model_tests.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from tensorflow import keras\n",
    "from tensorflow.keras import layers\n",
    "from tensorflow.keras import Sequential\n",
    "from tensorflow.keras.layers import Dense\n",
    "from tensorflow.keras.wrappers.scikit_learn import KerasClassifier\n",
    "from tensorflow.keras.callbacks import EarlyStopping\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.metrics import confusion_matrix, precision_recall_fscore_support, accuracy_score\n",
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.model_selection import GridSearchCV\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.base import BaseEstimator, ClassifierMixin\n",
    "from itertools import combinations\n",
    "import numpy as np\n",
    "SEED = 25\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>av_depth</th>\n",
       "      <th>max_depth</th>\n",
       "      <th>min_depth</th>\n",
       "      <th>num_acl</th>\n",
       "      <th>num_rel_cl</th>\n",
       "      <th>num_advcl</th>\n",
       "      <th>num_sent</th>\n",
       "      <th>num_tok</th>\n",
       "      <th>av_tok_before_root</th>\n",
       "      <th>av_len_sent</th>\n",
       "      <th>...</th>\n",
       "      <th>num_linkings</th>\n",
       "      <th>num_4grams</th>\n",
       "      <th>num_func_ngrams</th>\n",
       "      <th>num_shell_noun</th>\n",
       "      <th>num_misspelled_tokens</th>\n",
       "      <th>million_mistake</th>\n",
       "      <th>text</th>\n",
       "      <th>class</th>\n",
       "      <th>type</th>\n",
       "      <th>sum_punct</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3.56</td>\n",
       "      <td>6</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>18</td>\n",
       "      <td>244</td>\n",
       "      <td>4.39</td>\n",
       "      <td>13.56</td>\n",
       "      <td>...</td>\n",
       "      <td>5</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>This two line graphs illustrates monthly avera...</td>\n",
       "      <td>6</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>5.43</td>\n",
       "      <td>10</td>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>175</td>\n",
       "      <td>6.29</td>\n",
       "      <td>25.00</td>\n",
       "      <td>...</td>\n",
       "      <td>6</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>﻿We have two graphics, which show us the popul...</td>\n",
       "      <td>6</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2 rows × 62 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   av_depth  max_depth  min_depth  num_acl  num_rel_cl  num_advcl  num_sent  \\\n",
       "0      3.56          6          2        2           0          1        18   \n",
       "1      5.43         10          4        1           3          1         7   \n",
       "\n",
       "   num_tok  av_tok_before_root  av_len_sent  ...  num_linkings  num_4grams  \\\n",
       "0      244                4.39        13.56  ...             5           1   \n",
       "1      175                6.29        25.00  ...             6           1   \n",
       "\n",
       "   num_func_ngrams  num_shell_noun  num_misspelled_tokens  million_mistake  \\\n",
       "0                0               0                      2                0   \n",
       "1                0               0                      3                0   \n",
       "\n",
       "                                                text  class  type  sum_punct  \n",
       "0  This two line graphs illustrates monthly avera...      6     1          0  \n",
       "1  ﻿We have two graphics, which show us the popul...      6     1          0  \n",
       "\n",
       "[2 rows x 62 columns]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.read_csv(r'D:\\hse_iot\\VKR\\my_inspector\\data\\classification_dataset.csv')\n",
    "df['class'] = round(df['class'] / 10).astype(int)\n",
    "df = df[df['class'] != 1]\n",
    "_type = df['type'].to_list()\n",
    "df = df.drop(['name', 'side_mistake',\n",
    "              'vs', 'squared_vs', 'ttr', 'root_ttr',\n",
    "              'log_ttr', 'uber_ttr', 'd', 'vvi',\n",
    "              'squared_vv', 'punct_mistakes_pp',\n",
    "              'punct_mistakes_because',\n",
    "              'punct_mistakes_but', 'punct_mistakes_compare'], axis=1)\n",
    "\n",
    "df.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(3440, 62)"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 3440 entries, 0 to 3440\n",
      "Data columns (total 62 columns):\n",
      " #   Column                 Non-Null Count  Dtype  \n",
      "---  ------                 --------------  -----  \n",
      " 0   av_depth               3440 non-null   float64\n",
      " 1   max_depth              3440 non-null   int64  \n",
      " 2   min_depth              3440 non-null   int64  \n",
      " 3   num_acl                3440 non-null   int64  \n",
      " 4   num_rel_cl             3440 non-null   int64  \n",
      " 5   num_advcl              3440 non-null   int64  \n",
      " 6   num_sent               3440 non-null   int64  \n",
      " 7   num_tok                3440 non-null   int64  \n",
      " 8   av_tok_before_root     3440 non-null   float64\n",
      " 9   av_len_sent            3440 non-null   float64\n",
      " 10  num_cl                 3440 non-null   int64  \n",
      " 11  num_tu                 3440 non-null   int64  \n",
      " 12  num_compl_tu           3440 non-null   int64  \n",
      " 13  num_coord              3440 non-null   int64  \n",
      " 14  num_poss               3440 non-null   int64  \n",
      " 15  num_prep               3440 non-null   int64  \n",
      " 16  num_adj_noun           3440 non-null   int64  \n",
      " 17  num_part_noun          3440 non-null   int64  \n",
      " 18  num_noun_inf           3440 non-null   int64  \n",
      " 19  pos_sim_nei            3440 non-null   float64\n",
      " 20  pos_sim_all            3440 non-null   float64\n",
      " 21  lemma_sim_all          3440 non-null   float64\n",
      " 22  lemma_sim_nei          3440 non-null   float64\n",
      " 23  density                3440 non-null   float64\n",
      " 24  ls                     3440 non-null   float64\n",
      " 25  corrected_vs           3440 non-null   float64\n",
      " 26  lfp_1000               3440 non-null   float64\n",
      " 27  lfp_2000               3440 non-null   float64\n",
      " 28  lfp_uwl                3440 non-null   float64\n",
      " 29  lfp_rest               3440 non-null   float64\n",
      " 30  ndw                    3440 non-null   int64  \n",
      " 31  corrected_ttr          3440 non-null   float64\n",
      " 32  lv                     3440 non-null   float64\n",
      " 33  corrected_vv           3440 non-null   float64\n",
      " 34  vvii                   3440 non-null   float64\n",
      " 35  nv                     3440 non-null   float64\n",
      " 36  adjv                   3440 non-null   float64\n",
      " 37  advv                   3440 non-null   float64\n",
      " 38  modv                   3440 non-null   float64\n",
      " 39  der_level3             3440 non-null   float64\n",
      " 40  der_level4             3440 non-null   float64\n",
      " 41  der_level5             3440 non-null   float64\n",
      " 42  der_level6             3440 non-null   float64\n",
      " 43  mci                    3440 non-null   float64\n",
      " 44  freq_finite_forms      3440 non-null   float64\n",
      " 45  freq_aux               3440 non-null   float64\n",
      " 46  num_inf                3440 non-null   int64  \n",
      " 47  num_gerunds            3440 non-null   int64  \n",
      " 48  num_pres_sing          3440 non-null   int64  \n",
      " 49  num_pres_plur          3440 non-null   int64  \n",
      " 50  num_past_part          3440 non-null   int64  \n",
      " 51  num_past_simple        3440 non-null   int64  \n",
      " 52  num_linkings           3440 non-null   int64  \n",
      " 53  num_4grams             3440 non-null   int64  \n",
      " 54  num_func_ngrams        3440 non-null   int64  \n",
      " 55  num_shell_noun         3440 non-null   int64  \n",
      " 56  num_misspelled_tokens  3440 non-null   int64  \n",
      " 57  million_mistake        3440 non-null   int64  \n",
      " 58  text                   3440 non-null   object \n",
      " 59  class                  3440 non-null   int32  \n",
      " 60  type                   3440 non-null   int64  \n",
      " 61  sum_punct              3440 non-null   int64  \n",
      "dtypes: float64(29), int32(1), int64(31), object(1)\n",
      "memory usage: 1.6+ MB\n"
     ]
    }
   ],
   "source": [
    "df.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "6    2680\n",
       "7     308\n",
       "5     215\n",
       "4     122\n",
       "8      77\n",
       "2      24\n",
       "3      14\n",
       "Name: class, dtype: int64"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['class'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x720 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.set(rc={'figure.figsize':(20,10)})\n",
    "sns_plt = sns.countplot(x='class', data=df, palette=\"Set2\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "from umap import UMAP"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "feature_columns = [col for col in df.columns if col not in ['class', 'name', 'text', 'type']]\n",
