detection_facial_expression_Metodo2.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Dataset (Load , exploration and  Cleaning)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "d:\\ProgramData\\Anaconda3\\lib\\site-packages\\h5py\\__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  from ._conv import register_converters as _register_converters\n",
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "# calculations and Dataframe\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "# plot facial expression \n",
    "import matplotlib.pyplot as plt\n",
    "from skimage import io\n",
    "# TensorFlow and tf.keras\n",
    "import tensorflow as tf\n",
    "import keras\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout\n",
    "from keras.optimizers import RMSprop\n",
    "# Paquetes necesarios para la morfología matemática\n",
    "from skimage.morphology import erosion, dilation, opening, closing\n",
    "# Elementos estructurales\n",
    "from skimage.morphology import disk, diamond, ball, rectangle\n",
    "# Paquetes necesarios para la conversión de imágenes de color a escala de grises\n",
    "from skimage.color import rgb2gray\n",
    "from skimage.filters import gaussian\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# load dataset\n",
    "dataset_facial_expression = pd.read_csv(\"../fer2013.csv\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Data exploration"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "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>emotion</th>\n",
       "      <th>pixels</th>\n",
       "      <th>Usage</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>70 80 82 72 58 58 60 63 54 58 60 48 89 115 121...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>151 150 147 155 148 133 111 140 170 174 182 15...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>231 212 156 164 174 138 161 173 182 200 106 38...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>24 32 36 30 32 23 19 20 30 41 21 22 32 34 21 1...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>6</td>\n",
       "      <td>4 0 0 0 0 0 0 0 0 0 0 0 3 15 23 28 48 50 58 84...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   emotion                                             pixels     Usage\n",
       "0        0  70 80 82 72 58 58 60 63 54 58 60 48 89 115 121...  Training\n",
       "1        0  151 150 147 155 148 133 111 140 170 174 182 15...  Training\n",
       "2        2  231 212 156 164 174 138 161 173 182 200 106 38...  Training\n",
       "3        4  24 32 36 30 32 23 19 20 30 41 21 22 32 34 21 1...  Training\n",
       "4        6  4 0 0 0 0 0 0 0 0 0 0 0 3 15 23 28 48 50 58 84...  Training"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# display the top five\n",
    "dataset_facial_expression.head()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "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>emotion</th>\n",
       "      <th>pixels</th>\n",
       "      <th>Usage</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>35882</th>\n",
       "      <td>6</td>\n",
       "      <td>50 36 17 22 23 29 33 39 34 37 37 37 39 43 48 5...</td>\n",
       "      <td>PrivateTest</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35883</th>\n",
       "      <td>3</td>\n",
       "      <td>178 174 172 173 181 188 191 194 196 199 200 20...</td>\n",
       "      <td>PrivateTest</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35884</th>\n",
       "      <td>0</td>\n",
       "      <td>17 17 16 23 28 22 19 17 25 26 20 24 31 19 27 9...</td>\n",
       "      <td>PrivateTest</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35885</th>\n",
       "      <td>3</td>\n",
       "      <td>30 28 28 29 31 30 42 68 79 81 77 67 67 71 63 6...</td>\n",
       "      <td>PrivateTest</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35886</th>\n",
       "      <td>2</td>\n",
       "      <td>19 13 14 12 13 16 21 33 50 57 71 84 97 108 122...</td>\n",
       "      <td>PrivateTest</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       emotion                                             pixels        Usage\n",
       "35882        6  50 36 17 22 23 29 33 39 34 37 37 37 39 43 48 5...  PrivateTest\n",
       "35883        3  178 174 172 173 181 188 191 194 196 199 200 20...  PrivateTest\n",
       "35884        0  17 17 16 23 28 22 19 17 25 26 20 24 31 19 27 9...  PrivateTest\n",
       "35885        3  30 28 28 29 31 30 42 68 79 81 77 67 67 71 63 6...  PrivateTest\n",
       "35886        2  19 13 14 12 13 16 21 33 50 57 71 84 97 108 122...  PrivateTest"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# display the last five\n",
    "dataset_facial_expression.tail()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Detecting missing, nan or null"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "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>emotion</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>35887.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>3.323265</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>1.873819</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>2.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>3.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>5.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>6.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            emotion\n",
       "count  35887.000000\n",
       "mean       3.323265\n",
       "std        1.873819\n",
       "min        0.000000\n",
       "25%        2.000000\n",
       "50%        3.000000\n",
       "75%        5.000000\n",
       "max        6.000000"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Summary statistics: Numeric data\n",
    "dataset_facial_expression.describe()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "pandas.core.frame.DataFrame"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "type(dataset_facial_expression)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "35887"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(dataset_facial_expression)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'pandas.core.frame.DataFrame'>\n",
      "RangeIndex: 35887 entries, 0 to 35886\n",
      "Data columns (total 3 columns):\n",
      "emotion    35887 non-null int64\n",
      "pixels     35887 non-null object\n",
      "Usage      35887 non-null object\n",
      "dtypes: int64(1), object(2)\n",
      "memory usage: 841.2+ KB\n"
     ]
    }
   ],
   "source": [
    "dataset_facial_expression.info()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "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>emotion</th>\n",
       "      <th>pixels</th>\n",
       "      <th>Usage</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>5</td>\n",
       "      <td>6 9 12 38 77 60 28 21 17 21 36 54 70 77 87 96 ...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>6</td>\n",
       "      <td>139 51 92 106 111 122 129 130 131 144 153 158 ...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>6</td>\n",
       "      <td>64 62 70 77 77 94 108 118 127 136 138 139 153 ...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2</td>\n",
       "      <td>11 10 7 3 0 1 0 15 71 98 112 122 141 151 163 1...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>188 186 190 161 38 13 35 49 50 51 51 48 51 52 ...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>3</td>\n",
       "      <td>81 72 65 54 41 22 19 25 23 43 70 75 84 96 105 ...</td>\n",
       "      <td>PublicTest</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>3</td>\n",
       "      <td>218 216 226 174 23 23 16 10 14 55 43 77 67 45 ...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>4</td>\n",
       "      <td>234 229 239 173 116 144 144 145 113 96 165 180...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>5</td>\n",
       "      <td>233 162 97 80 89 103 107 111 93 62 70 76 76 63...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>3</td>\n",
       "      <td>228 237 236 238 239 239 237 239 240 241 242 24...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>0</td>\n",
       "      <td>169 156 129 53 6 47 45 46 52 60 55 53 50 38 55...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>3</td>\n",
       "      <td>57 58 56 55 56 56 55 52 34 49 42 64 59 47 55 5...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>4</td>\n",
       "      <td>57 65 54 62 50 48 56 59 65 65 63 63 68 72 73 7...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>6</td>\n",
       "      <td>90 90 91 92 92 93 94 92 91 92 92 93 95 98 97 9...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>0</td>\n",
       "      <td>54 58 62 64 64 65 64 57 65 73 78 83 88 97 104 ...</td>\n",
       "      <td>PrivateTest</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>2</td>\n",
       "      <td>255 255 255 255 255 255 255 255 250 240 237 23...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>4</td>\n",
       "      <td>10 10 10 10 10 10 10 10 11 11 10 12 14 8 23 45...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>2</td>\n",
       "      <td>40 28 28 29 32 31 31 31 25 27 33 35 31 32 37 3...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>6</td>\n",
       "      <td>19 20 18 16 31 26 15 37 38 29 42 47 40 41 40 4...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>3</td>\n",
       "      <td>221 223 225 226 226 225 223 225 223 222 219 21...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>0</td>\n",
       "      <td>63 65 72 81 87 100 116 138 154 165 170 174 181...</td>\n",
       "      <td>PublicTest</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>2</td>\n",
       "      <td>106 104 136 160 163 167 180 189 194 200 201 20...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>3</td>\n",
       "      <td>218 207 217 219 217 229 204 161 147 131 124 11...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>5</td>\n",
       "      <td>50 55 65 69 80 86 91 98 110 115 120 131 138 14...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>0</td>\n",
       "      <td>17 15 18 62 70 29 39 25 52 68 68 69 105 120 12...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>3</td>\n",
       "      <td>244 244 244 244 244 246 246 246 244 241 240 23...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>3</td>\n",
       "      <td>209 207 203 203 202 196 191 192 187 176 164 13...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>4</td>\n",
       "      <td>182 185 191 186 186 196 193 201 194 201 201 18...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>2</td>\n",
       "      <td>147 133 140 139 122 106 108 112 104 95 101 106...</td>\n",
       "      <td>PublicTest</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>3</td>\n",
       "      <td>190 192 191 190 179 167 172 180 195 212 217 21...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35857</th>\n",
       "      <td>3</td>\n",
       "      <td>225 232 235 204 171 200 185 175 130 115 165 16...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35858</th>\n",
       "      <td>2</td>\n",
       "      <td>227 227 227 227 227 229 229 228 235 202 93 84 ...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35859</th>\n",
       "      <td>0</td>\n",
       "      <td>27 26 29 27 61 43 64 36 12 37 29 41 46 29 23 2...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35860</th>\n",
       "      <td>2</td>\n",
       "      <td>207 207 208 209 210 214 200 94 75 81 54 46 45 ...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35861</th>\n",
       "      <td>3</td>\n",
       "      <td>228 217 122 134 134 142 138 124 122 87 124 108...</td>\n",
       "      <td>PrivateTest</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35862</th>\n",
       "      <td>6</td>\n",
       "      <td>151 152 151 160 83 18 34 37 42 42 43 50 55 56 ...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35863</th>\n",
       "      <td>2</td>\n",
       "      <td>50 48 46 41 53 51 33 42 41 36 36 33 33 23 23 3...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35864</th>\n",
       "      <td>6</td>\n",
       "      <td>245 246 246 246 247 247 248 246 246 234 81 35 ...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35865</th>\n",
       "      <td>2</td>\n",
       "      <td>8 10 12 13 17 20 20 22 27 32 47 68 77 84 94 10...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35866</th>\n",
       "      <td>6</td>\n",
       "      <td>3 0 0 1 4 6 0 2 11 25 55 70 62 92 132 152 183 ...</td>\n",
       "      <td>PrivateTest</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35867</th>\n",
       "      <td>2</td>\n",
       "      <td>207 11 125 235 208 211 218 219 207 214 223 198...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35868</th>\n",
       "      <td>6</td>\n",
       "      <td>50 53 56 49 48 50 59 62 83 84 89 89 117 149 17...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35869</th>\n",
       "      <td>3</td>\n",
       "      <td>180 151 162 179 185 168 127 80 58 84 130 100 5...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35870</th>\n",
       "      <td>3</td>\n",
       "      <td>244 224 198 164 217 214 144 177 213 232 222 21...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35871</th>\n",
       "      <td>6</td>\n",
       "      <td>69 62 62 63 98 120 64 75 73 87 91 92 92 103 96...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35872</th>\n",
       "      <td>4</td>\n",
       "      <td>27 24 26 34 52 67 84 97 97 102 111 124 126 125...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35873</th>\n",
       "      <td>6</td>\n",
       "      <td>120 125 107 95 95 116 113 113 109 97 94 90 87 ...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35874</th>\n",
       "      <td>5</td>\n",
       "      <td>25 30 40 52 69 80 86 102 121 133 145 153 158 1...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35875</th>\n",
       "      <td>6</td>\n",
       "      <td>188 166 176 177 154 154 173 181 204 206 196 20...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35876</th>\n",
       "      <td>5</td>\n",
       "      <td>49 69 83 98 86 74 82 88 88 91 103 112 122 124 ...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35877</th>\n",
       "      <td>6</td>\n",
       "      <td>63 60 17 9 22 21 32 51 60 59 46 50 70 91 98 96...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35878</th>\n",
       "      <td>3</td>\n",
       "      <td>112 57 83 127 45 80 167 143 41 67 65 59 62 58 ...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35879</th>\n",
       "      <td>0</td>\n",
       "      <td>129 169 187 165 189 200 196 199 202 198 204 20...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35880</th>\n",
       "      <td>4</td>\n",
       "      <td>97 103 114 120 127 136 141 145 145 145 143 143...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35881</th>\n",
       "      <td>4</td>\n",
       "      <td>97 95 96 96 96 95 96 95 95 93 90 89 90 97 132 ...</td>\n",
       "      <td>PublicTest</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35882</th>\n",
       "      <td>2</td>\n",
       "      <td>61 100 114 111 117 112 113 123 124 125 128 131...</td>\n",
       "      <td>PrivateTest</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35883</th>\n",
       "      <td>3</td>\n",
       "      <td>234 168 152 74 33 36 33 33 27 25 33 54 70 80 8...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35884</th>\n",
       "      <td>4</td>\n",
       "      <td>22 33 40 42 36 45 47 46 49 61 76 87 96 108 113...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35885</th>\n",
       "      <td>4</td>\n",
       "      <td>5 5 5 5 5 5 5 5 5 5 5 5 5 5 0 48 87 71 70 42 2...</td>\n",
       "      <td>PrivateTest</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35886</th>\n",
       "      <td>3</td>\n",
       "      <td>215 219 197 192 195 196 212 227 223 205 182 16...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>35887 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       emotion                                             pixels        Usage\n",
       "0            5  6 9 12 38 77 60 28 21 17 21 36 54 70 77 87 96 ...     Training\n",
       "1            6  139 51 92 106 111 122 129 130 131 144 153 158 ...     Training\n",
       "2            6  64 62 70 77 77 94 108 118 127 136 138 139 153 ...     Training\n",
       "3            2  11 10 7 3 0 1 0 15 71 98 112 122 141 151 163 1...     Training\n",
       "4            5  188 186 190 161 38 13 35 49 50 51 51 48 51 52 ...     Training\n",
       "5            3  81 72 65 54 41 22 19 25 23 43 70 75 84 96 105 ...   PublicTest\n",
       "6            3  218 216 226 174 23 23 16 10 14 55 43 77 67 45 ...     Training\n",
       "7            4  234 229 239 173 116 144 144 145 113 96 165 180...     Training\n",
       "8            5  233 162 97 80 89 103 107 111 93 62 70 76 76 63...     Training\n",
       "9            3  228 237 236 238 239 239 237 239 240 241 242 24...     Training\n",
       "10           0  169 156 129 53 6 47 45 46 52 60 55 53 50 38 55...     Training\n",
       "11           3  57 58 56 55 56 56 55 52 34 49 42 64 59 47 55 5...     Training\n",
       "12           4  57 65 54 62 50 48 56 59 65 65 63 63 68 72 73 7...     Training\n",
       "13           6  90 90 91 92 92 93 94 92 91 92 92 93 95 98 97 9...     Training\n",
       "14           0  54 58 62 64 64 65 64 57 65 73 78 83 88 97 104 ...  PrivateTest\n",
       "15           2  255 255 255 255 255 255 255 255 250 240 237 23...     Training\n",
       "16           4  10 10 10 10 10 10 10 10 11 11 10 12 14 8 23 45...     Training\n",
       "17           2  40 28 28 29 32 31 31 31 25 27 33 35 31 32 37 3...     Training\n",
       "18           6  19 20 18 16 31 26 15 37 38 29 42 47 40 41 40 4...     Training\n",
       "19           3  221 223 225 226 226 225 223 225 223 222 219 21...     Training\n",
       "20           0  63 65 72 81 87 100 116 138 154 165 170 174 181...   PublicTest\n",
       "21           2  106 104 136 160 163 167 180 189 194 200 201 20...     Training\n",
       "22           3  218 207 217 219 217 229 204 161 147 131 124 11...     Training\n",
       "23           5  50 55 65 69 80 86 91 98 110 115 120 131 138 14...     Training\n",
       "24           0  17 15 18 62 70 29 39 25 52 68 68 69 105 120 12...     Training\n",
       "25           3  244 244 244 244 244 246 246 246 244 241 240 23...     Training\n",
       "26           3  209 207 203 203 202 196 191 192 187 176 164 13...     Training\n",
       "27           4  182 185 191 186 186 196 193 201 194 201 201 18...     Training\n",
       "28           2  147 133 140 139 122 106 108 112 104 95 101 106...   PublicTest\n",
       "29           3  190 192 191 190 179 167 172 180 195 212 217 21...     Training\n",
       "...        ...                                                ...          ...\n",
       "35857        3  225 232 235 204 171 200 185 175 130 115 165 16...     Training\n",
       "35858        2  227 227 227 227 227 229 229 228 235 202 93 84 ...     Training\n",
       "35859        0  27 26 29 27 61 43 64 36 12 37 29 41 46 29 23 2...     Training\n",
       "35860        2  207 207 208 209 210 214 200 94 75 81 54 46 45 ...     Training\n",
       "35861        3  228 217 122 134 134 142 138 124 122 87 124 108...  PrivateTest\n",
       "35862        6  151 152 151 160 83 18 34 37 42 42 43 50 55 56 ...     Training\n",
       "35863        2  50 48 46 41 53 51 33 42 41 36 36 33 33 23 23 3...     Training\n",
       "35864        6  245 246 246 246 247 247 248 246 246 234 81 35 ...     Training\n",
       "35865        2  8 10 12 13 17 20 20 22 27 32 47 68 77 84 94 10...     Training\n",
       "35866        6  3 0 0 1 4 6 0 2 11 25 55 70 62 92 132 152 183 ...  PrivateTest\n",
       "35867        2  207 11 125 235 208 211 218 219 207 214 223 198...     Training\n",
       "35868        6  50 53 56 49 48 50 59 62 83 84 89 89 117 149 17...     Training\n",
       "35869        3  180 151 162 179 185 168 127 80 58 84 130 100 5...     Training\n",
       "35870        3  244 224 198 164 217 214 144 177 213 232 222 21...     Training\n",
       "35871        6  69 62 62 63 98 120 64 75 73 87 91 92 92 103 96...     Training\n",
       "35872        4  27 24 26 34 52 67 84 97 97 102 111 124 126 125...     Training\n",
       "35873        6  120 125 107 95 95 116 113 113 109 97 94 90 87 ...     Training\n",
       "35874        5  25 30 40 52 69 80 86 102 121 133 145 153 158 1...     Training\n",
       "35875        6  188 166 176 177 154 154 173 181 204 206 196 20...     Training\n",
       "35876        5  49 69 83 98 86 74 82 88 88 91 103 112 122 124 ...     Training\n",
       "35877        6  63 60 17 9 22 21 32 51 60 59 46 50 70 91 98 96...     Training\n",
       "35878        3  112 57 83 127 45 80 167 143 41 67 65 59 62 58 ...     Training\n",
       "35879        0  129 169 187 165 189 200 196 199 202 198 204 20...     Training\n",
       "35880        4  97 103 114 120 127 136 141 145 145 145 143 143...     Training\n",
       "35881        4  97 95 96 96 96 95 96 95 95 93 90 89 90 97 132 ...   PublicTest\n",
       "35882        2  61 100 114 111 117 112 113 123 124 125 128 131...  PrivateTest\n",
       "35883        3  234 168 152 74 33 36 33 33 27 25 33 54 70 80 8...     Training\n",
       "35884        4  22 33 40 42 36 45 47 46 49 61 76 87 96 108 113...     Training\n",
       "35885        4  5 5 5 5 5 5 5 5 5 5 5 5 5 5 0 48 87 71 70 42 2...  PrivateTest\n",
       "35886        3  215 219 197 192 195 196 212 227 223 205 182 16...     Training\n",
       "\n",
       "[35887 rows x 3 columns]"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Modify a sequence in-place by shuffling its contents. \n",
    "dataset_facial_expression = dataset_facial_expression.sample(frac=1).reset_index(drop=True)\n",
    "dataset_facial_expression\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Data Exploration and plot pixels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2e10536b38>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "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>emotion</th>\n",
       "      <th>count</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Happy</td>\n",
       "      <td>8989</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>Neutral</td>\n",
       "      <td>6198</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Sad</td>\n",
       "      <td>6077</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Fear</td>\n",
       "      <td>5121</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Angry</td>\n",
       "      <td>4953</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Surprise</td>\n",
       "      <td>4002</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Disgust</td>\n",
       "      <td>547</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    emotion  count\n",
       "3     Happy   8989\n",
       "6   Neutral   6198\n",
       "4       Sad   6077\n",
       "2      Fear   5121\n",
       "0     Angry   4953\n",
       "5  Surprise   4002\n",
       "1   Disgust    547"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "delais_emotion_label = (\"Happy\" , \"Neutral\", \"Sad\", \"Fear\" , \"Angry\", \"Surprise\" ,\"Disgust\")\n",
    "delais_emotion = pd.DataFrame(data=dataset_facial_expression[\"emotion\"].value_counts())\n",
    "\n",
    "delais_emotion[\"count\"] = delais_emotion[\"emotion\"]\n",
    "delais_emotion[\"emotion\"] = delais_emotion_label\n",
    "delais_emotion\n",
    "%matplotlib inline\n",
    "fig, ax = plt.subplots()\n",
    "\n",
    "ax.bar(delais_emotion[\"emotion\"], delais_emotion[\"count\"] )\n",
    "ax.set_title(\"Data Exploration\")\n",
    "ax.set_ylabel(\"Counts\")\n",
    "ax.set_xlabel(\"Emotions\")\n",
    "ax.grid()\n",
    "\n",
    "fig.tight_layout()\n",
    "plt.show()\n",
    "delais_emotion\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "y = np.zeros((35887, 48, 48))\n",
    "#35887\n",
    "for i in range(35887):\n",
    "    imagen = dataset_facial_expression[\"pixels\"][0+i]\n",
    "    imagen_list_pixel = imagen.split(\" \")\n",
    "    imagen_array_pixel = np.asarray(imagen_list_pixel, dtype=float)\n",
    "    imagen_rezise_pixel = imagen_array_pixel.reshape(-1,48,48)    \n",
    "    image_gray = rgb2gray(imagen_rezise_pixel[0])\n",
    "    image_gray = gaussian(image_gray, 1)\n",
    "    image_eroded = image_gray-erosion(image_gray,disk(1.8))\n",
    "    imagen_array_pixel = np.asarray(image_eroded, dtype=float)\n",
    "    imagen_rezise_pixel = imagen_array_pixel.reshape(-1,48,48)   \n",
