",
       "      <td>[6.1978855e-05, 0.9999176, 2.0371614e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>[0.021710116, 2.5961535e-05, 0.9782639]</td>\n",
       "      <td>[0.00081079284, 0.0007934927, 0.99839574]</td>\n",
       "      <td>[7.8857935e-05, 0.0003978529, 0.9995233]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>[0.00060418097, 0.99912363, 0.0002722138]</td>\n",
       "      <td>[6.452492e-05, 0.99992394, 1.1570731e-05]</td>\n",
       "      <td>[7.210481e-05, 0.9999161, 1.1803859e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>[0.9535876, 0.0369543, 0.009458121]</td>\n",
       "      <td>[0.06544465, 0.4085732, 0.52598214]</td>\n",
       "      <td>[0.00020762562, 0.00060981687, 0.9991825]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>[0.010873717, 0.98802453, 0.0011016902]</td>\n",
       "      <td>[1.33834255e-05, 0.99996865, 1.796761e-05]</td>\n",
       "      <td>[4.5374585e-05, 0.9999213, 3.3305845e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>[1.2437995e-05, 0.99998605, 1.4981332e-06]</td>\n",
       "      <td>[6.814804e-06, 0.99998903, 4.2227966e-06]</td>\n",
       "      <td>[5.612093e-05, 0.99992216, 2.1637394e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>[0.00022437192, 0.40065616, 0.5991194]</td>\n",
       "      <td>[0.0018520205, 0.23281698, 0.76533103]</td>\n",
       "      <td>[0.00013584517, 0.00024865704, 0.99961543]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>[0.42409116, 0.3661773, 0.2097315]</td>\n",
       "      <td>[0.00018075421, 0.00020491573, 0.99961436]</td>\n",
       "      <td>[8.948143e-05, 0.00019811906, 0.9997124]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>[0.00028815673, 0.02189521, 0.97781664]</td>\n",
       "      <td>[0.0020194512, 0.35163933, 0.6463412]</td>\n",
       "      <td>[0.00022410582, 0.013034229, 0.9867417]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>[0.0049450295, 0.99437827, 0.0006767206]</td>\n",
       "      <td>[0.0011020844, 0.9986578, 0.00024008311]</td>\n",
       "      <td>[5.753277e-05, 0.99992466, 1.777614e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>[0.0021360728, 0.9973984, 0.00046561955]</td>\n",
       "      <td>[0.06276215, 0.010787978, 0.9264499]</td>\n",
       "      <td>[5.8154856e-05, 0.00014961355, 0.9997923]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>[0.012432584, 0.9875252, 4.218152e-05]</td>\n",
       "      <td>[6.9983995e-05, 0.9999213, 8.655985e-06]</td>\n",
       "      <td>[7.0063914e-05, 0.9999087, 2.1275844e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>[3.4952702e-06, 0.99999654, 5.4307318e-08]</td>\n",
       "      <td>[3.8394963e-05, 0.9999201, 4.1535608e-05]</td>\n",
       "      <td>[4.8331603e-05, 0.99993205, 1.9714593e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>[0.009834621, 0.031162515, 0.95900285]</td>\n",
       "      <td>[3.8877777e-05, 0.99868363, 0.0012774161]</td>\n",
       "      <td>[5.3785814e-05, 0.00030585722, 0.99964035]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>[2.863268e-05, 0.9999547, 1.6737697e-05]</td>\n",
       "      <td>[9.463277e-06, 0.9999757, 1.4949457e-05]</td>\n",
       "      <td>[5.3389886e-05, 0.99991834, 2.8288976e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>[0.00046800202, 0.99953043, 1.5787145e-06]</td>\n",
       "      <td>[7.325024e-06, 0.9999875, 5.193338e-06]</td>\n",
       "      <td>[4.9465325e-05, 0.9999325, 1.801631e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>[0.0009937334, 0.9990043, 2.0188118e-06]</td>\n",
       "      <td>[0.0066787857, 0.99322104, 0.00010020443]</td>\n",
       "      <td>[5.763131e-05, 0.999928, 1.4367954e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>[0.00067893823, 0.999321, 1.2991575e-07]</td>\n",
       "      <td>[1.5814856e-05, 0.99997985, 4.2840243e-06]</td>\n",
       "      <td>[7.19162e-05, 0.9999155, 1.2658613e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>[0.002937752, 0.9970304, 3.1891446e-05]</td>\n",
       "      <td>[0.000570712, 0.9993119, 0.000117387346]</td>\n",
       "      <td>[5.7950838e-05, 0.9999305, 1.1548351e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>[0.00021700047, 0.12237597, 0.8774071]</td>\n",
       "      <td>[0.00014342695, 0.0009917042, 0.9988649]</td>\n",
       "      <td>[7.705059e-05, 0.0055413716, 0.9943816]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>[0.00041335513, 0.99958247, 4.15565e-06]</td>\n",
       "      <td>[1.3541771e-05, 0.9999796, 6.7871615e-06]</td>\n",
       "      <td>[5.4944056e-05, 0.99992704, 1.7942342e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2970</th>\n",
       "      <td>[0.00013102862, 0.999859, 1.002201e-05]</td>\n",
       "      <td>[8.479956e-06, 0.9999844, 7.094274e-06]</td>\n",
       "      <td>[5.0472172e-05, 0.9999002, 4.9328337e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2971</th>\n",
       "      <td>[0.0063016093, 0.9936958, 2.5687439e-06]</td>\n",
       "      <td>[2.0126467e-05, 0.9999597, 2.0086696e-05]</td>\n",
       "      <td>[6.4939864e-05, 0.9999056, 2.9467843e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2972</th>\n",
       "      <td>[8.079995e-05, 0.9999192, 5.4082644e-08]</td>\n",
       "      <td>[1.0396791e-05, 0.9999838, 5.8043006e-06]</td>\n",
       "      <td>[4.5115954e-05, 0.9999186, 3.6302445e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2973</th>\n",
       "      <td>[0.011751247, 0.9881732, 7.558385e-05]</td>\n",
       "      <td>[1.2453335e-05, 0.9999789, 8.749407e-06]</td>\n",
       "      <td>[4.8256217e-05, 0.99992466, 2.707365e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2974</th>\n",
       "      <td>[0.00568506, 0.9927377, 0.0015771401]</td>\n",
       "      <td>[0.0001500005, 0.9998109, 3.910941e-05]</td>\n",
       "      <td>[5.503107e-05, 0.9999263, 1.8559715e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2975</th>\n",
       "      <td>[0.0013939402, 0.9986022, 3.8363874e-06]</td>\n",
       "      <td>[6.4183373e-06, 0.9999894, 4.162622e-06]</td>\n",
       "      <td>[5.2535448e-05, 0.9999291, 1.8341178e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2976</th>\n",
       "      <td>[6.0102193e-05, 0.00019435708, 0.99974555]</td>\n",
       "      <td>[9.494994e-05, 9.341852e-05, 0.9998117]</td>\n",
       "      <td>[6.442183e-05, 0.00026567586, 0.9996699]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2977</th>\n",
       "      <td>[0.0014636873, 0.99853516, 1.1816319e-06]</td>\n",
       "      <td>[1.3079727e-05, 0.99998057, 6.3716625e-06]</td>\n",
       "      <td>[8.573858e-05, 0.99989474, 1.9587023e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2978</th>\n",
       "      <td>[0.16959986, 0.6882856, 0.14211453]</td>\n",
       "      <td>[0.0028511407, 0.9726916, 0.024457194]</td>\n",
       "      <td>[0.048414215, 0.036332503, 0.91525334]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2979</th>\n",
       "      <td>[0.0017004324, 0.998131, 0.00016857816]</td>\n",
       "      <td>[0.001993633, 0.99798137, 2.5018555e-05]</td>\n",
       "      <td>[0.00042199215, 0.9995654, 1.2613746e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2980</th>\n",
       "      <td>[0.00012538987, 0.9989166, 0.0009579694]</td>\n",
       "      <td>[1.2080088e-05, 0.9999558, 3.2193548e-05]</td>\n",
       "      <td>[4.9621558e-05, 0.9999237, 2.6679521e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2981</th>\n",
       "      <td>[0.009910916, 0.98633945, 0.0037496267]</td>\n",
       "      <td>[7.797504e-06, 0.9999833, 8.987446e-06]</td>\n",
       "      <td>[4.5276945e-05, 0.999912, 4.2622516e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2982</th>\n",
       "      <td>[0.010100571, 0.98953, 0.00036947953]</td>\n",
       "      <td>[2.1385e-05, 0.9999696, 9.035759e-06]</td>\n",
       "      <td>[0.000110848836, 0.9998785, 1.0575883e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2983</th>\n",
       "      <td>[4.693483e-06, 0.99999535, 5.278013e-09]</td>\n",
       "      <td>[1.6061853e-05, 0.99997807, 5.8738624e-06]</td>\n",
       "      <td>[4.7603855e-05, 0.9999274, 2.5006837e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2984</th>\n",
       "      <td>[0.00018049637, 0.9998196, 3.956059e-08]</td>\n",
       "      <td>[1.3465946e-05, 0.99998164, 4.930999e-06]</td>\n",
       "      <td>[5.0970968e-05, 0.9999248, 2.4255407e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2985</th>\n",
       "      <td>[6.223519e-05, 0.99993634, 1.4155079e-06]</td>\n",
       "      <td>[2.096273e-05, 0.9999702, 8.876558e-06]</td>\n",
       "      <td>[6.0911327e-05, 0.99992037, 1.8680157e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2986</th>\n",
       "      <td>[0.15090811, 0.84902453, 6.727478e-05]</td>\n",
       "      <td>[1.6015809e-05, 0.9999778, 6.1886635e-06]</td>\n",
       "      <td>[5.6788038e-05, 0.99992776, 1.5480824e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2987</th>\n",
       "      <td>[5.171855e-05, 0.99994826, 5.354327e-09]</td>\n",
       "      <td>[0.0001350763, 0.99985313, 1.1741929e-05]</td>\n",
       "      <td>[0.00010627753, 0.999884, 9.6272715e-06]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2988</th>\n",
       "      <td>[0.05583445, 0.944164, 1.5643257e-06]</td>\n",
       "      <td>[0.015329741, 0.98452073, 0.00014957698]</td>\n",
       "      <td>[9.4243944e-05, 0.999892, 1.3700564e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2989</th>\n",
       "      <td>[0.7011099, 0.03436372, 0.26452637]</td>\n",
       "      <td>[0.19836606, 0.009145228, 0.7924887]</td>\n",
       "      <td>[0.0008914261, 0.9989819, 0.00012670354]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2990</th>\n",
       "      <td>[0.42243278, 0.57580805, 0.0017591851]</td>\n",
       "      <td>[0.00037372633, 0.9994367, 0.00018960005]</td>\n",
       "      <td>[0.0006359286, 0.9993395, 2.4561226e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2991</th>\n",
       "      <td>[0.12673624, 0.87326056, 3.2519938e-06]</td>\n",
       "      <td>[0.020055577, 0.9798353, 0.000109053195]</td>\n",
       "      <td>[0.026576467, 0.9733123, 0.00011121639]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2992</th>\n",
       "      <td>[0.0004849217, 0.82130843, 0.17820662]</td>\n",
       "      <td>[0.00086266483, 0.16405062, 0.8350867]</td>\n",
       "      <td>[0.00011355382, 0.0038210351, 0.99606544]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2993</th>\n",
       "      <td>[0.43858352, 0.35041612, 0.21100032]</td>\n",
       "      <td>[0.011102119, 0.98604786, 0.002850076]</td>\n",
       "      <td>[0.0002352526, 0.999734, 3.0689174e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2994</th>\n",
       "      <td>[0.007363762, 0.03469444, 0.9579418]</td>\n",
       "      <td>[0.00038859833, 0.0002748932, 0.99933654]</td>\n",
       "      <td>[0.000146678, 0.00027509205, 0.9995782]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2995</th>\n",
       "      <td>[5.5074986e-07, 0.9999994, 3.1386886e-09]</td>\n",
       "      <td>[1.1006394e-05, 0.9999789, 1.0096202e-05]</td>\n",
       "      <td>[5.4615495e-05, 0.99992585, 1.9504161e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2996</th>\n",
       "      <td>[0.00054795184, 0.3214301, 0.67802197]</td>\n",
       "      <td>[9.5943695e-05, 0.002728676, 0.99717534]</td>\n",
       "      <td>[3.364683e-05, 0.0001740074, 0.9997924]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2997</th>\n",
       "      <td>[1.0340856e-05, 0.9999889, 7.5271475e-07]</td>\n",
       "      <td>[0.00015814147, 0.99613684, 0.0037049148]</td>\n",
       "      <td>[0.00016778223, 0.99978524, 4.694916e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2998</th>\n",
       "      <td>[0.02169743, 0.95809954, 0.020203082]</td>\n",
       "      <td>[0.00021270107, 0.9985361, 0.0012511375]</td>\n",
       "      <td>[6.2617466e-05, 0.99991655, 2.0800484e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2999</th>\n",
       "      <td>[3.4449095e-05, 0.99996555, 3.870269e-09]</td>\n",
       "      <td>[6.391212e-05, 0.9999279, 8.209448e-06]</td>\n",
       "      <td>[6.275527e-05, 0.9999261, 1.1155761e-05]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3000 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                    nnlm-128_tsize_18783_epoch30  \\\n",
       "0       [0.00013908364, 0.9998254, 3.547703e-05]   \n",
       "1       [0.002030536, 0.99750996, 0.00045953243]   \n",
       "2       [0.0006725391, 0.99926144, 6.598066e-05]   \n",
       "3       [0.0003392862, 0.9996592, 1.5918495e-06]   \n",
       "4      [4.3006254e-05, 0.99995697, 8.022627e-09]   \n",
       "5            [0.08427183, 0.8487675, 0.06696065]   \n",
       "6          [0.001172376, 0.9711367, 0.027690923]   \n",
       "7     [2.1608248e-06, 0.99978226, 0.00021563252]   \n",
       "8       [0.00057175703, 0.9981173, 0.0013109304]   \n",
       "9      [7.477624e-05, 0.99992514, 1.1235904e-07]   \n",
       "10       [0.021710116, 2.5961535e-05, 0.9782639]   \n",
       "11     [0.00060418097, 0.99912363, 0.0002722138]   \n",
       "12           [0.9535876, 0.0369543, 0.009458121]   \n",
       "13       [0.010873717, 0.98802453, 0.0011016902]   \n",
       "14    [1.2437995e-05, 0.99998605, 1.4981332e-06]   \n",
       "15        [0.00022437192, 0.40065616, 0.5991194]   \n",
       "16            [0.42409116, 0.3661773, 0.2097315]   \n",
       "17       [0.00028815673, 0.02189521, 0.97781664]   \n",
       "18      [0.0049450295, 0.99437827, 0.0006767206]   \n",
       "19      [0.0021360728, 0.9973984, 0.00046561955]   \n",
       "20        [0.012432584, 0.9875252, 4.218152e-05]   \n",
       "21    [3.4952702e-06, 0.99999654, 5.4307318e-08]   \n",
       "22        [0.009834621, 0.031162515, 0.95900285]   \n",
       "23      [2.863268e-05, 0.9999547, 1.6737697e-05]   \n",
       "24    [0.00046800202, 0.99953043, 1.5787145e-06]   \n",
       "25      [0.0009937334, 0.9990043, 2.0188118e-06]   \n",
       "26      [0.00067893823, 0.999321, 1.2991575e-07]   \n",
       "27       [0.002937752, 0.9970304, 3.1891446e-05]   \n",
       "28        [0.00021700047, 0.12237597, 0.8774071]   \n",
       "29      [0.00041335513, 0.99958247, 4.15565e-06]   \n",
       "...                                          ...   \n",
       "2970     [0.00013102862, 0.999859, 1.002201e-05]   \n",
       "2971    [0.0063016093, 0.9936958, 2.5687439e-06]   \n",
       "2972    [8.079995e-05, 0.9999192, 5.4082644e-08]   \n",
       "2973      [0.011751247, 0.9881732, 7.558385e-05]   \n",
       "2974       [0.00568506, 0.9927377, 0.0015771401]   \n",
       "2975    [0.0013939402, 0.9986022, 3.8363874e-06]   \n",
       "2976  [6.0102193e-05, 0.00019435708, 0.99974555]   \n",
       "2977   [0.0014636873, 0.99853516, 1.1816319e-06]   \n",
       "2978         [0.16959986, 0.6882856, 0.14211453]   \n",
       "2979     [0.0017004324, 0.998131, 0.00016857816]   \n",
       "2980    [0.00012538987, 0.9989166, 0.0009579694]   \n",
       "2981     [0.009910916, 0.98633945, 0.0037496267]   \n",
       "2982       [0.010100571, 0.98953, 0.00036947953]   \n",
       "2983    [4.693483e-06, 0.99999535, 5.278013e-09]   \n",
       "2984    [0.00018049637, 0.9998196, 3.956059e-08]   \n",
       "2985   [6.223519e-05, 0.99993634, 1.4155079e-06]   \n",
       "2986      [0.15090811, 0.84902453, 6.727478e-05]   \n",
       "2987    [5.171855e-05, 0.99994826, 5.354327e-09]   \n",
       "2988       [0.05583445, 0.944164, 1.5643257e-06]   \n",
       "2989         [0.7011099, 0.03436372, 0.26452637]   \n",
       "2990      [0.42243278, 0.57580805, 0.0017591851]   \n",
       "2991     [0.12673624, 0.87326056, 3.2519938e-06]   \n",
       "2992      [0.0004849217, 0.82130843, 0.17820662]   \n",
       "2993        [0.43858352, 0.35041612, 0.21100032]   \n",
       "2994        [0.007363762, 0.03469444, 0.9579418]   \n",
       "2995   [5.5074986e-07, 0.9999994, 3.1386886e-09]   \n",
       "2996      [0.00054795184, 0.3214301, 0.67802197]   \n",
       "2997   [1.0340856e-05, 0.9999889, 7.5271475e-07]   \n",
       "2998       [0.02169743, 0.95809954, 0.020203082]   \n",
       "2999   [3.4449095e-05, 0.99996555, 3.870269e-09]   \n",
       "\n",
       "                         DAN_tsize_18783_epoch30  \\\n",
       "0     [2.5715732e-05, 0.9998716, 0.000102616505]   \n",
       "1        [2.547616e-05, 0.9989743, 0.0010002507]   \n",
       "2        [1.7972814e-05, 0.999974, 8.038951e-06]   \n",
       "3      [1.8328115e-05, 0.9999672, 1.4433712e-05]   \n",
       "4      [1.7900435e-05, 0.9999765, 5.5927703e-06]   \n",
       "5          [0.0026555285, 0.06760905, 0.9297355]   \n",
       "6      [2.252558e-05, 0.99995816, 1.9321673e-05]   \n",
       "7         [0.00030092022, 0.9585573, 0.04114188]   \n",
       "8      [1.6481239e-05, 0.99988806, 9.542799e-05]   \n",
       "9      [5.6351914e-06, 0.9999902, 4.1621734e-06]   \n",
       "10     [0.00081079284, 0.0007934927, 0.99839574]   \n",
       "11     [6.452492e-05, 0.99992394, 1.1570731e-05]   \n",
       "12           [0.06544465, 0.4085732, 0.52598214]   \n",
       "13    [1.33834255e-05, 0.99996865, 1.796761e-05]   \n",
       "14     [6.814804e-06, 0.99998903, 4.2227966e-06]   \n",
       "15        [0.0018520205, 0.23281698, 0.76533103]   \n",
       "16    [0.00018075421, 0.00020491573, 0.99961436]   \n",
       "17         [0.0020194512, 0.35163933, 0.6463412]   \n",
       "18      [0.0011020844, 0.9986578, 0.00024008311]   \n",
       "19          [0.06276215, 0.010787978, 0.9264499]   \n",
       "20      [6.9983995e-05, 0.9999213, 8.655985e-06]   \n",
       "21     [3.8394963e-05, 0.9999201, 4.1535608e-05]   \n",
       "22     [3.8877777e-05, 0.99868363, 0.0012774161]   \n",
       "23      [9.463277e-06, 0.9999757, 1.4949457e-05]   \n",
       "24       [7.325024e-06, 0.9999875, 5.193338e-06]   \n",
       "25     [0.0066787857, 0.99322104, 0.00010020443]   \n",
       "26    [1.5814856e-05, 0.99997985, 4.2840243e-06]   \n",
       "27      [0.000570712, 0.9993119, 0.000117387346]   \n",
       "28      [0.00014342695, 0.0009917042, 0.9988649]   \n",
       "29     [1.3541771e-05, 0.9999796, 6.7871615e-06]   \n",
       "...                                          ...   \n",
       "2970     [8.479956e-06, 0.9999844, 7.094274e-06]   \n",
       "2971   [2.0126467e-05, 0.9999597, 2.0086696e-05]   \n",
       "2972   [1.0396791e-05, 0.9999838, 5.8043006e-06]   \n",
       "2973    [1.2453335e-05, 0.9999789, 8.749407e-06]   \n",
       "2974     [0.0001500005, 0.9998109, 3.910941e-05]   \n",
       "2975    [6.4183373e-06, 0.9999894, 4.162622e-06]   \n",
       "2976     [9.494994e-05, 9.341852e-05, 0.9998117]   \n",
       "2977  [1.3079727e-05, 0.99998057, 6.3716625e-06]   \n",
       "2978      [0.0028511407, 0.9726916, 0.024457194]   \n",
       "2979    [0.001993633, 0.99798137, 2.5018555e-05]   \n",
       "2980   [1.2080088e-05, 0.9999558, 3.2193548e-05]   \n",
       "2981     [7.797504e-06, 0.9999833, 8.987446e-06]   \n",
       "2982       [2.1385e-05, 0.9999696, 9.035759e-06]   \n",
       "2983  [1.6061853e-05, 0.99997807, 5.8738624e-06]   \n",
       "2984   [1.3465946e-05, 0.99998164, 4.930999e-06]   \n",
       "2985     [2.096273e-05, 0.9999702, 8.876558e-06]   \n",
       "2986   [1.6015809e-05, 0.9999778, 6.1886635e-06]   \n",
       "2987   [0.0001350763, 0.99985313, 1.1741929e-05]   \n",
       "2988    [0.015329741, 0.98452073, 0.00014957698]   \n",
       "2989        [0.19836606, 0.009145228, 0.7924887]   \n",
       "2990   [0.00037372633, 0.9994367, 0.00018960005]   \n",
       "2991    [0.020055577, 0.9798353, 0.000109053195]   \n",
       "2992      [0.00086266483, 0.16405062, 0.8350867]   \n",
       "2993      [0.011102119, 0.98604786, 0.002850076]   \n",
       "2994   [0.00038859833, 0.0002748932, 0.99933654]   \n",
       "2995   [1.1006394e-05, 0.9999789, 1.0096202e-05]   \n",
       "2996    [9.5943695e-05, 0.002728676, 0.99717534]   \n",
       "2997   [0.00015814147, 0.99613684, 0.0037049148]   \n",
       "2998    [0.00021270107, 0.9985361, 0.0012511375]   \n",
       "2999     [6.391212e-05, 0.9999279, 8.209448e-06]   \n",
       "\n",
       "                       univ3_tsize_18783_epoch30  \n",
       "0         [7.91436e-05, 0.9998338, 8.701929e-05]  \n",
       "1         [7.870038e-05, 0.9997701, 0.000151296]  \n",
       "2       [6.0237904e-05, 0.9999143, 2.548163e-05]  \n",
       "3       [4.4898494e-05, 0.9999287, 2.633073e-05]  \n",
       "4        [7.507247e-05, 0.9999107, 1.421926e-05]  \n",
       "5     [0.00016169962, 0.99982136, 1.6898892e-05]  \n",
       "6      [5.2135645e-05, 0.99991536, 3.253769e-05]  \n",
       "7       [6.164825e-05, 0.9998791, 5.9188635e-05]  \n",
       "8      [6.790512e-05, 0.99990237, 2.9676981e-05]  \n",
       "9      [6.1978855e-05, 0.9999176, 2.0371614e-05]  \n",
       "10      [7.8857935e-05, 0.0003978529, 0.9995233]  \n",
       "11      [7.210481e-05, 0.9999161, 1.1803859e-05]  \n",
       "12     [0.00020762562, 0.00060981687, 0.9991825]  \n",
       "13     [4.5374585e-05, 0.9999213, 3.3305845e-05]  \n",
       "14     [5.612093e-05, 0.99992216, 2.1637394e-05]  \n",
       "15    [0.00013584517, 0.00024865704, 0.99961543]  \n",
       "16      [8.948143e-05, 0.00019811906, 0.9997124]  \n",
       "17       [0.00022410582, 0.013034229, 0.9867417]  \n",
       "18      [5.753277e-05, 0.99992466, 1.777614e-05]  \n",
       "19     [5.8154856e-05, 0.00014961355, 0.9997923]  \n",
       "20     [7.0063914e-05, 0.9999087, 2.1275844e-05]  \n",
       "21    [4.8331603e-05, 0.99993205, 1.9714593e-05]  \n",
       "22    [5.3785814e-05, 0.00030585722, 0.99964035]  \n",
       "23    [5.3389886e-05, 0.99991834, 2.8288976e-05]  \n",
       "24      [4.9465325e-05, 0.9999325, 1.801631e-05]  \n",
       "25       [5.763131e-05, 0.999928, 1.4367954e-05]  \n",
       "26       [7.19162e-05, 0.9999155, 1.2658613e-05]  \n",
       "27     [5.7950838e-05, 0.9999305, 1.1548351e-05]  \n",
       "28       [7.705059e-05, 0.0055413716, 0.9943816]  \n",
       "29    [5.4944056e-05, 0.99992704, 1.7942342e-05]  \n",
       "...                                          ...  \n",
       "2970   [5.0472172e-05, 0.9999002, 4.9328337e-05]  \n",
       "2971   [6.4939864e-05, 0.9999056, 2.9467843e-05]  \n",
       "2972   [4.5115954e-05, 0.9999186, 3.6302445e-05]  \n",
       "2973   [4.8256217e-05, 0.99992466, 2.707365e-05]  \n",
       "2974    [5.503107e-05, 0.9999263, 1.8559715e-05]  \n",
       "2975   [5.2535448e-05, 0.9999291, 1.8341178e-05]  \n",
       "2976    [6.442183e-05, 0.00026567586, 0.9996699]  \n",
       "2977   [8.573858e-05, 0.99989474, 1.9587023e-05]  \n",
       "2978      [0.048414215, 0.036332503, 0.91525334]  \n",
       "2979   [0.00042199215, 0.9995654, 1.2613746e-05]  \n",
       "2980   [4.9621558e-05, 0.9999237, 2.6679521e-05]  \n",
       "2981    [4.5276945e-05, 0.999912, 4.2622516e-05]  \n",
       "2982  [0.000110848836, 0.9998785, 1.0575883e-05]  \n",
       "2983   [4.7603855e-05, 0.9999274, 2.5006837e-05]  \n",
       "2984   [5.0970968e-05, 0.9999248, 2.4255407e-05]  \n",
       "2985  [6.0911327e-05, 0.99992037, 1.8680157e-05]  \n",
       "2986  [5.6788038e-05, 0.99992776, 1.5480824e-05]  \n",
       "2987    [0.00010627753, 0.999884, 9.6272715e-06]  \n",
       "2988    [9.4243944e-05, 0.999892, 1.3700564e-05]  \n",
       "2989    [0.0008914261, 0.9989819, 0.00012670354]  \n",
       "2990    [0.0006359286, 0.9993395, 2.4561226e-05]  \n",
       "2991     [0.026576467, 0.9733123, 0.00011121639]  \n",
       "2992   [0.00011355382, 0.0038210351, 0.99606544]  \n",
       "2993     [0.0002352526, 0.999734, 3.0689174e-05]  \n",
       "2994     [0.000146678, 0.00027509205, 0.9995782]  \n",
       "2995  [5.4615495e-05, 0.99992585, 1.9504161e-05]  \n",
       "2996     [3.364683e-05, 0.0001740074, 0.9997924]  \n",
       "2997   [0.00016778223, 0.99978524, 4.694916e-05]  \n",
       "2998  [6.2617466e-05, 0.99991655, 2.0800484e-05]  \n",
       "2999    [6.275527e-05, 0.9999261, 1.1155761e-05]  \n",
       "\n",
       "[3000 rows x 3 columns]"
      ]
     },
     "execution_count": 131,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "probs = probs.iloc[:,3:6]\n",
    "probs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "nnlm-128_tsize_18783_epoch30      [0.00013908364, 0.9998254, 3.547703e-05]\n",
       "DAN_tsize_18783_epoch30         [2.5715732e-05, 0.9998716, 0.000102616505]\n",
       "univ3_tsize_18783_epoch30           [7.91436e-05, 0.9998338, 8.701929e-05]\n",
       "Name: 0, dtype: object"
      ]
     },
     "execution_count": 134,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "probs.iloc[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.0001, 0.9998, 0.0000], dtype=float32)"
      ]
     },
     "execution_count": 135,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "probs.iloc[0][0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 178,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.0001 0.9998 0.0000]\n",
      "[0.0000 0.9999 0.0001]\n"
     ]
    }
   ],
   "source": [
    "a = probs.iloc[0][0]\n",
    "b = probs.iloc[0][1]\n",
    "print(a)\n",
    "print(b)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 188,
   "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>nnlm-128_tsize_18783_epoch30</th>\n",
       "      <th>DAN_tsize_18783_epoch30</th>\n",
       "      <th>univ3_tsize_18783_epoch30</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>[0.00013908364, 0.9998254, 3.547703e-05]</td>\n",
       "      <td>[2.5715732e-05, 0.9998716, 0.000102616505]</td>\n",
       "      <td>[7.91436e-05, 0.9998338, 8.701929e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>[0.002030536, 0.99750996, 0.00045953243]</td>\n",
       "      <td>[2.547616e-05, 0.9989743, 0.0010002507]</td>\n",
       "      <td>[7.870038e-05, 0.9997701, 0.000151296]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>[0.0006725391, 0.99926144, 6.598066e-05]</td>\n",
       "      <td>[1.7972814e-05, 0.999974, 8.038951e-06]</td>\n",
       "      <td>[6.0237904e-05, 0.9999143, 2.548163e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>[0.0003392862, 0.9996592, 1.5918495e-06]</td>\n",
       "      <td>[1.8328115e-05, 0.9999672, 1.4433712e-05]</td>\n",
       "      <td>[4.4898494e-05, 0.9999287, 2.633073e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>[4.3006254e-05, 0.99995697, 8.022627e-09]</td>\n",
       "      <td>[1.7900435e-05, 0.9999765, 5.5927703e-06]</td>\n",
       "      <td>[7.507247e-05, 0.9999107, 1.421926e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>[0.08427183, 0.8487675, 0.06696065]</td>\n",
       "      <td>[0.0026555285, 0.06760905, 0.9297355]</td>\n",
       "      <td>[0.00016169962, 0.99982136, 1.6898892e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>[0.001172376, 0.9711367, 0.027690923]</td>\n",
       "      <td>[2.252558e-05, 0.99995816, 1.9321673e-05]</td>\n",
       "      <td>[5.2135645e-05, 0.99991536, 3.253769e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>[2.1608248e-06, 0.99978226, 0.00021563252]</td>\n",
       "      <td>[0.00030092022, 0.9585573, 0.04114188]</td>\n",
       "      <td>[6.164825e-05, 0.9998791, 5.9188635e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>[0.00057175703, 0.9981173, 0.0013109304]</td>\n",
       "      <td>[1.6481239e-05, 0.99988806, 9.542799e-05]</td>\n",
       "      <td>[6.790512e-05, 0.99990237, 2.9676981e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>[7.477624e-05, 0.99992514, 1.1235904e-07]</td>\n",
       "      <td>[5.6351914e-06, 0.9999902, 4.1621734e-06]</td>\n",
       "      <td>[6.1978855e-05, 0.9999176, 2.0371614e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>[0.021710116, 2.5961535e-05, 0.9782639]</td>\n",
       "      <td>[0.00081079284, 0.0007934927, 0.99839574]</td>\n",
       "      <td>[7.8857935e-05, 0.0003978529, 0.9995233]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>[0.00060418097, 0.99912363, 0.0002722138]</td>\n",