    "X = df[feature_columns]\n",
    "y = df['class'] - 2\n",
    "y.loc[y == 1] = 3\n",
    "y.loc[y == 0] = 3\n",
    "y.loc[y == 2] = 3\n",
    "y.loc[y == 5] = 6\n",
    "\n",
    "y.loc[y == 3] = 0\n",
    "y.loc[y == 4] = 1\n",
    "y.loc[y == 6] = 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x720 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "umap = UMAP()\n",
    "umaped_vct=umap.fit_transform(X)\n",
    "fig, ax = plt.subplots()\n",
    "ax.scatter(umaped_vct[:, 0], umaped_vct[:, 1], edgecolor='b', s=3)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_x, test_x, train_y, test_y = train_test_split(X, y, test_size=0.2, random_state = SEED)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Так как количество элементов в классах не равномерно то объединим классы 2,3,4,5; 7, 8; и уберем часть элементов из 6 класса"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1    2139\n",
       "2     322\n",
       "0     291\n",
       "Name: class, dtype: int64"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\n",
    "train_y.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2    322\n",
       "1    315\n",
       "0    291\n",
       "Name: class, dtype: int64"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dummy = train_y.loc[train_y == 1][:315]\n",
    "train_y.drop(train_y.loc[train_y == 1].index, inplace=True)\n",
    "train_y = pd.concat([train_y, dummy], ignore_index=True)\n",
    "train_y.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_idx = train_y.index\n",
    "train_x = train_x.iloc[y_idx.to_list()]\n",
    "train_x.reset_index(inplace=True, drop=True)\n",
    "test_x.reset_index(inplace=True, drop=True)\n",
    "train_y = train_y.reset_index(drop=True)\n",
    "test_y = test_y.reset_index(drop=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_y2 = []\n",
    "for y in train_y:\n",
    "    y2 = np.zeros(3)\n",
    "    y2[y] = 1\n",
    "    train_y2.append(y2)\n",
    "    \n",
    "test_y2 = []\n",
    "for y in test_y:\n",
    "    y2 = np.zeros(3)\n",
    "    y2[y] = 1\n",
    "    test_y2.append(y2)\n",
    "    \n",
    "train_y2 = np.array(train_y2)\n",
    "test_y2 = np.array(test_y2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "def create_model(neurons1=None, neurons2=None, neurons3=None, activation=\"relu\", optimizer='Adam', loss=keras.losses.MeanSquaredError()):\n",
    "    model = keras.Sequential()\n",
    "    if neurons1:\n",
    "        model.add(Dense(neurons1, activation=activation, kernel_regularizer=keras.regularizers.l2(0.001)))\n",
    "    if neurons2:\n",
    "        model.add(Dense(neurons2, activation=activation, kernel_regularizer=keras.regularizers.l2(0.001)))\n",
    "    if neurons3:\n",
    "        model.add(Dense(neurons3, activation=activation, kernel_regularizer=keras.regularizers.l2(0.001)))\n",
    "    model.add(Dense(3, activation=\"softmax\"))\n",
    "    model.compile(optimizer=optimizer, \n",
    "                  loss=loss,\n",
    "                  metrics=[keras.metrics.categorical_accuracy]\n",
    "         )\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = KerasClassifier(build_fn=create_model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import make_scorer, accuracy_score\n",
    "def custom_scorer(y_true, y_pred):\n",
    "    y_true = [np.argmax(y) for y in y_true]\n",
    "    y_true = np.array(y_true)\n",
    "    return accuracy_score(y_true, y_pred)\n",
    "scorer = make_scorer(custom_scorer, greater_is_better=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'neurons1': [16, 32, 64, 128, 256],\n",
       " 'neurons2': [16, 32, 64, 128, 256],\n",
       " 'neurons3': [16, 32, 64, 128, 256],\n",
       " 'activation': ['relu', 'tanh'],\n",
       " 'optimizer': ['Adam', 'SGD'],\n",
       " 'loss': ['mse'],\n",
       " 'epochs': [20, 50, 100],\n",
       " 'batch_size': [5, 10, 15, 20]}"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "neurons1 = [2**i for i in range(4, 9)]\n",
    "neurons2 = [2**i for i in range(4, 9)]\n",
    "neurons3 = [2**i for i in range(4, 9)]\n",
    "activation = [\"relu\", \"tanh\"]\n",
    "optimizer = ['Adam', 'SGD']\n",
    "loss = ['mse']\n",
    "epochs = [20, 50, 100]\n",
    "batch_size= [5, 10, 15, 20]\n",
    "\n",
    "param_grid = dict(neurons1=neurons1, \n",
    "                  neurons2=neurons2, \n",
    "                  neurons3=neurons3, \n",
    "                  activation=activation, \n",
    "                  optimizer=optimizer,\n",
    "                  loss=loss,\n",
    "                  epochs=epochs,\n",
    "                  batch_size=batch_size)\n",
    "param_grid"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 6000 candidates, totalling 18000 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 12 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  26 tasks      | elapsed:   49.8s\n",
      "[Parallel(n_jobs=-1)]: Done 176 tasks      | elapsed:  2.9min\n",
      "[Parallel(n_jobs=-1)]: Done 426 tasks      | elapsed:  6.6min\n",
      "[Parallel(n_jobs=-1)]: Done 776 tasks      | elapsed: 13.8min\n",
      "[Parallel(n_jobs=-1)]: Done 1226 tasks      | elapsed: 28.3min\n",
      "[Parallel(n_jobs=-1)]: Done 1776 tasks      | elapsed: 54.7min\n",
      "[Parallel(n_jobs=-1)]: Done 2426 tasks      | elapsed: 89.1min\n",
      "[Parallel(n_jobs=-1)]: Done 3176 tasks      | elapsed: 97.7min\n",
      "[Parallel(n_jobs=-1)]: Done 4026 tasks      | elapsed: 117.0min\n",
      "[Parallel(n_jobs=-1)]: Done 4976 tasks      | elapsed: 138.6min\n",
      "[Parallel(n_jobs=-1)]: Done 6026 tasks      | elapsed: 152.8min\n",
      "[Parallel(n_jobs=-1)]: Done 7176 tasks      | elapsed: 174.1min\n",
      "[Parallel(n_jobs=-1)]: Done 8426 tasks      | elapsed: 189.0min\n",
      "[Parallel(n_jobs=-1)]: Done 9776 tasks      | elapsed: 214.3min\n",
      "[Parallel(n_jobs=-1)]: Done 11226 tasks      | elapsed: 284.6min\n",
      "[Parallel(n_jobs=-1)]: Done 12776 tasks      | elapsed: 315.8min\n",
      "[Parallel(n_jobs=-1)]: Done 14426 tasks      | elapsed: 358.3min\n",
      "[Parallel(n_jobs=-1)]: Done 16176 tasks      | elapsed: 399.9min\n",
      "[Parallel(n_jobs=-1)]: Done 18000 out of 18000 | elapsed: 512.6min finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/50\n",
      "WARNING:tensorflow:Layer dense is casting an input tensor from dtype float64 to the layer's dtype of float32, which is new behavior in TensorFlow 2.  The layer has dtype float32 because its dtype defaults to floatx.\n",
      "\n",
      "If you intended to run this layer in float32, you can safely ignore this warning. If in doubt, this warning is likely only an issue if you are porting a TensorFlow 1.X model to TensorFlow 2.\n",
      "\n",
      "To change all layers to have dtype float64 by default, call `tf.keras.backend.set_floatx('float64')`. To change just this layer, pass dtype='float64' to the layer constructor. If you are the author of this layer, you can disable autocasting by passing autocast=False to the base Layer constructor.\n",
      "\n",
      "93/93 [==============================] - 0s 2ms/step - loss: 0.5450 - categorical_accuracy: 0.3362\n",
      "Epoch 2/50\n",
      "93/93 [==============================] - 0s 718us/step - loss: 0.5444 - categorical_accuracy: 0.3394\n",
      "Epoch 3/50\n",
      "93/93 [==============================] - 0s 463us/step - loss: 0.5441 - categorical_accuracy: 0.3394\n",
      "Epoch 4/50\n",
      "93/93 [==============================] - 0s 358us/step - loss: 0.5437 - categorical_accuracy: 0.3394\n",
      "Epoch 5/50\n",
      "93/93 [==============================] - 0s 548us/step - loss: 0.5433 - categorical_accuracy: 0.3394\n",
      "Epoch 6/50\n",
      "93/93 [==============================] - 0s 336us/step - loss: 0.5429 - categorical_accuracy: 0.3394\n",
      "Epoch 7/50\n",
      "93/93 [==============================] - 0s 336us/step - loss: 0.5425 - categorical_accuracy: 0.3394\n",
      "Epoch 8/50\n",
      "93/93 [==============================] - 0s 720us/step - loss: 0.5421 - categorical_accuracy: 0.3394\n",
      "Epoch 9/50\n",
      "93/93 [==============================] - 0s 491us/step - loss: 0.5418 - categorical_accuracy: 0.3394\n",
      "Epoch 10/50\n",
      "93/93 [==============================] - 0s 466us/step - loss: 0.5414 - categorical_accuracy: 0.3394\n",
      "Epoch 11/50\n",
      "93/93 [==============================] - 0s 442us/step - loss: 0.5410 - categorical_accuracy: 0.3394\n",
      "Epoch 12/50\n",
      "93/93 [==============================] - 0s 442us/step - loss: 0.5406 - categorical_accuracy: 0.3394\n",
      "Epoch 13/50\n",
      "93/93 [==============================] - 0s 447us/step - loss: 0.5403 - categorical_accuracy: 0.3394\n",
      "Epoch 14/50\n",
      "93/93 [==============================] - 0s 444us/step - loss: 0.5399 - categorical_accuracy: 0.3394\n",
      "Epoch 15/50\n",
      "93/93 [==============================] - 0s 474us/step - loss: 0.5395 - categorical_accuracy: 0.3394\n",
      "Epoch 16/50\n",
      "93/93 [==============================] - 0s 447us/step - loss: 0.5391 - categorical_accuracy: 0.3394\n",
      "Epoch 17/50\n",
      "93/93 [==============================] - 0s 454us/step - loss: 0.5388 - categorical_accuracy: 0.3394\n",
      "Epoch 18/50\n",
      "93/93 [==============================] - 0s 442us/step - loss: 0.5384 - categorical_accuracy: 0.3394\n",