    "    y[i] = y[i]+imagen_rezise_pixel[0]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "7177"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_images = y[:28709]\n",
    "train_labels = np.array(dataset_facial_expression[\"emotion\"][:28709])\n",
    "test_images = y[28710:]\n",
    "test_labels = np.array( dataset_facial_expression[\"emotion\"][28710:])\n",
    "len(test_images)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2e10536400>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure()\n",
    "plt.imshow(train_images[0], cmap=\"gray\")\n",
    "plt.colorbar()\n",
    "plt.grid(False)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "28709 train examples\n",
      "7177 test examples\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(28709, 2304)"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "batch_size = 1024\n",
    "num_classes = 7\n",
    "epochs = 1000\n",
    "\n",
    "train_images = train_images.reshape(28709 , 2304)\n",
    "test_images = test_images.reshape(7177 , 2304)\n",
    "train_images = train_images.astype(\"float32\")\n",
    "test_images = test_images.astype(\"float32\")\n",
    "train_images /= 255\n",
    "test_images /= 255\n",
    "\n",
    "print(train_images.shape[0], \"train examples\") \n",
    "print(test_images.shape[0], \"test examples\") \n",
    "\n",
    "train_labels = keras.utils.to_categorical(train_labels, num_classes)\n",
    "test_labels = keras.utils.to_categorical(test_labels, num_classes)\n",
    "\n",
    "train_images.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = Sequential()\n",
    "model.add(Dense(512, activation=\"relu\", input_shape=(2304,) ))\n",
    "model.add(Dropout(1))\n",
    "model.add(Dense(512, activation=\"relu\"))\n",
    "model.add(Dropout(1))\n",
    "model.add(Dense(num_classes, activation=\"softmax\"))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_7 (Dense)              (None, 512)               1180160   \n",
      "_________________________________________________________________\n",
      "dropout_5 (Dropout)          (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dense_8 (Dense)              (None, 512)               262656    \n",
      "_________________________________________________________________\n",
      "dropout_6 (Dropout)          (None, 512)               0         \n",
      "_________________________________________________________________\n",
      "dense_9 (Dense)              (None, 7)                 3591      \n",
      "=================================================================\n",
      "Total params: 1,446,407\n",
      "Trainable params: 1,446,407\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(loss=\"categorical_crossentropy\",\n",
    "             optimizer=RMSprop(),\n",
    "             metrics=[\"accuracy\"])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 28709 samples, validate on 7177 samples\n",
      "Epoch 1/1000\n",
      "28709/28709 [==============================] - 6s 204us/step - loss: 1.8717 - acc: 0.2432 - val_loss: 1.8484 - val_acc: 0.2448\n",
      "Epoch 2/1000\n",
      "28709/28709 [==============================] - 5s 187us/step - loss: 1.7828 - acc: 0.2751 - val_loss: 1.8647 - val_acc: 0.1909\n",
      "Epoch 3/1000\n",
      "28709/28709 [==============================] - 5s 186us/step - loss: 1.7474 - acc: 0.2959 - val_loss: 1.8871 - val_acc: 0.2037\n",
      "Epoch 4/1000\n",
      "28709/28709 [==============================] - 5s 186us/step - loss: 1.7290 - acc: 0.3046 - val_loss: 1.7461 - val_acc: 0.2927\n",
      "Epoch 5/1000\n",
      "28709/28709 [==============================] - 5s 188us/step - loss: 1.7133 - acc: 0.3142 - val_loss: 1.8409 - val_acc: 0.2367\n",
      "Epoch 6/1000\n",
      "28709/28709 [==============================] - 5s 188us/step - loss: 1.6992 - acc: 0.3220 - val_loss: 1.7921 - val_acc: 0.3001\n",
      "Epoch 7/1000\n",
      "28709/28709 [==============================] - 5s 190us/step - loss: 1.6949 - acc: 0.3288 - val_loss: 1.7943 - val_acc: 0.2451\n",
      "Epoch 8/1000\n",
      "28709/28709 [==============================] - 5s 187us/step - loss: 1.6682 - acc: 0.3388 - val_loss: 1.7664 - val_acc: 0.2874\n",
      "Epoch 9/1000\n",
      "28709/28709 [==============================] - 5s 188us/step - loss: 1.6564 - acc: 0.3425 - val_loss: 1.8345 - val_acc: 0.2437\n",
      "Epoch 10/1000\n",
      "28709/28709 [==============================] - 5s 188us/step - loss: 1.6460 - acc: 0.3495 - val_loss: 1.8596 - val_acc: 0.2434\n",
      "Epoch 11/1000\n",
      "28709/28709 [==============================] - 5s 186us/step - loss: 1.6363 - acc: 0.3577 - val_loss: 1.8190 - val_acc: 0.3455\n",
      "Epoch 12/1000\n",
      "28709/28709 [==============================] - 5s 186us/step - loss: 1.6173 - acc: 0.3681 - val_loss: 1.7003 - val_acc: 0.3209\n",
      "Epoch 13/1000\n",
      "28709/28709 [==============================] - 5s 191us/step - loss: 1.6068 - acc: 0.3675 - val_loss: 1.9014 - val_acc: 0.2913\n",
      "Epoch 14/1000\n",
      "28709/28709 [==============================] - 8s 282us/step - loss: 1.5919 - acc: 0.3760 - val_loss: 1.7164 - val_acc: 0.3539\n",
      "Epoch 15/1000\n",
      "28709/28709 [==============================] - 6s 217us/step - loss: 1.5651 - acc: 0.3866 - val_loss: 1.7335 - val_acc: 0.2936\n",
      "Epoch 16/1000\n",
      "28709/28709 [==============================] - 8s 284us/step - loss: 1.5636 - acc: 0.3902 - val_loss: 1.9372 - val_acc: 0.2877\n",
      "Epoch 17/1000\n",
      "28709/28709 [==============================] - 8s 266us/step - loss: 1.5340 - acc: 0.4012 - val_loss: 1.8971 - val_acc: 0.2483\n",
      "Epoch 18/1000\n",
      "28709/28709 [==============================] - 10s 357us/step - loss: 1.5282 - acc: 0.4104 - val_loss: 1.7664 - val_acc: 0.3145\n",
      "Epoch 19/1000\n",
      "28709/28709 [==============================] - 5s 187us/step - loss: 1.5010 - acc: 0.4198 - val_loss: 1.7420 - val_acc: 0.3280\n",
      "Epoch 20/1000\n",
      "28709/28709 [==============================] - 10s 352us/step - loss: 1.4838 - acc: 0.4304 - val_loss: 1.7132 - val_acc: 0.3329\n",
      "Epoch 21/1000\n",
      "28709/28709 [==============================] - 5s 188us/step - loss: 1.4732 - acc: 0.4334 - val_loss: 1.7429 - val_acc: 0.3613\n",
      "Epoch 22/1000\n",
      "28709/28709 [==============================] - 11s 373us/step - loss: 1.4461 - acc: 0.4511 - val_loss: 1.7385 - val_acc: 0.3504\n",
      "Epoch 23/1000\n",
      "28709/28709 [==============================] - 9s 307us/step - loss: 1.4252 - acc: 0.4548 - val_loss: 1.8959 - val_acc: 0.3109\n",
      "Epoch 24/1000\n",
      "28709/28709 [==============================] - 7s 257us/step - loss: 1.4056 - acc: 0.4639 - val_loss: 1.8247 - val_acc: 0.3044\n",
      "Epoch 25/1000\n",
      "28709/28709 [==============================] - 12s 407us/step - loss: 1.4090 - acc: 0.4633 - val_loss: 1.7528 - val_acc: 0.3529\n",
      "Epoch 26/1000\n",
      "28709/28709 [==============================] - 9s 311us/step - loss: 1.3757 - acc: 0.4779 - val_loss: 1.7848 - val_acc: 0.3574\n",
      "Epoch 27/1000\n",
      "28709/28709 [==============================] - 9s 297us/step - loss: 1.3462 - acc: 0.4870 - val_loss: 1.7671 - val_acc: 0.3564\n",
      "Epoch 28/1000\n",
      "28709/28709 [==============================] - 11s 390us/step - loss: 1.3395 - acc: 0.4912 - val_loss: 1.8275 - val_acc: 0.3357\n",
      "Epoch 29/1000\n",
      "28709/28709 [==============================] - 8s 267us/step - loss: 1.3266 - acc: 0.5012 - val_loss: 2.2547 - val_acc: 0.3287\n",
      "Epoch 30/1000\n",
      "28709/28709 [==============================] - 10s 338us/step - loss: 1.3120 - acc: 0.5084 - val_loss: 2.0221 - val_acc: 0.3366\n",
      "Epoch 31/1000\n",
      "28709/28709 [==============================] - 12s 430us/step - loss: 1.2925 - acc: 0.5150 - val_loss: 1.7864 - val_acc: 0.3546\n",
      "Epoch 32/1000\n",
      "28709/28709 [==============================] - 8s 266us/step - loss: 1.2678 - acc: 0.5234 - val_loss: 1.7157 - val_acc: 0.3736\n",
      "Epoch 33/1000\n",
      "28709/28709 [==============================] - 10s 351us/step - loss: 1.2436 - acc: 0.5301 - val_loss: 1.7673 - val_acc: 0.3730\n",
      "Epoch 34/1000\n",
      "28709/28709 [==============================] - 12s 411us/step - loss: 1.2238 - acc: 0.5440 - val_loss: 1.8785 - val_acc: 0.3405\n",
      "Epoch 35/1000\n",
      "28709/28709 [==============================] - 10s 343us/step - loss: 1.2016 - acc: 0.5523 - val_loss: 1.8328 - val_acc: 0.3394\n",
      "Epoch 36/1000\n",
      "28709/28709 [==============================] - 8s 265us/step - loss: 1.1842 - acc: 0.5621 - val_loss: 1.8889 - val_acc: 0.3398\n",
      "Epoch 37/1000\n",
      "28709/28709 [==============================] - 12s 413us/step - loss: 1.1959 - acc: 0.5578 - val_loss: 1.8611 - val_acc: 0.3464\n",
      "Epoch 38/1000\n",
      "28709/28709 [==============================] - 7s 244us/step - loss: 1.1558 - acc: 0.5712 - val_loss: 1.9518 - val_acc: 0.3631\n",
      "Epoch 39/1000\n",
      "28709/28709 [==============================] - 10s 356us/step - loss: 1.1530 - acc: 0.5781 - val_loss: 2.1293 - val_acc: 0.3035\n",
      "Epoch 40/1000\n",
      "28709/28709 [==============================] - 12s 407us/step - loss: 1.1314 - acc: 0.5821 - val_loss: 2.1902 - val_acc: 0.3534\n",
      "Epoch 41/1000\n",
      "28709/28709 [==============================] - 6s 202us/step - loss: 1.1272 - acc: 0.5835 - val_loss: 1.9492 - val_acc: 0.3560\n",
      "Epoch 42/1000\n",
      "28709/28709 [==============================] - 11s 397us/step - loss: 1.0876 - acc: 0.6000 - val_loss: 2.3007 - val_acc: 0.2557\n",
      "Epoch 43/1000\n",
      "28709/28709 [==============================] - 10s 352us/step - loss: 1.0892 - acc: 0.6010 - val_loss: 2.1127 - val_acc: 0.3083\n",
      "Epoch 44/1000\n",
      "28709/28709 [==============================] - 7s 239us/step - loss: 1.0693 - acc: 0.6071 - val_loss: 2.0670 - val_acc: 0.3681\n",
      "Epoch 45/1000\n",
      "28709/28709 [==============================] - 12s 406us/step - loss: 1.0459 - acc: 0.6141 - val_loss: 1.9811 - val_acc: 0.3496\n",
      "Epoch 46/1000\n",
      "28709/28709 [==============================] - 9s 300us/step - loss: 1.0391 - acc: 0.6178 - val_loss: 2.0605 - val_acc: 0.3326\n",
      "Epoch 47/1000\n",
      "28709/28709 [==============================] - 8s 290us/step - loss: 1.0069 - acc: 0.6310 - val_loss: 1.9634 - val_acc: 0.3651\n",
      "Epoch 48/1000\n",
      "28709/28709 [==============================] - 12s 414us/step - loss: 1.0090 - acc: 0.6295 - val_loss: 2.5140 - val_acc: 0.2702\n",
      "Epoch 49/1000\n",
      "28709/28709 [==============================] - 8s 293us/step - loss: 0.9801 - acc: 0.6415 - val_loss: 1.9772 - val_acc: 0.3858\n",
      "Epoch 50/1000\n",
      "28709/28709 [==============================] - 9s 296us/step - loss: 0.9682 - acc: 0.6436 - val_loss: 2.2032 - val_acc: 0.3614\n",
      "Epoch 51/1000\n",
      "28709/28709 [==============================] - 12s 427us/step - loss: 0.9482 - acc: 0.6520 - val_loss: 2.0039 - val_acc: 0.3617\n",
      "Epoch 52/1000\n",
      "28709/28709 [==============================] - 6s 217us/step - loss: 0.9200 - acc: 0.6574 - val_loss: 1.9210 - val_acc: 0.3567\n",
      "Epoch 53/1000\n",
      "28709/28709 [==============================] - 11s 383us/step - loss: 0.9185 - acc: 0.6611 - val_loss: 2.0975 - val_acc: 0.3157\n",
      "Epoch 54/1000\n",
      "28709/28709 [==============================] - 11s 394us/step - loss: 0.8872 - acc: 0.6730 - val_loss: 2.0770 - val_acc: 0.3614\n",
      "Epoch 55/1000\n",
      "28709/28709 [==============================] - 6s 212us/step - loss: 0.9058 - acc: 0.6730 - val_loss: 2.0838 - val_acc: 0.3595\n",
      "Epoch 56/1000\n",
      "28709/28709 [==============================] - 12s 427us/step - loss: 0.8755 - acc: 0.6821 - val_loss: 2.0829 - val_acc: 0.3593\n",
      "Epoch 57/1000\n",
      "28709/28709 [==============================] - 9s 298us/step - loss: 0.8511 - acc: 0.6855 - val_loss: 2.2225 - val_acc: 0.3252\n",
      "Epoch 58/1000\n",
      "28709/28709 [==============================] - 9s 314us/step - loss: 0.8424 - acc: 0.6929 - val_loss: 2.4886 - val_acc: 0.3329\n",
      "Epoch 59/1000\n",
      "28709/28709 [==============================] - 12s 409us/step - loss: 0.8278 - acc: 0.7001 - val_loss: 2.6405 - val_acc: 0.3298\n",
      "Epoch 60/1000\n",
      "28709/28709 [==============================] - 7s 238us/step - loss: 0.8446 - acc: 0.6956 - val_loss: 2.2511 - val_acc: 0.3582\n",
      "Epoch 61/1000\n",
      "28709/28709 [==============================] - 10s 357us/step - loss: 0.8228 - acc: 0.7028 - val_loss: 2.2296 - val_acc: 0.3814\n",
      "Epoch 62/1000\n",
      "28709/28709 [==============================] - 12s 403us/step - loss: 0.7924 - acc: 0.7084 - val_loss: 2.1651 - val_acc: 0.3443\n",
      "Epoch 63/1000\n",
      "28709/28709 [==============================] - 6s 205us/step - loss: 0.7683 - acc: 0.7171 - val_loss: 2.2413 - val_acc: 0.3577\n",
      "Epoch 64/1000\n",
      "28709/28709 [==============================] - 12s 407us/step - loss: 0.7973 - acc: 0.7193 - val_loss: 2.5073 - val_acc: 0.3761\n",
      "Epoch 65/1000\n",
      "28709/28709 [==============================] - 10s 354us/step - loss: 0.7550 - acc: 0.7276 - val_loss: 2.5977 - val_acc: 0.3290\n",
      "Epoch 66/1000\n",
      "28709/28709 [==============================] - 7s 260us/step - loss: 0.7553 - acc: 0.7260 - val_loss: 2.4788 - val_acc: 0.3684\n",
      "Epoch 67/1000\n",
      "28709/28709 [==============================] - 12s 426us/step - loss: 0.7248 - acc: 0.7407 - val_loss: 2.3735 - val_acc: 0.3352\n",
      "Epoch 68/1000\n",
      "28709/28709 [==============================] - 8s 296us/step - loss: 0.7210 - acc: 0.7416 - val_loss: 2.1070 - val_acc: 0.3931\n",
      "Epoch 69/1000\n",
      "28709/28709 [==============================] - 9s 306us/step - loss: 0.7119 - acc: 0.7418 - val_loss: 2.3200 - val_acc: 0.3776\n",
      "Epoch 70/1000\n",
      "28709/28709 [==============================] - 12s 424us/step - loss: 0.6812 - acc: 0.7544 - val_loss: 2.3148 - val_acc: 0.3947\n",
      "Epoch 71/1000\n",
      "28709/28709 [==============================] - 7s 227us/step - loss: 0.7062 - acc: 0.7540 - val_loss: 2.2062 - val_acc: 0.3756\n",
      "Epoch 72/1000\n",
      "28709/28709 [==============================] - 11s 371us/step - loss: 0.6607 - acc: 0.7619 - val_loss: 2.4479 - val_acc: 0.3766\n",
      "Epoch 73/1000\n",
      "28709/28709 [==============================] - 11s 392us/step - loss: 0.6473 - acc: 0.7698 - val_loss: 2.5507 - val_acc: 0.3660\n",
      "Epoch 74/1000\n",
      "28709/28709 [==============================] - 6s 219us/step - loss: 0.6529 - acc: 0.7775 - val_loss: 2.4860 - val_acc: 0.3591\n",
      "Epoch 75/1000\n",
      "28709/28709 [==============================] - 13s 442us/step - loss: 0.6520 - acc: 0.7674 - val_loss: 2.4706 - val_acc: 0.3908\n",
      "Epoch 76/1000\n",
      "28709/28709 [==============================] - 12s 423us/step - loss: 0.6227 - acc: 0.7746 - val_loss: 2.5037 - val_acc: 0.3574\n",
      "Epoch 77/1000\n",
      "28709/28709 [==============================] - 9s 329us/step - loss: 0.6003 - acc: 0.7838 - val_loss: 2.7297 - val_acc: 0.3532\n",
      "Epoch 78/1000\n",
      "28709/28709 [==============================] - 11s 396us/step - loss: 0.6453 - acc: 0.7808 - val_loss: 2.6504 - val_acc: 0.3802\n",
      "Epoch 79/1000\n",
      "28709/28709 [==============================] - 13s 455us/step - loss: 0.5750 - acc: 0.7947 - val_loss: 2.7722 - val_acc: 0.3592\n",
      "Epoch 80/1000\n",
      "28709/28709 [==============================] - 11s 372us/step - loss: 0.6233 - acc: 0.7795 - val_loss: 2.4159 - val_acc: 0.3974\n",
      "Epoch 81/1000\n",
      "28709/28709 [==============================] - 9s 308us/step - loss: 0.5583 - acc: 0.8027 - val_loss: 2.6308 - val_acc: 0.3607\n",
      "Epoch 82/1000\n",
      "28709/28709 [==============================] - 14s 490us/step - loss: 0.5456 - acc: 0.8092 - val_loss: 2.6482 - val_acc: 0.3600\n",
      "Epoch 83/1000\n",
      "28709/28709 [==============================] - 15s 519us/step - loss: 0.5793 - acc: 0.7944 - val_loss: 3.1115 - val_acc: 0.3132\n",
      "Epoch 84/1000\n",
      "28709/28709 [==============================] - 13s 437us/step - loss: 0.5268 - acc: 0.8214 - val_loss: 2.9344 - val_acc: 0.3609\n",
      "Epoch 85/1000\n",
      "28709/28709 [==============================] - 10s 333us/step - loss: 0.5943 - acc: 0.8038 - val_loss: 3.0800 - val_acc: 0.3453\n",
      "Epoch 86/1000\n",
      "28709/28709 [==============================] - 11s 390us/step - loss: 0.5208 - acc: 0.8218 - val_loss: 2.7470 - val_acc: 0.3460\n",
      "Epoch 87/1000\n",
      "28709/28709 [==============================] - 13s 456us/step - loss: 0.5037 - acc: 0.8242 - val_loss: 3.3760 - val_acc: 0.3687\n",
      "Epoch 88/1000\n",
      "28709/28709 [==============================] - 8s 274us/step - loss: 0.5291 - acc: 0.8334 - val_loss: 3.5818 - val_acc: 0.3458\n",
      "Epoch 89/1000\n",
      "28709/28709 [==============================] - 12s 424us/step - loss: 0.5442 - acc: 0.8270 - val_loss: 3.2020 - val_acc: 0.3568\n",
      "Epoch 90/1000\n",
      "28709/28709 [==============================] - 14s 496us/step - loss: 0.5029 - acc: 0.8470 - val_loss: 3.0583 - val_acc: 0.3408\n",
      "Epoch 91/1000\n",
      "28709/28709 [==============================] - 10s 350us/step - loss: 0.5140 - acc: 0.8242 - val_loss: 2.7605 - val_acc: 0.3762\n",
      "Epoch 92/1000\n",
      "28709/28709 [==============================] - 9s 329us/step - loss: 0.5057 - acc: 0.8275 - val_loss: 2.7315 - val_acc: 0.3979\n",
      "Epoch 93/1000\n",
      "28709/28709 [==============================] - 13s 448us/step - loss: 0.4736 - acc: 0.8355 - val_loss: 2.7605 - val_acc: 0.3623\n",
      "Epoch 94/1000\n",
      "28709/28709 [==============================] - 9s 317us/step - loss: 0.4719 - acc: 0.8391 - val_loss: 2.9605 - val_acc: 0.3964\n",
      "Epoch 95/1000\n",
      "28709/28709 [==============================] - 10s 341us/step - loss: 0.4327 - acc: 0.8583 - val_loss: 3.3846 - val_acc: 0.3812\n",
      "Epoch 96/1000\n",
      "28709/28709 [==============================] - 13s 462us/step - loss: 0.4833 - acc: 0.8538 - val_loss: 3.0838 - val_acc: 0.3922\n",
      "Epoch 97/1000\n",
      "28709/28709 [==============================] - 10s 361us/step - loss: 0.5278 - acc: 0.8391 - val_loss: 2.6403 - val_acc: 0.3991\n",
      "Epoch 98/1000\n",
      "28709/28709 [==============================] - 8s 289us/step - loss: 0.4649 - acc: 0.8643 - val_loss: 2.6396 - val_acc: 0.3809\n",
      "Epoch 99/1000\n",
      "28709/28709 [==============================] - 14s 477us/step - loss: 0.4171 - acc: 0.8611 - val_loss: 3.0466 - val_acc: 0.3690\n",
      "Epoch 100/1000\n",
      "28709/28709 [==============================] - 11s 368us/step - loss: 0.6073 - acc: 0.8247 - val_loss: 2.9174 - val_acc: 0.3816\n",
      "Epoch 101/1000\n",
      "28709/28709 [==============================] - 8s 296us/step - loss: 0.3081 - acc: 0.9047 - val_loss: 3.2060 - val_acc: 0.3886\n",
      "Epoch 102/1000\n",
      "28709/28709 [==============================] - 13s 449us/step - loss: 0.5282 - acc: 0.8490 - val_loss: 3.3425 - val_acc: 0.3444\n",
      "Epoch 103/1000\n",
      "28709/28709 [==============================] - 12s 409us/step - loss: 0.3820 - acc: 0.8859 - val_loss: 3.7014 - val_acc: 0.3748\n",
      "Epoch 104/1000\n",
      "28709/28709 [==============================] - 8s 284us/step - loss: 0.5041 - acc: 0.8546 - val_loss: 2.7969 - val_acc: 0.4056\n",
      "Epoch 105/1000\n",
      "28709/28709 [==============================] - 14s 476us/step - loss: 0.3853 - acc: 0.8949 - val_loss: 3.1856 - val_acc: 0.3687\n",
      "Epoch 106/1000\n",
      "28709/28709 [==============================] - 14s 472us/step - loss: 0.6989 - acc: 0.8233 - val_loss: 2.7304 - val_acc: 0.3886\n",
      "Epoch 107/1000\n",
      "28709/28709 [==============================] - 8s 262us/step - loss: 0.2670 - acc: 0.9293 - val_loss: 2.9971 - val_acc: 0.3989\n",
      "Epoch 108/1000\n",
      "28709/28709 [==============================] - 12s 434us/step - loss: 0.5742 - acc: 0.8538 - val_loss: 2.9991 - val_acc: 0.3607\n",
      "Epoch 109/1000\n",
      "28709/28709 [==============================] - 14s 473us/step - loss: 0.2882 - acc: 0.9143 - val_loss: 3.3181 - val_acc: 0.3549\n",
      "Epoch 110/1000\n",
      "28709/28709 [==============================] - 8s 272us/step - loss: 0.6385 - acc: 0.8389 - val_loss: 2.8371 - val_acc: 0.3991\n",
      "Epoch 111/1000\n",
      "28709/28709 [==============================] - 12s 429us/step - loss: 0.3481 - acc: 0.9148 - val_loss: 3.0215 - val_acc: 0.4173\n",
      "Epoch 112/1000\n",
      "28709/28709 [==============================] - 15s 520us/step - loss: 0.4999 - acc: 0.8674 - val_loss: 2.7765 - val_acc: 0.3942\n",
      "Epoch 113/1000\n",
      "28709/28709 [==============================] - 9s 317us/step - loss: 0.5180 - acc: 0.8804 - val_loss: 2.8024 - val_acc: 0.4023\n",
      "Epoch 114/1000\n",
      "28709/28709 [==============================] - 11s 373us/step - loss: 0.3544 - acc: 0.9045 - val_loss: 3.1727 - val_acc: 0.3982\n",
      "Epoch 115/1000\n",
      "28709/28709 [==============================] - 14s 495us/step - loss: 0.5024 - acc: 0.8737 - val_loss: 2.8268 - val_acc: 0.4094\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 116/1000\n",
      "28709/28709 [==============================] - 9s 322us/step - loss: 0.3607 - acc: 0.9056 - val_loss: 3.1450 - val_acc: 0.3887\n",
      "Epoch 117/1000\n",
      "28709/28709 [==============================] - 10s 354us/step - loss: 0.5188 - acc: 0.8650 - val_loss: 2.8156 - val_acc: 0.4050\n",
      "Epoch 118/1000\n",
      "28709/28709 [==============================] - 15s 508us/step - loss: 0.4550 - acc: 0.8961 - val_loss: 2.9666 - val_acc: 0.3906\n",
      "Epoch 119/1000\n",
      "28709/28709 [==============================] - 8s 290us/step - loss: 0.6697 - acc: 0.8781 - val_loss: 2.9369 - val_acc: 0.3968\n",
      "Epoch 120/1000\n",
      "28709/28709 [==============================] - 11s 376us/step - loss: 0.3416 - acc: 0.9230 - val_loss: 2.9186 - val_acc: 0.3997\n",
      "Epoch 121/1000\n",
      "28709/28709 [==============================] - 13s 468us/step - loss: 0.3975 - acc: 0.9036 - val_loss: 3.2991 - val_acc: 0.4050\n",
      "Epoch 122/1000\n",
      "28709/28709 [==============================] - 9s 325us/step - loss: 0.4909 - acc: 0.8977 - val_loss: 3.2838 - val_acc: 0.4037\n",
      "Epoch 123/1000\n",
      "28709/28709 [==============================] - 10s 350us/step - loss: 0.5648 - acc: 0.8865 - val_loss: 2.9154 - val_acc: 0.3807\n",
      "Epoch 124/1000\n",
      "28709/28709 [==============================] - 13s 470us/step - loss: 0.4785 - acc: 0.9121 - val_loss: 2.8325 - val_acc: 0.3464\n",
      "Epoch 125/1000\n",
      "28709/28709 [==============================] - 10s 338us/step - loss: 0.1667 - acc: 0.9560 - val_loss: 3.8445 - val_acc: 0.3518\n",
      "Epoch 126/1000\n",
      "28709/28709 [==============================] - 9s 321us/step - loss: 0.3958 - acc: 0.9021 - val_loss: 4.7536 - val_acc: 0.3836\n",
      "Epoch 127/1000\n",
      "28709/28709 [==============================] - 14s 472us/step - loss: 0.6368 - acc: 0.8746 - val_loss: 3.5131 - val_acc: 0.4067\n",
      "Epoch 128/1000\n",
      "28709/28709 [==============================] - 10s 346us/step - loss: 0.2789 - acc: 0.9258 - val_loss: 6.3866 - val_acc: 0.3287\n",
      "Epoch 129/1000\n",
      "28709/28709 [==============================] - 9s 313us/step - loss: 0.5683 - acc: 0.8897 - val_loss: 3.7780 - val_acc: 0.4039\n",
      "Epoch 130/1000\n",
      "28709/28709 [==============================] - 14s 496us/step - loss: 0.5302 - acc: 0.8898 - val_loss: 3.7757 - val_acc: 0.4085\n",
      "Epoch 131/1000\n",
      "28709/28709 [==============================] - 12s 409us/step - loss: 0.2433 - acc: 0.9310 - val_loss: 5.8105 - val_acc: 0.2738\n",
      "Epoch 132/1000\n",
      "28709/28709 [==============================] - 8s 284us/step - loss: 0.4317 - acc: 0.9145 - val_loss: 3.8133 - val_acc: 0.3932\n",
      "Epoch 133/1000\n",
      "28709/28709 [==============================] - 14s 504us/step - loss: 0.6200 - acc: 0.8933 - val_loss: 3.7408 - val_acc: 0.4074\n",