       "      <td>[6.452492e-05, 0.99992394, 1.1570731e-05]</td>\n",
       "      <td>[7.210481e-05, 0.9999161, 1.1803859e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>[0.9535876, 0.0369543, 0.009458121]</td>\n",
       "      <td>[0.06544465, 0.4085732, 0.52598214]</td>\n",
       "      <td>[0.00020762562, 0.00060981687, 0.9991825]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>[0.010873717, 0.98802453, 0.0011016902]</td>\n",
       "      <td>[1.33834255e-05, 0.99996865, 1.796761e-05]</td>\n",
       "      <td>[4.5374585e-05, 0.9999213, 3.3305845e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>[1.2437995e-05, 0.99998605, 1.4981332e-06]</td>\n",
       "      <td>[6.814804e-06, 0.99998903, 4.2227966e-06]</td>\n",
       "      <td>[5.612093e-05, 0.99992216, 2.1637394e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>[0.00022437192, 0.40065616, 0.5991194]</td>\n",
       "      <td>[0.0018520205, 0.23281698, 0.76533103]</td>\n",
       "      <td>[0.00013584517, 0.00024865704, 0.99961543]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>[0.42409116, 0.3661773, 0.2097315]</td>\n",
       "      <td>[0.00018075421, 0.00020491573, 0.99961436]</td>\n",
       "      <td>[8.948143e-05, 0.00019811906, 0.9997124]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>[0.00028815673, 0.02189521, 0.97781664]</td>\n",
       "      <td>[0.0020194512, 0.35163933, 0.6463412]</td>\n",
       "      <td>[0.00022410582, 0.013034229, 0.9867417]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>[0.0049450295, 0.99437827, 0.0006767206]</td>\n",
       "      <td>[0.0011020844, 0.9986578, 0.00024008311]</td>\n",
       "      <td>[5.753277e-05, 0.99992466, 1.777614e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>[0.0021360728, 0.9973984, 0.00046561955]</td>\n",
       "      <td>[0.06276215, 0.010787978, 0.9264499]</td>\n",
       "      <td>[5.8154856e-05, 0.00014961355, 0.9997923]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>[0.012432584, 0.9875252, 4.218152e-05]</td>\n",
       "      <td>[6.9983995e-05, 0.9999213, 8.655985e-06]</td>\n",
       "      <td>[7.0063914e-05, 0.9999087, 2.1275844e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>[3.4952702e-06, 0.99999654, 5.4307318e-08]</td>\n",
       "      <td>[3.8394963e-05, 0.9999201, 4.1535608e-05]</td>\n",
       "      <td>[4.8331603e-05, 0.99993205, 1.9714593e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>[0.009834621, 0.031162515, 0.95900285]</td>\n",
       "      <td>[3.8877777e-05, 0.99868363, 0.0012774161]</td>\n",
       "      <td>[5.3785814e-05, 0.00030585722, 0.99964035]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>[2.863268e-05, 0.9999547, 1.6737697e-05]</td>\n",
       "      <td>[9.463277e-06, 0.9999757, 1.4949457e-05]</td>\n",
       "      <td>[5.3389886e-05, 0.99991834, 2.8288976e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>[0.00046800202, 0.99953043, 1.5787145e-06]</td>\n",
       "      <td>[7.325024e-06, 0.9999875, 5.193338e-06]</td>\n",
       "      <td>[4.9465325e-05, 0.9999325, 1.801631e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>[0.0009937334, 0.9990043, 2.0188118e-06]</td>\n",
       "      <td>[0.0066787857, 0.99322104, 0.00010020443]</td>\n",
       "      <td>[5.763131e-05, 0.999928, 1.4367954e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>[0.00067893823, 0.999321, 1.2991575e-07]</td>\n",
       "      <td>[1.5814856e-05, 0.99997985, 4.2840243e-06]</td>\n",
       "      <td>[7.19162e-05, 0.9999155, 1.2658613e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>[0.002937752, 0.9970304, 3.1891446e-05]</td>\n",
       "      <td>[0.000570712, 0.9993119, 0.000117387346]</td>\n",
       "      <td>[5.7950838e-05, 0.9999305, 1.1548351e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>[0.00021700047, 0.12237597, 0.8774071]</td>\n",
       "      <td>[0.00014342695, 0.0009917042, 0.9988649]</td>\n",
       "      <td>[7.705059e-05, 0.0055413716, 0.9943816]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>[0.00041335513, 0.99958247, 4.15565e-06]</td>\n",
       "      <td>[1.3541771e-05, 0.9999796, 6.7871615e-06]</td>\n",
       "      <td>[5.4944056e-05, 0.99992704, 1.7942342e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2970</th>\n",
       "      <td>[0.00013102862, 0.999859, 1.002201e-05]</td>\n",
       "      <td>[8.479956e-06, 0.9999844, 7.094274e-06]</td>\n",
       "      <td>[5.0472172e-05, 0.9999002, 4.9328337e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2971</th>\n",
       "      <td>[0.0063016093, 0.9936958, 2.5687439e-06]</td>\n",
       "      <td>[2.0126467e-05, 0.9999597, 2.0086696e-05]</td>\n",
       "      <td>[6.4939864e-05, 0.9999056, 2.9467843e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2972</th>\n",
       "      <td>[8.079995e-05, 0.9999192, 5.4082644e-08]</td>\n",
       "      <td>[1.0396791e-05, 0.9999838, 5.8043006e-06]</td>\n",
       "      <td>[4.5115954e-05, 0.9999186, 3.6302445e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2973</th>\n",
       "      <td>[0.011751247, 0.9881732, 7.558385e-05]</td>\n",
       "      <td>[1.2453335e-05, 0.9999789, 8.749407e-06]</td>\n",
       "      <td>[4.8256217e-05, 0.99992466, 2.707365e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2974</th>\n",
       "      <td>[0.00568506, 0.9927377, 0.0015771401]</td>\n",
       "      <td>[0.0001500005, 0.9998109, 3.910941e-05]</td>\n",
       "      <td>[5.503107e-05, 0.9999263, 1.8559715e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2975</th>\n",
       "      <td>[0.0013939402, 0.9986022, 3.8363874e-06]</td>\n",
       "      <td>[6.4183373e-06, 0.9999894, 4.162622e-06]</td>\n",
       "      <td>[5.2535448e-05, 0.9999291, 1.8341178e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2976</th>\n",
       "      <td>[6.0102193e-05, 0.00019435708, 0.99974555]</td>\n",
       "      <td>[9.494994e-05, 9.341852e-05, 0.9998117]</td>\n",
       "      <td>[6.442183e-05, 0.00026567586, 0.9996699]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2977</th>\n",
       "      <td>[0.0014636873, 0.99853516, 1.1816319e-06]</td>\n",
       "      <td>[1.3079727e-05, 0.99998057, 6.3716625e-06]</td>\n",
       "      <td>[8.573858e-05, 0.99989474, 1.9587023e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2978</th>\n",
       "      <td>[0.16959986, 0.6882856, 0.14211453]</td>\n",
       "      <td>[0.0028511407, 0.9726916, 0.024457194]</td>\n",
       "      <td>[0.048414215, 0.036332503, 0.91525334]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2979</th>\n",
       "      <td>[0.0017004324, 0.998131, 0.00016857816]</td>\n",
       "      <td>[0.001993633, 0.99798137, 2.5018555e-05]</td>\n",
       "      <td>[0.00042199215, 0.9995654, 1.2613746e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2980</th>\n",
       "      <td>[0.00012538987, 0.9989166, 0.0009579694]</td>\n",
       "      <td>[1.2080088e-05, 0.9999558, 3.2193548e-05]</td>\n",
       "      <td>[4.9621558e-05, 0.9999237, 2.6679521e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2981</th>\n",
       "      <td>[0.009910916, 0.98633945, 0.0037496267]</td>\n",
       "      <td>[7.797504e-06, 0.9999833, 8.987446e-06]</td>\n",
       "      <td>[4.5276945e-05, 0.999912, 4.2622516e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2982</th>\n",
       "      <td>[0.010100571, 0.98953, 0.00036947953]</td>\n",
       "      <td>[2.1385e-05, 0.9999696, 9.035759e-06]</td>\n",
       "      <td>[0.000110848836, 0.9998785, 1.0575883e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2983</th>\n",
       "      <td>[4.693483e-06, 0.99999535, 5.278013e-09]</td>\n",
       "      <td>[1.6061853e-05, 0.99997807, 5.8738624e-06]</td>\n",
       "      <td>[4.7603855e-05, 0.9999274, 2.5006837e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2984</th>\n",
       "      <td>[0.00018049637, 0.9998196, 3.956059e-08]</td>\n",
       "      <td>[1.3465946e-05, 0.99998164, 4.930999e-06]</td>\n",
       "      <td>[5.0970968e-05, 0.9999248, 2.4255407e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2985</th>\n",
       "      <td>[6.223519e-05, 0.99993634, 1.4155079e-06]</td>\n",
       "      <td>[2.096273e-05, 0.9999702, 8.876558e-06]</td>\n",
       "      <td>[6.0911327e-05, 0.99992037, 1.8680157e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2986</th>\n",
       "      <td>[0.15090811, 0.84902453, 6.727478e-05]</td>\n",
       "      <td>[1.6015809e-05, 0.9999778, 6.1886635e-06]</td>\n",
       "      <td>[5.6788038e-05, 0.99992776, 1.5480824e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2987</th>\n",
       "      <td>[5.171855e-05, 0.99994826, 5.354327e-09]</td>\n",
       "      <td>[0.0001350763, 0.99985313, 1.1741929e-05]</td>\n",
       "      <td>[0.00010627753, 0.999884, 9.6272715e-06]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2988</th>\n",
       "      <td>[0.05583445, 0.944164, 1.5643257e-06]</td>\n",
       "      <td>[0.015329741, 0.98452073, 0.00014957698]</td>\n",
       "      <td>[9.4243944e-05, 0.999892, 1.3700564e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2989</th>\n",
       "      <td>[0.7011099, 0.03436372, 0.26452637]</td>\n",
       "      <td>[0.19836606, 0.009145228, 0.7924887]</td>\n",
       "      <td>[0.0008914261, 0.9989819, 0.00012670354]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2990</th>\n",
       "      <td>[0.42243278, 0.57580805, 0.0017591851]</td>\n",
       "      <td>[0.00037372633, 0.9994367, 0.00018960005]</td>\n",
       "      <td>[0.0006359286, 0.9993395, 2.4561226e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2991</th>\n",
       "      <td>[0.12673624, 0.87326056, 3.2519938e-06]</td>\n",
       "      <td>[0.020055577, 0.9798353, 0.000109053195]</td>\n",
       "      <td>[0.026576467, 0.9733123, 0.00011121639]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2992</th>\n",
       "      <td>[0.0004849217, 0.82130843, 0.17820662]</td>\n",
       "      <td>[0.00086266483, 0.16405062, 0.8350867]</td>\n",
       "      <td>[0.00011355382, 0.0038210351, 0.99606544]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2993</th>\n",
       "      <td>[0.43858352, 0.35041612, 0.21100032]</td>\n",
       "      <td>[0.011102119, 0.98604786, 0.002850076]</td>\n",
       "      <td>[0.0002352526, 0.999734, 3.0689174e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2994</th>\n",
       "      <td>[0.007363762, 0.03469444, 0.9579418]</td>\n",
       "      <td>[0.00038859833, 0.0002748932, 0.99933654]</td>\n",
       "      <td>[0.000146678, 0.00027509205, 0.9995782]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2995</th>\n",
       "      <td>[5.5074986e-07, 0.9999994, 3.1386886e-09]</td>\n",
       "      <td>[1.1006394e-05, 0.9999789, 1.0096202e-05]</td>\n",
       "      <td>[5.4615495e-05, 0.99992585, 1.9504161e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2996</th>\n",
       "      <td>[0.00054795184, 0.3214301, 0.67802197]</td>\n",
       "      <td>[9.5943695e-05, 0.002728676, 0.99717534]</td>\n",
       "      <td>[3.364683e-05, 0.0001740074, 0.9997924]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2997</th>\n",
       "      <td>[1.0340856e-05, 0.9999889, 7.5271475e-07]</td>\n",
       "      <td>[0.00015814147, 0.99613684, 0.0037049148]</td>\n",
       "      <td>[0.00016778223, 0.99978524, 4.694916e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2998</th>\n",
       "      <td>[0.02169743, 0.95809954, 0.020203082]</td>\n",
       "      <td>[0.00021270107, 0.9985361, 0.0012511375]</td>\n",
       "      <td>[6.2617466e-05, 0.99991655, 2.0800484e-05]</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2999</th>\n",
       "      <td>[3.4449095e-05, 0.99996555, 3.870269e-09]</td>\n",
       "      <td>[6.391212e-05, 0.9999279, 8.209448e-06]</td>\n",
       "      <td>[6.275527e-05, 0.9999261, 1.1155761e-05]</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3000 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                    nnlm-128_tsize_18783_epoch30  \\\n",
       "0       [0.00013908364, 0.9998254, 3.547703e-05]   \n",
       "1       [0.002030536, 0.99750996, 0.00045953243]   \n",
       "2       [0.0006725391, 0.99926144, 6.598066e-05]   \n",
       "3       [0.0003392862, 0.9996592, 1.5918495e-06]   \n",
       "4      [4.3006254e-05, 0.99995697, 8.022627e-09]   \n",
       "5            [0.08427183, 0.8487675, 0.06696065]   \n",
       "6          [0.001172376, 0.9711367, 0.027690923]   \n",
       "7     [2.1608248e-06, 0.99978226, 0.00021563252]   \n",
       "8       [0.00057175703, 0.9981173, 0.0013109304]   \n",
       "9      [7.477624e-05, 0.99992514, 1.1235904e-07]   \n",
       "10       [0.021710116, 2.5961535e-05, 0.9782639]   \n",
       "11     [0.00060418097, 0.99912363, 0.0002722138]   \n",
       "12           [0.9535876, 0.0369543, 0.009458121]   \n",
       "13       [0.010873717, 0.98802453, 0.0011016902]   \n",
       "14    [1.2437995e-05, 0.99998605, 1.4981332e-06]   \n",
       "15        [0.00022437192, 0.40065616, 0.5991194]   \n",
       "16            [0.42409116, 0.3661773, 0.2097315]   \n",
       "17       [0.00028815673, 0.02189521, 0.97781664]   \n",
       "18      [0.0049450295, 0.99437827, 0.0006767206]   \n",
       "19      [0.0021360728, 0.9973984, 0.00046561955]   \n",
       "20        [0.012432584, 0.9875252, 4.218152e-05]   \n",
       "21    [3.4952702e-06, 0.99999654, 5.4307318e-08]   \n",
       "22        [0.009834621, 0.031162515, 0.95900285]   \n",
       "23      [2.863268e-05, 0.9999547, 1.6737697e-05]   \n",
       "24    [0.00046800202, 0.99953043, 1.5787145e-06]   \n",
       "25      [0.0009937334, 0.9990043, 2.0188118e-06]   \n",
       "26      [0.00067893823, 0.999321, 1.2991575e-07]   \n",
       "27       [0.002937752, 0.9970304, 3.1891446e-05]   \n",
       "28        [0.00021700047, 0.12237597, 0.8774071]   \n",
       "29      [0.00041335513, 0.99958247, 4.15565e-06]   \n",
       "...                                          ...   \n",
       "2970     [0.00013102862, 0.999859, 1.002201e-05]   \n",
       "2971    [0.0063016093, 0.9936958, 2.5687439e-06]   \n",
       "2972    [8.079995e-05, 0.9999192, 5.4082644e-08]   \n",
       "2973      [0.011751247, 0.9881732, 7.558385e-05]   \n",
       "2974       [0.00568506, 0.9927377, 0.0015771401]   \n",
       "2975    [0.0013939402, 0.9986022, 3.8363874e-06]   \n",
       "2976  [6.0102193e-05, 0.00019435708, 0.99974555]   \n",
       "2977   [0.0014636873, 0.99853516, 1.1816319e-06]   \n",
       "2978         [0.16959986, 0.6882856, 0.14211453]   \n",
       "2979     [0.0017004324, 0.998131, 0.00016857816]   \n",
       "2980    [0.00012538987, 0.9989166, 0.0009579694]   \n",
       "2981     [0.009910916, 0.98633945, 0.0037496267]   \n",
       "2982       [0.010100571, 0.98953, 0.00036947953]   \n",
       "2983    [4.693483e-06, 0.99999535, 5.278013e-09]   \n",
       "2984    [0.00018049637, 0.9998196, 3.956059e-08]   \n",
       "2985   [6.223519e-05, 0.99993634, 1.4155079e-06]   \n",
       "2986      [0.15090811, 0.84902453, 6.727478e-05]   \n",
       "2987    [5.171855e-05, 0.99994826, 5.354327e-09]   \n",
       "2988       [0.05583445, 0.944164, 1.5643257e-06]   \n",
       "2989         [0.7011099, 0.03436372, 0.26452637]   \n",
       "2990      [0.42243278, 0.57580805, 0.0017591851]   \n",
       "2991     [0.12673624, 0.87326056, 3.2519938e-06]   \n",
       "2992      [0.0004849217, 0.82130843, 0.17820662]   \n",
       "2993        [0.43858352, 0.35041612, 0.21100032]   \n",
       "2994        [0.007363762, 0.03469444, 0.9579418]   \n",
       "2995   [5.5074986e-07, 0.9999994, 3.1386886e-09]   \n",
       "2996      [0.00054795184, 0.3214301, 0.67802197]   \n",
       "2997   [1.0340856e-05, 0.9999889, 7.5271475e-07]   \n",
       "2998       [0.02169743, 0.95809954, 0.020203082]   \n",
       "2999   [3.4449095e-05, 0.99996555, 3.870269e-09]   \n",
       "\n",
       "                         DAN_tsize_18783_epoch30  \\\n",
       "0     [2.5715732e-05, 0.9998716, 0.000102616505]   \n",
       "1        [2.547616e-05, 0.9989743, 0.0010002507]   \n",
       "2        [1.7972814e-05, 0.999974, 8.038951e-06]   \n",
       "3      [1.8328115e-05, 0.9999672, 1.4433712e-05]   \n",
       "4      [1.7900435e-05, 0.9999765, 5.5927703e-06]   \n",
       "5          [0.0026555285, 0.06760905, 0.9297355]   \n",
       "6      [2.252558e-05, 0.99995816, 1.9321673e-05]   \n",
       "7         [0.00030092022, 0.9585573, 0.04114188]   \n",
       "8      [1.6481239e-05, 0.99988806, 9.542799e-05]   \n",
       "9      [5.6351914e-06, 0.9999902, 4.1621734e-06]   \n",
       "10     [0.00081079284, 0.0007934927, 0.99839574]   \n",
       "11     [6.452492e-05, 0.99992394, 1.1570731e-05]   \n",
       "12           [0.06544465, 0.4085732, 0.52598214]   \n",
       "13    [1.33834255e-05, 0.99996865, 1.796761e-05]   \n",
       "14     [6.814804e-06, 0.99998903, 4.2227966e-06]   \n",
       "15        [0.0018520205, 0.23281698, 0.76533103]   \n",
       "16    [0.00018075421, 0.00020491573, 0.99961436]   \n",
       "17         [0.0020194512, 0.35163933, 0.6463412]   \n",
       "18      [0.0011020844, 0.9986578, 0.00024008311]   \n",
       "19          [0.06276215, 0.010787978, 0.9264499]   \n",
       "20      [6.9983995e-05, 0.9999213, 8.655985e-06]   \n",
       "21     [3.8394963e-05, 0.9999201, 4.1535608e-05]   \n",
       "22     [3.8877777e-05, 0.99868363, 0.0012774161]   \n",
       "23      [9.463277e-06, 0.9999757, 1.4949457e-05]   \n",
       "24       [7.325024e-06, 0.9999875, 5.193338e-06]   \n",
       "25     [0.0066787857, 0.99322104, 0.00010020443]   \n",
       "26    [1.5814856e-05, 0.99997985, 4.2840243e-06]   \n",
       "27      [0.000570712, 0.9993119, 0.000117387346]   \n",
       "28      [0.00014342695, 0.0009917042, 0.9988649]   \n",
       "29     [1.3541771e-05, 0.9999796, 6.7871615e-06]   \n",
       "...                                          ...   \n",
       "2970     [8.479956e-06, 0.9999844, 7.094274e-06]   \n",
       "2971   [2.0126467e-05, 0.9999597, 2.0086696e-05]   \n",
       "2972   [1.0396791e-05, 0.9999838, 5.8043006e-06]   \n",
       "2973    [1.2453335e-05, 0.9999789, 8.749407e-06]   \n",
       "2974     [0.0001500005, 0.9998109, 3.910941e-05]   \n",
       "2975    [6.4183373e-06, 0.9999894, 4.162622e-06]   \n",
       "2976     [9.494994e-05, 9.341852e-05, 0.9998117]   \n",
       "2977  [1.3079727e-05, 0.99998057, 6.3716625e-06]   \n",
       "2978      [0.0028511407, 0.9726916, 0.024457194]   \n",
       "2979    [0.001993633, 0.99798137, 2.5018555e-05]   \n",
       "2980   [1.2080088e-05, 0.9999558, 3.2193548e-05]   \n",
       "2981     [7.797504e-06, 0.9999833, 8.987446e-06]   \n",
       "2982       [2.1385e-05, 0.9999696, 9.035759e-06]   \n",
       "2983  [1.6061853e-05, 0.99997807, 5.8738624e-06]   \n",
       "2984   [1.3465946e-05, 0.99998164, 4.930999e-06]   \n",
       "2985     [2.096273e-05, 0.9999702, 8.876558e-06]   \n",
       "2986   [1.6015809e-05, 0.9999778, 6.1886635e-06]   \n",
       "2987   [0.0001350763, 0.99985313, 1.1741929e-05]   \n",
       "2988    [0.015329741, 0.98452073, 0.00014957698]   \n",
       "2989        [0.19836606, 0.009145228, 0.7924887]   \n",
       "2990   [0.00037372633, 0.9994367, 0.00018960005]   \n",
       "2991    [0.020055577, 0.9798353, 0.000109053195]   \n",
       "2992      [0.00086266483, 0.16405062, 0.8350867]   \n",
       "2993      [0.011102119, 0.98604786, 0.002850076]   \n",
       "2994   [0.00038859833, 0.0002748932, 0.99933654]   \n",
       "2995   [1.1006394e-05, 0.9999789, 1.0096202e-05]   \n",
       "2996    [9.5943695e-05, 0.002728676, 0.99717534]   \n",
       "2997   [0.00015814147, 0.99613684, 0.0037049148]   \n",
       "2998    [0.00021270107, 0.9985361, 0.0012511375]   \n",
       "2999     [6.391212e-05, 0.9999279, 8.209448e-06]   \n",
       "\n",
       "                       univ3_tsize_18783_epoch30  \n",
       "0         [7.91436e-05, 0.9998338, 8.701929e-05]  \n",
       "1         [7.870038e-05, 0.9997701, 0.000151296]  \n",
       "2       [6.0237904e-05, 0.9999143, 2.548163e-05]  \n",
       "3       [4.4898494e-05, 0.9999287, 2.633073e-05]  \n",
       "4        [7.507247e-05, 0.9999107, 1.421926e-05]  \n",
       "5     [0.00016169962, 0.99982136, 1.6898892e-05]  \n",
       "6      [5.2135645e-05, 0.99991536, 3.253769e-05]  \n",
       "7       [6.164825e-05, 0.9998791, 5.9188635e-05]  \n",
       "8      [6.790512e-05, 0.99990237, 2.9676981e-05]  \n",
       "9      [6.1978855e-05, 0.9999176, 2.0371614e-05]  \n",
       "10      [7.8857935e-05, 0.0003978529, 0.9995233]  \n",
       "11      [7.210481e-05, 0.9999161, 1.1803859e-05]  \n",
       "12     [0.00020762562, 0.00060981687, 0.9991825]  \n",
       "13     [4.5374585e-05, 0.9999213, 3.3305845e-05]  \n",
       "14     [5.612093e-05, 0.99992216, 2.1637394e-05]  \n",
       "15    [0.00013584517, 0.00024865704, 0.99961543]  \n",
       "16      [8.948143e-05, 0.00019811906, 0.9997124]  \n",
       "17       [0.00022410582, 0.013034229, 0.9867417]  \n",
       "18      [5.753277e-05, 0.99992466, 1.777614e-05]  \n",
       "19     [5.8154856e-05, 0.00014961355, 0.9997923]  \n",
       "20     [7.0063914e-05, 0.9999087, 2.1275844e-05]  \n",
       "21    [4.8331603e-05, 0.99993205, 1.9714593e-05]  \n",
       "22    [5.3785814e-05, 0.00030585722, 0.99964035]  \n",
       "23    [5.3389886e-05, 0.99991834, 2.8288976e-05]  \n",
       "24      [4.9465325e-05, 0.9999325, 1.801631e-05]  \n",
       "25       [5.763131e-05, 0.999928, 1.4367954e-05]  \n",
       "26       [7.19162e-05, 0.9999155, 1.2658613e-05]  \n",
       "27     [5.7950838e-05, 0.9999305, 1.1548351e-05]  \n",
       "28       [7.705059e-05, 0.0055413716, 0.9943816]  \n",
       "29    [5.4944056e-05, 0.99992704, 1.7942342e-05]  \n",
       "...                                          ...  \n",
       "2970   [5.0472172e-05, 0.9999002, 4.9328337e-05]  \n",
       "2971   [6.4939864e-05, 0.9999056, 2.9467843e-05]  \n",
       "2972   [4.5115954e-05, 0.9999186, 3.6302445e-05]  \n",
       "2973   [4.8256217e-05, 0.99992466, 2.707365e-05]  \n",
       "2974    [5.503107e-05, 0.9999263, 1.8559715e-05]  \n",
       "2975   [5.2535448e-05, 0.9999291, 1.8341178e-05]  \n",
       "2976    [6.442183e-05, 0.00026567586, 0.9996699]  \n",
       "2977   [8.573858e-05, 0.99989474, 1.9587023e-05]  \n",
       "2978      [0.048414215, 0.036332503, 0.91525334]  \n",
       "2979   [0.00042199215, 0.9995654, 1.2613746e-05]  \n",
       "2980   [4.9621558e-05, 0.9999237, 2.6679521e-05]  \n",
       "2981    [4.5276945e-05, 0.999912, 4.2622516e-05]  \n",
       "2982  [0.000110848836, 0.9998785, 1.0575883e-05]  \n",
       "2983   [4.7603855e-05, 0.9999274, 2.5006837e-05]  \n",
       "2984   [5.0970968e-05, 0.9999248, 2.4255407e-05]  \n",
       "2985  [6.0911327e-05, 0.99992037, 1.8680157e-05]  \n",
       "2986  [5.6788038e-05, 0.99992776, 1.5480824e-05]  \n",
       "2987    [0.00010627753, 0.999884, 9.6272715e-06]  \n",
       "2988    [9.4243944e-05, 0.999892, 1.3700564e-05]  \n",
       "2989    [0.0008914261, 0.9989819, 0.00012670354]  \n",
       "2990    [0.0006359286, 0.9993395, 2.4561226e-05]  \n",
       "2991     [0.026576467, 0.9733123, 0.00011121639]  \n",
       "2992   [0.00011355382, 0.0038210351, 0.99606544]  \n",
       "2993     [0.0002352526, 0.999734, 3.0689174e-05]  \n",
       "2994     [0.000146678, 0.00027509205, 0.9995782]  \n",
       "2995  [5.4615495e-05, 0.99992585, 1.9504161e-05]  \n",
       "2996     [3.364683e-05, 0.0001740074, 0.9997924]  \n",
       "2997   [0.00016778223, 0.99978524, 4.694916e-05]  \n",
       "2998  [6.2617466e-05, 0.99991655, 2.0800484e-05]  \n",
       "2999    [6.275527e-05, 0.9999261, 1.1155761e-05]  \n",
       "\n",
       "[3000 rows x 3 columns]"
      ]
     },
     "execution_count": 188,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "probs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Multiclass logloss \n",
    "http://wiki.fast.ai/index.php/Log_Loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 141,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import log_loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 145,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "numpy.ndarray"
      ]
     },
     "execution_count": 145,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "type(a)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 148,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2.1094237467877998e-15"
      ]
     },
     "execution_count": 148,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "log_loss(np.array([0,1,2]),np.array([[1,0,0],[0,1,0],[0,0,1]]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Cross entropy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 290,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True\n"
     ]
    }
   ],
   "source": [
    "def cross_entropy(predictions, targets, epsilon=1e-12):\n",
    "    \"\"\"\n",
    "    Computes cross entropy between targets (encoded as one-hot vectors)\n",
    "    and predictions. \n",
    "    Input: predictions (N, k) ndarray\n",
    "           targets (N, k) ndarray        \n",
    "    Returns: scalar\n",
    "    \"\"\"\n",
    "    predictions = np.clip(predictions, epsilon, 1. - epsilon)\n",
    "    N = predictions.shape[0]\n",
    "    ce = -np.sum(targets*np.log(predictions+1e-9))/N\n",
    "    return ce\n",
    "\n",
    "predictions = np.array([[0.25,0.25,0.25,0.25],\n",
    "                        [0.01,0.01,0.01,0.96]])\n",
    "targets = np.array([[0,0,0,1],\n",
    "                   [0,0,0,1]])\n",
    "ans = 0.71355817782  #Correct answer\n",
    "x = cross_entropy(predictions, targets)\n",
    "print(np.isclose(x,ans))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 291,
   "metadata": {},
   "outputs": [],
   "source": [
    "def mean_cross_entropy(a, b):\n",
    "    ''' input: takes list of items instead of single softmax'''\n",
    "    sum = 0\n",
    "    for i in range(len(a)):\n",
    "        sum += cross_entropy(a[i], b[i])\n",
    "    # mean\n",
    "    return (sum/ len(a))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 292,
   "metadata": {},
   "outputs": [],
   "source": [
    "a = probs.iloc[0][0]\n",
    "b = probs.iloc[0][1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 293,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.0004848038467268149"
      ]
     },
     "execution_count": 293,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cross_entropy(a,b)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 180,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.41588830713596714"
      ]
     },
     "execution_count": 180,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cross_entropy([0.25,0.25,0.5],[0.7, 0.1, 0.2] )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 181,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "6.907422112204443"
      ]
     },
     "execution_count": 181,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cross_entropy([1,0,0],[0, 1,0] )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 196,
   "metadata": {},
   "outputs": [],
   "source": [
    "a = probs.iloc[:,0]\n",
    "b = probs.iloc[:,1]\n",
    "c = probs.iloc[:,2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 195,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.13363556317385797"
      ]
     },
     "execution_count": 195,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mean_cross_entropy(a, b)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 197,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.11561411556630624"
      ]
     },
     "execution_count": 197,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mean_cross_entropy(a,c)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 198,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.10730167065954697"
      ]
     },
     "execution_count": 198,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mean_cross_entropy(b,c)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 161,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[  70   93   10]\n",
