neural_network.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np # linear algebra\n",
    "import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "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>Survived</th>\n",
       "      <th>Pclass</th>\n",
       "      <th>Name</th>\n",
       "      <th>Gender</th>\n",
       "      <th>Age</th>\n",
       "      <th>SibSp</th>\n",
       "      <th>Parch</th>\n",
       "      <th>Embarked</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Survived</td>\n",
       "      <td>Pclass</td>\n",
       "      <td>Name</td>\n",
       "      <td>Gender</td>\n",
       "      <td>Age</td>\n",
       "      <td>SibSp</td>\n",
       "      <td>Parch</td>\n",
       "      <td>Embarked</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Braund, Mr. Owen Harris</td>\n",
       "      <td>male</td>\n",
       "      <td>22</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>Cumings, Mrs. John Bradley (Florence Briggs Th...</td>\n",
       "      <td>female</td>\n",
       "      <td>38</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>C</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>Heikkinen, Miss. Laina</td>\n",
       "      <td>female</td>\n",
       "      <td>26</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>Futrelle, Mrs. Jacques Heath (Lily May Peel)</td>\n",
       "      <td>female</td>\n",
       "      <td>35</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Allen, Mr. William Henry</td>\n",
       "      <td>male</td>\n",
       "      <td>35</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Moran, Mr. James</td>\n",
       "      <td>male</td>\n",
       "      <td></td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>Q</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>McCarthy, Mr. Timothy J</td>\n",
       "      <td>male</td>\n",
       "      <td>54</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0</td>\n",
       "      <td>3</td>\n",
       "      <td>Palsson, Master. Gosta Leonard</td>\n",
       "      <td>male</td>\n",
       "      <td>2</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>1</td>\n",
       "      <td>3</td>\n",
       "      <td>Johnson, Mrs. Oscar W (Elisabeth Vilhelmina Berg)</td>\n",
       "      <td>female</td>\n",
       "      <td>27</td>\n",
       "      <td>0</td>\n",
       "      <td>2</td>\n",
       "      <td>S</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Survived  Pclass                                               Name  \\\n",
       "0  Survived  Pclass                                               Name   \n",
       "1         0       3                            Braund, Mr. Owen Harris   \n",
       "2         1       1  Cumings, Mrs. John Bradley (Florence Briggs Th...   \n",
       "3         1       3                             Heikkinen, Miss. Laina   \n",
       "4         1       1       Futrelle, Mrs. Jacques Heath (Lily May Peel)   \n",
       "5         0       3                           Allen, Mr. William Henry   \n",
       "6         0       3                                   Moran, Mr. James   \n",
       "7         0       1                            McCarthy, Mr. Timothy J   \n",
       "8         0       3                     Palsson, Master. Gosta Leonard   \n",
       "9         1       3  Johnson, Mrs. Oscar W (Elisabeth Vilhelmina Berg)   \n",
       "\n",
       "   Gender  Age  SibSp  Parch  Embarked  \n",
       "0  Gender  Age  SibSp  Parch  Embarked  \n",
       "1    male   22      1      0         S  \n",
       "2  female   38      1      0         C  \n",
       "3  female   26      0      0         S  \n",
       "4  female   35      1      0         S  \n",
       "5    male   35      0      0         S  \n",
       "6    male           0      0         Q  \n",
       "7    male   54      0      0         S  \n",
       "8    male    2      3      1         S  \n",
       "9  female   27      0      2         S  "
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# get the data\n",
    "filname = 'train.csv'\n",
    "label_map = ['Not Survived', 'Survived']\n",
    "names=['Survived', 'Pclass', 'Name', 'Gender', 'Age', 'SibSp', 'Parch', 'Embarked']\n",
    "df=pd.read_csv('train.csv',names=names, na_filter=False)\n",
    "df.head(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "def getData(filname):\n",
    "    # images are 48x48\n",
    "    # N = 35882\n",
    "    Y = []\n",
    "    X = []\n",
    "    first = True\n",
    "    i = 0\n",
    "    for line in open(filname):\n",
    "        if first:\n",
    "            first = False\n",
    "        else:\n",
    "            row = line.split(',')\n",
    "            assert(len(row) == 9)\n",
    "            \n",
    "            Y.append(int(row[0]))\n",
    "            \n",
    "            X.append(list())\n",
    "            \n",
    "            X[i].append(int(row[1]) / 3)\n",
    "            if row[4] == 'male':\n",
    "                row[4] = 1\n",
    "            else:\n",