      "Epoch 134/1000\n",
      "28709/28709 [==============================] - 10s 351us/step - loss: 0.2462 - acc: 0.9359 - val_loss: 4.1693 - val_acc: 0.4046\n",
      "Epoch 135/1000\n",
      "28709/28709 [==============================] - 10s 355us/step - loss: 0.7945 - acc: 0.8705 - val_loss: 3.8473 - val_acc: 0.4028\n",
      "Epoch 136/1000\n",
      "28709/28709 [==============================] - 13s 459us/step - loss: 0.4180 - acc: 0.9270 - val_loss: 5.0405 - val_acc: 0.3262\n",
      "Epoch 137/1000\n",
      "28709/28709 [==============================] - 11s 372us/step - loss: 0.7064 - acc: 0.8773 - val_loss: 4.0330 - val_acc: 0.3836\n",
      "Epoch 138/1000\n",
      "28709/28709 [==============================] - 9s 327us/step - loss: 0.2671 - acc: 0.9308 - val_loss: 6.3897 - val_acc: 0.3457\n",
      "Epoch 139/1000\n",
      "28709/28709 [==============================] - 13s 459us/step - loss: 0.5364 - acc: 0.9023 - val_loss: 3.9702 - val_acc: 0.3854\n",
      "Epoch 140/1000\n",
      "28709/28709 [==============================] - 13s 440us/step - loss: 0.6602 - acc: 0.8804 - val_loss: 3.7772 - val_acc: 0.3997\n",
      "Epoch 141/1000\n",
      "28709/28709 [==============================] - 8s 270us/step - loss: 0.3390 - acc: 0.9280 - val_loss: 3.9667 - val_acc: 0.4050\n",
      "Epoch 142/1000\n",
      "28709/28709 [==============================] - 13s 464us/step - loss: 0.4609 - acc: 0.9095 - val_loss: 3.8057 - val_acc: 0.4060\n",
      "Epoch 143/1000\n",
      "28709/28709 [==============================] - 11s 399us/step - loss: 0.5419 - acc: 0.9012 - val_loss: 4.1455 - val_acc: 0.3890\n",
      "Epoch 144/1000\n",
      "28709/28709 [==============================] - 8s 263us/step - loss: 0.2025 - acc: 0.9472 - val_loss: 4.3953 - val_acc: 0.3531\n",
      "Epoch 145/1000\n",
      "28709/28709 [==============================] - 14s 472us/step - loss: 0.5520 - acc: 0.9021 - val_loss: 4.1075 - val_acc: 0.3935\n",
      "Epoch 146/1000\n",
      "28709/28709 [==============================] - 11s 374us/step - loss: 0.4572 - acc: 0.9126 - val_loss: 4.1546 - val_acc: 0.4052\n",
      "Epoch 147/1000\n",
      "28709/28709 [==============================] - 8s 293us/step - loss: 0.6958 - acc: 0.8941 - val_loss: 3.9164 - val_acc: 0.4052\n",
      "Epoch 148/1000\n",
      "28709/28709 [==============================] - 14s 478us/step - loss: 0.3532 - acc: 0.9274 - val_loss: 4.4756 - val_acc: 0.4115\n",
      "Epoch 149/1000\n",
      "28709/28709 [==============================] - 9s 298us/step - loss: 0.9760 - acc: 0.8398 - val_loss: 4.0907 - val_acc: 0.3960\n",
      "Epoch 150/1000\n",
      "28709/28709 [==============================] - 11s 372us/step - loss: 0.2544 - acc: 0.9344 - val_loss: 4.2690 - val_acc: 0.4034\n",
      "Epoch 151/1000\n",
      "28709/28709 [==============================] - 13s 464us/step - loss: 0.5531 - acc: 0.8975 - val_loss: 3.8572 - val_acc: 0.4025\n",
      "Epoch 152/1000\n",
      "28709/28709 [==============================] - 8s 287us/step - loss: 0.4689 - acc: 0.9082 - val_loss: 4.2175 - val_acc: 0.4032\n",
      "Epoch 153/1000\n",
      "28709/28709 [==============================] - 11s 372us/step - loss: 0.6087 - acc: 0.9022 - val_loss: 3.8944 - val_acc: 0.3871\n",
      "Epoch 154/1000\n",
      "28709/28709 [==============================] - 13s 463us/step - loss: 0.2315 - acc: 0.9469 - val_loss: 4.4711 - val_acc: 0.3900\n",
      "Epoch 155/1000\n",
      "28709/28709 [==============================] - 7s 230us/step - loss: 0.5937 - acc: 0.8979 - val_loss: 3.9603 - val_acc: 0.4046\n",
      "Epoch 156/1000\n",
      "28709/28709 [==============================] - 12s 429us/step - loss: 0.3417 - acc: 0.9187 - val_loss: 3.8717 - val_acc: 0.4071\n",
      "Epoch 157/1000\n",
      "28709/28709 [==============================] - 11s 384us/step - loss: 1.0011 - acc: 0.8694 - val_loss: 4.0397 - val_acc: 0.3929\n",
      "Epoch 158/1000\n",
      "28709/28709 [==============================] - 8s 274us/step - loss: 0.5273 - acc: 0.9189 - val_loss: 2.9686 - val_acc: 0.3493\n",
      "Epoch 159/1000\n",
      "28709/28709 [==============================] - 13s 458us/step - loss: 0.1056 - acc: 0.9747 - val_loss: 4.5681 - val_acc: 0.4059\n",
      "Epoch 160/1000\n",
      "28709/28709 [==============================] - 10s 341us/step - loss: 0.8425 - acc: 0.8836 - val_loss: 3.9378 - val_acc: 0.4059\n",
      "Epoch 161/1000\n",
      "28709/28709 [==============================] - 9s 328us/step - loss: 0.6327 - acc: 0.8807 - val_loss: 3.5116 - val_acc: 0.3938\n",
      "Epoch 162/1000\n",
      "28709/28709 [==============================] - 14s 488us/step - loss: 1.1206 - acc: 0.8598 - val_loss: 2.9072 - val_acc: 0.4000\n",
      "Epoch 163/1000\n",
      "28709/28709 [==============================] - 9s 312us/step - loss: 0.0686 - acc: 0.9869 - val_loss: 5.3145 - val_acc: 0.3979\n",
      "Epoch 164/1000\n",
      "28709/28709 [==============================] - 10s 343us/step - loss: 0.6846 - acc: 0.8962 - val_loss: 4.1545 - val_acc: 0.3867\n",
      "Epoch 165/1000\n",
      "28709/28709 [==============================] - 13s 458us/step - loss: 0.6587 - acc: 0.8819 - val_loss: 4.0133 - val_acc: 0.4011\n",
      "Epoch 166/1000\n",
      "28709/28709 [==============================] - 9s 301us/step - loss: 1.1465 - acc: 0.8464 - val_loss: 3.8862 - val_acc: 0.3995\n",
      "Epoch 167/1000\n",
      "28709/28709 [==============================] - 11s 385us/step - loss: 0.2008 - acc: 0.9698 - val_loss: 5.4458 - val_acc: 0.3042\n",
      "Epoch 168/1000\n",
      "28709/28709 [==============================] - 13s 465us/step - loss: 0.2683 - acc: 0.9417 - val_loss: 4.6696 - val_acc: 0.3968\n",
      "Epoch 169/1000\n",
      "28709/28709 [==============================] - 7s 246us/step - loss: 0.7141 - acc: 0.8933 - val_loss: 4.2682 - val_acc: 0.4052\n",
      "Epoch 170/1000\n",
      "28709/28709 [==============================] - 11s 400us/step - loss: 0.2629 - acc: 0.9393 - val_loss: 4.4282 - val_acc: 0.3982\n",
      "Epoch 171/1000\n",
      "28709/28709 [==============================] - 11s 381us/step - loss: 0.8508 - acc: 0.8782 - val_loss: 3.8353 - val_acc: 0.3997\n",
      "Epoch 172/1000\n",
      "28709/28709 [==============================] - 8s 277us/step - loss: 0.8567 - acc: 0.8839 - val_loss: 3.2730 - val_acc: 0.3996\n",
      "Epoch 173/1000\n",
      "28709/28709 [==============================] - 13s 464us/step - loss: 0.6128 - acc: 0.9266 - val_loss: 4.5633 - val_acc: 0.2721\n",
      "Epoch 174/1000\n",
      "28709/28709 [==============================] - 10s 348us/step - loss: 0.1557 - acc: 0.9638 - val_loss: 4.4712 - val_acc: 0.4011\n",
      "Epoch 175/1000\n",
      "28709/28709 [==============================] - 9s 320us/step - loss: 0.2423 - acc: 0.9418 - val_loss: 4.9104 - val_acc: 0.3999\n",
      "Epoch 176/1000\n",
      "28709/28709 [==============================] - 13s 454us/step - loss: 1.0400 - acc: 0.8631 - val_loss: 4.5366 - val_acc: 0.4081\n",
      "Epoch 177/1000\n",
      "28709/28709 [==============================] - 9s 312us/step - loss: 1.3944 - acc: 0.8547 - val_loss: 3.5927 - val_acc: 0.4049\n",
      "Epoch 178/1000\n",
      "28709/28709 [==============================] - 10s 344us/step - loss: 0.6226 - acc: 0.9048 - val_loss: 3.3436 - val_acc: 0.3627\n",
      "Epoch 179/1000\n",
      "28709/28709 [==============================] - 13s 454us/step - loss: 0.0799 - acc: 0.9837 - val_loss: 4.3218 - val_acc: 0.4074\n",
      "Epoch 180/1000\n",
      "28709/28709 [==============================] - 6s 218us/step - loss: 0.8891 - acc: 0.8830 - val_loss: 3.9818 - val_acc: 0.3890\n",
      "Epoch 181/1000\n",
      "28709/28709 [==============================] - 14s 481us/step - loss: 0.2309 - acc: 0.9596 - val_loss: 10.2364 - val_acc: 0.2286\n",
      "Epoch 182/1000\n",
      "28709/28709 [==============================] - 14s 488us/step - loss: 0.7793 - acc: 0.9014 - val_loss: 4.7058 - val_acc: 0.4077\n",
      "Epoch 183/1000\n",
      "28709/28709 [==============================] - 9s 325us/step - loss: 1.0004 - acc: 0.8661 - val_loss: 3.7470 - val_acc: 0.4088\n",
      "Epoch 184/1000\n",
      "28709/28709 [==============================] - 9s 329us/step - loss: 0.8737 - acc: 0.8902 - val_loss: 3.3212 - val_acc: 0.3956\n",
      "Epoch 185/1000\n",
      "28709/28709 [==============================] - 13s 469us/step - loss: 0.4954 - acc: 0.9395 - val_loss: 5.9637 - val_acc: 0.2270\n",
      "Epoch 186/1000\n",
      "28709/28709 [==============================] - 10s 361us/step - loss: 0.3800 - acc: 0.9245 - val_loss: 4.7520 - val_acc: 0.3939\n",
      "Epoch 187/1000\n",
      "28709/28709 [==============================] - 9s 314us/step - loss: 0.6223 - acc: 0.9019 - val_loss: 4.5208 - val_acc: 0.3965\n",
      "Epoch 188/1000\n",
      "28709/28709 [==============================] - 13s 456us/step - loss: 0.9886 - acc: 0.8777 - val_loss: 3.5301 - val_acc: 0.3935\n",
      "Epoch 189/1000\n",
      "28709/28709 [==============================] - 9s 312us/step - loss: 0.0361 - acc: 0.9947 - val_loss: 4.6497 - val_acc: 0.4064\n",
      "Epoch 190/1000\n",
      "28709/28709 [==============================] - 10s 336us/step - loss: 1.1976 - acc: 0.8747 - val_loss: 4.0686 - val_acc: 0.4084\n",
      "Epoch 191/1000\n",
      "28709/28709 [==============================] - 13s 470us/step - loss: 0.7536 - acc: 0.8858 - val_loss: 3.5682 - val_acc: 0.3943\n",
      "Epoch 192/1000\n",
      "28709/28709 [==============================] - 8s 264us/step - loss: 0.0361 - acc: 0.9946 - val_loss: 5.1613 - val_acc: 0.3932\n",
      "Epoch 193/1000\n",
      "28709/28709 [==============================] - 11s 382us/step - loss: 0.7947 - acc: 0.8860 - val_loss: 4.4520 - val_acc: 0.3979\n",
      "Epoch 194/1000\n",
      "28709/28709 [==============================] - 13s 446us/step - loss: 0.7256 - acc: 0.8988 - val_loss: 4.8571 - val_acc: 0.4030\n",
      "Epoch 195/1000\n",
      "28709/28709 [==============================] - 6s 226us/step - loss: 1.1928 - acc: 0.8622 - val_loss: 3.5065 - val_acc: 0.3975\n",
      "Epoch 196/1000\n",
      "28709/28709 [==============================] - 13s 440us/step - loss: 0.0513 - acc: 0.9900 - val_loss: 4.8776 - val_acc: 0.3138\n",
      "Epoch 197/1000\n",
      "28709/28709 [==============================] - 11s 374us/step - loss: 0.2435 - acc: 0.9421 - val_loss: 4.4802 - val_acc: 0.3823\n",
      "Epoch 198/1000\n",
      "28709/28709 [==============================] - 8s 292us/step - loss: 0.5526 - acc: 0.9029 - val_loss: 4.1658 - val_acc: 0.3844\n",
      "Epoch 199/1000\n",
      "28709/28709 [==============================] - 13s 450us/step - loss: 1.2092 - acc: 0.8599 - val_loss: 3.4984 - val_acc: 0.4048\n",
      "Epoch 200/1000\n",
      "28709/28709 [==============================] - 9s 327us/step - loss: 0.0405 - acc: 0.9932 - val_loss: 4.6182 - val_acc: 0.4034\n",
      "Epoch 201/1000\n",
      "28709/28709 [==============================] - 9s 323us/step - loss: 0.7103 - acc: 0.8995 - val_loss: 4.1638 - val_acc: 0.3992\n",
      "Epoch 202/1000\n",
      "28709/28709 [==============================] - 13s 451us/step - loss: 0.5447 - acc: 0.9112 - val_loss: 3.5205 - val_acc: 0.3993\n",
      "Epoch 203/1000\n",
      "28709/28709 [==============================] - 7s 245us/step - loss: 0.0283 - acc: 0.9956 - val_loss: 5.5685 - val_acc: 0.3625\n",
      "Epoch 204/1000\n",
      "28709/28709 [==============================] - 11s 395us/step - loss: 1.1253 - acc: 0.8525 - val_loss: 4.9032 - val_acc: 0.3971\n",
      "Epoch 205/1000\n",
      "28709/28709 [==============================] - 12s 434us/step - loss: 0.6699 - acc: 0.9102 - val_loss: 3.6479 - val_acc: 0.4004\n",
      "Epoch 206/1000\n",
      "28709/28709 [==============================] - 6s 226us/step - loss: 0.1037 - acc: 0.9850 - val_loss: 7.1967 - val_acc: 0.2763\n",
      "Epoch 207/1000\n",
      "28709/28709 [==============================] - 13s 444us/step - loss: 0.3955 - acc: 0.9305 - val_loss: 5.1750 - val_acc: 0.4010\n",
      "Epoch 208/1000\n",
      "28709/28709 [==============================] - 10s 360us/step - loss: 0.8418 - acc: 0.8770 - val_loss: 3.9996 - val_acc: 0.3981\n",
      "Epoch 209/1000\n",
      "28709/28709 [==============================] - 7s 258us/step - loss: 0.6458 - acc: 0.9046 - val_loss: 3.3205 - val_acc: 0.3953\n",
      "Epoch 210/1000\n",
      "28709/28709 [==============================] - 14s 475us/step - loss: 0.0330 - acc: 0.9957 - val_loss: 4.9566 - val_acc: 0.3844\n",
      "Epoch 211/1000\n",
      "28709/28709 [==============================] - 10s 335us/step - loss: 0.8805 - acc: 0.8807 - val_loss: 5.0401 - val_acc: 0.4066\n",
      "Epoch 212/1000\n",
      "28709/28709 [==============================] - 9s 306us/step - loss: 1.0452 - acc: 0.8682 - val_loss: 3.2305 - val_acc: 0.3984\n",
      "Epoch 213/1000\n",
      "28709/28709 [==============================] - 13s 461us/step - loss: 0.0589 - acc: 0.9887 - val_loss: 4.9866 - val_acc: 0.3908\n",
      "Epoch 214/1000\n",
      "28709/28709 [==============================] - 9s 302us/step - loss: 0.6724 - acc: 0.8976 - val_loss: 4.8024 - val_acc: 0.4045\n",
      "Epoch 215/1000\n",
      "28709/28709 [==============================] - 10s 331us/step - loss: 0.7346 - acc: 0.8924 - val_loss: 3.4774 - val_acc: 0.3940\n",
      "Epoch 216/1000\n",
      "28709/28709 [==============================] - 13s 443us/step - loss: 0.0382 - acc: 0.9935 - val_loss: 8.0659 - val_acc: 0.2685\n",
      "Epoch 217/1000\n",
      "28709/28709 [==============================] - 8s 263us/step - loss: 0.5431 - acc: 0.9125 - val_loss: 4.3211 - val_acc: 0.4020\n",
      "Epoch 218/1000\n",
      "28709/28709 [==============================] - 11s 391us/step - loss: 0.5671 - acc: 0.8999 - val_loss: 3.4937 - val_acc: 0.4010\n",
      "Epoch 219/1000\n",
      "28709/28709 [==============================] - 13s 461us/step - loss: 0.0245 - acc: 0.9961 - val_loss: 5.0933 - val_acc: 0.3918\n",
      "Epoch 220/1000\n",
      "28709/28709 [==============================] - 7s 248us/step - loss: 0.5821 - acc: 0.9033 - val_loss: 5.1099 - val_acc: 0.3984\n",
      "Epoch 221/1000\n",
      "28709/28709 [==============================] - 12s 403us/step - loss: 0.9199 - acc: 0.8739 - val_loss: 3.4673 - val_acc: 0.3988\n",
      "Epoch 222/1000\n",
      "28709/28709 [==============================] - 12s 418us/step - loss: 0.0316 - acc: 0.9942 - val_loss: 6.6131 - val_acc: 0.2834\n",
      "Epoch 223/1000\n",
      "28709/28709 [==============================] - 7s 247us/step - loss: 0.5127 - acc: 0.9184 - val_loss: 4.4538 - val_acc: 0.3991\n",
      "Epoch 224/1000\n",
      "28709/28709 [==============================] - 13s 448us/step - loss: 1.0267 - acc: 0.8765 - val_loss: 3.5649 - val_acc: 0.3977\n",
      "Epoch 225/1000\n",
      "28709/28709 [==============================] - 10s 354us/step - loss: 0.0641 - acc: 0.9871 - val_loss: 7.3380 - val_acc: 0.3226\n",
      "Epoch 226/1000\n",
      "28709/28709 [==============================] - 8s 289us/step - loss: 0.5545 - acc: 0.9006 - val_loss: 4.5256 - val_acc: 0.4080\n",
      "Epoch 227/1000\n",
      "28709/28709 [==============================] - 13s 460us/step - loss: 0.5011 - acc: 0.9184 - val_loss: 4.9063 - val_acc: 0.4043\n",
      "Epoch 228/1000\n",
      "28709/28709 [==============================] - 8s 277us/step - loss: 0.6652 - acc: 0.9077 - val_loss: 3.7306 - val_acc: 0.4023\n",
      "Epoch 229/1000\n",
      "28709/28709 [==============================] - 10s 357us/step - loss: 0.4292 - acc: 0.9377 - val_loss: 3.4515 - val_acc: 0.3897\n",
      "Epoch 230/1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "28709/28709 [==============================] - 13s 444us/step - loss: 0.0738 - acc: 0.9823 - val_loss: 4.9905 - val_acc: 0.4023\n",
      "Epoch 231/1000\n",
      "28709/28709 [==============================] - 6s 216us/step - loss: 0.8417 - acc: 0.8871 - val_loss: 4.1875 - val_acc: 0.4028\n",
      "Epoch 232/1000\n",
      "28709/28709 [==============================] - 13s 437us/step - loss: 0.6726 - acc: 0.8846 - val_loss: 3.3744 - val_acc: 0.4009\n",
      "Epoch 233/1000\n",
      "28709/28709 [==============================] - 11s 388us/step - loss: 0.0338 - acc: 0.9953 - val_loss: 5.0424 - val_acc: 0.3912\n",
      "Epoch 234/1000\n",
      "28709/28709 [==============================] - 8s 278us/step - loss: 0.7612 - acc: 0.9011 - val_loss: 4.9572 - val_acc: 0.3915\n",
      "Epoch 235/1000\n",
      "28709/28709 [==============================] - 14s 494us/step - loss: 1.0426 - acc: 0.8809 - val_loss: 3.8282 - val_acc: 0.3945\n",
      "Epoch 236/1000\n",
      "28709/28709 [==============================] - 10s 332us/step - loss: 0.1166 - acc: 0.9823 - val_loss: 10.1754 - val_acc: 0.2006\n",
      "Epoch 237/1000\n",
      "28709/28709 [==============================] - 10s 345us/step - loss: 0.8616 - acc: 0.8823 - val_loss: 4.9683 - val_acc: 0.3802\n",
      "Epoch 238/1000\n",
      "28709/28709 [==============================] - 13s 456us/step - loss: 0.4206 - acc: 0.9302 - val_loss: 4.2794 - val_acc: 0.4028\n",
      "Epoch 239/1000\n",
      "28709/28709 [==============================] - 7s 230us/step - loss: 1.1464 - acc: 0.8697 - val_loss: 3.7411 - val_acc: 0.3918\n",
      "Epoch 240/1000\n",
      "28709/28709 [==============================] - 12s 430us/step - loss: 0.7633 - acc: 0.9065 - val_loss: 3.2608 - val_acc: 0.3701\n",
      "Epoch 241/1000\n",
      "28709/28709 [==============================] - 12s 403us/step - loss: 0.0746 - acc: 0.9834 - val_loss: 4.4824 - val_acc: 0.4042\n",
      "Epoch 242/1000\n",
      "28709/28709 [==============================] - 7s 238us/step - loss: 0.5726 - acc: 0.9115 - val_loss: 4.8051 - val_acc: 0.3995\n",
      "Epoch 243/1000\n",
      "28709/28709 [==============================] - 13s 460us/step - loss: 0.7776 - acc: 0.8869 - val_loss: 3.3411 - val_acc: 0.4014\n",
      "Epoch 244/1000\n",
      "28709/28709 [==============================] - 9s 311us/step - loss: 0.0323 - acc: 0.9953 - val_loss: 5.2907 - val_acc: 0.3899\n",
      "Epoch 245/1000\n",
      "28709/28709 [==============================] - 9s 324us/step - loss: 0.8208 - acc: 0.8826 - val_loss: 3.8759 - val_acc: 0.3989\n",
      "Epoch 246/1000\n",
      "28709/28709 [==============================] - 13s 466us/step - loss: 0.1721 - acc: 0.9779 - val_loss: 10.0178 - val_acc: 0.3044\n",
      "Epoch 247/1000\n",
      "28709/28709 [==============================] - 7s 250us/step - loss: 1.2478 - acc: 0.8727 - val_loss: 5.2146 - val_acc: 0.3915\n",
      "Epoch 248/1000\n",
      "28709/28709 [==============================] - 11s 390us/step - loss: 0.3942 - acc: 0.9386 - val_loss: 5.0912 - val_acc: 0.3981\n",
      "Epoch 249/1000\n",
      "28709/28709 [==============================] - 13s 447us/step - loss: 0.9168 - acc: 0.8865 - val_loss: 3.4865 - val_acc: 0.3879\n",
      "Epoch 250/1000\n",
      "28709/28709 [==============================] - 7s 249us/step - loss: 0.0410 - acc: 0.9940 - val_loss: 5.9913 - val_acc: 0.3476\n",
      "Epoch 251/1000\n",
      "28709/28709 [==============================] - 11s 390us/step - loss: 0.5095 - acc: 0.9125 - val_loss: 5.0289 - val_acc: 0.4059\n",
      "Epoch 252/1000\n",
      "28709/28709 [==============================] - 13s 457us/step - loss: 0.7735 - acc: 0.8928 - val_loss: 3.6368 - val_acc: 0.3986\n",
      "Epoch 253/1000\n",
      "28709/28709 [==============================] - 7s 229us/step - loss: 0.0297 - acc: 0.9946 - val_loss: 5.2646 - val_acc: 0.3871\n",
      "Epoch 254/1000\n",
      "28709/28709 [==============================] - 12s 433us/step - loss: 0.8553 - acc: 0.8955 - val_loss: 4.4758 - val_acc: 0.4006\n",
      "Epoch 255/1000\n",
      "28709/28709 [==============================] - 11s 379us/step - loss: 0.8432 - acc: 0.8946 - val_loss: 3.6776 - val_acc: 0.4034\n",
      "Epoch 256/1000\n",
      "28709/28709 [==============================] - 7s 252us/step - loss: 0.5025 - acc: 0.9451 - val_loss: 7.1544 - val_acc: 0.3152\n",
      "Epoch 257/1000\n",
      "28709/28709 [==============================] - 13s 442us/step - loss: 0.2338 - acc: 0.9554 - val_loss: 5.2626 - val_acc: 0.3821\n",
      "Epoch 258/1000\n",
      "28709/28709 [==============================] - 11s 396us/step - loss: 0.3701 - acc: 0.9279 - val_loss: 5.3030 - val_acc: 0.4042\n",
      "Epoch 259/1000\n",
      "28709/28709 [==============================] - 7s 255us/step - loss: 1.0905 - acc: 0.8834 - val_loss: 3.7559 - val_acc: 0.4050\n",
      "Epoch 260/1000\n",
      "28709/28709 [==============================] - 13s 458us/step - loss: 0.0259 - acc: 0.9959 - val_loss: 9.9084 - val_acc: 0.2034\n",
      "Epoch 261/1000\n",
      "28709/28709 [==============================] - 10s 336us/step - loss: 1.0149 - acc: 0.8703 - val_loss: 4.0701 - val_acc: 0.3950\n",
      "Epoch 262/1000\n",
      "28709/28709 [==============================] - 9s 308us/step - loss: 1.2399 - acc: 0.8689 - val_loss: 3.6085 - val_acc: 0.4006\n",
      "Epoch 263/1000\n",
      "28709/28709 [==============================] - 13s 452us/step - loss: 0.0379 - acc: 0.9935 - val_loss: 4.4998 - val_acc: 0.3981\n",
      "Epoch 264/1000\n",
      "28709/28709 [==============================] - 8s 269us/step - loss: 1.6059 - acc: 0.8423 - val_loss: 3.8800 - val_acc: 0.3943\n",
      "Epoch 265/1000\n",
      "28709/28709 [==============================] - 11s 373us/step - loss: 0.0256 - acc: 0.9954 - val_loss: 5.2950 - val_acc: 0.4007\n",
      "Epoch 266/1000\n",
      "28709/28709 [==============================] - 13s 441us/step - loss: 0.7140 - acc: 0.9097 - val_loss: 5.4561 - val_acc: 0.3982\n",
      "Epoch 267/1000\n",
      "28709/28709 [==============================] - 6s 225us/step - loss: 1.4861 - acc: 0.8575 - val_loss: 3.8085 - val_acc: 0.3961\n",
      "Epoch 268/1000\n",
      "28709/28709 [==============================] - 13s 440us/step - loss: 0.0283 - acc: 0.9952 - val_loss: 5.2071 - val_acc: 0.3943\n",
      "Epoch 269/1000\n",
      "28709/28709 [==============================] - 12s 420us/step - loss: 0.9277 - acc: 0.8902 - val_loss: 4.6730 - val_acc: 0.3931\n",
      "Epoch 270/1000\n",
      "28709/28709 [==============================] - 7s 233us/step - loss: 0.6147 - acc: 0.9401 - val_loss: 5.5024 - val_acc: 0.3510\n",
      "Epoch 271/1000\n",
      "28709/28709 [==============================] - 13s 457us/step - loss: 0.2214 - acc: 0.9498 - val_loss: 5.6714 - val_acc: 0.3965\n",
      "Epoch 272/1000\n",
      "28709/28709 [==============================] - 11s 370us/step - loss: 0.5470 - acc: 0.9208 - val_loss: 3.8101 - val_acc: 0.4035\n",
      "Epoch 273/1000\n",
      "28709/28709 [==============================] - 8s 274us/step - loss: 0.0208 - acc: 0.9961 - val_loss: 6.8110 - val_acc: 0.2993\n",
      "Epoch 274/1000\n",
      "28709/28709 [==============================] - 13s 445us/step - loss: 0.7005 - acc: 0.8931 - val_loss: 4.3903 - val_acc: 0.3963\n",
      "Epoch 275/1000\n",
      "28709/28709 [==============================] - 10s 360us/step - loss: 1.0037 - acc: 0.8929 - val_loss: 3.7965 - val_acc: 0.4035\n",
      "Epoch 276/1000\n",
      "28709/28709 [==============================] - 9s 306us/step - loss: 0.0289 - acc: 0.9936 - val_loss: 5.7252 - val_acc: 0.3479\n",
      "Epoch 277/1000\n",
      "28709/28709 [==============================] - 14s 473us/step - loss: 0.5574 - acc: 0.9005 - val_loss: 4.8659 - val_acc: 0.3968\n",
      "Epoch 278/1000\n",
      "28709/28709 [==============================] - 9s 327us/step - loss: 0.8298 - acc: 0.9038 - val_loss: 3.9541 - val_acc: 0.4027\n",
      "Epoch 279/1000\n",
      "28709/28709 [==============================] - 10s 334us/step - loss: 0.0203 - acc: 0.9962 - val_loss: 5.4291 - val_acc: 0.3713\n",
      "Epoch 280/1000\n",
      "28709/28709 [==============================] - 13s 456us/step - loss: 0.5623 - acc: 0.9179 - val_loss: 4.5359 - val_acc: 0.3985\n",
      "Epoch 281/1000\n",
      "28709/28709 [==============================] - 10s 348us/step - loss: 0.2626 - acc: 0.9486 - val_loss: 4.6886 - val_acc: 0.3847\n",
      "Epoch 282/1000\n",
      "28709/28709 [==============================] - 9s 299us/step - loss: 0.4893 - acc: 0.9315 - val_loss: 5.5162 - val_acc: 0.4032\n",
      "Epoch 283/1000\n",