      " [  60 2168   51]\n",
      " [   8   41  499]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.metrics import confusion_matrix\n",
    "\n",
    "cm = confusion_matrix(true_labels, yhat)\n",
    "CLASSES = [\"hatespeech\", \"offensive language\", \"neither\"]\n",
    "plot_confusion_matrix(cm, CLASSES)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Analyze biggest errors\n",
    "\n",
    "Turn labels into 1-hot. Multiply by predicted prob for labels.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def oneHot(Y, classes=3):\n",
    "    \"\"\"\n",
    "    Turn input of 1-dim values into 1-hot vector\n",
    "    in: array([5, 0, 4, 1, 9,...])\n",
    "    out[0] array([0., 0., 0., 0., 0., 1., 0., 0., 0., 0.])\n",
    "    \"\"\"\n",
    "    #classes = 10\n",
    "    samples = len(Y)\n",
    "\n",
    "    #a = np.array([1, 0, 3])\n",
    "    b = np.zeros((samples, classes))\n",
    "    b[np.arange(samples), Y] = 1\n",
    "    return b"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 420,
   "metadata": {},
   "outputs": [],
   "source": [
    "def getErrors(model_softmax, y_onehot):\n",
    "    ''' input predictions of model as softmax, true labels\n",
    "        count errors for each item'''\n",
    "    \n",
    "\n",
    "    comb = model_softmax * y_onehot\n",
    "    comb = comb.sum(axis=1)\n",
    "\n",
    "    comb = pd.DataFrame(comb)\n",
    "    comb.columns=['weight_correct']\n",
    "    comb['error'] = 1 - comb['weight_correct']\n",
    "    comb['true_label']=true_labels\n",
    "    comb['predicted'] = predicted_labels\n",
    "    \n",
    "    comb = pd.concat([comb, test[['id', 'text']] ], axis=1)\n",
    "    return comb "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def addText(comb)\n",
    "\n",
    "    return comb = pd.concat([comb, test[['id', 'text']] ], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 421,
   "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>weight_correct</th>\n",
       "      <th>error</th>\n",
       "      <th>true_label</th>\n",
       "      <th>predicted</th>\n",
       "      <th>id</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.999834</td>\n",
       "      <td>0.000166</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>23437</td>\n",
       "      <td>Wizard turned into a hoe after a break up</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.999770</td>\n",
       "      <td>0.000230</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>17813</td>\n",
       "      <td>RT  MENTION :  Captain Kirk hoe !</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.999914</td>\n",
       "      <td>0.000086</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>2561</td>\n",
       "      <td>MENTION bruh did go out like a bitch thoe lol that should not be his final match</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.999929</td>\n",
       "      <td>0.000071</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>18228</td>\n",
       "      <td>RT  MENTION :  RT  MENTION :  when you tryna get bitches on LinkedIn  URL</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.999911</td>\n",
       "      <td>0.000089</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>23030</td>\n",
       "      <td>Well I wish I could shoot the shit wit u bitches but daddy gotta get dis money  . one</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.999821</td>\n",
       "      <td>0.000179</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>16373</td>\n",
       "      <td>RT  MENTION :  Niggas can do slope intercept form equations but cant make scrambled fuckin eggs thats a lil backwards</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.999915</td>\n",
       "      <td>0.000085</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>3560</td>\n",
       "      <td>MENTION He must think you bump pussy .  I no mind .  I ' ll eat your  # thong and make you  # squirt like the  # geyser .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0.999879</td>\n",
       "      <td>0.000121</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>2692</td>\n",
       "      <td>MENTION always eatin good wit my niggah ,  and good lookin on the cd my boy got some heat now i can bump some bangers</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0.999902</td>\n",
       "      <td>0.000098</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>23549</td>\n",
       "      <td>Ya last nicca couldn ' t beat it  .  .  I hit it like a free throw !</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>0.999918</td>\n",
       "      <td>0.000082</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>12180</td>\n",
       "      <td>It ain ' t nothin to cut that bitch off</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.999523</td>\n",
       "      <td>0.000477</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>24918</td>\n",
       "      <td>she was telling me about this kid who ' s soles of his shoes are falling off .  I am going to get him some shoes .  hope his parents don ' t get mad</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>0.999916</td>\n",
       "      <td>0.000084</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>16493</td>\n",
       "      <td>RT  MENTION :  smh u got Ebola now RT MENTION ;  MENTION :  I stick my foot out to feel some freshness and this bitch ass mosquito has the nerve to  MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>0.999183</td>\n",
       "      <td>0.000817</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>16532</td>\n",
       "      <td>RT  MENTION :  trash liars RT  MENTION :  these law students are terrible liars  # HTGAWM</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>0.999921</td>\n",
       "      <td>0.000079</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>15319</td>\n",
       "      <td>RT  MENTION :  All hoes are friendly .  Althea wants to say hi .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>0.999922</td>\n",
       "      <td>0.000078</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>21838</td>\n",
       "      <td>That bitch mouth made the honor roll</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>0.999615</td>\n",
       "      <td>0.000385</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>3523</td>\n",
       "      <td>MENTION I wonder if I have my very own NSA spook following me by now ?</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>0.000198</td>\n",
       "      <td>0.999802</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>6008</td>\n",
       "      <td>MENTION when you realize you ' re trash</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>0.013034</td>\n",
       "      <td>0.986966</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1429</td>\n",
       "      <td>MENTION ;  MENTION :  I was told never fold  &amp; amp ;  never sell my sole quote They Should Have Told Yu Stay In School  !   * Soul MENTION ;  autocorrect hoed me</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>0.999925</td>\n",
       "      <td>0.000075</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>3240</td>\n",
       "      <td>MENTION khaye I can ' t see your freakin emojis I ' m a basic bitch today .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>0.999792</td>\n",
       "      <td>0.000208</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>21523</td>\n",
       "      <td>Someone should explain to Blake griffin that talking trash to an opposing 3rd string is unbecoming ,  even amongst clippers .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>0.999909</td>\n",
       "      <td>0.000091</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>2285</td>\n",
       "      <td>24 _ rozay _ idgt mad cuz Twan fuck his bitch MENTION ;   URL</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>0.999932</td>\n",
       "      <td>0.000068</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>496</td>\n",
       "      <td>quote I used to dress up like a cop too .  quote   quote Oh really ?  quote   quote Yeah .  .  . right after 9 / 11 .  .  . mad pussy .  quote   quote Really ?  quote   quote Yes .  .  . I ' ll NEVER FORGET .  quote</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>0.999640</td>\n",
       "      <td>0.000360</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>5757</td>\n",
       "      <td>MENTION Today was Talk Like A Pirate Day ,  and absolutely NOBODY knew Id been doing a Charlie Morton impersonation all day .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>0.999918</td>\n",
       "      <td>0.000082</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>21014</td>\n",
       "      <td>Running dangerously low on cigarettes these bitches better last me till Monday or Tuesday</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>0.999933</td>\n",
       "      <td>0.000067</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>15650</td>\n",
       "      <td>RT  MENTION :  bitches be like  quote I ' m a virgin quote   URL</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>0.999928</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>16003</td>\n",
       "      <td>RT  MENTION :  RT  MENTION :  Hispanic women just have the great pussy and wife skills ,  white girls own the dick sucking and black chicks .  .  .  . ca MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>0.999915</td>\n",
       "      <td>0.000085</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>15343</td>\n",
       "      <td>RT  MENTION :  fuck a bitch we never loved a thot</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>0.999931</td>\n",
       "      <td>0.000069</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>530</td>\n",
       "      <td>quote Lmao  MENTION :   MENTION ;  MENTION ;  RT  MENTION :  Pound cake look like pussy  MENTION ;   URL  quote</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>0.994382</td>\n",
       "      <td>0.005618</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>16457</td>\n",
       "      <td>RT  MENTION :  Sometimes I turn my phone on loud to tweet just to hear the bird chirp .   # TrueStory  # TeamiPhone</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>0.999927</td>\n",
       "      <td>0.000073</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>24671</td>\n",
       "      <td>lmaoooooo RT  MENTION :  Dude said bitches from NY fuck to Redman ' s greatest hits  MENTION ;  MENTION ;  MENTION ;</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",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2970</th>\n",
       "      <td>0.999900</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>18678</td>\n",
       "      <td>RT  MENTION :   MENTION ;  MENTION :  How I lose this sweet .  .  MENTION ; you ain ' t lose it you hide it lmfao Viri tryna smoke that hoe to his self</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2971</th>\n",
       "      <td>0.999906</td>\n",
       "      <td>0.000094</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>12041</td>\n",
       "      <td>Im not tongue kissing these bitches tho</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2972</th>\n",
       "      <td>0.999919</td>\n",
       "      <td>0.000081</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>17528</td>\n",
       "      <td>RT  MENTION :  Drunk inlove with my gun bae bitch Beyonc MENTION ;   MENTION ;  MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2973</th>\n",
       "      <td>0.999925</td>\n",
       "      <td>0.000075</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>19865</td>\n",
       "      <td>RT  MENTION :   MENTION :  when ur with ur friend and a bitch u both hate walks past  URL  quote   MENTION us as n MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2974</th>\n",
       "      <td>0.999926</td>\n",
       "      <td>0.000074</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>25042</td>\n",
       "      <td>there ' s 13 people in my house and they ' re all yelling pls god give me the strength not to beat a bitch .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2975</th>\n",
       "      <td>0.999929</td>\n",
       "      <td>0.000071</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>17334</td>\n",
       "      <td>RT  MENTION :  It ain nun to cut dat bitch awf  -  Verizon</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2976</th>\n",
       "      <td>0.999670</td>\n",
       "      <td>0.000330</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>23406</td>\n",
       "      <td>Will Hillary  MENTION ; dodo - bird MENTION ;  Be Dems 2016 Candidate To Take Over Obama ' s Job As US Prez ?  DoG  URL  via  MENTION</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2977</th>\n",
       "      <td>0.000086</td>\n",
       "      <td>0.999914</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>24984</td>\n",
       "      <td>stupid loose pussyed bitch</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2978</th>\n",
       "      <td>0.036333</td>\n",
       "      <td>0.963667</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>4903</td>\n",
       "      <td>MENTION  MENTION oh I see .  Part of a giant govt take over eh .  Stockboy you are a joke of a representative .   # dumb teabagger</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2979</th>\n",
       "      <td>0.999565</td>\n",
       "      <td>0.000435</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>20387</td>\n",
       "      <td>RT  MENTION :  Not .  A .  Fucking .  Charge .  Even .  As .  A .  Juvenile .  If you call  # MichaelBrown a thug at this point ,  just say  quote nigger quote  cuz it ' s  MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2980</th>\n",
       "      <td>0.999924</td>\n",
       "      <td>0.000076</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>8678</td>\n",
       "      <td>Chase the pussy ,  it ' s a sin .  Falls in ya lap ,  it ' s a blessing .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2981</th>\n",
       "      <td>0.999912</td>\n",
       "      <td>0.000088</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>19319</td>\n",
       "      <td>RT  MENTION :  She used to be more fun as a hoe  .  . this new you is to positive</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2982</th>\n",
       "      <td>0.000111</td>\n",
       "      <td>0.999889</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>3034</td>\n",
       "      <td>MENTION  MENTION get on cunt</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2983</th>\n",
       "      <td>0.999927</td>\n",
       "      <td>0.000073</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>12959</td>\n",
       "      <td>Look at y ' all bitches only Hmu when ya wifi on bitches</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2984</th>\n",
       "      <td>0.999925</td>\n",
       "      <td>0.000075</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>22222</td>\n",
       "      <td>These groupie bitches wild enough to suck a babies dick</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2985</th>\n",
       "      <td>0.999920</td>\n",
       "      <td>0.000080</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>22246</td>\n",
       "      <td>These hoes be actin up and these niccas be actin tough !   MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2986</th>\n",
       "      <td>0.999928</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>3553</td>\n",
       "      <td>MENTION i love seing new ones cause i save them and then youse them bitches lmfao</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2987</th>\n",
       "      <td>0.999884</td>\n",
       "      <td>0.000116</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>11272</td>\n",
       "      <td>I told that little bitch ass nigga ,   quote do it look like I ' m about to eat your gun charge ?  quote  I ' ll slap you and your hoe ass lawyer .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2988</th>\n",
       "      <td>0.999892</td>\n",
       "      <td>0.000108</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>7969</td>\n",
       "      <td>And now lookin ass niggas and lookin ass bitches will be the terms of 2014 .  Will be overused  &amp; amp ;  abused in every fucking way</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2989</th>\n",
       "      <td>0.998982</td>\n",
       "      <td>0.001018</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>16705</td>\n",
       "      <td>RT  MENTION :  I don ' t understand how so many Americans can hate on Obama for saluting with a latte .  Nobody tells these swagfags anything abo MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2990</th>\n",
       "      <td>0.999340</td>\n",
       "      <td>0.000660</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>21057</td>\n",
       "      <td>Same people bitchin about jetpacks and shit in AW are counting down the days until it comes out buncha fags .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2991</th>\n",
       "      <td>0.973312</td>\n",
       "      <td>0.026688</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>146</td>\n",
       "      <td>quote  MENTION :  Lets get drunk ,  get blowed ,  spit shit spark blunts  &amp; amp ;  fuck hoes !  quote  Bro white dudes have no chill .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2992</th>\n",
       "      <td>0.996065</td>\n",
       "      <td>0.003935</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>21902</td>\n",
       "      <td>That was trash .  I ' m just gonna put my phone down .  Goodnight</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2993</th>\n",
       "      <td>0.999734</td>\n",
       "      <td>0.000266</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>19708</td>\n",
       "      <td>RT  MENTION :  just seen this little yellow nigga who lives in a pineapple .  .  .  .  .  thats some crazy shit .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2994</th>\n",
       "      <td>0.999578</td>\n",
       "      <td>0.000422</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>8033</td>\n",
       "      <td>Arroz con pollo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2995</th>\n",
       "      <td>0.999926</td>\n",
       "      <td>0.000074</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>11238</td>\n",
       "      <td>I think I just gotta accept the fact that all these bitches gonna be hoes til they around the age of 27 and life will be a lot easier</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2996</th>\n",
       "      <td>0.999792</td>\n",
       "      <td>0.000208</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>17684</td>\n",
       "      <td>RT  MENTION :  Great start the morning !  My daughter danced herself into the coffee table and gave herself a nice shiner .  .  .  Yea she ' s a wh MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2997</th>\n",
       "      <td>0.999785</td>\n",
       "      <td>0.000215</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>5346</td>\n",
       "      <td>MENTION all the great QB ' s fold under pressure gshit if u blitz Peyton ass he gon gook out</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2998</th>\n",
       "      <td>0.999917</td>\n",
       "      <td>0.000083</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>12481</td>\n",
       "      <td>Just touch down in Jamaica tryna smoke two acres  .  Send a nigguh to his makerr</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2999</th>\n",
       "      <td>0.999926</td>\n",
       "      <td>0.000074</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>237</td>\n",
       "      <td>quote  MENTION :   MENTION ;  MENTION ;  MENTION ;  MENTION ; RT  MENTION :  Lmao !  This bitch come scooting in the tent like she about to fuck some shit up !  quote   MENTION ;  MENTION ;  MENTION ;</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3000 rows × 6 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      weight_correct    error  true_label  predicted     id  \\\n",
       "0    0.999834        0.000166  1           1          23437   \n",
       "1    0.999770        0.000230  1           1          17813   \n",
       "2    0.999914        0.000086  1           1          2561    \n",
       "3    0.999929        0.000071  1           1          18228   \n",
       "4    0.999911        0.000089  1           1          23030   \n",
       "5    0.999821        0.000179  1           1          16373   \n",
       "6    0.999915        0.000085  1           1          3560    \n",
       "7    0.999879        0.000121  1           1          2692    \n",
       "8    0.999902        0.000098  1           1          23549   \n",
       "9    0.999918        0.000082  1           1          12180   \n",
       "10   0.999523        0.000477  2           2          24918   \n",
       "11   0.999916        0.000084  1           1          16493   \n",
       "12   0.999183        0.000817  2           2          16532   \n",
       "13   0.999921        0.000079  1           1          15319   \n",
       "14   0.999922        0.000078  1           1          21838   \n",
       "15   0.999615        0.000385  2           2          3523    \n",
       "16   0.000198        0.999802  1           2          6008    \n",
       "17   0.013034        0.986966  1           2          1429    \n",
       "18   0.999925        0.000075  1           1          3240    \n",
       "19   0.999792        0.000208  2           2          21523   \n",
       "20   0.999909        0.000091  1           1          2285    \n",
       "21   0.999932        0.000068  1           1          496     \n",
       "22   0.999640        0.000360  2           2          5757    \n",
       "23   0.999918        0.000082  1           1          21014   \n",
       "24   0.999933        0.000067  1           1          15650   \n",
       "25   0.999928        0.000072  1           1          16003   \n",
       "26   0.999915        0.000085  1           1          15343   \n",
       "27   0.999931        0.000069  1           1          530     \n",
       "28   0.994382        0.005618  2           2          16457   \n",
       "29   0.999927        0.000073  1           1          24671   \n",
       "...       ...             ... ..          ..            ...   \n",
       "2970 0.999900        0.000100  1           1          18678   \n",
       "2971 0.999906        0.000094  1           1          12041   \n",
       "2972 0.999919        0.000081  1           1          17528   \n",
       "2973 0.999925        0.000075  1           1          19865   \n",
       "2974 0.999926        0.000074  1           1          25042   \n",
       "2975 0.999929        0.000071  1           1          17334   \n",
       "2976 0.999670        0.000330  2           2          23406   \n",
       "2977 0.000086        0.999914  0           1          24984   \n",
       "2978 0.036333        0.963667  1           2          4903    \n",
       "2979 0.999565        0.000435  1           1          20387   \n",
       "2980 0.999924        0.000076  1           1          8678    \n",
       "2981 0.999912        0.000088  1           1          19319   \n",
       "2982 0.000111        0.999889  0           1          3034    \n",
       "2983 0.999927        0.000073  1           1          12959   \n",
       "2984 0.999925        0.000075  1           1          22222   \n",
       "2985 0.999920        0.000080  1           1          22246   \n",
       "2986 0.999928        0.000072  1           1          3553    \n",
       "2987 0.999884        0.000116  1           1          11272   \n",
       "2988 0.999892        0.000108  1           1          7969    \n",
       "2989 0.998982        0.001018  1           1          16705   \n",
       "2990 0.999340        0.000660  1           1          21057   \n",
       "2991 0.973312        0.026688  1           1          146     \n",
       "2992 0.996065        0.003935  2           2          21902   \n",
       "2993 0.999734        0.000266  1           1          19708   \n",
       "2994 0.999578        0.000422  2           2          8033    \n",
       "2995 0.999926        0.000074  1           1          11238   \n",
       "2996 0.999792        0.000208  2           2          17684   \n",
       "2997 0.999785        0.000215  1           1          5346    \n",
       "2998 0.999917        0.000083  1           1          12481   \n",
       "2999 0.999926        0.000074  1           1          237     \n",
       "\n",
       "                                                                                                                                                                                                                           text  \n",
       "0     Wizard turned into a hoe after a break up                                                                                                                                                                                  \n",
       "1     RT  MENTION :  Captain Kirk hoe !                                                                                                                                                                                          \n",
       "2      MENTION bruh did go out like a bitch thoe lol that should not be his final match                                                                                                                                          \n",
       "3     RT  MENTION :  RT  MENTION :  when you tryna get bitches on LinkedIn  URL                                                                                                                                                  \n",
       "4     Well I wish I could shoot the shit wit u bitches but daddy gotta get dis money  . one                                                                                                                                      \n",
       "5     RT  MENTION :  Niggas can do slope intercept form equations but cant make scrambled fuckin eggs thats a lil backwards                                                                                                      \n",
       "6      MENTION He must think you bump pussy .  I no mind .  I ' ll eat your  # thong and make you  # squirt like the  # geyser .                                                                                                 \n",
       "7      MENTION always eatin good wit my niggah ,  and good lookin on the cd my boy got some heat now i can bump some bangers                                                                                                     \n",
       "8     Ya last nicca couldn ' t beat it  .  .  I hit it like a free throw !                                                                                                                                                       \n",
       "9     It ain ' t nothin to cut that bitch off                                                                                                                                                                                    \n",
       "10    she was telling me about this kid who ' s soles of his shoes are falling off .  I am going to get him some shoes .  hope his parents don ' t get mad                                                                       \n",
       "11    RT  MENTION :  smh u got Ebola now RT MENTION ;  MENTION :  I stick my foot out to feel some freshness and this bitch ass mosquito has the nerve to  MENTION ;                                                             \n",
       "12    RT  MENTION :  trash liars RT  MENTION :  these law students are terrible liars  # HTGAWM                                                                                                                                  \n",
       "13    RT  MENTION :  All hoes are friendly .  Althea wants to say hi .                                                                                                                                                           \n",