    "                row[4] = 0\n",
    "            X[i].append(int(row[4]))\n",
    "            \n",
    "            if row[5] == '':\n",
    "                row[5] = 30\n",
    "            X[i].append(float(row[5]) / 80.0)\n",
    "            \n",
    "            X[i].append(int(row[6]))\n",
    "            \n",
    "            X[i].append(int(row[7]))\n",
    "            \n",
    "            if row[8] == 'S':\n",
    "                row[8] = 0\n",
    "            elif row[8] == 'Q':\n",
    "                row[8] = 0.5\n",
    "            else:\n",
    "                row[8] = 1\n",
    "            X[i].append(int(row[8]))\n",
    "            i = i + 1\n",
    "\n",
    "    npX, npY = np.array(X), np.array(Y)\n",
    "    return npX, npY"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "X, Y = getData(filname)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of categories = 2\n",
      "number of passengers = 891\n",
      "Passengers' info shape = 6\n"
     ]
    }
   ],
   "source": [
    "num_class = len(set(Y))\n",
    "print('number of categories = ' + str(num_class))\n",
    "\n",
    "N, D = X.shape\n",
    "print('number of passengers = ' + str(N))\n",
    "print('Passengers\\' info shape = ' + str(D))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(891, 6, 1, 1)\n"
     ]
    }
   ],
   "source": [
    "X = X.reshape(N, 6, 1, 1)\n",
    "print(X.shape)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, Y, test_size=0.1, random_state=0)\n",
    "y_train = (np.arange(num_class) == y_train[:, None]).astype(np.float32)\n",
    "y_test = (np.arange(num_class) == y_test[:, None]).astype(np.float32)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense , Activation , Dropout ,Flatten\n",
    "from keras.layers.convolutional import Conv2D\n",
    "from keras.layers.convolutional import MaxPooling2D\n",
    "from keras.metrics import categorical_accuracy\n",
    "from keras.models import model_from_json\n",
    "from keras.callbacks import ModelCheckpoint\n",
    "from keras.optimizers import *\n",
    "from keras.layers.normalization import BatchNormalization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_2\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_5 (Conv2D)            (None, 6, 1, 64)          1664      \n",
      "_________________________________________________________________\n",
      "conv2d_6 (Conv2D)            (None, 6, 1, 64)          102464    \n",
      "_________________________________________________________________\n",
      "batch_normalization_4 (Batch (None, 6, 1, 64)          256       \n",
      "_________________________________________________________________\n",
      "conv2d_7 (Conv2D)            (None, 6, 1, 128)         204928    \n",
      "_________________________________________________________________\n",
      "conv2d_8 (Conv2D)            (None, 6, 1, 128)         409728    \n",
      "_________________________________________________________________\n",
      "batch_normalization_5 (Batch (None, 6, 1, 128)         512       \n",
      "_________________________________________________________________\n",
      "flatten_2 (Flatten)          (None, 768)               0         \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 128)               98432     \n",
      "_________________________________________________________________\n",
      "batch_normalization_6 (Batch (None, 128)               512       \n",
      "_________________________________________________________________\n",
      "activation_3 (Activation)    (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 2)                 258       \n",
      "_________________________________________________________________\n",
      "activation_4 (Activation)    (None, 2)                 0         \n",
      "=================================================================\n",
      "Total params: 818,754\n",
      "Trainable params: 818,114\n",
      "Non-trainable params: 640\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "def my_model():\n",
    "    model = Sequential()\n",
    "    input_shape = (6, 1, 1)\n",
    "    model.add(Conv2D(64, (5, 5), input_shape=input_shape,activation='relu', padding='same'))\n",
    "    model.add(Conv2D(64, (5, 5), activation='relu', padding='same'))\n",
    "    model.add(BatchNormalization())\n",
    "\n",
    "    model.add(Conv2D(128, (5, 5),activation='relu',padding='same'))\n",
    "    model.add(Conv2D(128, (5, 5),activation='relu',padding='same'))\n",
    "    model.add(BatchNormalization())\n",
    "\n",
    "    model.add(Flatten())\n",
    "    model.add(Dense(128))\n",
    "    model.add(BatchNormalization())\n",
    "    model.add(Activation('relu'))\n",
    "    model.add(Dropout(0.2))\n",
    "    model.add(Dense(2))\n",
    "    model.add(Activation('softmax'))\n",
    "    \n",
    "    model.compile(loss='categorical_crossentropy', metrics=['accuracy'],optimizer='adam')\n",
    "    # UNCOMMENT THIS TO VIEW THE ARCHITECTURE\n",
    "    #model.summary()\n",