      "Epoch 19/50\n",
      "93/93 [==============================] - 0s 454us/step - loss: 0.5380 - categorical_accuracy: 0.3394\n",
      "Epoch 20/50\n",
      "93/93 [==============================] - 0s 315us/step - loss: 0.5377 - categorical_accuracy: 0.3394\n",
      "Epoch 21/50\n",
      "93/93 [==============================] - 0s 498us/step - loss: 0.5373 - categorical_accuracy: 0.3394\n",
      "Epoch 22/50\n",
      "93/93 [==============================] - 0s 501us/step - loss: 0.5370 - categorical_accuracy: 0.3394\n",
      "Epoch 23/50\n",
      "93/93 [==============================] - 0s 443us/step - loss: 0.5366 - categorical_accuracy: 0.3394\n",
      "Epoch 24/50\n",
      "93/93 [==============================] - 0s 508us/step - loss: 0.5362 - categorical_accuracy: 0.3394\n",
      "Epoch 25/50\n",
      "93/93 [==============================] - 0s 522us/step - loss: 0.5359 - categorical_accuracy: 0.3394\n",
      "Epoch 26/50\n",
      "93/93 [==============================] - 0s 615us/step - loss: 0.5355 - categorical_accuracy: 0.3394\n",
      "Epoch 27/50\n",
      "93/93 [==============================] - 0s 505us/step - loss: 0.5352 - categorical_accuracy: 0.3394\n",
      "Epoch 28/50\n",
      "93/93 [==============================] - 0s 351us/step - loss: 0.5348 - categorical_accuracy: 0.3394\n",
      "Epoch 29/50\n",
      "93/93 [==============================] - 0s 524us/step - loss: 0.5344 - categorical_accuracy: 0.3394\n",
      "Epoch 30/50\n",
      "93/93 [==============================] - 0s 388us/step - loss: 0.5341 - categorical_accuracy: 0.3394\n",
      "Epoch 31/50\n",
      "93/93 [==============================] - 0s 506us/step - loss: 0.5337 - categorical_accuracy: 0.3394\n",
      "Epoch 32/50\n",
      "93/93 [==============================] - 0s 676us/step - loss: 0.5334 - categorical_accuracy: 0.3394\n",
      "Epoch 33/50\n",
      "93/93 [==============================] - 0s 650us/step - loss: 0.5330 - categorical_accuracy: 0.3394\n",
      "Epoch 34/50\n",
      "93/93 [==============================] - 0s 453us/step - loss: 0.5327 - categorical_accuracy: 0.3394\n",
      "Epoch 35/50\n",
      "93/93 [==============================] - 0s 681us/step - loss: 0.5323 - categorical_accuracy: 0.3394\n",
      "Epoch 36/50\n",
      "93/93 [==============================] - 0s 653us/step - loss: 0.5319 - categorical_accuracy: 0.3394\n",
      "Epoch 37/50\n",
      "93/93 [==============================] - 0s 433us/step - loss: 0.5316 - categorical_accuracy: 0.3394\n",
      "Epoch 38/50\n",
      "93/93 [==============================] - 0s 533us/step - loss: 0.5312 - categorical_accuracy: 0.3394\n",
      "Epoch 39/50\n",
      "93/93 [==============================] - 0s 668us/step - loss: 0.5307 - categorical_accuracy: 0.3394\n",
      "Epoch 40/50\n",
      "93/93 [==============================] - 0s 481us/step - loss: 0.5204 - categorical_accuracy: 0.3524\n",
      "Epoch 41/50\n",
      "93/93 [==============================] - 0s 448us/step - loss: 0.5234 - categorical_accuracy: 0.3448\n",
      "Epoch 42/50\n",
      "93/93 [==============================] - 0s 442us/step - loss: 0.5248 - categorical_accuracy: 0.3470\n",
      "Epoch 43/50\n",
      "93/93 [==============================] - 0s 431us/step - loss: 0.5245 - categorical_accuracy: 0.3470\n",
      "Epoch 44/50\n",
      "93/93 [==============================] - 0s 461us/step - loss: 0.5241 - categorical_accuracy: 0.3470\n",
      "Epoch 45/50\n",
      "93/93 [==============================] - 0s 440us/step - loss: 0.5237 - categorical_accuracy: 0.3470\n",
      "Epoch 46/50\n",
      "93/93 [==============================] - 0s 457us/step - loss: 0.5231 - categorical_accuracy: 0.3470\n",
      "Epoch 47/50\n",
      "93/93 [==============================] - 0s 464us/step - loss: 0.5214 - categorical_accuracy: 0.3470\n",
      "Epoch 48/50\n",
      "93/93 [==============================] - 0s 440us/step - loss: 0.5133 - categorical_accuracy: 0.3502\n",
      "Epoch 49/50\n",
      "93/93 [==============================] - 0s 366us/step - loss: 0.4207 - categorical_accuracy: 0.3211\n",
      "Epoch 50/50\n",
      "93/93 [==============================] - 0s 498us/step - loss: 0.3451 - categorical_accuracy: 0.3427\n",
      "Wall time: 8h 33min 21s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "grid = GridSearchCV(estimator=model, param_grid=param_grid, scoring=scorer, n_jobs=-1, verbose=1, cv=3)\n",
    "grid_result = grid.fit(train_x, train_y2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best: 0.618708 using {'activation': 'relu', 'batch_size': 10, 'epochs': 50, 'loss': 'mse', 'neurons1': 32, 'neurons2': 32, 'neurons3': 32, 'optimizer': 'SGD'}\n"
     ]
    }
   ],
   "source": [
    "print(\"Best: %f using %s\" % (grid_result.best_score_, grid_result.best_params_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From C:\\Users\\Konstantin\\.conda\\envs\\ml_hse\\lib\\site-packages\\tensorflow\\python\\keras\\wrappers\\scikit_learn.py:241: Sequential.predict_classes (from tensorflow.python.keras.engine.sequential) is deprecated and will be removed after 2021-01-01.\n",
      "Instructions for updating:\n",
      "Please use instead:* `np.argmax(model.predict(x), axis=-1)`,   if your model does multi-class classification   (e.g. if it uses a `softmax` last-layer activation).* `(model.predict(x) > 0.5).astype(\"int32\")`,   if your model does binary classification   (e.g. if it uses a `sigmoid` last-layer activation).\n"
     ]
    }
   ],
   "source": [
    "pred_y = grid_result.best_estimator_.predict(test_x)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 20,  36,  28],\n",
       "       [ 54, 283, 204],\n",
       "       [  7,  35,  21]], dtype=int64)"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "confusion_matrix(test_y, pred_y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.47093023255813954"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracy_score(test_y, pred_y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Теперь сверточные сети."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_x2 = np.array(train_x)\n",
    "test_x2 = np.array(test_x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_x3 = train_x2.reshape(-1, 59, 1)\n",
    "test_x3 = test_x2.reshape(-1, 59, 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "def create_model_conv(neurons1=None, neurons2=None, neurons3=None, core=3, activation=\"relu\", optimizer='Adam', loss=keras.losses.MeanSquaredError()):\n",
    "    model = keras.Sequential()\n",
    "    if neurons1:\n",
    "        model.add(layers.Conv1D(neurons1, core, activation=activation))\n",
    "    if neurons2:\n",
    "        model.add(layers.Conv1D(neurons2, core, activation=activation))\n",
    "    if neurons3:\n",
    "        model.add(layers.Conv1D(neurons3, core, activation=activation))\n",
    "    model.add(layers.Flatten())\n",
    "    model.add(Dense(3, activation=\"softmax\"))\n",
    "    model.compile(optimizer=optimizer, \n",
    "                  loss=loss,\n",
    "                  metrics=[keras.metrics.categorical_accuracy]\n",
    "         )\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'neurons1': [8, 16, 32, 64],\n",
       " 'neurons2': [8, 16, 32, 64],\n",
       " 'neurons3': [8, 16, 32, 64],\n",
       " 'core': [3, 4, 5, 10],\n",
       " 'activation': ['relu', 'tanh'],\n",
       " 'optimizer': ['Adam', 'SGD'],\n",
       " 'loss': ['mse'],\n",
       " 'epochs': [100],\n",
       " 'batch_size': [5, 10, 15, 20]}"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "neurons1 = [2**i for i in range(3, 7)]\n",
    "neurons2 = [2**i for i in range(3, 7)]\n",
    "neurons3 = [2**i for i in range(3, 7)]\n",
    "core = [3, 4, 5, 10]\n",
    "activation = [\"relu\", \"tanh\"]\n",
    "optimizer = ['Adam', 'SGD']\n",
    "loss = ['mse']\n",
    "epochs = [100,]\n",
    "batch_size= [5, 10, 15, 20]\n",
    "\n",
    "param_grid_conv = dict(neurons1=neurons1, \n",
    "                  neurons2=neurons2, \n",
    "                  neurons3=neurons3,\n",
    "                  core=core,\n",
    "                  activation=activation, \n",
    "                  optimizer=optimizer,\n",
    "                  loss=loss,\n",
    "                  epochs=epochs,\n",
    "                  batch_size=batch_size)\n",
    "param_grid_conv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_conv = KerasClassifier(build_fn=create_model_conv)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 3 folds for each of 4096 candidates, totalling 12288 fits\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[Parallel(n_jobs=-1)]: Using backend LokyBackend with 12 concurrent workers.\n",
      "[Parallel(n_jobs=-1)]: Done  26 tasks      | elapsed:  4.1min\n",
      "[Parallel(n_jobs=-1)]: Done 176 tasks      | elapsed: 21.6min\n",
      "[Parallel(n_jobs=-1)]: Done 426 tasks      | elapsed: 56.8min\n",
      "[Parallel(n_jobs=-1)]: Done 776 tasks      | elapsed: 109.1min\n",
      "[Parallel(n_jobs=-1)]: Done 1226 tasks      | elapsed: 179.5min\n",
      "[Parallel(n_jobs=-1)]: Done 1776 tasks      | elapsed: 258.6min\n",
      "[Parallel(n_jobs=-1)]: Done 2426 tasks      | elapsed: 326.7min\n",
      "[Parallel(n_jobs=-1)]: Done 3176 tasks      | elapsed: 423.1min\n",
      "[Parallel(n_jobs=-1)]: Done 4026 tasks      | elapsed: 498.4min\n",