       "14    That bitch mouth made the honor roll                                                                                                                                                                                       \n",
       "15     MENTION I wonder if I have my very own NSA spook following me by now ?                                                                                                                                                    \n",
       "16     MENTION when you realize you ' re trash                                                                                                                                                                                   \n",
       "17     MENTION ;  MENTION :  I was told never fold  & amp ;  never sell my sole quote They Should Have Told Yu Stay In School  !   * Soul MENTION ;  autocorrect hoed me                                                         \n",
       "18     MENTION khaye I can ' t see your freakin emojis I ' m a basic bitch today .                                                                                                                                               \n",
       "19    Someone should explain to Blake griffin that talking trash to an opposing 3rd string is unbecoming ,  even amongst clippers .                                                                                              \n",
       "20    24 _ rozay _ idgt mad cuz Twan fuck his bitch MENTION ;   URL                                                                                                                                                              \n",
       "21     quote I used to dress up like a cop too .  quote   quote Oh really ?  quote   quote Yeah .  .  . right after 9 / 11 .  .  . mad pussy .  quote   quote Really ?  quote   quote Yes .  .  . I ' ll NEVER FORGET .  quote   \n",
       "22     MENTION Today was Talk Like A Pirate Day ,  and absolutely NOBODY knew Id been doing a Charlie Morton impersonation all day .                                                                                             \n",
       "23    Running dangerously low on cigarettes these bitches better last me till Monday or Tuesday                                                                                                                                  \n",
       "24    RT  MENTION :  bitches be like  quote I ' m a virgin quote   URL                                                                                                                                                           \n",
       "25    RT  MENTION :  RT  MENTION :  Hispanic women just have the great pussy and wife skills ,  white girls own the dick sucking and black chicks .  .  .  . ca MENTION ;                                                        \n",
       "26    RT  MENTION :  fuck a bitch we never loved a thot                                                                                                                                                                          \n",
       "27     quote Lmao  MENTION :   MENTION ;  MENTION ;  RT  MENTION :  Pound cake look like pussy  MENTION ;   URL  quote                                                                                                           \n",
       "28    RT  MENTION :  Sometimes I turn my phone on loud to tweet just to hear the bird chirp .   # TrueStory  # TeamiPhone                                                                                                        \n",
       "29    lmaoooooo RT  MENTION :  Dude said bitches from NY fuck to Redman ' s greatest hits  MENTION ;  MENTION ;  MENTION ;                                                                                                       \n",
       "...                                                                                                                     ...                                                                                                      \n",
       "2970  RT  MENTION :   MENTION ;  MENTION :  How I lose this sweet .  .  MENTION ; you ain ' t lose it you hide it lmfao Viri tryna smoke that hoe to his self                                                                    \n",
       "2971  Im not tongue kissing these bitches tho                                                                                                                                                                                    \n",
       "2972  RT  MENTION :  Drunk inlove with my gun bae bitch Beyonc MENTION ;   MENTION ;  MENTION ;                                                                                                                                  \n",
       "2973  RT  MENTION :   MENTION :  when ur with ur friend and a bitch u both hate walks past  URL  quote   MENTION us as n MENTION ;                                                                                               \n",
       "2974  there ' s 13 people in my house and they ' re all yelling pls god give me the strength not to beat a bitch .                                                                                                               \n",
       "2975  RT  MENTION :  It ain nun to cut dat bitch awf  -  Verizon                                                                                                                                                                 \n",
       "2976  Will Hillary  MENTION ; dodo - bird MENTION ;  Be Dems 2016 Candidate To Take Over Obama ' s Job As US Prez ?  DoG  URL  via  MENTION                                                                                      \n",
       "2977  stupid loose pussyed bitch                                                                                                                                                                                                 \n",
       "2978   MENTION  MENTION oh I see .  Part of a giant govt take over eh .  Stockboy you are a joke of a representative .   # dumb teabagger                                                                                        \n",
       "2979  RT  MENTION :  Not .  A .  Fucking .  Charge .  Even .  As .  A .  Juvenile .  If you call  # MichaelBrown a thug at this point ,  just say  quote nigger quote  cuz it ' s  MENTION ;                                     \n",
       "2980  Chase the pussy ,  it ' s a sin .  Falls in ya lap ,  it ' s a blessing .                                                                                                                                                  \n",
       "2981  RT  MENTION :  She used to be more fun as a hoe  .  . this new you is to positive                                                                                                                                          \n",
       "2982   MENTION  MENTION get on cunt                                                                                                                                                                                              \n",
       "2983  Look at y ' all bitches only Hmu when ya wifi on bitches                                                                                                                                                                   \n",
       "2984  These groupie bitches wild enough to suck a babies dick                                                                                                                                                                    \n",
       "2985  These hoes be actin up and these niccas be actin tough !   MENTION ;                                                                                                                                                       \n",
       "2986   MENTION i love seing new ones cause i save them and then youse them bitches lmfao                                                                                                                                         \n",
       "2987  I told that little bitch ass nigga ,   quote do it look like I ' m about to eat your gun charge ?  quote  I ' ll slap you and your hoe ass lawyer .                                                                        \n",
       "2988  And now lookin ass niggas and lookin ass bitches will be the terms of 2014 .  Will be overused  & amp ;  abused in every fucking way                                                                                       \n",
       "2989  RT  MENTION :  I don ' t understand how so many Americans can hate on Obama for saluting with a latte .  Nobody tells these swagfags anything abo MENTION ;                                                                \n",
       "2990  Same people bitchin about jetpacks and shit in AW are counting down the days until it comes out buncha fags .                                                                                                              \n",
       "2991   quote  MENTION :  Lets get drunk ,  get blowed ,  spit shit spark blunts  & amp ;  fuck hoes !  quote  Bro white dudes have no chill .                                                                                    \n",
       "2992  That was trash .  I ' m just gonna put my phone down .  Goodnight                                                                                                                                                          \n",
       "2993  RT  MENTION :  just seen this little yellow nigga who lives in a pineapple .  .  .  .  .  thats some crazy shit .                                                                                                          \n",
       "2994  Arroz con pollo                                                                                                                                                                                                            \n",
       "2995  I think I just gotta accept the fact that all these bitches gonna be hoes til they around the age of 27 and life will be a lot easier                                                                                      \n",
       "2996  RT  MENTION :  Great start the morning !  My daughter danced herself into the coffee table and gave herself a nice shiner .  .  .  Yea she ' s a wh MENTION ;                                                              \n",
       "2997   MENTION all the great QB ' s fold under pressure gshit if u blitz Peyton ass he gon gook out                                                                                                                              \n",
       "2998  Just touch down in Jamaica tryna smoke two acres  .  Send a nigguh to his makerr                                                                                                                                           \n",
       "2999   quote  MENTION :   MENTION ;  MENTION ;  MENTION ;  MENTION ; RT  MENTION :  Lmao !  This bitch come scooting in the tent like she about to fuck some shit up !  quote   MENTION ;  MENTION ;  MENTION ;                  \n",
       "\n",
       "[3000 rows x 6 columns]"
      ]
     },
     "execution_count": 421,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# onehot\n",
    "onehot = oneHot(true_labels)\n",
    "\n",
    "model = 5\n",
    "model_softmax = all[model]['probs']\n",
    "model_softmax = model_softmax.apply(pd.Series)\n",
    "\n",
    "comb = getErrors(model_softmax, onehot)\n",
    "comb\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 422,
   "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>weight_correct</th>\n",
       "      <th>error</th>\n",
       "      <th>true_label</th>\n",
       "      <th>predicted</th>\n",
       "      <th>id</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>931</th>\n",
       "      <td>0.000027</td>\n",
       "      <td>0.999973</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>15612</td>\n",
       "      <td>RT  MENTION :  Todays Horoscope :  you ' re still a hoe That is all .  have a nice day</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2947</th>\n",
       "      <td>0.000035</td>\n",
       "      <td>0.999965</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>8932</td>\n",
       "      <td>Define wigga</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2083</th>\n",
       "      <td>0.000038</td>\n",
       "      <td>0.999962</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>23076</td>\n",
       "      <td>What else come on on mondays im not finna watch that trash ,  ive destroyed enuf brain cells .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1459</th>\n",
       "      <td>0.000043</td>\n",
       "      <td>0.999957</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>9708</td>\n",
       "      <td>Giving you that trailer park trash .   # transformthursday  # ladykimora  # vegasqueens  # vegasshowgirls  URL</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>640</th>\n",
       "      <td>0.000044</td>\n",
       "      <td>0.999956</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>10498</td>\n",
       "      <td>I don ' t know what you been told but I love them project hoes</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2617</th>\n",
       "      <td>0.000049</td>\n",
       "      <td>0.999951</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>24158</td>\n",
       "      <td>bitches be like i ' m mixed wit Italian Chinese nd African American I be like bitch just cuz you ate ur MENTION ;   URL</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1082</th>\n",
       "      <td>0.000049</td>\n",
       "      <td>0.999951</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>5688</td>\n",
       "      <td>MENTION May I jump in u ball - less dick - wad ?  As a conservative ,  feminist ,  disabled vet bitch ,  I wish u luck with a capital F .   MENTION</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1132</th>\n",
       "      <td>0.000051</td>\n",
       "      <td>0.999949</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>19439</td>\n",
       "      <td>RT  MENTION :  There might be a couple cotton pickers on the spurs but they all play unselfish white man basketball</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1231</th>\n",
       "      <td>0.000059</td>\n",
       "      <td>0.999941</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>20120</td>\n",
       "      <td>RT  MENTION :  realest tweet i seen today RT  MENTION :   MENTION ;  .  Dyke bitches are rehab for hoes . And once im healed , back to the dick I go</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1651</th>\n",
       "      <td>0.000059</td>\n",
       "      <td>0.999941</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>14053</td>\n",
       "      <td>Or she just cant afford winter clothes RT  MENTION Summer hoe  =  winter Muslimah</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1992</th>\n",
       "      <td>0.000067</td>\n",
       "      <td>0.999933</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>18670</td>\n",
       "      <td>RT  MENTION :  DEAD !  !  MENTION ;  MENTION ;  MENTION :   MENTION ;  RT  MENTION :  When ur fuckin and her pussy stank but u have really good friends .  .  .  .  .  .  .  .  .  .  .  ht MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1345</th>\n",
       "      <td>0.000076</td>\n",
       "      <td>0.999924</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>22352</td>\n",
       "      <td>They slut for me they kill for me they steal for me and of course it ' ll be yo cash and I ' ll murder that bitch send the body back to yo ass</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2565</th>\n",
       "      <td>0.000078</td>\n",
       "      <td>0.999922</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>7026</td>\n",
       "      <td>MENTION yeah you right and you know them Mexicans bringing all they family that hoe finna be packed</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2555</th>\n",
       "      <td>0.000079</td>\n",
       "      <td>0.999921</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>9979</td>\n",
       "      <td>He ain ' t shit girl ,   MENTION ; he a bitch made nigga  !</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2881</th>\n",
       "      <td>0.000079</td>\n",
       "      <td>0.999921</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>10030</td>\n",
       "      <td>Heard that bitch gtta thang for them boss niggas</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2775</th>\n",
       "      <td>0.000079</td>\n",
       "      <td>0.999921</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>8679</td>\n",
       "      <td>Chasing these hoes ,  better get cash nigga</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2222</th>\n",
       "      <td>0.000080</td>\n",
       "      <td>0.999920</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>6177</td>\n",
       "      <td>MENTION I don ' t hate the Japs .  But ,  I AINT FORGETTING GETTIN JAPPED EITHER .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2369</th>\n",
       "      <td>0.000083</td>\n",
       "      <td>0.999917</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>17888</td>\n",
       "      <td>RT  MENTION :   MENTION  MENTION  MENTION  MENTION but cant nobody at yall pep rally out jigg me</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1860</th>\n",
       "      <td>0.000084</td>\n",
       "      <td>0.999916</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>21182</td>\n",
       "      <td>She got every niggah lookin like dam</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2977</th>\n",
       "      <td>0.000086</td>\n",
       "      <td>0.999914</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>24984</td>\n",
       "      <td>stupid loose pussyed bitch</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1545</th>\n",
       "      <td>0.000090</td>\n",
       "      <td>0.999910</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>9597</td>\n",
       "      <td>Fucking with these hoes nigga don ' t be no dumby</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1911</th>\n",
       "      <td>0.000091</td>\n",
       "      <td>0.999909</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>6643</td>\n",
       "      <td>MENTION lmao why ur dad raised u to be a bitch made nigga</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1960</th>\n",
       "      <td>0.000092</td>\n",
       "      <td>0.999908</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>3219</td>\n",
       "      <td>MENTION  MENTION you have a redneck attitude .  The same people that think Benghazzi is real think Fox fake nws is real .  Your one</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1278</th>\n",
       "      <td>0.000097</td>\n",
       "      <td>0.999903</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>15921</td>\n",
       "      <td>RT  MENTION :  Nigga screamed like a Lil bitch lol  URL</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1199</th>\n",
       "      <td>0.000098</td>\n",
       "      <td>0.999902</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>25067</td>\n",
       "      <td>they ' re working on a bill to prevent retards from voting .  who knew retards COULD vote ?  things are starting to make sense now .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1998</th>\n",
       "      <td>0.000100</td>\n",
       "      <td>0.999900</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>8080</td>\n",
       "      <td>Awk .  .  .  U don ' t like your own nephew  /  cousins bc our race .  .  LMAO broke ugly bitches fuck them we don ' t need them</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1239</th>\n",
       "      <td>0.000104</td>\n",
       "      <td>0.999896</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>14319</td>\n",
       "      <td>RT  quote  MENTION :   &amp; amp ;  alla my niggas hot boys  ,  so don ' t bring no ice in this bitch  MENTION ;  MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2405</th>\n",
       "      <td>0.000104</td>\n",
       "      <td>0.999896</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>12618</td>\n",
       "      <td>Lakers are trash right now .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>406</th>\n",
       "      <td>0.000105</td>\n",
       "      <td>0.999895</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>8564</td>\n",
       "      <td>Can MENTION ; t wait to humble these niggas in the playoffs again .  Until then ,  good job hoes</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1266</th>\n",
       "      <td>0.000106</td>\n",
       "      <td>0.999894</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>19899</td>\n",
       "      <td>RT  MENTION :  No mom .  Im not  quote dramatic  / immature quote  i just have fucking mental problems and youre a bitch and a cunt .</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",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2025</th>\n",
       "      <td>0.999934</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>15120</td>\n",
       "      <td>RT  MENTION :  Stop fuckin with a  quote take me out somewhere quote  type of bitch .  .  .  Get you a  quote baby I cooked for you quote  type of chick !  !   MENTION ;  MENTION ;  MENTION ;  MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1435</th>\n",
       "      <td>0.999934</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>19329</td>\n",
       "      <td>RT  MENTION :  I do know one thing tho ,  bitches they come  &amp; amp ;  go .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>664</th>\n",
       "      <td>0.999934</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>24115</td>\n",
       "      <td>bitch betta get it right .  .  . shout out .  .  .  . ddaaawwgggffaaaccee cccaalldddeerooonn .  .  .  .  .  . dese hoes be aktin .  .  .   URL</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>272</th>\n",
       "      <td>0.999934</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>6220</td>\n",
       "      <td>MENTION  MENTION no bitch lol</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>626</th>\n",
       "      <td>0.999934</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1014</td>\n",
       "      <td>MENTION ;  MENTION ;  MENTION ;  MENTION ;   MENTION bitch was mad af  URL</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1910</th>\n",
       "      <td>0.999934</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>5372</td>\n",
       "      <td>MENTION bitches better watch out because we ' re running the tables tonight !   MENTION ;  MENTION ;  MENTION ;  MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1121</th>\n",
       "      <td>0.999934</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>20001</td>\n",
       "      <td>RT  MENTION :  im not  quote shy quote  or  quote quiet quote  im just chilling bitch</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2379</th>\n",
       "      <td>0.999934</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1158</td>\n",
       "      <td>MENTION ;  MENTION :  Bitches love throwing that irrelevant shade  MENTION ;  MENTION ;  go sit lil bitch  .  MENTION ;  MENTION ;  MENTION ;  MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>232</th>\n",
       "      <td>0.999934</td>\n",
       "      <td>0.000066</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>24455</td>\n",
       "      <td>URL  Evilyn Fierce  MENTION ;  Naughty girl next door pummels her pussy with a purple MENTION ; toy  URL</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2404</th>\n",
       "      <td>0.999935</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>52</td>\n",
       "      <td>quote  you dodge a bullet  quote   MENTION ;   MENTION ;  MENTION :   quote All da bitches I cut off pregnant or bound to be  .  .  .  . thank God  MENTION ;  quote  MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2794</th>\n",
       "      <td>0.999935</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>16153</td>\n",
       "      <td>RT  MENTION :  If a hoe give my lil Sista a eye like that she betta call me so we can go jump the bitch  MENTION ;  MENTION ;  MENTION ;  MENTION ;  MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1846</th>\n",
       "      <td>0.999935</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>16515</td>\n",
       "      <td>RT  MENTION :  I dive in tha pussy like .  .  .   MENTION ;  MENTION ;  MENTION ;  MENTION ;  MENTION ;  MENTION ;   URL</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2003</th>\n",
       "      <td>0.999935</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>20439</td>\n",
       "      <td>RT  MENTION  quote  MENTION :  I almost ran this fine red bitch over .  quote  Damn  +  lol</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1837</th>\n",
       "      <td>0.999935</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1980</td>\n",
       "      <td>MENTION ;  MENTION :   MENTION ;  MENTION :   MENTION  MENTION ;  nah lemme chill i love the bitches MENTION ;  MENTION ;  MENTION ;  like trick love the kids MENTION ;   &amp; amp ;  you know this  MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>956</th>\n",
       "      <td>0.999935</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>16953</td>\n",
       "      <td>RT  MENTION :   MENTION ;  MENTION :   MENTION it ' s a sleepover ?  Can I come robe ?  MENTION ;  Only if that bitch silk .  .  .   MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>249</th>\n",
       "      <td>0.999935</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1934</td>\n",
       "      <td>MENTION ;  MENTION :   MENTION ;  MENTION :  Dat bird bitch MENTION ; HEAVY BIRDCAGE BITCH MENTION ;  byyyyyyrrrrrrrrrrrrr bitch  MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>887</th>\n",
       "      <td>0.999935</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>14636</td>\n",
       "      <td>RT  MENTION :   MENTION this bitch think look like Rihanna  ?   MENTION ;  MENTION ;  MENTION ;  MENTION ;  MENTION ;  MENTION ;   URL</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1491</th>\n",
       "      <td>0.999935</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>20548</td>\n",
       "      <td>RT  MENTION :  Bitches tweets be like .  .  .   quote I wanna suck some dick  MENTION ;  quote  then be like .  .  .   quote  MENTION ;  Chill it ' s just a song !   quote  bitch what song is MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1485</th>\n",
       "      <td>0.999935</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>19017</td>\n",
       "      <td>RT  MENTION :   quote Drop that bitch quote   -  MENTION on the atomic bomb</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2004</th>\n",
       "      <td>0.999935</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>14463</td>\n",
       "      <td>RT  MENTION :   MENTION ;  MENTION ;  MENTION ;  RT  MENTION :  Lmfao  quote  MENTION :  How side bitches be when you scoop them up in the daytime !   URL</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2789</th>\n",
       "      <td>0.999935</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>5511</td>\n",
       "      <td>MENTION  MENTION  MENTION I ' m lmao fa real cuz I really saw her throw dat bitch in a circle lol</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>960</th>\n",
       "      <td>0.999936</td>\n",
       "      <td>0.000064</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>25186</td>\n",
       "      <td>when nino said  quote cancel that bitch ill buy a new one quote   &amp; gt ;  &amp; gt ;  &amp; gt ;  &amp; gt ;  &amp; gt ;  &amp; gt ;  &amp; gt ;  &amp; gt ;  &amp; gt ;  &amp; gt ;  &amp; gt ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1042</th>\n",
       "      <td>0.999936</td>\n",
