NN_Model_Doctors.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 92,
   "metadata": {},
   "outputs": [],
   "source": [
    "import utils\n",
    "\n",
    "import nltk\n",
    "from nltk.corpus import stopwords\n",
    "\n",
    "from sklearn.feature_extraction.text import CountVectorizer\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from sklearn.pipeline import Pipeline, FeatureUnion\n",
    "from sklearn.base import BaseEstimator, TransformerMixin\n",
    "from sklearn.svm import SVC\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import jaccard_similarity_score\n",
    "\n",
    "import keras\n",
    "from keras.callbacks import TensorBoard, ModelCheckpoint\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Activation, InputLayer\n",
    "from keras.optimizers import RMSprop, Adam, SGD\n",
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from scipy.sparse import csr_matrix\n",
    "import scipy\n",
    "from tqdm import tqdm\n",
    "\n",
    "from os.path import join, basename, exists\n",
    "from os import makedirs, listdir\n",
    "import json"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def preproc_target_train(data, reduce_classes=False):\n",
    "    diagnoses = data['Код_диагноза'].copy()\n",
    "    \n",
    "    if reduce_classes:\n",
    "        pop_diagnoses = set(utils.get_most_popular_diagnoses(diagnoses, percent=.80))\n",
    "        most_pop_diagnose = scipy.stats.mode(diagnoses)[0][0]\n",
    "    else:\n",
    "        pop_diagnoses = set(diagnoses)\n",
    "        most_pop_diagnose = scipy.stats.mode(diagnoses)[0][0]\n",
    "    \n",
    "    diagnoses = diagnoses.apply(\n",
    "        lambda diag: diag if diag in pop_diagnoses else most_pop_diagnose\n",
    "    )\n",
    "    \n",
    "    return diagnoses, pop_diagnoses, most_pop_diagnose\n",
    "\n",
    "def preproc_target_test(data, pop_diagnoses, most_pop_diagnose):\n",
    "    diagnoses = data['Код_диагноза'].copy()\n",
    "    \n",
    "    diagnoses = diagnoses.apply(\n",
    "        lambda diag: diag if diag in pop_diagnoses else most_pop_diagnose\n",
    "    )\n",
    "    \n",
    "    return diagnoses, pop_diagnoses, most_pop_diagnose\n",
    "\n",
    "def join_topics(data, topics):\n",
    "    data = data.copy()\n",
    "    \n",
    "    topics_df = pd.DataFrame(dict(zip(\n",
    "        ['topic' + str(i) for i in range(topics.shape[1])],\n",
    "        [topics[:, i] for i in range(topics.shape[1])])))\n",
    "    topics_df['Id_Записи'] = data['Id_Записи']\n",
    "\n",
    "    data = data.join(topics_df, on='Id_Записи', rsuffix='_topics', how='outer')\n",
    "    data = data.drop(columns=['Id_Записи_topics'])\n",
    "    \n",
    "    return data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [],
   "source": [
    "class ItemSelector(BaseEstimator, TransformerMixin):\n",
    "    def __init__(self, key):\n",
    "        self.key = key\n",
    "\n",
    "    def fit(self, x, y=None):\n",
    "        return self\n",
    "\n",
    "    def transform(self, data_dict):\n",
    "        return data_dict[self.key]\n",
    "\n",
    "class DoctorsPopularityTransformator(BaseEstimator, TransformerMixin):\n",
    "    def fit(self, x, y=None):\n",
    "        doctors = x.fillna('sss')\n",
    "        doctors_voc, counts = np.unique(doctors, return_counts=True)\n",
    "        self.pop_doctor = doctors_voc[np.argsort(counts)[::-1][0]]\n",
    "        \n",
    "        return self\n",
    "\n",
    "    def transform(self, x):\n",
    "        x = x.fillna('sss')\n",
    "        x[x == 'sss'] = self.pop_doctor\n",
    "        \n",
    "        return x\n",
    "    \n",
    "class GenderTransformator(BaseEstimator, TransformerMixin):\n",
    "    def fit(self, x, y=None):\n",
    "        return self\n",
    "    \n",
    "    def transform(self, x):        \n",
    "        x = x.copy()\n",
    "        x[x == 1] = 0\n",
    "        x[x == 2] = 1\n",
    "        x = x.astype(np.float32)\n",
    "        \n",
    "        return np.expand_dims(x, axis=1)\n",
    "    \n",
    "class AgeTransformator(BaseEstimator, TransformerMixin):\n",
    "    def fit(self, x, y=None):\n",
    "        return self\n",
    "    \n",
    "    def transform(self, x):\n",
    "        x = x.astype(np.float32)\n",
    "        return np.expand_dims(x, axis=1)\n",
    "    \n",
    "class TopicsTransformator(BaseEstimator, TransformerMixin):\n",
    "    def fit(self, x, y=None):\n",
    "        return self\n",
    "    \n",
    "    def transform(self, x):\n",
    "        return x.astype(np.float32)\n",
    "\n",
    "class ClinicTransformator(BaseEstimator, TransformerMixin):\n",
    "    def fit(self, x, y=None):\n",
    "        clinics, counts = np.unique(x, return_counts=True)\n",
    "        self.clinics_to_idx = dict(zip(clinics, range(len(clinics))))\n",
    "        \n",
    "        self.most_pop_clinic_idx = np.argmax(counts)\n",
    "        \n",
    "        return self\n",
    "    \n",
    "    def transform(self, x):\n",
    "        x_vec = np.zeros((x.shape[0], len(self.clinics_to_idx)), dtype=np.float32)\n",
    "        for i, clinic in enumerate(x):\n",
    "            if clinic in self.clinics_to_idx:\n",
    "                x_vec[i, self.clinics_to_idx[clinic]] = 1.\n",
    "            else:\n",
    "                x_vec[i, self.most_pop_clinic_idx] = 1.\n",
    "        \n",
    "        return x_vec\n",
    "    \n",
    "class RepeatsTransformator(BaseEstimator, TransformerMixin):\n",
    "    def fit(self, x, y=None):\n",
    "        return self\n",
    "    \n",
    "    def transform(self, x):\n",
    "        x = x.astype(np.float32)\n",
    "        return np.expand_dims(x, axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [],
   "source": [
    "class NNModel(BaseEstimator, TransformerMixin):\n",
    "    def __init__(self, batch_size, epochs, tb_log_dir, model_path, validation_split=0.3):\n",
    "        self.batch_size = batch_size\n",
    "        self.epochs = epochs\n",
    "        self.tb_log_dir = tb_log_dir\n",
    "        self.model_path = model_path\n",
    "        self.validation_split = validation_split\n",
    "    \n",
    "    def fit(self, x, y=None):\n",
    "        self.classes = np.unique(y)\n",
    "        self.classes_voc = dict(zip(self.classes, range(self.classes.shape[0])))\n",
    "        self.voc_classes = dict(zip(range(self.classes.shape[0]), self.classes))\n",
    "        y_proc = np.zeros((y.shape[0], self.classes.shape[0]), dtype=np.float32)\n",
    "        for i, yc in enumerate(y):\n",
    "            y_proc[i, self.classes_voc[yc]] = 1.\n",
    "        \n",
    "        self.model = Sequential([\n",
    "            InputLayer(input_shape=(x.shape[1],)),\n",
    "            Dense(1024, activation='sigmoid'),\n",
    "#             Dense(2048, activation='sigmoid'),\n",
    "#             Dense(2048, activation='sigmoid'),\n",
    "#             Dense(4096, activation='sigmoid'),\n",
    "#             Dense(4096, activation='sigmoid'),\n",
    "            Dense(self.classes.shape[0], activation='softmax')\n",
    "        ])\n",
    "        \n",
    "        optim = RMSprop(lr=1e-4, decay=1e-6)\n",
    "        self.model.compile(optimizer=optim, loss='categorical_crossentropy', metrics=['accuracy'])\n",
    "        \n",
    "        self.model.fit(x, y_proc,\n",
    "                  batch_size=self.batch_size, epochs=self.epochs,\n",
    "                  callbacks=[\n",
    "                      TensorBoard(log_dir=self.tb_log_dir, batch_size=self.batch_size),\n",
    "                      ModelCheckpoint(filepath=self.model_path, monitor='acc', period=5)\n",
    "                  ],\n",
    "                  validation_split=self.validation_split)\n",
    "        \n",
    "        return self\n",
    "        \n",
    "    def transform(self, x):\n",
    "        pred = self.model.predict(x)\n",
    "        max_idxs = np.argmax(pred, axis=1)\n",
    "        \n",
    "        return np.array(list(map(lambda max_idx: self.voc_classes[max_idx], max_idxs)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [],
   "source": [
    "def create_pipeline(experiment_dir, model_fname_prefix, no_validation=False, validation_split=0.3):\n",
    "    model_name = model_fname_prefix + str(utils.get_next_model_id(experiment_dir))\n",
    "    tb_log_dir = join(experiment_dir, 'log', model_name)\n",
    "    models_dir = join(experiment_dir, 'models')\n",
    "\n",
    "    if no_validation:\n",
    "        model_path = utils.get_model_fname_pattern(models_dir, model_name, no_validation=True)\n",
    "        validation_split = 0.0\n",
    "    else:\n",
    "        model_path = utils.get_model_fname_pattern(models_dir, model_name, no_validation=False)\n",
    "\n",
    "    pipe = Pipeline([\n",
    "        ('union', FeatureUnion(\n",
    "            transformer_list = [\n",
    "                ('complaints_pipe', Pipeline([\n",
    "                    ('complaint_selector', ItemSelector(key='Жалобы (unigramm)')),\n",
    "                    ('tfidf', TfidfVectorizer(ngram_range=(1,1), min_df=1, stop_words=stopwords.words('russian')))\n",
    "                ])),\n",
    "                ('complaints_n_pipe', Pipeline([\n",
    "                    ('complaint_n_selector', ItemSelector(key='Жалобы (ngramm)')),\n",
    "                    ('tfidf', TfidfVectorizer(ngram_range=(1,1), min_df=1, stop_words=stopwords.words('russian')))\n",
    "                ])),\n",
    "                ('doctor_pipe', Pipeline([\n",
    "                    ('doctor_selector', ItemSelector(key='Врач')),\n",
    "                    ('doc_pop', DoctorsPopularityTransformator()),\n",
    "                    ('count_vect', CountVectorizer())\n",
    "                ])),\n",
    "                ('gender_pipe', Pipeline([\n",
    "                    ('gender_selector', ItemSelector(key='Пол')),\n",
    "                    ('gender_transform', GenderTransformator())\n",
    "                ])),\n",
    "                ('age_pipe', Pipeline([\n",
    "                    ('age_selector', ItemSelector(key='Возраст')),\n",
    "                    ('age_transformator', AgeTransformator())\n",
    "                ])),\n",
    "                ('topics_pipe', Pipeline([\n",
    "                    ('topic_selector', ItemSelector(key=['topic' + str(i) for i in range(355)])),\n",
    "                    ('topics_transform', TopicsTransformator())\n",
    "                ])),\n",
    "                ('clinic_pipe', Pipeline([\n",
    "                    ('clinic_selector', ItemSelector(key='Клиника')),\n",
    "                    ('clinic_transform', ClinicTransformator())\n",
    "                ])),\n",
    "                ('repeats_pipe', Pipeline([\n",
    "                    ('repeats_selector', ItemSelector(key='Повторный приём')),\n",
    "                    ('repeats_transform', RepeatsTransformator())\n",
    "                ]))\n",
    "            ]\n",
    "        )),\n",
    "        ('clf', NNModel(batch_size=128, epochs=50, tb_log_dir=tb_log_dir,\n",
    "                        model_path=model_path, validation_split=validation_split))\n",
    "    ])\n",
    "    \n",
    "    return pipe, model_name"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [],
   "source": [
    "train = utils.load_data('data/train_data_complaints_repeats_doctors.csv')\n",
    "train_topics = np.load('data/topics_train_ngramm.npy')\n",
    "train = utils.join_topics(train, train_topics)\n",
    "doctors = train['Врач'].fillna('sss')\n",
    "doctors_voc, counts = np.unique(doctors, return_counts=True)\n",
    "pop_doctor = doctors_voc[np.argsort(counts)[::-1][0]]\n",
    "doctors[doctors == 'sss'] = pop_doctor\n",
    "train['Врач'] = doctors\n",
    "train.iloc[np.where(train['Врач'] == 'ЛФК невролог')[0][0], train.columns.get_loc('Врач')] = 'невролог'\n",
    "# train, valid = train_test_split(train, test_size=0.3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {},
   "outputs": [],
   "source": [
    "test = utils.load_data('data/test_data_complaints_repeats_doctors.csv')\n",
    "test_topics = np.load('data/topics_test_ngramm.npy')\n",
    "test = utils.join_topics(test, test_topics)\n",
    "doctors = test['Врач'].fillna('sss')\n",
    "doctors_voc, counts = np.unique(doctors, return_counts=True)\n",
    "pop_doctor = doctors_voc[np.argsort(counts)[::-1][0]]\n",
    "doctors[doctors == 'sss'] = pop_doctor\n",
    "test['Врач'] = doctors\n",
    "test.iloc[np.where(test['Врач'] == 'ЛФК невролог')[0][0], test.columns.get_loc('Врач')] = 'невролог'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = train\n",
    "doctors_all = np.unique(data['Врач'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array(['акушер гинеколог', 'аллерголог иммунолог', 'ангиохирург хирург',\n",
       "        'врачебный', 'гастроэнтеролог', 'гинеколог', 'дерматовенеролог',\n",
       "        'дерматовенеролог трихолог', 'кардиолог', 'колопроктолог',\n",
       "        'маммолог онколог', 'невролог', 'невролог травматолог',\n",
       "        'нейрохирург хирург', 'ортопед', 'ортопед травматолог',\n",
       "        'оториноларинголог', 'офтальмолог', 'педиатр',\n",
       "        'психотерапевт терапевт', 'пульмонолог', 'реабилитолог ревматолог',\n",
       "        'ревматолог', 'рефлексотерапевт терапевт', 'стоматолог',\n",
       "        'терапевт', 'уролог', 'флеболог хирург', 'хирург', 'эндокринолог'],\n",
       "       dtype=object),\n",
       " array([11242,   513,     3,     9,  4208,     4,  4064,    72,  2459,\n",
       "          129,   979, 12636,    92,   194,     5,  2808,  4794,  2154,\n",
       "          106,    74,   121,    52,   388,   246,    22,  6291,  3711,\n",
       "          472,  1641,  2487]))"
      ]
     },
     "execution_count": 97,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.unique(data['Врач'], return_counts=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([], dtype=object)"
      ]
     },
     "execution_count": 98,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.unique(train[(train['Пол'] != 1) & (train['Пол'] != 2)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "11242/11242 [==============================] - 1s 80us/step - loss: 3.8789 - acc: 0.1399\n",
      "Epoch 2/50\n",
      "11242/11242 [==============================] - 1s 61us/step - loss: 3.3984 - acc: 0.1986\n",
      "Epoch 3/50\n",
      "11242/11242 [==============================] - 1s 55us/step - loss: 3.3443 - acc: 0.2252\n",
      "Epoch 4/50\n",
      "11242/11242 [==============================] - 1s 56us/step - loss: 3.2929 - acc: 0.2374\n",
      "Epoch 5/50\n",
      "11242/11242 [==============================] - 1s 55us/step - loss: 3.2445 - acc: 0.2461\n",
      "Epoch 6/50\n",
      "11242/11242 [==============================] - 1s 63us/step - loss: 3.1981 - acc: 0.2511\n",
      "Epoch 7/50\n",
      "11242/11242 [==============================] - 1s 51us/step - loss: 3.1532 - acc: 0.2582\n",
      "Epoch 8/50\n",
      "11242/11242 [==============================] - 1s 57us/step - loss: 3.1089 - acc: 0.2624\n",
      "Epoch 9/50\n",
      "11242/11242 [==============================] - 1s 58us/step - loss: 3.0681 - acc: 0.2685\n",
      "Epoch 10/50\n",
      "11242/11242 [==============================] - 1s 57us/step - loss: 3.0303 - acc: 0.2693\n",
      "Epoch 11/50\n",
      "11242/11242 [==============================] - 1s 63us/step - loss: 2.9964 - acc: 0.2787\n",
      "Epoch 12/50\n",
      "11242/11242 [==============================] - 1s 60us/step - loss: 2.9660 - acc: 0.2818\n",
      "Epoch 13/50\n",
      "11242/11242 [==============================] - 1s 52us/step - loss: 2.9396 - acc: 0.2888\n",
      "Epoch 14/50\n",
      "11242/11242 [==============================] - 1s 53us/step - loss: 2.9134 - acc: 0.2951\n",
      "Epoch 15/50\n",
      "11242/11242 [==============================] - 1s 50us/step - loss: 2.8918 - acc: 0.2960\n",
      "Epoch 16/50\n",
      "11242/11242 [==============================] - 1s 57us/step - loss: 2.8698 - acc: 0.3005\n",
      "Epoch 17/50\n",
      "11242/11242 [==============================] - 1s 54us/step - loss: 2.8512 - acc: 0.3038\n",
      "Epoch 18/50\n",
      "11242/11242 [==============================] - 1s 53us/step - loss: 2.8340 - acc: 0.3057\n",
      "Epoch 19/50\n",
      "11242/11242 [==============================] - 1s 60us/step - loss: 2.8199 - acc: 0.3087\n",
      "Epoch 20/50\n",
      "11242/11242 [==============================] - 1s 54us/step - loss: 2.8037 - acc: 0.3105\n",
      "Epoch 21/50\n",
      "11242/11242 [==============================] - 1s 57us/step - loss: 2.7891 - acc: 0.3137\n",
      "Epoch 22/50\n",
      "11242/11242 [==============================] - 1s 54us/step - loss: 2.7781 - acc: 0.3172\n",
      "Epoch 23/50\n",
      "11242/11242 [==============================] - 1s 58us/step - loss: 2.7645 - acc: 0.3192\n",
      "Epoch 24/50\n",
      "11242/11242 [==============================] - 1s 57us/step - loss: 2.7538 - acc: 0.3218\n",
      "Epoch 25/50\n",
      "11242/11242 [==============================] - 1s 56us/step - loss: 2.7413 - acc: 0.3229\n",
      "Epoch 26/50\n",
      "11242/11242 [==============================] - 1s 61us/step - loss: 2.7310 - acc: 0.3262\n",
      "Epoch 27/50\n",
      "11242/11242 [==============================] - 1s 55us/step - loss: 2.7198 - acc: 0.3290\n",
      "Epoch 28/50\n",
      "11242/11242 [==============================] - 1s 60us/step - loss: 2.7109 - acc: 0.3276\n",
      "Epoch 29/50\n",
      "11242/11242 [==============================] - 1s 56us/step - loss: 2.7005 - acc: 0.3317\n",
      "Epoch 30/50\n",
      "11242/11242 [==============================] - 1s 56us/step - loss: 2.6919 - acc: 0.3321\n",
      "Epoch 31/50\n",
      "11242/11242 [==============================] - 1s 58us/step - loss: 2.6840 - acc: 0.3309\n",
      "Epoch 32/50\n",
      "11242/11242 [==============================] - 1s 61us/step - loss: 2.6738 - acc: 0.3367\n",
      "Epoch 33/50\n",
      "11242/11242 [==============================] - 1s 65us/step - loss: 2.6672 - acc: 0.3365\n",
      "Epoch 34/50\n",
      "11242/11242 [==============================] - 1s 55us/step - loss: 2.6585 - acc: 0.3367\n",
      "Epoch 35/50\n",
      "11242/11242 [==============================] - 1s 54us/step - loss: 2.6506 - acc: 0.3388\n",
      "Epoch 36/50\n",
      "11242/11242 [==============================] - 1s 54us/step - loss: 2.6419 - acc: 0.3392\n",
      "Epoch 37/50\n",
      "11242/11242 [==============================] - 1s 53us/step - loss: 2.6353 - acc: 0.3405\n",
      "Epoch 38/50\n",
      "11242/11242 [==============================] - 1s 48us/step - loss: 2.6278 - acc: 0.3428\n",
      "Epoch 39/50\n",
      "11242/11242 [==============================] - 1s 50us/step - loss: 2.6211 - acc: 0.3424\n",
      "Epoch 40/50\n",
      "11242/11242 [==============================] - 1s 53us/step - loss: 2.6146 - acc: 0.3442\n",
      "Epoch 41/50\n",
      "11242/11242 [==============================] - 1s 58us/step - loss: 2.6082 - acc: 0.3457\n",
      "Epoch 42/50\n",
      "11242/11242 [==============================] - 1s 58us/step - loss: 2.6023 - acc: 0.3452\n",
      "Epoch 43/50\n",
      "11242/11242 [==============================] - 1s 55us/step - loss: 2.5955 - acc: 0.3475\n",
      "Epoch 44/50\n",
      "11242/11242 [==============================] - 1s 54us/step - loss: 2.5889 - acc: 0.3480\n",
      "Epoch 45/50\n",
      "11242/11242 [==============================] - 1s 56us/step - loss: 2.5836 - acc: 0.3481\n",
      "Epoch 46/50\n",
      "11242/11242 [==============================] - 1s 57us/step - loss: 2.5767 - acc: 0.3508\n",
      "Epoch 47/50\n",
      "11242/11242 [==============================] - 1s 59us/step - loss: 2.5726 - acc: 0.3498\n",
      "Epoch 48/50\n",
      "11242/11242 [==============================] - 1s 62us/step - loss: 2.5659 - acc: 0.3498\n",
      "Epoch 49/50\n",
      "11242/11242 [==============================] - 1s 51us/step - loss: 2.5604 - acc: 0.3523\n",
      "Epoch 50/50\n",
      "11242/11242 [==============================] - 1s 58us/step - loss: 2.5553 - acc: 0.3515\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "513/513 [==============================] - 0s 331us/step - loss: 4.3686 - acc: 0.0175\n",
      "Epoch 2/50\n",
      "513/513 [==============================] - 0s 43us/step - loss: 3.9466 - acc: 0.0526\n",
      "Epoch 3/50\n",
      "513/513 [==============================] - 0s 43us/step - loss: 3.7122 - acc: 0.0663\n",
      "Epoch 4/50\n",
      "513/513 [==============================] - 0s 45us/step - loss: 3.5376 - acc: 0.0897\n",
      "Epoch 5/50\n",
      "513/513 [==============================] - 0s 45us/step - loss: 3.3718 - acc: 0.2612\n",
      "Epoch 6/50\n",
      "513/513 [==============================] - 0s 63us/step - loss: 3.2583 - acc: 0.2612\n",
      "Epoch 7/50\n",
      "513/513 [==============================] - 0s 57us/step - loss: 3.1951 - acc: 0.2612\n",
      "Epoch 8/50\n",
      "513/513 [==============================] - 0s 65us/step - loss: 3.1449 - acc: 0.2612\n",
      "Epoch 9/50\n",
      "513/513 [==============================] - 0s 64us/step - loss: 3.1080 - acc: 0.2612\n",
