tensorflow/machine_learning_CNN_Keras.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Machine learning with TensorFlow: CNN\n",
    "\n",
    "Date: November 20, 2018\n",
    "\n",
    "Neural network: convolution, fully-connected and softmax\n",
    "\n",
    "Classification of human body motion:\n",
    "- walking\n",
    "- sitting down\n",
    "- turning right while walking\n",
    "- turning left while walking"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "\n",
    "import matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import scipy.signal as sg\n",
    "import serial\n",
    "import time\n",
    "import random"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Preparing data set for training CNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "SAMPLE_NUMS = 260  # per sec\n",
    "SAMPLING_RATE = 80.0  # Hz\n",
    "GYRO_RESOLUTION = 250.0 / 32768.0\n",
    "ACCEL_RESOLUTION = 2.0 / 32768.0\n",
    "\n",
    "TRAINING_SET = [['./20180821_walking_straight.csv', 0],\n",
    "               ['./20180821_sitting.csv', 1],\n",
    "               ['./20180821_turning_left.csv', 2],\n",
    "               ['./20180821_turning_right.csv', 3],                \n",
    "               ['./20180823_walking_straight.csv', 0],\n",
    "               ['./20180823_sitting.csv', 1],\n",
    "               ['./20180823_turning_left.csv', 2],\n",
    "               ['./20180823_turning_right.csv', 3]]\n",
    "\n",
    "TEST_SET = [['./20180822_walking_straight.csv', 0],\n",
    "            ['./20180822_sitting.csv', 1],\n",
    "            ['./20180822_turning_left.csv', 2],\n",
    "            ['./20180822_turning_right.csv', 3]]\n",
    "\n",
    "TIME_INTERVAL = 260.0 / SAMPLING_RATE  # sec\n",
    "\n",
    "MEASUREMENTS = 8\n",
    "RECORDS = 260"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "gres = lambda v: v * GYRO_RESOLUTION\n",
    "ares = lambda v: v * ACCEL_RESOLUTION\n",
    "to_time = lambda v: v / SAMPLING_RATE\n",
    "\n",
    "def conv(df, lpf=False):\n",
    "    df[['gx', 'gy', 'gz']] = df[['gx', 'gy', 'gz']].apply(gres)\n",
    "    df[['ax', 'ay', 'az']] = df[['ax', 'ay', 'az']].apply(ares)\n",
    "    if lpf:\n",
    "        CUTOFF = 10.0\n",
    "        b, a = sg.butter(5, CUTOFF/SAMPLING_RATE, btype='low')\n",
    "        df[['ax', 'ay', 'az']] = df[['ax', 'ay', 'az']].apply(lambda row: sg.lfilter(b, a, row))\n",
    "        CUTOFF = 10.0\n",
    "        b, a = sg.butter(5, CUTOFF/SAMPLING_RATE, btype='low')\n",
    "        df[['gx', 'gy', 'gz']] = df[['gx', 'gy', 'gz']].apply(lambda row: sg.lfilter(b, a, row))\n",
    "    df[['cnt']] = df[['cnt']].apply(to_time)\n",
    "    df.set_index('cnt', drop=True, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "training_set = []\n",
    "test_set = []\n",
    "\n",
    "for file, l in TRAINING_SET:\n",
    "    df = pd.read_csv(file, dtype=np.int16)\n",
    "    conv(df, lpf=True)\n",
    "    for i in range(MEASUREMENTS):\n",
    "        training_set.append([df[df['id']==i], l])\n",
    "        \n",
    "for file, l in TEST_SET:\n",
    "    df = pd.read_csv(file, dtype=np.int16)\n",
    "    conv(df, lpf=True)\n",
    "    for i in range(MEASUREMENTS):\n",
    "        test_set.append([df[df['id']==i], l])\n",
    "    \n",
    "random.shuffle(training_set)\n",
    "random.shuffle(test_set)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "# use accel x axis and gyro z axis values for classification of human body motion\n",
    "train_x, train_t = [], []\n",
    "test_x, test_t = [], []\n",
    "for df, label in training_set:\n",
    "    values = []\n",
    "    for index, item in df.iterrows():\n",
    "        values.append([item['ax'],item['ay'],item['az'],item['gx'],item['gy'],item['gz']])\n",
    "    train_x.append(values)\n",
    "    train_t.append(label)\n",
    "for df, label in test_set:\n",
    "    values = []\n",
    "    for index, item in df.iterrows():\n",
    "        values.append([item['ax'],item['ay'],item['az'],item['gx'],item['gy'],item['gz']])                     \n",
    "    test_x.append(values)\n",
    "    test_t.append(label)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(64, 260, 6)"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from keras.utils import to_categorical\n",
    "\n",
    "train_labels=to_categorical(train_t)\n",
    "test_labels=to_categorical(test_t)\n",
    "\n",
    "train_data = np.array(train_x)\n",
    "test_data = np.array(test_x)\n",
    "\n",
    "train_data.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Keras/TensorFlow"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "num_samples = 260\n",
    "num_classes = 4\n",
    "num_layers = 6"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv1d_4 (Conv1D)            (None, 249, 4)            292       \n",
      "_________________________________________________________________\n",
      "max_pooling1d_4 (MaxPooling1 (None, 124, 4)            0         \n",
      "_________________________________________________________________\n",