       "      <td>0.000064</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1100</td>\n",
       "      <td>MENTION ;  MENTION :   MENTION  MENTION ;  MENTION ;  a bitch is all ears .  .  .  MENTION ;  My mama told me not to talk to bitches so .  .  .  MENTION ;  MENTION ; jk</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2707</th>\n",
       "      <td>0.999936</td>\n",
       "      <td>0.000064</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>19644</td>\n",
       "      <td>RT  MENTION :  Yall bitches don ' t know how to be friends  MENTION ;  MENTION ;   URL</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2202</th>\n",
       "      <td>0.999936</td>\n",
       "      <td>0.000064</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>17556</td>\n",
       "      <td>RT  MENTION :  3 point on a bitch like Dirk Nowitzki ,  up in da club like Dirk Nowitzki .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1682</th>\n",
       "      <td>0.999936</td>\n",
       "      <td>0.000064</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>14627</td>\n",
       "      <td>RT  MENTION :  really bitch ?   URL</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>832</th>\n",
       "      <td>0.999937</td>\n",
       "      <td>0.000063</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1225</td>\n",
       "      <td>MENTION ;  MENTION :  bitch that ' s ham on her face  URL  # 8221 ;   MENTION</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1315</th>\n",
       "      <td>0.999937</td>\n",
       "      <td>0.000063</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1674</td>\n",
       "      <td>MENTION ;  MENTION :   MENTION ;  MENTION :  Kyrie fast than a bitch with the ball MENTION ;  look like you out there lol MENTION ; lol I wish</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2427</th>\n",
       "      <td>0.999937</td>\n",
       "      <td>0.000063</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>71</td>\n",
       "      <td>quote  MENTION :  Hennessey Venom GT  MENTION ;   URL  quote  that ' s one sexy bitch</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>974</th>\n",
       "      <td>0.999938</td>\n",
       "      <td>0.000062</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>24454</td>\n",
       "      <td>URL  Crystal Star  MENTION ;  Leggy and playful babe Crystal Star spreads her pussy for MENTION ; you  URL</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3000 rows × 6 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      weight_correct    error  true_label  predicted     id  \\\n",
       "931  0.000027        0.999973  2           1          15612   \n",
       "2947 0.000035        0.999965  2           1          8932    \n",
       "2083 0.000038        0.999962  2           1          23076   \n",
       "1459 0.000043        0.999957  0           2          9708    \n",
       "640  0.000044        0.999956  2           1          10498   \n",
       "2617 0.000049        0.999951  0           1          24158   \n",
       "1082 0.000049        0.999951  0           1          5688    \n",
       "1132 0.000051        0.999949  0           2          19439   \n",
       "1231 0.000059        0.999941  0           1          20120   \n",
       "1651 0.000059        0.999941  2           1          14053   \n",
       "1992 0.000067        0.999933  0           1          18670   \n",
       "1345 0.000076        0.999924  0           1          22352   \n",
       "2565 0.000078        0.999922  0           1          7026    \n",
       "2555 0.000079        0.999921  0           1          9979    \n",
       "2881 0.000079        0.999921  0           1          10030   \n",
       "2775 0.000079        0.999921  0           1          8679    \n",
       "2222 0.000080        0.999920  0           1          6177    \n",
       "2369 0.000083        0.999917  2           1          17888   \n",
       "1860 0.000084        0.999916  0           1          21182   \n",
       "2977 0.000086        0.999914  0           1          24984   \n",
       "1545 0.000090        0.999910  0           1          9597    \n",
       "1911 0.000091        0.999909  0           1          6643    \n",
       "1960 0.000092        0.999908  0           2          3219    \n",
       "1278 0.000097        0.999903  0           1          15921   \n",
       "1199 0.000098        0.999902  0           1          25067   \n",
       "1998 0.000100        0.999900  0           1          8080    \n",
       "1239 0.000104        0.999896  0           1          14319   \n",
       "2405 0.000104        0.999896  0           2          12618   \n",
       "406  0.000105        0.999895  0           1          8564    \n",
       "1266 0.000106        0.999894  0           1          19899   \n",
       "...       ...             ... ..          ..            ...   \n",
       "2025 0.999934        0.000066  1           1          15120   \n",
       "1435 0.999934        0.000066  1           1          19329   \n",
       "664  0.999934        0.000066  1           1          24115   \n",
       "272  0.999934        0.000066  1           1          6220    \n",
       "626  0.999934        0.000066  1           1          1014    \n",
       "1910 0.999934        0.000066  1           1          5372    \n",
       "1121 0.999934        0.000066  1           1          20001   \n",
       "2379 0.999934        0.000066  1           1          1158    \n",
       "232  0.999934        0.000066  1           1          24455   \n",
       "2404 0.999935        0.000065  1           1          52      \n",
       "2794 0.999935        0.000065  1           1          16153   \n",
       "1846 0.999935        0.000065  1           1          16515   \n",
       "2003 0.999935        0.000065  1           1          20439   \n",
       "1837 0.999935        0.000065  1           1          1980    \n",
       "956  0.999935        0.000065  1           1          16953   \n",
       "249  0.999935        0.000065  1           1          1934    \n",
       "887  0.999935        0.000065  1           1          14636   \n",
       "1491 0.999935        0.000065  1           1          20548   \n",
       "1485 0.999935        0.000065  1           1          19017   \n",
       "2004 0.999935        0.000065  1           1          14463   \n",
       "2789 0.999935        0.000065  1           1          5511    \n",
       "960  0.999936        0.000064  1           1          25186   \n",
       "1042 0.999936        0.000064  1           1          1100    \n",
       "2707 0.999936        0.000064  1           1          19644   \n",
       "2202 0.999936        0.000064  1           1          17556   \n",
       "1682 0.999936        0.000064  1           1          14627   \n",
       "832  0.999937        0.000063  1           1          1225    \n",
       "1315 0.999937        0.000063  1           1          1674    \n",
       "2427 0.999937        0.000063  1           1          71      \n",
       "974  0.999938        0.000062  1           1          24454   \n",
       "\n",
       "                                                                                                                                                                                                                text  \n",
       "931   RT  MENTION :  Todays Horoscope :  you ' re still a hoe That is all .  have a nice day                                                                                                                          \n",
       "2947  Define wigga                                                                                                                                                                                                    \n",
       "2083  What else come on on mondays im not finna watch that trash ,  ive destroyed enuf brain cells .                                                                                                                  \n",
       "1459  Giving you that trailer park trash .   # transformthursday  # ladykimora  # vegasqueens  # vegasshowgirls  URL                                                                                                  \n",
       "640   I don ' t know what you been told but I love them project hoes                                                                                                                                                  \n",
       "2617  bitches be like i ' m mixed wit Italian Chinese nd African American I be like bitch just cuz you ate ur MENTION ;   URL                                                                                         \n",
       "1082   MENTION May I jump in u ball - less dick - wad ?  As a conservative ,  feminist ,  disabled vet bitch ,  I wish u luck with a capital F .   MENTION                                                            \n",
       "1132  RT  MENTION :  There might be a couple cotton pickers on the spurs but they all play unselfish white man basketball                                                                                             \n",
       "1231  RT  MENTION :  realest tweet i seen today RT  MENTION :   MENTION ;  .  Dyke bitches are rehab for hoes . And once im healed , back to the dick I go                                                            \n",
       "1651  Or she just cant afford winter clothes RT  MENTION Summer hoe  =  winter Muslimah                                                                                                                               \n",
       "1992  RT  MENTION :  DEAD !  !  MENTION ;  MENTION ;  MENTION :   MENTION ;  RT  MENTION :  When ur fuckin and her pussy stank but u have really good friends .  .  .  .  .  .  .  .  .  .  .  ht MENTION ;           \n",
       "1345  They slut for me they kill for me they steal for me and of course it ' ll be yo cash and I ' ll murder that bitch send the body back to yo ass                                                                  \n",
       "2565   MENTION yeah you right and you know them Mexicans bringing all they family that hoe finna be packed                                                                                                            \n",
       "2555  He ain ' t shit girl ,   MENTION ; he a bitch made nigga  !                                                                                                                                                     \n",
       "2881  Heard that bitch gtta thang for them boss niggas                                                                                                                                                                \n",
       "2775  Chasing these hoes ,  better get cash nigga                                                                                                                                                                     \n",
       "2222   MENTION I don ' t hate the Japs .  But ,  I AINT FORGETTING GETTIN JAPPED EITHER .                                                                                                                             \n",
       "2369  RT  MENTION :   MENTION  MENTION  MENTION  MENTION but cant nobody at yall pep rally out jigg me                                                                                                                \n",
       "1860  She got every niggah lookin like dam                                                                                                                                                                            \n",
       "2977  stupid loose pussyed bitch                                                                                                                                                                                      \n",
       "1545  Fucking with these hoes nigga don ' t be no dumby                                                                                                                                                               \n",
       "1911   MENTION lmao why ur dad raised u to be a bitch made nigga                                                                                                                                                      \n",
       "1960   MENTION  MENTION you have a redneck attitude .  The same people that think Benghazzi is real think Fox fake nws is real .  Your one                                                                            \n",
       "1278  RT  MENTION :  Nigga screamed like a Lil bitch lol  URL                                                                                                                                                         \n",
       "1199  they ' re working on a bill to prevent retards from voting .  who knew retards COULD vote ?  things are starting to make sense now .                                                                            \n",
       "1998  Awk .  .  .  U don ' t like your own nephew  /  cousins bc our race .  .  LMAO broke ugly bitches fuck them we don ' t need them                                                                                \n",
       "1239  RT  quote  MENTION :   & amp ;  alla my niggas hot boys  ,  so don ' t bring no ice in this bitch  MENTION ;  MENTION ;                                                                                         \n",
       "2405  Lakers are trash right now .                                                                                                                                                                                    \n",
       "406   Can MENTION ; t wait to humble these niggas in the playoffs again .  Until then ,  good job hoes                                                                                                                \n",
       "1266  RT  MENTION :  No mom .  Im not  quote dramatic  / immature quote  i just have fucking mental problems and youre a bitch and a cunt .                                                                           \n",
       "...                                                                                                                                      ...                                                                          \n",
       "2025  RT  MENTION :  Stop fuckin with a  quote take me out somewhere quote  type of bitch .  .  .  Get you a  quote baby I cooked for you quote  type of chick !  !   MENTION ;  MENTION ;  MENTION ;  MENTION ;      \n",
       "1435  RT  MENTION :  I do know one thing tho ,  bitches they come  & amp ;  go .                                                                                                                                      \n",
       "664   bitch betta get it right .  .  . shout out .  .  .  . ddaaawwgggffaaaccee cccaalldddeerooonn .  .  .  .  .  . dese hoes be aktin .  .  .   URL                                                                  \n",
       "272    MENTION  MENTION no bitch lol                                                                                                                                                                                  \n",
       "626    MENTION ;  MENTION ;  MENTION ;  MENTION ;   MENTION bitch was mad af  URL                                                                                                                                     \n",
       "1910   MENTION bitches better watch out because we ' re running the tables tonight !   MENTION ;  MENTION ;  MENTION ;  MENTION ;                                                                                     \n",
       "1121  RT  MENTION :  im not  quote shy quote  or  quote quiet quote  im just chilling bitch                                                                                                                           \n",
       "2379   MENTION ;  MENTION :  Bitches love throwing that irrelevant shade  MENTION ;  MENTION ;  go sit lil bitch  .  MENTION ;  MENTION ;  MENTION ;  MENTION ;                                                       \n",
       "232    URL  Evilyn Fierce  MENTION ;  Naughty girl next door pummels her pussy with a purple MENTION ; toy  URL                                                                                                       \n",
       "2404   quote  you dodge a bullet  quote   MENTION ;   MENTION ;  MENTION :   quote All da bitches I cut off pregnant or bound to be  .  .  .  . thank God  MENTION ;  quote  MENTION ;                                \n",
       "2794  RT  MENTION :  If a hoe give my lil Sista a eye like that she betta call me so we can go jump the bitch  MENTION ;  MENTION ;  MENTION ;  MENTION ;  MENTION ;                                                  \n",
       "1846  RT  MENTION :  I dive in tha pussy like .  .  .   MENTION ;  MENTION ;  MENTION ;  MENTION ;  MENTION ;  MENTION ;   URL                                                                                        \n",
       "2003  RT  MENTION  quote  MENTION :  I almost ran this fine red bitch over .  quote  Damn  +  lol                                                                                                                     \n",
       "1837   MENTION ;  MENTION :   MENTION ;  MENTION :   MENTION  MENTION ;  nah lemme chill i love the bitches MENTION ;  MENTION ;  MENTION ;  like trick love the kids MENTION ;   & amp ;  you know this  MENTION ;   \n",
       "956   RT  MENTION :   MENTION ;  MENTION :   MENTION it ' s a sleepover ?  Can I come robe ?  MENTION ;  Only if that bitch silk .  .  .   MENTION ;                                                                  \n",
       "249    MENTION ;  MENTION :   MENTION ;  MENTION :  Dat bird bitch MENTION ; HEAVY BIRDCAGE BITCH MENTION ;  byyyyyyrrrrrrrrrrrrr bitch  MENTION ;                                                                    \n",
       "887   RT  MENTION :   MENTION this bitch think look like Rihanna  ?   MENTION ;  MENTION ;  MENTION ;  MENTION ;  MENTION ;  MENTION ;   URL                                                                          \n",
       "1491  RT  MENTION :  Bitches tweets be like .  .  .   quote I wanna suck some dick  MENTION ;  quote  then be like .  .  .   quote  MENTION ;  Chill it ' s just a song !   quote  bitch what song is MENTION ;       \n",
       "1485  RT  MENTION :   quote Drop that bitch quote   -  MENTION on the atomic bomb                                                                                                                                     \n",
       "2004  RT  MENTION :   MENTION ;  MENTION ;  MENTION ;  RT  MENTION :  Lmfao  quote  MENTION :  How side bitches be when you scoop them up in the daytime !   URL                                                      \n",
       "2789   MENTION  MENTION  MENTION I ' m lmao fa real cuz I really saw her throw dat bitch in a circle lol                                                                                                              \n",
       "960   when nino said  quote cancel that bitch ill buy a new one quote   & gt ;  & gt ;  & gt ;  & gt ;  & gt ;  & gt ;  & gt ;  & gt ;  & gt ;  & gt ;  & gt ;                                                        \n",
       "1042   MENTION ;  MENTION :   MENTION  MENTION ;  MENTION ;  a bitch is all ears .  .  .  MENTION ;  My mama told me not to talk to bitches so .  .  .  MENTION ;  MENTION ; jk                                       \n",
       "2707  RT  MENTION :  Yall bitches don ' t know how to be friends  MENTION ;  MENTION ;   URL                                                                                                                          \n",
       "2202  RT  MENTION :  3 point on a bitch like Dirk Nowitzki ,  up in da club like Dirk Nowitzki .                                                                                                                      \n",
       "1682  RT  MENTION :  really bitch ?   URL                                                                                                                                                                             \n",
       "832    MENTION ;  MENTION :  bitch that ' s ham on her face  URL  # 8221 ;   MENTION                                                                                                                                  \n",
       "1315   MENTION ;  MENTION :   MENTION ;  MENTION :  Kyrie fast than a bitch with the ball MENTION ;  look like you out there lol MENTION ; lol I wish                                                                 \n",
       "2427   quote  MENTION :  Hennessey Venom GT  MENTION ;   URL  quote  that ' s one sexy bitch                                                                                                                          \n",
       "974    URL  Crystal Star  MENTION ;  Leggy and playful babe Crystal Star spreads her pussy for MENTION ; you  URL                                                                                                     \n",
       "\n",
       "[3000 rows x 6 columns]"
      ]
     },
     "execution_count": 422,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 0 - hate speech\n",
    "# 1 - offensive language\n",
    "# 2 - neither\n",
    "\n",
    "sorted = comb.sort_values(by=['error'], ascending=False)\n",
    "sorted"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 409,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'univ3_tsize_18783'"
      ]
     },
     "execution_count": 409,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# predicted probabilities for model X\n",
    "# dataframe of softmax values, one row per item, cols are classes\n",
    "model = 5\n",
    "model_softmax = all[model]['probs']\n",
    "model_softmax = model_softmax.apply(pd.Series)\n",
    "\n",
    "# onehot\n",
    "onehot = oneHot(true_labels)\n",
    "\n",
    "combined = model_softmax * onehot\n",
    "# sum 3 classes to one\n",
    "comb = combined.sum(axis=1)\n",
    "\n",
    "# error is the amount how far predicted weight for real class was. if given 1 prob to real class error is 0.\n",
    "comb = pd.DataFrame(comb)\n",
    "comb.columns=['weight_correct']\n",
    "comb['error'] = 1 - comb['weight_correct']\n",
    "comb['true_label']=true_labels\n",
    "comb['predicted'] = predicted_labels\n",
    "#comb.head()\n",
    "\n",
    "\n",
    "predicted_labels = labels.iloc[:,5]\n",
    "all[5]['model']\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 408,
   "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>weight_correct</th>\n",
       "      <th>error</th>\n",
       "      <th>true_label</th>\n",
       "      <th>predicted</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.999834</td>\n",
       "      <td>0.000166</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.999770</td>\n",
       "      <td>0.000230</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.999914</td>\n",
       "      <td>0.000086</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.999929</td>\n",
       "      <td>0.000071</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.999911</td>\n",
       "      <td>0.000089</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.999821</td>\n",
       "      <td>0.000179</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.999915</td>\n",
       "      <td>0.000085</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0.999879</td>\n",
       "      <td>0.000121</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0.999902</td>\n",
       "      <td>0.000098</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>0.999918</td>\n",
       "      <td>0.000082</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.999523</td>\n",
       "      <td>0.000477</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>0.999916</td>\n",
       "      <td>0.000084</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>0.999183</td>\n",
       "      <td>0.000817</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>0.999921</td>\n",
       "      <td>0.000079</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>0.999922</td>\n",
       "      <td>0.000078</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>0.999615</td>\n",
       "      <td>0.000385</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>0.000198</td>\n",
       "      <td>0.999802</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>0.013034</td>\n",
       "      <td>0.986966</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>0.999925</td>\n",
       "      <td>0.000075</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>0.999792</td>\n",
       "      <td>0.000208</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>0.999909</td>\n",
       "      <td>0.000091</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>0.999932</td>\n",
       "      <td>0.000068</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>0.999640</td>\n",
       "      <td>0.000360</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>0.999918</td>\n",
       "      <td>0.000082</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>0.999933</td>\n",
       "      <td>0.000067</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>0.999928</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>0.999915</td>\n",
       "      <td>0.000085</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>0.999931</td>\n",
       "      <td>0.000069</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>0.994382</td>\n",
       "      <td>0.005618</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>0.999927</td>\n",
       "      <td>0.000073</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2970</th>\n",
       "      <td>0.999900</td>\n",
       "      <td>0.000100</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2971</th>\n",
       "      <td>0.999906</td>\n",
       "      <td>0.000094</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2972</th>\n",
       "      <td>0.999919</td>\n",
       "      <td>0.000081</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2973</th>\n",
       "      <td>0.999925</td>\n",
       "      <td>0.000075</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2974</th>\n",
       "      <td>0.999926</td>\n",
       "      <td>0.000074</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2975</th>\n",
       "      <td>0.999929</td>\n",
       "      <td>0.000071</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2976</th>\n",
       "      <td>0.999670</td>\n",
       "      <td>0.000330</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2977</th>\n",
       "      <td>0.000086</td>\n",
       "      <td>0.999914</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2978</th>\n",
       "      <td>0.036333</td>\n",
       "      <td>0.963667</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2979</th>\n",
       "      <td>0.999565</td>\n",
       "      <td>0.000435</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2980</th>\n",
       "      <td>0.999924</td>\n",
       "      <td>0.000076</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2981</th>\n",
       "      <td>0.999912</td>\n",
       "      <td>0.000088</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2982</th>\n",
       "      <td>0.000111</td>\n",
       "      <td>0.999889</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2983</th>\n",
       "      <td>0.999927</td>\n",
       "      <td>0.000073</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2984</th>\n",
       "      <td>0.999925</td>\n",
       "      <td>0.000075</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2985</th>\n",
       "      <td>0.999920</td>\n",
       "      <td>0.000080</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2986</th>\n",
       "      <td>0.999928</td>\n",
       "      <td>0.000072</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2987</th>\n",
       "      <td>0.999884</td>\n",
       "      <td>0.000116</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2988</th>\n",