      "Epoch 10/50\n",
      "513/513 [==============================] - 0s 61us/step - loss: 3.0793 - acc: 0.2710\n",
      "Epoch 11/50\n",
      "513/513 [==============================] - 0s 55us/step - loss: 3.0622 - acc: 0.2690\n",
      "Epoch 12/50\n",
      "513/513 [==============================] - 0s 59us/step - loss: 3.0489 - acc: 0.2749\n",
      "Epoch 13/50\n",
      "513/513 [==============================] - 0s 60us/step - loss: 3.0324 - acc: 0.2729\n",
      "Epoch 14/50\n",
      "513/513 [==============================] - 0s 58us/step - loss: 3.0185 - acc: 0.2768\n",
      "Epoch 15/50\n",
      "513/513 [==============================] - 0s 53us/step - loss: 3.0104 - acc: 0.2749\n",
      "Epoch 16/50\n",
      "513/513 [==============================] - 0s 54us/step - loss: 2.9997 - acc: 0.2768\n",
      "Epoch 17/50\n",
      "513/513 [==============================] - 0s 63us/step - loss: 2.9915 - acc: 0.2885\n",
      "Epoch 18/50\n",
      "513/513 [==============================] - 0s 64us/step - loss: 2.9901 - acc: 0.2904\n",
      "Epoch 19/50\n",
      "513/513 [==============================] - 0s 59us/step - loss: 2.9885 - acc: 0.2768\n",
      "Epoch 20/50\n",
      "513/513 [==============================] - 0s 56us/step - loss: 2.9900 - acc: 0.2339\n",
      "Epoch 21/50\n",
      "513/513 [==============================] - 0s 59us/step - loss: 2.9758 - acc: 0.2710\n",
      "Epoch 22/50\n",
      "513/513 [==============================] - 0s 56us/step - loss: 2.9630 - acc: 0.2885\n",
      "Epoch 23/50\n",
      "513/513 [==============================] - 0s 51us/step - loss: 2.9575 - acc: 0.2943\n",
      "Epoch 24/50\n",
      "513/513 [==============================] - 0s 58us/step - loss: 2.9479 - acc: 0.2924\n",
      "Epoch 25/50\n",
      "513/513 [==============================] - 0s 61us/step - loss: 2.9390 - acc: 0.2846\n",
      "Epoch 26/50\n",
      "513/513 [==============================] - 0s 54us/step - loss: 2.9332 - acc: 0.2768\n",
      "Epoch 27/50\n",
      "513/513 [==============================] - 0s 52us/step - loss: 2.9279 - acc: 0.2768\n",
      "Epoch 28/50\n",
      "513/513 [==============================] - 0s 49us/step - loss: 2.9280 - acc: 0.2749\n",
      "Epoch 29/50\n",
      "513/513 [==============================] - 0s 46us/step - loss: 2.9186 - acc: 0.2768\n",
      "Epoch 30/50\n",
      "513/513 [==============================] - 0s 45us/step - loss: 2.9129 - acc: 0.2749\n",
      "Epoch 31/50\n",
      "513/513 [==============================] - 0s 59us/step - loss: 2.9047 - acc: 0.2749\n",
      "Epoch 32/50\n",
      "513/513 [==============================] - 0s 63us/step - loss: 2.8969 - acc: 0.2749\n",
      "Epoch 33/50\n",
      "513/513 [==============================] - 0s 58us/step - loss: 2.8899 - acc: 0.2749\n",
      "Epoch 34/50\n",
      "513/513 [==============================] - 0s 58us/step - loss: 2.8899 - acc: 0.2749\n",
      "Epoch 35/50\n",
      "513/513 [==============================] - 0s 62us/step - loss: 2.8806 - acc: 0.2807\n",
      "Epoch 36/50\n",
      "513/513 [==============================] - 0s 59us/step - loss: 2.8788 - acc: 0.2729\n",
      "Epoch 37/50\n",
      "513/513 [==============================] - 0s 53us/step - loss: 2.8740 - acc: 0.2729\n",
      "Epoch 38/50\n",
      "513/513 [==============================] - 0s 55us/step - loss: 2.8665 - acc: 0.2788\n",
      "Epoch 39/50\n",
      "513/513 [==============================] - 0s 61us/step - loss: 2.8574 - acc: 0.2885\n",
      "Epoch 40/50\n",
      "513/513 [==============================] - 0s 60us/step - loss: 2.8507 - acc: 0.2865\n",
      "Epoch 41/50\n",
      "513/513 [==============================] - 0s 60us/step - loss: 2.8461 - acc: 0.2865\n",
      "Epoch 42/50\n",
      "513/513 [==============================] - 0s 59us/step - loss: 2.8439 - acc: 0.2885\n",
      "Epoch 43/50\n",
      "513/513 [==============================] - 0s 69us/step - loss: 2.8423 - acc: 0.2865\n",
      "Epoch 44/50\n",
      "513/513 [==============================] - 0s 63us/step - loss: 2.8355 - acc: 0.2788\n",
      "Epoch 45/50\n",
      "513/513 [==============================] - 0s 60us/step - loss: 2.8276 - acc: 0.2788\n",
      "Epoch 46/50\n",
      "513/513 [==============================] - 0s 62us/step - loss: 2.8270 - acc: 0.2807\n",
      "Epoch 47/50\n",
      "513/513 [==============================] - 0s 66us/step - loss: 2.8238 - acc: 0.2865\n",
      "Epoch 48/50\n",
      "513/513 [==============================] - 0s 63us/step - loss: 2.8169 - acc: 0.2885\n",
      "Epoch 49/50\n",
      "513/513 [==============================] - 0s 65us/step - loss: 2.8186 - acc: 0.2904\n",
      "Epoch 50/50\n",
      "513/513 [==============================] - 0s 60us/step - loss: 2.8102 - acc: 0.2788\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "3/3 [==============================] - 0s 74ms/step - loss: 1.4559 - acc: 0.3333\n",
      "Epoch 2/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.3358 - acc: 0.3333\n",
      "Epoch 3/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.2622 - acc: 0.3333\n",
      "Epoch 4/50\n",
      "3/3 [==============================] - 0s 1ms/step - loss: 1.2103 - acc: 0.3333\n",
      "Epoch 5/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.1722 - acc: 0.3333\n",
      "Epoch 6/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.1437 - acc: 0.3333\n",
      "Epoch 7/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.1224 - acc: 0.3333\n",
      "Epoch 8/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.1063 - acc: 0.3333\n",
      "Epoch 9/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0941 - acc: 0.3333\n",
      "Epoch 10/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0848 - acc: 0.3333\n",
      "Epoch 11/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0776 - acc: 0.3333\n",
      "Epoch 12/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0717 - acc: 0.6667\n",
      "Epoch 13/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0668 - acc: 0.6667\n",
      "Epoch 14/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0626 - acc: 0.6667\n",
      "Epoch 15/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0587 - acc: 0.6667\n",
      "Epoch 16/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0551 - acc: 0.6667\n",
      "Epoch 17/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0517 - acc: 0.6667\n",
      "Epoch 18/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0484 - acc: 0.6667\n",
      "Epoch 19/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0451 - acc: 1.0000\n",
      "Epoch 20/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0419 - acc: 1.0000\n",
      "Epoch 21/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0387 - acc: 1.0000\n",
      "Epoch 22/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0355 - acc: 1.0000\n",
      "Epoch 23/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0323 - acc: 1.0000\n",
      "Epoch 24/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0292 - acc: 1.0000\n",
      "Epoch 25/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0260 - acc: 1.0000\n",
      "Epoch 26/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0228 - acc: 1.0000\n",
      "Epoch 27/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0197 - acc: 1.0000\n",
      "Epoch 28/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0165 - acc: 1.0000\n",
      "Epoch 29/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0133 - acc: 1.0000\n",
      "Epoch 30/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0101 - acc: 1.0000\n",
      "Epoch 31/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0069 - acc: 1.0000\n",
      "Epoch 32/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0037 - acc: 1.0000\n",
      "Epoch 33/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 1.0004 - acc: 1.0000\n",
      "Epoch 34/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 0.9972 - acc: 1.0000\n",
      "Epoch 35/50\n",
      "3/3 [==============================] - 0s 3ms/step - loss: 0.9939 - acc: 1.0000\n",
      "Epoch 36/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 0.9906 - acc: 1.0000\n",
      "Epoch 37/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 0.9873 - acc: 1.0000\n",
      "Epoch 38/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 0.9840 - acc: 1.0000\n",
      "Epoch 39/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 0.9807 - acc: 1.0000\n",
      "Epoch 40/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 0.9774 - acc: 1.0000\n",
      "Epoch 41/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 0.9740 - acc: 1.0000\n",
      "Epoch 42/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 0.9707 - acc: 1.0000\n",
      "Epoch 43/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 0.9673 - acc: 1.0000\n",
      "Epoch 44/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 0.9639 - acc: 1.0000\n",
      "Epoch 45/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 0.9604 - acc: 1.0000\n",
      "Epoch 46/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 0.9570 - acc: 1.0000\n",
      "Epoch 47/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 0.9536 - acc: 1.0000\n",
      "Epoch 48/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 0.9501 - acc: 1.0000\n",
      "Epoch 49/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 0.9466 - acc: 1.0000\n",
      "Epoch 50/50\n",
      "3/3 [==============================] - 0s 2ms/step - loss: 0.9431 - acc: 1.0000\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "9/9 [==============================] - 0s 18ms/step - loss: 1.1247 - acc: 0.2222\n",
      "Epoch 2/50\n",
      "9/9 [==============================] - 0s 668us/step - loss: 1.0063 - acc: 0.2222\n",
      "Epoch 3/50\n",
      "9/9 [==============================] - 0s 607us/step - loss: 0.9469 - acc: 0.6667\n",
      "Epoch 4/50\n",
      "9/9 [==============================] - 0s 488us/step - loss: 0.9106 - acc: 0.6667\n",
      "Epoch 5/50\n",
      "9/9 [==============================] - 0s 489us/step - loss: 0.8867 - acc: 0.6667\n",
      "Epoch 6/50\n",
      "9/9 [==============================] - 0s 802us/step - loss: 0.8704 - acc: 0.6667\n",
      "Epoch 7/50\n",
      "9/9 [==============================] - 0s 668us/step - loss: 0.8589 - acc: 0.6667\n",
      "Epoch 8/50\n",
      "9/9 [==============================] - 0s 724us/step - loss: 0.8507 - acc: 0.6667\n",
      "Epoch 9/50\n",
      "9/9 [==============================] - 0s 675us/step - loss: 0.8447 - acc: 0.6667\n",
      "Epoch 10/50\n",
      "9/9 [==============================] - 0s 604us/step - loss: 0.8402 - acc: 0.6667\n",
      "Epoch 11/50\n",
      "9/9 [==============================] - 0s 723us/step - loss: 0.8368 - acc: 0.6667\n",
      "Epoch 12/50\n",
      "9/9 [==============================] - 0s 633us/step - loss: 0.8341 - acc: 0.6667\n",
      "Epoch 13/50\n",
      "9/9 [==============================] - 0s 664us/step - loss: 0.8319 - acc: 0.6667\n",
      "Epoch 14/50\n",
      "9/9 [==============================] - 0s 701us/step - loss: 0.8301 - acc: 0.6667\n",
      "Epoch 15/50\n",
      "9/9 [==============================] - 0s 668us/step - loss: 0.8285 - acc: 0.6667\n",
      "Epoch 16/50\n",
      "9/9 [==============================] - 0s 622us/step - loss: 0.8271 - acc: 0.6667\n",
      "Epoch 17/50\n",
      "9/9 [==============================] - 0s 621us/step - loss: 0.8257 - acc: 0.6667\n",
      "Epoch 18/50\n",
      "9/9 [==============================] - 0s 757us/step - loss: 0.8245 - acc: 0.6667\n",
      "Epoch 19/50\n",
      "9/9 [==============================] - 0s 706us/step - loss: 0.8233 - acc: 0.6667\n",
      "Epoch 20/50\n",
      "9/9 [==============================] - 0s 690us/step - loss: 0.8221 - acc: 0.6667\n",
      "Epoch 21/50\n",
      "9/9 [==============================] - 0s 670us/step - loss: 0.8210 - acc: 0.6667\n",
      "Epoch 22/50\n",
      "9/9 [==============================] - 0s 669us/step - loss: 0.8198 - acc: 0.6667\n",
      "Epoch 23/50\n",
      "9/9 [==============================] - 0s 634us/step - loss: 0.8187 - acc: 0.6667\n",
      "Epoch 24/50\n",
      "9/9 [==============================] - 0s 624us/step - loss: 0.8175 - acc: 0.6667\n",
      "Epoch 25/50\n",
      "9/9 [==============================] - 0s 799us/step - loss: 0.8164 - acc: 0.6667\n",
      "Epoch 26/50\n",
      "9/9 [==============================] - 0s 569us/step - loss: 0.8153 - acc: 0.6667\n",
      "Epoch 27/50\n",
      "9/9 [==============================] - 0s 565us/step - loss: 0.8141 - acc: 0.6667\n",
      "Epoch 28/50\n",
      "9/9 [==============================] - 0s 525us/step - loss: 0.8130 - acc: 0.6667\n",
      "Epoch 29/50\n",
      "9/9 [==============================] - 0s 590us/step - loss: 0.8118 - acc: 0.6667\n",
      "Epoch 30/50\n",
      "9/9 [==============================] - 0s 732us/step - loss: 0.8107 - acc: 0.6667\n",
      "Epoch 31/50\n",
      "9/9 [==============================] - 0s 574us/step - loss: 0.8095 - acc: 0.6667\n",
      "Epoch 32/50\n",
      "9/9 [==============================] - 0s 570us/step - loss: 0.8084 - acc: 0.6667\n",
      "Epoch 33/50\n",
      "9/9 [==============================] - 0s 570us/step - loss: 0.8072 - acc: 0.6667\n",
      "Epoch 34/50\n",
      "9/9 [==============================] - 0s 599us/step - loss: 0.8060 - acc: 0.6667\n",
      "Epoch 35/50\n",
      "9/9 [==============================] - 0s 599us/step - loss: 0.8049 - acc: 0.6667\n",
      "Epoch 36/50\n",
      "9/9 [==============================] - 0s 634us/step - loss: 0.8037 - acc: 0.6667\n",
      "Epoch 37/50\n",
      "9/9 [==============================] - 0s 730us/step - loss: 0.8025 - acc: 0.6667\n",
      "Epoch 38/50\n",
      "9/9 [==============================] - 0s 635us/step - loss: 0.8013 - acc: 0.6667\n",
      "Epoch 39/50\n",
      "9/9 [==============================] - 0s 710us/step - loss: 0.8001 - acc: 0.6667\n",
      "Epoch 40/50\n",
      "9/9 [==============================] - 0s 693us/step - loss: 0.7989 - acc: 0.6667\n",
      "Epoch 41/50\n",
      "9/9 [==============================] - 0s 685us/step - loss: 0.7977 - acc: 0.6667\n",
      "Epoch 42/50\n",
      "9/9 [==============================] - 0s 639us/step - loss: 0.7965 - acc: 0.6667\n",
      "Epoch 43/50\n",
      "9/9 [==============================] - 0s 597us/step - loss: 0.7953 - acc: 0.6667\n",
      "Epoch 44/50\n",
      "9/9 [==============================] - 0s 548us/step - loss: 0.7941 - acc: 0.6667\n",
      "Epoch 45/50\n",
      "9/9 [==============================] - 0s 564us/step - loss: 0.7928 - acc: 0.6667\n",
      "Epoch 46/50\n",
      "9/9 [==============================] - 0s 572us/step - loss: 0.7916 - acc: 0.6667\n",
      "Epoch 47/50\n",
      "9/9 [==============================] - 0s 571us/step - loss: 0.7904 - acc: 0.6667\n",
      "Epoch 48/50\n",
      "9/9 [==============================] - 0s 557us/step - loss: 0.7891 - acc: 0.6667\n",
      "Epoch 49/50\n",
      "9/9 [==============================] - 0s 505us/step - loss: 0.7879 - acc: 0.6667\n",
      "Epoch 50/50\n",
      "9/9 [==============================] - 0s 504us/step - loss: 0.7867 - acc: 0.6667\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "4208/4208 [==============================] - 0s 99us/step - loss: 4.2072 - acc: 0.0965\n",
      "Epoch 2/50\n",
      "4208/4208 [==============================] - 0s 62us/step - loss: 3.1862 - acc: 0.1500\n",
      "Epoch 3/50\n",
      "4208/4208 [==============================] - 0s 63us/step - loss: 3.0123 - acc: 0.2058\n",
      "Epoch 4/50\n",
      "4208/4208 [==============================] - 0s 62us/step - loss: 2.9680 - acc: 0.2203\n",
      "Epoch 5/50\n",
      "4208/4208 [==============================] - 0s 65us/step - loss: 2.9388 - acc: 0.2255\n",
      "Epoch 6/50\n",
      "4208/4208 [==============================] - 0s 61us/step - loss: 2.9104 - acc: 0.2600\n",
      "Epoch 7/50\n",
      "4208/4208 [==============================] - 0s 56us/step - loss: 2.8865 - acc: 0.2681\n",
      "Epoch 8/50\n",
      "4208/4208 [==============================] - 0s 59us/step - loss: 2.8661 - acc: 0.2757\n",
      "Epoch 9/50\n",
      "4208/4208 [==============================] - 0s 56us/step - loss: 2.8467 - acc: 0.2890\n",
      "Epoch 10/50\n",
      "4208/4208 [==============================] - 0s 58us/step - loss: 2.8249 - acc: 0.2956\n",
      "Epoch 11/50\n",
      "4208/4208 [==============================] - 0s 48us/step - loss: 2.8061 - acc: 0.2966\n",
      "Epoch 12/50\n",
      "4208/4208 [==============================] - 0s 56us/step - loss: 2.7868 - acc: 0.3032\n",
      "Epoch 13/50\n",
      "4208/4208 [==============================] - 0s 56us/step - loss: 2.7695 - acc: 0.3056\n",
      "Epoch 14/50\n",
      "4208/4208 [==============================] - 0s 49us/step - loss: 2.7510 - acc: 0.3094\n",
      "Epoch 15/50\n",
      "4208/4208 [==============================] - 0s 57us/step - loss: 2.7334 - acc: 0.3089\n",
      "12636/12636 [==============================] - 1s 83us/step - loss: 3.2352 - acc: 0.2806\n",
      "Epoch 40/50\n",
      "12636/12636 [==============================] - 1s 85us/step - loss: 3.2205 - acc: 0.2843\n",
      "Epoch 41/50\n",
      "12636/12636 [==============================] - 1s 81us/step - loss: 3.2083 - acc: 0.2863\n",
      "Epoch 42/50\n",
      "12636/12636 [==============================] - 1s 82us/step - loss: 3.1950 - acc: 0.2893\n",
      "Epoch 43/50\n",
      "12636/12636 [==============================] - 1s 82us/step - loss: 3.1823 - acc: 0.2897\n",
      "Epoch 44/50\n",
      "12636/12636 [==============================] - 1s 89us/step - loss: 3.1704 - acc: 0.2904\n",
      "Epoch 45/50\n",
      "12636/12636 [==============================] - 1s 86us/step - loss: 3.1579 - acc: 0.2878\n",
      "Epoch 46/50\n",
      "12636/12636 [==============================] - 1s 83us/step - loss: 3.1477 - acc: 0.2912\n",
      "Epoch 47/50\n",
      "12636/12636 [==============================] - 1s 81us/step - loss: 3.1370 - acc: 0.2919\n",
      "Epoch 48/50\n",
      "12636/12636 [==============================] - 1s 79us/step - loss: 3.1249 - acc: 0.2927\n",
      "Epoch 49/50\n",
      "12636/12636 [==============================] - 1s 79us/step - loss: 3.1152 - acc: 0.2919\n",
      "Epoch 50/50\n",
      "12636/12636 [==============================] - 1s 85us/step - loss: 3.1040 - acc: 0.2954\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "92/92 [==============================] - 0s 3ms/step - loss: 4.0555 - acc: 0.0109\n",
      "Epoch 2/50\n",
      "92/92 [==============================] - 0s 63us/step - loss: 3.8768 - acc: 0.0109\n",
      "Epoch 3/50\n",
      "92/92 [==============================] - 0s 59us/step - loss: 3.7685 - acc: 0.0109\n",
      "Epoch 4/50\n",
      "92/92 [==============================] - 0s 59us/step - loss: 3.6874 - acc: 0.0109\n",
      "Epoch 5/50\n",
      "92/92 [==============================] - 0s 56us/step - loss: 3.6220 - acc: 0.0109\n",
      "Epoch 6/50\n",
      "92/92 [==============================] - 0s 87us/step - loss: 3.5668 - acc: 0.0109\n",
      "Epoch 7/50\n",
      "92/92 [==============================] - 0s 67us/step - loss: 3.5190 - acc: 0.0109\n",
      "Epoch 8/50\n",
      "92/92 [==============================] - 0s 68us/step - loss: 3.4767 - acc: 0.0109\n",
      "Epoch 9/50\n",
      "92/92 [==============================] - 0s 69us/step - loss: 3.4389 - acc: 0.0109\n",
      "Epoch 10/50\n",
      "92/92 [==============================] - 0s 85us/step - loss: 3.4048 - acc: 0.0109\n",
      "Epoch 11/50\n",
      "92/92 [==============================] - 0s 73us/step - loss: 3.3737 - acc: 0.0109\n",
      "Epoch 12/50\n",
      "92/92 [==============================] - 0s 71us/step - loss: 3.3452 - acc: 0.0435\n",
      "Epoch 13/50\n",
      "92/92 [==============================] - 0s 76us/step - loss: 3.3191 - acc: 0.0543\n",
      "Epoch 14/50\n",
      "92/92 [==============================] - 0s 74us/step - loss: 3.2949 - acc: 0.0870\n",
      "Epoch 15/50\n",
      "92/92 [==============================] - 0s 72us/step - loss: 3.2726 - acc: 0.0870\n",
      "Epoch 16/50\n",
      "92/92 [==============================] - 0s 71us/step - loss: 3.2520 - acc: 0.0978\n",
      "Epoch 17/50\n",
      "92/92 [==============================] - 0s 64us/step - loss: 3.2329 - acc: 0.1087\n",
      "Epoch 18/50\n",
      "92/92 [==============================] - 0s 72us/step - loss: 3.2151 - acc: 0.1087\n",
      "Epoch 19/50\n",
      "92/92 [==============================] - 0s 73us/step - loss: 3.1987 - acc: 0.1087\n",
      "Epoch 20/50\n",
      "92/92 [==============================] - 0s 79us/step - loss: 3.1834 - acc: 0.1087\n",
      "Epoch 21/50\n",
      "92/92 [==============================] - 0s 74us/step - loss: 3.1693 - acc: 0.1087\n",
      "Epoch 22/50\n",
      "92/92 [==============================] - 0s 71us/step - loss: 3.1562 - acc: 0.1087\n",
      "Epoch 23/50\n",
      "92/92 [==============================] - 0s 79us/step - loss: 3.1440 - acc: 0.1522\n",
      "Epoch 24/50\n",