      "28709/28709 [==============================] - 13s 463us/step - loss: 0.5163 - acc: 0.9313 - val_loss: 5.1788 - val_acc: 0.3938\n",
      "Epoch 284/1000\n",
      "28709/28709 [==============================] - 8s 284us/step - loss: 0.6175 - acc: 0.9142 - val_loss: 5.2033 - val_acc: 0.3958\n",
      "Epoch 285/1000\n",
      "28709/28709 [==============================] - 11s 375us/step - loss: 0.7120 - acc: 0.9084 - val_loss: 3.5724 - val_acc: 0.3968\n",
      "Epoch 286/1000\n",
      "28709/28709 [==============================] - 13s 458us/step - loss: 0.0249 - acc: 0.9959 - val_loss: 5.3547 - val_acc: 0.4057\n",
      "Epoch 287/1000\n",
      "28709/28709 [==============================] - 7s 249us/step - loss: 0.8267 - acc: 0.8980 - val_loss: 4.7202 - val_acc: 0.3942\n",
      "Epoch 288/1000\n",
      "28709/28709 [==============================] - 12s 401us/step - loss: 0.2812 - acc: 0.9641 - val_loss: 8.8973 - val_acc: 0.2543\n",
      "Epoch 289/1000\n",
      "28709/28709 [==============================] - 12s 433us/step - loss: 0.5455 - acc: 0.9350 - val_loss: 4.3695 - val_acc: 0.4007\n",
      "Epoch 290/1000\n",
      "28709/28709 [==============================] - 7s 261us/step - loss: 0.5340 - acc: 0.9151 - val_loss: 3.4491 - val_acc: 0.4024\n",
      "Epoch 291/1000\n",
      "28709/28709 [==============================] - 12s 421us/step - loss: 0.0240 - acc: 0.9966 - val_loss: 5.4906 - val_acc: 0.3777\n",
      "Epoch 292/1000\n",
      "28709/28709 [==============================] - 12s 432us/step - loss: 0.4786 - acc: 0.9298 - val_loss: 4.3900 - val_acc: 0.3997\n",
      "Epoch 293/1000\n",
      "28709/28709 [==============================] - 10s 334us/step - loss: 0.8149 - acc: 0.8860 - val_loss: 3.8996 - val_acc: 0.3768\n",
      "Epoch 294/1000\n",
      "28709/28709 [==============================] - 12s 423us/step - loss: 0.0264 - acc: 0.9943 - val_loss: 5.9196 - val_acc: 0.4043\n",
      "Epoch 295/1000\n",
      "28709/28709 [==============================] - 15s 511us/step - loss: 0.7256 - acc: 0.9217 - val_loss: 4.6577 - val_acc: 0.4004\n",
      "Epoch 296/1000\n",
      "28709/28709 [==============================] - 10s 358us/step - loss: 0.7922 - acc: 0.9011 - val_loss: 3.8104 - val_acc: 0.3985\n",
      "Epoch 297/1000\n",
      "28709/28709 [==============================] - 10s 352us/step - loss: 0.0223 - acc: 0.9966 - val_loss: 5.2520 - val_acc: 0.3991\n",
      "Epoch 298/1000\n",
      "28709/28709 [==============================] - 14s 493us/step - loss: 0.7223 - acc: 0.9065 - val_loss: 5.5059 - val_acc: 0.3964\n",
      "Epoch 299/1000\n",
      "28709/28709 [==============================] - 13s 446us/step - loss: 0.6644 - acc: 0.9048 - val_loss: 3.6605 - val_acc: 0.3965\n",
      "Epoch 300/1000\n",
      "28709/28709 [==============================] - 7s 254us/step - loss: 0.7249 - acc: 0.9212 - val_loss: 4.1823 - val_acc: 0.3708\n",
      "Epoch 301/1000\n",
      "28709/28709 [==============================] - 12s 433us/step - loss: 0.0649 - acc: 0.9837 - val_loss: 4.6711 - val_acc: 0.3938\n",
      "Epoch 302/1000\n",
      "28709/28709 [==============================] - 12s 406us/step - loss: 0.7705 - acc: 0.8975 - val_loss: 4.8259 - val_acc: 0.3978\n",
      "Epoch 303/1000\n",
      "28709/28709 [==============================] - 7s 260us/step - loss: 0.0150 - acc: 0.9965 - val_loss: 6.0379 - val_acc: 0.3783\n",
      "Epoch 304/1000\n",
      "28709/28709 [==============================] - 13s 448us/step - loss: 0.9102 - acc: 0.8984 - val_loss: 4.9292 - val_acc: 0.3993\n",
      "Epoch 305/1000\n",
      "28709/28709 [==============================] - 11s 368us/step - loss: 0.6877 - acc: 0.9073 - val_loss: 3.4983 - val_acc: 0.3929\n",
      "Epoch 306/1000\n",
      "28709/28709 [==============================] - 9s 326us/step - loss: 0.0272 - acc: 0.9954 - val_loss: 5.4458 - val_acc: 0.4010\n",
      "Epoch 307/1000\n",
      "28709/28709 [==============================] - 14s 484us/step - loss: 0.7385 - acc: 0.9035 - val_loss: 4.0454 - val_acc: 0.3985\n",
      "Epoch 308/1000\n",
      "28709/28709 [==============================] - 10s 340us/step - loss: 0.0140 - acc: 0.9968 - val_loss: 5.5507 - val_acc: 0.3744\n",
      "Epoch 309/1000\n",
      "28709/28709 [==============================] - 10s 357us/step - loss: 0.9040 - acc: 0.8910 - val_loss: 4.7069 - val_acc: 0.3982\n",
      "Epoch 310/1000\n",
      "28709/28709 [==============================] - 14s 487us/step - loss: 0.6008 - acc: 0.9179 - val_loss: 3.7732 - val_acc: 0.3887\n",
      "Epoch 311/1000\n",
      "28709/28709 [==============================] - 8s 275us/step - loss: 0.0254 - acc: 0.9957 - val_loss: 4.9171 - val_acc: 0.3893\n",
      "Epoch 312/1000\n",
      "28709/28709 [==============================] - 12s 425us/step - loss: 0.8090 - acc: 0.8974 - val_loss: 5.3073 - val_acc: 0.4017\n",
      "Epoch 313/1000\n",
      "28709/28709 [==============================] - 13s 444us/step - loss: 0.8806 - acc: 0.8959 - val_loss: 3.5053 - val_acc: 0.3910\n",
      "Epoch 314/1000\n",
      "28709/28709 [==============================] - 8s 267us/step - loss: 0.0294 - acc: 0.9951 - val_loss: 5.6321 - val_acc: 0.4014\n",
      "Epoch 315/1000\n",
      "28709/28709 [==============================] - 14s 474us/step - loss: 0.9350 - acc: 0.8893 - val_loss: 5.2749 - val_acc: 0.3984\n",
      "Epoch 316/1000\n",
      "28709/28709 [==============================] - 13s 449us/step - loss: 0.0146 - acc: 0.9960 - val_loss: 5.6677 - val_acc: 0.4039\n",
      "Epoch 317/1000\n",
      "28709/28709 [==============================] - 8s 285us/step - loss: 1.0911 - acc: 0.8772 - val_loss: 5.0051 - val_acc: 0.3954\n",
      "Epoch 318/1000\n",
      "28709/28709 [==============================] - 13s 458us/step - loss: 1.2736 - acc: 0.8720 - val_loss: 4.2940 - val_acc: 0.3940\n",
      "Epoch 319/1000\n",
      "28709/28709 [==============================] - 11s 394us/step - loss: 0.0203 - acc: 0.9948 - val_loss: 5.7271 - val_acc: 0.3963\n",
      "Epoch 320/1000\n",
      "28709/28709 [==============================] - 9s 301us/step - loss: 0.7114 - acc: 0.9216 - val_loss: 4.5107 - val_acc: 0.3926\n",
      "Epoch 321/1000\n",
      "28709/28709 [==============================] - 14s 473us/step - loss: 0.2757 - acc: 0.9411 - val_loss: 4.2188 - val_acc: 0.4018\n",
      "Epoch 322/1000\n",
      "28709/28709 [==============================] - 10s 358us/step - loss: 0.8403 - acc: 0.8936 - val_loss: 3.7666 - val_acc: 0.3875\n",
      "Epoch 323/1000\n",
      "28709/28709 [==============================] - 9s 329us/step - loss: 0.0258 - acc: 0.9943 - val_loss: 6.5787 - val_acc: 0.3591\n",
      "Epoch 324/1000\n",
      "28709/28709 [==============================] - 14s 489us/step - loss: 0.6740 - acc: 0.9162 - val_loss: 4.4746 - val_acc: 0.3978\n",
      "Epoch 325/1000\n",
      "28709/28709 [==============================] - 11s 371us/step - loss: 0.8437 - acc: 0.9025 - val_loss: 3.6600 - val_acc: 0.4002\n",
      "Epoch 326/1000\n",
      "28709/28709 [==============================] - 9s 308us/step - loss: 0.0163 - acc: 0.9967 - val_loss: 5.6612 - val_acc: 0.3963\n",
      "Epoch 327/1000\n",
      "28709/28709 [==============================] - 13s 465us/step - loss: 1.0004 - acc: 0.8841 - val_loss: 4.1488 - val_acc: 0.4066\n",
      "Epoch 328/1000\n",
      "28709/28709 [==============================] - 10s 349us/step - loss: 0.0162 - acc: 0.9961 - val_loss: 5.3589 - val_acc: 0.3584\n",
      "Epoch 329/1000\n",
      "28709/28709 [==============================] - 9s 329us/step - loss: 0.7862 - acc: 0.9016 - val_loss: 5.6951 - val_acc: 0.3942\n",
      "Epoch 330/1000\n",
      "28709/28709 [==============================] - 13s 468us/step - loss: 0.4737 - acc: 0.9286 - val_loss: 4.0533 - val_acc: 0.3928\n",
      "Epoch 331/1000\n",
      "28709/28709 [==============================] - 9s 326us/step - loss: 0.0141 - acc: 0.9961 - val_loss: 7.0934 - val_acc: 0.3433\n",
      "Epoch 332/1000\n",
      "28709/28709 [==============================] - 11s 370us/step - loss: 0.6282 - acc: 0.9140 - val_loss: 4.8507 - val_acc: 0.3967\n",
      "Epoch 333/1000\n",
      "28709/28709 [==============================] - 14s 499us/step - loss: 0.5505 - acc: 0.9271 - val_loss: 3.9040 - val_acc: 0.3991\n",
      "Epoch 334/1000\n",
      "28709/28709 [==============================] - 10s 347us/step - loss: 0.0274 - acc: 0.9944 - val_loss: 8.3098 - val_acc: 0.3242\n",
      "Epoch 335/1000\n",
      "28709/28709 [==============================] - 10s 363us/step - loss: 0.6264 - acc: 0.9163 - val_loss: 5.4092 - val_acc: 0.3897\n",
      "Epoch 336/1000\n",
      "28709/28709 [==============================] - 14s 499us/step - loss: 0.3525 - acc: 0.9429 - val_loss: 4.3682 - val_acc: 0.4025\n",
      "Epoch 337/1000\n",
      "28709/28709 [==============================] - 12s 404us/step - loss: 0.7288 - acc: 0.9106 - val_loss: 4.1597 - val_acc: 0.3981\n",
      "Epoch 338/1000\n",
      "28709/28709 [==============================] - 9s 320us/step - loss: 0.0170 - acc: 0.9957 - val_loss: 5.9841 - val_acc: 0.3953\n",
      "Epoch 339/1000\n",
      "28709/28709 [==============================] - 15s 511us/step - loss: 0.6390 - acc: 0.9252 - val_loss: 5.3498 - val_acc: 0.3968\n",
      "Epoch 340/1000\n",
      "28709/28709 [==============================] - 12s 433us/step - loss: 0.7525 - acc: 0.9122 - val_loss: 3.7517 - val_acc: 0.3977\n",
      "Epoch 341/1000\n",
      "28709/28709 [==============================] - 8s 293us/step - loss: 0.0180 - acc: 0.9965 - val_loss: 5.9320 - val_acc: 0.4037\n",
      "Epoch 342/1000\n",
      "28709/28709 [==============================] - 15s 513us/step - loss: 0.9025 - acc: 0.8987 - val_loss: 4.1017 - val_acc: 0.3977\n",
      "Epoch 343/1000\n",
      "28709/28709 [==============================] - 12s 420us/step - loss: 0.4782 - acc: 0.9547 - val_loss: 6.9252 - val_acc: 0.3255\n",
      "Epoch 344/1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "28709/28709 [==============================] - 9s 311us/step - loss: 0.3213 - acc: 0.9469 - val_loss: 5.2149 - val_acc: 0.3932\n",
      "Epoch 345/1000\n",
      "28709/28709 [==============================] - 14s 490us/step - loss: 1.2636 - acc: 0.8760 - val_loss: 3.8909 - val_acc: 0.4016\n",
      "Epoch 346/1000\n",
      "28709/28709 [==============================] - 13s 436us/step - loss: 0.0241 - acc: 0.9947 - val_loss: 5.6789 - val_acc: 0.3986\n",
      "Epoch 347/1000\n",
      "28709/28709 [==============================] - 8s 287us/step - loss: 0.9396 - acc: 0.9000 - val_loss: 3.9960 - val_acc: 0.4013\n",
      "Epoch 348/1000\n",
      "28709/28709 [==============================] - 14s 505us/step - loss: 0.0152 - acc: 0.9967 - val_loss: 5.7201 - val_acc: 0.3949\n",
      "Epoch 349/1000\n",
      "28709/28709 [==============================] - 12s 425us/step - loss: 0.9407 - acc: 0.8878 - val_loss: 4.6037 - val_acc: 0.3970\n",
      "Epoch 350/1000\n",
      "28709/28709 [==============================] - 8s 292us/step - loss: 0.1137 - acc: 0.9803 - val_loss: 10.0781 - val_acc: 0.2175\n",
      "Epoch 351/1000\n",
      "28709/28709 [==============================] - 14s 485us/step - loss: 0.6056 - acc: 0.9247 - val_loss: 4.5671 - val_acc: 0.3981\n",
      "Epoch 352/1000\n",
      "28709/28709 [==============================] - 13s 456us/step - loss: 0.6974 - acc: 0.9117 - val_loss: 3.8464 - val_acc: 0.3873\n",
      "Epoch 353/1000\n",
      "28709/28709 [==============================] - 8s 291us/step - loss: 0.0477 - acc: 0.9895 - val_loss: 4.3980 - val_acc: 0.4021\n",
      "Epoch 354/1000\n",
      "28709/28709 [==============================] - 14s 477us/step - loss: 0.9179 - acc: 0.8998 - val_loss: 3.8804 - val_acc: 0.3922\n",
      "Epoch 355/1000\n",
      "28709/28709 [==============================] - 13s 449us/step - loss: 0.0256 - acc: 0.9962 - val_loss: 4.7130 - val_acc: 0.3942\n",
      "Epoch 356/1000\n",
      "28709/28709 [==============================] - 10s 342us/step - loss: 0.7814 - acc: 0.9189 - val_loss: 4.4479 - val_acc: 0.3972\n",
      "Epoch 357/1000\n",
      "28709/28709 [==============================] - 14s 479us/step - loss: 0.0131 - acc: 0.9966 - val_loss: 5.7420 - val_acc: 0.3977\n",
      "Epoch 358/1000\n",
      "28709/28709 [==============================] - 15s 533us/step - loss: 0.8742 - acc: 0.8990 - val_loss: 4.3804 - val_acc: 0.3967\n",
      "Epoch 359/1000\n",
      "28709/28709 [==============================] - 10s 366us/step - loss: 0.0254 - acc: 0.9948 - val_loss: 8.6614 - val_acc: 0.3043\n",
      "Epoch 360/1000\n",
      "28709/28709 [==============================] - 11s 397us/step - loss: 0.6156 - acc: 0.9242 - val_loss: 5.7608 - val_acc: 0.3956\n",
      "Epoch 361/1000\n",
      "28709/28709 [==============================] - 14s 499us/step - loss: 0.6789 - acc: 0.9161 - val_loss: 4.0359 - val_acc: 0.4009\n",
      "Epoch 362/1000\n",
      "28709/28709 [==============================] - 10s 355us/step - loss: 0.0178 - acc: 0.9963 - val_loss: 5.9342 - val_acc: 0.3561\n",
      "Epoch 363/1000\n",
      "28709/28709 [==============================] - 10s 333us/step - loss: 0.4696 - acc: 0.9257 - val_loss: 4.5357 - val_acc: 0.4009\n",
      "Epoch 364/1000\n",
      "28709/28709 [==============================] - 14s 491us/step - loss: 0.6787 - acc: 0.9122 - val_loss: 3.8938 - val_acc: 0.4035\n",
      "Epoch 365/1000\n",
      "28709/28709 [==============================] - 9s 305us/step - loss: 0.0222 - acc: 0.9966 - val_loss: 6.0258 - val_acc: 0.3939\n",
      "Epoch 366/1000\n",
      "28709/28709 [==============================] - 11s 398us/step - loss: 0.7909 - acc: 0.9145 - val_loss: 5.0312 - val_acc: 0.3952\n",
      "Epoch 367/1000\n",
      "28709/28709 [==============================] - 14s 482us/step - loss: 0.0117 - acc: 0.9968 - val_loss: 6.0310 - val_acc: 0.3926\n",
      "Epoch 368/1000\n",
      "28709/28709 [==============================] - 8s 279us/step - loss: 1.0869 - acc: 0.8937 - val_loss: 5.0072 - val_acc: 0.3982\n",
      "Epoch 369/1000\n",
      "28709/28709 [==============================] - 14s 474us/step - loss: 1.0375 - acc: 0.8961 - val_loss: 4.3656 - val_acc: 0.3736\n",
      "Epoch 370/1000\n",
      "28709/28709 [==============================] - 11s 382us/step - loss: 0.0358 - acc: 0.9911 - val_loss: 5.6749 - val_acc: 0.3974\n",
      "Epoch 371/1000\n",
      "28709/28709 [==============================] - 10s 343us/step - loss: 0.7014 - acc: 0.9094 - val_loss: 4.0248 - val_acc: 0.3950\n",
      "Epoch 372/1000\n",
      "28709/28709 [==============================] - 15s 516us/step - loss: 0.0233 - acc: 0.9955 - val_loss: 5.7062 - val_acc: 0.3964\n",
      "Epoch 373/1000\n",
      "28709/28709 [==============================] - 8s 262us/step - loss: 0.8695 - acc: 0.8996 - val_loss: 4.1818 - val_acc: 0.4025\n",
      "Epoch 374/1000\n",
      "28709/28709 [==============================] - 13s 437us/step - loss: 0.0152 - acc: 0.9965 - val_loss: 6.6985 - val_acc: 0.3364\n",
      "Epoch 375/1000\n",
      "28709/28709 [==============================] - 12s 419us/step - loss: 0.9333 - acc: 0.8948 - val_loss: 4.4844 - val_acc: 0.3971\n",
      "Epoch 376/1000\n",
      "28709/28709 [==============================] - 9s 303us/step - loss: 0.8478 - acc: 0.9087 - val_loss: 4.2385 - val_acc: 0.3847\n",
      "Epoch 377/1000\n",
      "28709/28709 [==============================] - 15s 506us/step - loss: 0.0266 - acc: 0.9952 - val_loss: 4.8590 - val_acc: 0.3972\n",
      "Epoch 378/1000\n",
      "28709/28709 [==============================] - 10s 350us/step - loss: 1.3037 - acc: 0.8746 - val_loss: 4.4162 - val_acc: 0.3970\n",
      "Epoch 379/1000\n",
      "28709/28709 [==============================] - 11s 373us/step - loss: 0.0129 - acc: 0.9966 - val_loss: 5.8468 - val_acc: 0.3837\n",
      "Epoch 380/1000\n",
      "28709/28709 [==============================] - 13s 449us/step - loss: 0.8237 - acc: 0.9056 - val_loss: 4.9907 - val_acc: 0.4018\n",
      "Epoch 381/1000\n",
      "28709/28709 [==============================] - 8s 290us/step - loss: 0.7066 - acc: 0.9251 - val_loss: 4.1739 - val_acc: 0.3925\n",
      "Epoch 382/1000\n",
      "28709/28709 [==============================] - 13s 469us/step - loss: 0.0192 - acc: 0.9968 - val_loss: 5.4494 - val_acc: 0.3970\n",
      "Epoch 383/1000\n",
      "28709/28709 [==============================] - 12s 403us/step - loss: 0.9077 - acc: 0.8996 - val_loss: 5.5277 - val_acc: 0.3960\n",
      "Epoch 384/1000\n",
      "28709/28709 [==============================] - 9s 315us/step - loss: 0.3984 - acc: 0.9509 - val_loss: 5.4800 - val_acc: 0.3095\n",
      "Epoch 385/1000\n",
      "28709/28709 [==============================] - 15s 514us/step - loss: 0.1016 - acc: 0.9755 - val_loss: 5.7566 - val_acc: 0.3967\n",
      "Epoch 386/1000\n",
      "28709/28709 [==============================] - 7s 259us/step - loss: 0.8753 - acc: 0.8970 - val_loss: 4.1158 - val_acc: 0.3946\n",
      "Epoch 387/1000\n",
      "28709/28709 [==============================] - 13s 464us/step - loss: 0.0174 - acc: 0.9969 - val_loss: 5.8972 - val_acc: 0.3717\n",
      "Epoch 388/1000\n",
      "28709/28709 [==============================] - 11s 384us/step - loss: 0.4399 - acc: 0.9323 - val_loss: 4.5165 - val_acc: 0.4053\n",
      "Epoch 389/1000\n",
      "28709/28709 [==============================] - 10s 337us/step - loss: 1.3478 - acc: 0.8665 - val_loss: 4.3352 - val_acc: 0.3936\n",
      "Epoch 390/1000\n",
      "28709/28709 [==============================] - 14s 490us/step - loss: 0.0249 - acc: 0.9946 - val_loss: 5.3045 - val_acc: 0.3963\n",
      "Epoch 391/1000\n",
      "28709/28709 [==============================] - 8s 269us/step - loss: 0.6504 - acc: 0.9209 - val_loss: 5.4904 - val_acc: 0.3972\n",
      "Epoch 392/1000\n",
      "28709/28709 [==============================] - 13s 459us/step - loss: 0.0116 - acc: 0.9970 - val_loss: 6.7785 - val_acc: 0.3685\n",
      "Epoch 393/1000\n",
      "28709/28709 [==============================] - 11s 399us/step - loss: 1.9438 - acc: 0.8285 - val_loss: 5.9373 - val_acc: 0.3963\n",
      "Epoch 394/1000\n",
      "28709/28709 [==============================] - 10s 348us/step - loss: 1.2527 - acc: 0.8950 - val_loss: 4.6641 - val_acc: 0.3652\n",
      "Epoch 395/1000\n",
      "28709/28709 [==============================] - 15s 526us/step - loss: 0.0464 - acc: 0.9881 - val_loss: 5.9976 - val_acc: 0.4002\n",
      "Epoch 396/1000\n",
      "28709/28709 [==============================] - 10s 332us/step - loss: 0.9522 - acc: 0.8866 - val_loss: 5.8029 - val_acc: 0.3936\n",
      "Epoch 397/1000\n",
      "28709/28709 [==============================] - 11s 394us/step - loss: 0.0114 - acc: 0.9966 - val_loss: 5.9244 - val_acc: 0.3981\n",
      "Epoch 398/1000\n",
      "28709/28709 [==============================] - 13s 466us/step - loss: 0.7382 - acc: 0.9109 - val_loss: 4.4123 - val_acc: 0.3999\n",
      "Epoch 399/1000\n",
      "28709/28709 [==============================] - 9s 319us/step - loss: 0.0126 - acc: 0.9973 - val_loss: 6.0606 - val_acc: 0.3875\n",
      "Epoch 400/1000\n",
      "28709/28709 [==============================] - 14s 483us/step - loss: 0.7603 - acc: 0.9056 - val_loss: 5.7237 - val_acc: 0.4009\n",
      "Epoch 401/1000\n",
      "28709/28709 [==============================] - 12s 403us/step - loss: 0.9223 - acc: 0.9016 - val_loss: 4.6956 - val_acc: 0.3971\n",
      "Epoch 402/1000\n",
      "28709/28709 [==============================] - 10s 349us/step - loss: 0.0121 - acc: 0.9965 - val_loss: 6.4758 - val_acc: 0.3986\n",
      "Epoch 403/1000\n",
      "28709/28709 [==============================] - 15s 529us/step - loss: 0.9167 - acc: 0.8891 - val_loss: 4.2725 - val_acc: 0.3953\n",
      "Epoch 404/1000\n",
      "28709/28709 [==============================] - 7s 258us/step - loss: 0.4506 - acc: 0.9543 - val_loss: 5.3739 - val_acc: 0.3676\n",
      "Epoch 405/1000\n",
      "28709/28709 [==============================] - 14s 496us/step - loss: 0.0995 - acc: 0.9744 - val_loss: 5.5433 - val_acc: 0.3989\n",
      "Epoch 406/1000\n",
      "28709/28709 [==============================] - 12s 434us/step - loss: 0.6584 - acc: 0.9221 - val_loss: 4.2165 - val_acc: 0.3991\n",
      "Epoch 407/1000\n",
      "28709/28709 [==============================] - 10s 342us/step - loss: 0.0146 - acc: 0.9968 - val_loss: 5.7355 - val_acc: 0.3991\n",
      "Epoch 408/1000\n",
      "28709/28709 [==============================] - 16s 562us/step - loss: 0.8460 - acc: 0.9051 - val_loss: 4.2361 - val_acc: 0.3925\n",
      "Epoch 409/1000\n",
      "28709/28709 [==============================] - 12s 402us/step - loss: 0.0206 - acc: 0.9966 - val_loss: 5.0535 - val_acc: 0.3979\n",
      "Epoch 410/1000\n",
      "28709/28709 [==============================] - 11s 394us/step - loss: 0.8108 - acc: 0.8968 - val_loss: 5.2156 - val_acc: 0.4037\n",
      "Epoch 411/1000\n",
      "28709/28709 [==============================] - 16s 564us/step - loss: 0.0560 - acc: 0.9877 - val_loss: 9.3977 - val_acc: 0.2776\n",
      "Epoch 412/1000\n",
      "28709/28709 [==============================] - 10s 343us/step - loss: 0.8659 - acc: 0.9062 - val_loss: 5.5770 - val_acc: 0.3946\n",
      "Epoch 413/1000\n",
      "28709/28709 [==============================] - 13s 436us/step - loss: 0.7199 - acc: 0.9182 - val_loss: 4.2888 - val_acc: 0.3800\n",
      "Epoch 414/1000\n",
      "28709/28709 [==============================] - 14s 497us/step - loss: 0.0541 - acc: 0.9841 - val_loss: 5.8831 - val_acc: 0.4013\n",
      "Epoch 415/1000\n",
      "28709/28709 [==============================] - 9s 310us/step - loss: 1.1311 - acc: 0.8878 - val_loss: 4.5431 - val_acc: 0.3993\n",
      "Epoch 416/1000\n",
      "28709/28709 [==============================] - 15s 524us/step - loss: 0.0171 - acc: 0.9966 - val_loss: 5.1596 - val_acc: 0.3815\n",
      "Epoch 417/1000\n",
      "28709/28709 [==============================] - 13s 464us/step - loss: 0.3429 - acc: 0.9413 - val_loss: 5.5532 - val_acc: 0.3992\n",
      "Epoch 418/1000\n",
      "28709/28709 [==============================] - 10s 362us/step - loss: 0.6712 - acc: 0.9168 - val_loss: 5.1588 - val_acc: 0.3887\n",
      "Epoch 419/1000\n",
      "28709/28709 [==============================] - 15s 532us/step - loss: 0.0100 - acc: 0.9964 - val_loss: 6.0389 - val_acc: 0.3922\n",
      "Epoch 420/1000\n",
      "28709/28709 [==============================] - 14s 493us/step - loss: 1.5765 - acc: 0.8520 - val_loss: 5.2940 - val_acc: 0.4016\n",
      "Epoch 421/1000\n",
      "28709/28709 [==============================] - 11s 382us/step - loss: 0.0151 - acc: 0.9958 - val_loss: 5.8673 - val_acc: 0.4017\n",
      "Epoch 422/1000\n",
      "28709/28709 [==============================] - 15s 529us/step - loss: 0.7479 - acc: 0.9134 - val_loss: 5.5916 - val_acc: 0.3975\n",
      "Epoch 423/1000\n",
      "28709/28709 [==============================] - 17s 601us/step - loss: 0.0103 - acc: 0.9965 - val_loss: 6.0322 - val_acc: 0.3984\n",
      "Epoch 424/1000\n",
      "28709/28709 [==============================] - 10s 360us/step - loss: 1.1105 - acc: 0.8824 - val_loss: 5.5309 - val_acc: 0.3946\n",
      "Epoch 425/1000\n",
      "28709/28709 [==============================] - 12s 422us/step - loss: 0.6114 - acc: 0.9258 - val_loss: 4.0186 - val_acc: 0.3953\n",
      "Epoch 426/1000\n",
      "28709/28709 [==============================] - 16s 540us/step - loss: 0.0199 - acc: 0.9968 - val_loss: 5.7586 - val_acc: 0.3974\n",
      "Epoch 427/1000\n",
      "28709/28709 [==============================] - 10s 352us/step - loss: 1.5565 - acc: 0.8533 - val_loss: 4.3943 - val_acc: 0.4016\n",
      "Epoch 428/1000\n",
      "28709/28709 [==============================] - 15s 511us/step - loss: 0.0164 - acc: 0.9966 - val_loss: 6.1975 - val_acc: 0.4027\n",
      "Epoch 429/1000\n",
      "28709/28709 [==============================] - 16s 574us/step - loss: 0.6379 - acc: 0.9168 - val_loss: 5.1417 - val_acc: 0.3986\n",
      "Epoch 430/1000\n",
      "28709/28709 [==============================] - 11s 372us/step - loss: 0.0095 - acc: 0.9968 - val_loss: 6.0223 - val_acc: 0.3861\n",
      "Epoch 431/1000\n",
      "28709/28709 [==============================] - 13s 460us/step - loss: 0.7204 - acc: 0.9203 - val_loss: 4.7275 - val_acc: 0.4021\n",
      "Epoch 432/1000\n",
      "28709/28709 [==============================] - 17s 593us/step - loss: 0.0217 - acc: 0.9947 - val_loss: 7.5295 - val_acc: 0.3297\n",
      "Epoch 433/1000\n",