      "[Parallel(n_jobs=-1)]: Done 4976 tasks      | elapsed: 603.4min\n",
      "[Parallel(n_jobs=-1)]: Done 6026 tasks      | elapsed: 706.3min\n",
      "[Parallel(n_jobs=-1)]: Done 7176 tasks      | elapsed: 874.3min\n",
      "[Parallel(n_jobs=-1)]: Done 8426 tasks      | elapsed: 1044.4min\n",
      "[Parallel(n_jobs=-1)]: Done 9776 tasks      | elapsed: 1218.0min\n",
      "[Parallel(n_jobs=-1)]: Done 11226 tasks      | elapsed: 1412.7min\n",
      "[Parallel(n_jobs=-1)]: Done 12288 out of 12288 | elapsed: 1643.0min finished\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "186/186 [==============================] - 7s 37ms/step - loss: 0.4397 - categorical_accuracy: 0.3384\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 6ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4552 - categorical_accuracy: 0.3168\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 6ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 4ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 4ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 7/100\n",
      "186/186 [==============================] - 1s 6ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 8/100\n",
      "186/186 [==============================] - 1s 7ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 9/100\n",
      "186/186 [==============================] - 1s 6ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 10/100\n",
      "186/186 [==============================] - 1s 6ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 11/100\n",
      "186/186 [==============================] - 1s 4ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 12/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 13/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 14/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 15/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 16/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 17/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 18/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 19/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 20/100\n",
      "186/186 [==============================] - 1s 6ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 21/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 22/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 23/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 24/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 25/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 26/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 27/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 28/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 29/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 30/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 31/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 32/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 33/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 34/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 35/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 36/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 37/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 38/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 39/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 40/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 41/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 42/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 43/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 44/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 45/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 46/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 47/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 48/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 49/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 50/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 51/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 52/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 53/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 54/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 55/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 56/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 57/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 58/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 59/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 60/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 61/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 62/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 63/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 64/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 65/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 66/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 67/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 68/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 69/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 70/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 71/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 72/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 73/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 74/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 75/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 76/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 77/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 78/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 79/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 80/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 81/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 82/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 83/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 84/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 85/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 86/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 87/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 88/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 89/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 90/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 91/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 92/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 93/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 94/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 95/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 96/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 97/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 98/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 99/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 100/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Wall time: 1d 3h 32min 3s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "early_stopping = EarlyStopping(monitor='categorical_accuracy', patience=5)\n",
    "grid_conv = GridSearchCV(estimator=model_conv, param_grid=param_grid_conv, scoring=scorer, verbose=1, cv=3, n_jobs=-1)\n",
    "grid_result_conv = grid_conv.fit(train_x3, train_y2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best: 0.708153 using {'activation': 'relu', 'batch_size': 5, 'core': 10, 'epochs': 100, 'loss': 'mse', 'neurons1': 64, 'neurons2': 64, 'neurons3': 16, 'optimizer': 'Adam'}\n"
     ]
    }
   ],
   "source": [
    "print(\"Best: %f using %s\" % (grid_result_conv.best_score_, grid_result_conv.best_params_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [],
   "source": [
    "pred_y = grid_result_conv.best_estimator_.predict(test_x3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 84,   0,   0],\n",
       "       [541,   0,   0],\n",
       "       [ 63,   0,   0]], dtype=int64)"
      ]
     },
     "execution_count": 59,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred_y = [np.argmax(y) for y in pred_y]\n",
    "pred_y = np.array(pred_y)\n",
    "confusion_matrix(test_y, pred_y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.12209302325581395"
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracy_score(test_y, pred_y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4253 - categorical_accuracy: 0.3588\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 7/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 8/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 9/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 10/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 11/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 12/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 13/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 14/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 15/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 16/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 17/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 18/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 19/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 20/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 21/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 22/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 23/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 24/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 25/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 26/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 27/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 28/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 29/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 30/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 31/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 32/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 33/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 34/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 35/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 36/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 37/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 38/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 39/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 40/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 41/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 42/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 43/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 44/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 45/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 46/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 47/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 48/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 49/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4339 - categorical_accuracy: 0.3491\n",