      "92/92 [==============================] - 0s 74us/step - loss: 3.1327 - acc: 0.2609\n",
      "Epoch 25/50\n",
      "92/92 [==============================] - 0s 69us/step - loss: 3.1222 - acc: 0.2826\n",
      "Epoch 26/50\n",
      "92/92 [==============================] - 0s 74us/step - loss: 3.1125 - acc: 0.2391\n",
      "Epoch 27/50\n",
      "92/92 [==============================] - 0s 67us/step - loss: 3.1034 - acc: 0.1739\n",
      "Epoch 28/50\n",
      "92/92 [==============================] - 0s 68us/step - loss: 3.0949 - acc: 0.1739\n",
      "Epoch 29/50\n",
      "92/92 [==============================] - 0s 60us/step - loss: 3.0871 - acc: 0.1739\n",
      "Epoch 30/50\n",
      "92/92 [==============================] - 0s 71us/step - loss: 3.0797 - acc: 0.1739\n",
      "Epoch 31/50\n",
      "92/92 [==============================] - 0s 85us/step - loss: 3.0727 - acc: 0.1739\n",
      "Epoch 32/50\n",
      "92/92 [==============================] - 0s 82us/step - loss: 3.0662 - acc: 0.1739\n",
      "Epoch 33/50\n",
      "92/92 [==============================] - 0s 75us/step - loss: 3.0600 - acc: 0.1739\n",
      "Epoch 34/50\n",
      "92/92 [==============================] - 0s 71us/step - loss: 3.0542 - acc: 0.1739\n",
      "Epoch 35/50\n",
      "92/92 [==============================] - 0s 70us/step - loss: 3.0486 - acc: 0.1739\n",
      "Epoch 36/50\n",
      "92/92 [==============================] - 0s 81us/step - loss: 3.0433 - acc: 0.1739\n",
      "Epoch 37/50\n",
      "92/92 [==============================] - 0s 77us/step - loss: 3.0382 - acc: 0.1739\n",
      "Epoch 38/50\n",
      "92/92 [==============================] - 0s 82us/step - loss: 3.0333 - acc: 0.1739\n",
      "Epoch 39/50\n",
      "92/92 [==============================] - 0s 82us/step - loss: 3.0285 - acc: 0.1739\n",
      "Epoch 40/50\n",
      "92/92 [==============================] - 0s 74us/step - loss: 3.0239 - acc: 0.1739\n",
      "Epoch 41/50\n",
      "92/92 [==============================] - 0s 77us/step - loss: 3.0194 - acc: 0.1739\n",
      "Epoch 42/50\n",
      "92/92 [==============================] - 0s 75us/step - loss: 3.0151 - acc: 0.1739\n",
      "Epoch 43/50\n",
      "92/92 [==============================] - 0s 76us/step - loss: 3.0108 - acc: 0.1739\n",
      "Epoch 44/50\n",
      "92/92 [==============================] - 0s 70us/step - loss: 3.0066 - acc: 0.1739\n",
      "Epoch 45/50\n",
      "92/92 [==============================] - 0s 70us/step - loss: 3.0025 - acc: 0.1848\n",
      "Epoch 46/50\n",
      "92/92 [==============================] - 0s 71us/step - loss: 2.9985 - acc: 0.1848\n",
      "Epoch 47/50\n",
      "92/92 [==============================] - 0s 66us/step - loss: 2.9945 - acc: 0.1848\n",
      "Epoch 48/50\n",
      "92/92 [==============================] - 0s 66us/step - loss: 2.9905 - acc: 0.1848\n",
      "Epoch 49/50\n",
      "92/92 [==============================] - 0s 67us/step - loss: 2.9866 - acc: 0.1848\n",
      "Epoch 50/50\n",
      "92/92 [==============================] - 0s 71us/step - loss: 2.9828 - acc: 0.1848\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "194/194 [==============================] - 0s 1ms/step - loss: 3.5743 - acc: 0.1443\n",
      "Epoch 2/50\n",
      "194/194 [==============================] - 0s 73us/step - loss: 3.3046 - acc: 0.1753\n",
      "Epoch 3/50\n",
      "194/194 [==============================] - 0s 74us/step - loss: 3.1604 - acc: 0.1753\n",
      "Epoch 4/50\n",
      "194/194 [==============================] - 0s 71us/step - loss: 3.0613 - acc: 0.1701\n",
      "Epoch 5/50\n",
      "194/194 [==============================] - 0s 70us/step - loss: 2.9848 - acc: 0.1598\n",
      "Epoch 6/50\n",
      "194/194 [==============================] - 0s 91us/step - loss: 2.9244 - acc: 0.1598\n",
      "Epoch 7/50\n",
      "194/194 [==============================] - 0s 70us/step - loss: 2.8773 - acc: 0.1495\n",
      "Epoch 8/50\n",
      "194/194 [==============================] - 0s 70us/step - loss: 2.8375 - acc: 0.1598\n",
      "Epoch 9/50\n",
      "194/194 [==============================] - 0s 68us/step - loss: 2.8044 - acc: 0.1753\n",
      "Epoch 10/50\n",
      "194/194 [==============================] - 0s 68us/step - loss: 2.7732 - acc: 0.1959\n",
      "Epoch 11/50\n",
      "194/194 [==============================] - 0s 77us/step - loss: 2.7470 - acc: 0.1856\n",
      "Epoch 12/50\n",
      "194/194 [==============================] - 0s 67us/step - loss: 2.7253 - acc: 0.2010\n",
      "Epoch 13/50\n",
      "194/194 [==============================] - 0s 69us/step - loss: 2.7058 - acc: 0.1959\n",
      "Epoch 14/50\n",
      "194/194 [==============================] - 0s 70us/step - loss: 2.6875 - acc: 0.2062\n",
      "Epoch 15/50\n",
      "194/194 [==============================] - 0s 71us/step - loss: 2.6735 - acc: 0.1959\n",
      "Epoch 16/50\n",
      "194/194 [==============================] - 0s 81us/step - loss: 2.6610 - acc: 0.2165\n",
      "Epoch 17/50\n",
      "194/194 [==============================] - 0s 75us/step - loss: 2.6530 - acc: 0.2165\n",
      "Epoch 18/50\n",
      "194/194 [==============================] - 0s 76us/step - loss: 2.6446 - acc: 0.2113\n",
      "Epoch 19/50\n",
      "194/194 [==============================] - 0s 79us/step - loss: 2.6385 - acc: 0.2165\n",
      "Epoch 20/50\n",
      "194/194 [==============================] - 0s 77us/step - loss: 2.6301 - acc: 0.2216\n",
      "Epoch 21/50\n",
      "194/194 [==============================] - 0s 78us/step - loss: 2.6235 - acc: 0.2268\n",
      "Epoch 22/50\n",
      "194/194 [==============================] - 0s 75us/step - loss: 2.6195 - acc: 0.2165\n",
      "Epoch 23/50\n",
      "194/194 [==============================] - 0s 69us/step - loss: 2.6147 - acc: 0.2423\n",
      "Epoch 24/50\n",
      "194/194 [==============================] - 0s 61us/step - loss: 2.6141 - acc: 0.2526\n",
      "Epoch 25/50\n",
      "194/194 [==============================] - 0s 69us/step - loss: 2.6067 - acc: 0.2526\n",
      "Epoch 26/50\n",
      "194/194 [==============================] - 0s 74us/step - loss: 2.6031 - acc: 0.2887\n",
      "Epoch 27/50\n",
      "194/194 [==============================] - 0s 72us/step - loss: 2.5999 - acc: 0.3041\n",
      "Epoch 28/50\n",
      "194/194 [==============================] - 0s 73us/step - loss: 2.5950 - acc: 0.2680\n",
      "Epoch 29/50\n",
      "194/194 [==============================] - 0s 59us/step - loss: 2.5920 - acc: 0.2784\n",
      "Epoch 30/50\n",
      "194/194 [==============================] - 0s 58us/step - loss: 2.5889 - acc: 0.2887\n",
      "Epoch 31/50\n",
      "194/194 [==============================] - 0s 73us/step - loss: 2.5863 - acc: 0.2835\n",
      "Epoch 32/50\n",
      "194/194 [==============================] - 0s 62us/step - loss: 2.5824 - acc: 0.2938\n",
      "Epoch 33/50\n",
      "194/194 [==============================] - 0s 72us/step - loss: 2.5798 - acc: 0.3144\n",
      "Epoch 34/50\n",
      "194/194 [==============================] - 0s 66us/step - loss: 2.5750 - acc: 0.2887\n",
      "Epoch 35/50\n",
      "194/194 [==============================] - 0s 73us/step - loss: 2.5730 - acc: 0.2887\n",
      "Epoch 36/50\n",
      "194/194 [==============================] - 0s 72us/step - loss: 2.5706 - acc: 0.3299\n",
      "Epoch 37/50\n",
      "194/194 [==============================] - 0s 61us/step - loss: 2.5721 - acc: 0.2938\n",
      "Epoch 38/50\n",
      "194/194 [==============================] - 0s 59us/step - loss: 2.5653 - acc: 0.2938\n",
      "Epoch 39/50\n",
      "194/194 [==============================] - 0s 59us/step - loss: 2.5630 - acc: 0.3041\n",
      "Epoch 40/50\n",
      "194/194 [==============================] - 0s 55us/step - loss: 2.5568 - acc: 0.3402\n",
      "Epoch 41/50\n",
      "194/194 [==============================] - 0s 61us/step - loss: 2.5541 - acc: 0.3454\n",
      "Epoch 42/50\n",
      "194/194 [==============================] - 0s 55us/step - loss: 2.5526 - acc: 0.3247\n",
      "Epoch 43/50\n",
      "194/194 [==============================] - 0s 51us/step - loss: 2.5488 - acc: 0.3351\n",
      "Epoch 44/50\n",
      "194/194 [==============================] - 0s 58us/step - loss: 2.5454 - acc: 0.3299\n",
      "Epoch 45/50\n",
      "194/194 [==============================] - 0s 49us/step - loss: 2.5434 - acc: 0.3402\n",
      "Epoch 46/50\n",
      "194/194 [==============================] - 0s 54us/step - loss: 2.5414 - acc: 0.3299\n",
      "Epoch 47/50\n",
      "194/194 [==============================] - 0s 47us/step - loss: 2.5400 - acc: 0.3351\n",
      "Epoch 48/50\n",
      "194/194 [==============================] - 0s 51us/step - loss: 2.5362 - acc: 0.3247\n",
      "Epoch 49/50\n",
      "194/194 [==============================] - 0s 67us/step - loss: 2.5304 - acc: 0.3247\n",
      "Epoch 50/50\n",
      "194/194 [==============================] - 0s 68us/step - loss: 2.5323 - acc: 0.3247\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "5/5 [==============================] - 0s 36ms/step - loss: 0.5164 - acc: 0.8000\n",
      "Epoch 2/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.5057 - acc: 0.8000\n",
      "Epoch 3/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.4955 - acc: 0.8000\n",
      "Epoch 4/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.4876 - acc: 0.8000\n",
      "Epoch 5/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.4808 - acc: 0.8000\n",
      "Epoch 6/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.4746 - acc: 0.8000\n",
      "Epoch 7/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.4689 - acc: 0.8000\n",
      "Epoch 8/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.4635 - acc: 0.8000\n",
      "Epoch 9/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.4583 - acc: 0.8000\n",
      "Epoch 10/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.4534 - acc: 0.8000\n",
      "Epoch 11/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.4486 - acc: 0.8000\n",
      "Epoch 12/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.4440 - acc: 0.8000\n",
      "Epoch 13/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.4394 - acc: 0.8000\n",
      "Epoch 14/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.4350 - acc: 0.8000\n",
      "Epoch 15/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.4307 - acc: 0.8000\n",
      "Epoch 16/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.4265 - acc: 0.8000\n",
      "Epoch 17/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.4224 - acc: 0.8000\n",
      "Epoch 18/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.4183 - acc: 0.8000\n",
      "Epoch 19/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.4143 - acc: 0.8000\n",
      "Epoch 20/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.4103 - acc: 0.8000\n",
      "Epoch 21/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.4064 - acc: 0.8000\n",
      "Epoch 22/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.4026 - acc: 0.8000\n",
      "Epoch 23/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3988 - acc: 0.8000\n",
      "Epoch 24/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3951 - acc: 0.8000\n",
      "Epoch 25/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3913 - acc: 0.8000\n",
      "Epoch 26/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3877 - acc: 0.8000\n",
      "Epoch 27/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3840 - acc: 0.8000\n",
      "Epoch 28/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3805 - acc: 0.8000\n",
      "Epoch 29/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3769 - acc: 0.8000\n",
      "Epoch 30/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3734 - acc: 0.8000\n",
      "Epoch 31/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3699 - acc: 0.8000\n",
      "Epoch 32/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3664 - acc: 0.8000\n",
      "Epoch 33/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3630 - acc: 0.8000\n",
      "Epoch 34/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3596 - acc: 0.8000\n",
      "Epoch 35/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3562 - acc: 0.8000\n",
      "Epoch 36/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3529 - acc: 0.8000\n",
      "Epoch 37/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3496 - acc: 0.8000\n",
      "Epoch 38/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3463 - acc: 0.8000\n",
      "Epoch 39/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3431 - acc: 0.8000\n",
      "Epoch 40/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3401 - acc: 0.8000\n",
      "Epoch 41/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3370 - acc: 0.8000\n",
      "Epoch 42/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3340 - acc: 0.8000\n",
      "Epoch 43/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3308 - acc: 0.8000\n",
      "Epoch 44/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3277 - acc: 0.8000\n",
      "Epoch 45/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3247 - acc: 0.8000\n",
      "Epoch 46/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3217 - acc: 0.8000\n",
      "Epoch 47/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3187 - acc: 0.8000\n",
      "Epoch 48/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3158 - acc: 0.8000\n",
      "Epoch 49/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3128 - acc: 0.8000\n",
      "Epoch 50/50\n",
      "5/5 [==============================] - 0s 1ms/step - loss: 0.3099 - acc: 0.8000\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "2808/2808 [==============================] - 0s 118us/step - loss: 5.4177 - acc: 0.0331\n",
      "Epoch 2/50\n",
      "2808/2808 [==============================] - 0s 57us/step - loss: 4.6868 - acc: 0.1652\n",
      "Epoch 3/50\n",
      "2808/2808 [==============================] - 0s 56us/step - loss: 4.4823 - acc: 0.1905\n",
      "Epoch 4/50\n",
      "2808/2808 [==============================] - 0s 59us/step - loss: 4.4379 - acc: 0.1905\n",
      "Epoch 5/50\n",
      "2808/2808 [==============================] - 0s 58us/step - loss: 4.4156 - acc: 0.1905\n",
      "Epoch 6/50\n",
      "2808/2808 [==============================] - 0s 53us/step - loss: 4.3992 - acc: 0.1905\n",
      "Epoch 7/50\n",
      "2808/2808 [==============================] - 0s 54us/step - loss: 4.3840 - acc: 0.1905\n",
      "Epoch 8/50\n",
      "2808/2808 [==============================] - 0s 46us/step - loss: 4.3693 - acc: 0.1905\n",
      "Epoch 9/50\n",
      "2808/2808 [==============================] - 0s 45us/step - loss: 4.3551 - acc: 0.1912\n",
      "Epoch 10/50\n",
      "2808/2808 [==============================] - 0s 48us/step - loss: 4.3417 - acc: 0.1930\n",
      "Epoch 11/50\n",
      "2808/2808 [==============================] - 0s 53us/step - loss: 4.3289 - acc: 0.1944\n",
      "Epoch 12/50\n",
      "2808/2808 [==============================] - 0s 47us/step - loss: 4.3162 - acc: 0.1966\n",
      "Epoch 13/50\n",
      "2808/2808 [==============================] - 0s 42us/step - loss: 4.3036 - acc: 0.1980\n",
      "Epoch 14/50\n",
      "2808/2808 [==============================] - 0s 43us/step - loss: 4.2926 - acc: 0.1998\n",
      "Epoch 15/50\n",
      "2808/2808 [==============================] - 0s 46us/step - loss: 4.2804 - acc: 0.2019\n",
      "Epoch 16/50\n",
      "2808/2808 [==============================] - 0s 45us/step - loss: 4.2679 - acc: 0.2048\n",
      "Epoch 17/50\n",
      "2808/2808 [==============================] - 0s 52us/step - loss: 4.2570 - acc: 0.2055\n",
      "Epoch 18/50\n",
      "2808/2808 [==============================] - 0s 50us/step - loss: 4.2447 - acc: 0.2058\n",
      "Epoch 19/50\n",
      "2808/2808 [==============================] - 0s 50us/step - loss: 4.2362 - acc: 0.2069\n",
      "Epoch 20/50\n",
      "2808/2808 [==============================] - 0s 48us/step - loss: 4.2245 - acc: 0.2076\n",
      "Epoch 21/50\n",
      "2808/2808 [==============================] - 0s 48us/step - loss: 4.2146 - acc: 0.2080\n",
      "Epoch 22/50\n",
      "2808/2808 [==============================] - 0s 44us/step - loss: 4.2038 - acc: 0.2080\n",
      "Epoch 23/50\n",
      "2808/2808 [==============================] - 0s 55us/step - loss: 4.1938 - acc: 0.2087\n",
      "Epoch 24/50\n",
      "2808/2808 [==============================] - 0s 51us/step - loss: 4.1820 - acc: 0.2090\n",
      "Epoch 25/50\n",
      "2808/2808 [==============================] - 0s 56us/step - loss: 4.1731 - acc: 0.2098\n",
      "Epoch 26/50\n",
      "2808/2808 [==============================] - 0s 51us/step - loss: 4.1627 - acc: 0.2105\n",
      "Epoch 27/50\n",
      "2808/2808 [==============================] - 0s 53us/step - loss: 4.1535 - acc: 0.2094\n",
      "Epoch 28/50\n",
      "2808/2808 [==============================] - 0s 53us/step - loss: 4.1433 - acc: 0.2115\n",
      "Epoch 29/50\n",
      "2808/2808 [==============================] - 0s 52us/step - loss: 4.1330 - acc: 0.2101\n",
      "Epoch 30/50\n",
      "2808/2808 [==============================] - 0s 58us/step - loss: 4.1228 - acc: 0.2126\n",
      "Epoch 31/50\n",
      "2808/2808 [==============================] - 0s 54us/step - loss: 4.1142 - acc: 0.2123\n",
      "Epoch 32/50\n",
      "2808/2808 [==============================] - 0s 57us/step - loss: 4.1034 - acc: 0.2130\n",
      "Epoch 33/50\n",
      "2808/2808 [==============================] - 0s 61us/step - loss: 4.0934 - acc: 0.2155\n",
      "Epoch 34/50\n",
      "2808/2808 [==============================] - 0s 60us/step - loss: 4.0834 - acc: 0.2140\n",
      "Epoch 35/50\n",
      "2808/2808 [==============================] - 0s 58us/step - loss: 4.0741 - acc: 0.2176\n",
      "Epoch 36/50\n",
      "2808/2808 [==============================] - 0s 57us/step - loss: 4.0637 - acc: 0.2201\n",
      "Epoch 37/50\n",
      "2808/2808 [==============================] - 0s 58us/step - loss: 4.0542 - acc: 0.2222\n",
      "Epoch 38/50\n",
      "2808/2808 [==============================] - 0s 55us/step - loss: 4.0440 - acc: 0.2269\n",
      "Epoch 39/50\n",
      "2808/2808 [==============================] - 0s 56us/step - loss: 4.0348 - acc: 0.2329\n",
      "Epoch 40/50\n",
      "2808/2808 [==============================] - 0s 50us/step - loss: 4.0255 - acc: 0.2276\n",
      "Epoch 41/50\n",
      "2808/2808 [==============================] - 0s 54us/step - loss: 4.0156 - acc: 0.2361\n",
      "Epoch 42/50\n",
      "2808/2808 [==============================] - 0s 51us/step - loss: 4.0057 - acc: 0.2386\n",
      "Epoch 43/50\n",
      "2808/2808 [==============================] - 0s 49us/step - loss: 3.9965 - acc: 0.2390\n",
      "Epoch 44/50\n",
      "2808/2808 [==============================] - 0s 52us/step - loss: 3.9859 - acc: 0.2400\n",
      "Epoch 45/50\n",
      "2808/2808 [==============================] - 0s 53us/step - loss: 3.9769 - acc: 0.2454\n",
      "Epoch 46/50\n",
      "2808/2808 [==============================] - 0s 57us/step - loss: 3.9677 - acc: 0.2464\n",
      "Epoch 47/50\n",
      "2808/2808 [==============================] - 0s 54us/step - loss: 3.9583 - acc: 0.2500\n",
      "Epoch 48/50\n",
      "2808/2808 [==============================] - 0s 56us/step - loss: 3.9497 - acc: 0.2486\n",
      "Epoch 49/50\n",
      "2808/2808 [==============================] - 0s 54us/step - loss: 3.9398 - acc: 0.2514\n",
      "Epoch 50/50\n",
      "2808/2808 [==============================] - 0s 57us/step - loss: 3.9306 - acc: 0.2532\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "4794/4794 [==============================] - 1s 105us/step - loss: 4.0317 - acc: 0.1089\n",
      "Epoch 2/50\n",
      "4794/4794 [==============================] - 0s 56us/step - loss: 3.3770 - acc: 0.1792\n",
      "Epoch 3/50\n",
      "4794/4794 [==============================] - 0s 51us/step - loss: 3.2938 - acc: 0.2136\n",
      "Epoch 4/50\n",
      "4794/4794 [==============================] - 0s 49us/step - loss: 3.2553 - acc: 0.2253\n",
      "Epoch 5/50\n",
      "4794/4794 [==============================] - 0s 45us/step - loss: 3.2230 - acc: 0.2441\n",
      "Epoch 6/50\n",
      "4794/4794 [==============================] - 0s 50us/step - loss: 3.1919 - acc: 0.2441\n",
      "Epoch 7/50\n",
      "4794/4794 [==============================] - 0s 49us/step - loss: 3.1638 - acc: 0.2599\n",
      "Epoch 8/50\n",
      "4794/4794 [==============================] - 0s 53us/step - loss: 3.1377 - acc: 0.2639\n",
      "Epoch 9/50\n",
      "4794/4794 [==============================] - 0s 46us/step - loss: 3.1107 - acc: 0.2701\n",
      "Epoch 10/50\n",
      "4794/4794 [==============================] - 0s 47us/step - loss: 3.0854 - acc: 0.2829\n",
      "Epoch 11/50\n",
      "4794/4794 [==============================] - 0s 46us/step - loss: 3.0607 - acc: 0.2833\n",
      "Epoch 12/50\n",
      "4794/4794 [==============================] - 0s 55us/step - loss: 3.0389 - acc: 0.2904\n",
      "Epoch 13/50\n",
      "4794/4794 [==============================] - 0s 50us/step - loss: 3.0157 - acc: 0.2897\n",
      "Epoch 14/50\n",
      "4794/4794 [==============================] - 0s 40us/step - loss: 2.9940 - acc: 0.2952\n",