      "28709/28709 [==============================] - 12s 402us/step - loss: 0.3630 - acc: 0.9386 - val_loss: 5.5861 - val_acc: 0.4032\n",
      "Epoch 434/1000\n",
      "28709/28709 [==============================] - 12s 408us/step - loss: 1.0410 - acc: 0.8957 - val_loss: 4.0515 - val_acc: 0.3971\n",
      "Epoch 435/1000\n",
      "28709/28709 [==============================] - 16s 547us/step - loss: 0.0253 - acc: 0.9955 - val_loss: 4.8987 - val_acc: 0.4080\n",
      "Epoch 436/1000\n",
      "28709/28709 [==============================] - 9s 315us/step - loss: 0.5899 - acc: 0.9267 - val_loss: 4.5241 - val_acc: 0.3986\n",
      "Epoch 437/1000\n",
      "28709/28709 [==============================] - 12s 401us/step - loss: 0.6072 - acc: 0.9437 - val_loss: 6.9676 - val_acc: 0.3368\n",
      "Epoch 438/1000\n",
      "28709/28709 [==============================] - 14s 499us/step - loss: 0.1957 - acc: 0.9639 - val_loss: 5.8824 - val_acc: 0.3953\n",
      "Epoch 439/1000\n",
      "28709/28709 [==============================] - 9s 323us/step - loss: 1.1151 - acc: 0.8882 - val_loss: 4.6651 - val_acc: 0.3586\n",
      "Epoch 440/1000\n",
      "28709/28709 [==============================] - 13s 469us/step - loss: 0.0414 - acc: 0.9889 - val_loss: 6.0289 - val_acc: 0.3967\n",
      "Epoch 441/1000\n",
      "28709/28709 [==============================] - 14s 479us/step - loss: 1.2068 - acc: 0.8929 - val_loss: 4.7310 - val_acc: 0.3875\n",
      "Epoch 442/1000\n",
      "28709/28709 [==============================] - 9s 296us/step - loss: 0.0162 - acc: 0.9952 - val_loss: 6.0596 - val_acc: 0.3965\n",
      "Epoch 443/1000\n",
      "28709/28709 [==============================] - 16s 542us/step - loss: 0.7543 - acc: 0.9180 - val_loss: 5.3322 - val_acc: 0.4030\n",
      "Epoch 444/1000\n",
      "28709/28709 [==============================] - 13s 464us/step - loss: 0.0120 - acc: 0.9964 - val_loss: 6.3249 - val_acc: 0.3574\n",
      "Epoch 445/1000\n",
      "28709/28709 [==============================] - 10s 356us/step - loss: 0.9274 - acc: 0.9043 - val_loss: 5.3937 - val_acc: 0.3952\n",
      "Epoch 446/1000\n",
      "28709/28709 [==============================] - 16s 563us/step - loss: 0.8049 - acc: 0.9031 - val_loss: 4.3401 - val_acc: 0.3928\n",
      "Epoch 447/1000\n",
      "28709/28709 [==============================] - 13s 438us/step - loss: 0.0155 - acc: 0.9967 - val_loss: 5.9241 - val_acc: 0.3960\n",
      "Epoch 448/1000\n",
      "28709/28709 [==============================] - 11s 369us/step - loss: 1.5394 - acc: 0.8683 - val_loss: 5.4595 - val_acc: 0.3453\n",
      "Epoch 449/1000\n",
      "28709/28709 [==============================] - 16s 567us/step - loss: 0.0958 - acc: 0.9766 - val_loss: 5.0322 - val_acc: 0.4027\n",
      "Epoch 450/1000\n",
      "28709/28709 [==============================] - 10s 339us/step - loss: 0.8888 - acc: 0.9009 - val_loss: 4.2861 - val_acc: 0.3988\n",
      "Epoch 451/1000\n",
      "28709/28709 [==============================] - 13s 438us/step - loss: 0.0156 - acc: 0.9968 - val_loss: 6.3283 - val_acc: 0.4055\n",
      "Epoch 452/1000\n",
      "28709/28709 [==============================] - 15s 510us/step - loss: 0.6382 - acc: 0.9292 - val_loss: 4.7038 - val_acc: 0.3982\n",
      "Epoch 453/1000\n",
      "28709/28709 [==============================] - 10s 349us/step - loss: 0.5152 - acc: 0.9452 - val_loss: 6.7962 - val_acc: 0.3163\n",
      "Epoch 454/1000\n",
      "28709/28709 [==============================] - 13s 442us/step - loss: 0.1585 - acc: 0.9701 - val_loss: 5.5377 - val_acc: 0.3889\n",
      "Epoch 455/1000\n",
      "28709/28709 [==============================] - 14s 477us/step - loss: 0.3873 - acc: 0.9394 - val_loss: 4.6866 - val_acc: 0.4020\n",
      "Epoch 456/1000\n",
      "28709/28709 [==============================] - 9s 325us/step - loss: 0.2555 - acc: 0.9704 - val_loss: 8.4110 - val_acc: 0.2898\n",
      "Epoch 457/1000\n",
      "28709/28709 [==============================] - 15s 515us/step - loss: 0.3216 - acc: 0.9588 - val_loss: 4.9506 - val_acc: 0.4013\n",
      "Epoch 458/1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "28709/28709 [==============================] - 12s 402us/step - loss: 0.8249 - acc: 0.9230 - val_loss: 5.4778 - val_acc: 0.3717\n",
      "Epoch 459/1000\n",
      "28709/28709 [==============================] - 10s 356us/step - loss: 0.0489 - acc: 0.9863 - val_loss: 5.7392 - val_acc: 0.3993\n",
      "Epoch 460/1000\n",
      "28709/28709 [==============================] - 16s 551us/step - loss: 0.9349 - acc: 0.9081 - val_loss: 4.8585 - val_acc: 0.3943\n",
      "Epoch 461/1000\n",
      "28709/28709 [==============================] - 9s 298us/step - loss: 0.0145 - acc: 0.9965 - val_loss: 5.9797 - val_acc: 0.3786\n",
      "Epoch 462/1000\n",
      "28709/28709 [==============================] - 14s 484us/step - loss: 0.5059 - acc: 0.9349 - val_loss: 5.8094 - val_acc: 0.3791\n",
      "Epoch 463/1000\n",
      "28709/28709 [==============================] - 13s 458us/step - loss: 0.3735 - acc: 0.9411 - val_loss: 5.0475 - val_acc: 0.3975\n",
      "Epoch 464/1000\n",
      "28709/28709 [==============================] - 9s 324us/step - loss: 1.1923 - acc: 0.8994 - val_loss: 6.6720 - val_acc: 0.3037\n",
      "Epoch 465/1000\n",
      "28709/28709 [==============================] - 14s 497us/step - loss: 0.1503 - acc: 0.9703 - val_loss: 5.1474 - val_acc: 0.3988\n",
      "Epoch 466/1000\n",
      "28709/28709 [==============================] - 12s 431us/step - loss: 0.5188 - acc: 0.9485 - val_loss: 5.6517 - val_acc: 0.3625\n",
      "Epoch 467/1000\n",
      "28709/28709 [==============================] - 9s 330us/step - loss: 0.0736 - acc: 0.9803 - val_loss: 5.2459 - val_acc: 0.4006\n",
      "Epoch 468/1000\n",
      "28709/28709 [==============================] - 16s 547us/step - loss: 0.5265 - acc: 0.9290 - val_loss: 4.6480 - val_acc: 0.4048\n",
      "Epoch 469/1000\n",
      "28709/28709 [==============================] - 12s 401us/step - loss: 0.0113 - acc: 0.9971 - val_loss: 6.0060 - val_acc: 0.4035\n",
      "Epoch 470/1000\n",
      "28709/28709 [==============================] - 10s 357us/step - loss: 1.4180 - acc: 0.8677 - val_loss: 4.9303 - val_acc: 0.3996\n",
      "Epoch 471/1000\n",
      "28709/28709 [==============================] - 15s 534us/step - loss: 0.0112 - acc: 0.9970 - val_loss: 5.4889 - val_acc: 0.3945\n",
      "Epoch 472/1000\n",
      "28709/28709 [==============================] - 9s 321us/step - loss: 0.7285 - acc: 0.9215 - val_loss: 5.1133 - val_acc: 0.3928\n",
      "Epoch 473/1000\n",
      "28709/28709 [==============================] - 13s 462us/step - loss: 0.0104 - acc: 0.9968 - val_loss: 6.0420 - val_acc: 0.3972\n",
      "Epoch 474/1000\n",
      "28709/28709 [==============================] - 17s 583us/step - loss: 1.1059 - acc: 0.8897 - val_loss: 6.6143 - val_acc: 0.3981\n",
      "Epoch 475/1000\n",
      "28709/28709 [==============================] - 14s 482us/step - loss: 0.2848 - acc: 0.9670 - val_loss: 9.2864 - val_acc: 0.3099\n",
      "Epoch 476/1000\n",
      "28709/28709 [==============================] - 8s 281us/step - loss: 0.7897 - acc: 0.9244 - val_loss: 6.0179 - val_acc: 0.3975\n",
      "Epoch 477/1000\n",
      "28709/28709 [==============================] - 14s 474us/step - loss: 0.4139 - acc: 0.9417 - val_loss: 4.7734 - val_acc: 0.3691\n",
      "Epoch 478/1000\n",
      "28709/28709 [==============================] - 11s 377us/step - loss: 0.0186 - acc: 0.9944 - val_loss: 6.2008 - val_acc: 0.3914\n",
      "Epoch 479/1000\n",
      "28709/28709 [==============================] - 10s 338us/step - loss: 1.0474 - acc: 0.8967 - val_loss: 4.7839 - val_acc: 0.4031\n",
      "Epoch 480/1000\n",
      "28709/28709 [==============================] - 15s 519us/step - loss: 0.0130 - acc: 0.9971 - val_loss: 5.7405 - val_acc: 0.3993\n",
      "Epoch 481/1000\n",
      "28709/28709 [==============================] - 8s 288us/step - loss: 1.4644 - acc: 0.8670 - val_loss: 4.9163 - val_acc: 0.3982\n",
      "Epoch 482/1000\n",
      "28709/28709 [==============================] - 12s 431us/step - loss: 0.0144 - acc: 0.9966 - val_loss: 5.4135 - val_acc: 0.3931\n",
      "Epoch 483/1000\n",
      "28709/28709 [==============================] - 14s 482us/step - loss: 0.7377 - acc: 0.9242 - val_loss: 5.4615 - val_acc: 0.3931\n",
      "Epoch 484/1000\n",
      "28709/28709 [==============================] - 8s 268us/step - loss: 0.0103 - acc: 0.9966 - val_loss: 6.3213 - val_acc: 0.3946\n",
      "Epoch 485/1000\n",
      "28709/28709 [==============================] - 14s 471us/step - loss: 0.8971 - acc: 0.9046 - val_loss: 4.9240 - val_acc: 0.4009\n",
      "Epoch 486/1000\n",
      "28709/28709 [==============================] - 11s 371us/step - loss: 0.0088 - acc: 0.9971 - val_loss: 6.0578 - val_acc: 0.3943\n",
      "Epoch 487/1000\n",
      "28709/28709 [==============================] - 10s 347us/step - loss: 1.0145 - acc: 0.8951 - val_loss: 5.0292 - val_acc: 0.3991\n",
      "Epoch 488/1000\n",
      "28709/28709 [==============================] - 14s 499us/step - loss: 0.0122 - acc: 0.9970 - val_loss: 6.0870 - val_acc: 0.4021\n",
      "Epoch 489/1000\n",
      "28709/28709 [==============================] - 9s 300us/step - loss: 1.0923 - acc: 0.8940 - val_loss: 5.1619 - val_acc: 0.4009\n",
      "Epoch 490/1000\n",
      "28709/28709 [==============================] - 13s 457us/step - loss: 0.0106 - acc: 0.9968 - val_loss: 6.1843 - val_acc: 0.3984\n",
      "Epoch 491/1000\n",
      "28709/28709 [==============================] - 12s 419us/step - loss: 0.6936 - acc: 0.9250 - val_loss: 6.2859 - val_acc: 0.3999\n",
      "Epoch 492/1000\n",
      "28709/28709 [==============================] - 9s 328us/step - loss: 1.1390 - acc: 0.9057 - val_loss: 6.6628 - val_acc: 0.3316\n",
      "Epoch 493/1000\n",
      "28709/28709 [==============================] - 14s 502us/step - loss: 0.1487 - acc: 0.9695 - val_loss: 5.2275 - val_acc: 0.4053\n",
      "Epoch 494/1000\n",
      "28709/28709 [==============================] - 9s 305us/step - loss: 0.7754 - acc: 0.9301 - val_loss: 6.1617 - val_acc: 0.3348\n",
      "Epoch 495/1000\n",
      "28709/28709 [==============================] - 13s 440us/step - loss: 0.0926 - acc: 0.9749 - val_loss: 6.0734 - val_acc: 0.3970\n",
      "Epoch 496/1000\n",
      "28709/28709 [==============================] - 13s 444us/step - loss: 0.4001 - acc: 0.9405 - val_loss: 4.9551 - val_acc: 0.3963\n",
      "Epoch 497/1000\n",
      "28709/28709 [==============================] - 9s 311us/step - loss: 0.0100 - acc: 0.9969 - val_loss: 6.3238 - val_acc: 0.4041\n",
      "Epoch 498/1000\n",
      "28709/28709 [==============================] - 17s 588us/step - loss: 0.6524 - acc: 0.9251 - val_loss: 5.1193 - val_acc: 0.4000\n",
      "Epoch 499/1000\n",
      "28709/28709 [==============================] - 14s 490us/step - loss: 0.0119 - acc: 0.9966 - val_loss: 7.1889 - val_acc: 0.3357\n",
      "Epoch 500/1000\n",
      "28709/28709 [==============================] - 9s 304us/step - loss: 0.8509 - acc: 0.9118 - val_loss: 5.2515 - val_acc: 0.3993\n",
      "Epoch 501/1000\n",
      "28709/28709 [==============================] - 15s 508us/step - loss: 0.7967 - acc: 0.9117 - val_loss: 4.9232 - val_acc: 0.3972\n",
      "Epoch 502/1000\n",
      "28709/28709 [==============================] - 10s 343us/step - loss: 0.0135 - acc: 0.9968 - val_loss: 5.2205 - val_acc: 0.3996\n",
      "Epoch 503/1000\n",
      "28709/28709 [==============================] - 10s 366us/step - loss: 1.1960 - acc: 0.8843 - val_loss: 5.2104 - val_acc: 0.3875\n",
      "Epoch 504/1000\n",
      "28709/28709 [==============================] - 13s 455us/step - loss: 0.0162 - acc: 0.9956 - val_loss: 5.9553 - val_acc: 0.3861\n",
      "Epoch 505/1000\n",
      "28709/28709 [==============================] - 8s 269us/step - loss: 0.5096 - acc: 0.9317 - val_loss: 5.9784 - val_acc: 0.3981\n",
      "Epoch 506/1000\n",
      "28709/28709 [==============================] - 15s 516us/step - loss: 0.0231 - acc: 0.9940 - val_loss: 9.8846 - val_acc: 0.2300\n",
      "Epoch 507/1000\n",
      "28709/28709 [==============================] - 10s 352us/step - loss: 0.6575 - acc: 0.9174 - val_loss: 5.2816 - val_acc: 0.4024\n",
      "Epoch 508/1000\n",
      "28709/28709 [==============================] - 12s 402us/step - loss: 0.2592 - acc: 0.9722 - val_loss: 9.2073 - val_acc: 0.2805\n",
      "Epoch 509/1000\n",
      "28709/28709 [==============================] - 16s 561us/step - loss: 0.5971 - acc: 0.9327 - val_loss: 5.2269 - val_acc: 0.4010\n",
      "Epoch 510/1000\n",
      "28709/28709 [==============================] - 9s 327us/step - loss: 0.0100 - acc: 0.9967 - val_loss: 5.9169 - val_acc: 0.3815\n",
      "Epoch 511/1000\n",
      "28709/28709 [==============================] - 13s 459us/step - loss: 0.7332 - acc: 0.9181 - val_loss: 6.1645 - val_acc: 0.3958\n",
      "Epoch 512/1000\n",
      "28709/28709 [==============================] - 15s 528us/step - loss: 0.0096 - acc: 0.9970 - val_loss: 6.7282 - val_acc: 0.3801\n",
      "Epoch 513/1000\n",
      "28709/28709 [==============================] - 9s 325us/step - loss: 0.7661 - acc: 0.9118 - val_loss: 5.4997 - val_acc: 0.3964\n",
      "Epoch 514/1000\n",
      "28709/28709 [==============================] - 14s 483us/step - loss: 0.0091 - acc: 0.9968 - val_loss: 6.1843 - val_acc: 0.3921\n",
      "Epoch 515/1000\n",
      "28709/28709 [==============================] - 14s 495us/step - loss: 0.9068 - acc: 0.9031 - val_loss: 5.8206 - val_acc: 0.3967\n",
      "Epoch 516/1000\n",
      "28709/28709 [==============================] - 9s 305us/step - loss: 0.0091 - acc: 0.9971 - val_loss: 6.3888 - val_acc: 0.3954\n",
      "Epoch 517/1000\n",
      "28709/28709 [==============================] - 14s 499us/step - loss: 1.0151 - acc: 0.9014 - val_loss: 5.1861 - val_acc: 0.3965\n",
      "Epoch 518/1000\n",
      "28709/28709 [==============================] - 12s 434us/step - loss: 0.0107 - acc: 0.9966 - val_loss: 5.4209 - val_acc: 0.3984\n",
      "Epoch 519/1000\n",
      "28709/28709 [==============================] - 10s 353us/step - loss: 0.9984 - acc: 0.8987 - val_loss: 5.6413 - val_acc: 0.3965\n",
      "Epoch 520/1000\n",
      "28709/28709 [==============================] - 16s 557us/step - loss: 0.0105 - acc: 0.9966 - val_loss: 6.3308 - val_acc: 0.3922\n",
      "Epoch 521/1000\n",
      "28709/28709 [==============================] - 11s 392us/step - loss: 0.8441 - acc: 0.9085 - val_loss: 4.9838 - val_acc: 0.4003\n",
      "Epoch 522/1000\n",
      "28709/28709 [==============================] - 11s 398us/step - loss: 0.0108 - acc: 0.9972 - val_loss: 5.3775 - val_acc: 0.3982\n",
      "Epoch 523/1000\n",
      "28709/28709 [==============================] - 17s 586us/step - loss: 0.8652 - acc: 0.9127 - val_loss: 5.2911 - val_acc: 0.3931\n",
      "Epoch 524/1000\n",
      "28709/28709 [==============================] - 14s 476us/step - loss: 0.0109 - acc: 0.9972 - val_loss: 5.4781 - val_acc: 0.4000\n",
      "Epoch 525/1000\n",
      "28709/28709 [==============================] - 11s 392us/step - loss: 0.8878 - acc: 0.9075 - val_loss: 5.3647 - val_acc: 0.3975\n",
      "Epoch 526/1000\n",
      "28709/28709 [==============================] - 15s 538us/step - loss: 0.0113 - acc: 0.9967 - val_loss: 6.1214 - val_acc: 0.3970\n",
      "Epoch 527/1000\n",
      "28709/28709 [==============================] - 14s 497us/step - loss: 1.3443 - acc: 0.8823 - val_loss: 5.4294 - val_acc: 0.3911\n",
      "Epoch 528/1000\n",
      "28709/28709 [==============================] - 10s 354us/step - loss: 0.0126 - acc: 0.9961 - val_loss: 5.6117 - val_acc: 0.3932\n",
      "Epoch 529/1000\n",
      "28709/28709 [==============================] - 15s 519us/step - loss: 1.0907 - acc: 0.8858 - val_loss: 5.4042 - val_acc: 0.3968\n",
      "Epoch 530/1000\n",
      "28709/28709 [==============================] - 13s 454us/step - loss: 0.0124 - acc: 0.9964 - val_loss: 5.5319 - val_acc: 0.3991\n",
      "Epoch 531/1000\n",
      "28709/28709 [==============================] - 12s 409us/step - loss: 1.1128 - acc: 0.8984 - val_loss: 5.7367 - val_acc: 0.3712\n",
      "Epoch 532/1000\n",
      "28709/28709 [==============================] - 16s 540us/step - loss: 0.0407 - acc: 0.9881 - val_loss: 6.1265 - val_acc: 0.3984\n",
      "Epoch 533/1000\n",
      "28709/28709 [==============================] - 14s 470us/step - loss: 0.3901 - acc: 0.9642 - val_loss: 8.3574 - val_acc: 0.2913\n",
      "Epoch 534/1000\n",
      "28709/28709 [==============================] - 10s 337us/step - loss: 0.2571 - acc: 0.9617 - val_loss: 6.1855 - val_acc: 0.3965\n",
      "Epoch 535/1000\n",
      "28709/28709 [==============================] - 17s 589us/step - loss: 0.7788 - acc: 0.9302 - val_loss: 5.7803 - val_acc: 0.3599\n",
      "Epoch 536/1000\n",
      "28709/28709 [==============================] - 13s 454us/step - loss: 0.0542 - acc: 0.9849 - val_loss: 5.9701 - val_acc: 0.3997\n",
      "Epoch 537/1000\n",
      "28709/28709 [==============================] - 11s 390us/step - loss: 0.6202 - acc: 0.9241 - val_loss: 5.0537 - val_acc: 0.3967\n",
      "Epoch 538/1000\n",
      "28709/28709 [==============================] - 17s 609us/step - loss: 0.0097 - acc: 0.9973 - val_loss: 5.3849 - val_acc: 0.3945\n",
      "Epoch 539/1000\n",
      "28709/28709 [==============================] - 11s 394us/step - loss: 1.1009 - acc: 0.8903 - val_loss: 5.3477 - val_acc: 0.3940\n",
      "Epoch 540/1000\n",
      "28709/28709 [==============================] - 13s 446us/step - loss: 0.0148 - acc: 0.9961 - val_loss: 5.7758 - val_acc: 0.3903\n",
      "Epoch 541/1000\n",
      "28709/28709 [==============================] - 17s 594us/step - loss: 0.7869 - acc: 0.9104 - val_loss: 5.2874 - val_acc: 0.4007\n",
      "Epoch 542/1000\n",
      "28709/28709 [==============================] - 12s 418us/step - loss: 0.0092 - acc: 0.9973 - val_loss: 5.5450 - val_acc: 0.3947\n",
      "Epoch 543/1000\n",
      "28709/28709 [==============================] - 13s 436us/step - loss: 0.7483 - acc: 0.9186 - val_loss: 5.2979 - val_acc: 0.4006\n",
      "Epoch 544/1000\n",
      "28709/28709 [==============================] - 16s 564us/step - loss: 0.0125 - acc: 0.9964 - val_loss: 6.1515 - val_acc: 0.3986\n",
      "Epoch 545/1000\n",
      "28709/28709 [==============================] - 10s 365us/step - loss: 0.7889 - acc: 0.9176 - val_loss: 5.3403 - val_acc: 0.4007\n",
      "Epoch 546/1000\n",
      "28709/28709 [==============================] - 16s 546us/step - loss: 0.0107 - acc: 0.9967 - val_loss: 5.7381 - val_acc: 0.4000\n",
      "Epoch 547/1000\n",
      "28709/28709 [==============================] - 14s 495us/step - loss: 0.9638 - acc: 0.9040 - val_loss: 5.3930 - val_acc: 0.4024\n",
      "Epoch 548/1000\n",
      "28709/28709 [==============================] - 10s 355us/step - loss: 0.0095 - acc: 0.9968 - val_loss: 6.1085 - val_acc: 0.3956\n",
      "Epoch 549/1000\n",
      "28709/28709 [==============================] - 17s 603us/step - loss: 0.8429 - acc: 0.9074 - val_loss: 5.4858 - val_acc: 0.3960\n",
      "Epoch 550/1000\n",
      "28709/28709 [==============================] - 11s 396us/step - loss: 0.0096 - acc: 0.9967 - val_loss: 6.5001 - val_acc: 0.3775\n",
      "Epoch 551/1000\n",
      "28709/28709 [==============================] - 14s 491us/step - loss: 0.5316 - acc: 0.9321 - val_loss: 6.4278 - val_acc: 0.3901\n",
      "Epoch 552/1000\n",
      "28709/28709 [==============================] - 15s 528us/step - loss: 0.0098 - acc: 0.9965 - val_loss: 6.2771 - val_acc: 0.4052\n",
      "Epoch 553/1000\n",
      "28709/28709 [==============================] - 10s 333us/step - loss: 0.7169 - acc: 0.9183 - val_loss: 5.4367 - val_acc: 0.3865\n",
      "Epoch 554/1000\n",
      "28709/28709 [==============================] - 15s 519us/step - loss: 0.0116 - acc: 0.9965 - val_loss: 5.8111 - val_acc: 0.3978\n",
      "Epoch 555/1000\n",
      "28709/28709 [==============================] - 13s 455us/step - loss: 1.5185 - acc: 0.8693 - val_loss: 5.6611 - val_acc: 0.3982\n",
      "Epoch 556/1000\n",
      "28709/28709 [==============================] - 11s 394us/step - loss: 0.0103 - acc: 0.9966 - val_loss: 6.2761 - val_acc: 0.4017\n",
      "Epoch 557/1000\n",
      "28709/28709 [==============================] - 16s 548us/step - loss: 0.5550 - acc: 0.9313 - val_loss: 5.2851 - val_acc: 0.4037\n",
      "Epoch 558/1000\n",
      "28709/28709 [==============================] - 9s 310us/step - loss: 0.0098 - acc: 0.9967 - val_loss: 5.6647 - val_acc: 0.3984\n",
      "Epoch 559/1000\n",
      "28709/28709 [==============================] - 15s 506us/step - loss: 0.8027 - acc: 0.9190 - val_loss: 5.3909 - val_acc: 0.3958\n",
      "Epoch 560/1000\n",
      "28709/28709 [==============================] - 10s 354us/step - loss: 0.0089 - acc: 0.9971 - val_loss: 5.8699 - val_acc: 0.3939\n",
      "Epoch 561/1000\n",
      "28709/28709 [==============================] - 13s 443us/step - loss: 0.7360 - acc: 0.9184 - val_loss: 6.4196 - val_acc: 0.3956\n",
      "Epoch 562/1000\n",
      "28709/28709 [==============================] - 12s 424us/step - loss: 0.0076 - acc: 0.9970 - val_loss: 6.3844 - val_acc: 0.3924\n",
      "Epoch 563/1000\n",
      "28709/28709 [==============================] - 11s 391us/step - loss: 0.5797 - acc: 0.9330 - val_loss: 6.2122 - val_acc: 0.3932\n",
      "Epoch 564/1000\n",
      "28709/28709 [==============================] - 18s 612us/step - loss: 0.0081 - acc: 0.9969 - val_loss: 6.3920 - val_acc: 0.3912\n",
      "Epoch 565/1000\n",
      "28709/28709 [==============================] - 10s 350us/step - loss: 0.7071 - acc: 0.9258 - val_loss: 6.4358 - val_acc: 0.4009\n",
      "Epoch 566/1000\n",
      "28709/28709 [==============================] - 14s 474us/step - loss: 0.6391 - acc: 0.9233 - val_loss: 5.2097 - val_acc: 0.4007\n",
      "Epoch 567/1000\n",
      "28709/28709 [==============================] - 11s 397us/step - loss: 0.0167 - acc: 0.9955 - val_loss: 6.1262 - val_acc: 0.3953\n",
      "Epoch 568/1000\n",
      "28709/28709 [==============================] - 12s 420us/step - loss: 0.1947 - acc: 0.9770 - val_loss: 9.9694 - val_acc: 0.3015\n",
      "Epoch 569/1000\n",
      "28709/28709 [==============================] - 13s 451us/step - loss: 0.7575 - acc: 0.9248 - val_loss: 6.1596 - val_acc: 0.3952\n",
      "Epoch 570/1000\n",
      "28709/28709 [==============================] - 10s 363us/step - loss: 1.0512 - acc: 0.9021 - val_loss: 5.3724 - val_acc: 0.3939\n",
      "Epoch 571/1000\n",
      "28709/28709 [==============================] - 15s 510us/step - loss: 0.0201 - acc: 0.9938 - val_loss: 6.4925 - val_acc: 0.3952\n",
      "Epoch 572/1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "28709/28709 [==============================] - 9s 328us/step - loss: 0.6027 - acc: 0.9331 - val_loss: 5.0567 - val_acc: 0.3910\n",
      "Epoch 573/1000\n",
      "28709/28709 [==============================] - 15s 530us/step - loss: 0.0125 - acc: 0.9966 - val_loss: 6.4886 - val_acc: 0.3984\n",
      "Epoch 574/1000\n",
      "28709/28709 [==============================] - 9s 304us/step - loss: 0.7851 - acc: 0.9138 - val_loss: 5.2192 - val_acc: 0.4013\n",
      "Epoch 575/1000\n",
      "28709/28709 [==============================] - 16s 562us/step - loss: 0.0116 - acc: 0.9965 - val_loss: 5.7170 - val_acc: 0.4045\n",
      "Epoch 576/1000\n",
      "28709/28709 [==============================] - 9s 307us/step - loss: 0.2045 - acc: 0.9767 - val_loss: 9.1260 - val_acc: 0.2926\n",
      "Epoch 577/1000\n",
      "28709/28709 [==============================] - 16s 561us/step - loss: 0.5056 - acc: 0.9387 - val_loss: 6.0579 - val_acc: 0.3957\n",
      "Epoch 578/1000\n",
      "28709/28709 [==============================] - 9s 326us/step - loss: 0.0098 - acc: 0.9966 - val_loss: 6.3552 - val_acc: 0.3952\n",
      "Epoch 579/1000\n",
      "28709/28709 [==============================] - 14s 499us/step - loss: 0.8681 - acc: 0.9097 - val_loss: 5.9483 - val_acc: 0.3995\n",
      "Epoch 580/1000\n",
      "28709/28709 [==============================] - 11s 387us/step - loss: 0.0088 - acc: 0.9968 - val_loss: 7.2358 - val_acc: 0.3447\n",
      "Epoch 581/1000\n",
      "28709/28709 [==============================] - 12s 429us/step - loss: 0.7873 - acc: 0.9091 - val_loss: 6.1134 - val_acc: 0.3965\n",
      "Epoch 582/1000\n",
      "28709/28709 [==============================] - 12s 432us/step - loss: 0.0279 - acc: 0.9929 - val_loss: 9.2650 - val_acc: 0.3051\n",