       "      <td>0.999892</td>\n",
       "      <td>0.000108</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2989</th>\n",
       "      <td>0.998982</td>\n",
       "      <td>0.001018</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2990</th>\n",
       "      <td>0.999340</td>\n",
       "      <td>0.000660</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2991</th>\n",
       "      <td>0.973312</td>\n",
       "      <td>0.026688</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2992</th>\n",
       "      <td>0.996065</td>\n",
       "      <td>0.003935</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2993</th>\n",
       "      <td>0.999734</td>\n",
       "      <td>0.000266</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2994</th>\n",
       "      <td>0.999578</td>\n",
       "      <td>0.000422</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2995</th>\n",
       "      <td>0.999926</td>\n",
       "      <td>0.000074</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2996</th>\n",
       "      <td>0.999792</td>\n",
       "      <td>0.000208</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2997</th>\n",
       "      <td>0.999785</td>\n",
       "      <td>0.000215</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2998</th>\n",
       "      <td>0.999917</td>\n",
       "      <td>0.000083</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2999</th>\n",
       "      <td>0.999926</td>\n",
       "      <td>0.000074</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3000 rows × 4 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      weight_correct    error  true_label  predicted\n",
       "0    0.999834        0.000166  1           1        \n",
       "1    0.999770        0.000230  1           1        \n",
       "2    0.999914        0.000086  1           1        \n",
       "3    0.999929        0.000071  1           1        \n",
       "4    0.999911        0.000089  1           1        \n",
       "5    0.999821        0.000179  1           1        \n",
       "6    0.999915        0.000085  1           1        \n",
       "7    0.999879        0.000121  1           1        \n",
       "8    0.999902        0.000098  1           1        \n",
       "9    0.999918        0.000082  1           1        \n",
       "10   0.999523        0.000477  2           2        \n",
       "11   0.999916        0.000084  1           1        \n",
       "12   0.999183        0.000817  2           2        \n",
       "13   0.999921        0.000079  1           1        \n",
       "14   0.999922        0.000078  1           1        \n",
       "15   0.999615        0.000385  2           2        \n",
       "16   0.000198        0.999802  1           2        \n",
       "17   0.013034        0.986966  1           2        \n",
       "18   0.999925        0.000075  1           1        \n",
       "19   0.999792        0.000208  2           2        \n",
       "20   0.999909        0.000091  1           1        \n",
       "21   0.999932        0.000068  1           1        \n",
       "22   0.999640        0.000360  2           2        \n",
       "23   0.999918        0.000082  1           1        \n",
       "24   0.999933        0.000067  1           1        \n",
       "25   0.999928        0.000072  1           1        \n",
       "26   0.999915        0.000085  1           1        \n",
       "27   0.999931        0.000069  1           1        \n",
       "28   0.994382        0.005618  2           2        \n",
       "29   0.999927        0.000073  1           1        \n",
       "...       ...             ... ..          ..        \n",
       "2970 0.999900        0.000100  1           1        \n",
       "2971 0.999906        0.000094  1           1        \n",
       "2972 0.999919        0.000081  1           1        \n",
       "2973 0.999925        0.000075  1           1        \n",
       "2974 0.999926        0.000074  1           1        \n",
       "2975 0.999929        0.000071  1           1        \n",
       "2976 0.999670        0.000330  2           2        \n",
       "2977 0.000086        0.999914  0           1        \n",
       "2978 0.036333        0.963667  1           2        \n",
       "2979 0.999565        0.000435  1           1        \n",
       "2980 0.999924        0.000076  1           1        \n",
       "2981 0.999912        0.000088  1           1        \n",
       "2982 0.000111        0.999889  0           1        \n",
       "2983 0.999927        0.000073  1           1        \n",
       "2984 0.999925        0.000075  1           1        \n",
       "2985 0.999920        0.000080  1           1        \n",
       "2986 0.999928        0.000072  1           1        \n",
       "2987 0.999884        0.000116  1           1        \n",
       "2988 0.999892        0.000108  1           1        \n",
       "2989 0.998982        0.001018  1           1        \n",
       "2990 0.999340        0.000660  1           1        \n",
       "2991 0.973312        0.026688  1           1        \n",
       "2992 0.996065        0.003935  2           2        \n",
       "2993 0.999734        0.000266  1           1        \n",
       "2994 0.999578        0.000422  2           2        \n",
       "2995 0.999926        0.000074  1           1        \n",
       "2996 0.999792        0.000208  2           2        \n",
       "2997 0.999785        0.000215  1           1        \n",
       "2998 0.999917        0.000083  1           1        \n",
       "2999 0.999926        0.000074  1           1        \n",
       "\n",
       "[3000 rows x 4 columns]"
      ]
     },
     "execution_count": 408,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "comb"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 390,
   "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>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.999919</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.997979</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.998711</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.999517</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.999968</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         0        1        2\n",
       "0 0.000000 0.999919 0.000000\n",
       "1 0.000000 0.997979 0.000000\n",
       "2 0.000000 0.998711 0.000000\n",
       "3 0.000000 0.999517 0.000000\n",
       "4 0.000000 0.999968 0.000000"
      ]
     },
     "execution_count": 390,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 401,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0   0.999919\n",
       "1   0.997979\n",
       "2   0.998711\n",
       "3   0.999517\n",
       "4   0.999968\n",
       "dtype: float64"
      ]
     },
     "execution_count": 401,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 402,
   "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>weight_correct</th>\n",
       "      <th>error</th>\n",
       "      <th>true_label</th>\n",
       "      <th>predicted</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.999919</td>\n",
       "      <td>0.000081</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.997979</td>\n",
       "      <td>0.002021</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.998711</td>\n",
       "      <td>0.001289</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.999517</td>\n",
       "      <td>0.000483</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.999968</td>\n",
       "      <td>0.000032</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   weight_correct    error  true_label  predicted\n",
       "0 0.999919        0.000081  1           1        \n",
       "1 0.997979        0.002021  1           1        \n",
       "2 0.998711        0.001289  1           1        \n",
       "3 0.999517        0.000483  1           1        \n",
       "4 0.999968        0.000032  1           1        "
      ]
     },
     "execution_count": 402,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# error is the amount how far predicted weight for real class was. if given 1 prob to real class error is 0.\n",
    "comb = pd.DataFrame(comb)\n",
    "comb.columns=['weight_correct']\n",
    "comb['error'] = 1 - comb['weight_correct']\n",
    "comb['true_label']=true_labels\n",
    "comb['predicted'] = predicted_labels\n",
    "comb.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 403,
   "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>weight_correct</th>\n",
       "      <th>error</th>\n",
       "      <th>true_label</th>\n",
       "      <th>predicted</th>\n",
       "      <th>id</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.999919</td>\n",
       "      <td>0.000081</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>23437</td>\n",
       "      <td>Wizard turned into a hoe after a break up</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.997979</td>\n",
       "      <td>0.002021</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>17813</td>\n",
       "      <td>RT  MENTION :  Captain Kirk hoe !</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.998711</td>\n",
       "      <td>0.001289</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>2561</td>\n",
       "      <td>MENTION bruh did go out like a bitch thoe lol that should not be his final match</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.999517</td>\n",
       "      <td>0.000483</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>18228</td>\n",
       "      <td>RT  MENTION :  RT  MENTION :  when you tryna get bitches on LinkedIn  URL</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.999968</td>\n",
       "      <td>0.000032</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>23030</td>\n",
       "      <td>Well I wish I could shoot the shit wit u bitches but daddy gotta get dis money  . one</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0.738689</td>\n",
       "      <td>0.261311</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>16373</td>\n",
       "      <td>RT  MENTION :  Niggas can do slope intercept form equations but cant make scrambled fuckin eggs thats a lil backwards</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.959847</td>\n",
       "      <td>0.040153</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>3560</td>\n",
       "      <td>MENTION He must think you bump pussy .  I no mind .  I ' ll eat your  # thong and make you  # squirt like the  # geyser .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0.999865</td>\n",
       "      <td>0.000135</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>2692</td>\n",
       "      <td>MENTION always eatin good wit my niggah ,  and good lookin on the cd my boy got some heat now i can bump some bangers</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0.996946</td>\n",
       "      <td>0.003054</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>23549</td>\n",
       "      <td>Ya last nicca couldn ' t beat it  .  .  I hit it like a free throw !</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>0.999955</td>\n",
       "      <td>0.000045</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>12180</td>\n",
       "      <td>It ain ' t nothin to cut that bitch off</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0.971522</td>\n",
       "      <td>0.028478</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>24918</td>\n",
       "      <td>she was telling me about this kid who ' s soles of his shoes are falling off .  I am going to get him some shoes .  hope his parents don ' t get mad</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>0.999442</td>\n",
       "      <td>0.000558</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>16493</td>\n",
       "      <td>RT  MENTION :  smh u got Ebola now RT MENTION ;  MENTION :  I stick my foot out to feel some freshness and this bitch ass mosquito has the nerve to  MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>0.004137</td>\n",
       "      <td>0.995863</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>16532</td>\n",
       "      <td>RT  MENTION :  trash liars RT  MENTION :  these law students are terrible liars  # HTGAWM</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>0.993414</td>\n",
       "      <td>0.006586</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>15319</td>\n",
       "      <td>RT  MENTION :  All hoes are friendly .  Althea wants to say hi .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>0.999990</td>\n",
       "      <td>0.000010</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>21838</td>\n",
       "      <td>That bitch mouth made the honor roll</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>0.858086</td>\n",
       "      <td>0.141914</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>3523</td>\n",
       "      <td>MENTION I wonder if I have my very own NSA spook following me by now ?</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>0.282035</td>\n",
       "      <td>0.717965</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>6008</td>\n",
       "      <td>MENTION when you realize you ' re trash</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>0.082147</td>\n",
       "      <td>0.917853</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>1429</td>\n",
       "      <td>MENTION ;  MENTION :  I was told never fold  &amp; amp ;  never sell my sole quote They Should Have Told Yu Stay In School  !   * Soul MENTION ;  autocorrect hoed me</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>0.990445</td>\n",
       "      <td>0.009555</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>3240</td>\n",
       "      <td>MENTION khaye I can ' t see your freakin emojis I ' m a basic bitch today .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>0.000367</td>\n",
       "      <td>0.999633</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>21523</td>\n",
       "      <td>Someone should explain to Blake griffin that talking trash to an opposing 3rd string is unbecoming ,  even amongst clippers .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>0.981430</td>\n",
       "      <td>0.018570</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>2285</td>\n",
       "      <td>24 _ rozay _ idgt mad cuz Twan fuck his bitch MENTION ;   URL</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>0.999997</td>\n",
       "      <td>0.000003</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>496</td>\n",
       "      <td>quote I used to dress up like a cop too .  quote   quote Oh really ?  quote   quote Yeah .  .  . right after 9 / 11 .  .  . mad pussy .  quote   quote Really ?  quote   quote Yes .  .  . I ' ll NEVER FORGET .  quote</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>0.904808</td>\n",
       "      <td>0.095192</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>5757</td>\n",
       "      <td>MENTION Today was Talk Like A Pirate Day ,  and absolutely NOBODY knew Id been doing a Charlie Morton impersonation all day .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>0.999888</td>\n",
       "      <td>0.000112</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>21014</td>\n",
       "      <td>Running dangerously low on cigarettes these bitches better last me till Monday or Tuesday</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>0.999820</td>\n",
       "      <td>0.000180</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>15650</td>\n",
       "      <td>RT  MENTION :  bitches be like  quote I ' m a virgin quote   URL</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>0.999715</td>\n",
       "      <td>0.000285</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>16003</td>\n",
       "      <td>RT  MENTION :  RT  MENTION :  Hispanic women just have the great pussy and wife skills ,  white girls own the dick sucking and black chicks .  .  .  . ca MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>0.999859</td>\n",
       "      <td>0.000141</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>15343</td>\n",
       "      <td>RT  MENTION :  fuck a bitch we never loved a thot</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>0.997741</td>\n",
       "      <td>0.002259</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>530</td>\n",
       "      <td>quote Lmao  MENTION :   MENTION ;  MENTION ;  RT  MENTION :  Pound cake look like pussy  MENTION ;   URL  quote</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>0.913211</td>\n",
       "      <td>0.086789</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>16457</td>\n",
       "      <td>RT  MENTION :  Sometimes I turn my phone on loud to tweet just to hear the bird chirp .   # TrueStory  # TeamiPhone</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>0.999935</td>\n",
       "      <td>0.000065</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>24671</td>\n",
       "      <td>lmaoooooo RT  MENTION :  Dude said bitches from NY fuck to Redman ' s greatest hits  MENTION ;  MENTION ;  MENTION ;</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",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2970</th>\n",
       "      <td>0.999768</td>\n",
       "      <td>0.000232</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>18678</td>\n",
       "      <td>RT  MENTION :   MENTION ;  MENTION :  How I lose this sweet .  .  MENTION ; you ain ' t lose it you hide it lmfao Viri tryna smoke that hoe to his self</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2971</th>\n",
       "      <td>0.996143</td>\n",
       "      <td>0.003857</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>12041</td>\n",
       "      <td>Im not tongue kissing these bitches tho</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2972</th>\n",
       "      <td>0.999912</td>\n",
       "      <td>0.000088</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>17528</td>\n",
       "      <td>RT  MENTION :  Drunk inlove with my gun bae bitch Beyonc MENTION ;   MENTION ;  MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2973</th>\n",
       "      <td>0.998389</td>\n",
       "      <td>0.001611</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>19865</td>\n",
       "      <td>RT  MENTION :   MENTION :  when ur with ur friend and a bitch u both hate walks past  URL  quote   MENTION us as n MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2974</th>\n",
       "      <td>0.994836</td>\n",
       "      <td>0.005164</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>25042</td>\n",
       "      <td>there ' s 13 people in my house and they ' re all yelling pls god give me the strength not to beat a bitch .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2975</th>\n",
       "      <td>0.996965</td>\n",
       "      <td>0.003035</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>17334</td>\n",
       "      <td>RT  MENTION :  It ain nun to cut dat bitch awf  -  Verizon</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2976</th>\n",
       "      <td>0.998894</td>\n",
       "      <td>0.001106</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>23406</td>\n",
       "      <td>Will Hillary  MENTION ; dodo - bird MENTION ;  Be Dems 2016 Candidate To Take Over Obama ' s Job As US Prez ?  DoG  URL  via  MENTION</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2977</th>\n",
       "      <td>0.002201</td>\n",
       "      <td>0.997799</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>24984</td>\n",
       "      <td>stupid loose pussyed bitch</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2978</th>\n",
       "      <td>0.852594</td>\n",
       "      <td>0.147406</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>4903</td>\n",
       "      <td>MENTION  MENTION oh I see .  Part of a giant govt take over eh .  Stockboy you are a joke of a representative .   # dumb teabagger</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2979</th>\n",
       "      <td>0.997528</td>\n",
       "      <td>0.002472</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>20387</td>\n",
       "      <td>RT  MENTION :  Not .  A .  Fucking .  Charge .  Even .  As .  A .  Juvenile .  If you call  # MichaelBrown a thug at this point ,  just say  quote nigger quote  cuz it ' s  MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2980</th>\n",
       "      <td>0.999093</td>\n",
       "      <td>0.000907</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>8678</td>\n",
       "      <td>Chase the pussy ,  it ' s a sin .  Falls in ya lap ,  it ' s a blessing .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2981</th>\n",
       "      <td>0.997548</td>\n",
       "      <td>0.002452</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>19319</td>\n",
       "      <td>RT  MENTION :  She used to be more fun as a hoe  .  . this new you is to positive</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2982</th>\n",
       "      <td>0.010658</td>\n",
       "      <td>0.989342</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>3034</td>\n",
       "      <td>MENTION  MENTION get on cunt</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2983</th>\n",
       "      <td>0.999993</td>\n",
       "      <td>0.000007</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>12959</td>\n",
       "      <td>Look at y ' all bitches only Hmu when ya wifi on bitches</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2984</th>\n",
       "      <td>0.999946</td>\n",
       "      <td>0.000054</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>22222</td>\n",
       "      <td>These groupie bitches wild enough to suck a babies dick</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2985</th>\n",
       "      <td>0.999970</td>\n",
       "      <td>0.000030</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>22246</td>\n",
       "      <td>These hoes be actin up and these niccas be actin tough !   MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2986</th>\n",
       "      <td>0.931661</td>\n",
       "      <td>0.068339</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>3553</td>\n",
       "      <td>MENTION i love seing new ones cause i save them and then youse them bitches lmfao</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2987</th>\n",
       "      <td>0.999914</td>\n",
       "      <td>0.000086</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>11272</td>\n",
       "      <td>I told that little bitch ass nigga ,   quote do it look like I ' m about to eat your gun charge ?  quote  I ' ll slap you and your hoe ass lawyer .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2988</th>\n",
       "      <td>0.986510</td>\n",
       "      <td>0.013490</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>7969</td>\n",
       "      <td>And now lookin ass niggas and lookin ass bitches will be the terms of 2014 .  Will be overused  &amp; amp ;  abused in every fucking way</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2989</th>\n",
       "      <td>0.064884</td>\n",
       "      <td>0.935116</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>16705</td>\n",
       "      <td>RT  MENTION :  I don ' t understand how so many Americans can hate on Obama for saluting with a latte .  Nobody tells these swagfags anything abo MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2990</th>\n",
       "      <td>0.621399</td>\n",
       "      <td>0.378601</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>21057</td>\n",
       "      <td>Same people bitchin about jetpacks and shit in AW are counting down the days until it comes out buncha fags .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2991</th>\n",
       "      <td>0.900514</td>\n",
       "      <td>0.099486</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>146</td>\n",
       "      <td>quote  MENTION :  Lets get drunk ,  get blowed ,  spit shit spark blunts  &amp; amp ;  fuck hoes !  quote  Bro white dudes have no chill .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2992</th>\n",
       "      <td>0.352312</td>\n",
       "      <td>0.647688</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>21902</td>\n",
       "      <td>That was trash .  I ' m just gonna put my phone down .  Goodnight</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2993</th>\n",
       "      <td>0.730325</td>\n",
       "      <td>0.269675</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>19708</td>\n",
       "      <td>RT  MENTION :  just seen this little yellow nigga who lives in a pineapple .  .  .  .  .  thats some crazy shit .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2994</th>\n",
       "      <td>0.984619</td>\n",
       "      <td>0.015381</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>8033</td>\n",
       "      <td>Arroz con pollo</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2995</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>11238</td>\n",
       "      <td>I think I just gotta accept the fact that all these bitches gonna be hoes til they around the age of 27 and life will be a lot easier</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2996</th>\n",
       "      <td>0.822015</td>\n",
       "      <td>0.177985</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>17684</td>\n",
       "      <td>RT  MENTION :  Great start the morning !  My daughter danced herself into the coffee table and gave herself a nice shiner .  .  .  Yea she ' s a wh MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2997</th>\n",
       "      <td>0.999997</td>\n",
       "      <td>0.000003</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>5346</td>\n",
       "      <td>MENTION all the great QB ' s fold under pressure gshit if u blitz Peyton ass he gon gook out</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2998</th>\n",
       "      <td>0.880950</td>\n",
       "      <td>0.119050</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>12481</td>\n",
       "      <td>Just touch down in Jamaica tryna smoke two acres  .  Send a nigguh to his makerr</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2999</th>\n",
       "      <td>0.999987</td>\n",
       "      <td>0.000013</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>237</td>\n",
       "      <td>quote  MENTION :   MENTION ;  MENTION ;  MENTION ;  MENTION ; RT  MENTION :  Lmao !  This bitch come scooting in the tent like she about to fuck some shit up !  quote   MENTION ;  MENTION ;  MENTION ;</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3000 rows × 6 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      weight_correct    error  true_label  predicted     id  \\\n",
       "0    0.999919        0.000081  1           1          23437   \n",
       "1    0.997979        0.002021  1           1          17813   \n",
       "2    0.998711        0.001289  1           1          2561    \n",
       "3    0.999517        0.000483  1           1          18228   \n",
       "4    0.999968        0.000032  1           1          23030   \n",
       "5    0.738689        0.261311  1           1          16373   \n",
       "6    0.959847        0.040153  1           1          3560    \n",
       "7    0.999865        0.000135  1           1          2692    \n",
       "8    0.996946        0.003054  1           1          23549   \n",
       "9    0.999955        0.000045  1           1          12180   \n",
       "10   0.971522        0.028478  2           2          24918   \n",
       "11   0.999442        0.000558  1           1          16493   \n",
       "12   0.004137        0.995863  2           2          16532   \n",
       "13   0.993414        0.006586  1           1          15319   \n",
       "14   0.999990        0.000010  1           1          21838   \n",
       "15   0.858086        0.141914  2           2          3523    \n",
       "16   0.282035        0.717965  1           2          6008    \n",
       "17   0.082147        0.917853  1           2          1429    \n",
       "18   0.990445        0.009555  1           1          3240    \n",