      "Epoch 15/50\n",
      "4794/4794 [==============================] - 0s 50us/step - loss: 2.9717 - acc: 0.2983\n",
      "Epoch 16/50\n",
      "4794/4794 [==============================] - 0s 50us/step - loss: 2.9510 - acc: 0.3045\n",
      "Epoch 17/50\n",
      "4794/4794 [==============================] - 0s 50us/step - loss: 2.9302 - acc: 0.3075\n",
      "Epoch 18/50\n",
      "4794/4794 [==============================] - 0s 42us/step - loss: 2.9095 - acc: 0.3114\n",
      "Epoch 19/50\n",
      "4794/4794 [==============================] - 0s 42us/step - loss: 2.8898 - acc: 0.3133\n",
      "Epoch 20/50\n",
      "4794/4794 [==============================] - 0s 47us/step - loss: 2.8704 - acc: 0.3202\n",
      "Epoch 21/50\n",
      "4794/4794 [==============================] - 0s 46us/step - loss: 2.8504 - acc: 0.3267\n",
      "Epoch 22/50\n",
      "4794/4794 [==============================] - 0s 50us/step - loss: 2.8338 - acc: 0.3346\n",
      "Epoch 23/50\n",
      "4794/4794 [==============================] - 0s 47us/step - loss: 2.8135 - acc: 0.3369\n",
      "Epoch 24/50\n",
      "4794/4794 [==============================] - 0s 56us/step - loss: 2.7963 - acc: 0.3406\n",
      "Epoch 25/50\n",
      "4794/4794 [==============================] - 0s 55us/step - loss: 2.7789 - acc: 0.3436\n",
      "Epoch 26/50\n",
      "4794/4794 [==============================] - 0s 55us/step - loss: 2.7624 - acc: 0.3475\n",
      "Epoch 27/50\n",
      "4794/4794 [==============================] - 0s 53us/step - loss: 2.7456 - acc: 0.3492\n",
      "Epoch 28/50\n",
      "4794/4794 [==============================] - 0s 54us/step - loss: 2.7289 - acc: 0.3600\n",
      "Epoch 29/50\n",
      "4794/4794 [==============================] - 0s 43us/step - loss: 2.7143 - acc: 0.3561\n",
      "Epoch 30/50\n",
      "4794/4794 [==============================] - 0s 52us/step - loss: 2.6985 - acc: 0.3565\n",
      "Epoch 31/50\n",
      "4794/4794 [==============================] - 0s 50us/step - loss: 2.6838 - acc: 0.3611\n",
      "Epoch 32/50\n",
      "4794/4794 [==============================] - 0s 45us/step - loss: 2.6698 - acc: 0.3634\n",
      "Epoch 33/50\n",
      "4794/4794 [==============================] - 0s 45us/step - loss: 2.6550 - acc: 0.3636\n",
      "Epoch 34/50\n",
      "4794/4794 [==============================] - 0s 46us/step - loss: 2.6413 - acc: 0.3659\n",
      "Epoch 35/50\n",
      "4794/4794 [==============================] - 0s 49us/step - loss: 2.6265 - acc: 0.3678\n",
      "Epoch 36/50\n",
      "4794/4794 [==============================] - 0s 49us/step - loss: 2.6140 - acc: 0.3730\n",
      "Epoch 37/50\n",
      "4794/4794 [==============================] - 0s 44us/step - loss: 2.6008 - acc: 0.3771\n",
      "Epoch 38/50\n",
      "4794/4794 [==============================] - 0s 40us/step - loss: 2.5877 - acc: 0.3757\n",
      "Epoch 39/50\n",
      "4794/4794 [==============================] - 0s 48us/step - loss: 2.5746 - acc: 0.3830\n",
      "Epoch 40/50\n",
      "4794/4794 [==============================] - 0s 45us/step - loss: 2.5627 - acc: 0.3846\n",
      "Epoch 41/50\n",
      "4794/4794 [==============================] - 0s 48us/step - loss: 2.5498 - acc: 0.3851\n",
      "Epoch 42/50\n",
      "4794/4794 [==============================] - 0s 45us/step - loss: 2.5387 - acc: 0.3874\n",
      "Epoch 43/50\n",
      "4794/4794 [==============================] - 0s 43us/step - loss: 2.5278 - acc: 0.3913\n",
      "Epoch 44/50\n",
      "4794/4794 [==============================] - 0s 48us/step - loss: 2.5153 - acc: 0.3899\n",
      "Epoch 45/50\n",
      "4794/4794 [==============================] - 0s 45us/step - loss: 2.5056 - acc: 0.3888\n",
      "Epoch 46/50\n",
      "4794/4794 [==============================] - 0s 54us/step - loss: 2.4944 - acc: 0.3928\n",
      "Epoch 47/50\n",
      "4794/4794 [==============================] - 0s 54us/step - loss: 2.4837 - acc: 0.3911\n",
      "Epoch 48/50\n",
      "4794/4794 [==============================] - 0s 51us/step - loss: 2.4738 - acc: 0.3911\n",
      "Epoch 49/50\n",
      "4794/4794 [==============================] - 0s 45us/step - loss: 2.4625 - acc: 0.3942\n",
      "Epoch 50/50\n",
      "4794/4794 [==============================] - 0s 48us/step - loss: 2.4537 - acc: 0.3930\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "2154/2154 [==============================] - 0s 131us/step - loss: 4.5248 - acc: 0.0659\n",
      "Epoch 2/50\n",
      "2154/2154 [==============================] - 0s 51us/step - loss: 3.8004 - acc: 0.1012\n",
      "Epoch 3/50\n",
      "2154/2154 [==============================] - 0s 56us/step - loss: 3.5308 - acc: 0.2159\n",
      "Epoch 4/50\n",
      "2154/2154 [==============================] - 0s 57us/step - loss: 3.4556 - acc: 0.2154\n",
      "Epoch 5/50\n",
      "2154/2154 [==============================] - 0s 58us/step - loss: 3.4268 - acc: 0.2154\n",
      "Epoch 6/50\n",
      "2154/2154 [==============================] - 0s 51us/step - loss: 3.4044 - acc: 0.2154\n",
      "Epoch 7/50\n",
      "2154/2154 [==============================] - 0s 56us/step - loss: 3.3852 - acc: 0.2154\n",
      "Epoch 8/50\n",
      "2154/2154 [==============================] - 0s 49us/step - loss: 3.3690 - acc: 0.2154\n",
      "Epoch 9/50\n",
      "2154/2154 [==============================] - 0s 52us/step - loss: 3.3572 - acc: 0.2154\n",
      "Epoch 10/50\n",
      "2154/2154 [==============================] - 0s 54us/step - loss: 3.3419 - acc: 0.2154\n",
      "Epoch 11/50\n",
      "2154/2154 [==============================] - 0s 53us/step - loss: 3.3304 - acc: 0.2168\n",
      "Epoch 12/50\n",
      "2154/2154 [==============================] - 0s 55us/step - loss: 3.3161 - acc: 0.2173\n",
      "Epoch 13/50\n",
      "2154/2154 [==============================] - 0s 50us/step - loss: 3.3070 - acc: 0.2214\n",
      "Epoch 14/50\n",
      "2154/2154 [==============================] - 0s 43us/step - loss: 3.2948 - acc: 0.2224\n",
      "Epoch 15/50\n",
      "2154/2154 [==============================] - 0s 44us/step - loss: 3.2834 - acc: 0.2205\n",
      "Epoch 16/50\n",
      "2154/2154 [==============================] - 0s 54us/step - loss: 3.2729 - acc: 0.2256\n",
      "Epoch 17/50\n",
      "2154/2154 [==============================] - 0s 55us/step - loss: 3.2639 - acc: 0.2284\n",
      "Epoch 18/50\n",
      "2154/2154 [==============================] - 0s 44us/step - loss: 3.2539 - acc: 0.2279\n",
      "Epoch 19/50\n",
      "2154/2154 [==============================] - 0s 42us/step - loss: 3.2430 - acc: 0.2266\n",
      "Epoch 20/50\n",
      "2154/2154 [==============================] - 0s 38us/step - loss: 3.2341 - acc: 0.2340\n",
      "Epoch 21/50\n",
      "2154/2154 [==============================] - 0s 41us/step - loss: 3.2245 - acc: 0.2307\n",
      "Epoch 22/50\n",
      "2154/2154 [==============================] - 0s 38us/step - loss: 3.2147 - acc: 0.2307\n",
      "Epoch 23/50\n",
      "2154/2154 [==============================] - 0s 45us/step - loss: 3.2070 - acc: 0.2349\n",
      "Epoch 24/50\n",
      "2154/2154 [==============================] - 0s 46us/step - loss: 3.1959 - acc: 0.2307\n",
      "Epoch 25/50\n",
      "2154/2154 [==============================] - 0s 51us/step - loss: 3.1867 - acc: 0.2344\n",
      "Epoch 26/50\n",
      "2154/2154 [==============================] - 0s 53us/step - loss: 3.1799 - acc: 0.2354\n",
      "Epoch 27/50\n",
      "2154/2154 [==============================] - 0s 52us/step - loss: 3.1697 - acc: 0.2372\n",
      "Epoch 28/50\n",
      "2154/2154 [==============================] - 0s 46us/step - loss: 3.1605 - acc: 0.2396\n",
      "Epoch 29/50\n",
      "2154/2154 [==============================] - 0s 54us/step - loss: 3.1531 - acc: 0.2414\n",
      "Epoch 30/50\n",
      "2154/2154 [==============================] - 0s 46us/step - loss: 3.1441 - acc: 0.2423\n",
      "Epoch 31/50\n",
      "2154/2154 [==============================] - 0s 43us/step - loss: 3.1343 - acc: 0.2451\n",
      "Epoch 32/50\n",
      "2154/2154 [==============================] - 0s 40us/step - loss: 3.1265 - acc: 0.2479\n",
      "Epoch 33/50\n",
      "2154/2154 [==============================] - 0s 40us/step - loss: 3.1182 - acc: 0.2447\n",
      "Epoch 34/50\n",
      "2154/2154 [==============================] - 0s 42us/step - loss: 3.1100 - acc: 0.2493\n",
      "Epoch 35/50\n",
      "2154/2154 [==============================] - 0s 42us/step - loss: 3.1021 - acc: 0.2456\n",
      "Epoch 36/50\n",
      "2154/2154 [==============================] - 0s 41us/step - loss: 3.0929 - acc: 0.2502\n",
      "Epoch 37/50\n",
      "2154/2154 [==============================] - 0s 40us/step - loss: 3.0883 - acc: 0.2493\n",
      "Epoch 38/50\n",
      "2154/2154 [==============================] - 0s 40us/step - loss: 3.0775 - acc: 0.2539\n",
      "Epoch 39/50\n",
      "2154/2154 [==============================] - 0s 41us/step - loss: 3.0698 - acc: 0.2553\n",
      "Epoch 40/50\n",
      "2154/2154 [==============================] - 0s 40us/step - loss: 3.0601 - acc: 0.2526\n",
      "Epoch 41/50\n",
      "2154/2154 [==============================] - 0s 44us/step - loss: 3.0516 - acc: 0.2609\n",
      "Epoch 42/50\n",
      "2154/2154 [==============================] - 0s 42us/step - loss: 3.0451 - acc: 0.2665\n",
      "Epoch 43/50\n",
      "2154/2154 [==============================] - 0s 47us/step - loss: 3.0359 - acc: 0.2674\n",
      "Epoch 44/50\n",
      "2154/2154 [==============================] - 0s 46us/step - loss: 3.0280 - acc: 0.2656\n",
      "Epoch 45/50\n",
      "2154/2154 [==============================] - 0s 44us/step - loss: 3.0202 - acc: 0.2748\n",
      "Epoch 46/50\n",
      "2154/2154 [==============================] - 0s 46us/step - loss: 3.0134 - acc: 0.2707\n",
      "Epoch 47/50\n",
      "2154/2154 [==============================] - 0s 47us/step - loss: 3.0044 - acc: 0.2772\n",
      "Epoch 48/50\n",
      "2154/2154 [==============================] - 0s 46us/step - loss: 2.9958 - acc: 0.2702\n",
      "Epoch 49/50\n",
      "2154/2154 [==============================] - 0s 45us/step - loss: 2.9890 - acc: 0.2799\n",
      "Epoch 50/50\n",
      "2154/2154 [==============================] - 0s 45us/step - loss: 2.9810 - acc: 0.2818\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "106/106 [==============================] - 0s 2ms/step - loss: 4.2469 - acc: 0.0094\n",
      "Epoch 2/50\n",
      "106/106 [==============================] - 0s 71us/step - loss: 4.0878 - acc: 0.0660\n",
      "Epoch 3/50\n",
      "106/106 [==============================] - 0s 61us/step - loss: 3.9859 - acc: 0.1038\n",
      "Epoch 4/50\n",
      "106/106 [==============================] - 0s 59us/step - loss: 3.9071 - acc: 0.0943\n",
      "Epoch 5/50\n",
      "106/106 [==============================] - 0s 60us/step - loss: 3.8413 - acc: 0.0943\n",
      "Epoch 6/50\n",
      "106/106 [==============================] - 0s 69us/step - loss: 3.7842 - acc: 0.0943\n",
      "Epoch 7/50\n",
      "106/106 [==============================] - 0s 52us/step - loss: 3.7332 - acc: 0.0943\n",
      "Epoch 8/50\n",
      "106/106 [==============================] - 0s 48us/step - loss: 3.6870 - acc: 0.0943\n",
      "Epoch 9/50\n",
      "106/106 [==============================] - 0s 50us/step - loss: 3.6446 - acc: 0.0943\n",
      "Epoch 10/50\n",
      "106/106 [==============================] - 0s 49us/step - loss: 3.6053 - acc: 0.0943\n",
      "Epoch 11/50\n",
      "106/106 [==============================] - 0s 58us/step - loss: 3.5687 - acc: 0.0943\n",
      "Epoch 12/50\n",
      "106/106 [==============================] - 0s 52us/step - loss: 3.5346 - acc: 0.0943\n",
      "Epoch 13/50\n",
      "106/106 [==============================] - 0s 48us/step - loss: 3.5027 - acc: 0.0943\n",
      "Epoch 14/50\n",
      "106/106 [==============================] - 0s 51us/step - loss: 3.4728 - acc: 0.0943\n",
      "Epoch 15/50\n",
      "106/106 [==============================] - 0s 52us/step - loss: 3.4449 - acc: 0.0943\n",
      "Epoch 16/50\n",
      "106/106 [==============================] - 0s 55us/step - loss: 3.4189 - acc: 0.0943\n",
      "Epoch 17/50\n",
      "106/106 [==============================] - 0s 60us/step - loss: 3.3946 - acc: 0.1132\n",
      "Epoch 18/50\n",
      "106/106 [==============================] - 0s 67us/step - loss: 3.3719 - acc: 0.1132\n",
      "Epoch 19/50\n",
      "106/106 [==============================] - 0s 54us/step - loss: 3.3508 - acc: 0.1132\n",
      "Epoch 20/50\n",
      "106/106 [==============================] - 0s 60us/step - loss: 3.3312 - acc: 0.1226\n",
      "Epoch 21/50\n",
      "106/106 [==============================] - 0s 60us/step - loss: 3.3130 - acc: 0.1226\n",
      "Epoch 22/50\n",
      "106/106 [==============================] - 0s 58us/step - loss: 3.2961 - acc: 0.1226\n",
      "Epoch 23/50\n",
      "106/106 [==============================] - 0s 67us/step - loss: 3.2805 - acc: 0.1226\n",
      "Epoch 24/50\n",
      "106/106 [==============================] - 0s 66us/step - loss: 3.2659 - acc: 0.1321\n",
      "Epoch 25/50\n",
      "106/106 [==============================] - 0s 64us/step - loss: 3.2525 - acc: 0.1321\n",
      "Epoch 26/50\n",
      "106/106 [==============================] - 0s 63us/step - loss: 3.2401 - acc: 0.1321\n",
      "Epoch 27/50\n",
      "106/106 [==============================] - 0s 65us/step - loss: 3.2286 - acc: 0.1415\n",
      "Epoch 28/50\n",
      "106/106 [==============================] - 0s 66us/step - loss: 3.2180 - acc: 0.1604\n",
      "Epoch 29/50\n",
      "106/106 [==============================] - 0s 65us/step - loss: 3.2081 - acc: 0.1981\n",
      "Epoch 30/50\n",
      "106/106 [==============================] - 0s 60us/step - loss: 3.1991 - acc: 0.2547\n",
      "Epoch 31/50\n",
      "106/106 [==============================] - 0s 68us/step - loss: 3.1907 - acc: 0.2547\n",
      "Epoch 32/50\n",
      "106/106 [==============================] - 0s 61us/step - loss: 3.1829 - acc: 0.2170\n",
      "Epoch 33/50\n",
      "106/106 [==============================] - 0s 66us/step - loss: 3.1757 - acc: 0.2264\n",
      "Epoch 34/50\n",
      "106/106 [==============================] - 0s 66us/step - loss: 3.1690 - acc: 0.2264\n",
      "Epoch 35/50\n",
      "106/106 [==============================] - 0s 66us/step - loss: 3.1629 - acc: 0.2453\n",
      "Epoch 36/50\n",
      "106/106 [==============================] - 0s 62us/step - loss: 3.1571 - acc: 0.2453\n",
      "Epoch 37/50\n",
      "106/106 [==============================] - 0s 68us/step - loss: 3.1518 - acc: 0.2264\n",
      "Epoch 38/50\n",
      "106/106 [==============================] - 0s 73us/step - loss: 3.1468 - acc: 0.2358\n",
      "Epoch 39/50\n",
      "106/106 [==============================] - 0s 68us/step - loss: 3.1421 - acc: 0.2358\n",
      "Epoch 40/50\n",
      "106/106 [==============================] - 0s 63us/step - loss: 3.1377 - acc: 0.2358\n",
      "Epoch 41/50\n",
      "106/106 [==============================] - 0s 66us/step - loss: 3.1335 - acc: 0.2358\n",
      "Epoch 42/50\n",
      "106/106 [==============================] - 0s 67us/step - loss: 3.1295 - acc: 0.2358\n",
      "Epoch 43/50\n",
      "106/106 [==============================] - 0s 64us/step - loss: 3.1257 - acc: 0.2358\n",
      "Epoch 44/50\n",
      "106/106 [==============================] - 0s 65us/step - loss: 3.1221 - acc: 0.2358\n",
      "Epoch 45/50\n",
      "106/106 [==============================] - 0s 59us/step - loss: 3.1186 - acc: 0.2264\n",
      "Epoch 46/50\n",
      "106/106 [==============================] - 0s 75us/step - loss: 3.1152 - acc: 0.2264\n",
      "Epoch 47/50\n",
      "106/106 [==============================] - 0s 66us/step - loss: 3.1119 - acc: 0.2264\n",
      "Epoch 48/50\n",
      "106/106 [==============================] - 0s 62us/step - loss: 3.1087 - acc: 0.2264\n",
      "Epoch 49/50\n",
      "106/106 [==============================] - 0s 66us/step - loss: 3.1056 - acc: 0.2453\n",
      "Epoch 50/50\n",
      "106/106 [==============================] - 0s 60us/step - loss: 3.1025 - acc: 0.2453\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "74/74 [==============================] - 0s 4ms/step - loss: 3.5438 - acc: 0.0270\n",
      "Epoch 2/50\n",
      "74/74 [==============================] - 0s 102us/step - loss: 3.3412 - acc: 0.0270\n",
      "Epoch 3/50\n",
      "74/74 [==============================] - 0s 71us/step - loss: 3.2060 - acc: 0.0270\n",
      "Epoch 4/50\n",
      "74/74 [==============================] - 0s 69us/step - loss: 3.1006 - acc: 0.0405\n",
      "Epoch 5/50\n",
      "74/74 [==============================] - 0s 68us/step - loss: 3.0131 - acc: 0.1351\n",
      "Epoch 6/50\n",
      "74/74 [==============================] - 0s 122us/step - loss: 2.9378 - acc: 0.1486\n",
      "Epoch 7/50\n",
      "74/74 [==============================] - 0s 83us/step - loss: 2.8715 - acc: 0.1486\n",
      "Epoch 8/50\n",
      "74/74 [==============================] - 0s 83us/step - loss: 2.8122 - acc: 0.1486\n",
      "Epoch 9/50\n",
      "74/74 [==============================] - 0s 83us/step - loss: 2.7585 - acc: 0.1622\n",
      "Epoch 10/50\n",
      "74/74 [==============================] - 0s 89us/step - loss: 2.7095 - acc: 0.1622\n",
      "Epoch 11/50\n",
      "74/74 [==============================] - 0s 98us/step - loss: 2.6643 - acc: 0.1622\n",
      "Epoch 12/50\n",
      "74/74 [==============================] - 0s 95us/step - loss: 2.6226 - acc: 0.1757\n",
      "Epoch 13/50\n",
      "74/74 [==============================] - 0s 95us/step - loss: 2.5839 - acc: 0.1757\n",
      "Epoch 14/50\n",
      "74/74 [==============================] - 0s 89us/step - loss: 2.5479 - acc: 0.1622\n",
      "Epoch 15/50\n",
      "74/74 [==============================] - 0s 88us/step - loss: 2.5144 - acc: 0.1622\n",
      "Epoch 16/50\n",
      "74/74 [==============================] - 0s 94us/step - loss: 2.4833 - acc: 0.1622\n",
      "Epoch 17/50\n",
      "74/74 [==============================] - 0s 100us/step - loss: 2.4544 - acc: 0.1622\n",
      "Epoch 18/50\n",
      "74/74 [==============================] - 0s 83us/step - loss: 2.4276 - acc: 0.1757\n",
      "Epoch 19/50\n",
      "74/74 [==============================] - 0s 93us/step - loss: 2.4027 - acc: 0.1622\n",
      "Epoch 20/50\n",
      "74/74 [==============================] - 0s 87us/step - loss: 2.3797 - acc: 0.1622\n",
      "Epoch 21/50\n",
      "74/74 [==============================] - 0s 95us/step - loss: 2.3584 - acc: 0.1622\n",
      "Epoch 22/50\n",
      "74/74 [==============================] - 0s 86us/step - loss: 2.3388 - acc: 0.1892\n",
      "Epoch 23/50\n",
      "74/74 [==============================] - 0s 103us/step - loss: 2.3208 - acc: 0.2297\n",
      "Epoch 24/50\n",
      "74/74 [==============================] - 0s 103us/step - loss: 2.3042 - acc: 0.2568\n",
      "Epoch 25/50\n",
      "74/74 [==============================] - 0s 95us/step - loss: 2.2890 - acc: 0.3243\n",
      "Epoch 26/50\n",
      "74/74 [==============================] - 0s 100us/step - loss: 2.2750 - acc: 0.3514\n",
      "Epoch 27/50\n",
      "74/74 [==============================] - 0s 93us/step - loss: 2.2623 - acc: 0.3784\n",
      "Epoch 28/50\n",
      "74/74 [==============================] - 0s 88us/step - loss: 2.2505 - acc: 0.3919\n",
      "Epoch 29/50\n",
      "74/74 [==============================] - 0s 86us/step - loss: 2.2397 - acc: 0.3919\n",
      "Epoch 30/50\n",
      "74/74 [==============================] - 0s 88us/step - loss: 2.2298 - acc: 0.3514\n",
      "Epoch 31/50\n",
      "74/74 [==============================] - 0s 101us/step - loss: 2.2207 - acc: 0.3378\n",
      "Epoch 32/50\n",
      "74/74 [==============================] - 0s 87us/step - loss: 2.2122 - acc: 0.3108\n",
      "Epoch 33/50\n",
      "74/74 [==============================] - 0s 90us/step - loss: 2.2043 - acc: 0.3108\n",
      "Epoch 34/50\n",
      "74/74 [==============================] - 0s 91us/step - loss: 2.1970 - acc: 0.2973\n",
      "Epoch 35/50\n",
      "74/74 [==============================] - 0s 89us/step - loss: 2.1902 - acc: 0.2838\n",
      "Epoch 36/50\n",
      "74/74 [==============================] - 0s 85us/step - loss: 2.1837 - acc: 0.2838\n",
      "Epoch 37/50\n",
      "74/74 [==============================] - 0s 76us/step - loss: 2.1776 - acc: 0.2838\n",
      "Epoch 38/50\n",
      "74/74 [==============================] - 0s 71us/step - loss: 2.1718 - acc: 0.2838\n",
      "Epoch 39/50\n",
      "74/74 [==============================] - 0s 91us/step - loss: 2.1663 - acc: 0.2838\n",
      "Epoch 40/50\n",
      "74/74 [==============================] - 0s 88us/step - loss: 2.1610 - acc: 0.2838\n",
      "Epoch 41/50\n",
      "74/74 [==============================] - 0s 101us/step - loss: 2.1558 - acc: 0.2838\n",
      "Epoch 42/50\n",
      "74/74 [==============================] - 0s 87us/step - loss: 2.1509 - acc: 0.2838\n",
      "Epoch 43/50\n",
      "74/74 [==============================] - 0s 78us/step - loss: 2.1461 - acc: 0.2838\n",
      "Epoch 44/50\n",