    "    \n",
    "    return model\n",
    "\n",
    "model=my_model()\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 801 samples, validate on 90 samples\n",
      "Epoch 1/60\n",
      "801/801 [==============================] - 2s 2ms/step - loss: 0.5990 - accuracy: 0.7191 - val_loss: 0.7444 - val_accuracy: 0.4333\n",
      "Epoch 2/60\n",
      "801/801 [==============================] - 1s 898us/step - loss: 0.5720 - accuracy: 0.7516 - val_loss: 0.7241 - val_accuracy: 0.4333\n",
      "Epoch 3/60\n",
      "801/801 [==============================] - 1s 900us/step - loss: 0.5481 - accuracy: 0.7553 - val_loss: 0.6764 - val_accuracy: 0.4333\n",
      "Epoch 4/60\n",
      "801/801 [==============================] - 1s 894us/step - loss: 0.5230 - accuracy: 0.7740 - val_loss: 0.5511 - val_accuracy: 0.8111\n",
      "Epoch 5/60\n",
      "801/801 [==============================] - 1s 923us/step - loss: 0.5034 - accuracy: 0.7828 - val_loss: 0.4738 - val_accuracy: 0.8111\n",
      "Epoch 6/60\n",
      "801/801 [==============================] - 1s 906us/step - loss: 0.5016 - accuracy: 0.7828 - val_loss: 0.4339 - val_accuracy: 0.7889\n",
      "Epoch 7/60\n",
      "801/801 [==============================] - 1s 903us/step - loss: 0.4924 - accuracy: 0.7878 - val_loss: 0.4004 - val_accuracy: 0.8222\n",
      "Epoch 8/60\n",
      "801/801 [==============================] - 1s 900us/step - loss: 0.4851 - accuracy: 0.7915 - val_loss: 0.3940 - val_accuracy: 0.8333\n",
      "Epoch 9/60\n",
      "801/801 [==============================] - 1s 899us/step - loss: 0.4691 - accuracy: 0.8052 - val_loss: 0.3896 - val_accuracy: 0.8556\n",
      "Epoch 10/60\n",
      "801/801 [==============================] - 1s 905us/step - loss: 0.4622 - accuracy: 0.8040 - val_loss: 0.3713 - val_accuracy: 0.8333\n",
      "Epoch 11/60\n",
      "801/801 [==============================] - 1s 909us/step - loss: 0.4570 - accuracy: 0.8040 - val_loss: 0.4269 - val_accuracy: 0.8111\n",
      "Epoch 12/60\n",
      "801/801 [==============================] - 1s 914us/step - loss: 0.4445 - accuracy: 0.8052 - val_loss: 0.4033 - val_accuracy: 0.8444\n",
      "Epoch 13/60\n",
      "801/801 [==============================] - 1s 900us/step - loss: 0.4321 - accuracy: 0.8027 - val_loss: 0.4033 - val_accuracy: 0.8000\n",
      "Epoch 14/60\n",
      "801/801 [==============================] - 1s 904us/step - loss: 0.4333 - accuracy: 0.8165 - val_loss: 0.3940 - val_accuracy: 0.8222\n",
      "Epoch 15/60\n",
      "801/801 [==============================] - 1s 905us/step - loss: 0.4285 - accuracy: 0.8152 - val_loss: 0.3720 - val_accuracy: 0.8333\n",
      "Epoch 16/60\n",
      "801/801 [==============================] - 1s 909us/step - loss: 0.4413 - accuracy: 0.8165 - val_loss: 0.3898 - val_accuracy: 0.8333\n",
      "Epoch 17/60\n",
      "801/801 [==============================] - 1s 914us/step - loss: 0.4212 - accuracy: 0.8265 - val_loss: 0.3888 - val_accuracy: 0.8222\n",
      "Epoch 18/60\n",
      "801/801 [==============================] - 1s 910us/step - loss: 0.4276 - accuracy: 0.8227 - val_loss: 0.3830 - val_accuracy: 0.8111\n",
      "Epoch 19/60\n",
      "801/801 [==============================] - 1s 923us/step - loss: 0.4227 - accuracy: 0.8290 - val_loss: 0.3794 - val_accuracy: 0.8333\n",
      "Epoch 20/60\n",
      "801/801 [==============================] - 1s 904us/step - loss: 0.4167 - accuracy: 0.8265 - val_loss: 0.4133 - val_accuracy: 0.7889\n",
      "Epoch 21/60\n",
      "801/801 [==============================] - 1s 908us/step - loss: 0.4191 - accuracy: 0.8227 - val_loss: 0.3949 - val_accuracy: 0.7889\n",
      "Epoch 22/60\n",
      "801/801 [==============================] - 1s 905us/step - loss: 0.4136 - accuracy: 0.8227 - val_loss: 0.4251 - val_accuracy: 0.7889\n",
      "Epoch 23/60\n",
      "801/801 [==============================] - 1s 913us/step - loss: 0.4035 - accuracy: 0.8365 - val_loss: 0.3767 - val_accuracy: 0.8111\n",
      "Epoch 24/60\n",
      "801/801 [==============================] - 1s 908us/step - loss: 0.4234 - accuracy: 0.8140 - val_loss: 0.3828 - val_accuracy: 0.8222\n",
      "Epoch 25/60\n",
      "801/801 [==============================] - 1s 926us/step - loss: 0.3916 - accuracy: 0.8427 - val_loss: 0.3964 - val_accuracy: 0.8111\n",
      "Epoch 26/60\n",
      "801/801 [==============================] - 1s 964us/step - loss: 0.3756 - accuracy: 0.8414 - val_loss: 0.3985 - val_accuracy: 0.8222\n",
      "Epoch 27/60\n",
      "801/801 [==============================] - 1s 900us/step - loss: 0.3645 - accuracy: 0.8577 - val_loss: 0.4011 - val_accuracy: 0.8333\n",
      "Epoch 28/60\n",
      "801/801 [==============================] - 1s 920us/step - loss: 0.3816 - accuracy: 0.8290 - val_loss: 0.4040 - val_accuracy: 0.8111\n",
      "Epoch 29/60\n",
      "801/801 [==============================] - 1s 916us/step - loss: 0.3628 - accuracy: 0.8502 - val_loss: 0.4087 - val_accuracy: 0.8333\n",
      "Epoch 30/60\n",