      "Epoch 50/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 51/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 52/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 53/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 54/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 55/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 56/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 57/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 58/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 59/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 60/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 61/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 62/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 63/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 64/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 65/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 66/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 67/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 68/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 69/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 70/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 71/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 72/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 73/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 74/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 75/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 76/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 77/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 78/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 79/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 80/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 81/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 82/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 83/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 84/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 85/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 86/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 87/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 88/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 89/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 90/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 91/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 92/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 93/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 94/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 95/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 96/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 97/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 98/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 99/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 100/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[ 84,   0,   0],\n",
       "       [541,   0,   0],\n",
       "       [ 63,   0,   0]], dtype=int64)"
      ]
     },
     "execution_count": 65,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = create_model_conv(neurons1=64, neurons2=64, neurons3=16, core=10, activation=\"relu\", optimizer='Adam', loss='mse')\n",
    "hist = model.fit(train_x3, train_y2, epochs=100, batch_size=5)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 93,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "--------------------0-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4368 - categorical_accuracy: 0.3427\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4302 - categorical_accuracy: 0.3545\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------1-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4432 - categorical_accuracy: 0.3341\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------2-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4394 - categorical_accuracy: 0.3405\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4389 - categorical_accuracy: 0.3416\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 7/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------3-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4442 - categorical_accuracy: 0.3330\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4406 - categorical_accuracy: 0.3384\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------4-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4345 - categorical_accuracy: 0.3459\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------5-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4429 - categorical_accuracy: 0.3341\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------6-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4343 - categorical_accuracy: 0.3459\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------7-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4356 - categorical_accuracy: 0.3438\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4597 - categorical_accuracy: 0.3103\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------8-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 1s 4ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4339 - categorical_accuracy: 0.3491\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------9-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4355 - categorical_accuracy: 0.3459\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------10-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4581 - categorical_accuracy: 0.3125\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------11-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4394 - categorical_accuracy: 0.3405\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4358 - categorical_accuracy: 0.3459\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 7/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 8/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------12-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4323 - categorical_accuracy: 0.3502\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------13-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4555 - categorical_accuracy: 0.3136\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------14-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4351 - categorical_accuracy: 0.3459\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------15-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4388 - categorical_accuracy: 0.3416\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4407 - categorical_accuracy: 0.3373\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------16-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4568 - categorical_accuracy: 0.3147\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------17-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4345 - categorical_accuracy: 0.3481\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4357 - categorical_accuracy: 0.3459\n",
      "--------------------18-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4445 - categorical_accuracy: 0.3297\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4543 - categorical_accuracy: 0.3179\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------19-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4367 - categorical_accuracy: 0.3427\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------20-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.2526 - categorical_accuracy: 0.3297\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.2223 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.2222 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.2221 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.2221 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.2221 - categorical_accuracy: 0.3470\n",