      "74/74 [==============================] - 0s 80us/step - loss: 2.1414 - acc: 0.2838\n",
      "Epoch 45/50\n",
      "74/74 [==============================] - 0s 88us/step - loss: 2.1368 - acc: 0.2838\n",
      "Epoch 46/50\n",
      "74/74 [==============================] - 0s 102us/step - loss: 2.1323 - acc: 0.2838\n",
      "Epoch 47/50\n",
      "74/74 [==============================] - 0s 85us/step - loss: 2.1279 - acc: 0.2838\n",
      "Epoch 48/50\n",
      "74/74 [==============================] - 0s 82us/step - loss: 2.1235 - acc: 0.2838\n",
      "Epoch 49/50\n",
      "74/74 [==============================] - 0s 88us/step - loss: 2.1192 - acc: 0.2838\n",
      "Epoch 50/50\n",
      "74/74 [==============================] - 0s 79us/step - loss: 2.1149 - acc: 0.2838\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "121/121 [==============================] - 0s 2ms/step - loss: 3.5760 - acc: 0.0083\n",
      "Epoch 2/50\n",
      "121/121 [==============================] - 0s 73us/step - loss: 3.3819 - acc: 0.0083\n",
      "Epoch 3/50\n",
      "121/121 [==============================] - 0s 68us/step - loss: 3.2586 - acc: 0.0083\n",
      "Epoch 4/50\n",
      "121/121 [==============================] - 0s 61us/step - loss: 3.1669 - acc: 0.0083\n",
      "Epoch 5/50\n",
      "121/121 [==============================] - 0s 56us/step - loss: 3.0938 - acc: 0.0083\n",
      "Epoch 6/50\n",
      "121/121 [==============================] - 0s 71us/step - loss: 3.0333 - acc: 0.0579\n",
      "Epoch 7/50\n",
      "121/121 [==============================] - 0s 58us/step - loss: 2.9820 - acc: 0.0992\n",
      "Epoch 8/50\n",
      "121/121 [==============================] - 0s 81us/step - loss: 2.9376 - acc: 0.1322\n",
      "Epoch 9/50\n",
      "121/121 [==============================] - 0s 55us/step - loss: 2.8988 - acc: 0.1240\n",
      "Epoch 10/50\n",
      "121/121 [==============================] - 0s 51us/step - loss: 2.8645 - acc: 0.1157\n",
      "Epoch 11/50\n",
      "121/121 [==============================] - 0s 60us/step - loss: 2.8339 - acc: 0.1157\n",
      "Epoch 12/50\n",
      "121/121 [==============================] - 0s 62us/step - loss: 2.8064 - acc: 0.1157\n",
      "Epoch 13/50\n",
      "121/121 [==============================] - 0s 55us/step - loss: 2.7815 - acc: 0.1074\n",
      "Epoch 14/50\n",
      "121/121 [==============================] - 0s 55us/step - loss: 2.7590 - acc: 0.1074\n",
      "Epoch 15/50\n",
      "121/121 [==============================] - 0s 61us/step - loss: 2.7385 - acc: 0.1074\n",
      "Epoch 16/50\n",
      "121/121 [==============================] - 0s 66us/step - loss: 2.7198 - acc: 0.1074\n",
      "Epoch 17/50\n",
      "121/121 [==============================] - 0s 56us/step - loss: 2.7027 - acc: 0.1074\n",
      "Epoch 18/50\n",
      "121/121 [==============================] - 0s 56us/step - loss: 2.6871 - acc: 0.1074\n",
      "Epoch 19/50\n",
      "121/121 [==============================] - 0s 56us/step - loss: 2.6727 - acc: 0.1074\n",
      "Epoch 20/50\n",
      "121/121 [==============================] - 0s 61us/step - loss: 2.6594 - acc: 0.1157\n",
      "Epoch 21/50\n",
      "121/121 [==============================] - 0s 64us/step - loss: 2.6473 - acc: 0.1157\n",
      "Epoch 22/50\n",
      "121/121 [==============================] - 0s 52us/step - loss: 2.6360 - acc: 0.1240\n",
      "Epoch 23/50\n",
      "121/121 [==============================] - 0s 50us/step - loss: 2.6256 - acc: 0.1901\n",
      "Epoch 24/50\n",
      "121/121 [==============================] - 0s 49us/step - loss: 2.6160 - acc: 0.1983\n",
      "Epoch 25/50\n",
      "121/121 [==============================] - 0s 48us/step - loss: 2.6071 - acc: 0.2149\n",
      "Epoch 26/50\n",
      "121/121 [==============================] - 0s 57us/step - loss: 2.5988 - acc: 0.2066\n",
      "Epoch 27/50\n",
      "121/121 [==============================] - 0s 50us/step - loss: 2.5911 - acc: 0.1983\n",
      "Epoch 28/50\n",
      "121/121 [==============================] - 0s 75us/step - loss: 2.5839 - acc: 0.1901\n",
      "Epoch 29/50\n",
      "121/121 [==============================] - 0s 115us/step - loss: 2.5772 - acc: 0.1570\n",
      "Epoch 30/50\n",
      "121/121 [==============================] - 0s 56us/step - loss: 2.5709 - acc: 0.1570\n",
      "Epoch 31/50\n",
      "121/121 [==============================] - 0s 61us/step - loss: 2.5650 - acc: 0.1653\n",
      "Epoch 32/50\n",
      "121/121 [==============================] - 0s 59us/step - loss: 2.5595 - acc: 0.1736\n",
      "Epoch 33/50\n",
      "121/121 [==============================] - 0s 60us/step - loss: 2.5542 - acc: 0.1818\n",
      "Epoch 34/50\n",
      "121/121 [==============================] - 0s 61us/step - loss: 2.5492 - acc: 0.1818\n",
      "Epoch 35/50\n",
      "121/121 [==============================] - 0s 61us/step - loss: 2.5445 - acc: 0.1818\n",
      "Epoch 36/50\n",
      "121/121 [==============================] - 0s 62us/step - loss: 2.5400 - acc: 0.1818\n",
      "Epoch 37/50\n",
      "121/121 [==============================] - 0s 72us/step - loss: 2.5356 - acc: 0.1818\n",
      "Epoch 38/50\n",
      "121/121 [==============================] - 0s 85us/step - loss: 2.5315 - acc: 0.1818\n",
      "Epoch 39/50\n",
      "121/121 [==============================] - 0s 52us/step - loss: 2.5275 - acc: 0.1818\n",
      "Epoch 40/50\n",
      "121/121 [==============================] - 0s 56us/step - loss: 2.5236 - acc: 0.1818\n",
      "Epoch 41/50\n",
      "121/121 [==============================] - 0s 56us/step - loss: 2.5199 - acc: 0.1818\n",
      "Epoch 42/50\n",
      "121/121 [==============================] - 0s 53us/step - loss: 2.5163 - acc: 0.1818\n",
      "Epoch 43/50\n",
      "121/121 [==============================] - 0s 47us/step - loss: 2.5127 - acc: 0.1818\n",
      "Epoch 44/50\n",
      "121/121 [==============================] - 0s 53us/step - loss: 2.5093 - acc: 0.1818\n",
      "Epoch 45/50\n",
      "121/121 [==============================] - 0s 50us/step - loss: 2.5059 - acc: 0.1818\n",
      "Epoch 46/50\n",
      "121/121 [==============================] - 0s 54us/step - loss: 2.5026 - acc: 0.1818\n",
      "Epoch 47/50\n",
      "121/121 [==============================] - 0s 46us/step - loss: 2.4993 - acc: 0.1818\n",
      "Epoch 48/50\n",
      "121/121 [==============================] - 0s 48us/step - loss: 2.4961 - acc: 0.1736\n",
      "Epoch 49/50\n",
      "121/121 [==============================] - 0s 44us/step - loss: 2.4930 - acc: 0.1736\n",
      "Epoch 50/50\n",
      "121/121 [==============================] - 0s 46us/step - loss: 2.4898 - acc: 0.1736\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "52/52 [==============================] - 0s 5ms/step - loss: 3.2862 - acc: 0.0192\n",
      "Epoch 2/50\n",
      "52/52 [==============================] - 0s 145us/step - loss: 3.1441 - acc: 0.0192\n",
      "Epoch 3/50\n",
      "52/52 [==============================] - 0s 114us/step - loss: 3.0631 - acc: 0.0192\n",
      "Epoch 4/50\n",
      "52/52 [==============================] - 0s 115us/step - loss: 3.0069 - acc: 0.0385\n",
      "Epoch 5/50\n",
      "52/52 [==============================] - 0s 123us/step - loss: 2.9639 - acc: 0.1154\n",
      "Epoch 6/50\n",
      "52/52 [==============================] - 0s 150us/step - loss: 2.9292 - acc: 0.1154\n",
      "Epoch 7/50\n",
      "52/52 [==============================] - 0s 148us/step - loss: 2.9002 - acc: 0.1154\n",
      "Epoch 8/50\n",
      "52/52 [==============================] - 0s 123us/step - loss: 2.8755 - acc: 0.1154\n",
      "Epoch 9/50\n",
      "52/52 [==============================] - 0s 113us/step - loss: 2.8541 - acc: 0.1154\n",
      "Epoch 10/50\n",
      "52/52 [==============================] - 0s 115us/step - loss: 2.8353 - acc: 0.1154\n",
      "Epoch 11/50\n",
      "52/52 [==============================] - 0s 107us/step - loss: 2.8186 - acc: 0.1154\n",
      "Epoch 12/50\n",
      "52/52 [==============================] - 0s 98us/step - loss: 2.8038 - acc: 0.1154\n",
      "Epoch 13/50\n",
      "52/52 [==============================] - 0s 95us/step - loss: 2.7906 - acc: 0.1154\n",
      "Epoch 14/50\n",
      "52/52 [==============================] - 0s 93us/step - loss: 2.7787 - acc: 0.1154\n",
      "Epoch 15/50\n",
      "52/52 [==============================] - 0s 102us/step - loss: 2.7680 - acc: 0.1923\n",
      "Epoch 16/50\n",
      "52/52 [==============================] - 0s 110us/step - loss: 2.7582 - acc: 0.2500\n",
      "Epoch 17/50\n",
      "52/52 [==============================] - 0s 111us/step - loss: 2.7494 - acc: 0.1923\n",
      "Epoch 18/50\n",
      "52/52 [==============================] - 0s 109us/step - loss: 2.7413 - acc: 0.1731\n",
      "Epoch 19/50\n",
      "52/52 [==============================] - 0s 103us/step - loss: 2.7338 - acc: 0.1538\n",
      "Epoch 20/50\n",
      "52/52 [==============================] - 0s 99us/step - loss: 2.7270 - acc: 0.1538\n",
      "Epoch 21/50\n",
      "52/52 [==============================] - 0s 117us/step - loss: 2.7206 - acc: 0.1538\n",
      "Epoch 22/50\n",
      "52/52 [==============================] - 0s 134us/step - loss: 2.7147 - acc: 0.1538\n",
      "Epoch 23/50\n",
      "52/52 [==============================] - 0s 127us/step - loss: 2.7091 - acc: 0.1538\n",
      "Epoch 24/50\n",
      "52/52 [==============================] - 0s 124us/step - loss: 2.7038 - acc: 0.1538\n",
      "Epoch 25/50\n",
      "52/52 [==============================] - 0s 122us/step - loss: 2.6989 - acc: 0.1538\n",
      "Epoch 26/50\n",
      "52/52 [==============================] - 0s 128us/step - loss: 2.6941 - acc: 0.1538\n",
      "Epoch 27/50\n",
      "52/52 [==============================] - 0s 124us/step - loss: 2.6896 - acc: 0.1538\n",
      "Epoch 28/50\n",
      "52/52 [==============================] - 0s 117us/step - loss: 2.6853 - acc: 0.1538\n",
      "Epoch 29/50\n",
      "52/52 [==============================] - 0s 118us/step - loss: 2.6811 - acc: 0.1538\n",
      "Epoch 30/50\n",
      "52/52 [==============================] - 0s 130us/step - loss: 2.6770 - acc: 0.1538\n",
      "Epoch 31/50\n",
      "52/52 [==============================] - 0s 133us/step - loss: 2.6731 - acc: 0.1538\n",
      "Epoch 32/50\n",
      "52/52 [==============================] - 0s 122us/step - loss: 2.6693 - acc: 0.1538\n",
      "Epoch 33/50\n",
      "52/52 [==============================] - 0s 118us/step - loss: 2.6655 - acc: 0.1538\n",
      "Epoch 34/50\n",
      "52/52 [==============================] - 0s 116us/step - loss: 2.6618 - acc: 0.1538\n",
      "Epoch 35/50\n",
      "52/52 [==============================] - 0s 116us/step - loss: 2.6582 - acc: 0.1538\n",
      "Epoch 36/50\n",
      "52/52 [==============================] - 0s 107us/step - loss: 2.6546 - acc: 0.1538\n",
      "Epoch 37/50\n",
      "52/52 [==============================] - 0s 125us/step - loss: 2.6511 - acc: 0.1538\n",
      "Epoch 38/50\n",
      "52/52 [==============================] - 0s 118us/step - loss: 2.6476 - acc: 0.1538\n",
      "Epoch 39/50\n",
      "52/52 [==============================] - 0s 103us/step - loss: 2.6441 - acc: 0.1538\n",
      "Epoch 40/50\n",
      "52/52 [==============================] - 0s 94us/step - loss: 2.6406 - acc: 0.1538\n",
      "Epoch 41/50\n",
      "52/52 [==============================] - 0s 104us/step - loss: 2.6372 - acc: 0.1538\n",
      "Epoch 42/50\n",
      "52/52 [==============================] - 0s 94us/step - loss: 2.6338 - acc: 0.1538\n",
      "Epoch 43/50\n",
      "52/52 [==============================] - 0s 102us/step - loss: 2.6304 - acc: 0.1538\n",
      "Epoch 44/50\n",
      "52/52 [==============================] - 0s 110us/step - loss: 2.6271 - acc: 0.1538\n",
      "Epoch 45/50\n",
      "52/52 [==============================] - 0s 122us/step - loss: 2.6238 - acc: 0.1538\n",
      "Epoch 46/50\n",
      "52/52 [==============================] - 0s 138us/step - loss: 2.6205 - acc: 0.1538\n",
      "Epoch 47/50\n",
      "52/52 [==============================] - 0s 124us/step - loss: 2.6171 - acc: 0.1538\n",
      "Epoch 48/50\n",
      "52/52 [==============================] - 0s 122us/step - loss: 2.6137 - acc: 0.1538\n",
      "Epoch 49/50\n",
      "52/52 [==============================] - 0s 125us/step - loss: 2.6104 - acc: 0.1538\n",
      "Epoch 50/50\n",
      "52/52 [==============================] - 0s 106us/step - loss: 2.6071 - acc: 0.1731\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "388/388 [==============================] - 0s 666us/step - loss: 4.1229 - acc: 0.0103\n",
      "Epoch 2/50\n",
      "388/388 [==============================] - 0s 69us/step - loss: 3.6697 - acc: 0.2552\n",
      "Epoch 3/50\n",
      "388/388 [==============================] - 0s 65us/step - loss: 3.4471 - acc: 0.3711\n",
      "Epoch 4/50\n",
      "388/388 [==============================] - 0s 68us/step - loss: 3.2844 - acc: 0.3711\n",
      "Epoch 5/50\n",
      "388/388 [==============================] - 0s 69us/step - loss: 3.1652 - acc: 0.3711\n",
      "Epoch 6/50\n",
      "388/388 [==============================] - 0s 78us/step - loss: 3.0961 - acc: 0.3711\n",
      "Epoch 7/50\n",
      "388/388 [==============================] - 0s 63us/step - loss: 3.0400 - acc: 0.3711\n",
      "Epoch 8/50\n",
      "388/388 [==============================] - 0s 76us/step - loss: 3.0170 - acc: 0.3711\n",
      "Epoch 9/50\n",
      "388/388 [==============================] - 0s 70us/step - loss: 2.9867 - acc: 0.3711\n",
      "Epoch 10/50\n",
      "388/388 [==============================] - 0s 72us/step - loss: 2.9723 - acc: 0.3711\n",
      "Epoch 11/50\n",
      "388/388 [==============================] - 0s 71us/step - loss: 2.9564 - acc: 0.3711\n",
      "Epoch 12/50\n",
      "388/388 [==============================] - 0s 71us/step - loss: 2.9430 - acc: 0.3711\n",
      "Epoch 13/50\n",
      "388/388 [==============================] - 0s 71us/step - loss: 2.9370 - acc: 0.3711\n",
      "Epoch 14/50\n",
      "388/388 [==============================] - 0s 58us/step - loss: 2.9268 - acc: 0.3711\n",
      "Epoch 15/50\n",
      "388/388 [==============================] - 0s 58us/step - loss: 2.9210 - acc: 0.3711\n",
      "Epoch 16/50\n",
      "388/388 [==============================] - 0s 69us/step - loss: 2.9159 - acc: 0.3711\n",
      "Epoch 17/50\n",
      "388/388 [==============================] - 0s 73us/step - loss: 2.9127 - acc: 0.3711\n",
      "Epoch 18/50\n",
      "388/388 [==============================] - 0s 70us/step - loss: 2.9078 - acc: 0.3711\n",
      "Epoch 19/50\n",
      "388/388 [==============================] - 0s 66us/step - loss: 2.9038 - acc: 0.3711\n",
      "Epoch 20/50\n",
      "388/388 [==============================] - 0s 65us/step - loss: 2.8983 - acc: 0.3711\n",
      "Epoch 21/50\n",
      "388/388 [==============================] - 0s 64us/step - loss: 2.9001 - acc: 0.3711\n",
      "Epoch 22/50\n",
      "388/388 [==============================] - 0s 72us/step - loss: 2.8945 - acc: 0.3711\n",
      "Epoch 23/50\n",
      "388/388 [==============================] - 0s 60us/step - loss: 2.8926 - acc: 0.3711\n",
      "Epoch 24/50\n",
      "388/388 [==============================] - 0s 72us/step - loss: 2.8869 - acc: 0.3711\n",
      "Epoch 25/50\n",
      "388/388 [==============================] - 0s 66us/step - loss: 2.8853 - acc: 0.3711\n",
      "Epoch 26/50\n",
      "388/388 [==============================] - 0s 68us/step - loss: 2.8829 - acc: 0.3711\n",
      "Epoch 27/50\n",
      "388/388 [==============================] - 0s 60us/step - loss: 2.8819 - acc: 0.3711\n",
      "Epoch 28/50\n",
      "388/388 [==============================] - 0s 61us/step - loss: 2.8808 - acc: 0.3711\n",
      "Epoch 29/50\n",
      "388/388 [==============================] - 0s 70us/step - loss: 2.8750 - acc: 0.3711\n",
      "Epoch 30/50\n",
      "388/388 [==============================] - 0s 68us/step - loss: 2.8711 - acc: 0.3711\n",
      "Epoch 31/50\n",
      "388/388 [==============================] - 0s 71us/step - loss: 2.8668 - acc: 0.3711\n",
      "Epoch 32/50\n",
      "388/388 [==============================] - 0s 66us/step - loss: 2.8634 - acc: 0.3711\n",
      "Epoch 33/50\n",
      "388/388 [==============================] - 0s 58us/step - loss: 2.8601 - acc: 0.3711\n",
      "Epoch 34/50\n",
      "388/388 [==============================] - 0s 74us/step - loss: 2.8584 - acc: 0.3711\n",
      "Epoch 35/50\n",
      "388/388 [==============================] - 0s 62us/step - loss: 2.8571 - acc: 0.3711\n",
      "Epoch 36/50\n",
      "388/388 [==============================] - 0s 58us/step - loss: 2.8537 - acc: 0.3711\n",
      "Epoch 37/50\n",
      "388/388 [==============================] - 0s 51us/step - loss: 2.8487 - acc: 0.3711\n",
      "Epoch 38/50\n",
      "388/388 [==============================] - 0s 48us/step - loss: 2.8493 - acc: 0.3711\n",
      "Epoch 39/50\n",
      "388/388 [==============================] - 0s 48us/step - loss: 2.8465 - acc: 0.3711\n",
      "Epoch 40/50\n",
      "388/388 [==============================] - 0s 53us/step - loss: 2.8414 - acc: 0.3711\n",
      "Epoch 41/50\n",
      "388/388 [==============================] - 0s 54us/step - loss: 2.8377 - acc: 0.3711\n",
      "Epoch 42/50\n",
      "388/388 [==============================] - 0s 49us/step - loss: 2.8358 - acc: 0.3711\n",
      "Epoch 43/50\n",
      "388/388 [==============================] - 0s 48us/step - loss: 2.8323 - acc: 0.3711\n",
      "Epoch 44/50\n",
      "388/388 [==============================] - 0s 50us/step - loss: 2.8308 - acc: 0.3711\n",
      "Epoch 45/50\n",
      "388/388 [==============================] - 0s 72us/step - loss: 2.8296 - acc: 0.3711\n",
      "Epoch 46/50\n",
      "388/388 [==============================] - 0s 62us/step - loss: 2.8245 - acc: 0.3711\n",
      "Epoch 47/50\n",
      "388/388 [==============================] - 0s 68us/step - loss: 2.8235 - acc: 0.3711\n",
      "Epoch 48/50\n",
      "388/388 [==============================] - 0s 59us/step - loss: 2.8244 - acc: 0.3711\n",
      "Epoch 49/50\n",
      "388/388 [==============================] - 0s 58us/step - loss: 2.8166 - acc: 0.3711\n",
      "Epoch 50/50\n",
      "388/388 [==============================] - 0s 57us/step - loss: 2.8151 - acc: 0.3711\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "246/246 [==============================] - 0s 1ms/step - loss: 3.6397 - acc: 0.0813\n",
      "Epoch 2/50\n",
      "246/246 [==============================] - 0s 47us/step - loss: 3.4349 - acc: 0.1220\n",
      "Epoch 3/50\n",
      "246/246 [==============================] - 0s 47us/step - loss: 3.3260 - acc: 0.1220\n",
      "Epoch 4/50\n",
      "246/246 [==============================] - 0s 55us/step - loss: 3.2425 - acc: 0.1220\n",
      "Epoch 5/50\n",
      "246/246 [==============================] - 0s 56us/step - loss: 3.1804 - acc: 0.1220\n",
      "Epoch 6/50\n",
      "246/246 [==============================] - 0s 90us/step - loss: 3.1289 - acc: 0.1220\n",
      "Epoch 7/50\n",
      "246/246 [==============================] - 0s 68us/step - loss: 3.0847 - acc: 0.1341\n",
      "Epoch 8/50\n",
      "246/246 [==============================] - 0s 65us/step - loss: 3.0459 - acc: 0.1220\n",
      "Epoch 9/50\n",
      "246/246 [==============================] - 0s 57us/step - loss: 3.0164 - acc: 0.1260\n",
      "Epoch 10/50\n",
      "246/246 [==============================] - 0s 59us/step - loss: 2.9873 - acc: 0.1220\n",
      "Epoch 11/50\n",
      "246/246 [==============================] - 0s 61us/step - loss: 2.9678 - acc: 0.1220\n",
      "Epoch 12/50\n",
      "246/246 [==============================] - 0s 64us/step - loss: 2.9485 - acc: 0.1220\n",
      "Epoch 13/50\n",
      "246/246 [==============================] - 0s 60us/step - loss: 2.9329 - acc: 0.1220\n",
      "Epoch 14/50\n",
      "246/246 [==============================] - 0s 57us/step - loss: 2.9181 - acc: 0.1260\n",
      "Epoch 15/50\n",
      "246/246 [==============================] - 0s 53us/step - loss: 2.9089 - acc: 0.1382\n",
      "Epoch 16/50\n",
      "246/246 [==============================] - 0s 51us/step - loss: 2.9004 - acc: 0.1667\n",
      "Epoch 17/50\n",
      "246/246 [==============================] - 0s 50us/step - loss: 2.8910 - acc: 0.1545\n",
      "Epoch 18/50\n",
      "246/246 [==============================] - 0s 80us/step - loss: 2.8849 - acc: 0.1667\n",
      "Epoch 19/50\n",
      "246/246 [==============================] - 0s 61us/step - loss: 2.8802 - acc: 0.1707\n",
      "Epoch 20/50\n",
      "246/246 [==============================] - 0s 51us/step - loss: 2.8774 - acc: 0.1585\n",
      "Epoch 21/50\n",
      "246/246 [==============================] - 0s 58us/step - loss: 2.8707 - acc: 0.1463\n",
      "Epoch 22/50\n",
      "246/246 [==============================] - 0s 59us/step - loss: 2.8663 - acc: 0.1829\n",
      "Epoch 23/50\n",
      "246/246 [==============================] - 0s 55us/step - loss: 2.8694 - acc: 0.1545\n",
      "Epoch 24/50\n",
      "246/246 [==============================] - 0s 58us/step - loss: 2.8593 - acc: 0.1585\n",
      "Epoch 25/50\n",