      "Epoch 583/1000\n",
      "28709/28709 [==============================] - 11s 371us/step - loss: 0.6594 - acc: 0.9193 - val_loss: 6.0206 - val_acc: 0.3995\n",
      "Epoch 584/1000\n",
      "28709/28709 [==============================] - 15s 535us/step - loss: 0.0098 - acc: 0.9963 - val_loss: 6.1104 - val_acc: 0.4007\n",
      "Epoch 585/1000\n",
      "28709/28709 [==============================] - 9s 328us/step - loss: 1.2091 - acc: 0.8796 - val_loss: 5.3647 - val_acc: 0.3915\n",
      "Epoch 586/1000\n",
      "28709/28709 [==============================] - 16s 565us/step - loss: 0.0115 - acc: 0.9966 - val_loss: 6.3102 - val_acc: 0.3968\n",
      "Epoch 587/1000\n",
      "28709/28709 [==============================] - 11s 366us/step - loss: 0.6526 - acc: 0.9316 - val_loss: 5.2109 - val_acc: 0.4014\n",
      "Epoch 588/1000\n",
      "28709/28709 [==============================] - 14s 472us/step - loss: 0.4624 - acc: 0.9536 - val_loss: 5.9774 - val_acc: 0.3621\n",
      "Epoch 589/1000\n",
      "28709/28709 [==============================] - 13s 451us/step - loss: 0.0525 - acc: 0.9867 - val_loss: 5.3396 - val_acc: 0.4035\n",
      "Epoch 590/1000\n",
      "28709/28709 [==============================] - 11s 379us/step - loss: 0.0336 - acc: 0.9913 - val_loss: 9.1892 - val_acc: 0.3244\n",
      "Epoch 591/1000\n",
      "28709/28709 [==============================] - 16s 555us/step - loss: 0.7427 - acc: 0.9199 - val_loss: 6.4201 - val_acc: 0.4000\n",
      "Epoch 592/1000\n",
      "28709/28709 [==============================] - 9s 300us/step - loss: 0.0089 - acc: 0.9967 - val_loss: 6.6156 - val_acc: 0.3919\n",
      "Epoch 593/1000\n",
      "28709/28709 [==============================] - 17s 584us/step - loss: 0.5283 - acc: 0.9434 - val_loss: 5.6334 - val_acc: 0.3926\n",
      "Epoch 594/1000\n",
      "28709/28709 [==============================] - 9s 323us/step - loss: 0.6365 - acc: 0.9285 - val_loss: 5.2563 - val_acc: 0.3908\n",
      "Epoch 595/1000\n",
      "28709/28709 [==============================] - 14s 480us/step - loss: 0.0155 - acc: 0.9956 - val_loss: 5.9399 - val_acc: 0.3999\n",
      "Epoch 596/1000\n",
      "28709/28709 [==============================] - 13s 436us/step - loss: 0.7358 - acc: 0.9230 - val_loss: 5.3884 - val_acc: 0.3997\n",
      "Epoch 597/1000\n",
      "28709/28709 [==============================] - 11s 377us/step - loss: 0.0120 - acc: 0.9964 - val_loss: 5.9688 - val_acc: 0.3970\n",
      "Epoch 598/1000\n",
      "28709/28709 [==============================] - 15s 539us/step - loss: 0.8506 - acc: 0.9143 - val_loss: 5.4001 - val_acc: 0.4032\n",
      "Epoch 599/1000\n",
      "28709/28709 [==============================] - 9s 307us/step - loss: 0.0161 - acc: 0.9951 - val_loss: 6.4112 - val_acc: 0.4052\n",
      "Epoch 600/1000\n",
      "28709/28709 [==============================] - 17s 592us/step - loss: 0.6594 - acc: 0.9278 - val_loss: 5.4670 - val_acc: 0.4049\n",
      "Epoch 601/1000\n",
      "28709/28709 [==============================] - 10s 338us/step - loss: 0.0135 - acc: 0.9959 - val_loss: 5.4877 - val_acc: 0.4055\n",
      "Epoch 602/1000\n",
      "28709/28709 [==============================] - 14s 483us/step - loss: 0.7299 - acc: 0.9239 - val_loss: 5.6215 - val_acc: 0.3841\n",
      "Epoch 603/1000\n",
      "28709/28709 [==============================] - 12s 421us/step - loss: 0.0132 - acc: 0.9962 - val_loss: 5.6068 - val_acc: 0.4002\n",
      "Epoch 604/1000\n",
      "28709/28709 [==============================] - 11s 366us/step - loss: 0.8030 - acc: 0.9223 - val_loss: 5.8267 - val_acc: 0.3922\n",
      "Epoch 605/1000\n",
      "28709/28709 [==============================] - 16s 565us/step - loss: 0.0160 - acc: 0.9947 - val_loss: 6.2960 - val_acc: 0.4000\n",
      "Epoch 606/1000\n",
      "28709/28709 [==============================] - 9s 325us/step - loss: 0.6340 - acc: 0.9325 - val_loss: 5.4866 - val_acc: 0.3991\n",
      "Epoch 607/1000\n",
      "28709/28709 [==============================] - 13s 464us/step - loss: 0.0113 - acc: 0.9967 - val_loss: 5.6840 - val_acc: 0.4003\n",
      "Epoch 608/1000\n",
      "28709/28709 [==============================] - 13s 445us/step - loss: 1.0159 - acc: 0.9183 - val_loss: 7.5770 - val_acc: 0.3677\n",
      "Epoch 609/1000\n",
      "28709/28709 [==============================] - 10s 337us/step - loss: 0.1803 - acc: 0.9697 - val_loss: 5.7995 - val_acc: 0.4017\n",
      "Epoch 610/1000\n",
      "28709/28709 [==============================] - 16s 562us/step - loss: 0.0109 - acc: 0.9962 - val_loss: 8.4819 - val_acc: 0.3203\n",
      "Epoch 611/1000\n",
      "28709/28709 [==============================] - 9s 318us/step - loss: 0.6680 - acc: 0.9271 - val_loss: 5.7674 - val_acc: 0.3954\n",
      "Epoch 612/1000\n",
      "28709/28709 [==============================] - 14s 492us/step - loss: 0.0077 - acc: 0.9968 - val_loss: 6.9789 - val_acc: 0.3840\n",
      "Epoch 613/1000\n",
      "28709/28709 [==============================] - 13s 452us/step - loss: 0.8467 - acc: 0.9068 - val_loss: 5.6330 - val_acc: 0.4016\n",
      "Epoch 614/1000\n",
      "28709/28709 [==============================] - 10s 357us/step - loss: 0.0088 - acc: 0.9969 - val_loss: 6.7177 - val_acc: 0.3896\n",
      "Epoch 615/1000\n",
      "28709/28709 [==============================] - 17s 608us/step - loss: 0.6744 - acc: 0.9287 - val_loss: 5.5233 - val_acc: 0.4018\n",
      "Epoch 616/1000\n",
      "28709/28709 [==============================] - 10s 363us/step - loss: 0.0081 - acc: 0.9971 - val_loss: 5.8557 - val_acc: 0.3925\n",
      "Epoch 617/1000\n",
      "28709/28709 [==============================] - 14s 480us/step - loss: 0.8350 - acc: 0.9114 - val_loss: 5.6792 - val_acc: 0.3981\n",
      "Epoch 618/1000\n",
      "28709/28709 [==============================] - 15s 516us/step - loss: 0.0087 - acc: 0.9969 - val_loss: 5.9503 - val_acc: 0.4010\n",
      "Epoch 619/1000\n",
      "28709/28709 [==============================] - 11s 376us/step - loss: 1.0679 - acc: 0.9018 - val_loss: 5.7833 - val_acc: 0.4004\n",
      "Epoch 620/1000\n",
      "28709/28709 [==============================] - 16s 544us/step - loss: 0.0082 - acc: 0.9968 - val_loss: 6.4569 - val_acc: 0.4021\n",
      "Epoch 621/1000\n",
      "28709/28709 [==============================] - 12s 418us/step - loss: 0.8183 - acc: 0.9190 - val_loss: 5.8675 - val_acc: 0.4009\n",
      "Epoch 622/1000\n",
      "28709/28709 [==============================] - 13s 456us/step - loss: 0.0108 - acc: 0.9960 - val_loss: 8.2848 - val_acc: 0.3086\n",
      "Epoch 623/1000\n",
      "28709/28709 [==============================] - 19s 655us/step - loss: 0.6282 - acc: 0.9290 - val_loss: 5.6631 - val_acc: 0.4031\n",
      "Epoch 624/1000\n",
      "28709/28709 [==============================] - 12s 411us/step - loss: 0.0085 - acc: 0.9970 - val_loss: 6.1705 - val_acc: 0.3942\n",
      "Epoch 625/1000\n",
      "28709/28709 [==============================] - 15s 513us/step - loss: 0.6965 - acc: 0.9232 - val_loss: 6.4129 - val_acc: 0.3981\n",
      "Epoch 626/1000\n",
      "28709/28709 [==============================] - 16s 573us/step - loss: 0.0079 - acc: 0.9970 - val_loss: 6.4343 - val_acc: 0.3981\n",
      "Epoch 627/1000\n",
      "28709/28709 [==============================] - 10s 354us/step - loss: 0.8037 - acc: 0.9120 - val_loss: 5.7855 - val_acc: 0.4028\n",
      "Epoch 628/1000\n",
      "28709/28709 [==============================] - 17s 592us/step - loss: 0.0093 - acc: 0.9966 - val_loss: 6.1371 - val_acc: 0.4048\n",
      "Epoch 629/1000\n",
      "28709/28709 [==============================] - 13s 440us/step - loss: 0.6467 - acc: 0.9297 - val_loss: 6.4820 - val_acc: 0.3992\n",
      "Epoch 630/1000\n",
      "28709/28709 [==============================] - 13s 464us/step - loss: 0.0086 - acc: 0.9967 - val_loss: 6.5712 - val_acc: 0.3956\n",
      "Epoch 631/1000\n",
      "28709/28709 [==============================] - 16s 547us/step - loss: 0.7189 - acc: 0.9195 - val_loss: 5.7480 - val_acc: 0.3825\n",
      "Epoch 632/1000\n",
      "28709/28709 [==============================] - 10s 349us/step - loss: 0.0142 - acc: 0.9957 - val_loss: 5.5983 - val_acc: 0.4018\n",
      "Epoch 633/1000\n",
      "28709/28709 [==============================] - 17s 577us/step - loss: 0.7933 - acc: 0.9225 - val_loss: 6.1470 - val_acc: 0.3919\n",
      "Epoch 634/1000\n",
      "28709/28709 [==============================] - 9s 319us/step - loss: 0.0352 - acc: 0.9901 - val_loss: 5.7337 - val_acc: 0.3977\n",
      "Epoch 635/1000\n",
      "28709/28709 [==============================] - 15s 529us/step - loss: 0.2869 - acc: 0.9696 - val_loss: 8.3036 - val_acc: 0.2913\n",
      "Epoch 636/1000\n",
      "28709/28709 [==============================] - 12s 417us/step - loss: 0.2276 - acc: 0.9652 - val_loss: 5.7878 - val_acc: 0.3968\n",
      "Epoch 637/1000\n",
      "28709/28709 [==============================] - 11s 395us/step - loss: 0.0196 - acc: 0.9948 - val_loss: 10.2780 - val_acc: 0.2501\n",
      "Epoch 638/1000\n",
      "28709/28709 [==============================] - 13s 463us/step - loss: 0.6167 - acc: 0.9314 - val_loss: 6.0132 - val_acc: 0.4004\n",
      "Epoch 639/1000\n",
      "28709/28709 [==============================] - 10s 355us/step - loss: 0.0084 - acc: 0.9970 - val_loss: 6.4914 - val_acc: 0.3783\n",
      "Epoch 640/1000\n",
      "28709/28709 [==============================] - 14s 487us/step - loss: 0.7742 - acc: 0.9213 - val_loss: 5.7809 - val_acc: 0.3999\n",
      "Epoch 641/1000\n",
      "28709/28709 [==============================] - 9s 328us/step - loss: 0.0081 - acc: 0.9972 - val_loss: 6.0218 - val_acc: 0.3957\n",
      "Epoch 642/1000\n",
      "28709/28709 [==============================] - 15s 514us/step - loss: 0.7080 - acc: 0.9252 - val_loss: 5.7663 - val_acc: 0.3949\n",
      "Epoch 643/1000\n",
      "28709/28709 [==============================] - 8s 291us/step - loss: 0.0078 - acc: 0.9970 - val_loss: 5.9460 - val_acc: 0.3995\n",
      "Epoch 644/1000\n",
      "28709/28709 [==============================] - 16s 549us/step - loss: 0.7232 - acc: 0.9218 - val_loss: 5.8110 - val_acc: 0.3939\n",
      "Epoch 645/1000\n",
      "28709/28709 [==============================] - 8s 278us/step - loss: 0.0084 - acc: 0.9970 - val_loss: 6.4865 - val_acc: 0.3974\n",
      "Epoch 646/1000\n",
      "28709/28709 [==============================] - 16s 549us/step - loss: 0.7148 - acc: 0.9236 - val_loss: 5.6092 - val_acc: 0.4024\n",
      "Epoch 647/1000\n",
      "28709/28709 [==============================] - 8s 290us/step - loss: 0.0085 - acc: 0.9972 - val_loss: 5.9117 - val_acc: 0.4009\n",
      "Epoch 648/1000\n",
      "28709/28709 [==============================] - 16s 544us/step - loss: 0.7052 - acc: 0.9249 - val_loss: 5.6922 - val_acc: 0.3839\n",
      "Epoch 649/1000\n",
      "28709/28709 [==============================] - 8s 290us/step - loss: 0.0145 - acc: 0.9949 - val_loss: 6.3677 - val_acc: 0.3963\n",
      "Epoch 650/1000\n",
      "28709/28709 [==============================] - 14s 488us/step - loss: 0.7397 - acc: 0.9259 - val_loss: 5.7449 - val_acc: 0.3816\n",
      "Epoch 651/1000\n",
      "28709/28709 [==============================] - 10s 351us/step - loss: 0.0142 - acc: 0.9953 - val_loss: 5.7978 - val_acc: 0.4009\n",
      "Epoch 652/1000\n",
      "28709/28709 [==============================] - 12s 417us/step - loss: 0.9538 - acc: 0.9228 - val_loss: 6.2076 - val_acc: 0.3853\n",
      "Epoch 653/1000\n",
      "28709/28709 [==============================] - 12s 413us/step - loss: 0.0472 - acc: 0.9865 - val_loss: 5.9040 - val_acc: 0.4011\n",
      "Epoch 654/1000\n",
      "28709/28709 [==============================] - 11s 371us/step - loss: 0.0555 - acc: 0.9900 - val_loss: 7.7885 - val_acc: 0.3538\n",
      "Epoch 655/1000\n",
      "28709/28709 [==============================] - 13s 468us/step - loss: 0.1799 - acc: 0.9729 - val_loss: 5.8888 - val_acc: 0.3960\n",
      "Epoch 656/1000\n",
      "28709/28709 [==============================] - 9s 307us/step - loss: 0.0261 - acc: 0.9935 - val_loss: 6.7502 - val_acc: 0.3731\n",
      "Epoch 657/1000\n",
      "28709/28709 [==============================] - 16s 540us/step - loss: 0.2611 - acc: 0.9565 - val_loss: 5.8021 - val_acc: 0.4025\n",
      "Epoch 658/1000\n",
      "28709/28709 [==============================] - 8s 274us/step - loss: 0.4184 - acc: 0.9567 - val_loss: 8.4944 - val_acc: 0.3331\n",
      "Epoch 659/1000\n",
      "28709/28709 [==============================] - 15s 531us/step - loss: 0.2182 - acc: 0.9712 - val_loss: 6.3074 - val_acc: 0.3996\n",
      "Epoch 660/1000\n",
      "28709/28709 [==============================] - 9s 305us/step - loss: 0.0084 - acc: 0.9964 - val_loss: 7.0249 - val_acc: 0.3641\n",
      "Epoch 661/1000\n",
      "28709/28709 [==============================] - 13s 462us/step - loss: 0.8429 - acc: 0.9138 - val_loss: 6.7845 - val_acc: 0.3963\n",
      "Epoch 662/1000\n",
      "28709/28709 [==============================] - 10s 358us/step - loss: 0.0089 - acc: 0.9966 - val_loss: 6.9265 - val_acc: 0.3869\n",
      "Epoch 663/1000\n",
      "28709/28709 [==============================] - 12s 407us/step - loss: 0.7175 - acc: 0.9298 - val_loss: 6.3159 - val_acc: 0.3956\n",
      "Epoch 664/1000\n",
      "28709/28709 [==============================] - 12s 429us/step - loss: 0.1716 - acc: 0.9807 - val_loss: 9.3567 - val_acc: 0.2704\n",
      "Epoch 665/1000\n",
      "28709/28709 [==============================] - 9s 322us/step - loss: 0.3788 - acc: 0.9542 - val_loss: 6.7373 - val_acc: 0.3847\n",
      "Epoch 666/1000\n",
      "28709/28709 [==============================] - 15s 512us/step - loss: 0.4068 - acc: 0.9615 - val_loss: 7.6343 - val_acc: 0.3428\n",
      "Epoch 667/1000\n",
      "28709/28709 [==============================] - 8s 296us/step - loss: 0.2299 - acc: 0.9633 - val_loss: 6.5880 - val_acc: 0.3988\n",
      "Epoch 668/1000\n",
      "28709/28709 [==============================] - 13s 444us/step - loss: 0.6818 - acc: 0.9272 - val_loss: 5.5831 - val_acc: 0.3924\n",
      "Epoch 669/1000\n",
      "28709/28709 [==============================] - 11s 374us/step - loss: 0.0099 - acc: 0.9972 - val_loss: 5.7914 - val_acc: 0.3988\n",
      "Epoch 670/1000\n",
      "28709/28709 [==============================] - 10s 365us/step - loss: 0.0413 - acc: 0.9919 - val_loss: 8.5637 - val_acc: 0.3479\n",
      "Epoch 671/1000\n",
      "28709/28709 [==============================] - 13s 458us/step - loss: 0.4542 - acc: 0.9442 - val_loss: 5.9593 - val_acc: 0.3872\n",
      "Epoch 672/1000\n",
      "28709/28709 [==============================] - 8s 292us/step - loss: 0.0174 - acc: 0.9950 - val_loss: 8.1677 - val_acc: 0.3631\n",
      "Epoch 673/1000\n",
      "28709/28709 [==============================] - 16s 553us/step - loss: 0.3339 - acc: 0.9592 - val_loss: 6.6843 - val_acc: 0.4063\n",
      "Epoch 674/1000\n",
      "28709/28709 [==============================] - 8s 268us/step - loss: 0.1497 - acc: 0.9825 - val_loss: 9.9779 - val_acc: 0.2514\n",
      "Epoch 675/1000\n",
      "28709/28709 [==============================] - 13s 458us/step - loss: 0.3691 - acc: 0.9544 - val_loss: 6.6450 - val_acc: 0.3970\n",
      "Epoch 676/1000\n",
      "28709/28709 [==============================] - 10s 354us/step - loss: 0.0084 - acc: 0.9968 - val_loss: 6.5803 - val_acc: 0.4035\n",
      "Epoch 677/1000\n",
      "28709/28709 [==============================] - 11s 399us/step - loss: 1.0154 - acc: 0.9028 - val_loss: 7.0060 - val_acc: 0.3894\n",
      "Epoch 678/1000\n",
      "28709/28709 [==============================] - 11s 400us/step - loss: 0.0104 - acc: 0.9963 - val_loss: 6.5971 - val_acc: 0.4045\n",
      "Epoch 679/1000\n",
      "28709/28709 [==============================] - 10s 340us/step - loss: 0.6620 - acc: 0.9280 - val_loss: 5.7332 - val_acc: 0.4041\n",
      "Epoch 680/1000\n",
      "28709/28709 [==============================] - 13s 467us/step - loss: 0.0074 - acc: 0.9972 - val_loss: 6.3679 - val_acc: 0.4037\n",
      "Epoch 681/1000\n",
      "28709/28709 [==============================] - 8s 285us/step - loss: 0.8979 - acc: 0.9170 - val_loss: 5.8009 - val_acc: 0.4024\n",
      "Epoch 682/1000\n",
      "28709/28709 [==============================] - 14s 500us/step - loss: 0.0095 - acc: 0.9970 - val_loss: 6.0671 - val_acc: 0.3853\n",
      "Epoch 683/1000\n",
      "28709/28709 [==============================] - 9s 305us/step - loss: 0.0125 - acc: 0.9961 - val_loss: 6.1040 - val_acc: 0.3995\n",
      "Epoch 684/1000\n",
      "28709/28709 [==============================] - 12s 424us/step - loss: 0.8075 - acc: 0.9131 - val_loss: 5.9319 - val_acc: 0.3997\n",
      "Epoch 685/1000\n",
      "28709/28709 [==============================] - 11s 376us/step - loss: 0.0076 - acc: 0.9970 - val_loss: 6.7149 - val_acc: 0.4031\n",
      "Epoch 686/1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "28709/28709 [==============================] - 10s 355us/step - loss: 0.8120 - acc: 0.9155 - val_loss: 5.8738 - val_acc: 0.3993\n",
      "Epoch 687/1000\n",
      "28709/28709 [==============================] - 13s 460us/step - loss: 0.0077 - acc: 0.9968 - val_loss: 6.6378 - val_acc: 0.3991\n",
      "Epoch 688/1000\n",
      "28709/28709 [==============================] - 8s 277us/step - loss: 0.6438 - acc: 0.9318 - val_loss: 5.8321 - val_acc: 0.3974\n",
      "Epoch 689/1000\n",
      "28709/28709 [==============================] - 14s 472us/step - loss: 0.0080 - acc: 0.9969 - val_loss: 6.0111 - val_acc: 0.3988\n",
      "Epoch 690/1000\n",
      "28709/28709 [==============================] - 10s 334us/step - loss: 0.6442 - acc: 0.9333 - val_loss: 5.7878 - val_acc: 0.3967\n",
      "Epoch 691/1000\n",
      "28709/28709 [==============================] - 11s 375us/step - loss: 0.0077 - acc: 0.9974 - val_loss: 5.8580 - val_acc: 0.4021\n",
      "Epoch 692/1000\n",
      "28709/28709 [==============================] - 13s 457us/step - loss: 0.8934 - acc: 0.9125 - val_loss: 5.9398 - val_acc: 0.3981\n",
      "Epoch 693/1000\n",
      "28709/28709 [==============================] - 8s 278us/step - loss: 0.0109 - acc: 0.9966 - val_loss: 5.9807 - val_acc: 0.4003\n",
      "Epoch 694/1000\n",
      "28709/28709 [==============================] - 14s 482us/step - loss: 0.6375 - acc: 0.9351 - val_loss: 5.8520 - val_acc: 0.3989\n",
      "Epoch 695/1000\n",
      "28709/28709 [==============================] - 10s 356us/step - loss: 0.0101 - acc: 0.9967 - val_loss: 6.2501 - val_acc: 0.4049\n",
      "Epoch 696/1000\n",
      "28709/28709 [==============================] - 11s 382us/step - loss: 0.0089 - acc: 0.9966 - val_loss: 6.4406 - val_acc: 0.4045\n",
      "Epoch 697/1000\n",
      "28709/28709 [==============================] - 15s 518us/step - loss: 0.9433 - acc: 0.9099 - val_loss: 5.9825 - val_acc: 0.3981\n",
      "Epoch 698/1000\n",
      "28709/28709 [==============================] - 8s 264us/step - loss: 0.0078 - acc: 0.9972 - val_loss: 6.1212 - val_acc: 0.4000\n",
      "Epoch 699/1000\n",
      "28709/28709 [==============================] - 13s 463us/step - loss: 0.5201 - acc: 0.9384 - val_loss: 5.8955 - val_acc: 0.4007\n",
      "Epoch 700/1000\n",
      "28709/28709 [==============================] - 10s 356us/step - loss: 0.0081 - acc: 0.9971 - val_loss: 5.9843 - val_acc: 0.4016\n",
      "Epoch 701/1000\n",
      "28709/28709 [==============================] - 11s 375us/step - loss: 0.6446 - acc: 0.9306 - val_loss: 5.9306 - val_acc: 0.4011\n",
      "Epoch 702/1000\n",
      "28709/28709 [==============================] - 13s 452us/step - loss: 0.0078 - acc: 0.9971 - val_loss: 6.0623 - val_acc: 0.4016\n",
      "Epoch 703/1000\n",
      "28709/28709 [==============================] - 8s 269us/step - loss: 0.6706 - acc: 0.9324 - val_loss: 5.9102 - val_acc: 0.3924\n",
      "Epoch 704/1000\n",
      "28709/28709 [==============================] - 15s 523us/step - loss: 0.0093 - acc: 0.9970 - val_loss: 6.0383 - val_acc: 0.4004\n",
      "Epoch 705/1000\n",
      "28709/28709 [==============================] - 8s 276us/step - loss: 0.7182 - acc: 0.9332 - val_loss: 6.6737 - val_acc: 0.3612\n",
      "Epoch 706/1000\n",
      "28709/28709 [==============================] - 12s 427us/step - loss: 0.0438 - acc: 0.9886 - val_loss: 6.7785 - val_acc: 0.4078\n",
      "Epoch 707/1000\n",
      "28709/28709 [==============================] - 11s 392us/step - loss: 0.0083 - acc: 0.9968 - val_loss: 6.5526 - val_acc: 0.3938\n",
      "Epoch 708/1000\n",
      "28709/28709 [==============================] - 9s 305us/step - loss: 0.6894 - acc: 0.9305 - val_loss: 6.0775 - val_acc: 0.3996\n",
      "Epoch 709/1000\n",
      "28709/28709 [==============================] - 14s 499us/step - loss: 0.0077 - acc: 0.9970 - val_loss: 6.5145 - val_acc: 0.3939\n",
      "Epoch 710/1000\n",
      "28709/28709 [==============================] - 8s 292us/step - loss: 0.5242 - acc: 0.9457 - val_loss: 6.0141 - val_acc: 0.4030\n",
      "Epoch 711/1000\n",
      "28709/28709 [==============================] - 12s 406us/step - loss: 0.0078 - acc: 0.9970 - val_loss: 6.3572 - val_acc: 0.3961\n",
      "Epoch 712/1000\n",
      "28709/28709 [==============================] - 11s 386us/step - loss: 0.6865 - acc: 0.9273 - val_loss: 6.0802 - val_acc: 0.4031\n",
      "Epoch 713/1000\n",
      "28709/28709 [==============================] - 9s 312us/step - loss: 0.0082 - acc: 0.9969 - val_loss: 6.3316 - val_acc: 0.3992\n",
      "Epoch 714/1000\n",
      "28709/28709 [==============================] - 14s 498us/step - loss: 0.5079 - acc: 0.9378 - val_loss: 5.9299 - val_acc: 0.3997\n",
      "Epoch 715/1000\n",
      "28709/28709 [==============================] - 8s 270us/step - loss: 0.0071 - acc: 0.9974 - val_loss: 6.0521 - val_acc: 0.3992\n",
      "Epoch 716/1000\n",
      "28709/28709 [==============================] - 12s 431us/step - loss: 0.8206 - acc: 0.9301 - val_loss: 7.2895 - val_acc: 0.3762\n",
      "Epoch 717/1000\n",
      "28709/28709 [==============================] - 11s 381us/step - loss: 0.0737 - acc: 0.9856 - val_loss: 6.1753 - val_acc: 0.4000\n",
      "Epoch 718/1000\n",
      "28709/28709 [==============================] - 9s 307us/step - loss: 0.1823 - acc: 0.9785 - val_loss: 8.5699 - val_acc: 0.3453\n",
      "Epoch 719/1000\n",
      "28709/28709 [==============================] - 14s 497us/step - loss: 0.3580 - acc: 0.9547 - val_loss: 6.1358 - val_acc: 0.4018\n",
      "Epoch 720/1000\n",
      "28709/28709 [==============================] - 7s 248us/step - loss: 0.0162 - acc: 0.9956 - val_loss: 8.5503 - val_acc: 0.3060\n",
      "Epoch 721/1000\n",
      "28709/28709 [==============================] - 13s 463us/step - loss: 0.3790 - acc: 0.9540 - val_loss: 6.0874 - val_acc: 0.4011\n",
      "Epoch 722/1000\n",
      "28709/28709 [==============================] - 10s 346us/step - loss: 0.0071 - acc: 0.9971 - val_loss: 6.4987 - val_acc: 0.3992\n",
      "Epoch 723/1000\n",
      "28709/28709 [==============================] - 9s 329us/step - loss: 0.6356 - acc: 0.9310 - val_loss: 6.0874 - val_acc: 0.4018\n",
      "Epoch 724/1000\n",
      "28709/28709 [==============================] - 14s 485us/step - loss: 0.0072 - acc: 0.9969 - val_loss: 6.1405 - val_acc: 0.4030\n",
      "Epoch 725/1000\n",
      "28709/28709 [==============================] - 7s 250us/step - loss: 0.6337 - acc: 0.9333 - val_loss: 5.9337 - val_acc: 0.3988\n",
      "Epoch 726/1000\n",
      "28709/28709 [==============================] - 13s 461us/step - loss: 0.0074 - acc: 0.9972 - val_loss: 6.2170 - val_acc: 0.3996\n",
      "Epoch 727/1000\n",
      "28709/28709 [==============================] - 10s 345us/step - loss: 0.9487 - acc: 0.9104 - val_loss: 6.1610 - val_acc: 0.3991\n",
      "Epoch 728/1000\n",
      "28709/28709 [==============================] - 10s 340us/step - loss: 0.0138 - acc: 0.9954 - val_loss: 6.6658 - val_acc: 0.4009\n",
      "Epoch 729/1000\n",
      "28709/28709 [==============================] - 13s 465us/step - loss: 0.0563 - acc: 0.9890 - val_loss: 9.0286 - val_acc: 0.3237\n",
      "Epoch 730/1000\n",
      "28709/28709 [==============================] - 8s 264us/step - loss: 0.6876 - acc: 0.9327 - val_loss: 6.6491 - val_acc: 0.3971\n",
      "Epoch 731/1000\n",
      "28709/28709 [==============================] - 12s 429us/step - loss: 0.0084 - acc: 0.9965 - val_loss: 6.7436 - val_acc: 0.3950\n",
      "Epoch 732/1000\n",