      "801/801 [==============================] - 1s 908us/step - loss: 0.3562 - accuracy: 0.8514 - val_loss: 0.4115 - val_accuracy: 0.8444\n",
      "Epoch 31/60\n",
      "801/801 [==============================] - 1s 898us/step - loss: 0.3495 - accuracy: 0.8527 - val_loss: 0.4964 - val_accuracy: 0.7778\n",
      "Epoch 32/60\n",
      "801/801 [==============================] - 1s 913us/step - loss: 0.3651 - accuracy: 0.8340 - val_loss: 0.4490 - val_accuracy: 0.7667\n",
      "Epoch 33/60\n",
      "801/801 [==============================] - 1s 904us/step - loss: 0.3610 - accuracy: 0.8614 - val_loss: 0.4781 - val_accuracy: 0.8222\n",
      "Epoch 34/60\n",
      "801/801 [==============================] - 1s 906us/step - loss: 0.3384 - accuracy: 0.8589 - val_loss: 0.4196 - val_accuracy: 0.8111\n",
      "Epoch 35/60\n",
      "801/801 [==============================] - 1s 909us/step - loss: 0.3451 - accuracy: 0.8614 - val_loss: 0.4354 - val_accuracy: 0.8444\n",
      "Epoch 36/60\n",
      "801/801 [==============================] - 1s 913us/step - loss: 0.3254 - accuracy: 0.8727 - val_loss: 0.4240 - val_accuracy: 0.8333\n",
      "Epoch 37/60\n",
      "801/801 [==============================] - 1s 898us/step - loss: 0.3331 - accuracy: 0.8764 - val_loss: 0.3969 - val_accuracy: 0.7889\n",
      "Epoch 38/60\n",
      "801/801 [==============================] - 1s 895us/step - loss: 0.3096 - accuracy: 0.8777 - val_loss: 0.4517 - val_accuracy: 0.8000\n",
      "Epoch 39/60\n",
      "801/801 [==============================] - 1s 910us/step - loss: 0.2925 - accuracy: 0.8777 - val_loss: 0.4412 - val_accuracy: 0.8222\n",
      "Epoch 40/60\n",
      "801/801 [==============================] - 1s 916us/step - loss: 0.3061 - accuracy: 0.8826 - val_loss: 0.4694 - val_accuracy: 0.7778\n",
      "Epoch 41/60\n",
      "801/801 [==============================] - 1s 909us/step - loss: 0.3022 - accuracy: 0.8839 - val_loss: 0.4441 - val_accuracy: 0.8111\n",
      "Epoch 42/60\n",
      "801/801 [==============================] - 1s 919us/step - loss: 0.2826 - accuracy: 0.8876 - val_loss: 0.4194 - val_accuracy: 0.8333\n",
      "Epoch 43/60\n",
      "801/801 [==============================] - 1s 916us/step - loss: 0.2982 - accuracy: 0.8864 - val_loss: 0.4012 - val_accuracy: 0.8222\n",
      "Epoch 44/60\n",
      "801/801 [==============================] - 1s 918us/step - loss: 0.2640 - accuracy: 0.8939 - val_loss: 0.4402 - val_accuracy: 0.7778\n",
      "Epoch 45/60\n",
      "801/801 [==============================] - 1s 910us/step - loss: 0.2746 - accuracy: 0.8839 - val_loss: 0.4664 - val_accuracy: 0.7778\n",
      "Epoch 46/60\n",
      "801/801 [==============================] - 1s 946us/step - loss: 0.2605 - accuracy: 0.8989 - val_loss: 0.4191 - val_accuracy: 0.8000\n",
      "Epoch 47/60\n",
      "801/801 [==============================] - 1s 895us/step - loss: 0.2647 - accuracy: 0.8926 - val_loss: 0.4507 - val_accuracy: 0.8444\n",
      "Epoch 48/60\n",
      "801/801 [==============================] - 1s 899us/step - loss: 0.2670 - accuracy: 0.8976 - val_loss: 0.4310 - val_accuracy: 0.8222\n",
      "Epoch 49/60\n",
      "801/801 [==============================] - 1s 911us/step - loss: 0.2313 - accuracy: 0.9101 - val_loss: 0.4810 - val_accuracy: 0.8000\n",
      "Epoch 50/60\n",
      "801/801 [==============================] - 1s 923us/step - loss: 0.2414 - accuracy: 0.9101 - val_loss: 0.4414 - val_accuracy: 0.8111\n",
      "Epoch 51/60\n",
      "801/801 [==============================] - 1s 915us/step - loss: 0.2459 - accuracy: 0.9014 - val_loss: 0.4936 - val_accuracy: 0.7778\n",
      "Epoch 52/60\n",
      "801/801 [==============================] - 1s 919us/step - loss: 0.2250 - accuracy: 0.9189 - val_loss: 0.3923 - val_accuracy: 0.8222\n",
      "Epoch 53/60\n",
      "801/801 [==============================] - 1s 911us/step - loss: 0.1996 - accuracy: 0.9251 - val_loss: 0.4083 - val_accuracy: 0.8556\n",
      "Epoch 54/60\n",
      "801/801 [==============================] - 1s 914us/step - loss: 0.1908 - accuracy: 0.9301 - val_loss: 0.4454 - val_accuracy: 0.8000\n",
      "Epoch 55/60\n",
      "801/801 [==============================] - 1s 911us/step - loss: 0.2015 - accuracy: 0.9126 - val_loss: 0.4944 - val_accuracy: 0.7889\n",
      "Epoch 56/60\n",
      "801/801 [==============================] - 1s 908us/step - loss: 0.2228 - accuracy: 0.9126 - val_loss: 0.4220 - val_accuracy: 0.8556\n",
      "Epoch 57/60\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "801/801 [==============================] - 1s 916us/step - loss: 0.2040 - accuracy: 0.9226 - val_loss: 0.4038 - val_accuracy: 0.8444\n",
      "Epoch 58/60\n",
      "801/801 [==============================] - 1s 909us/step - loss: 0.2291 - accuracy: 0.9051 - val_loss: 0.4064 - val_accuracy: 0.8333\n",
      "Epoch 59/60\n",
      "801/801 [==============================] - 1s 908us/step - loss: 0.2173 - accuracy: 0.9189 - val_loss: 0.4639 - val_accuracy: 0.8333\n",
      "Epoch 60/60\n",