      "246/246 [==============================] - 0s 54us/step - loss: 2.8539 - acc: 0.1382\n",
      "Epoch 26/50\n",
      "246/246 [==============================] - 0s 53us/step - loss: 2.8523 - acc: 0.1911\n",
      "Epoch 27/50\n",
      "246/246 [==============================] - 0s 50us/step - loss: 2.8528 - acc: 0.1789\n",
      "Epoch 28/50\n",
      "246/246 [==============================] - 0s 54us/step - loss: 2.8438 - acc: 0.1789\n",
      "Epoch 29/50\n",
      "246/246 [==============================] - 0s 47us/step - loss: 2.8430 - acc: 0.1585\n",
      "Epoch 30/50\n",
      "246/246 [==============================] - 0s 45us/step - loss: 2.8405 - acc: 0.1382\n",
      "Epoch 31/50\n",
      "246/246 [==============================] - 0s 45us/step - loss: 2.8376 - acc: 0.1585\n",
      "Epoch 32/50\n",
      "246/246 [==============================] - 0s 44us/step - loss: 2.8377 - acc: 0.1789\n",
      "Epoch 33/50\n",
      "246/246 [==============================] - 0s 49us/step - loss: 2.8295 - acc: 0.1992\n",
      "Epoch 34/50\n",
      "246/246 [==============================] - 0s 51us/step - loss: 2.8325 - acc: 0.1545\n",
      "Epoch 35/50\n",
      "246/246 [==============================] - 0s 76us/step - loss: 2.8247 - acc: 0.1911\n",
      "Epoch 36/50\n",
      "246/246 [==============================] - 0s 61us/step - loss: 2.8221 - acc: 0.2033\n",
      "Epoch 37/50\n",
      "246/246 [==============================] - 0s 58us/step - loss: 2.8177 - acc: 0.1992\n",
      "Epoch 38/50\n",
      "246/246 [==============================] - 0s 58us/step - loss: 2.8172 - acc: 0.1992\n",
      "Epoch 39/50\n",
      "246/246 [==============================] - 0s 58us/step - loss: 2.8152 - acc: 0.2033\n",
      "Epoch 40/50\n",
      "246/246 [==============================] - 0s 56us/step - loss: 2.8116 - acc: 0.2033\n",
      "Epoch 41/50\n",
      "246/246 [==============================] - 0s 63us/step - loss: 2.8078 - acc: 0.1707\n",
      "Epoch 42/50\n",
      "246/246 [==============================] - 0s 62us/step - loss: 2.8045 - acc: 0.2154\n",
      "Epoch 43/50\n",
      "246/246 [==============================] - 0s 62us/step - loss: 2.8015 - acc: 0.2154\n",
      "Epoch 44/50\n",
      "246/246 [==============================] - 0s 62us/step - loss: 2.8007 - acc: 0.2195\n",
      "Epoch 45/50\n",
      "246/246 [==============================] - 0s 59us/step - loss: 2.7965 - acc: 0.2114\n",
      "Epoch 46/50\n",
      "246/246 [==============================] - 0s 59us/step - loss: 2.7903 - acc: 0.2236\n",
      "Epoch 47/50\n",
      "246/246 [==============================] - 0s 56us/step - loss: 2.7931 - acc: 0.2195\n",
      "Epoch 48/50\n",
      "246/246 [==============================] - 0s 57us/step - loss: 2.7874 - acc: 0.2114\n",
      "Epoch 49/50\n",
      "246/246 [==============================] - 0s 54us/step - loss: 2.7832 - acc: 0.2195\n",
      "Epoch 50/50\n",
      "246/246 [==============================] - 0s 52us/step - loss: 2.7829 - acc: 0.2154\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "22/22 [==============================] - 0s 13ms/step - loss: 2.5066 - acc: 0.0455\n",
      "Epoch 2/50\n",
      "22/22 [==============================] - 0s 305us/step - loss: 2.3294 - acc: 0.0455\n",
      "Epoch 3/50\n",
      "22/22 [==============================] - 0s 273us/step - loss: 2.2277 - acc: 0.0455\n",
      "Epoch 4/50\n",
      "22/22 [==============================] - 0s 316us/step - loss: 2.1571 - acc: 0.0455\n",
      "Epoch 5/50\n",
      "22/22 [==============================] - 0s 296us/step - loss: 2.1043 - acc: 0.0455\n",
      "Epoch 6/50\n",
      "22/22 [==============================] - 0s 824us/step - loss: 2.0631 - acc: 0.0455\n",
      "Epoch 7/50\n",
      "22/22 [==============================] - 0s 294us/step - loss: 2.0301 - acc: 0.1818\n",
      "Epoch 8/50\n",
      "22/22 [==============================] - 0s 267us/step - loss: 2.0032 - acc: 0.1818\n",
      "Epoch 9/50\n",
      "22/22 [==============================] - 0s 301us/step - loss: 1.9807 - acc: 0.2273\n",
      "Epoch 10/50\n",
      "22/22 [==============================] - 0s 295us/step - loss: 1.9619 - acc: 0.2273\n",
      "Epoch 11/50\n",
      "22/22 [==============================] - 0s 279us/step - loss: 1.9458 - acc: 0.2727\n",
      "Epoch 12/50\n",
      "22/22 [==============================] - 0s 273us/step - loss: 1.9320 - acc: 0.2727\n",
      "Epoch 13/50\n",
      "22/22 [==============================] - 0s 275us/step - loss: 1.9200 - acc: 0.2727\n",
      "Epoch 14/50\n",
      "22/22 [==============================] - 0s 299us/step - loss: 1.9095 - acc: 0.2727\n",
      "Epoch 15/50\n",
      "22/22 [==============================] - 0s 298us/step - loss: 1.9003 - acc: 0.2727\n",
      "Epoch 16/50\n",
      "22/22 [==============================] - 0s 304us/step - loss: 1.8920 - acc: 0.2727\n",
      "Epoch 17/50\n",
      "22/22 [==============================] - 0s 291us/step - loss: 1.8846 - acc: 0.2727\n",
      "Epoch 18/50\n",
      "22/22 [==============================] - 0s 291us/step - loss: 1.8779 - acc: 0.2727\n",
      "Epoch 19/50\n",
      "22/22 [==============================] - 0s 286us/step - loss: 1.8718 - acc: 0.2727\n",
      "Epoch 20/50\n",
      "22/22 [==============================] - 0s 307us/step - loss: 1.8661 - acc: 0.2727\n",
      "Epoch 21/50\n",
      "22/22 [==============================] - 0s 304us/step - loss: 1.8609 - acc: 0.2727\n",
      "Epoch 22/50\n",
      "22/22 [==============================] - 0s 283us/step - loss: 1.8559 - acc: 0.2727\n",
      "Epoch 23/50\n",
      "22/22 [==============================] - 0s 276us/step - loss: 1.8513 - acc: 0.2727\n",
      "Epoch 24/50\n",
      "22/22 [==============================] - 0s 273us/step - loss: 1.8468 - acc: 0.2727\n",
      "Epoch 25/50\n",
      "22/22 [==============================] - 0s 287us/step - loss: 1.8426 - acc: 0.2727\n",
      "Epoch 26/50\n",
      "22/22 [==============================] - 0s 281us/step - loss: 1.8385 - acc: 0.2727\n",
      "Epoch 27/50\n",
      "22/22 [==============================] - 0s 291us/step - loss: 1.8345 - acc: 0.2727\n",
      "Epoch 28/50\n",
      "22/22 [==============================] - 0s 306us/step - loss: 1.8306 - acc: 0.2727\n",
      "Epoch 29/50\n",
      "22/22 [==============================] - 0s 320us/step - loss: 1.8269 - acc: 0.2727\n",
      "Epoch 30/50\n",
      "22/22 [==============================] - 0s 297us/step - loss: 1.8231 - acc: 0.2727\n",
      "Epoch 31/50\n",
      "22/22 [==============================] - 0s 296us/step - loss: 1.8195 - acc: 0.2727\n",
      "Epoch 32/50\n",
      "22/22 [==============================] - 0s 296us/step - loss: 1.8159 - acc: 0.2727\n",
      "Epoch 33/50\n",
      "22/22 [==============================] - 0s 289us/step - loss: 1.8123 - acc: 0.2727\n",
      "Epoch 34/50\n",
      "22/22 [==============================] - 0s 267us/step - loss: 1.8087 - acc: 0.2727\n",
      "Epoch 35/50\n",
      "22/22 [==============================] - 0s 245us/step - loss: 1.8052 - acc: 0.2727\n",
      "Epoch 36/50\n",
      "22/22 [==============================] - 0s 261us/step - loss: 1.8017 - acc: 0.2727\n",
      "Epoch 37/50\n",
      "22/22 [==============================] - 0s 298us/step - loss: 1.7982 - acc: 0.2727\n",
      "Epoch 38/50\n",
      "22/22 [==============================] - 0s 273us/step - loss: 1.7947 - acc: 0.2727\n",
      "Epoch 39/50\n",
      "22/22 [==============================] - 0s 252us/step - loss: 1.7912 - acc: 0.2727\n",
      "Epoch 40/50\n",
      "22/22 [==============================] - 0s 311us/step - loss: 1.7877 - acc: 0.2727\n",
      "Epoch 41/50\n",
      "22/22 [==============================] - 0s 291us/step - loss: 1.7842 - acc: 0.2727\n",
      "Epoch 42/50\n",
      "22/22 [==============================] - 0s 252us/step - loss: 1.7807 - acc: 0.2727\n",
      "Epoch 43/50\n",
      "22/22 [==============================] - 0s 253us/step - loss: 1.7772 - acc: 0.2727\n",
      "Epoch 44/50\n",
      "22/22 [==============================] - 0s 279us/step - loss: 1.7737 - acc: 0.2727\n",
      "Epoch 45/50\n",
      "22/22 [==============================] - 0s 283us/step - loss: 1.7702 - acc: 0.2727\n",
      "Epoch 46/50\n",
      "22/22 [==============================] - 0s 269us/step - loss: 1.7666 - acc: 0.2727\n",
      "Epoch 47/50\n",
      "22/22 [==============================] - 0s 272us/step - loss: 1.7631 - acc: 0.2727\n",
      "Epoch 48/50\n",
      "22/22 [==============================] - 0s 321us/step - loss: 1.7597 - acc: 0.2727\n",
      "Epoch 49/50\n",
      "22/22 [==============================] - 0s 289us/step - loss: 1.7563 - acc: 0.2727\n",
      "Epoch 50/50\n",
      "22/22 [==============================] - 0s 310us/step - loss: 1.7527 - acc: 0.3182\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "6291/6291 [==============================] - 1s 102us/step - loss: 4.0679 - acc: 0.2629\n",
      "Epoch 2/50\n",
      "6291/6291 [==============================] - 0s 66us/step - loss: 3.1852 - acc: 0.3319\n",
      "Epoch 3/50\n",
      "6291/6291 [==============================] - 0s 69us/step - loss: 3.1243 - acc: 0.3319\n",
      "Epoch 4/50\n",
      "6291/6291 [==============================] - 0s 73us/step - loss: 3.0877 - acc: 0.3319\n",
      "Epoch 5/50\n",
      "6291/6291 [==============================] - 0s 73us/step - loss: 3.0604 - acc: 0.3319\n",
      "Epoch 6/50\n",
      "6291/6291 [==============================] - 0s 73us/step - loss: 3.0339 - acc: 0.3319\n",
      "Epoch 7/50\n",
      "6291/6291 [==============================] - 0s 76us/step - loss: 3.0105 - acc: 0.3325\n",
      "Epoch 8/50\n",
      "6291/6291 [==============================] - 0s 69us/step - loss: 2.9875 - acc: 0.3344\n",
      "Epoch 9/50\n",
      "6291/6291 [==============================] - 0s 67us/step - loss: 2.9642 - acc: 0.3346\n",
      "Epoch 10/50\n",
      "6291/6291 [==============================] - 0s 70us/step - loss: 2.9417 - acc: 0.3387\n",
      "Epoch 11/50\n",
      "6291/6291 [==============================] - 0s 73us/step - loss: 2.9213 - acc: 0.3437\n",
      "Epoch 12/50\n",
      "6291/6291 [==============================] - 0s 69us/step - loss: 2.9004 - acc: 0.3480\n",
      "Epoch 13/50\n",
      "6291/6291 [==============================] - 0s 68us/step - loss: 2.8818 - acc: 0.3526\n",
      "Epoch 14/50\n",
      "6291/6291 [==============================] - 0s 63us/step - loss: 2.8621 - acc: 0.3567\n",
      "Epoch 15/50\n",
      "6291/6291 [==============================] - 0s 64us/step - loss: 2.8420 - acc: 0.3608\n",
      "Epoch 16/50\n",
      "6291/6291 [==============================] - 0s 71us/step - loss: 2.8232 - acc: 0.3634\n",
      "Epoch 17/50\n",
      "6291/6291 [==============================] - 0s 69us/step - loss: 2.8051 - acc: 0.3681\n",
      "Epoch 18/50\n",
      "6291/6291 [==============================] - 0s 72us/step - loss: 2.7857 - acc: 0.3704\n",
      "Epoch 19/50\n",
      "6291/6291 [==============================] - 0s 70us/step - loss: 2.7683 - acc: 0.3720\n",
      "Epoch 20/50\n",
      "6291/6291 [==============================] - 0s 67us/step - loss: 2.7507 - acc: 0.3755\n",
      "Epoch 21/50\n",
      "6291/6291 [==============================] - 0s 70us/step - loss: 2.7346 - acc: 0.3780\n",
      "Epoch 22/50\n",
      "6291/6291 [==============================] - 0s 64us/step - loss: 2.7181 - acc: 0.3793\n",
      "Epoch 23/50\n",
      "6291/6291 [==============================] - 0s 63us/step - loss: 2.7023 - acc: 0.3825\n",
      "Epoch 24/50\n",
      "6291/6291 [==============================] - 0s 68us/step - loss: 2.6855 - acc: 0.3882\n",
      "Epoch 25/50\n",
      "6291/6291 [==============================] - 0s 71us/step - loss: 2.6706 - acc: 0.3906\n",
      "Epoch 26/50\n",
      "6291/6291 [==============================] - 0s 65us/step - loss: 2.6553 - acc: 0.3945\n",
      "Epoch 27/50\n",
      "6291/6291 [==============================] - 0s 65us/step - loss: 2.6395 - acc: 0.3998\n",
      "Epoch 28/50\n",
      "6291/6291 [==============================] - 0s 68us/step - loss: 2.6267 - acc: 0.3995\n",
      "Epoch 29/50\n",
      "6291/6291 [==============================] - 0s 73us/step - loss: 2.6121 - acc: 0.4052\n",
      "Epoch 30/50\n",
      "6291/6291 [==============================] - 0s 64us/step - loss: 2.5988 - acc: 0.4085\n",
      "Epoch 31/50\n",
      "6291/6291 [==============================] - 0s 68us/step - loss: 2.5852 - acc: 0.4125\n",
      "Epoch 32/50\n",
      "6291/6291 [==============================] - 0s 64us/step - loss: 2.5720 - acc: 0.4128\n",
      "Epoch 33/50\n",
      "6291/6291 [==============================] - 0s 67us/step - loss: 2.5581 - acc: 0.4187\n",
      "Epoch 34/50\n",
      "6291/6291 [==============================] - 0s 71us/step - loss: 2.5471 - acc: 0.4189\n",
      "Epoch 35/50\n",
      "6291/6291 [==============================] - 0s 72us/step - loss: 2.5343 - acc: 0.4227\n",
      "Epoch 36/50\n",
      "6291/6291 [==============================] - 0s 71us/step - loss: 2.5219 - acc: 0.4217\n",
      "Epoch 37/50\n",
      "6291/6291 [==============================] - 0s 69us/step - loss: 2.5101 - acc: 0.4247\n",
      "Epoch 38/50\n",
      "6291/6291 [==============================] - 0s 71us/step - loss: 2.4986 - acc: 0.4268\n",
      "Epoch 39/50\n",
      "6291/6291 [==============================] - 0s 62us/step - loss: 2.4868 - acc: 0.4297\n",
      "Epoch 40/50\n",
      "6291/6291 [==============================] - 0s 70us/step - loss: 2.4771 - acc: 0.4297\n",
      "Epoch 41/50\n",
      "6291/6291 [==============================] - 0s 62us/step - loss: 2.4668 - acc: 0.4319\n",
      "Epoch 42/50\n",
      "6291/6291 [==============================] - 0s 61us/step - loss: 2.4557 - acc: 0.4335\n",
      "Epoch 43/50\n",
      "6291/6291 [==============================] - 0s 63us/step - loss: 2.4450 - acc: 0.4330\n",
      "Epoch 44/50\n",
      "6291/6291 [==============================] - 0s 68us/step - loss: 2.4344 - acc: 0.4362\n",
      "Epoch 45/50\n",
      "6291/6291 [==============================] - 0s 69us/step - loss: 2.4263 - acc: 0.4365\n",
      "Epoch 46/50\n",
      "6291/6291 [==============================] - 0s 68us/step - loss: 2.4160 - acc: 0.4379\n",
      "Epoch 47/50\n",
      "6291/6291 [==============================] - 0s 68us/step - loss: 2.4066 - acc: 0.4373\n",
      "Epoch 48/50\n",
      "6291/6291 [==============================] - 0s 67us/step - loss: 2.3967 - acc: 0.4402\n",
      "Epoch 49/50\n",
      "6291/6291 [==============================] - 0s 71us/step - loss: 2.3885 - acc: 0.4440\n",
      "Epoch 50/50\n",
      "6291/6291 [==============================] - 0s 67us/step - loss: 2.3788 - acc: 0.4422\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "3711/3711 [==============================] - 0s 106us/step - loss: 4.2125 - acc: 0.2137\n",
      "Epoch 2/50\n",
      "3711/3711 [==============================] - 0s 58us/step - loss: 3.3778 - acc: 0.2953\n",
      "Epoch 3/50\n",
      "3711/3711 [==============================] - 0s 62us/step - loss: 3.2538 - acc: 0.2953\n",
      "Epoch 4/50\n",
      "3711/3711 [==============================] - 0s 58us/step - loss: 3.2133 - acc: 0.2953\n",
      "Epoch 5/50\n",
      "3711/3711 [==============================] - 0s 51us/step - loss: 3.1875 - acc: 0.2953\n",
      "Epoch 6/50\n",
      "3711/3711 [==============================] - 0s 62us/step - loss: 3.1627 - acc: 0.2967\n",
      "Epoch 7/50\n",
      "3711/3711 [==============================] - 0s 61us/step - loss: 3.1448 - acc: 0.2999\n",
      "Epoch 8/50\n",
      "3711/3711 [==============================] - 0s 59us/step - loss: 3.1243 - acc: 0.3042\n",
      "Epoch 9/50\n",
      "3711/3711 [==============================] - 0s 60us/step - loss: 3.1069 - acc: 0.3058\n",
      "Epoch 10/50\n",
      "3711/3711 [==============================] - 0s 59us/step - loss: 3.0893 - acc: 0.3129\n",
      "Epoch 11/50\n",
      "3711/3711 [==============================] - 0s 61us/step - loss: 3.0729 - acc: 0.3131\n",
      "Epoch 12/50\n",
      "3711/3711 [==============================] - 0s 55us/step - loss: 3.0582 - acc: 0.3134\n",
      "Epoch 13/50\n",
      "3711/3711 [==============================] - 0s 50us/step - loss: 3.0414 - acc: 0.3188\n",
      "Epoch 14/50\n",
      "3711/3711 [==============================] - 0s 50us/step - loss: 3.0273 - acc: 0.3196\n",
      "Epoch 15/50\n",
      "3711/3711 [==============================] - 0s 54us/step - loss: 3.0130 - acc: 0.3217\n",
      "Epoch 16/50\n",
      "3711/3711 [==============================] - 0s 61us/step - loss: 2.9976 - acc: 0.3247\n",
      "Epoch 17/50\n",
      "3711/3711 [==============================] - 0s 64us/step - loss: 2.9859 - acc: 0.3228\n",
      "Epoch 18/50\n",
      "3711/3711 [==============================] - 0s 57us/step - loss: 2.9729 - acc: 0.3282\n",
      "Epoch 19/50\n",
      "3711/3711 [==============================] - 0s 49us/step - loss: 2.9594 - acc: 0.3352\n",
      "Epoch 20/50\n",
      "3711/3711 [==============================] - 0s 59us/step - loss: 2.9455 - acc: 0.3363\n",
      "Epoch 21/50\n",
      "3711/3711 [==============================] - 0s 61us/step - loss: 2.9335 - acc: 0.3422\n",
      "Epoch 22/50\n",
      "3711/3711 [==============================] - 0s 58us/step - loss: 2.9200 - acc: 0.3455\n",
      "Epoch 23/50\n",
      "3711/3711 [==============================] - 0s 57us/step - loss: 2.9088 - acc: 0.3525\n",
      "Epoch 24/50\n",
      "3711/3711 [==============================] - 0s 63us/step - loss: 2.8953 - acc: 0.3554\n",
      "Epoch 25/50\n",
      "3711/3711 [==============================] - 0s 56us/step - loss: 2.8846 - acc: 0.3587\n",
      "Epoch 26/50\n",
      "3711/3711 [==============================] - 0s 51us/step - loss: 2.8701 - acc: 0.3643\n",
      "Epoch 27/50\n",
      "3711/3711 [==============================] - 0s 54us/step - loss: 2.8581 - acc: 0.3711\n",
      "Epoch 28/50\n",
      "3711/3711 [==============================] - 0s 52us/step - loss: 2.8460 - acc: 0.3708\n",
      "Epoch 29/50\n",
      "3711/3711 [==============================] - 0s 59us/step - loss: 2.8344 - acc: 0.3794\n",
      "Epoch 30/50\n",
      "3711/3711 [==============================] - 0s 57us/step - loss: 2.8225 - acc: 0.3810\n",
      "Epoch 31/50\n",
      "3711/3711 [==============================] - 0s 60us/step - loss: 2.8109 - acc: 0.3845\n",
      "Epoch 32/50\n",
      "3711/3711 [==============================] - 0s 60us/step - loss: 2.8000 - acc: 0.3915\n",
      "Epoch 33/50\n",
      "3711/3711 [==============================] - 0s 64us/step - loss: 2.7879 - acc: 0.3934\n",
      "Epoch 34/50\n",
      "3711/3711 [==============================] - 0s 54us/step - loss: 2.7757 - acc: 0.3977\n",
      "Epoch 35/50\n",
      "3711/3711 [==============================] - 0s 55us/step - loss: 2.7637 - acc: 0.3980\n",
      "Epoch 36/50\n",
      "3711/3711 [==============================] - 0s 54us/step - loss: 2.7537 - acc: 0.4037\n",
      "Epoch 37/50\n",
      "3711/3711 [==============================] - 0s 51us/step - loss: 2.7413 - acc: 0.4042\n",
      "Epoch 38/50\n",
      "3711/3711 [==============================] - 0s 49us/step - loss: 2.7310 - acc: 0.4064\n",
      "Epoch 39/50\n",
      "3711/3711 [==============================] - 0s 58us/step - loss: 2.7189 - acc: 0.4117\n",
      "Epoch 40/50\n",
      "3711/3711 [==============================] - 0s 55us/step - loss: 2.7083 - acc: 0.4161\n",
      "Epoch 41/50\n",
      "3711/3711 [==============================] - 0s 61us/step - loss: 2.6972 - acc: 0.4155\n",
      "Epoch 42/50\n",
      "3711/3711 [==============================] - 0s 58us/step - loss: 2.6870 - acc: 0.4241\n",
      "Epoch 43/50\n",
      "3711/3711 [==============================] - 0s 59us/step - loss: 2.6763 - acc: 0.4214\n",
      "Epoch 44/50\n",
      "3711/3711 [==============================] - 0s 59us/step - loss: 2.6659 - acc: 0.4252\n",
      "Epoch 45/50\n",
      "3711/3711 [==============================] - 0s 63us/step - loss: 2.6547 - acc: 0.4285\n",
      "Epoch 46/50\n",
      "3711/3711 [==============================] - 0s 56us/step - loss: 2.6446 - acc: 0.4325\n",
      "Epoch 47/50\n",
      "3711/3711 [==============================] - 0s 54us/step - loss: 2.6352 - acc: 0.4322\n",
      "Epoch 48/50\n",
      "3711/3711 [==============================] - 0s 51us/step - loss: 2.6241 - acc: 0.4344\n",
      "Epoch 49/50\n",
      "3711/3711 [==============================] - 0s 54us/step - loss: 2.6142 - acc: 0.4384\n",
      "Epoch 50/50\n",
      "3711/3711 [==============================] - 0s 57us/step - loss: 2.6050 - acc: 0.4400\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "472/472 [==============================] - 0s 590us/step - loss: 4.0170 - acc: 0.0021\n",
      "Epoch 2/50\n",
      "472/472 [==============================] - 0s 55us/step - loss: 3.4015 - acc: 0.0593\n",
      "Epoch 3/50\n",
      "472/472 [==============================] - 0s 55us/step - loss: 3.0116 - acc: 0.5784\n",
      "Epoch 4/50\n",