      "--------------------21-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4571 - categorical_accuracy: 0.3114\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4580 - categorical_accuracy: 0.3125\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 7/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 8/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 9/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 10/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------22-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4174 - categorical_accuracy: 0.3254\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------23-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4355 - categorical_accuracy: 0.3459\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4450 - categorical_accuracy: 0.3319\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------24-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4523 - categorical_accuracy: 0.3200\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------25-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4546 - categorical_accuracy: 0.3157\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------26-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4589 - categorical_accuracy: 0.3103\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------27-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4445 - categorical_accuracy: 0.3297\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4341 - categorical_accuracy: 0.3481\n",
      "Epoch 7/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 8/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 9/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 10/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------28-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.2473 - categorical_accuracy: 0.3276\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.2222 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.2221 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.2221 - categorical_accuracy: 0.3276\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.2221 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.2221 - categorical_accuracy: 0.3470\n",
      "--------------------29-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4582 - categorical_accuracy: 0.3125\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------30-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4572 - categorical_accuracy: 0.3125\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------31-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4566 - categorical_accuracy: 0.3125\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------32-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4409 - categorical_accuracy: 0.3351\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------33-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4434 - categorical_accuracy: 0.3341\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------34-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4560 - categorical_accuracy: 0.3136\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4439 - categorical_accuracy: 0.3330\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 7/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 8/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 9/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 10/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------35-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4499 - categorical_accuracy: 0.3244\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4583 - categorical_accuracy: 0.3125\n",
      "--------------------36-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4462 - categorical_accuracy: 0.3265\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------37-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4347 - categorical_accuracy: 0.3470\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------38-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4484 - categorical_accuracy: 0.3254\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------39-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4342 - categorical_accuracy: 0.3481\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4298 - categorical_accuracy: 0.3545\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------40-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4445 - categorical_accuracy: 0.3265\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4567 - categorical_accuracy: 0.3147\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4422 - categorical_accuracy: 0.3351\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 7/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 8/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------41-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4596 - categorical_accuracy: 0.3082\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4545 - categorical_accuracy: 0.3179\n",
      "Epoch 7/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 8/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 9/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 10/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 11/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------42-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4354 - categorical_accuracy: 0.3448\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------43-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4579 - categorical_accuracy: 0.3103\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------44-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4625 - categorical_accuracy: 0.3060\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------45-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4347 - categorical_accuracy: 0.3459\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------46-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4411 - categorical_accuracy: 0.3373\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4408 - categorical_accuracy: 0.3384\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------47-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4343 - categorical_accuracy: 0.3459\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4436 - categorical_accuracy: 0.3330\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------48-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4331 - categorical_accuracy: 0.3502\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------49-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4377 - categorical_accuracy: 0.3405\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------50-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4550 - categorical_accuracy: 0.3147\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4581 - categorical_accuracy: 0.3125\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 7/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 8/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------51-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4365 - categorical_accuracy: 0.3394\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------52-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4546 - categorical_accuracy: 0.3168\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------53-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4425 - categorical_accuracy: 0.3351\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------54-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4349 - categorical_accuracy: 0.3459\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------55-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4351 - categorical_accuracy: 0.3459\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------56-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.2491 - categorical_accuracy: 0.3028\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.2222 - categorical_accuracy: 0.3276\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.2221 - categorical_accuracy: 0.3276\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.2221 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.2221 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.2221 - categorical_accuracy: 0.3470\n",
      "Epoch 7/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.2221 - categorical_accuracy: 0.3470\n",
      "Epoch 8/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.2221 - categorical_accuracy: 0.3470\n",
      "--------------------57-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4582 - categorical_accuracy: 0.3114\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------58-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4395 - categorical_accuracy: 0.3405\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------59-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4567 - categorical_accuracy: 0.3147\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4433 - categorical_accuracy: 0.3341\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 7/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------60-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4449 - categorical_accuracy: 0.3308\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------61-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4360 - categorical_accuracy: 0.3448\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------62-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4412 - categorical_accuracy: 0.3373\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4403 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------63-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4423 - categorical_accuracy: 0.3330\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------64-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4314 - categorical_accuracy: 0.3448\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136: 