      "28709/28709 [==============================] - 11s 372us/step - loss: 0.5911 - acc: 0.9323 - val_loss: 5.9702 - val_acc: 0.4002\n",
      "Epoch 733/1000\n",
      "28709/28709 [==============================] - 9s 322us/step - loss: 0.0068 - acc: 0.9970 - val_loss: 6.3815 - val_acc: 0.4071\n",
      "Epoch 734/1000\n",
      "28709/28709 [==============================] - 14s 477us/step - loss: 0.6398 - acc: 0.9337 - val_loss: 6.0897 - val_acc: 0.3921\n",
      "Epoch 735/1000\n",
      "28709/28709 [==============================] - 8s 263us/step - loss: 0.0076 - acc: 0.9970 - val_loss: 6.1038 - val_acc: 0.3995\n",
      "Epoch 736/1000\n",
      "28709/28709 [==============================] - 12s 419us/step - loss: 0.5986 - acc: 0.9348 - val_loss: 6.0848 - val_acc: 0.3988\n",
      "Epoch 737/1000\n",
      "28709/28709 [==============================] - 11s 377us/step - loss: 0.0070 - acc: 0.9973 - val_loss: 6.1671 - val_acc: 0.4014\n",
      "Epoch 738/1000\n",
      "28709/28709 [==============================] - 9s 313us/step - loss: 0.8276 - acc: 0.9218 - val_loss: 6.2497 - val_acc: 0.4098\n",
      "Epoch 739/1000\n",
      "28709/28709 [==============================] - 14s 489us/step - loss: 0.0106 - acc: 0.9961 - val_loss: 6.4430 - val_acc: 0.4064\n",
      "Epoch 740/1000\n",
      "28709/28709 [==============================] - 7s 249us/step - loss: 0.4219 - acc: 0.9599 - val_loss: 7.3846 - val_acc: 0.3653\n",
      "Epoch 741/1000\n",
      "28709/28709 [==============================] - 13s 442us/step - loss: 0.0826 - acc: 0.9823 - val_loss: 6.2271 - val_acc: 0.4016\n",
      "Epoch 742/1000\n",
      "28709/28709 [==============================] - 10s 345us/step - loss: 0.0078 - acc: 0.9969 - val_loss: 6.5919 - val_acc: 0.4023\n",
      "Epoch 743/1000\n",
      "28709/28709 [==============================] - 10s 341us/step - loss: 0.8176 - acc: 0.9185 - val_loss: 6.5319 - val_acc: 0.4032\n",
      "Epoch 744/1000\n",
      "28709/28709 [==============================] - 13s 467us/step - loss: 0.0068 - acc: 0.9971 - val_loss: 6.4366 - val_acc: 0.3940\n",
      "Epoch 745/1000\n",
      "28709/28709 [==============================] - 7s 235us/step - loss: 0.8093 - acc: 0.9202 - val_loss: 6.1852 - val_acc: 0.3977\n",
      "Epoch 746/1000\n",
      "28709/28709 [==============================] - 14s 475us/step - loss: 0.0074 - acc: 0.9969 - val_loss: 6.7284 - val_acc: 0.4021\n",
      "Epoch 747/1000\n",
      "28709/28709 [==============================] - 10s 331us/step - loss: 0.7581 - acc: 0.9241 - val_loss: 6.0455 - val_acc: 0.4069\n",
      "Epoch 748/1000\n",
      "28709/28709 [==============================] - 10s 355us/step - loss: 0.0071 - acc: 0.9972 - val_loss: 6.1637 - val_acc: 0.4041\n",
      "Epoch 749/1000\n",
      "28709/28709 [==============================] - 12s 433us/step - loss: 0.7652 - acc: 0.9265 - val_loss: 6.1492 - val_acc: 0.4045\n",
      "Epoch 750/1000\n",
      "28709/28709 [==============================] - 7s 254us/step - loss: 0.0087 - acc: 0.9966 - val_loss: 6.7641 - val_acc: 0.4021\n",
      "Epoch 751/1000\n",
      "28709/28709 [==============================] - 14s 479us/step - loss: 0.8380 - acc: 0.9275 - val_loss: 6.7093 - val_acc: 0.3839\n",
      "Epoch 752/1000\n",
      "28709/28709 [==============================] - 11s 372us/step - loss: 0.0296 - acc: 0.9916 - val_loss: 6.6508 - val_acc: 0.3975\n",
      "Epoch 753/1000\n",
      "28709/28709 [==============================] - 11s 396us/step - loss: 0.6132 - acc: 0.9442 - val_loss: 6.8831 - val_acc: 0.3726\n",
      "Epoch 754/1000\n",
      "28709/28709 [==============================] - 14s 481us/step - loss: 0.0551 - acc: 0.9855 - val_loss: 6.2894 - val_acc: 0.3999\n",
      "Epoch 755/1000\n",
      "28709/28709 [==============================] - 8s 290us/step - loss: 0.0080 - acc: 0.9970 - val_loss: 6.4853 - val_acc: 0.3896\n",
      "Epoch 756/1000\n",
      "28709/28709 [==============================] - 12s 416us/step - loss: 1.4625 - acc: 0.8733 - val_loss: 6.5768 - val_acc: 0.4010\n",
      "Epoch 757/1000\n",
      "28709/28709 [==============================] - 12s 431us/step - loss: 0.0086 - acc: 0.9968 - val_loss: 6.8227 - val_acc: 0.3975\n",
      "Epoch 758/1000\n",
      "28709/28709 [==============================] - 8s 271us/step - loss: 0.7092 - acc: 0.9254 - val_loss: 6.2148 - val_acc: 0.3981\n",
      "Epoch 759/1000\n",
      "28709/28709 [==============================] - 14s 499us/step - loss: 0.0072 - acc: 0.9970 - val_loss: 6.6452 - val_acc: 0.3963\n",
      "Epoch 760/1000\n",
      "28709/28709 [==============================] - 10s 350us/step - loss: 1.0051 - acc: 0.9105 - val_loss: 6.3969 - val_acc: 0.3952\n",
      "Epoch 761/1000\n",
      "28709/28709 [==============================] - 10s 347us/step - loss: 0.0094 - acc: 0.9969 - val_loss: 6.3798 - val_acc: 0.4011\n",
      "Epoch 762/1000\n",
      "28709/28709 [==============================] - 14s 495us/step - loss: 0.7253 - acc: 0.9296 - val_loss: 6.1585 - val_acc: 0.4037\n",
      "Epoch 763/1000\n",
      "28709/28709 [==============================] - 8s 279us/step - loss: 0.0082 - acc: 0.9968 - val_loss: 6.2335 - val_acc: 0.4028\n",
      "Epoch 764/1000\n",
      "28709/28709 [==============================] - 12s 420us/step - loss: 0.2119 - acc: 0.9754 - val_loss: 9.4409 - val_acc: 0.3005\n",
      "Epoch 765/1000\n",
      "28709/28709 [==============================] - 13s 439us/step - loss: 0.3361 - acc: 0.9591 - val_loss: 6.3035 - val_acc: 0.4048\n",
      "Epoch 766/1000\n",
      "28709/28709 [==============================] - 9s 318us/step - loss: 0.0863 - acc: 0.9887 - val_loss: 8.3004 - val_acc: 0.3492\n",
      "Epoch 767/1000\n",
      "28709/28709 [==============================] - 15s 506us/step - loss: 0.2307 - acc: 0.9677 - val_loss: 6.3935 - val_acc: 0.4013\n",
      "Epoch 768/1000\n",
      "28709/28709 [==============================] - 11s 386us/step - loss: 0.5190 - acc: 0.9419 - val_loss: 6.1240 - val_acc: 0.4048\n",
      "Epoch 769/1000\n",
      "28709/28709 [==============================] - 10s 343us/step - loss: 0.0077 - acc: 0.9971 - val_loss: 6.6205 - val_acc: 0.4011\n",
      "Epoch 770/1000\n",
      "28709/28709 [==============================] - 15s 531us/step - loss: 0.6530 - acc: 0.9330 - val_loss: 6.0860 - val_acc: 0.3986\n",
      "Epoch 771/1000\n",
      "28709/28709 [==============================] - 10s 332us/step - loss: 0.0083 - acc: 0.9970 - val_loss: 6.1270 - val_acc: 0.4011\n",
      "Epoch 772/1000\n",
      "28709/28709 [==============================] - 12s 409us/step - loss: 0.4532 - acc: 0.9596 - val_loss: 9.5580 - val_acc: 0.3093\n",
      "Epoch 773/1000\n",
      "28709/28709 [==============================] - 14s 484us/step - loss: 0.5110 - acc: 0.9495 - val_loss: 6.3512 - val_acc: 0.4006\n",
      "Epoch 774/1000\n",
      "28709/28709 [==============================] - 9s 302us/step - loss: 0.0083 - acc: 0.9970 - val_loss: 6.4008 - val_acc: 0.3992\n",
      "Epoch 775/1000\n",
      "28709/28709 [==============================] - 13s 453us/step - loss: 0.6260 - acc: 0.9343 - val_loss: 6.2297 - val_acc: 0.4018\n",
      "Epoch 776/1000\n",
      "28709/28709 [==============================] - 13s 453us/step - loss: 0.0065 - acc: 0.9972 - val_loss: 6.6201 - val_acc: 0.3950\n",
      "Epoch 777/1000\n",
      "28709/28709 [==============================] - 8s 291us/step - loss: 0.5461 - acc: 0.9411 - val_loss: 6.2894 - val_acc: 0.4030\n",
      "Epoch 778/1000\n",
      "28709/28709 [==============================] - 14s 503us/step - loss: 0.0082 - acc: 0.9970 - val_loss: 6.3777 - val_acc: 0.4011\n",
      "Epoch 779/1000\n",
      "28709/28709 [==============================] - 12s 403us/step - loss: 0.6924 - acc: 0.9305 - val_loss: 6.2600 - val_acc: 0.4030\n",
      "Epoch 780/1000\n",
      "28709/28709 [==============================] - 9s 330us/step - loss: 0.0065 - acc: 0.9970 - val_loss: 6.5222 - val_acc: 0.4023\n",
      "Epoch 781/1000\n",
      "28709/28709 [==============================] - 14s 498us/step - loss: 0.6129 - acc: 0.9343 - val_loss: 6.2309 - val_acc: 0.3979\n",
      "Epoch 782/1000\n",
      "28709/28709 [==============================] - 11s 398us/step - loss: 0.0073 - acc: 0.9971 - val_loss: 6.2603 - val_acc: 0.3993\n",
      "Epoch 783/1000\n",
      "28709/28709 [==============================] - 10s 338us/step - loss: 0.5256 - acc: 0.9423 - val_loss: 6.2420 - val_acc: 0.4070\n",
      "Epoch 784/1000\n",
      "28709/28709 [==============================] - 15s 521us/step - loss: 0.0101 - acc: 0.9967 - val_loss: 6.2641 - val_acc: 0.4028\n",
      "Epoch 785/1000\n",
      "28709/28709 [==============================] - 10s 342us/step - loss: 0.0074 - acc: 0.9970 - val_loss: 6.4010 - val_acc: 0.4049\n",
      "Epoch 786/1000\n",
      "28709/28709 [==============================] - 11s 399us/step - loss: 0.6770 - acc: 0.9298 - val_loss: 6.2722 - val_acc: 0.3997\n",
      "Epoch 787/1000\n",
      "28709/28709 [==============================] - 15s 522us/step - loss: 0.0069 - acc: 0.9970 - val_loss: 6.4832 - val_acc: 0.3982\n",
      "Epoch 788/1000\n",
      "28709/28709 [==============================] - 9s 304us/step - loss: 1.6353 - acc: 0.8661 - val_loss: 6.3702 - val_acc: 0.4013\n",
      "Epoch 789/1000\n",
      "28709/28709 [==============================] - 12s 425us/step - loss: 0.0091 - acc: 0.9967 - val_loss: 6.4091 - val_acc: 0.3995\n",
      "Epoch 790/1000\n",
      "28709/28709 [==============================] - 14s 503us/step - loss: 0.0079 - acc: 0.9969 - val_loss: 6.5120 - val_acc: 0.3986\n",
      "Epoch 791/1000\n",
      "28709/28709 [==============================] - 9s 301us/step - loss: 0.6683 - acc: 0.9279 - val_loss: 6.2352 - val_acc: 0.4000\n",
      "Epoch 792/1000\n",
      "28709/28709 [==============================] - 13s 455us/step - loss: 0.0071 - acc: 0.9968 - val_loss: 6.3347 - val_acc: 0.3952\n",
      "Epoch 793/1000\n",
      "28709/28709 [==============================] - 14s 476us/step - loss: 0.5448 - acc: 0.9347 - val_loss: 6.1958 - val_acc: 0.4002\n",
      "Epoch 794/1000\n",
      "28709/28709 [==============================] - 9s 305us/step - loss: 0.0077 - acc: 0.9968 - val_loss: 6.2964 - val_acc: 0.4007\n",
      "Epoch 795/1000\n",
      "28709/28709 [==============================] - 14s 478us/step - loss: 0.5073 - acc: 0.9425 - val_loss: 6.1382 - val_acc: 0.4002\n",
      "Epoch 796/1000\n",
      "28709/28709 [==============================] - 12s 423us/step - loss: 0.0072 - acc: 0.9972 - val_loss: 6.2658 - val_acc: 0.3997\n",
      "Epoch 797/1000\n",
      "28709/28709 [==============================] - 9s 329us/step - loss: 0.6908 - acc: 0.9317 - val_loss: 6.3603 - val_acc: 0.3993\n",
      "Epoch 798/1000\n",
      "28709/28709 [==============================] - 16s 574us/step - loss: 0.0090 - acc: 0.9968 - val_loss: 6.3665 - val_acc: 0.3993\n",
      "Epoch 799/1000\n",
      "28709/28709 [==============================] - 13s 443us/step - loss: 0.0202 - acc: 0.9945 - val_loss: 9.7883 - val_acc: 0.3156\n",
      "Epoch 800/1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "28709/28709 [==============================] - 9s 321us/step - loss: 0.7411 - acc: 0.9289 - val_loss: 6.5346 - val_acc: 0.3993\n",
      "Epoch 801/1000\n",
      "28709/28709 [==============================] - 15s 515us/step - loss: 0.0071 - acc: 0.9974 - val_loss: 6.9501 - val_acc: 0.4059\n",
      "Epoch 802/1000\n",
      "28709/28709 [==============================] - 9s 323us/step - loss: 1.1856 - acc: 0.8970 - val_loss: 6.5814 - val_acc: 0.4020\n",
      "Epoch 803/1000\n",
      "28709/28709 [==============================] - 11s 386us/step - loss: 0.0081 - acc: 0.9969 - val_loss: 6.8209 - val_acc: 0.3981\n",
      "Epoch 804/1000\n",
      "28709/28709 [==============================] - 15s 512us/step - loss: 0.5883 - acc: 0.9386 - val_loss: 6.4465 - val_acc: 0.4006\n",
      "Epoch 805/1000\n",
      "28709/28709 [==============================] - 10s 353us/step - loss: 0.0076 - acc: 0.9969 - val_loss: 6.7149 - val_acc: 0.3988\n",
      "Epoch 806/1000\n",
      "28709/28709 [==============================] - 12s 405us/step - loss: 0.4153 - acc: 0.9469 - val_loss: 6.2169 - val_acc: 0.4024\n",
      "Epoch 807/1000\n",
      "28709/28709 [==============================] - 15s 523us/step - loss: 0.0069 - acc: 0.9972 - val_loss: 6.2765 - val_acc: 0.4010\n",
      "Epoch 808/1000\n",
      "28709/28709 [==============================] - 8s 273us/step - loss: 0.5623 - acc: 0.9445 - val_loss: 6.5895 - val_acc: 0.4007\n",
      "Epoch 809/1000\n",
      "28709/28709 [==============================] - 13s 453us/step - loss: 0.0083 - acc: 0.9966 - val_loss: 6.5683 - val_acc: 0.4098\n",
      "Epoch 810/1000\n",
      "28709/28709 [==============================] - 12s 419us/step - loss: 0.5129 - acc: 0.9407 - val_loss: 6.0393 - val_acc: 0.3979\n",
      "Epoch 811/1000\n",
      "28709/28709 [==============================] - 9s 309us/step - loss: 0.0076 - acc: 0.9970 - val_loss: 6.2613 - val_acc: 0.3894\n",
      "Epoch 812/1000\n",
      "28709/28709 [==============================] - 15s 537us/step - loss: 0.0291 - acc: 0.9925 - val_loss: 9.7697 - val_acc: 0.2841\n",
      "Epoch 813/1000\n",
      "28709/28709 [==============================] - 9s 299us/step - loss: 0.4844 - acc: 0.9475 - val_loss: 6.6936 - val_acc: 0.3964\n",
      "Epoch 814/1000\n",
      "28709/28709 [==============================] - 11s 399us/step - loss: 0.0077 - acc: 0.9968 - val_loss: 6.5936 - val_acc: 0.4032\n",
      "Epoch 815/1000\n",
      "28709/28709 [==============================] - 14s 473us/step - loss: 0.4886 - acc: 0.9429 - val_loss: 6.3008 - val_acc: 0.3924\n",
      "Epoch 816/1000\n",
      "28709/28709 [==============================] - 8s 279us/step - loss: 0.0082 - acc: 0.9971 - val_loss: 6.4664 - val_acc: 0.3943\n",
      "Epoch 817/1000\n",
      "28709/28709 [==============================] - 15s 509us/step - loss: 0.4283 - acc: 0.9471 - val_loss: 6.1025 - val_acc: 0.4048\n",
      "Epoch 818/1000\n",
      "28709/28709 [==============================] - 10s 343us/step - loss: 0.0068 - acc: 0.9972 - val_loss: 6.5953 - val_acc: 0.4003\n",
      "Epoch 819/1000\n",
      "28709/28709 [==============================] - 11s 384us/step - loss: 0.7151 - acc: 0.9269 - val_loss: 6.2716 - val_acc: 0.3978\n",
      "Epoch 820/1000\n",
      "28709/28709 [==============================] - 15s 516us/step - loss: 0.0075 - acc: 0.9969 - val_loss: 6.4174 - val_acc: 0.3979\n",
      "Epoch 821/1000\n",
      "28709/28709 [==============================] - 8s 279us/step - loss: 0.7485 - acc: 0.9288 - val_loss: 6.3745 - val_acc: 0.3988\n",
      "Epoch 822/1000\n",
      "28709/28709 [==============================] - 13s 457us/step - loss: 0.0079 - acc: 0.9970 - val_loss: 6.4006 - val_acc: 0.4030\n",
      "Epoch 823/1000\n",
      "28709/28709 [==============================] - 12s 427us/step - loss: 0.7300 - acc: 0.9304 - val_loss: 6.3661 - val_acc: 0.3979\n",
      "Epoch 824/1000\n",
      "28709/28709 [==============================] - 9s 300us/step - loss: 0.0068 - acc: 0.9970 - val_loss: 6.4043 - val_acc: 0.4000\n",
      "Epoch 825/1000\n",
      "28709/28709 [==============================] - 15s 529us/step - loss: 0.6185 - acc: 0.9366 - val_loss: 6.2233 - val_acc: 0.4021\n",
      "Epoch 826/1000\n",
      "28709/28709 [==============================] - 9s 310us/step - loss: 0.0067 - acc: 0.9972 - val_loss: 6.4140 - val_acc: 0.4006\n",
      "Epoch 827/1000\n",
      "28709/28709 [==============================] - 12s 413us/step - loss: 0.5849 - acc: 0.9401 - val_loss: 6.3090 - val_acc: 0.3963\n",
      "Epoch 828/1000\n",
      "28709/28709 [==============================] - 13s 459us/step - loss: 0.0080 - acc: 0.9971 - val_loss: 6.3247 - val_acc: 0.4042\n",
      "Epoch 829/1000\n",
      "28709/28709 [==============================] - 8s 282us/step - loss: 0.4171 - acc: 0.9600 - val_loss: 7.5930 - val_acc: 0.3620\n",
      "Epoch 830/1000\n",
      "28709/28709 [==============================] - 14s 490us/step - loss: 0.0879 - acc: 0.9833 - val_loss: 6.8938 - val_acc: 0.4018\n",
      "Epoch 831/1000\n",
      "28709/28709 [==============================] - 11s 390us/step - loss: 0.3885 - acc: 0.9526 - val_loss: 6.2014 - val_acc: 0.3945\n",
      "Epoch 832/1000\n",
      "28709/28709 [==============================] - 10s 332us/step - loss: 0.0073 - acc: 0.9970 - val_loss: 6.6579 - val_acc: 0.3988\n",
      "Epoch 833/1000\n",
      "28709/28709 [==============================] - 15s 523us/step - loss: 0.9383 - acc: 0.9106 - val_loss: 6.3663 - val_acc: 0.4030\n",
      "Epoch 834/1000\n",
      "28709/28709 [==============================] - 8s 285us/step - loss: 0.0079 - acc: 0.9970 - val_loss: 6.5554 - val_acc: 0.3940\n",
      "Epoch 835/1000\n",
      "28709/28709 [==============================] - 13s 436us/step - loss: 0.4353 - acc: 0.9467 - val_loss: 6.2100 - val_acc: 0.3929\n",
      "Epoch 836/1000\n",
      "28709/28709 [==============================] - 12s 426us/step - loss: 0.0070 - acc: 0.9974 - val_loss: 6.3551 - val_acc: 0.3953\n",
      "Epoch 837/1000\n",
      "28709/28709 [==============================] - 9s 298us/step - loss: 0.0095 - acc: 0.9965 - val_loss: 6.4314 - val_acc: 0.4049\n",
      "Epoch 838/1000\n",
      "28709/28709 [==============================] - 15s 517us/step - loss: 0.9218 - acc: 0.9138 - val_loss: 6.4672 - val_acc: 0.4011\n",
      "Epoch 839/1000\n",
      "28709/28709 [==============================] - 9s 317us/step - loss: 0.0077 - acc: 0.9968 - val_loss: 6.5462 - val_acc: 0.4055\n",
      "Epoch 840/1000\n",
      "28709/28709 [==============================] - 11s 399us/step - loss: 0.6458 - acc: 0.9289 - val_loss: 6.2727 - val_acc: 0.4023\n",
      "Epoch 841/1000\n",
      "28709/28709 [==============================] - 13s 457us/step - loss: 0.0073 - acc: 0.9972 - val_loss: 6.6413 - val_acc: 0.4046\n",
      "Epoch 842/1000\n",
      "28709/28709 [==============================] - 8s 280us/step - loss: 0.0891 - acc: 0.9858 - val_loss: 9.1979 - val_acc: 0.3481\n",
      "Epoch 843/1000\n",
      "28709/28709 [==============================] - 15s 525us/step - loss: 0.5346 - acc: 0.9447 - val_loss: 6.2949 - val_acc: 0.4028\n",
      "Epoch 844/1000\n",
      "28709/28709 [==============================] - 11s 380us/step - loss: 0.0385 - acc: 0.9916 - val_loss: 9.1061 - val_acc: 0.3205\n",
      "Epoch 845/1000\n",
      "28709/28709 [==============================] - 11s 368us/step - loss: 0.5303 - acc: 0.9416 - val_loss: 6.4667 - val_acc: 0.4006\n",
      "Epoch 846/1000\n",
      "28709/28709 [==============================] - 15s 533us/step - loss: 0.0080 - acc: 0.9969 - val_loss: 6.9234 - val_acc: 0.3949\n",
      "Epoch 847/1000\n",
      "28709/28709 [==============================] - 9s 326us/step - loss: 0.8156 - acc: 0.9216 - val_loss: 6.3709 - val_acc: 0.4003\n",
      "Epoch 848/1000\n",
      "28709/28709 [==============================] - 13s 442us/step - loss: 0.0068 - acc: 0.9971 - val_loss: 6.4243 - val_acc: 0.3942\n",
      "Epoch 849/1000\n",
      "28709/28709 [==============================] - 13s 462us/step - loss: 0.7654 - acc: 0.9213 - val_loss: 6.3483 - val_acc: 0.4010\n",
      "Epoch 850/1000\n",
      "28709/28709 [==============================] - 9s 306us/step - loss: 0.0075 - acc: 0.9971 - val_loss: 6.4957 - val_acc: 0.3932\n",
      "Epoch 851/1000\n",
      "28709/28709 [==============================] - 16s 558us/step - loss: 0.0094 - acc: 0.9965 - val_loss: 7.6611 - val_acc: 0.3830\n",
      "Epoch 852/1000\n",
      "28709/28709 [==============================] - 10s 341us/step - loss: 0.6443 - acc: 0.9337 - val_loss: 6.6352 - val_acc: 0.4034\n",
      "Epoch 853/1000\n",
      "28709/28709 [==============================] - 12s 409us/step - loss: 0.0069 - acc: 0.9970 - val_loss: 6.6748 - val_acc: 0.3958\n",
      "Epoch 854/1000\n",
      "28709/28709 [==============================] - 15s 528us/step - loss: 1.0029 - acc: 0.9085 - val_loss: 6.6520 - val_acc: 0.4006\n",
      "Epoch 855/1000\n",
      "28709/28709 [==============================] - 9s 306us/step - loss: 0.0070 - acc: 0.9971 - val_loss: 7.1212 - val_acc: 0.3769\n",
      "Epoch 856/1000\n",
      "28709/28709 [==============================] - 13s 455us/step - loss: 0.6083 - acc: 0.9370 - val_loss: 6.3569 - val_acc: 0.4003\n",
      "Epoch 857/1000\n",
      "28709/28709 [==============================] - 13s 442us/step - loss: 0.0068 - acc: 0.9971 - val_loss: 7.0114 - val_acc: 0.3900\n",
      "Epoch 858/1000\n",
      "28709/28709 [==============================] - 9s 302us/step - loss: 0.7067 - acc: 0.9307 - val_loss: 6.4109 - val_acc: 0.4028\n",
      "Epoch 859/1000\n",
      "28709/28709 [==============================] - 15s 506us/step - loss: 0.0069 - acc: 0.9970 - val_loss: 6.5228 - val_acc: 0.3972\n",
      "Epoch 860/1000\n",
      "28709/28709 [==============================] - 11s 372us/step - loss: 1.0461 - acc: 0.9042 - val_loss: 6.4410 - val_acc: 0.3945\n",
      "Epoch 861/1000\n",
      "28709/28709 [==============================] - 11s 368us/step - loss: 0.0081 - acc: 0.9968 - val_loss: 6.4389 - val_acc: 0.4018\n",
      "Epoch 862/1000\n",
      "28709/28709 [==============================] - 15s 523us/step - loss: 0.8625 - acc: 0.9212 - val_loss: 6.3601 - val_acc: 0.4000\n",
      "Epoch 863/1000\n",
      "28709/28709 [==============================] - 9s 320us/step - loss: 0.0081 - acc: 0.9968 - val_loss: 6.3932 - val_acc: 0.4037\n",
      "Epoch 864/1000\n",
      "28709/28709 [==============================] - 12s 425us/step - loss: 0.4543 - acc: 0.9593 - val_loss: 7.5365 - val_acc: 0.3645\n",
      "Epoch 865/1000\n",
      "28709/28709 [==============================] - 13s 464us/step - loss: 0.0842 - acc: 0.9834 - val_loss: 6.5400 - val_acc: 0.3997\n",
      "Epoch 866/1000\n",
      "28709/28709 [==============================] - 9s 329us/step - loss: 0.0072 - acc: 0.9970 - val_loss: 6.9299 - val_acc: 0.3912\n",
      "Epoch 867/1000\n",
      "28709/28709 [==============================] - 13s 448us/step - loss: 0.7873 - acc: 0.9224 - val_loss: 6.4788 - val_acc: 0.3965\n",
      "Epoch 868/1000\n",
      "28709/28709 [==============================] - 13s 459us/step - loss: 0.0071 - acc: 0.9971 - val_loss: 6.5945 - val_acc: 0.3996\n",
      "Epoch 869/1000\n",
      "28709/28709 [==============================] - 9s 302us/step - loss: 0.5412 - acc: 0.9412 - val_loss: 6.6986 - val_acc: 0.3936\n",
      "Epoch 870/1000\n",
      "28709/28709 [==============================] - 14s 500us/step - loss: 0.0080 - acc: 0.9969 - val_loss: 6.6132 - val_acc: 0.3952\n",
      "Epoch 871/1000\n",
      "28709/28709 [==============================] - 11s 396us/step - loss: 0.6141 - acc: 0.9409 - val_loss: 6.4945 - val_acc: 0.3971\n",
      "Epoch 872/1000\n",
      "28709/28709 [==============================] - 10s 359us/step - loss: 0.0072 - acc: 0.9967 - val_loss: 6.7524 - val_acc: 0.4004\n",
      "Epoch 873/1000\n",
      "28709/28709 [==============================] - 15s 526us/step - loss: 0.8306 - acc: 0.9203 - val_loss: 6.5730 - val_acc: 0.3972\n",
      "Epoch 874/1000\n",
      "28709/28709 [==============================] - 8s 285us/step - loss: 0.0076 - acc: 0.9970 - val_loss: 6.6352 - val_acc: 0.3915\n",
      "Epoch 875/1000\n",
      "28709/28709 [==============================] - 13s 469us/step - loss: 0.0086 - acc: 0.9964 - val_loss: 7.1205 - val_acc: 0.3807\n",
      "Epoch 876/1000\n",
      "28709/28709 [==============================] - 13s 443us/step - loss: 0.7615 - acc: 0.9225 - val_loss: 6.5070 - val_acc: 0.4034\n",
      "Epoch 877/1000\n",
      "28709/28709 [==============================] - 9s 306us/step - loss: 0.0072 - acc: 0.9970 - val_loss: 6.6150 - val_acc: 0.4007\n",
      "Epoch 878/1000\n",
      "28709/28709 [==============================] - 15s 536us/step - loss: 0.5218 - acc: 0.9423 - val_loss: 6.3690 - val_acc: 0.3975\n",
      "Epoch 879/1000\n",
      "28709/28709 [==============================] - 11s 389us/step - loss: 0.0069 - acc: 0.9974 - val_loss: 6.4486 - val_acc: 0.4028\n",
      "Epoch 880/1000\n",
      "28709/28709 [==============================] - 10s 339us/step - loss: 0.9972 - acc: 0.9177 - val_loss: 6.9412 - val_acc: 0.3960\n",
      "Epoch 881/1000\n",
      "28709/28709 [==============================] - 15s 539us/step - loss: 0.0168 - acc: 0.9950 - val_loss: 6.7204 - val_acc: 0.3963\n",
      "Epoch 882/1000\n",