       "19   0.000367        0.999633  2           2          21523   \n",
       "20   0.981430        0.018570  1           1          2285    \n",
       "21   0.999997        0.000003  1           1          496     \n",
       "22   0.904808        0.095192  2           2          5757    \n",
       "23   0.999888        0.000112  1           1          21014   \n",
       "24   0.999820        0.000180  1           1          15650   \n",
       "25   0.999715        0.000285  1           1          16003   \n",
       "26   0.999859        0.000141  1           1          15343   \n",
       "27   0.997741        0.002259  1           1          530     \n",
       "28   0.913211        0.086789  2           2          16457   \n",
       "29   0.999935        0.000065  1           1          24671   \n",
       "...       ...             ... ..          ..            ...   \n",
       "2970 0.999768        0.000232  1           1          18678   \n",
       "2971 0.996143        0.003857  1           1          12041   \n",
       "2972 0.999912        0.000088  1           1          17528   \n",
       "2973 0.998389        0.001611  1           1          19865   \n",
       "2974 0.994836        0.005164  1           1          25042   \n",
       "2975 0.996965        0.003035  1           1          17334   \n",
       "2976 0.998894        0.001106  2           2          23406   \n",
       "2977 0.002201        0.997799  0           1          24984   \n",
       "2978 0.852594        0.147406  1           2          4903    \n",
       "2979 0.997528        0.002472  1           1          20387   \n",
       "2980 0.999093        0.000907  1           1          8678    \n",
       "2981 0.997548        0.002452  1           1          19319   \n",
       "2982 0.010658        0.989342  0           1          3034    \n",
       "2983 0.999993        0.000007  1           1          12959   \n",
       "2984 0.999946        0.000054  1           1          22222   \n",
       "2985 0.999970        0.000030  1           1          22246   \n",
       "2986 0.931661        0.068339  1           1          3553    \n",
       "2987 0.999914        0.000086  1           1          11272   \n",
       "2988 0.986510        0.013490  1           1          7969    \n",
       "2989 0.064884        0.935116  1           1          16705   \n",
       "2990 0.621399        0.378601  1           1          21057   \n",
       "2991 0.900514        0.099486  1           1          146     \n",
       "2992 0.352312        0.647688  2           2          21902   \n",
       "2993 0.730325        0.269675  1           1          19708   \n",
       "2994 0.984619        0.015381  2           2          8033    \n",
       "2995 1.000000        0.000000  1           1          11238   \n",
       "2996 0.822015        0.177985  2           2          17684   \n",
       "2997 0.999997        0.000003  1           1          5346    \n",
       "2998 0.880950        0.119050  1           1          12481   \n",
       "2999 0.999987        0.000013  1           1          237     \n",
       "\n",
       "                                                                                                                                                                                                                           text  \n",
       "0     Wizard turned into a hoe after a break up                                                                                                                                                                                  \n",
       "1     RT  MENTION :  Captain Kirk hoe !                                                                                                                                                                                          \n",
       "2      MENTION bruh did go out like a bitch thoe lol that should not be his final match                                                                                                                                          \n",
       "3     RT  MENTION :  RT  MENTION :  when you tryna get bitches on LinkedIn  URL                                                                                                                                                  \n",
       "4     Well I wish I could shoot the shit wit u bitches but daddy gotta get dis money  . one                                                                                                                                      \n",
       "5     RT  MENTION :  Niggas can do slope intercept form equations but cant make scrambled fuckin eggs thats a lil backwards                                                                                                      \n",
       "6      MENTION He must think you bump pussy .  I no mind .  I ' ll eat your  # thong and make you  # squirt like the  # geyser .                                                                                                 \n",
       "7      MENTION always eatin good wit my niggah ,  and good lookin on the cd my boy got some heat now i can bump some bangers                                                                                                     \n",
       "8     Ya last nicca couldn ' t beat it  .  .  I hit it like a free throw !                                                                                                                                                       \n",
       "9     It ain ' t nothin to cut that bitch off                                                                                                                                                                                    \n",
       "10    she was telling me about this kid who ' s soles of his shoes are falling off .  I am going to get him some shoes .  hope his parents don ' t get mad                                                                       \n",
       "11    RT  MENTION :  smh u got Ebola now RT MENTION ;  MENTION :  I stick my foot out to feel some freshness and this bitch ass mosquito has the nerve to  MENTION ;                                                             \n",
       "12    RT  MENTION :  trash liars RT  MENTION :  these law students are terrible liars  # HTGAWM                                                                                                                                  \n",
       "13    RT  MENTION :  All hoes are friendly .  Althea wants to say hi .                                                                                                                                                           \n",
       "14    That bitch mouth made the honor roll                                                                                                                                                                                       \n",
       "15     MENTION I wonder if I have my very own NSA spook following me by now ?                                                                                                                                                    \n",
       "16     MENTION when you realize you ' re trash                                                                                                                                                                                   \n",
       "17     MENTION ;  MENTION :  I was told never fold  & amp ;  never sell my sole quote They Should Have Told Yu Stay In School  !   * Soul MENTION ;  autocorrect hoed me                                                         \n",
       "18     MENTION khaye I can ' t see your freakin emojis I ' m a basic bitch today .                                                                                                                                               \n",
       "19    Someone should explain to Blake griffin that talking trash to an opposing 3rd string is unbecoming ,  even amongst clippers .                                                                                              \n",
       "20    24 _ rozay _ idgt mad cuz Twan fuck his bitch MENTION ;   URL                                                                                                                                                              \n",
       "21     quote I used to dress up like a cop too .  quote   quote Oh really ?  quote   quote Yeah .  .  . right after 9 / 11 .  .  . mad pussy .  quote   quote Really ?  quote   quote Yes .  .  . I ' ll NEVER FORGET .  quote   \n",
       "22     MENTION Today was Talk Like A Pirate Day ,  and absolutely NOBODY knew Id been doing a Charlie Morton impersonation all day .                                                                                             \n",
       "23    Running dangerously low on cigarettes these bitches better last me till Monday or Tuesday                                                                                                                                  \n",
       "24    RT  MENTION :  bitches be like  quote I ' m a virgin quote   URL                                                                                                                                                           \n",
       "25    RT  MENTION :  RT  MENTION :  Hispanic women just have the great pussy and wife skills ,  white girls own the dick sucking and black chicks .  .  .  . ca MENTION ;                                                        \n",
       "26    RT  MENTION :  fuck a bitch we never loved a thot                                                                                                                                                                          \n",
       "27     quote Lmao  MENTION :   MENTION ;  MENTION ;  RT  MENTION :  Pound cake look like pussy  MENTION ;   URL  quote                                                                                                           \n",
       "28    RT  MENTION :  Sometimes I turn my phone on loud to tweet just to hear the bird chirp .   # TrueStory  # TeamiPhone                                                                                                        \n",
       "29    lmaoooooo RT  MENTION :  Dude said bitches from NY fuck to Redman ' s greatest hits  MENTION ;  MENTION ;  MENTION ;                                                                                                       \n",
       "...                                                                                                                     ...                                                                                                      \n",
       "2970  RT  MENTION :   MENTION ;  MENTION :  How I lose this sweet .  .  MENTION ; you ain ' t lose it you hide it lmfao Viri tryna smoke that hoe to his self                                                                    \n",
       "2971  Im not tongue kissing these bitches tho                                                                                                                                                                                    \n",
       "2972  RT  MENTION :  Drunk inlove with my gun bae bitch Beyonc MENTION ;   MENTION ;  MENTION ;                                                                                                                                  \n",
       "2973  RT  MENTION :   MENTION :  when ur with ur friend and a bitch u both hate walks past  URL  quote   MENTION us as n MENTION ;                                                                                               \n",
       "2974  there ' s 13 people in my house and they ' re all yelling pls god give me the strength not to beat a bitch .                                                                                                               \n",
       "2975  RT  MENTION :  It ain nun to cut dat bitch awf  -  Verizon                                                                                                                                                                 \n",
       "2976  Will Hillary  MENTION ; dodo - bird MENTION ;  Be Dems 2016 Candidate To Take Over Obama ' s Job As US Prez ?  DoG  URL  via  MENTION                                                                                      \n",
       "2977  stupid loose pussyed bitch                                                                                                                                                                                                 \n",
       "2978   MENTION  MENTION oh I see .  Part of a giant govt take over eh .  Stockboy you are a joke of a representative .   # dumb teabagger                                                                                        \n",
       "2979  RT  MENTION :  Not .  A .  Fucking .  Charge .  Even .  As .  A .  Juvenile .  If you call  # MichaelBrown a thug at this point ,  just say  quote nigger quote  cuz it ' s  MENTION ;                                     \n",
       "2980  Chase the pussy ,  it ' s a sin .  Falls in ya lap ,  it ' s a blessing .                                                                                                                                                  \n",
       "2981  RT  MENTION :  She used to be more fun as a hoe  .  . this new you is to positive                                                                                                                                          \n",
       "2982   MENTION  MENTION get on cunt                                                                                                                                                                                              \n",
       "2983  Look at y ' all bitches only Hmu when ya wifi on bitches                                                                                                                                                                   \n",
       "2984  These groupie bitches wild enough to suck a babies dick                                                                                                                                                                    \n",
       "2985  These hoes be actin up and these niccas be actin tough !   MENTION ;                                                                                                                                                       \n",
       "2986   MENTION i love seing new ones cause i save them and then youse them bitches lmfao                                                                                                                                         \n",
       "2987  I told that little bitch ass nigga ,   quote do it look like I ' m about to eat your gun charge ?  quote  I ' ll slap you and your hoe ass lawyer .                                                                        \n",
       "2988  And now lookin ass niggas and lookin ass bitches will be the terms of 2014 .  Will be overused  & amp ;  abused in every fucking way                                                                                       \n",
       "2989  RT  MENTION :  I don ' t understand how so many Americans can hate on Obama for saluting with a latte .  Nobody tells these swagfags anything abo MENTION ;                                                                \n",
       "2990  Same people bitchin about jetpacks and shit in AW are counting down the days until it comes out buncha fags .                                                                                                              \n",
       "2991   quote  MENTION :  Lets get drunk ,  get blowed ,  spit shit spark blunts  & amp ;  fuck hoes !  quote  Bro white dudes have no chill .                                                                                    \n",
       "2992  That was trash .  I ' m just gonna put my phone down .  Goodnight                                                                                                                                                          \n",
       "2993  RT  MENTION :  just seen this little yellow nigga who lives in a pineapple .  .  .  .  .  thats some crazy shit .                                                                                                          \n",
       "2994  Arroz con pollo                                                                                                                                                                                                            \n",
       "2995  I think I just gotta accept the fact that all these bitches gonna be hoes til they around the age of 27 and life will be a lot easier                                                                                      \n",
       "2996  RT  MENTION :  Great start the morning !  My daughter danced herself into the coffee table and gave herself a nice shiner .  .  .  Yea she ' s a wh MENTION ;                                                              \n",
       "2997   MENTION all the great QB ' s fold under pressure gshit if u blitz Peyton ass he gon gook out                                                                                                                              \n",
       "2998  Just touch down in Jamaica tryna smoke two acres  .  Send a nigguh to his makerr                                                                                                                                           \n",
       "2999   quote  MENTION :   MENTION ;  MENTION ;  MENTION ;  MENTION ; RT  MENTION :  Lmao !  This bitch come scooting in the tent like she about to fuck some shit up !  quote   MENTION ;  MENTION ;  MENTION ;                  \n",
       "\n",
       "[3000 rows x 6 columns]"
      ]
     },
     "execution_count": 403,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# combine each items id and text\n",
    "comb = pd.concat([comb, test[['id', 'text']] ], axis=1)\n",
    "comb"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 404,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Display whole text of dataframe field and don't cut it\n",
    "pd.set_option('display.max_colwidth', -1)   \n",
    "\n",
    "#np.set_printoptions(suppress=True)\n",
    "pd.set_option('float_format', '{:f}'.format)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 405,
   "metadata": {},
   "outputs": [],
   "source": [
    "# sort by error, decreasing order\n",
    "sorted = comb.sort_values(by=['error'], ascending=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 406,
   "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>weight_correct</th>\n",
       "      <th>error</th>\n",
       "      <th>true_label</th>\n",
       "      <th>predicted</th>\n",
       "      <th>id</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1231</th>\n",
       "      <td>0.000001</td>\n",
       "      <td>0.999999</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>20120</td>\n",
       "      <td>RT  MENTION :  realest tweet i seen today RT  MENTION :   MENTION ;  .  Dyke bitches are rehab for hoes . And once im healed , back to the dick I go</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>406</th>\n",
       "      <td>0.000003</td>\n",
       "      <td>0.999997</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>8564</td>\n",
       "      <td>Can MENTION ; t wait to humble these niggas in the playoffs again .  Until then ,  good job hoes</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1831</th>\n",
       "      <td>0.000009</td>\n",
       "      <td>0.999991</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>3124</td>\n",
       "      <td>MENTION  MENTION  MENTION ha booner has that twitter  MENTION ha</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1604</th>\n",
       "      <td>0.000009</td>\n",
       "      <td>0.999991</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>17299</td>\n",
       "      <td>RT  MENTION :  Green azz niggas !  !  !  !   &amp; amp ;  yall new bitches condone dat sucka activity dat why they keep doin it !  !  !</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2337</th>\n",
       "      <td>0.000011</td>\n",
       "      <td>0.999989</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>10949</td>\n",
       "      <td>I like how niggas try an come at me and im like do it then bitch and they dont do shit .  Pussies</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1002</th>\n",
       "      <td>0.000011</td>\n",
       "      <td>0.999989</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>9185</td>\n",
       "      <td>Eh bitch !  ?  How about you worry about your own psychotic ass .  Looks like you need to  quote be saved quote  from being a noisy bitch  MENTION ;  MENTION ;  MENTION ;   # cantStandUcunt</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1345</th>\n",
       "      <td>0.000013</td>\n",
       "      <td>0.999987</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>22352</td>\n",
       "      <td>They slut for me they kill for me they steal for me and of course it ' ll be yo cash and I ' ll murder that bitch send the body back to yo ass</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>710</th>\n",
       "      <td>0.000016</td>\n",
       "      <td>0.999984</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>532</td>\n",
       "      <td>quote Mam ,  Ren MENTION ; e doet in haar rol net of ze ongesteld is Wat is ongesteld ?  quote  Van dat soort vragen waar je denkt :  hoe leg ik die weer uit .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>273</th>\n",
       "      <td>0.000018</td>\n",
       "      <td>0.999982</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>12494</td>\n",
       "      <td>KaasPiet , bloemPiet .  ? lol men weet van gekkigheid niet meer hoe ze simpele oplossing tegen racisme moet maken . Zolang maar blanke beslissing is</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1806</th>\n",
       "      <td>0.000024</td>\n",
       "      <td>0.999976</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>5481</td>\n",
       "      <td>MENTION derrrrrrr ,  1st of all her songs are actually good compared to the other trash ,   +  +  +  shes SUCH a showman omg</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>390</th>\n",
       "      <td>0.000025</td>\n",
       "      <td>0.999975</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>11074</td>\n",
       "      <td>I need to chat with  MENTION  .  Where dat nicca at ?  Lol</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1738</th>\n",
       "      <td>0.000030</td>\n",
       "      <td>0.999970</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>16303</td>\n",
       "      <td>RT  MENTION :  Scholen waar directie leerkrachten pest ,  kunt u zich voorstellen hoe groot pestproblematiek onder leerlingen wel moet zijn ?  MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>931</th>\n",
       "      <td>0.000032</td>\n",
       "      <td>0.999968</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>15612</td>\n",
       "      <td>RT  MENTION :  Todays Horoscope :  you ' re still a hoe That is all .  have a nice day</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1082</th>\n",
       "      <td>0.000039</td>\n",
       "      <td>0.999961</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>5688</td>\n",
       "      <td>MENTION May I jump in u ball - less dick - wad ?  As a conservative ,  feminist ,  disabled vet bitch ,  I wish u luck with a capital F .   MENTION</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2914</th>\n",
       "      <td>0.000056</td>\n",
       "      <td>0.999944</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>7462</td>\n",
       "      <td>MENTION  MENTION  MENTION aye ,  I respect ya hustle and all but dont tweet me anymore trash ass videos .  .  . thank u kindly</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2555</th>\n",
       "      <td>0.000065</td>\n",
       "      <td>0.999935</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>9979</td>\n",
       "      <td>He ain ' t shit girl ,   MENTION ; he a bitch made nigga  !</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>551</th>\n",
       "      <td>0.000080</td>\n",
       "      <td>0.999920</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>10911</td>\n",
       "      <td>I know righttt RT  MENTION :  I hate fat loud bitches</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1156</th>\n",
       "      <td>0.000088</td>\n",
       "      <td>0.999912</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>24202</td>\n",
       "      <td>cant trust niggas when it come to bitches</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1614</th>\n",
       "      <td>0.000109</td>\n",
       "      <td>0.999891</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>15872</td>\n",
       "      <td>RT  MENTION :  I like how I found out that you can now edit Facebook comments through a trending topic on Twitter .  You think I ' d hear abo  .  .  .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2924</th>\n",
       "      <td>0.000148</td>\n",
       "      <td>0.999852</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>15629</td>\n",
       "      <td>RT  MENTION :  I wanna sleep but I ' m outside with this nigger lol trying to hang him</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>256</th>\n",
       "      <td>0.000151</td>\n",
       "      <td>0.999849</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>6375</td>\n",
       "      <td>MENTION  &amp; amp ;  4th my guinea pig will haunt you</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1029</th>\n",
       "      <td>0.000201</td>\n",
       "      <td>0.999799</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>7159</td>\n",
       "      <td>MENTION some negro amigo pulled a gun on Nate and I a few weeks ago .  I was STOKED !  MENTION ;  MENTION ;  MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1421</th>\n",
       "      <td>0.000208</td>\n",
       "      <td>0.999792</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>23303</td>\n",
       "      <td>Who would ' ve known that Jesus ' ll come back to the ghetto  /  on that level and that thorough like a black heroine .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2931</th>\n",
       "      <td>0.000221</td>\n",
       "      <td>0.999779</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>14429</td>\n",
       "      <td>RT  MENTION :  I hate when a nigga ask me  quote Where the hoes ?  quote   .  .  .  .  Nigga idfk  !</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2512</th>\n",
       "      <td>0.000225</td>\n",
       "      <td>0.999775</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>3068</td>\n",
       "      <td>MENTION  MENTION  MENTION another black man ?  What does that have to do with anything ?  Once a monkey ,  always a monkey ,  Chicago idiot</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>618</th>\n",
       "      <td>0.000229</td>\n",
       "      <td>0.999771</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>2217</td>\n",
       "      <td>.  .  .  .  .  .  .  .  .  . does nobody sell  # orange flavored and colored  # bubblegum ?  IS THAT YOUR FUCKIN '  PROBLEM ?   # Giants ?  Fuck it .  I ' m going  # 6Pack now</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1088</th>\n",
       "      <td>0.000234</td>\n",
       "      <td>0.999766</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>3448</td>\n",
       "      <td>MENTION Ass Oreos and his name is Thad Castle .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2088</th>\n",
       "      <td>0.000264</td>\n",
       "      <td>0.999736</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>333</td>\n",
       "      <td>quote  MENTION :  Roses are red and violets are blue and niggers are black ,  you know that ' s true .  quote   MENTION</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1763</th>\n",