0s - loss: 0.4433 - categorical_\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------65-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4398 - categorical_accuracy: 0.3384\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------66-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4370 - categorical_accuracy: 0.3416\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4343 - categorical_accuracy: 0.3481\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 7/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 8/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 9/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------67-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4363 - categorical_accuracy: 0.3448\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4296 - categorical_accuracy: 0.3556\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------68-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4335 - categorical_accuracy: 0.3362\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4575 - categorical_accuracy: 0.3125\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------69-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4589 - categorical_accuracy: 0.3103\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------70-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4397 - categorical_accuracy: 0.3394\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4497 - categorical_accuracy: 0.3254\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------71-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4512 - categorical_accuracy: 0.3211\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------72-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4405 - categorical_accuracy: 0.3384\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------73-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4400 - categorical_accuracy: 0.3362\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------74-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4348 - categorical_accuracy: 0.3459\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------75-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4360 - categorical_accuracy: 0.3448\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3459\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 7/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------76-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4349 - categorical_accuracy: 0.3416\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4197 - categorical_accuracy: 0.3696\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4481 - categorical_accuracy: 0.3276\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------77-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4376 - categorical_accuracy: 0.3427\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------78-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4384 - categorical_accuracy: 0.3416\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------79-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4359 - categorical_accuracy: 0.3448\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------80-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4344 - categorical_accuracy: 0.3459\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------81-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4482 - categorical_accuracy: 0.3265\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------82-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4352 - categorical_accuracy: 0.3459\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------83-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4399 - categorical_accuracy: 0.3384\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------84-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4338 - categorical_accuracy: 0.3481\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4350 - categorical_accuracy: 0.3459\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------85-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4590 - categorical_accuracy: 0.3103\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------86-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4424 - categorical_accuracy: 0.3341\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4445 - categorical_accuracy: 0.3330\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 7/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 8/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 9/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------87-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4356 - categorical_accuracy: 0.3416\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------88-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4397 - categorical_accuracy: 0.3362\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 2ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------89-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4393 - categorical_accuracy: 0.3373\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------90-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4578 - categorical_accuracy: 0.3114\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------91-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4340 - categorical_accuracy: 0.3448\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------92-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4584 - categorical_accuracy: 0.3103\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4488 - categorical_accuracy: 0.3265\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 7/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------93-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4295 - categorical_accuracy: 0.3502\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------94-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4432 - categorical_accuracy: 0.3341\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4489 - categorical_accuracy: 0.3265\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------95-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4387 - categorical_accuracy: 0.3416\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------96-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4306 - categorical_accuracy: 0.3513\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "--------------------97-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4565 - categorical_accuracy: 0.3147\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4576 - categorical_accuracy: 0.3136\n",
      "--------------------98-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4385 - categorical_accuracy: 0.3416\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4404 - categorical_accuracy: 0.3394\n",
      "--------------------99-------------------\n",
      "Epoch 1/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4345 - categorical_accuracy: 0.3438\n",
      "Epoch 2/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 3/100\n",
      "186/186 [==============================] - 0s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 4/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 5/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n",
      "Epoch 6/100\n",
      "186/186 [==============================] - 1s 3ms/step - loss: 0.4353 - categorical_accuracy: 0.3470\n"
     ]
    }
   ],
   "source": [
    "max_fit_count = 100\n",
    "early_stopping = EarlyStopping(monitor='categorical_accuracy', patience=5)\n",
    "best_model = None\n",
    "best_loss = 0\n",
    "for fit in range(max_fit_count):\n",
    "    print('--------------------{0}-------------------'.format(fit))\n",
    "    model = create_model_conv(neurons1=64, neurons2=64, neurons3=16, core=10, activation=\"relu\", optimizer='Adam', loss='mse')\n",
    "    hist = model.fit(train_x3, train_y2, epochs=100, batch_size=5, callbacks=[early_stopping])        \n",
    "    if not best_model or hist.history['categorical_accuracy'][-1] > best_loss:\n",
    "        best_model = model\n",
    "        best_loss = hist.history['categorical_accuracy'][-1]\n",
    "        if len(hist.history['categorical_accuracy']) > 70:\n",
    "            break\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[  0,   0,  84],\n",
       "       [  0,   0, 541],\n",
       "       [  0,   0,  63]], dtype=int64)"
      ]
     },
     "execution_count": 94,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred_y = best_model.predict(test_x3)\n",
    "pred_y = [np.argmax(y) for y in pred_y]\n",
    "pred_y = np.array(pred_y)\n",
    "confusion_matrix(test_y, pred_y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.313953488372093"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracy_score(test_y, pred_y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Попробуем обучить модели на сгенерированных данных"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>av_depth</th>\n",
       "      <th>max_depth</th>\n",
       "      <th>min_depth</th>\n",
       "      <th>num_acl</th>\n",
       "      <th>num_rel_cl</th>\n",
       "      <th>num_advcl</th>\n",
       "      <th>num_sent</th>\n",
       "      <th>num_tok</th>\n",
       "      <th>av_tok_before_root</th>\n",
       "      <th>av_len_sent</th>\n",
       "      <th>...</th>\n",
       "      <th>num_pres_plur</th>\n",
       "      <th>num_past_part</th>\n",
       "      <th>num_past_simple</th>\n",