      "472/472 [==============================] - 0s 58us/step - loss: 2.7133 - acc: 0.5911\n",
      "Epoch 5/50\n",
      "472/472 [==============================] - 0s 55us/step - loss: 2.4901 - acc: 0.5911\n",
      "Epoch 6/50\n",
      "472/472 [==============================] - 0s 63us/step - loss: 2.3189 - acc: 0.5911\n",
      "Epoch 7/50\n",
      "472/472 [==============================] - 0s 56us/step - loss: 2.1976 - acc: 0.5911\n",
      "Epoch 8/50\n",
      "472/472 [==============================] - 0s 60us/step - loss: 2.1133 - acc: 0.5911\n",
      "Epoch 9/50\n",
      "472/472 [==============================] - 0s 64us/step - loss: 2.0568 - acc: 0.5911\n",
      "Epoch 10/50\n",
      "472/472 [==============================] - 0s 58us/step - loss: 2.0201 - acc: 0.5911\n",
      "Epoch 11/50\n",
      "472/472 [==============================] - 0s 61us/step - loss: 1.9976 - acc: 0.5911\n",
      "Epoch 12/50\n",
      "472/472 [==============================] - 0s 56us/step - loss: 1.9804 - acc: 0.5911\n",
      "Epoch 13/50\n",
      "472/472 [==============================] - 0s 61us/step - loss: 1.9692 - acc: 0.5911\n",
      "Epoch 14/50\n",
      "472/472 [==============================] - 0s 54us/step - loss: 1.9607 - acc: 0.5911\n",
      "Epoch 15/50\n",
      "472/472 [==============================] - 0s 59us/step - loss: 1.9515 - acc: 0.5911\n",
      "Epoch 16/50\n",
      "472/472 [==============================] - 0s 55us/step - loss: 1.9463 - acc: 0.5911\n",
      "Epoch 17/50\n",
      "472/472 [==============================] - 0s 56us/step - loss: 1.9415 - acc: 0.5911\n",
      "Epoch 18/50\n",
      "472/472 [==============================] - 0s 51us/step - loss: 1.9375 - acc: 0.5911\n",
      "Epoch 19/50\n",
      "472/472 [==============================] - 0s 54us/step - loss: 1.9289 - acc: 0.5911\n",
      "Epoch 20/50\n",
      "472/472 [==============================] - 0s 51us/step - loss: 1.9269 - acc: 0.5911\n",
      "Epoch 21/50\n",
      "472/472 [==============================] - 0s 54us/step - loss: 1.9246 - acc: 0.5911\n",
      "Epoch 22/50\n",
      "472/472 [==============================] - 0s 49us/step - loss: 1.9185 - acc: 0.5911\n",
      "Epoch 23/50\n",
      "472/472 [==============================] - 0s 52us/step - loss: 1.9145 - acc: 0.5911\n",
      "Epoch 24/50\n",
      "472/472 [==============================] - 0s 57us/step - loss: 1.9097 - acc: 0.5911\n",
      "Epoch 25/50\n",
      "472/472 [==============================] - 0s 58us/step - loss: 1.9074 - acc: 0.5911\n",
      "Epoch 26/50\n",
      "472/472 [==============================] - 0s 56us/step - loss: 1.9031 - acc: 0.5911\n",
      "Epoch 27/50\n",
      "472/472 [==============================] - 0s 60us/step - loss: 1.8990 - acc: 0.5911\n",
      "Epoch 28/50\n",
      "472/472 [==============================] - 0s 56us/step - loss: 1.8976 - acc: 0.5911\n",
      "Epoch 29/50\n",
      "472/472 [==============================] - 0s 56us/step - loss: 1.8925 - acc: 0.5911\n",
      "Epoch 30/50\n",
      "472/472 [==============================] - 0s 65us/step - loss: 1.8914 - acc: 0.5911\n",
      "Epoch 31/50\n",
      "472/472 [==============================] - 0s 55us/step - loss: 1.8868 - acc: 0.5911\n",
      "Epoch 32/50\n",
      "472/472 [==============================] - 0s 56us/step - loss: 1.8854 - acc: 0.5911\n",
      "Epoch 33/50\n",
      "472/472 [==============================] - 0s 73us/step - loss: 1.8835 - acc: 0.5911\n",
      "Epoch 34/50\n",
      "472/472 [==============================] - 0s 54us/step - loss: 1.8773 - acc: 0.5911\n",
      "Epoch 35/50\n",
      "472/472 [==============================] - 0s 49us/step - loss: 1.8787 - acc: 0.5911\n",
      "Epoch 36/50\n",
      "472/472 [==============================] - 0s 57us/step - loss: 1.8756 - acc: 0.5911\n",
      "Epoch 37/50\n",
      "472/472 [==============================] - 0s 47us/step - loss: 1.8702 - acc: 0.5911\n",
      "Epoch 38/50\n",
      "472/472 [==============================] - 0s 47us/step - loss: 1.8678 - acc: 0.5911\n",
      "Epoch 39/50\n",
      "472/472 [==============================] - 0s 47us/step - loss: 1.8661 - acc: 0.5911\n",
      "Epoch 40/50\n",
      "472/472 [==============================] - 0s 48us/step - loss: 1.8620 - acc: 0.5911\n",
      "Epoch 41/50\n",
      "472/472 [==============================] - 0s 55us/step - loss: 1.8609 - acc: 0.5911\n",
      "Epoch 42/50\n",
      "472/472 [==============================] - 0s 63us/step - loss: 1.8591 - acc: 0.5911\n",
      "Epoch 43/50\n",
      "472/472 [==============================] - 0s 52us/step - loss: 1.8548 - acc: 0.5911\n",
      "Epoch 44/50\n",
      "472/472 [==============================] - 0s 52us/step - loss: 1.8535 - acc: 0.5911\n",
      "Epoch 45/50\n",
      "472/472 [==============================] - 0s 49us/step - loss: 1.8507 - acc: 0.5911\n",
      "Epoch 46/50\n",
      "472/472 [==============================] - 0s 53us/step - loss: 1.8482 - acc: 0.5911\n",
      "Epoch 47/50\n",
      "472/472 [==============================] - 0s 58us/step - loss: 1.8436 - acc: 0.5911\n",
      "Epoch 48/50\n",
      "472/472 [==============================] - 0s 57us/step - loss: 1.8437 - acc: 0.5911\n",
      "Epoch 49/50\n",
      "472/472 [==============================] - 0s 58us/step - loss: 1.8394 - acc: 0.5911\n",
      "Epoch 50/50\n",
      "472/472 [==============================] - 0s 61us/step - loss: 1.8382 - acc: 0.5911\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "1641/1641 [==============================] - 0s 178us/step - loss: 5.8056 - acc: 0.0079\n",
      "Epoch 2/50\n",
      "1641/1641 [==============================] - 0s 52us/step - loss: 5.3743 - acc: 0.0317\n",
      "Epoch 3/50\n",
      "1641/1641 [==============================] - 0s 58us/step - loss: 5.2031 - acc: 0.0573\n",
      "Epoch 4/50\n",
      "1641/1641 [==============================] - 0s 59us/step - loss: 5.1267 - acc: 0.0555\n",
      "Epoch 5/50\n",
      "1641/1641 [==============================] - 0s 58us/step - loss: 5.0926 - acc: 0.0603\n",
      "Epoch 6/50\n",
      "1641/1641 [==============================] - 0s 52us/step - loss: 5.0735 - acc: 0.0536\n",
      "Epoch 7/50\n",
      "1641/1641 [==============================] - 0s 53us/step - loss: 5.0602 - acc: 0.0591\n",
      "Epoch 8/50\n",
      "1641/1641 [==============================] - 0s 59us/step - loss: 5.0484 - acc: 0.0628\n",
      "Epoch 9/50\n",
      "1641/1641 [==============================] - 0s 57us/step - loss: 5.0405 - acc: 0.0652\n",
      "Epoch 10/50\n",
      "1641/1641 [==============================] - 0s 59us/step - loss: 5.0314 - acc: 0.0615\n",
      "Epoch 11/50\n",
      "1641/1641 [==============================] - 0s 59us/step - loss: 5.0219 - acc: 0.0695\n",
      "Epoch 12/50\n",
      "1641/1641 [==============================] - 0s 56us/step - loss: 5.0141 - acc: 0.0695\n",
      "Epoch 13/50\n",
      "1641/1641 [==============================] - 0s 56us/step - loss: 5.0081 - acc: 0.0658\n",
      "Epoch 14/50\n",
      "1641/1641 [==============================] - 0s 56us/step - loss: 4.9981 - acc: 0.0658\n",
      "Epoch 15/50\n",
      "1641/1641 [==============================] - 0s 54us/step - loss: 4.9898 - acc: 0.0652\n",
      "Epoch 16/50\n",
      "1641/1641 [==============================] - 0s 50us/step - loss: 4.9834 - acc: 0.0707\n",
      "Epoch 17/50\n",
      "1641/1641 [==============================] - 0s 50us/step - loss: 4.9754 - acc: 0.0707\n",
      "Epoch 18/50\n",
      "1641/1641 [==============================] - 0s 55us/step - loss: 4.9674 - acc: 0.0719\n",
      "Epoch 19/50\n",
      "1641/1641 [==============================] - 0s 46us/step - loss: 4.9584 - acc: 0.0689\n",
      "Epoch 20/50\n",
      "1641/1641 [==============================] - 0s 47us/step - loss: 4.9527 - acc: 0.0683\n",
      "Epoch 21/50\n",
      "1641/1641 [==============================] - 0s 56us/step - loss: 4.9445 - acc: 0.0689\n",
      "Epoch 22/50\n",
      "1641/1641 [==============================] - 0s 51us/step - loss: 4.9366 - acc: 0.0743\n",
      "Epoch 23/50\n",
      "1641/1641 [==============================] - 0s 54us/step - loss: 4.9290 - acc: 0.0725\n",
      "Epoch 24/50\n",
      "1641/1641 [==============================] - 0s 49us/step - loss: 4.9232 - acc: 0.0737\n",
      "Epoch 25/50\n",
      "1641/1641 [==============================] - 0s 50us/step - loss: 4.9148 - acc: 0.0786\n",
      "Epoch 26/50\n",
      "1641/1641 [==============================] - 0s 48us/step - loss: 4.9080 - acc: 0.0719\n",
      "Epoch 27/50\n",
      "1641/1641 [==============================] - 0s 44us/step - loss: 4.8990 - acc: 0.0859\n",
      "Epoch 28/50\n",
      "1641/1641 [==============================] - 0s 43us/step - loss: 4.8933 - acc: 0.0713\n",
      "Epoch 29/50\n",
      "1641/1641 [==============================] - 0s 46us/step - loss: 4.8858 - acc: 0.0859\n",
      "Epoch 30/50\n",
      "1641/1641 [==============================] - 0s 54us/step - loss: 4.8781 - acc: 0.0841\n",
      "Epoch 31/50\n",
      "1641/1641 [==============================] - 0s 53us/step - loss: 4.8713 - acc: 0.0932\n",
      "Epoch 32/50\n",
      "1641/1641 [==============================] - 0s 55us/step - loss: 4.8638 - acc: 0.0914\n",
      "Epoch 33/50\n",
      "1641/1641 [==============================] - 0s 56us/step - loss: 4.8566 - acc: 0.0823\n",
      "Epoch 34/50\n",
      "1641/1641 [==============================] - 0s 56us/step - loss: 4.8492 - acc: 0.0938\n",
      "Epoch 35/50\n",
      "1641/1641 [==============================] - 0s 53us/step - loss: 4.8425 - acc: 0.0896\n",
      "Epoch 36/50\n",
      "1641/1641 [==============================] - 0s 48us/step - loss: 4.8356 - acc: 0.1073\n",
      "Epoch 37/50\n",
      "1641/1641 [==============================] - 0s 43us/step - loss: 4.8298 - acc: 0.1079\n",
      "Epoch 38/50\n",
      "1641/1641 [==============================] - 0s 43us/step - loss: 4.8221 - acc: 0.0853\n",
      "Epoch 39/50\n",
      "1641/1641 [==============================] - 0s 45us/step - loss: 4.8145 - acc: 0.1091\n",
      "Epoch 40/50\n",
      "1641/1641 [==============================] - 0s 57us/step - loss: 4.8079 - acc: 0.1109\n",
      "Epoch 41/50\n",
      "1641/1641 [==============================] - 0s 57us/step - loss: 4.8002 - acc: 0.1200\n",
      "Epoch 42/50\n",
      "1641/1641 [==============================] - 0s 57us/step - loss: 4.7946 - acc: 0.1152\n",
      "Epoch 43/50\n",
      "1641/1641 [==============================] - 0s 51us/step - loss: 4.7864 - acc: 0.1103\n",
      "Epoch 44/50\n",
      "1641/1641 [==============================] - 0s 48us/step - loss: 4.7795 - acc: 0.1231\n",
      "Epoch 45/50\n",
      "1641/1641 [==============================] - 0s 48us/step - loss: 4.7734 - acc: 0.1194\n",
      "Epoch 46/50\n",
      "1641/1641 [==============================] - 0s 51us/step - loss: 4.7658 - acc: 0.1231\n",
      "Epoch 47/50\n",
      "1641/1641 [==============================] - 0s 49us/step - loss: 4.7598 - acc: 0.1243\n",
      "Epoch 48/50\n",
      "1641/1641 [==============================] - 0s 50us/step - loss: 4.7511 - acc: 0.1463\n",
      "Epoch 49/50\n",
      "1641/1641 [==============================] - 0s 55us/step - loss: 4.7447 - acc: 0.1371\n",
      "Epoch 50/50\n",
      "1641/1641 [==============================] - 0s 52us/step - loss: 4.7385 - acc: 0.1231\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "2487/2487 [==============================] - 0s 129us/step - loss: 4.0782 - acc: 0.0869\n",
      "Epoch 2/50\n",
      "2487/2487 [==============================] - 0s 54us/step - loss: 3.3253 - acc: 0.1122\n",
      "Epoch 3/50\n",
      "2487/2487 [==============================] - 0s 55us/step - loss: 3.1194 - acc: 0.1870\n",
      "Epoch 4/50\n",
      "2487/2487 [==============================] - 0s 58us/step - loss: 3.0736 - acc: 0.1878\n",
      "Epoch 5/50\n",
      "2487/2487 [==============================] - 0s 54us/step - loss: 3.0488 - acc: 0.1998\n",
      "Epoch 6/50\n",
      "2487/2487 [==============================] - 0s 58us/step - loss: 3.0288 - acc: 0.2079\n",
      "Epoch 7/50\n",
      "2487/2487 [==============================] - 0s 56us/step - loss: 3.0163 - acc: 0.2159\n",
      "Epoch 8/50\n",
      "2487/2487 [==============================] - 0s 48us/step - loss: 3.0025 - acc: 0.2139\n",
      "Epoch 9/50\n",
      "2487/2487 [==============================] - 0s 54us/step - loss: 2.9915 - acc: 0.2203\n",
      "Epoch 10/50\n",
      "2487/2487 [==============================] - 0s 57us/step - loss: 2.9815 - acc: 0.2272\n",
      "Epoch 11/50\n",
      "2487/2487 [==============================] - 0s 56us/step - loss: 2.9698 - acc: 0.2308\n",
      "Epoch 12/50\n",
      "2487/2487 [==============================] - 0s 52us/step - loss: 2.9623 - acc: 0.2336\n",
      "Epoch 13/50\n",
      "2487/2487 [==============================] - 0s 58us/step - loss: 2.9521 - acc: 0.2396\n",
      "Epoch 14/50\n",
      "2487/2487 [==============================] - 0s 52us/step - loss: 2.9435 - acc: 0.2477\n",
      "Epoch 15/50\n",
      "2487/2487 [==============================] - 0s 50us/step - loss: 2.9355 - acc: 0.2481\n",
      "Epoch 16/50\n",
      "2487/2487 [==============================] - 0s 55us/step - loss: 2.9276 - acc: 0.2392\n",
      "Epoch 17/50\n",
      "2487/2487 [==============================] - 0s 53us/step - loss: 2.9209 - acc: 0.2481\n",
      "Epoch 18/50\n",
      "2487/2487 [==============================] - 0s 58us/step - loss: 2.9134 - acc: 0.2485\n",
      "Epoch 19/50\n",
      "2487/2487 [==============================] - 0s 58us/step - loss: 2.9058 - acc: 0.2541\n",
      "Epoch 20/50\n",
      "2487/2487 [==============================] - 0s 61us/step - loss: 2.8996 - acc: 0.2485\n",
      "Epoch 21/50\n",
      "2487/2487 [==============================] - 0s 56us/step - loss: 2.8924 - acc: 0.2561\n",
      "Epoch 22/50\n",
      "2487/2487 [==============================] - 0s 48us/step - loss: 2.8881 - acc: 0.2565\n",
      "Epoch 23/50\n",
      "2487/2487 [==============================] - 0s 57us/step - loss: 2.8815 - acc: 0.2537\n",
      "Epoch 24/50\n",
      "2487/2487 [==============================] - 0s 54us/step - loss: 2.8755 - acc: 0.2577\n",
      "Epoch 25/50\n",
      "2487/2487 [==============================] - 0s 59us/step - loss: 2.8695 - acc: 0.2553\n",
      "Epoch 26/50\n",
      "2487/2487 [==============================] - 0s 49us/step - loss: 2.8625 - acc: 0.2666\n",
      "Epoch 27/50\n",
      "2487/2487 [==============================] - 0s 43us/step - loss: 2.8576 - acc: 0.2642\n",
      "Epoch 28/50\n",
      "2487/2487 [==============================] - 0s 43us/step - loss: 2.8525 - acc: 0.2642\n",
      "Epoch 29/50\n",
      "2487/2487 [==============================] - 0s 43us/step - loss: 2.8465 - acc: 0.2670\n",
      "Epoch 30/50\n",
      "2487/2487 [==============================] - 0s 55us/step - loss: 2.8419 - acc: 0.2670\n",
      "Epoch 31/50\n",
      "2487/2487 [==============================] - 0s 51us/step - loss: 2.8358 - acc: 0.2718\n",
      "Epoch 32/50\n",
      "2487/2487 [==============================] - 0s 57us/step - loss: 2.8289 - acc: 0.2662\n",
      "Epoch 33/50\n",
      "2487/2487 [==============================] - 0s 58us/step - loss: 2.8237 - acc: 0.2698\n",
      "Epoch 34/50\n",
      "2487/2487 [==============================] - 0s 56us/step - loss: 2.8204 - acc: 0.2690\n",
      "Epoch 35/50\n",
      "2487/2487 [==============================] - 0s 56us/step - loss: 2.8149 - acc: 0.2762\n",
      "Epoch 36/50\n",
      "2487/2487 [==============================] - 0s 49us/step - loss: 2.8111 - acc: 0.2722\n",
      "Epoch 37/50\n",
      "2487/2487 [==============================] - 0s 48us/step - loss: 2.8057 - acc: 0.2726\n",
      "Epoch 38/50\n",
      "2487/2487 [==============================] - 0s 44us/step - loss: 2.8002 - acc: 0.2766\n",
      "Epoch 39/50\n",
      "2487/2487 [==============================] - 0s 44us/step - loss: 2.7952 - acc: 0.2730\n",
      "Epoch 40/50\n",
      "2487/2487 [==============================] - 0s 49us/step - loss: 2.7907 - acc: 0.2782\n",
      "Epoch 41/50\n",
      "2487/2487 [==============================] - 0s 54us/step - loss: 2.7855 - acc: 0.2831\n",
      "Epoch 42/50\n",
      "2487/2487 [==============================] - 0s 55us/step - loss: 2.7800 - acc: 0.2859\n",
      "Epoch 43/50\n",
      "2487/2487 [==============================] - 0s 51us/step - loss: 2.7745 - acc: 0.2847\n",
      "Epoch 44/50\n",
      "2487/2487 [==============================] - 0s 57us/step - loss: 2.7712 - acc: 0.2839\n",
      "Epoch 45/50\n",
      "2487/2487 [==============================] - 0s 54us/step - loss: 2.7662 - acc: 0.2863\n",
      "Epoch 46/50\n",
      "2487/2487 [==============================] - 0s 44us/step - loss: 2.7629 - acc: 0.2891\n",
      "Epoch 47/50\n",
      "2487/2487 [==============================] - 0s 55us/step - loss: 2.7573 - acc: 0.2899\n",
      "Epoch 48/50\n",
      "2487/2487 [==============================] - 0s 53us/step - loss: 2.7526 - acc: 0.2847\n",
      "Epoch 49/50\n",
      "2487/2487 [==============================] - 0s 57us/step - loss: 2.7491 - acc: 0.2931\n",
      "Epoch 50/50\n",
      "2487/2487 [==============================] - 0s 58us/step - loss: 2.7449 - acc: 0.2899\n"
     ]
    }
   ],
   "source": [
    "experiment_dir = 'nn_doctors'\n",
    "test_submits = []\n",
    "model_names = dict()\n",
    "\n",
    "for doctor in doctors_all:\n",
    "    train_doctor = train[train['Врач'] == doctor]\n",
    "    pipe, model_name = create_pipeline(experiment_dir, 'nn_2dense', no_validation=True, validation_split=0.3)\n",
    "    \n",
    "    train_doctor_y, pop_diagnoses, most_pop_diagnose = preproc_target_train(train_doctor, reduce_classes=False)\n",
    "    \n",
    "    pipe.fit(train_doctor, train_doctor_y)\n",
    "    \n",
    "    model_names[doctor] = model_name\n",
    "    with open(join(experiment_dir, 'model_names.json'), 'w') as f:\n",
    "        json.dump(model_names, f, ensure_ascii=False)\n",
    "    \n",
    "    test_doctor = test[test['Врач'] == doctor]\n",
    "    test_pred = pipe.transform(test_doctor)\n",
    "    \n",
    "    test_doctor_submit = pd.DataFrame({'Id_Записи': test_doctor['Id_Записи'], 'Код_диагноза': test_pred})\n",
    "    test_submits.append(test_doctor_submit)\n",
    "    \n",
    "test_submit = pd.concat(test_submits)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_submit = pd.concat(test_submits)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(29997,)"
      ]
     },
     "execution_count": 111,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.unique(test_submit['Id_Записи']).shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    }
   ],
   "source": [
    "test_submit = pd.concat([\n",
    "    test_submit,\n",
    "    pd.DataFrame({\n",
    "        'Id_Записи': test[(test['Врач'] == 'ЛФК') | (test['Врач'] == 'ЛФК ортопед')]['Id_Записи'],\n",
    "        'Код_диагноза': scipy.stats.mode(train['Код_диагноза'])[0][0]\n",
    "    })\n",
    "])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array(['ЛФК', 'ЛФК ортопед', 'акушер гинеколог', 'аллерголог иммунолог',\n",
       "        'ангиохирург хирург', 'врачебный', 'гастроэнтеролог', 'гинеколог',\n",
       "        'дерматовенеролог', 'дерматовенеролог трихолог', 'кардиолог',\n",
       "        'колопроктолог', 'маммолог онколог', 'невролог',\n",
       "        'невролог травматолог', 'нейрохирург хирург', 'ортопед',\n",
       "        'ортопед травматолог', 'оториноларинголог', 'офтальмолог',\n",
       "        'педиатр', 'психотерапевт терапевт', 'пульмонолог',\n",
       "        'реабилитолог ревматолог', 'ревматолог',\n",
       "        'рефлексотерапевт терапевт', 'стоматолог', 'терапевт', 'уролог',\n",
       "        'флеболог хирург', 'хирург', 'эндокринолог'], dtype=object),\n",
       " array([   2,    1, 5557,  205,    2,    4, 2064,    1, 1967,   31, 1124,\n",
       "          62,  455, 6236,   35,   98,    6, 1210, 2360, 1052,   44,   34,\n",
       "          69,   10,  159,  120,   12, 3079, 1861,  248,  709, 1183]))"
      ]
     },
     "execution_count": 115,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.unique(test['Врач'], return_counts=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_submit.to_csv('submit/bow_simple_nn_doctors_3dense_diag_all_compl_uni-n_gram_doctor_topics_clinic_repeats.csv', header=True, index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    }
   ],
   "source": [
    "train_y, pop_diagnoses, most_pop_diagnose = preproc_target_train(train, reduce_classes=False)\n",
    "# valid_y, _, _ = preproc_target_test(valid, pop_diagnoses, most_pop_diagnose)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2302,)"
      ]
     },
     "execution_count": 121,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.unique(train_y).shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/200\n",
      "61976/61976 [==============================] - 10s 160us/step - loss: 5.8049 - acc: 0.0519\n",
      "Epoch 2/200\n",