       "      <td>0.000313</td>\n",
       "      <td>0.999687</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>6348</td>\n",
       "      <td>MENTION I ' m going to blame the black man ,  since they always blame  quote whitey quote  I ' m an equal opportunity hater .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2882</th>\n",
       "      <td>0.000325</td>\n",
       "      <td>0.999675</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>24595</td>\n",
       "      <td>is it ghetto to grease yo scalp at work</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",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2728</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>15913</td>\n",
       "      <td>RT  MENTION :  I never understood that whole pissin on a bitch trend that was going round .  .  . I like my bitches clean ,  mayne</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2260</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>11666</td>\n",
       "      <td>I ' m serious bruh !  She ' s too pretty I would eat both her pussy  &amp; amp ;  butt  MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2956</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>17314</td>\n",
       "      <td>RT  MENTION :   MENTION fr mane on some 100 shit cuff her if she made you that happy in one night cause these hoes for everybody fr</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>58</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>13992</td>\n",
       "      <td>One time in high school ,  this girl told my friend that he needed proactive .  So he told her she needed weight watchers .  HAHAHA bitch .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1748</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>20628</td>\n",
       "      <td>RT  MENTION :  if ya bitch mad tell ha I ' ll put ha own  MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2995</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>11238</td>\n",
       "      <td>I think I just gotta accept the fact that all these bitches gonna be hoes til they around the age of 27 and life will be a lot easier</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2112</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>18691</td>\n",
       "      <td>RT  MENTION :  If you can ' t eat pussy that ' s got hair on it I ' m not even sure you really know how to eat pussy .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>109</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>9272</td>\n",
       "      <td>Ew I hate when ratchet ass hoes post  ' stranger danger '  ewww bitch</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1657</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>9055</td>\n",
       "      <td>Don ' t love yo bitch ,  just fuck yo bitch when I got time to do it</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1728</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>482</td>\n",
       "      <td>quote I fucking hate that bitch so much I wanna smack her in the face then tongue kiss her quote</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2039</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>5450</td>\n",
       "      <td>MENTION  -  lol it ' s possible that a hoe can treat you the right way but will she love you the right way ?  No .  A hoe will be a hoe .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>556</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>8378</td>\n",
       "      <td>Bout to unfollow all the bitches I follow currently and follow all new bitches .  I need change in my life and I feel like that ' s a good start</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>424</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>15747</td>\n",
       "      <td>RT  MENTION :  I MENTION ; m a bitch ?  You MENTION ; re a bitch .  Your mom MENTION ; s a bitch for having a bitch ,  your dad MENTION ; s a bitch for fucking a bitch .  Now who MENTION ; s t MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1637</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>24429</td>\n",
       "      <td>hoes need to stay in hoes place fr</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>811</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>10336</td>\n",
       "      <td>I been getting bitches my whole life I been getting money my whole life  MENTION ;  MENTION ;  MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>942</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>9062</td>\n",
       "      <td>Don ' t put too much effort into these bitches if you ain ' t gettin none back .  Give them exactly what they give you</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2658</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>13303</td>\n",
       "      <td>My bitch pussy got hella powers</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1519</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>24149</td>\n",
       "      <td>bitches aint shit but hoes an tricks</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2303</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>21865</td>\n",
       "      <td>That last tweet was just song lyrics .  I don ' t really like bad bitches ,  I like pretty chicks .  Bad bitches cheat too easily and too often</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>414</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>16212</td>\n",
       "      <td>RT  MENTION :  You can ' t be a hoe saying that another girls a hoe .  Your both hoes ,  it cancels each other out .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1031</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>10113</td>\n",
       "      <td>Hoes always blaming hoes for being hoes .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1575</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>11568</td>\n",
       "      <td>I ' m going thru it my bitch ain ' t even my bitch she let the whole team hit</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2110</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>13817</td>\n",
       "      <td>Now my side bitch my main bitch cause my main bitch aint feelin me no moe</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>986</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>7641</td>\n",
       "      <td>A girlfriend is almost a bill lowkey .  So at this point in my life where I should be gettin outta my moms crib ,  I ' ll rather just have hoes .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1500</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>11392</td>\n",
       "      <td>I wish ,  I wish ,  I wish ,  I wish ,  I wish ,  I wish a bitch would .  .  Cept she already did ,  gonna have to handle this  MENTION ;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2252</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>20041</td>\n",
       "      <td>RT  MENTION :  Hoes be calling hoes ,  hoes ,  when they know they just as much of a hoe as the other hoe they calling a hoe .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>779</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>25004</td>\n",
       "      <td>that hoe that you in love with ?  that ' s that hoe I mess around with</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2176</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>11311</td>\n",
       "      <td>I want my bitch to answer if my ex calling ,  with my dick in her mouth</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>296</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>5866</td>\n",
       "      <td>MENTION bitch yessssssss MENTION ;  MENTION ;  omg I jus saw one of em yesterday I ' m in disbelief I rode by like oh that ' s court bro now I ' m lost</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1255</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>11807</td>\n",
       "      <td>If I ' ve never seen her do some  quote hoe quote  shit with my own eyes then I refuse to let another nigga influence my thoughts on her .  So what ?</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3000 rows × 6 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      weight_correct    error  true_label  predicted     id  \\\n",
       "1231 0.000001        0.999999  0           1          20120   \n",
       "406  0.000003        0.999997  0           1          8564    \n",
       "1831 0.000009        0.999991  2           2          3124    \n",
       "1604 0.000009        0.999991  0           1          17299   \n",
       "2337 0.000011        0.999989  0           1          10949   \n",
       "1002 0.000011        0.999989  0           1          9185    \n",
       "1345 0.000013        0.999987  0           1          22352   \n",
       "710  0.000016        0.999984  2           2          532     \n",
       "273  0.000018        0.999982  2           2          12494   \n",
       "1806 0.000024        0.999976  2           2          5481    \n",
       "390  0.000025        0.999975  0           1          11074   \n",
       "1738 0.000030        0.999970  2           2          16303   \n",
       "931  0.000032        0.999968  2           1          15612   \n",
       "1082 0.000039        0.999961  0           1          5688    \n",
       "2914 0.000056        0.999944  2           1          7462    \n",
       "2555 0.000065        0.999935  0           1          9979    \n",
       "551  0.000080        0.999920  0           1          10911   \n",
       "1156 0.000088        0.999912  0           1          24202   \n",
       "1614 0.000109        0.999891  2           2          15872   \n",
       "2924 0.000148        0.999852  0           1          15629   \n",
       "256  0.000151        0.999849  1           2          6375    \n",
       "1029 0.000201        0.999799  0           1          7159    \n",
       "1421 0.000208        0.999792  2           1          23303   \n",
       "2931 0.000221        0.999779  0           1          14429   \n",
       "2512 0.000225        0.999775  1           2          3068    \n",
       "618  0.000229        0.999771  1           1          2217    \n",
       "1088 0.000234        0.999766  1           1          3448    \n",
       "2088 0.000264        0.999736  1           0          333     \n",
       "1763 0.000313        0.999687  0           1          6348    \n",
       "2882 0.000325        0.999675  2           2          24595   \n",
       "...       ...             ... ..          ..            ...   \n",
       "2728 1.000000        0.000000  1           1          15913   \n",
       "2260 1.000000        0.000000  1           1          11666   \n",
       "2956 1.000000        0.000000  1           1          17314   \n",
       "58   1.000000        0.000000  1           1          13992   \n",
       "1748 1.000000        0.000000  1           1          20628   \n",
       "2995 1.000000        0.000000  1           1          11238   \n",
       "2112 1.000000        0.000000  1           1          18691   \n",
       "109  1.000000        0.000000  1           1          9272    \n",
       "1657 1.000000        0.000000  1           1          9055    \n",
       "1728 1.000000        0.000000  1           1          482     \n",
       "2039 1.000000        0.000000  1           1          5450    \n",
       "556  1.000000        0.000000  1           1          8378    \n",
       "424  1.000000        0.000000  1           1          15747   \n",
       "1637 1.000000        0.000000  1           1          24429   \n",
       "811  1.000000        0.000000  1           1          10336   \n",
       "942  1.000000        0.000000  1           1          9062    \n",
       "2658 1.000000        0.000000  1           1          13303   \n",
       "1519 1.000000        0.000000  1           1          24149   \n",
       "2303 1.000000        0.000000  1           1          21865   \n",
       "414  1.000000        0.000000  1           1          16212   \n",
       "1031 1.000000        0.000000  1           1          10113   \n",
       "1575 1.000000        0.000000  1           1          11568   \n",
       "2110 1.000000        0.000000  1           1          13817   \n",
       "986  1.000000        0.000000  1           1          7641    \n",
       "1500 1.000000        0.000000  1           1          11392   \n",
       "2252 1.000000        0.000000  1           1          20041   \n",
       "779  1.000000        0.000000  1           1          25004   \n",
       "2176 1.000000        0.000000  1           1          11311   \n",
       "296  1.000000        0.000000  1           1          5866    \n",
       "1255 1.000000        0.000000  1           1          11807   \n",
       "\n",
       "                                                                                                                                                                                                             text  \n",
       "1231  RT  MENTION :  realest tweet i seen today RT  MENTION :   MENTION ;  .  Dyke bitches are rehab for hoes . And once im healed , back to the dick I go                                                         \n",
       "406   Can MENTION ; t wait to humble these niggas in the playoffs again .  Until then ,  good job hoes                                                                                                             \n",
       "1831   MENTION  MENTION  MENTION ha booner has that twitter  MENTION ha                                                                                                                                            \n",
       "1604  RT  MENTION :  Green azz niggas !  !  !  !   & amp ;  yall new bitches condone dat sucka activity dat why they keep doin it !  !  !                                                                          \n",
       "2337  I like how niggas try an come at me and im like do it then bitch and they dont do shit .  Pussies                                                                                                            \n",
       "1002  Eh bitch !  ?  How about you worry about your own psychotic ass .  Looks like you need to  quote be saved quote  from being a noisy bitch  MENTION ;  MENTION ;  MENTION ;   # cantStandUcunt                \n",
       "1345  They slut for me they kill for me they steal for me and of course it ' ll be yo cash and I ' ll murder that bitch send the body back to yo ass                                                               \n",
       "710    quote Mam ,  Ren MENTION ; e doet in haar rol net of ze ongesteld is Wat is ongesteld ?  quote  Van dat soort vragen waar je denkt :  hoe leg ik die weer uit .                                             \n",
       "273   KaasPiet , bloemPiet .  ? lol men weet van gekkigheid niet meer hoe ze simpele oplossing tegen racisme moet maken . Zolang maar blanke beslissing is                                                         \n",
       "1806   MENTION derrrrrrr ,  1st of all her songs are actually good compared to the other trash ,   +  +  +  shes SUCH a showman omg                                                                                \n",
       "390   I need to chat with  MENTION  .  Where dat nicca at ?  Lol                                                                                                                                                   \n",
       "1738  RT  MENTION :  Scholen waar directie leerkrachten pest ,  kunt u zich voorstellen hoe groot pestproblematiek onder leerlingen wel moet zijn ?  MENTION ;                                                     \n",
       "931   RT  MENTION :  Todays Horoscope :  you ' re still a hoe That is all .  have a nice day                                                                                                                       \n",
       "1082   MENTION May I jump in u ball - less dick - wad ?  As a conservative ,  feminist ,  disabled vet bitch ,  I wish u luck with a capital F .   MENTION                                                         \n",
       "2914   MENTION  MENTION  MENTION aye ,  I respect ya hustle and all but dont tweet me anymore trash ass videos .  .  . thank u kindly                                                                              \n",
       "2555  He ain ' t shit girl ,   MENTION ; he a bitch made nigga  !                                                                                                                                                  \n",
       "551   I know righttt RT  MENTION :  I hate fat loud bitches                                                                                                                                                        \n",
       "1156  cant trust niggas when it come to bitches                                                                                                                                                                    \n",
       "1614  RT  MENTION :  I like how I found out that you can now edit Facebook comments through a trending topic on Twitter .  You think I ' d hear abo  .  .  .                                                       \n",
       "2924  RT  MENTION :  I wanna sleep but I ' m outside with this nigger lol trying to hang him                                                                                                                       \n",
       "256    MENTION  & amp ;  4th my guinea pig will haunt you                                                                                                                                                          \n",
       "1029   MENTION some negro amigo pulled a gun on Nate and I a few weeks ago .  I was STOKED !  MENTION ;  MENTION ;  MENTION ;                                                                                      \n",
       "1421  Who would ' ve known that Jesus ' ll come back to the ghetto  /  on that level and that thorough like a black heroine .                                                                                      \n",
       "2931  RT  MENTION :  I hate when a nigga ask me  quote Where the hoes ?  quote   .  .  .  .  Nigga idfk  !                                                                                                         \n",
       "2512   MENTION  MENTION  MENTION another black man ?  What does that have to do with anything ?  Once a monkey ,  always a monkey ,  Chicago idiot                                                                 \n",
       "618    .  .  .  .  .  .  .  .  .  . does nobody sell  # orange flavored and colored  # bubblegum ?  IS THAT YOUR FUCKIN '  PROBLEM ?   # Giants ?  Fuck it .  I ' m going  # 6Pack now                             \n",
       "1088   MENTION Ass Oreos and his name is Thad Castle .                                                                                                                                                             \n",
       "2088   quote  MENTION :  Roses are red and violets are blue and niggers are black ,  you know that ' s true .  quote   MENTION                                                                                     \n",
       "1763   MENTION I ' m going to blame the black man ,  since they always blame  quote whitey quote  I ' m an equal opportunity hater .                                                                               \n",
       "2882  is it ghetto to grease yo scalp at work                                                                                                                                                                      \n",
       "...                                       ...                                                                                                                                                                      \n",
       "2728  RT  MENTION :  I never understood that whole pissin on a bitch trend that was going round .  .  . I like my bitches clean ,  mayne                                                                           \n",
       "2260  I ' m serious bruh !  She ' s too pretty I would eat both her pussy  & amp ;  butt  MENTION ;                                                                                                                \n",
       "2956  RT  MENTION :   MENTION fr mane on some 100 shit cuff her if she made you that happy in one night cause these hoes for everybody fr                                                                          \n",
       "58    One time in high school ,  this girl told my friend that he needed proactive .  So he told her she needed weight watchers .  HAHAHA bitch .                                                                  \n",
       "1748  RT  MENTION :  if ya bitch mad tell ha I ' ll put ha own  MENTION ;                                                                                                                                          \n",
       "2995  I think I just gotta accept the fact that all these bitches gonna be hoes til they around the age of 27 and life will be a lot easier                                                                        \n",
       "2112  RT  MENTION :  If you can ' t eat pussy that ' s got hair on it I ' m not even sure you really know how to eat pussy .                                                                                       \n",
       "109   Ew I hate when ratchet ass hoes post  ' stranger danger '  ewww bitch                                                                                                                                        \n",
       "1657  Don ' t love yo bitch ,  just fuck yo bitch when I got time to do it                                                                                                                                         \n",
       "1728   quote I fucking hate that bitch so much I wanna smack her in the face then tongue kiss her quote                                                                                                            \n",
       "2039   MENTION  -  lol it ' s possible that a hoe can treat you the right way but will she love you the right way ?  No .  A hoe will be a hoe .                                                                   \n",
       "556   Bout to unfollow all the bitches I follow currently and follow all new bitches .  I need change in my life and I feel like that ' s a good start                                                             \n",
       "424   RT  MENTION :  I MENTION ; m a bitch ?  You MENTION ; re a bitch .  Your mom MENTION ; s a bitch for having a bitch ,  your dad MENTION ; s a bitch for fucking a bitch .  Now who MENTION ; s t MENTION ;   \n",
       "1637  hoes need to stay in hoes place fr                                                                                                                                                                           \n",
       "811   I been getting bitches my whole life I been getting money my whole life  MENTION ;  MENTION ;  MENTION ;                                                                                                     \n",
       "942   Don ' t put too much effort into these bitches if you ain ' t gettin none back .  Give them exactly what they give you                                                                                       \n",
       "2658  My bitch pussy got hella powers                                                                                                                                                                              \n",
       "1519  bitches aint shit but hoes an tricks                                                                                                                                                                         \n",
       "2303  That last tweet was just song lyrics .  I don ' t really like bad bitches ,  I like pretty chicks .  Bad bitches cheat too easily and too often                                                              \n",
       "414   RT  MENTION :  You can ' t be a hoe saying that another girls a hoe .  Your both hoes ,  it cancels each other out .                                                                                         \n",
       "1031  Hoes always blaming hoes for being hoes .                                                                                                                                                                    \n",
       "1575  I ' m going thru it my bitch ain ' t even my bitch she let the whole team hit                                                                                                                                \n",
       "2110  Now my side bitch my main bitch cause my main bitch aint feelin me no moe                                                                                                                                    \n",
       "986   A girlfriend is almost a bill lowkey .  So at this point in my life where I should be gettin outta my moms crib ,  I ' ll rather just have hoes .                                                            \n",
       "1500  I wish ,  I wish ,  I wish ,  I wish ,  I wish ,  I wish a bitch would .  .  Cept she already did ,  gonna have to handle this  MENTION ;                                                                    \n",
       "2252  RT  MENTION :  Hoes be calling hoes ,  hoes ,  when they know they just as much of a hoe as the other hoe they calling a hoe .                                                                               \n",
       "779   that hoe that you in love with ?  that ' s that hoe I mess around with                                                                                                                                       \n",
       "2176  I want my bitch to answer if my ex calling ,  with my dick in her mouth                                                                                                                                      \n",
       "296    MENTION bitch yessssssss MENTION ;  MENTION ;  omg I jus saw one of em yesterday I ' m in disbelief I rode by like oh that ' s court bro now I ' m lost                                                     \n",
       "1255  If I ' ve never seen her do some  quote hoe quote  shit with my own eyes then I refuse to let another nigga influence my thoughts on her .  So what ?                                                        \n",
       "\n",
       "[3000 rows x 6 columns]"
      ]
     },
     "execution_count": 406,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\n",
    "sorted"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 233,
   "metadata": {},
   "outputs": [],
   "source": [
    "# function from https://deepnotes.io/softmax-crossentropy\n",
    "def cross_entropy(y,X):\n",
    "#def crossEntropy(y,X):\n",
    "    \"\"\"\n",
    "    X is the output from fully connected layer (num_examples x num_classes)\n",
    "    y is labels (num_examples x 1)\n",
    "    \tNote that y is not one-hot encoded vector. \n",
    "    \tIt can be computed as y.argmax(axis=1) from one-hot encoded vectors of labels if required.\n",
    "    \"\"\"\n",
    "    m = y.shape[0]\n",
    "    #p = softmax(X)\n",
    "    p=X\n",
    "    \n",
    "    # We use multidimensional array indexing to extract \n",
    "    # softmax probability of the correct label for each sample.\n",
    "    # Refer to https://docs.scipy.org/doc/numpy/user/basics.indexing.html#indexing-multi-dimensional-arrays for understanding multidimensional array indexing.\n",
    "    log_likelihood = -np.log(  (p[range(m),y])+1e-15)\n",
    "    loss = np.sum(log_likelihood) / m\n",
    "    return loss"
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "Predicting Movie Reviews with BERT on TF Hub.ipynb",
   "provenance": [],
   "version": "0.3.2"
  },
  "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": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}