      "801/801 [==============================] - 1s 905us/step - loss: 0.1869 - accuracy: 0.9351 - val_loss: 0.4680 - val_accuracy: 0.8333\n"
     ]
    }
   ],
   "source": [
    "path_model='model_filter.h5' # save model at this location after each epoch\n",
    "K.tensorflow_backend.clear_session() # destroys the current graph and builds a new one\n",
    "model=my_model() # create the model\n",
    "K.set_value(model.optimizer.lr,1e-3) # set the learning rate\n",
    "# fit the model\n",
    "h=model.fit(x = X_train,     \n",
    "            y = y_train, \n",
    "            batch_size = 8, \n",
    "            epochs = 60, \n",
    "            verbose = 1, \n",
    "            validation_data = (X_test,y_test),\n",
    "            shuffle = False,\n",
    "            callbacks = [\n",
    "                ModelCheckpoint(filepath=path_model),\n",
    "            ]\n",
    "            )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAEICAYAAABPgw/pAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8vihELAAAACXBIWXMAAAsTAAALEwEAmpwYAAA1i0lEQVR4nO2de3gU5fXHvychAcJNEkCFQIIWilyTmIIgCIoXJIhKtYJUAa0IqFCtrSBVqZVWi8qlXihaRQVF2yr1Al6oUCnqT6KCgogi14gKBOSaQC7n98eZSXY3M7uzt+zu7Pk8zz67M/POO+ed3f3OmfOe9x1iZiiKoiiJT0qsDVAURVEigwq6oiiKS1BBVxRFcQkq6IqiKC5BBV1RFMUlqKAriqK4BBV0F0NEy4loTKTLxhIi2k5E50ehXiainxif5xPRXU7KhnCc0UT0dqh2Koo/SPPQ4wsiOuKxmAHgOIAqY/lGZl5c/1bFD0S0HcCvmHlFhOtlAJ2YeUukyhJRLoBtANKYuTIihiqKHxrE2gDFG2Zuan72J15E1EBFQokX9PcYH2jIJUEgokFEVEJEdxDR9wCeJqKWRPQ6Ee0logPG52yPfVYR0a+Mz2OJ6H9E9KBRdhsRXRxi2Y5E9B4RHSaiFUT0KBEtsrHbiY1/JKI1Rn1vE1Erj+3XENEOIioloul+zs9ZRPQ9EaV6rLuciD4zPvcmog+I6Eci+o6IHiGidJu6FhLRfR7LvzX22U1E1/mULSKiT4noEBHtIqIZHpvfM95/JKIjRNTXPLce+/cjorVEdNB47+f03AR5njOJ6GmjDQeIaKnHtkuJaJ3Rhm+IaIix3iu8RUQzzO+ZiHKN0NP1RLQTwLvG+n8Y38NB4zfSzWP/xkT0kPF9HjR+Y42J6A0iusWnPZ8R0WVWbVXsUUFPLE4BkAkgB8B4yPf3tLHcAUAZgEf87N8HwGYArQD8BcDfiYhCKPs8gI8AZAGYAeAaP8d0YuPVAMYBaAMgHcDtAEBEXQE8btTf1jheNixg5g8BHAVwnk+9zxufqwDcarSnL4DBACb5sRuGDUMMey4A0AmAb/z+KIBrAZwEoAjARA8hOsd4P4mZmzLzBz51ZwJ4A8A8o20PA3iDiLJ82lDn3FgQ6Dw/BwnhdTPqmm3Y0BvAswB+a7ThHADbbY5hxUAAZwC4yFheDjlPbQB8AsAzRPgggDMB9IP8jn8HoBrAMwB+aRYiol4A2gFYFoQdCgAws77i9AX5Y51vfB4E4ASARn7K5wE44LG8ChKyAYCxALZ4bMsAwABOCaYsRCwqAWR4bF8EYJHDNlnZ+HuP5UkA3jQ+3w1gice2JsY5ON+m7vsAPGV8bgYR2xybsr8G8IrHMgP4ifF5IYD7jM9PAbjfo1xnz7IW9c4BMNv4nGuUbeCxfSyA/xmfrwHwkc/+HwAYG+jcBHOeAZwKEc6WFuX+Ztrr7/dnLM8wv2ePtp3mx4aTjDItIBecMgC9LMo1BLAf0i8BiPA/Fo3/lNtf6qEnFnuZudxcIKIMIvqbcQt7CHKLf5Jn2MGH780PzHzM+Ng0yLJtAez3WAcAu+wMdmjj9x6fj3nY1NazbmY+CqDU7lgQb3wEETUEMALAJ8y8w7CjsxGG+N6w408Qbz0QXjYA2OHTvj5EtNIIdRwEMMFhvWbdO3zW7YB4pyZ258aLAOe5PeQ7O2Cxa3sA3zi014qac0NEqUR0vxG2OYRaT7+V8WpkdSxmPg7gJQC/JKIUAKMgdxRKkKigJxa+KUm/AfBTAH2YuTlqb/HtwiiR4DsAmUSU4bGuvZ/y4dj4nWfdxjGz7Aoz8xcQQbwY3uEWQEI3X0K8wOYA7gzFBsgdiifPA3gVQHtmbgFgvke9gVLIdkNCJJ50APCtA7t88Xeed0G+s5Ms9tsF4HSbOo9C7s5MTrEo49nGqwFcCglLtYB48aYN+wCU+znWMwBGQ0Jhx9gnPKU4QwU9sWkGuY390YjH3hPtAxoebzGAGUSUTkR9AVwSJRv/CWAYEfU3OjDvReDf7PMAJkME7R8+dhwCcISIugCY6NCGlwCMJaKuxgXF1/5mEO+33IhHX+2xbS8k1HGaTd3LAHQmoquJqAERXQWgK4DXHdrma4fleWbm7yCx7ceMztM0IjIF/+8AxhHRYCJKIaJ2xvkBgHUARhrlCwFc4cCG45C7qAzIXZBpQzUkfPUwEbU1vPm+xt0UDAGvBvAQ1DsPGRX0xGYOgMYQ7+dDAG/W03FHQzoWSyFx6xchf2Qr5iBEG5l5I4CbICL9HYADAEoC7PYCpL/hXWbe57H+dojYHgbwhGGzExuWG214F8AW492TSQDuJaLDkJj/Sx77HgMwE8Aakuyas3zqLgUwDOJdl0I6CYf52O2UOfB/nq8BUAG5S9kD6UMAM38E6XSdDeAggP+i9q7hLohHfQDAH+B9x2PFs5A7pG8BfGHY4cntAD4HsBYSM38A3hr0LIAekD4ZJQR0YJESNkT0IoAvmTnqdwiKeyGiawGMZ+b+sbYlUVEPXQkaIvoZEZ1u3KIPgcRNl8bYLCWBMcJZkwAsiLUtiYwKuhIKp0BS6o5AcqgnMvOnMbVISViI6CJIf8MPCBzWUfygIRdFURSXoB66oiiKS4jZ5FytWrXi3NzcWB1eURQlIfn444/3MXNrq20xE/Tc3FwUFxfH6vCKoigJCRH5ji6uQUMuiqIoLkEFXVEUxSWooCuKoriEuHpiUUVFBUpKSlBeXh64sBITGjVqhOzsbKSlpcXaFEVRfIgrQS8pKUGzZs2Qm5sL++cuKLGCmVFaWoqSkhJ07Ngx1uYoiuJDXIVcysvLkZWVpWIepxARsrKy9A5KUUJk8WIgNxdISZH3xRF+5HtceegAVMzjHP1+FCU0Fi8Gxo8HjhmPhtmxQ5YBYPToyBwjrjx0RVGUaBNtL9mO6dNrxdzk2DFZHylU0D0oLS1FXl4e8vLycMopp6Bdu3Y1yydOnPC7b3FxMSZPnhzwGP369QtYRlGU6GB6yTt2AMy1XnJ9iPrOncGtD4WEFvRIX2mzsrKwbt06rFu3DhMmTMCtt95as5yeno7KykrbfQsLCzFv3ryAx3j//ffDM1JRFMf4asSUKdH3ku3o4PvwwgDrQyFhBb2+rrRjx47FbbfdhnPPPRd33HEHPvroI/Tr1w/5+fno168fNm/eDABYtWoVhg0bBgCYMWMGrrvuOgwaNAinnXaal9A3bdq0pvygQYNwxRVXoEuXLhg9erT5BHQsW7YMXbp0Qf/+/TF58uSaej3Zvn07BgwYgIKCAhQUFHhdKP7yl7+gR48e6NWrF6ZOnQoA2LJlC84//3z06tULBQUF+OabcJ4LrCjxj5VGlNo8YtyplxyOEzlzJpCR4b0uI0PWRwxmjsnrzDPPZF+++OKLOuvsyMlhlq/J+5WT47gKv9xzzz08a9YsHjNmDBcVFXFlZSUzMx88eJArKiqYmfmdd97hESNGMDPzypUruaioqGbfvn37cnl5Oe/du5czMzP5xIkTzMzcpEmTmvLNmzfnXbt2cVVVFZ911lm8evVqLisr4+zsbN66dSszM48cObKmXk+OHj3KZWVlzMz81VdfsXk+ly1bxn379uWjR48yM3NpaSkzM/fu3ZtffvllZmYuKyur2R4KwXxPihIr7DQiVN1YtIg5I8N7v4wMWe+URYvkWETyHsy+JgCK2UZXE9ZDr494lMmVV16J1NRUAMDBgwdx5ZVXonv37rj11luxceNGy32KiorQsGFDtGrVCm3atMEPP/xQp0zv3r2RnZ2NlJQU5OXlYfv27fjyyy9x2mmn1eR5jxo1yrL+iooK3HDDDejRoweuvPJKfPHFFwCAFStWYNy4ccgwXIHMzEwcPnwY3377LS6//HIAMjgow9dVUJQEx9d73mE7hZU3Tr3kYDs1rbz50aOB7duB6mp5j1R2i0nCCnp9xKNMmjRpUvP5rrvuwrnnnosNGzbgtddes83JbtiwYc3n1NRUy/i7VRl2+MCR2bNn4+STT8b69etRXFxc02nLzHVSC53WqSiJilV4xS7DNisLyMmR7Tk5wIIFzoQ1GCcyVp2vCSvo9RKPsuDgwYNo164dAGDhwoURr79Lly7YunUrtm/fDgB48UXrh9MfPHgQp556KlJSUvDcc8+hqqoKAHDhhRfiqaeewjHDldi/fz+aN2+O7OxsLF26FABw/Pjxmu2K4gasvGfmuqKekQHMnVvXS7bypn3XZWZaH9vKiayPFEUrElbQR4+WK2soV9pw+N3vfodp06bh7LPPrhHRSNK4cWM89thjGDJkCPr374+TTz4ZLVq0qFNu0qRJeOaZZ3DWWWfhq6++qrmLGDJkCIYPH47CwkLk5eXhwQcfBAA899xzmDdvHnr27Il+/frh+++/j7jtihIOTkTVzsO1856ZA2uElTc9bhxw3XXe6w4dAtLTvfe1cyLrMyTshV1wPdqvcDtF3czhw4eZmbm6uponTpzIDz/8cIwt8ka/JyXSWHU4pqUxp6c764QMJ0kimM7TrCxnnZrRTNqAGztF3cwTTzyBvLw8dOvWDQcPHsSNN94Ya5MUJapYhSgqKgDf8Xx2YYtwQrDBeM379zsL18QqJKweuhI0+j0pkYbIuZdMZF1HqCmB4aQ3+ktljESKohUI10MnoiFEtJmIthDRVIvtLYnoFSL6jIg+IqLuEb/yKIriWoLJTsvMtI6rh5oSaOVNp6U5i5f76/yMdoqiFQEFnYhSATwK4GIAXQGMIqKuPsXuBLCOmXsCuBbA3EgbqihK/OO0E9O33NChzkQ1LQ04fDiy6YBWCRZPPw089VTgDtWYdX7aYee6my8AfQG85bE8DcA0nzJvAOjvsfwNgJP91ashl8RFv6fkwmnowOlISrtyEyfWPY7vsbOyotfZGArRHrFuBfyEXJzMh94OwC6P5RIAfXzKrAcwAsD/iKg3gBwA2QC8hkcS0XgA4wGgQzRGACmKElGCmcPbLvwwZYps27lTQitHjliXW7ZMQhO+eB4nxSamECuPeOZM7/MD1FPnpw1OYuhW4618hx7eD6AlEa0DcAuATwHUGRrJzAuYuZCZC1u3bh2srVFn0KBBeOutt7zWzZkzB5MmTfK7T3FxMQBg6NCh+PHHH+uUmTFjRk0+uB1Lly6tGb4PAHfffTdWrFgRhPWKEnmCGSBjJ6qlpZGbIKs+R4g7IVbjYexwIuglANp7LGcD2O1ZgJkPMfM4Zs6DxNBbA9gWKSPri1GjRmHJkiVe65YsWWI7n4ovy5Ytw0knnRTSsX0F/d5778X5558fUl2K4gQnA3ns5kOxEt9wRdXJ/jFLB/RDLDo/bbGLxZgvyGPqtgLoCCAdEl7p5lPmJADpxucbADwbqN54jKHv27ePW7VqxeXl5czMvG3bNm7fvj1XV1fzhAkT+Mwzz+SuXbvy3XffXbPPwIEDee3atczMnJOTw3v37mVm5vvuu487d+7MgwcP5pEjR/KsWbOYmXnBggVcWFjIPXv25BEjRvDRo0d5zZo13LJlS87NzeVevXrxli1beMyYMfyPf/yDmZlXrFjBeXl53L17dx43blyNfTk5OXz33Xdzfn4+d+/enTdt2lSnTdu2beP+/ftzfn4+5+fn85o1a2q2PfDAA9y9e3fu2bMn33HHHczM/PXXX/PgwYO5Z8+enJ+fz1u2bKlTZ6y/JyV4fGPREyc6G8hjl05oNcDGKjbu9BXMrIXRSgdMFOAnhu4oZxzAUABfQTo7pxvrJgCYwLUdp18D+BLAywBaBqozkKBPmcI8cGBkX1OmBD5ZQ4cO5aVLlzIz85///Ge+/fbbmbl2GtrKykoeOHAgr1+/npmtBb24uJi7d+/OR48e5YMHD/Lpp59eI+j79u2rOdb06dN53rx5zMxeAu65bE6nu3nzZmZmvuaaa3j27Nk1xzP3f/TRR/n666+v055oTLOrgh7fOBHvYPO+nY7gdNqJ6XTEpVIXf4LuKA+dmZcxc2dmPp2ZZxrr5jPzfOPzB8zciZm7MPMIZj4Q3n1D7PAMu3iGW1566SUUFBQgPz8fGzdu9AqP+LJ69WpcfvnlyMjIQPPmzTF8+PCabRs2bMCAAQPQo0cPLF682Hb6XZPNmzejY8eO6Ny5MwBgzJgxeO+992q2jxgxAgBw5pln1kzo5YlOs5tcWM1LMn++9cRVTmH2jhE3b24/gtM3/DB3rnWIxGqCLCV8nGS5xIQ5c2Jz3Msuuwy33XYbPvnkE5SVlaGgoADbtm3Dgw8+iLVr16Jly5YYO3as7bS5Jr5T2JqMHTsWS5cuRa9evbBw4UKsWrXKbz0c4J9nTsFrN0Wv5zS71dXVaNSoUU29Os1ufLJ4sXdWiBkfdrLObtbBcMjJ8c4+CSbTxBRqXztVwKODzuXiQ9OmTTFo0CBcd911Nd75oUOH0KRJE7Ro0QI//PADli9f7reOc845B6+88grKyspw+PBhvPbaazXbDh8+jFNPPRUVFRVY7DEaolmzZjh8+HCdurp06YLt27djy5YtAGTWxIEDBzpuj06zm1g4nfnPap25n1N8fQ6noyODzTSJq05Dl6OCbsGoUaOwfv16jBw5EgDQq1cv5Ofno1u3brjuuutw9tln+92/oKAAV111FfLy8vDzn/8cAwYMqNn2xz/+EX369MEFF1yALl261KwfOXIkZs2ahfz8fK/nfTZq1AhPP/00rrzySvTo0QMpKSmYMGGC47boNLvxjZOHGFtNUmU3cZXxYK06WM0LPmFCaKMj4zHTRDGwC65H+xWPWS6KM/R7igzhZIUEyhhxMgozXNu1UzM2IMyRooqiRAGreHe45OTUxtKjGbMePVpDJ/GICrqixAinw9XT0iQE4hlisVpnhj1UbJOXuIuhs2ZaxDX6/XjjdHZBq7J2z6j0fYixVWzbabxbSS7iykNv1KgRSktLkZWVZZv2p8QOZkZpaWlN6mOy42/iKsA77DF0KPDMM95lzawSXy977lxrYXa6TkleKFYeV2FhIZuTWplUVFSgpKQkYI63EjsaNWqE7OxspKWlxdoUL6xyt6MtdnZznWRlAWVl3vFxIut88KwsoGlTzdFWnENEHzNzodW2uPLQ09LS0LFjx1iboSQYwUzxGkn8zS7oi53ftH8/sG9f5GxSkpu4i6G7hUOHYm1B8hDMFK9WOJl10Co2HokpW/WxANGntFQGNYXCiRORz0SKJiroUWDTJuCUU4A//CHWliQHwTwGzFeoJ01yNjLT6jFndgNssrKs7bEa3KODcaLLzp1AdjawcGFo+0+fLv9ln8ckxC92CerRflkNLHID1dXM558vAzrS05ktZp9VIozTx4BZDeQJZtZBq8eKWQ2wCeYRa0p0ueceOf+/+EVo+xcUyP6pqcyPPhpR00IG4U6fG42XWwX9X/+Ss3rnncxNmzIPHx5ri9yP02dZ2gl/MNPIBmOTindsqaxkbt9evru2bcXZCobycpkq+KabmIcNk3qmTJF6Y4kKej1x7Jj8eXv0YK6oYH7gATnDy5fH2rLExM77tRJKJwIajDfu1ENX4pc33pDv7bzz5P2bb4Lb/+OPZb8XXxQRv/VWWS4qYj50KDo2O0EFvZ6YMUPO6KpVslxeztypE3PnzszHjzuv56WXmCdPjo6NiYKV1+3vwQpOcOqhh3scO6qrxdubNy94bzGRqaxkHjmSecWK+j3u5Zczt27NXFws3+EzzwS3/xNPyH5ff1277vHHJfzSowfzjh2RtdcpKuj1wLZtzI0aMV91lff6ZcvkLD/wgLN6qqqYTz9d9tm1K+JmxgVOvOlgwiNOPWeri0TjxszZ2d51+bsTCIf/+7/a4/zqV8wnToRfZyKwZk3tubV4AFZU+O475gYNmG+/Xf5TLVow33BDcHVMnCj7+V58336buXlz5pNPlu+0vlFBrwdGjBCxsBLhSy6RePq33wau5913a//08+dH3s5Y4zTeHewj0oI5vinUbdvKq0ED5sGDpa7vv3e2bygif8MN0tbf/EaONXgw8/79wdWRiEybxpySIm2+5576Oeb998vxvvxSlocOZe7SJbg6evdmHjTIetvGjcwdO4oT5/HkyHpBBT3KvPOOnMmZM623b9nC3LAh8y9/GbiuUaOYTzpJOnMuuSSydsYDTjNSouGhe/K//zG3asXcsiXzypUSJgPkjsoKpxciOw4flov6uHGyvHChhHZ++lP3Z0L17CnP9B05UgRw27boHq+6mvknP2E+55zadX/+s3xne/Y4q+PECfnP3nabfZk9e5j79ZN6//Sn+gujqaBHkfJy5jPOkDCJ8SxmS6ZPl7P9v//Zl9m3T2K3N98sr8aNpaM1GKqqmF94gfmvf/V+PfII8/btwdXFLIJlhiQiEXqw87x9vexgY+jHjzO/8op3X4WdR/3CC1JPp07MxrO3+ccfpb777rO22+mFyA4zHrtmTe26VauYMzPlgcnvv++snkhTXMwczentd+yQds+aJXevGRkS244mK1fKMZ99tnbd6tWyznj+e0A++0zKL17sv1xZGfPVV0vZkSPr/u/sXp6/g2BRQY8Se/cyDxggZ/H11/2XPXJEhDEvzz7tac4cqWvdOsmM8ecx2h3jssvsPdnMTOb33pOyTsIH0cinDkYYnWa57NsnHiAgF05/tj/4oIh5//6ynyedOknozAqnFyI7+vRh7tq1rhf31VfSjoICZ/VEkm+/lbuGxo2jFzZ4/HE5T+ZFY+ZMWX777egcj1kEtkULb2eorEy+99/8xlkdCxeKnZs2BS5bXS0JEcGECe+4I6SmMbMKelTYtEm88oYNxeNzwpIlcsYff7zutupq5u7dmX/2M1kuKxMBuukmZ3WXlIgopKTIhWHvXu/XZ5/J7X1aGvONN4aXt+37ww0m9BBu6MKXzZtFiNPT5dyZg7nsbG/cmLlZM+bdu+vWddVV9h53OB666e09/LD1djMc4KSPJZKMHi2/38JCjlrYYNgw5tNOq623vFz+N126BJf55ZTSUmmT1f+mf3+5sDrhlluYmzQJLuf80KG6/zu715Ejzuv1RQU9wqxYIXHuNm2YP/jA+X7V1dLJkplZ1zv88EP5Nv72t9p1w4eLYAT6k33yiXTuNW0qubd27N9fm5Nr9crK8vZ+nXobnuEYJx51pDJIVq6UGHirVhLKMj3OSy7x7y3NmmVdnzluwPe7YQ7vQjR5slxo9u613v7551LfE084bnrYmCGI6dPFeRg1SpbHjImc0B47JhfQW27xXv/aa3KsBx+MzHE8mTuXa+5yfZk6VTrAnWTanH22XADiERX0CLJggfwounULrXPns88kj3XiRO/1118vHoHngIW//U2+oQ0b7Ov7979FWNq3Z16/PvDxT5wITqiDeYWbN15dLX0ATnjqKfkezjjDe8CIKcpt2ljb2KCBvWCtWCFl3nnHeru/C5HdRbesTC46vumsnlRXM3fowHzppX4a7IdgverKSgn9tW9f6ylWV9cOkz/nHOuLWrCYA3vefLPutqFD7e+UQsW8yy0stN7++utiz8qV/uuprJT/YryOBVFBjxD//KecsYsukk60ULnlFgmNfPqpLB86JD+g66/3LldSIse7//66dSxaVCta6enS6WmuD+QRZ2ZGXsxTU52XtQpT/PCD9EeccYbEle2oqpL4I8B8wQXMBw54bz9+XAZynXKKeIe+x/YXuywtlTJOxwyYbNggdwnz5tXdtmiR1BloUM2kSfIb8NexbsWyZZIT7Tn4JRBmXPvFF63tTU+X7yJcJk60b9NXX8lxrr02/OOYrF3Lde5yPdm/X/4D997rv55Nm6SehQsjZ1skUUGPEAMHSjywoiK8evbvFwHo31+8igUL5JuwCt/k5dX9cy1aVFeszM5KJ16y1bpAAgzIn6FbN+vQQzDi79uRuHGjjOgzwyQpKcy//33dc3H0qHRaAtIPYDcwx+xQHjnS23YnP7nc3OAncjJzngGJ3Xr+PgYNknzlQHce/rxZO8rK5Pdod9G3Yt8+uaCfe669Z292XJaUOLfFFyd3HVOnynEileFz113y2/F3d9GjB/OFF/qvx7wIf/ZZZOyKNCroEWDzZjlbf/pTZOoz09gWL5YBDN27W//Bpk+XH2lpae06u/h2MF6yGS8397OLOXt602ZM/7nn6t4FhJo3/vbb1p60Kdomu3fLrTQR8+zZgcMMw4fXDuYaNUo6ypzM5fHzn0sOczAMGyZ3Bb/9rdg9ZAjzwYPihQL24xM8sYs3++O++2q/S6ce9cSJ8n37EyuzE3fBAue2+OKkX+DwYen7OfPMyEx4VVAgsW9/TJwovwt/Ttltt0m+fLiOW7RQQY8Av/ud/BEiFfOrqhKBatlSvoU5c6zLffCBbH/++dp1oYZF7LzkkhJrQW7c2DtObMb0P/+8rp2hzL1izouRlmZv5/Tp0umbnS2376++6uz8fvONiPjPfib13HWXs/1M79RpSK2qSr7DX/1Klp94oraP5ZprpH1Os1eGDRNv3klMfOdOOZc//7nczaSmel/0rfj0U3EOAl00wo3pMzvP3Fm8OPyLB7McB5Dj+uP556Xcxx/blxk0yHk2TCxQQffgsceYr7ii7uvpp+33OX5c4tVOf+BOszhMsW7YsPbP6Lvvs89KeObqq2v3a9LEWvzCiWMfPiyekrm9Vau6dvuL6du12+5cmAOthg4NbGtKigi62efglN//Xvbv0MH5HCJmuCZQx5mJ6Yl6xltXrJA8aCC46ZN9c7b98Ytf1I66tLro+1JdLV58q1bOphsww3fBxvRN+vdnzs8PXM6JXa+/bp+ZZGLe8QYKk+zcKeXmzrXeXlUlfRK+SQvxhAq6Bzk58mfr2rX21a6diIZVqhNz7RzngQYPMQc3GGfRoto/fk6OdQw8I0Nu5825ME4+Wd4bNLA+RjiZJpWVzHffzTxhgr2XaBXTDxZTgMaOldtau3BNhw5y53LJJaHdGR05wnzllcHN8vfDD3Lshx5yVt4UYd/h+5s2MV98cXCTN5liE0i8zPl+ZsyQ5cpK6YMYPdp+n48+kn2sOm6tCCWmb7Jvn/xend4VrVsn5W++2Xt9dXWtpw/4z/a67DLJ2nFyd9Ohg/wurPj6aznWk086sz0WhC3oAIYA2AxgC4CpFttbAHgNwHoAGwGMC1RnrAS9TRvv2CxzbSflgAHWP4ghQ0T0ncT5nA7GsRJapyPNiOQW3+m84JGcOfD3v68b0w8GM9R06qm1KZqRHmwULtnZ/sXRk9GjpS2RGpBjzntiR0WF9Lfk5nqPhLz2WunstPuNmhODOQ0lmTF9X5F1ghlG+fBD5/tMmiS/KzP19vhxmfcGkKkC0tKYf/1r633Ly+Wu1alXffXVkgVl9Z29+KIc85NPnNte34Ql6ABSAXwD4DQA6YZod/UpcyeAB4zPrQHsB5Dur95YCXqzZtY/DDPTxPe2dccOEUKn3ka4D1EINWRSXzi5vffHk0/K/r5iHU9P+Bk+XNInneDP2wuFO++U0JlvOqaJOXDm5Ze915tCZDVXkDkx2NixwdkSTEzfk1Gj5I7B6ZgCZnEQMjPlYrZvn8SxAbljrK6WsGhmpoi3L2+9xY7voJnt76qYJa01PT06o1gjRbiC3hfAWx7L0wBM8ykzDcBjAAhAR8OTT/FXb6wEPS1NpvP0pbJSYsht28ofwOSee0RknA4iCvcxZ05fwUwZG0kqK+vG9J1y4ID80c8+O74f8GDOy+H5O7DCnHjKLh4bCubc4UuW1N22Z4+E6C64oO75O3BAwnBTp9bdz7yIBjshlCl8Gzc636eiQjqJx4wJ7liex2vdWkTV86JuirbVNBu33BLcRHZW/R4m55/vLL01loQr6FcAeNJj+RoAj/iUaQZgJYDvABwBUGRT13gAxQCKO3ToUH9nwKCyUlpsN7Dg/fdlu/mnMJ9JGChv1ZNwH0TsW9ZJOmF9E+j23o7Jk6U98Xw7y1w7NN3fzJjMtRkTkWxPZaWkIV5zjff6nTuZe/USh8Su03TQIMmz9uWss+SOI9iLqBnT/8tfnO9jzlf0z38GdyxmaXthYe1UDp5UVclv/rzzvNdXV0suflGR8+NUVUlfVG6ud1y+ulp+18E+CKO+8SfoKQgMWaxjn+WLAKwD0BZAHoBHiKh5nZ2YFzBzITMXtm7d2sGhI0tZmbw3amS9vW9f4NprgYceAr7+Gnj7bWDXLuCGG+zrXLwYyM0FUlLkHQAWLABycgAieZ8wAcjI8N4vLQ1IT/del5EhZQPtm5EBzJzptNWRp6gI2L8f+PBD5/ts2AA8+ihw441Afn70bIsEBQXy/skn/sutXg00awb06BG5Y6emAhdfDCxfDlRVybriYqBPH2DrVuDVV4EzzrDed9gw4PPPgZ07a9dt2CDf0w03yG8qGNq3B3r2BN54w1n5o0eB228H8vKAyy4L7liAtH3lSuCbb4Czz/belpICXH898O67st1k82Y5L0VFzo+TkgK8/jpQXg706yf/c0DO2/79td9/QmKn9OYLzkIubwAY4LH8LoDe/uqNRchlzx7xHv76V/sy330ncfaiIhmV2Lq1xNPsOhudduaF01kZT/FlZv+391ZUV8vIRKtJyeKR6mrx4ALFnJ2MOgwF08t9/33JsGrcWL53q/x/T8wh6489VrtuyhT/E4MFYto0iek7SXU0U1FXrw7tWIHYtUs6Tj1DprNmyTFDeb7nzp3SCZ2aKufs5Zelrlg8Vi4YEGbIpQGArZDYuNkp2s2nzOMAZhifTwbwLYBW/uqNhaCbMc9AKUkPPlgb7rj9dnvhzsqKv3BIfWF3e2/FSy/VFZp4Z+hQ/+0z5wX54x8jf+wDB0RkevWS89anj/9H45mY4YehQ2W5rEwuov4mBguEv5i+J1u2yIXDaXZQqBQVSYaKOYozmN+hFYcOSZ2AdACnpoaee19fhCXosj+GAvgKku0y3Vg3AcAE43NbAG8D+BzABgC/DFRnLAT9yy+lxZ5PIbHyfhcurM3zbtvWXrjjrcOyPjEvem3aBH41bOj/wR7xiDn60q6jzczTdjoAKVjMB3ZcdVVwT62aPFkGHB09Whvjt5s90glmTD/Q4xODeW5uOCxdKm1aurT2TtEqySEYKisl8w0I7+JQX/gT9AYOwzLLACzzWTff4/NuABc6qSuWmDH0xo3lffFiYPx44NgxWd6xAxg3TmKNlZWybvfu4I/ToUP4tsY7Y8cCJSUShwxEgwbALbdIjDRRKCiQGPbnnwO9e9fdvnq19INYbYsEDz0EfPopcN11EvN1SlERMG+exJqffBLo2BE477zQ7fCN6Vt9h8uXA6+9BjzwANC2bejHckJREXDqqdK248flfxpM/NyK1FRg9myJ22dlRcbOmGGn9NF+xcJDN7NYli+X5XBTDLOy4mtAjBI5zPCc1dOlmGVo+1ln1a9NTjAH2Vx4odhv94zUYHjhBbZNezx+XJ4Y1blz/eVuT5smsfTBg0PLtkp0EGaWi2swvUnTQ/fMBgiWjAxg7ty6GS0LFgCjR4dvqxJb2rcXb+3jj+tuKy8HPvoI6N+//u0KRMOGwAUXSOZGSorccYbLRReJF2uV7TJnjmSEzZ1bN2srWlx/PVBdDfznP3L3kEh3ftHGUcjFLSwzgkaDBon4ZmYCpaXO9s3KApo2lYtAhw6SNmgKtwq4+yCSsMtHH8m9l2fKX3ExcOIEMGBA7OzzR1ERsHSpvEciBNKypYQjFi2qDUUCcl4efxy45BJgyJDwj+OU00+XMNK774YfbnEddq57tF/1HXIxn8TiGR5x+vAHDaMkJ7Nny/d/9dXemQ/mhFGhpgJGmx9+kDDIqlWRq/Ppp+V/0KiR9+u006yH0EebN9+UB03bTZHgZuAn5EKyvf4pLCzk4uLiejtebq50evpi5XkDwPTp1t64kjwwA/ffD9x5pwxAWboUaN1aBvBs3Qp88UWsLVSSESL6mJkLrbYlTcjFLl6+fz+wb1/d9SrgChEwbRrQqRNwzTUyWvPVV4E1a4Bf/CLW1ilKXZKmU9QulTAZUgyV8LjiCuC//5X01p/9DPjxx/jsEFWUpBH0mTMlb9iTWM+JoiQOvXtLB2mnTpI9Eq8dokpykzSCPnq0xD5NNMVQCZYOHYAPPpBURnMiNkWJJ1wr6L6zIC5eDHTuLHm6zMD27SrmSvA0aSKzCSpKPOLKTlGrIf3jx0vc0xxUpCiK4jZc4aH7euNTptSKucmxY8D779vPha4oipLoJLyHbuWN23HkiOQRK4qiuJGE99CnT6/rjduRkaEhF0VR3EvCe+hOJ9jKyJCUswYJ32JFURRrEt5DtxsYlJVVdxbEzEz10BVFcS8JL+gzZ1o/RHnuXElNrK6uTVEsK1NBVxTFvSS8oI8e7XxO8vJyFXRFUdyLKyLKo0c7GySkHrqiKG4m4T30YCgr0zx0RVHcS9IJunroiqK4FRV0RVEUl6CCriiK4hISStCtZlB0SkUFUFWlgq4ointJmCwXuxkUAWcZLuXl8q6CriiKW0kYD91qzpZjx2S9E8rK5F0FXVEUt5Iwgm43Z4vTuVxU0BVFcTsJI+jhPuTZFHTNQ1cUxa0kjKDbzdni9CHP6qEriuJ2EkbQg5mzxQoVdEVR3E7CZLkAzudssUIFXVEUt+PIQyeiIUS0mYi2ENFUi+2/JaJ1xmsDEVURUWbkzQ0dFXRFUdxOQEEnolQAjwK4GEBXAKOIqKtnGWaexcx5zJwHYBqA/zLz/ijYGzKah64oittx4qH3BrCFmbcy8wkASwBc6qf8KAAvRMK4SKIeuqIobseJoLcDsMtjucRYVwciygAwBMC/bLaPJ6JiIireu3dvsLaGhaYtKoridpwIOlmsY5uylwBYYxduYeYFzFzIzIWtW7d2amNEUA9dURS340TQSwC091jOBrDbpuxIxGG4BVBBVxTF/TgR9LUAOhFRRyJKh4j2q76FiKgFgIEA/h1ZEyODhlwURXE7AfPQmbmSiG4G8BaAVABPMfNGIppgbJ9vFL0cwNvMfDRq1oaB+fg5sgogKYqiuABHA4uYeRmAZT7r5vssLwSwMFKGRRp9uIWiKG4nYYb+h0t5uQq6oijuJmkEXT10RVHcTlIJunaIKoriZpJK0NVDVxTFzaigK4qiuAQVdEVRFJeggq4oiuISVNAVRVFcQtIIuuahK4ridpJG0NVDVxTF7SSVoGseuqIobiYpBJ1ZPXRFUdxPUgh6RQVQXa2CriiKu0kKQdeHWyiKkgyooCuKorgEFXRFURSXkBSCXl4u7yroiqK4maQQdPXQFUVJBpJK0DUPXVEUN5NUgq4euqIobkYFXVEUxSWooCuKorgEFXRFURSXkBSCrmmLiqIkA0kh6OqhK4qSDKigK4qiuISkEXQiID091pYoiqJEj6QR9EaNRNQVRVHcStIIuoZbFEVxOyroiqIoLsGRoBPRECLaTERbiGiqTZlBRLSOiDYS0X8ja2Z4qKAripIMNAhUgIhSATwK4AIAJQDWEtGrzPyFR5mTADwGYAgz7ySiNlGyNyTKy1XQFUVxP0489N4AtjDzVmY+AWAJgEt9ylwN4GVm3gkAzLwnsmaGh3roiqIkA04EvR2AXR7LJcY6TzoDaElEq4joYyK61qoiIhpPRMVEVLx3797QLA4BM8tFURTFzTgRdKtkP/ZZbgDgTABFAC4CcBcRda6zE/MCZi5k5sLWrVsHbWyoqIeuKEoy4ETQSwC091jOBrDbosybzHyUmfcBeA9Ar8iYGD4q6IqiJANOBH0tgE5E1JGI0gGMBPCqT5l/AxhARA2IKANAHwCbImtq6KigK4qSDATMcmHmSiK6GcBbAFIBPMXMG4logrF9PjNvIqI3AXwGoBrAk8y8IZqGB4MKuqIoyUBAQQcAZl4GYJnPuvk+y7MAzIqcaZFDBV1RlGQgKUaKah66oijJgOsFnVk9dEVRkgPXC/qJEyLqmoeuKIrbcb2g68MtFEVJFlTQFUVRXIIKuqIoiktQQVcURXEJKuiKoiguwfWCXl4u7yroiqK4HdcLunroiqIkC0kj6JqHriiK20kaQVcPXVEUt6OCriiK4hJU0BVFUVyCCrqiKIpLUEFXFEVxCa4X9PJyICUFSEuLtSWKoijRxfWCbs6FThRrSxRFUaJLUgi65qAripIMJIWga/xcUZRkQAVdURTFJaigK4qiuAQVdEVRFJfgekEvL1dBVxQlOXC9oKuHrihKsqCCriiK4hKSQtA1D11RlGQgKQRdPXRFUZIBFXRFURSXoIKuKIriEhwJOhENIaLNRLSFiKZabB9ERAeJaJ3xujvypgYPs6YtKoqSPDQIVICIUgE8CuACACUA1hLRq8z8hU/R1cw8LAo2hszx4/Kugq4oSjLgxEPvDWALM29l5hMAlgC4NLpmRQZ9uIWiKMmEE0FvB2CXx3KJsc6XvkS0noiWE1G3iFgXJqaga9qioijJQMCQCwCrR0Owz/InAHKY+QgRDQWwFECnOhURjQcwHgA6dOgQnKUhoB66oijJhBMPvQRAe4/lbAC7PQsw8yFmPmJ8XgYgjYha+VbEzAuYuZCZC1u3bh2G2c5QQVcUJZlwIuhrAXQioo5ElA5gJIBXPQsQ0SlE8pA3Iupt1FsaaWODRQVdUZRkImDIhZkriehmAG8BSAXwFDNvJKIJxvb5AK4AMJGIKgGUARjJzL5hmXpHBV1RlGTCSQzdDKMs81k33+PzIwAeiaxp4aOCrihKMuHqkaLl5fKugq4oSjLgakFXD11RlGQiKQRd89AVRUkGkkLQ1UNXFCUZUEFXFEVxCSroiqIoLsH1gp6aCqSlxdoSRVGU6ON6QVfvXFGUZMHVgq4Pt1AUJZlwtaCrh64oSjLhekHXHHRFUZIF1wu6euiKoiQLKuiKoiguQQVdURTFJaigK4qiuAQVdEVRFJfgakHXPHRFUZIJVwu6euiKoiQTrhd0zUNXFCVZcL2gq4euKEqy4Ogh0fHEW28Bt93mrOzx4yroiqIkDwkn6M2bA127OivbowcwYkR07VEURYkXEk7Q+/YF/vGPWFuhKIoSf7g6hq4oipJMqKAriqK4BBV0RVEUl6CCriiK4hJU0BVFUVyCCrqiKIpLUEFXFEVxCSroiqIoLoGYOTYHJtoLYEeIu7cCsC+C5sQabU/84qa2AO5qj5vaAjhvTw4zt7baEDNBDwciKmbmwljbESm0PfGLm9oCuKs9bmoLEJn2aMhFURTFJaigK4qiuIREFfQFsTYgwmh74hc3tQVwV3vc1BYgAu1JyBi6oiiKUpdE9dAVRVEUH1TQFUVRXELCCToRDSGizUS0hYimxtqeYCGip4hoDxFt8FiXSUTvENHXxnvLWNroFCJqT0QriWgTEW0koinG+kRtTyMi+oiI1hvt+YOxPiHbAwBElEpEnxLR68ZyIrdlOxF9TkTriKjYWJeQ7SGik4jon0T0pfH/6RuJtiSUoBNRKoBHAVwMoCuAUUTk8IF0ccNCAEN81k0F8B9m7gTgP8ZyIlAJ4DfMfAaAswDcZHwfidqe4wDOY+ZeAPIADCGis5C47QGAKQA2eSwnclsA4FxmzvPI107U9swF8CYzdwHQC/Idhd8WZk6YF4C+AN7yWJ4GYFqs7QqhHbkANngsbwZwqvH5VACbY21jiO36N4AL3NAeABkAPgHQJ1HbAyDbEIbzALxurEvIthj2bgfQymddwrUHQHMA22AkpUSyLQnloQNoB2CXx3KJsS7ROZmZvwMA471NjO0JGiLKBZAP4P+QwO0xQhTrAOwB8A4zJ3J75gD4HYBqj3WJ2hYAYABvE9HHRDTeWJeI7TkNwF4ATxvhsCeJqAki0JZEE3SyWKd5lzGGiJoC+BeAXzPzoVjbEw7MXMXMeRDvtjcRdY+xSSFBRMMA7GHmj2NtSwQ5m5kLICHXm4jonFgbFCINABQAeJyZ8wEcRYRCRYkm6CUA2nssZwPYHSNbIskPRHQqABjve2Jsj2OIKA0i5ouZ+WVjdcK2x4SZfwSwCtLfkYjtORvAcCLaDmAJgPOIaBESsy0AAGbebbzvAfAKgN5IzPaUACgx7v4A4J8QgQ+7LYkm6GsBdCKijkSUDmAkgFdjbFMkeBXAGOPzGEgsOu4hIgLwdwCbmPlhj02J2p7WRHSS8bkxgPMBfIkEbA8zT2PmbGbOhfxP3mXmXyIB2wIARNSEiJqZnwFcCGADErA9zPw9gF1E9FNj1WAAXyASbYl1B0EIHQpDAXwF4BsA02NtTwj2vwDgOwAVkCv19QCyIJ1XXxvvmbG202Fb+kNCXp8BWGe8hiZwe3oC+NRozwYAdxvrE7I9Hu0ahNpO0YRsCyTuvN54bTT/+wncnjwAxcZvbSmAlpFoiw79VxRFcQmJFnJRFEVRbFBBVxRFcQkq6IqiKC5BBV1RFMUlqKAriqK4BBV0RVEUl6CCriiK4hL+H+YQgs9p4BtvAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "acc = h.history['accuracy']\n",
    "val_acc = h.history['val_accuracy']\n",
    "loss = h.history['loss']\n",
    "val_loss = h.history['val_loss']\n",
    "\n",
    "epochs = range(len(acc))\n",
    "\n",
    "plt.plot(epochs, acc, 'bo', label='Training acc')\n",
    "plt.plot(epochs, val_acc, 'b', label='Validation acc')\n",
    "plt.title('Training and validation accuracy')\n",
    "plt.legend()\n",
    "\n",
    "plt.figure()\n",
    "\n",
    "plt.plot(epochs, loss, 'bo', label='Training loss')\n",
    "plt.plot(epochs, val_loss, 'b', label='Validation loss')\n",
    "plt.title('Training and validation loss')\n",
    "plt.legend()\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "tensorflow",
   "language": "python",
   "name": "tensorflow"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}