      "28709/28709 [==============================] - 10s 348us/step - loss: 0.2249 - acc: 0.9767 - val_loss: 9.1049 - val_acc: 0.2946\n",
      "Epoch 883/1000\n",
      "28709/28709 [==============================] - 12s 410us/step - loss: 0.5588 - acc: 0.9411 - val_loss: 6.6348 - val_acc: 0.3977\n",
      "Epoch 884/1000\n",
      "28709/28709 [==============================] - 15s 505us/step - loss: 0.0070 - acc: 0.9970 - val_loss: 6.7329 - val_acc: 0.3932\n",
      "Epoch 885/1000\n",
      "28709/28709 [==============================] - 9s 307us/step - loss: 0.6476 - acc: 0.9343 - val_loss: 6.7319 - val_acc: 0.4023\n",
      "Epoch 886/1000\n",
      "28709/28709 [==============================] - 14s 500us/step - loss: 0.0064 - acc: 0.9971 - val_loss: 6.9653 - val_acc: 0.4016\n",
      "Epoch 887/1000\n",
      "28709/28709 [==============================] - 11s 398us/step - loss: 0.5205 - acc: 0.9446 - val_loss: 6.4392 - val_acc: 0.3979\n",
      "Epoch 888/1000\n",
      "28709/28709 [==============================] - 10s 361us/step - loss: 0.0067 - acc: 0.9974 - val_loss: 6.4509 - val_acc: 0.4018\n",
      "Epoch 889/1000\n",
      "28709/28709 [==============================] - 15s 531us/step - loss: 0.5841 - acc: 0.9372 - val_loss: 6.4290 - val_acc: 0.3971\n",
      "Epoch 890/1000\n",
      "28709/28709 [==============================] - 9s 302us/step - loss: 0.0080 - acc: 0.9969 - val_loss: 6.4606 - val_acc: 0.3984\n",
      "Epoch 891/1000\n",
      "28709/28709 [==============================] - 13s 459us/step - loss: 0.0070 - acc: 0.9971 - val_loss: 6.5233 - val_acc: 0.3985\n",
      "Epoch 892/1000\n",
      "28709/28709 [==============================] - 13s 449us/step - loss: 0.8199 - acc: 0.9214 - val_loss: 6.4998 - val_acc: 0.3996\n",
      "Epoch 893/1000\n",
      "28709/28709 [==============================] - 9s 319us/step - loss: 0.0072 - acc: 0.9971 - val_loss: 6.5826 - val_acc: 0.3989\n",
      "Epoch 894/1000\n",
      "28709/28709 [==============================] - 14s 504us/step - loss: 0.5479 - acc: 0.9402 - val_loss: 6.4924 - val_acc: 0.3949\n",
      "Epoch 895/1000\n",
      "28709/28709 [==============================] - 12s 405us/step - loss: 0.0067 - acc: 0.9970 - val_loss: 6.4808 - val_acc: 0.3985\n",
      "Epoch 896/1000\n",
      "28709/28709 [==============================] - 10s 355us/step - loss: 0.0071 - acc: 0.9968 - val_loss: 6.6404 - val_acc: 0.3935\n",
      "Epoch 897/1000\n",
      "28709/28709 [==============================] - 15s 535us/step - loss: 0.6470 - acc: 0.9357 - val_loss: 6.5792 - val_acc: 0.3942\n",
      "Epoch 898/1000\n",
      "28709/28709 [==============================] - 9s 296us/step - loss: 0.0085 - acc: 0.9965 - val_loss: 6.7876 - val_acc: 0.3995\n",
      "Epoch 899/1000\n",
      "28709/28709 [==============================] - 14s 482us/step - loss: 0.5299 - acc: 0.9452 - val_loss: 6.5658 - val_acc: 0.3985\n",
      "Epoch 900/1000\n",
      "28709/28709 [==============================] - 13s 458us/step - loss: 0.0071 - acc: 0.9974 - val_loss: 6.6123 - val_acc: 0.3993\n",
      "Epoch 901/1000\n",
      "28709/28709 [==============================] - 9s 311us/step - loss: 0.3771 - acc: 0.9672 - val_loss: 8.3640 - val_acc: 0.3134\n",
      "Epoch 902/1000\n",
      "28709/28709 [==============================] - 15s 532us/step - loss: 0.1556 - acc: 0.9735 - val_loss: 6.5446 - val_acc: 0.4018\n",
      "Epoch 903/1000\n",
      "28709/28709 [==============================] - 11s 395us/step - loss: 0.2449 - acc: 0.9742 - val_loss: 8.9355 - val_acc: 0.3416\n",
      "Epoch 904/1000\n",
      "28709/28709 [==============================] - 10s 363us/step - loss: 0.2209 - acc: 0.9728 - val_loss: 6.5918 - val_acc: 0.3979\n",
      "Epoch 905/1000\n",
      "28709/28709 [==============================] - 16s 549us/step - loss: 0.0070 - acc: 0.9969 - val_loss: 6.8444 - val_acc: 0.4000\n",
      "Epoch 906/1000\n",
      "28709/28709 [==============================] - 9s 308us/step - loss: 1.0251 - acc: 0.9066 - val_loss: 6.8868 - val_acc: 0.3834\n",
      "Epoch 907/1000\n",
      "28709/28709 [==============================] - 14s 482us/step - loss: 0.0091 - acc: 0.9966 - val_loss: 6.6975 - val_acc: 0.3965\n",
      "Epoch 908/1000\n",
      "28709/28709 [==============================] - 13s 463us/step - loss: 0.7264 - acc: 0.9329 - val_loss: 6.5341 - val_acc: 0.3979\n",
      "Epoch 909/1000\n",
      "28709/28709 [==============================] - 9s 324us/step - loss: 0.0094 - acc: 0.9964 - val_loss: 6.8975 - val_acc: 0.3984\n",
      "Epoch 910/1000\n",
      "28709/28709 [==============================] - 16s 557us/step - loss: 0.6612 - acc: 0.9394 - val_loss: 6.5774 - val_acc: 0.3892\n",
      "Epoch 911/1000\n",
      "28709/28709 [==============================] - 10s 365us/step - loss: 0.0084 - acc: 0.9967 - val_loss: 6.6548 - val_acc: 0.4021\n",
      "Epoch 912/1000\n",
      "28709/28709 [==============================] - 12s 419us/step - loss: 0.0099 - acc: 0.9962 - val_loss: 8.2169 - val_acc: 0.3248\n",
      "Epoch 913/1000\n",
      "28709/28709 [==============================] - 15s 511us/step - loss: 0.6124 - acc: 0.9363 - val_loss: 6.6312 - val_acc: 0.3954\n",
      "Epoch 914/1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "28709/28709 [==============================] - 8s 295us/step - loss: 0.0067 - acc: 0.9969 - val_loss: 6.7050 - val_acc: 0.3960\n",
      "Epoch 915/1000\n",
      "28709/28709 [==============================] - 15s 533us/step - loss: 0.8516 - acc: 0.9201 - val_loss: 6.5801 - val_acc: 0.3970\n",
      "Epoch 916/1000\n",
      "28709/28709 [==============================] - 11s 369us/step - loss: 0.0069 - acc: 0.9970 - val_loss: 6.6179 - val_acc: 0.3974\n",
      "Epoch 917/1000\n",
      "28709/28709 [==============================] - 12s 403us/step - loss: 0.5622 - acc: 0.9481 - val_loss: 7.7164 - val_acc: 0.3552\n",
      "Epoch 918/1000\n",
      "28709/28709 [==============================] - 15s 506us/step - loss: 0.0758 - acc: 0.9844 - val_loss: 6.7443 - val_acc: 0.3988\n",
      "Epoch 919/1000\n",
      "28709/28709 [==============================] - 9s 314us/step - loss: 0.2701 - acc: 0.9740 - val_loss: 9.0317 - val_acc: 0.3297\n",
      "Epoch 920/1000\n",
      "28709/28709 [==============================] - 14s 493us/step - loss: 0.3084 - acc: 0.9623 - val_loss: 6.6596 - val_acc: 0.3984\n",
      "Epoch 921/1000\n",
      "28709/28709 [==============================] - 12s 429us/step - loss: 0.4104 - acc: 0.9487 - val_loss: 6.7029 - val_acc: 0.3784\n",
      "Epoch 922/1000\n",
      "28709/28709 [==============================] - 10s 359us/step - loss: 0.0094 - acc: 0.9964 - val_loss: 6.4612 - val_acc: 0.3979\n",
      "Epoch 923/1000\n",
      "28709/28709 [==============================] - 16s 547us/step - loss: 0.0071 - acc: 0.9967 - val_loss: 6.8312 - val_acc: 0.3996\n",
      "Epoch 924/1000\n",
      "28709/28709 [==============================] - 9s 316us/step - loss: 0.5943 - acc: 0.9362 - val_loss: 6.4715 - val_acc: 0.3997\n",
      "Epoch 925/1000\n",
      "28709/28709 [==============================] - 14s 482us/step - loss: 0.0059 - acc: 0.9971 - val_loss: 6.5394 - val_acc: 0.4032\n",
      "Epoch 926/1000\n",
      "28709/28709 [==============================] - 13s 442us/step - loss: 0.5522 - acc: 0.9410 - val_loss: 6.5636 - val_acc: 0.3950\n",
      "Epoch 927/1000\n",
      "28709/28709 [==============================] - 10s 354us/step - loss: 0.0069 - acc: 0.9972 - val_loss: 6.7580 - val_acc: 0.3964\n",
      "Epoch 928/1000\n",
      "28709/28709 [==============================] - 16s 558us/step - loss: 0.0122 - acc: 0.9957 - val_loss: 8.1321 - val_acc: 0.3603\n",
      "Epoch 929/1000\n",
      "28709/28709 [==============================] - 9s 329us/step - loss: 0.3698 - acc: 0.9537 - val_loss: 6.5791 - val_acc: 0.4021\n",
      "Epoch 930/1000\n",
      "28709/28709 [==============================] - 14s 491us/step - loss: 0.0067 - acc: 0.9973 - val_loss: 6.8242 - val_acc: 0.4020\n",
      "Epoch 931/1000\n",
      "28709/28709 [==============================] - 12s 414us/step - loss: 0.8157 - acc: 0.9250 - val_loss: 6.7144 - val_acc: 0.3964\n",
      "Epoch 932/1000\n",
      "28709/28709 [==============================] - 11s 399us/step - loss: 0.0069 - acc: 0.9972 - val_loss: 6.7603 - val_acc: 0.3979\n",
      "Epoch 933/1000\n",
      "28709/28709 [==============================] - 15s 519us/step - loss: 0.5784 - acc: 0.9399 - val_loss: 6.4602 - val_acc: 0.4050\n",
      "Epoch 934/1000\n",
      "28709/28709 [==============================] - 9s 311us/step - loss: 0.0073 - acc: 0.9972 - val_loss: 6.5222 - val_acc: 0.4059\n",
      "Epoch 935/1000\n",
      "28709/28709 [==============================] - 17s 586us/step - loss: 0.1706 - acc: 0.9809 - val_loss: 9.8683 - val_acc: 0.3075\n",
      "Epoch 936/1000\n",
      "28709/28709 [==============================] - 9s 330us/step - loss: 0.5915 - acc: 0.9407 - val_loss: 6.7310 - val_acc: 0.3972\n",
      "Epoch 937/1000\n",
      "28709/28709 [==============================] - 13s 464us/step - loss: 0.0070 - acc: 0.9968 - val_loss: 6.8613 - val_acc: 0.4000\n",
      "Epoch 938/1000\n",
      "28709/28709 [==============================] - 13s 458us/step - loss: 0.5758 - acc: 0.9434 - val_loss: 6.6604 - val_acc: 0.3971\n",
      "Epoch 939/1000\n",
      "28709/28709 [==============================] - 10s 336us/step - loss: 0.0064 - acc: 0.9970 - val_loss: 6.7348 - val_acc: 0.3947\n",
      "Epoch 940/1000\n",
      "28709/28709 [==============================] - 17s 597us/step - loss: 0.6740 - acc: 0.9363 - val_loss: 6.6075 - val_acc: 0.3933\n",
      "Epoch 941/1000\n",
      "28709/28709 [==============================] - 9s 328us/step - loss: 0.0070 - acc: 0.9971 - val_loss: 6.6605 - val_acc: 0.3952\n",
      "Epoch 942/1000\n",
      "28709/28709 [==============================] - 15s 523us/step - loss: 0.4274 - acc: 0.9519 - val_loss: 6.5505 - val_acc: 0.4006\n",
      "Epoch 943/1000\n",
      "28709/28709 [==============================] - 11s 390us/step - loss: 0.0076 - acc: 0.9971 - val_loss: 6.7268 - val_acc: 0.3952\n",
      "Epoch 944/1000\n",
      "28709/28709 [==============================] - 13s 456us/step - loss: 0.0572 - acc: 0.9898 - val_loss: 8.4702 - val_acc: 0.3501\n",
      "Epoch 945/1000\n",
      "28709/28709 [==============================] - 14s 502us/step - loss: 0.2671 - acc: 0.9659 - val_loss: 6.6203 - val_acc: 0.3993\n",
      "Epoch 946/1000\n",
      "28709/28709 [==============================] - 10s 351us/step - loss: 0.0070 - acc: 0.9969 - val_loss: 6.8707 - val_acc: 0.3956\n",
      "Epoch 947/1000\n",
      "28709/28709 [==============================] - 18s 613us/step - loss: 0.4804 - acc: 0.9463 - val_loss: 6.5604 - val_acc: 0.4002\n",
      "Epoch 948/1000\n",
      "28709/28709 [==============================] - 10s 348us/step - loss: 0.0068 - acc: 0.9972 - val_loss: 6.7343 - val_acc: 0.3979\n",
      "Epoch 949/1000\n",
      "28709/28709 [==============================] - 14s 503us/step - loss: 0.5951 - acc: 0.9397 - val_loss: 6.5367 - val_acc: 0.3979\n",
      "Epoch 950/1000\n",
      "28709/28709 [==============================] - 12s 418us/step - loss: 0.0063 - acc: 0.9970 - val_loss: 6.6405 - val_acc: 0.3942\n",
      "Epoch 951/1000\n",
      "28709/28709 [==============================] - 12s 406us/step - loss: 0.7300 - acc: 0.9293 - val_loss: 6.5782 - val_acc: 0.3961\n",
      "Epoch 952/1000\n",
      "28709/28709 [==============================] - 15s 513us/step - loss: 0.0071 - acc: 0.9969 - val_loss: 6.9263 - val_acc: 0.3861\n",
      "Epoch 953/1000\n",
      "28709/28709 [==============================] - 10s 338us/step - loss: 0.4674 - acc: 0.9470 - val_loss: 6.4596 - val_acc: 0.3989\n",
      "Epoch 954/1000\n",
      "28709/28709 [==============================] - 16s 563us/step - loss: 0.0061 - acc: 0.9974 - val_loss: 6.5398 - val_acc: 0.4010\n",
      "Epoch 955/1000\n",
      "28709/28709 [==============================] - 9s 324us/step - loss: 0.4812 - acc: 0.9481 - val_loss: 6.5809 - val_acc: 0.4043\n",
      "Epoch 956/1000\n",
      "28709/28709 [==============================] - 17s 586us/step - loss: 0.0073 - acc: 0.9972 - val_loss: 7.2820 - val_acc: 0.3919\n",
      "Epoch 957/1000\n",
      "28709/28709 [==============================] - 9s 328us/step - loss: 0.5387 - acc: 0.9456 - val_loss: 6.5396 - val_acc: 0.3997\n",
      "Epoch 958/1000\n",
      "28709/28709 [==============================] - 15s 540us/step - loss: 0.0066 - acc: 0.9969 - val_loss: 6.6126 - val_acc: 0.4049\n",
      "Epoch 959/1000\n",
      "28709/28709 [==============================] - 11s 373us/step - loss: 0.4899 - acc: 0.9463 - val_loss: 6.5107 - val_acc: 0.3960\n",
      "Epoch 960/1000\n",
      "28709/28709 [==============================] - 14s 472us/step - loss: 0.0065 - acc: 0.9973 - val_loss: 6.5337 - val_acc: 0.3986\n",
      "Epoch 961/1000\n",
      "28709/28709 [==============================] - 13s 446us/step - loss: 0.6935 - acc: 0.9362 - val_loss: 6.6018 - val_acc: 0.4000\n",
      "Epoch 962/1000\n",
      "28709/28709 [==============================] - 11s 390us/step - loss: 0.0078 - acc: 0.9971 - val_loss: 6.8403 - val_acc: 0.3929\n",
      "Epoch 963/1000\n",
      "28709/28709 [==============================] - 15s 530us/step - loss: 0.0216 - acc: 0.9943 - val_loss: 9.4840 - val_acc: 0.3404\n",
      "Epoch 964/1000\n",
      "28709/28709 [==============================] - 9s 327us/step - loss: 0.5228 - acc: 0.9478 - val_loss: 6.9227 - val_acc: 0.3986\n",
      "Epoch 965/1000\n",
      "28709/28709 [==============================] - 17s 587us/step - loss: 0.4274 - acc: 0.9575 - val_loss: 9.0380 - val_acc: 0.3173\n",
      "Epoch 966/1000\n",
      "28709/28709 [==============================] - 10s 347us/step - loss: 0.1888 - acc: 0.9772 - val_loss: 6.6966 - val_acc: 0.3991\n",
      "Epoch 967/1000\n",
      "28709/28709 [==============================] - 15s 509us/step - loss: 0.0062 - acc: 0.9969 - val_loss: 6.8533 - val_acc: 0.3965\n",
      "Epoch 968/1000\n",
      "28709/28709 [==============================] - 12s 408us/step - loss: 0.6513 - acc: 0.9356 - val_loss: 6.7827 - val_acc: 0.4017\n",
      "Epoch 969/1000\n",
      "28709/28709 [==============================] - 13s 452us/step - loss: 0.0067 - acc: 0.9971 - val_loss: 6.8356 - val_acc: 0.3981\n",
      "Epoch 970/1000\n",
      "28709/28709 [==============================] - 13s 442us/step - loss: 0.7212 - acc: 0.9277 - val_loss: 6.7306 - val_acc: 0.4017\n",
      "Epoch 971/1000\n",
      "28709/28709 [==============================] - 12s 427us/step - loss: 0.0080 - acc: 0.9966 - val_loss: 6.7976 - val_acc: 0.3919\n",
      "Epoch 972/1000\n",
      "28709/28709 [==============================] - 14s 486us/step - loss: 0.0079 - acc: 0.9967 - val_loss: 6.8777 - val_acc: 0.3936\n",
      "Epoch 973/1000\n",
      "28709/28709 [==============================] - 10s 360us/step - loss: 0.6701 - acc: 0.9311 - val_loss: 6.7910 - val_acc: 0.4027\n",
      "Epoch 974/1000\n",
      "28709/28709 [==============================] - 16s 567us/step - loss: 0.0067 - acc: 0.9973 - val_loss: 6.7981 - val_acc: 0.4069\n",
      "Epoch 975/1000\n",
      "28709/28709 [==============================] - 10s 331us/step - loss: 0.6977 - acc: 0.9318 - val_loss: 6.8569 - val_acc: 0.3897\n",
      "Epoch 976/1000\n",
      "28709/28709 [==============================] - 17s 590us/step - loss: 0.0080 - acc: 0.9969 - val_loss: 6.7870 - val_acc: 0.3992\n",
      "Epoch 977/1000\n",
      "28709/28709 [==============================] - 9s 318us/step - loss: 0.1475 - acc: 0.9824 - val_loss: 9.0858 - val_acc: 0.3472\n",
      "Epoch 978/1000\n",
      "28709/28709 [==============================] - 17s 589us/step - loss: 0.2843 - acc: 0.9663 - val_loss: 6.8600 - val_acc: 0.3984\n",
      "Epoch 979/1000\n",
      "28709/28709 [==============================] - 11s 392us/step - loss: 0.0065 - acc: 0.9969 - val_loss: 6.8665 - val_acc: 0.3985\n",
      "Epoch 980/1000\n",
      "28709/28709 [==============================] - 15s 525us/step - loss: 0.7203 - acc: 0.9326 - val_loss: 6.8108 - val_acc: 0.4020\n",
      "Epoch 981/1000\n",
      "28709/28709 [==============================] - 16s 544us/step - loss: 0.0062 - acc: 0.9974 - val_loss: 6.8458 - val_acc: 0.4010\n",
      "Epoch 982/1000\n",
      "28709/28709 [==============================] - 13s 454us/step - loss: 0.7207 - acc: 0.9301 - val_loss: 6.7472 - val_acc: 0.3968\n",
      "Epoch 983/1000\n",
      "28709/28709 [==============================] - 17s 606us/step - loss: 0.0070 - acc: 0.9972 - val_loss: 6.7298 - val_acc: 0.3982\n",
      "Epoch 984/1000\n",
      "28709/28709 [==============================] - 13s 451us/step - loss: 0.3042 - acc: 0.9681 - val_loss: 7.5628 - val_acc: 0.3766\n",
      "Epoch 985/1000\n",
      "28709/28709 [==============================] - 17s 601us/step - loss: 0.0951 - acc: 0.9801 - val_loss: 6.7327 - val_acc: 0.4009\n",
      "Epoch 986/1000\n",
      "28709/28709 [==============================] - 10s 349us/step - loss: 0.1198 - acc: 0.9840 - val_loss: 8.4748 - val_acc: 0.3673\n",
      "Epoch 987/1000\n",
      "28709/28709 [==============================] - 17s 581us/step - loss: 0.2792 - acc: 0.9668 - val_loss: 6.8115 - val_acc: 0.3981\n",
      "Epoch 988/1000\n",
      "28709/28709 [==============================] - 10s 339us/step - loss: 0.0064 - acc: 0.9970 - val_loss: 6.9799 - val_acc: 0.3893\n",
      "Epoch 989/1000\n",
      "28709/28709 [==============================] - 17s 577us/step - loss: 0.5500 - acc: 0.9420 - val_loss: 6.5695 - val_acc: 0.4030\n",
      "Epoch 990/1000\n",
      "28709/28709 [==============================] - 10s 346us/step - loss: 0.0060 - acc: 0.9971 - val_loss: 6.6545 - val_acc: 0.3989\n",
      "Epoch 991/1000\n",
      "28709/28709 [==============================] - 17s 575us/step - loss: 0.7579 - acc: 0.9247 - val_loss: 6.6619 - val_acc: 0.4017\n",
      "Epoch 992/1000\n",
      "28709/28709 [==============================] - 10s 347us/step - loss: 0.0065 - acc: 0.9972 - val_loss: 6.6761 - val_acc: 0.3999\n",
      "Epoch 993/1000\n",
      "28709/28709 [==============================] - 16s 558us/step - loss: 0.6098 - acc: 0.9447 - val_loss: 7.1435 - val_acc: 0.3748\n",
      "Epoch 994/1000\n",
      "28709/28709 [==============================] - 10s 361us/step - loss: 0.0158 - acc: 0.9950 - val_loss: 6.7870 - val_acc: 0.3958\n",
      "Epoch 995/1000\n",
      "28709/28709 [==============================] - 15s 540us/step - loss: 0.4704 - acc: 0.9593 - val_loss: 8.4560 - val_acc: 0.3552\n",
      "Epoch 996/1000\n",
      "28709/28709 [==============================] - 11s 390us/step - loss: 0.1744 - acc: 0.9760 - val_loss: 6.7870 - val_acc: 0.4004\n",
      "Epoch 997/1000\n",
      "28709/28709 [==============================] - 14s 479us/step - loss: 0.0066 - acc: 0.9970 - val_loss: 6.8641 - val_acc: 0.3993\n",
      "Epoch 998/1000\n",
      "28709/28709 [==============================] - 13s 436us/step - loss: 0.5790 - acc: 0.9418 - val_loss: 6.7093 - val_acc: 0.4003\n",
      "Epoch 999/1000\n",
      "28709/28709 [==============================] - 13s 447us/step - loss: 0.0065 - acc: 0.9971 - val_loss: 6.7722 - val_acc: 0.4020\n",
      "Epoch 1000/1000\n",
      "28709/28709 [==============================] - 13s 447us/step - loss: 0.7190 - acc: 0.9301 - val_loss: 6.8194 - val_acc: 0.3961\n",
      "Test loss 6.819397463379289\n",
      "Test accuracy 0.39612651526952863\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(train_images,train_labels,\n",
    "                    batch_size=batch_size,\n",
    "                    epochs=epochs,\n",
    "                    verbose=1,\n",
    "                    validation_data=(test_images,test_labels))\n",
    "score = model.evaluate(test_images,test_labels, verbose=0)\n",
    "print(\"Test loss\", score[0])\n",
    "print(\"Test accuracy\", score[1] )\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "7177"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predictions = model.predict(test_images)\n",
    "len(predictions)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.02301867, 0.0011747 , 0.2851315 , 0.03891646, 0.03718293,\n",
       "       0.52093405, 0.09364165], dtype=float32)"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predictions[5]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "5"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.argmax(predictions[5])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'tf' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-1-5f45bee47b71>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[1;32m----> 1\u001b[1;33m \u001b[0mmnist\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mtf\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcontrib\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mlearn\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdatasets\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mload_dataset\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m\"mnist\"\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[1;31mNameError\u001b[0m: name 'tf' is not defined"
     ]
    }
   ],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From <ipython-input-3-5f45bee47b71>:1: load_dataset (from tensorflow.contrib.learn.python.learn.datasets) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use tf.data.\n",
      "WARNING:tensorflow:From d:\\ProgramData\\Anaconda3\\lib\\site-packages\\tensorflow\\contrib\\learn\\python\\learn\\datasets\\__init__.py:80: load_mnist (from tensorflow.contrib.learn.python.learn.datasets.mnist) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use alternatives such as official/mnist/dataset.py from tensorflow/models.\n",
      "WARNING:tensorflow:From d:\\ProgramData\\Anaconda3\\lib\\site-packages\\tensorflow\\contrib\\learn\\python\\learn\\datasets\\mnist.py:300: read_data_sets (from tensorflow.contrib.learn.python.learn.datasets.mnist) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use alternatives such as official/mnist/dataset.py from tensorflow/models.\n",
      "WARNING:tensorflow:From d:\\ProgramData\\Anaconda3\\lib\\site-packages\\tensorflow\\contrib\\learn\\python\\learn\\datasets\\mnist.py:260: maybe_download (from tensorflow.contrib.learn.python.learn.datasets.base) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please write your own downloading logic.\n",
      "WARNING:tensorflow:From d:\\ProgramData\\Anaconda3\\lib\\site-packages\\tensorflow\\contrib\\learn\\python\\learn\\datasets\\base.py:252: _internal_retry.<locals>.wrap.<locals>.wrapped_fn (from tensorflow.contrib.learn.python.learn.datasets.base) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use urllib or similar directly.\n",
      "Successfully downloaded train-images-idx3-ubyte.gz 9912422 bytes.\n",
      "WARNING:tensorflow:From d:\\ProgramData\\Anaconda3\\lib\\site-packages\\tensorflow\\contrib\\learn\\python\\learn\\datasets\\mnist.py:262: extract_images (from tensorflow.contrib.learn.python.learn.datasets.mnist) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use tf.data to implement this functionality.\n",
      "Extracting MNIST-data\\train-images-idx3-ubyte.gz\n",
      "Successfully downloaded train-labels-idx1-ubyte.gz 28881 bytes.\n",
      "WARNING:tensorflow:From d:\\ProgramData\\Anaconda3\\lib\\site-packages\\tensorflow\\contrib\\learn\\python\\learn\\datasets\\mnist.py:267: extract_labels (from tensorflow.contrib.learn.python.learn.datasets.mnist) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use tf.data to implement this functionality.\n",
      "Extracting MNIST-data\\train-labels-idx1-ubyte.gz\n",
      "Successfully downloaded t10k-images-idx3-ubyte.gz 1648877 bytes.\n",
      "Extracting MNIST-data\\t10k-images-idx3-ubyte.gz\n",
      "Successfully downloaded t10k-labels-idx1-ubyte.gz 4542 bytes.\n",
      "Extracting MNIST-data\\t10k-labels-idx1-ubyte.gz\n",
      "WARNING:tensorflow:From d:\\ProgramData\\Anaconda3\\lib\\site-packages\\tensorflow\\contrib\\learn\\python\\learn\\datasets\\mnist.py:290: DataSet.__init__ (from tensorflow.contrib.learn.python.learn.datasets.mnist) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use alternatives such as official/mnist/dataset.py from tensorflow/models.\n"
     ]
    }
   ],
   "source": [
    "  mnist = tf.contrib.learn.datasets.load_dataset(\"mnist\")"
   ]
  },
  {
   "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}