      "45056/61976 [====================>.........] - ETA: 2s - loss: 5.5170 - acc: 0.0823"
     ]
    }
   ],
   "source": [
    "pipe.fit(train, train_y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "pred = pipe.transform(train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array(['A46', 'A63.0', 'B00', 'B00.1', 'B00.8', 'B02', 'B07', 'B08.1',\n",
       "        'B35.1', 'B35.3', 'B35.6', 'B36.0', 'B37', 'B37.3+', 'B97.1',\n",
       "        'B97.7', 'D17.0', 'D17.1', 'D17.2', 'D18.0', 'D23', 'D23.3',\n",
       "        'D23.5', 'D23.9', 'D24', 'D25', 'D50', 'D50.9', 'D89.8', 'E01.1',\n",
       "        'E01.8', 'E03.8', 'E03.9', 'E04.1', 'E04.2', 'E05.0', 'E06.3',\n",
       "        'E11', 'E22.1', 'E28', 'E28.1', 'E28.8', 'E28.9', 'E66.0', 'E89.0',\n",
       "        'F41.0', 'F41.2', 'F45.3', 'F95.0', 'G24', 'G24.9', 'G43.0',\n",
       "        'G44.2', 'G50.0', 'G51.0', 'G55*', 'G55.1*', 'G55.3*', 'G56.0',\n",
       "        'G56.2', 'G90', 'G90.8', 'G90.9', 'G93.4', 'H00', 'H00.0', 'H00.1',\n",
       "        'H01.0', 'H04.1', 'H10.0', 'H10.1', 'H10.2', 'H10.3', 'H10.5',\n",
       "        'H10.8', 'H11.3', 'H15.1', 'H16.1', 'H16.2', 'H20.0', 'H35.0',\n",
       "        'H35.3', 'H35.4', 'H40.1', 'H52.0', 'H52.1', 'H52.2', 'H52.4',\n",
       "        'H52.5', 'H60', 'H61.2', 'H65', 'H65.0', 'H65.1', 'H66.0', 'H66.1',\n",
       "        'H68.0', 'H68.1', 'H81.1', 'H81.2', 'H90.3', 'H93.1', 'I10', 'I11',\n",
       "        'I11.9', 'I20', 'I20.1', 'I20.8', 'I20.9', 'I25', 'I25.0', 'I25.1',\n",
       "        'I34.1', 'I47', 'I48', 'I49.8', 'I67.2', 'I67.4', 'I67.8', 'I70.2',\n",
       "        'I83', 'I83.9', 'I84.3', 'I84.4', 'I84.5', 'I86.1', 'I87.0',\n",
       "        'I87.2', 'J00', 'J01.0', 'J01.1', 'J01.8', 'J02', 'J02.9', 'J03',\n",
       "        'J03.9', 'J04.0', 'J04.1', 'J06.0', 'J06.9', 'J18', 'J18.0',\n",
       "        'J18.9', 'J20', 'J20.9', 'J30', 'J30.0', 'J30.1', 'J30.3', 'J31.0',\n",
       "        'J31.1', 'J31.2', 'J34.2', 'J35.0', 'J35.2', 'J35.3', 'J35.8',\n",
       "        'J41.0', 'J45.0', 'J45.8', 'J45.9', 'K02.1', 'K04.0', 'K05.1',\n",
       "        'K05.3', 'K21', 'K21.0', 'K21.9', 'K29.3', 'K29.5', 'K29.9', 'K30',\n",
       "        'K40', 'K40.9', 'K43', 'K46', 'K52.9', 'K58', 'K58.0', 'K80.1',\n",
       "        'K81', 'K82.8', 'K83.9', 'K86.1', 'K87*', 'L01.0', 'L02.0',\n",
       "        'L02.2', 'L02.4', 'L04.0', 'L08.0', 'L08.8', 'L20', 'L20.8', 'L21',\n",
       "        'L23', 'L23.9', 'L28.0', 'L30.0', 'L30.2', 'L30.3', 'L30.8',\n",
       "        'L30.9', 'L40.0', 'L52', 'L60.0', 'L60.3', 'L63', 'L65.0', 'L70.0',\n",
       "        'L71', 'L71.0', 'L72', 'L73.2', 'L84', 'M06.0', 'M12.5', 'M15-M19',\n",
       "        'M15.8', 'M17', 'M17.1', 'M17.4', 'M20-M25', 'M21.4', 'M22.4',\n",
       "        'M23.2', 'M41', 'M41.1', 'M42', 'M42.1', 'M47', 'M47.8', 'M50.1',\n",
       "        'M51', 'M51.1', 'M51.3', 'M51.8', 'M53.0', 'M53.1', 'M53.8',\n",
       "        'M53.9', 'M54', 'M54.1', 'M54.2', 'M54.4', 'M54.5', 'M54.6',\n",
       "        'M54.8', 'M54.9', 'M65', 'M67.4', 'M70.2', 'M71.8', 'M72.5',\n",
       "        'M75.3', 'M76.0', 'M77.1', 'M77.3', 'M77.9', 'M79.1', 'M79.2',\n",
       "        'M79.6', 'N10', 'N11', 'N20', 'N20.0', 'N20.2', 'N30', 'N30.0',\n",
       "        'N30.2', 'N34', 'N34.1', 'N34.2', 'N40', 'N41.0', 'N41.1', 'N42',\n",
       "        'N42.8', 'N46', 'N47', 'N48.1', 'N48.6', 'N48.8', 'N60.1', 'N64',\n",
       "        'N64.4', 'N70', 'N72', 'N75.0', 'N76.0', 'N76.1', 'N77.1*',\n",
       "        'N80.0', 'N83.0', 'N83.2', 'N84.0', 'N85.0', 'N86', 'N94.0',\n",
       "        'N94.9', 'N95.2', 'O04', 'R52.0', 'S06.0', 'S20.2', 'S22.3',\n",
       "        'S52.5', 'S61.0', 'S62.3', 'S80', 'S80.0', 'S83.5', 'S92.3',\n",
       "        'S92.5', 'S93.4', 'T15.1', 'Z00', 'Z00.0', 'Z00.8', 'Z01.0',\n",
       "        'Z01.4', 'Z01.8', 'Z03.3', 'Z04.8', 'Z10.0', 'Z32.0', 'Z32.1',\n",
       "        'Z34.0', 'Z34.8', 'Z35', 'Z35.8'], dtype='<U7'),\n",
       " array([  49,    1,  133,    2,   17,   76,  342,   31,   70,   54,   51,\n",
       "         112,   12,  288,    1,   22,   21,    8,   33,    9,   36,    2,\n",
       "          45,  765,   39,  300,   19,  104,   18,   27,    2,  311,   41,\n",
       "          51,  687,   10,   74,  242,    7,  197,    1,   19,    3,   48,\n",
       "          19,   16,    2,   61,    8,   48,    9,    8,  253,   70,    5,\n",
       "          41,   22,   47,    2,   42,  442,   76,  114,   31,   27,   59,\n",
       "          19,   12,    5,   22,   29,  202,   14,   37,    1,    5,   16,\n",
       "           2,    2,    1,  303,   14,   27,    1,   53,  800,   66,  123,\n",
       "           9,  252,  686,   10,   10,    6,    7,   17,    6,    7,    4,\n",
       "           7,   14,   27,  168, 2224,   96,    2,    7,   80,   71,   12,\n",
       "          16,   16,   85,    9,   18,   73,  222,    6,  233,    9,   14,\n",
       "         490,   45,   65,   79,   10,    5,    2, 1297,  839,    2,   13,\n",
       "          36,   15,   19,   19,   49,  379,   28, 3165,  222,    8,   85,\n",
       "           6,   58,    2,   79,  122,   87,  100,  147,   96,   61, 1013,\n",
       "          15,    3,  159,   34,   34,   20,    3,   59,   26,   29,    5,\n",
       "         139,  273,    1,  153,  681, 1137, 1159,   22,   10,   17,    5,\n",
       "         114,  126,   50,   20,    8,   63,   80,  120,  394,    1,  112,\n",
       "         298,  134,   10,  113,   19,  161,  156,  335,    9,    4,   20,\n",
       "          29,  334,   25,  489,   68,    6,    3,  178,   23,   31,   84,\n",
       "         108,   98,  250,   70,    1,   61,    8,   16,   59,  298,   23,\n",
       "           1,   10,    7,  259,    2,  270,  101,    7,  151, 5466,   56,\n",
       "           4,    6,  158,   97,  109,  248,  148,   52,  436,   24,  173,\n",
       "           7,  399,  438,  335,  192, 1228,   46, 1640,   27,   77,   19,\n",
       "           8,    6,    2,   23,   23,    2,  184,   33,  129,   17,   53,\n",
       "           5,  250,    4,   25,  379,  157,   14,   43,  109,  584,    5,\n",
       "        2223,    1,  105,   12,    6,  130,   10,    5, 1399,    4,    5,\n",
       "          53,  388,    3, 2488, 1913, 1849,   53,  347,   13,  304,   19,\n",
       "         295,    8,   13,   69,   14,    4,    9,   26,    4,   36,   53,\n",
       "          23,    4,   17,   46,   11,    6,   33,   43,    5, 1376,  124,\n",
       "          72,  409, 1157,    1,    9,   89,   10, 3089,  103,   11,   20,\n",
       "          14]))"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.unique(pred, return_counts=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Submit"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/root/anaconda3/lib/python3.6/site-packages/scipy/stats/stats.py:245: RuntimeWarning: The input array could not be properly checked for nan values. nan values will be ignored.\n",
      "  \"values. nan values will be ignored.\", RuntimeWarning)\n"
     ]
    }
   ],
   "source": [
    "train = utils.load_data('data/train_data_complaints_repeats_doctors.csv')\n",
    "train_topics = np.load('data/topics_train_ngramm.npy')\n",
    "train = join_topics(train, train_topics)\n",
    "train_y, pop_diagnoses, most_pop_diagnose = preproc_target_train(train, reduce_classes=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "61976/61976 [==============================] - 13s 215us/step - loss: 5.8733 - acc: 0.0998\n",
      "Epoch 2/100\n",
      "61976/61976 [==============================] - 13s 212us/step - loss: 5.5176 - acc: 0.1808\n",
      "Epoch 3/100\n",
      "61976/61976 [==============================] - 13s 211us/step - loss: 5.1080 - acc: 0.2237\n",
      "Epoch 4/100\n",
      "61976/61976 [==============================] - 13s 212us/step - loss: 4.7338 - acc: 0.2444\n",
      "Epoch 5/100\n",
      "61976/61976 [==============================] - 13s 211us/step - loss: 4.4498 - acc: 0.2577\n",
      "Epoch 6/100\n",
      "61976/61976 [==============================] - 13s 211us/step - loss: 4.2481 - acc: 0.2679\n",
      "Epoch 7/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 4.1090 - acc: 0.2773\n",
      "Epoch 8/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 4.0043 - acc: 0.2860\n",
      "Epoch 9/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.9136 - acc: 0.2952\n",
      "Epoch 10/100\n",
      "61976/61976 [==============================] - 13s 211us/step - loss: 3.8513 - acc: 0.3007\n",
      "Epoch 11/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.7937 - acc: 0.3045\n",
      "Epoch 12/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.7485 - acc: 0.3090\n",
      "Epoch 13/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.7130 - acc: 0.3135\n",
      "Epoch 14/100\n",
      "61976/61976 [==============================] - 13s 211us/step - loss: 3.6764 - acc: 0.3183\n",
      "Epoch 15/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.6508 - acc: 0.3216\n",
      "Epoch 16/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.6254 - acc: 0.3243\n",
      "Epoch 17/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.6010 - acc: 0.3273\n",
      "Epoch 18/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.5754 - acc: 0.3303\n",
      "Epoch 19/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.5600 - acc: 0.3320\n",
      "Epoch 20/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.5423 - acc: 0.3338\n",
      "Epoch 21/100\n",
      "61976/61976 [==============================] - 13s 210us/step - loss: 3.5206 - acc: 0.3362\n",
      "Epoch 22/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.5016 - acc: 0.3372\n",
      "Epoch 23/100\n",
      "61976/61976 [==============================] - 13s 210us/step - loss: 3.4885 - acc: 0.3391\n",
      "Epoch 24/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.4783 - acc: 0.3425\n",
      "Epoch 25/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.4684 - acc: 0.3435\n",
      "Epoch 26/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.4551 - acc: 0.3455\n",
      "Epoch 27/100\n",
      "61976/61976 [==============================] - 13s 211us/step - loss: 3.4406 - acc: 0.3463\n",
      "Epoch 28/100\n",
      "61976/61976 [==============================] - 13s 210us/step - loss: 3.4187 - acc: 0.3492\n",
      "Epoch 29/100\n",
      "61976/61976 [==============================] - 13s 210us/step - loss: 3.4105 - acc: 0.3511\n",
      "Epoch 30/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.4028 - acc: 0.3520\n",
      "Epoch 31/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.3930 - acc: 0.3535\n",
      "Epoch 32/100\n",
      "61976/61976 [==============================] - 13s 211us/step - loss: 3.3829 - acc: 0.3549\n",
      "Epoch 33/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.3741 - acc: 0.3567\n",
      "Epoch 34/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.3705 - acc: 0.3569\n",
      "Epoch 35/100\n",
      "61976/61976 [==============================] - 13s 212us/step - loss: 3.3674 - acc: 0.3575\n",
      "Epoch 36/100\n",
      "61976/61976 [==============================] - 13s 211us/step - loss: 3.3516 - acc: 0.3589\n",
      "Epoch 37/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.3422 - acc: 0.3609\n",
      "Epoch 38/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.3366 - acc: 0.3610\n",
      "Epoch 39/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.3263 - acc: 0.3618\n",
      "Epoch 40/100\n",
      "61976/61976 [==============================] - 13s 210us/step - loss: 3.3232 - acc: 0.3636\n",
      "Epoch 41/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.3110 - acc: 0.3647\n",
      "Epoch 42/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.3067 - acc: 0.3649\n",
      "Epoch 43/100\n",
      "61976/61976 [==============================] - 13s 207us/step - loss: 3.2982 - acc: 0.3659\n",
      "Epoch 44/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.2923 - acc: 0.3672\n",
      "Epoch 45/100\n",
      "61976/61976 [==============================] - 13s 207us/step - loss: 3.2898 - acc: 0.3678\n",
      "Epoch 46/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.2819 - acc: 0.3696\n",
      "Epoch 47/100\n",
      "61976/61976 [==============================] - 13s 207us/step - loss: 3.2814 - acc: 0.3711\n",
      "Epoch 48/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.2806 - acc: 0.3720\n",
      "Epoch 49/100\n",
      "61976/61976 [==============================] - 13s 211us/step - loss: 3.2746 - acc: 0.3714\n",
      "Epoch 50/100\n",
      "61976/61976 [==============================] - 13s 210us/step - loss: 3.2675 - acc: 0.3722\n",
      "Epoch 51/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.2674 - acc: 0.3726\n",
      "Epoch 52/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.2611 - acc: 0.3730\n",
      "Epoch 53/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.2531 - acc: 0.3746\n",
      "Epoch 54/100\n",
      "61976/61976 [==============================] - 13s 207us/step - loss: 3.2521 - acc: 0.3757\n",
      "Epoch 55/100\n",
      "61976/61976 [==============================] - 13s 210us/step - loss: 3.2457 - acc: 0.3763\n",
      "Epoch 56/100\n",
      "61976/61976 [==============================] - 13s 210us/step - loss: 3.2397 - acc: 0.3763\n",
      "Epoch 57/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.2336 - acc: 0.3774\n",
      "Epoch 58/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.2326 - acc: 0.3780\n",
      "Epoch 59/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.2253 - acc: 0.3797\n",
      "Epoch 60/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.2197 - acc: 0.3799\n",
      "Epoch 61/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.2116 - acc: 0.3807\n",
      "Epoch 62/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.2075 - acc: 0.3818\n",
      "Epoch 63/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.2062 - acc: 0.3832\n",
      "Epoch 64/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.2028 - acc: 0.3816\n",
      "Epoch 65/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.1995 - acc: 0.3822\n",
      "Epoch 66/100\n",
      "61976/61976 [==============================] - 13s 210us/step - loss: 3.1949 - acc: 0.3835\n",
      "Epoch 67/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.1904 - acc: 0.3836\n",
      "Epoch 68/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.1891 - acc: 0.3856\n",
      "Epoch 69/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.1808 - acc: 0.3866\n",
      "Epoch 70/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.1779 - acc: 0.3857\n",
      "Epoch 71/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.1716 - acc: 0.3875\n",
      "Epoch 72/100\n",
      "61976/61976 [==============================] - 13s 207us/step - loss: 3.1683 - acc: 0.3876\n",
      "Epoch 73/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.1660 - acc: 0.3878\n",
      "Epoch 74/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.1626 - acc: 0.3897\n",
      "Epoch 75/100\n",
      "61976/61976 [==============================] - 13s 211us/step - loss: 3.1570 - acc: 0.3903\n",
      "Epoch 76/100\n",
      "61976/61976 [==============================] - 13s 210us/step - loss: 3.1557 - acc: 0.3917\n",
      "Epoch 77/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.1534 - acc: 0.3918\n",
      "Epoch 78/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.1506 - acc: 0.3916\n",
      "Epoch 79/100\n",
      "61976/61976 [==============================] - 13s 207us/step - loss: 3.1456 - acc: 0.3923\n",
      "Epoch 80/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.1393 - acc: 0.3923\n",
      "Epoch 81/100\n",
      "61976/61976 [==============================] - 13s 210us/step - loss: 3.1382 - acc: 0.3930\n",
      "Epoch 82/100\n",
      "61976/61976 [==============================] - 13s 207us/step - loss: 3.1367 - acc: 0.3942\n",
      "Epoch 83/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.1333 - acc: 0.3947\n",
      "Epoch 84/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.1251 - acc: 0.3940\n",
      "Epoch 85/100\n",
      "61976/61976 [==============================] - 13s 207us/step - loss: 3.1240 - acc: 0.3963\n",
      "Epoch 86/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.1194 - acc: 0.3956\n",
      "Epoch 87/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.1168 - acc: 0.3966\n",
      "Epoch 88/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.1133 - acc: 0.3970\n",
      "Epoch 89/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.1103 - acc: 0.3963\n",
      "Epoch 90/100\n",
      "61976/61976 [==============================] - 13s 209us/step - loss: 3.1043 - acc: 0.3972\n",
      "Epoch 91/100\n",
      "61976/61976 [==============================] - 13s 207us/step - loss: 3.1059 - acc: 0.3976\n",
      "Epoch 92/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.1007 - acc: 0.3990\n",
      "Epoch 93/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.0974 - acc: 0.3981\n",
      "Epoch 94/100\n",
      "61976/61976 [==============================] - 13s 211us/step - loss: 3.0929 - acc: 0.3996\n",
      "Epoch 95/100\n",
      "61976/61976 [==============================] - 13s 207us/step - loss: 3.0878 - acc: 0.4004\n",
      "Epoch 96/100\n",
      "61976/61976 [==============================] - 13s 207us/step - loss: 3.0853 - acc: 0.4003\n",
      "Epoch 97/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.0840 - acc: 0.4004\n",
      "Epoch 98/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.0798 - acc: 0.4023\n",
      "Epoch 99/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.0758 - acc: 0.4015\n",
      "Epoch 100/100\n",
      "61976/61976 [==============================] - 13s 208us/step - loss: 3.0723 - acc: 0.4017\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Pipeline(memory=None,\n",
       "     steps=[('union', FeatureUnion(n_jobs=1,\n",
       "       transformer_list=[('complaints_pipe', Pipeline(memory=None,\n",
       "     steps=[('complaint_selector', ItemSelector(key='Жалобы (unigramm)')), ('tfidf', TfidfVectorizer(analyzer='word', binary=False, decode_error='strict',\n",
       "        dtype=<class 'numpy.int64'>, e...:48_2018/{epoch:02d}_{acc:.2f}.h5',\n",
       "    tb_log_dir='simple_models/log/nn14', validation_split=0.0))])"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pipe.fit(train, train_y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {},
   "outputs": [],
   "source": [
    "test = utils.load_data('data/test_data_complaints_repeats_doctors.csv')\n",
    "test_topics = np.load('data/topics_test_ngramm.npy')\n",
    "test = join_topics(test, test_topics)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_pred = pipe.transform(test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 115,
   "metadata": {},
   "outputs": [],
   "source": [
    "submit = pd.DataFrame({'Id_Записи': test['Id_Записи'], 'Код_диагноза': test_pred})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 116,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "M42.1      2346\n",
       "J06.9      1734\n",
       "N76.0      1504\n",
       "Z32.1      1201\n",
       "N77.1*     1022\n",
       "N41.1       950\n",
       "J00         778\n",
       "M51         735\n",
       "I11         722\n",
       "K29.9       584\n",
       "M65         575\n",
       "K30         523\n",
       "J35.0       520\n",
       "Z01.8       505\n",
       "Z00.0       498\n",
       "M54.8       480\n",
       "N60.1       427\n",
       "M53.8       403\n",
       "D23.9       398\n",
       "H52.1       378\n",
       "Z01.4       355\n",
       "M54.2       354\n",
       "J01.0       330\n",
       "K29.5       317\n",
       "L30.8       289\n",
       "D25         264\n",
       "I83.9       261\n",
       "E04.2       245\n",
       "G90         225\n",
       "J04.1       223\n",
       "           ... \n",
       "H81.2         2\n",
       "Z10           2\n",
       "J02.9         2\n",
       "M54.1         2\n",
       "Z00.1         2\n",
       "E22.1         2\n",
       "L91.0         2\n",
       "H00.1         1\n",
       "I87.2         1\n",
       "M43.9         1\n",
       "J35.3         1\n",
       "H40.1         1\n",
       "L08.8         1\n",
       "L01.0         1\n",
       "S91.3         1\n",
       "N94.9         1\n",
       "M17.0         1\n",
       "G43.0         1\n",
       "I49           1\n",
       "B00.1         1\n",
       "M65-M68       1\n",
       "K58.9         1\n",
       "H65.1         1\n",
       "L01           1\n",
       "E05.0         1\n",
       "N84.1         1\n",
       "I86.1         1\n",
       "L04.0         1\n",
       "G44.1         1\n",
       "K81           1\n",
       "Name: Код_диагноза, Length: 307, dtype: int64"
      ]
     },
     "execution_count": 116,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "submit['Код_диагноза'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 117,
   "metadata": {},
   "outputs": [],
   "source": [
    "submit.to_csv('submit/bow_simple_nn33_3dense_diag_all_compl_uni-n_gram_doctor_topics_clinic_repeats.csv', header=True, index=False)"
   ]
  }
 ],
 "metadata": {
  "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}