       "      <th>num_linkings</th>\n",
       "      <th>num_4grams</th>\n",
       "      <th>num_func_ngrams</th>\n",
       "      <th>million_mistake</th>\n",
       "      <th>text</th>\n",
       "      <th>sum_punct</th>\n",
       "      <th>cls</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>idx</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>3.69</td>\n",
       "      <td>7</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>13</td>\n",
       "      <td>189</td>\n",
       "      <td>4.23</td>\n",
       "      <td>14.54</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>4</td>\n",
       "      <td>6</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>The graph illustrates the market situation abo...</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3.93</td>\n",
       "      <td>7</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>15</td>\n",
       "      <td>244</td>\n",
       "      <td>4.80</td>\n",
       "      <td>16.27</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>13</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>There is a linegraph and piechart. The linegra...</td>\n",
       "      <td>0</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2 rows × 68 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "     av_depth  max_depth  min_depth  num_acl  num_rel_cl  num_advcl  num_sent  \\\n",
       "idx                                                                             \n",
       "0        3.69          7          2        2           2          0        13   \n",
       "1        3.93          7          2        1           3          1        15   \n",
       "\n",
       "     num_tok  av_tok_before_root  av_len_sent  ...  num_pres_plur  \\\n",
       "idx                                            ...                  \n",
       "0        189                4.23        14.54  ...              0   \n",
       "1        244                4.80        16.27  ...              0   \n",
       "\n",
       "     num_past_part  num_past_simple  num_linkings  num_4grams  \\\n",
       "idx                                                             \n",
       "0                2                4             6           3   \n",
       "1                3                0            13           2   \n",
       "\n",
       "     num_func_ngrams  million_mistake  \\\n",
       "idx                                     \n",
       "0                  1                0   \n",
       "1                  0                0   \n",
       "\n",
       "                                                  text  sum_punct  cls  \n",
       "idx                                                                     \n",
       "0    The graph illustrates the market situation abo...          0    4  \n",
       "1    There is a linegraph and piechart. The linegra...          0    4  \n",
       "\n",
       "[2 rows x 68 columns]"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_gen = pd.read_csv(r'D:\\hse_iot\\VKR\\my_inspector\\data\\generated_data.csv', index_col='idx')\n",
    "df_gen.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "Int64Index: 9100 entries, 0 to 9099\n",
      "Data columns (total 68 columns):\n",
      " #   Column              Non-Null Count  Dtype  \n",
      "---  ------              --------------  -----  \n",
      " 0   av_depth            9100 non-null   float64\n",
      " 1   max_depth           9100 non-null   int64  \n",
      " 2   min_depth           9100 non-null   int64  \n",
      " 3   num_acl             9100 non-null   int64  \n",
      " 4   num_rel_cl          9100 non-null   int64  \n",
      " 5   num_advcl           9100 non-null   int64  \n",
      " 6   num_sent            9100 non-null   int64  \n",
      " 7   num_tok             9100 non-null   int64  \n",
      " 8   av_tok_before_root  9100 non-null   float64\n",
      " 9   av_len_sent         9100 non-null   float64\n",
      " 10  num_cl              9100 non-null   int64  \n",
      " 11  num_tu              9100 non-null   int64  \n",
      " 12  num_compl_tu        9100 non-null   int64  \n",
      " 13  num_coord           9100 non-null   int64  \n",
      " 14  num_poss            9100 non-null   int64  \n",
      " 15  num_prep            9100 non-null   int64  \n",
      " 16  num_adj_noun        9100 non-null   int64  \n",
      " 17  num_part_noun       9100 non-null   int64  \n",
      " 18  num_noun_inf        9100 non-null   int64  \n",
      " 19  pos_sim_nei         9100 non-null   float64\n",
      " 20  pos_sim_all         9100 non-null   float64\n",
      " 21  lemma_sim_all       9100 non-null   float64\n",
      " 22  lemma_sim_nei       9100 non-null   float64\n",
      " 23  density             9100 non-null   float64\n",
      " 24  ls                  9100 non-null   float64\n",
      " 25  vs                  9100 non-null   float64\n",
      " 26  corrected_vs        9100 non-null   float64\n",
      " 27  squared_vs          9100 non-null   float64\n",
      " 28  lfp_1000            9100 non-null   float64\n",
      " 29  lfp_2000            9100 non-null   float64\n",
      " 30  lfp_uwl             9100 non-null   float64\n",
      " 31  lfp_rest            9100 non-null   float64\n",
      " 32  ndw                 9100 non-null   int64  \n",
      " 33  ttr                 9100 non-null   float64\n",
      " 34  corrected_ttr       9100 non-null   float64\n",
      " 35  root_ttr            9100 non-null   float64\n",
      " 36  log_ttr             9100 non-null   float64\n",
      " 37  uber_ttr            9100 non-null   float64\n",
      " 38  d                   9100 non-null   float64\n",
      " 39  lv                  9100 non-null   float64\n",
      " 40  vvi                 9100 non-null   float64\n",
      " 41  squared_vv          9100 non-null   float64\n",
      " 42  corrected_vv        9100 non-null   float64\n",
      " 43  vvii                9100 non-null   float64\n",
      " 44  nv                  9100 non-null   float64\n",
      " 45  adjv                9100 non-null   float64\n",
      " 46  advv                9100 non-null   float64\n",
      " 47  modv                9100 non-null   float64\n",
      " 48  der_level3          9100 non-null   float64\n",
      " 49  der_level4          9100 non-null   float64\n",
      " 50  der_level5          9100 non-null   float64\n",
      " 51  der_level6          9100 non-null   float64\n",
      " 52  mci                 9100 non-null   float64\n",
      " 53  freq_finite_forms   9100 non-null   float64\n",
      " 54  freq_aux            9100 non-null   float64\n",
      " 55  num_inf             9100 non-null   int64  \n",
      " 56  num_gerunds         9100 non-null   int64  \n",
      " 57  num_pres_sing       9100 non-null   int64  \n",
      " 58  num_pres_plur       9100 non-null   int64  \n",
      " 59  num_past_part       9100 non-null   int64  \n",
      " 60  num_past_simple     9100 non-null   int64  \n",
      " 61  num_linkings        9100 non-null   int64  \n",
      " 62  num_4grams          9100 non-null   int64  \n",
      " 63  num_func_ngrams     9100 non-null   int64  \n",
      " 64  million_mistake     9100 non-null   int64  \n",
      " 65  text                9100 non-null   object \n",
      " 66  sum_punct           9100 non-null   int64  \n",
      " 67  cls                 9100 non-null   int64  \n",
      "dtypes: float64(38), int64(29), object(1)\n",
      "memory usage: 4.8+ MB\n"
     ]
    }
   ],
   "source": [
    "df_gen.info()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "5    2441\n",
       "4    2386\n",
       "3    2257\n",
       "2    2016\n",
       "Name: cls, dtype: int64"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_gen['cls'].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x720 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sns.set(rc={'figure.figsize':(20,10)})\n",
    "sns_plt = sns.countplot(x='cls', data=df_gen, palette=\"Set2\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Т.к. сгенерированны данные только для тренировочной выборки, то тестовую выборку получим из начальных данных"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv(r'D:\\hse_iot\\VKR\\my_inspector\\data\\classification_dataset.csv')\n",
    "df.drop(['num_misspelled_tokens', 'num_shell_noun', 'name', 'type', 'punct_mistakes_pp', 'punct_mistakes_because', 'punct_mistakes_but', 'punct_mistakes_compare', 'side_mistake'],  inplace=True, axis=1)\n",
    "df['class'] = round(df['class'] / 10).astype(int)\n",
    "df.loc[df['class'] == 1, 'class'] = 2\n",
    "df.loc[df['class'] == 3, 'class'] = 2\n",
    "df.loc[df['class'] == 4, 'class'] = 3\n",
    "df.loc[df['class'] == 5, 'class'] = 3\n",
    "df.loc[df['class'] == 6, 'class'] = 4\n",
    "df.loc[df['class'] == 7, 'class'] = 4\n",
    "df.loc[df['class'] == 8, 'class'] = 5\n",
    "\n",
    "df = df.rename(columns = {'class': 'cls'})\n",
    "feature_columns = [col for col in df.columns if col not in ['cls',]]\n",
    "X = df[feature_columns]\n",
    "y = df['cls']\n",
    "train_x, test_x, train_y, test_y = train_test_split(X, y, test_size=0.2, random_state = SEED)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "feature_columns = [col for col in df_gen.columns if col not in ['cls',]]\n",
    "train_x = df_gen[feature_columns]\n",
    "train_y = df_gen['cls']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}