      "conv1d_5 (Conv1D)            (None, 113, 8)            392       \n",
      "_________________________________________________________________\n",
      "max_pooling1d_5 (MaxPooling1 (None, 56, 8)             0         \n",
      "_________________________________________________________________\n",
      "conv1d_6 (Conv1D)            (None, 45, 16)            1552      \n",
      "_________________________________________________________________\n",
      "max_pooling1d_6 (MaxPooling1 (None, 22, 16)            0         \n",
      "_________________________________________________________________\n",
      "flatten_2 (Flatten)          (None, 352)               0         \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 352)               0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 4)                 1412      \n",
      "=================================================================\n",
      "Total params: 3,648\n",
      "Trainable params: 3,648\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "from keras import layers\n",
    "from keras import models\n",
    "\n",
    "model = models.Sequential()\n",
    "\n",
    "model.add(layers.Conv1D(4, 12, activation='relu', input_shape=(num_samples, num_layers)))\n",
    "model.add(layers.MaxPooling1D(2))\n",
    "\n",
    "model.add(layers.Conv1D(8, 12, activation='relu'))\n",
    "model.add(layers.MaxPooling1D(2))\n",
    "\n",
    "model.add(layers.Conv1D(16, 12, activation='relu'))\n",
    "model.add(layers.MaxPooling1D(2))\n",
    "\n",
    "model.add(layers.Flatten())\n",
    "model.add(layers.Dropout(0.5))\n",
    "model.add(layers.Dense(num_classes, activation='softmax'))\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/svg+xml": [
       "<svg height=\"702pt\" viewBox=\"0.00 0.00 263.00 702.00\" width=\"263pt\" xmlns=\"http://www.w3.org/2000/svg\" xmlns:xlink=\"http://www.w3.org/1999/xlink\">\n",
       "<g class=\"graph\" id=\"graph0\" transform=\"scale(1 1) rotate(0) translate(4 698)\">\n",
       "<title>G</title>\n",
       "<polygon fill=\"white\" points=\"-4,4 -4,-698 259,-698 259,4 -4,4\" stroke=\"none\"/>\n",
       "<!-- 1599006864 -->\n",
       "<g class=\"node\" id=\"node1\"><title>1599006864</title>\n",
       "<polygon fill=\"none\" points=\"52,-584.5 52,-620.5 203,-620.5 203,-584.5 52,-584.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"127.5\" y=\"-598.8\">conv1d_4: Conv1D</text>\n",
       "</g>\n",
       "<!-- 1599007824 -->\n",
       "<g class=\"node\" id=\"node2\"><title>1599007824</title>\n",
       "<polygon fill=\"none\" points=\"0,-511.5 0,-547.5 255,-547.5 255,-511.5 0,-511.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"127.5\" y=\"-525.8\">max_pooling1d_4: MaxPooling1D</text>\n",
       "</g>\n",
       "<!-- 1599006864&#45;&gt;1599007824 -->\n",
       "<g class=\"edge\" id=\"edge2\"><title>1599006864-&gt;1599007824</title>\n",
       "<path d=\"M127.5,-584.313C127.5,-576.289 127.5,-566.547 127.5,-557.569\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"131,-557.529 127.5,-547.529 124,-557.529 131,-557.529\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 1676054032 -->\n",
       "<g class=\"node\" id=\"node3\"><title>1676054032</title>\n",
       "<polygon fill=\"none\" points=\"52,-438.5 52,-474.5 203,-474.5 203,-438.5 52,-438.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"127.5\" y=\"-452.8\">conv1d_5: Conv1D</text>\n",
       "</g>\n",
       "<!-- 1599007824&#45;&gt;1676054032 -->\n",
       "<g class=\"edge\" id=\"edge3\"><title>1599007824-&gt;1676054032</title>\n",
       "<path d=\"M127.5,-511.313C127.5,-503.289 127.5,-493.547 127.5,-484.569\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"131,-484.529 127.5,-474.529 124,-484.529 131,-484.529\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 1600014064 -->\n",
       "<g class=\"node\" id=\"node4\"><title>1600014064</title>\n",
       "<polygon fill=\"none\" points=\"0,-365.5 0,-401.5 255,-401.5 255,-365.5 0,-365.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"127.5\" y=\"-379.8\">max_pooling1d_5: MaxPooling1D</text>\n",
       "</g>\n",
       "<!-- 1676054032&#45;&gt;1600014064 -->\n",
       "<g class=\"edge\" id=\"edge4\"><title>1676054032-&gt;1600014064</title>\n",
       "<path d=\"M127.5,-438.313C127.5,-430.289 127.5,-420.547 127.5,-411.569\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"131,-411.529 127.5,-401.529 124,-411.529 131,-411.529\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 1599083888 -->\n",
       "<g class=\"node\" id=\"node5\"><title>1599083888</title>\n",
       "<polygon fill=\"none\" points=\"52,-292.5 52,-328.5 203,-328.5 203,-292.5 52,-292.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"127.5\" y=\"-306.8\">conv1d_6: Conv1D</text>\n",
       "</g>\n",
       "<!-- 1600014064&#45;&gt;1599083888 -->\n",
       "<g class=\"edge\" id=\"edge5\"><title>1600014064-&gt;1599083888</title>\n",
       "<path d=\"M127.5,-365.313C127.5,-357.289 127.5,-347.547 127.5,-338.569\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"131,-338.529 127.5,-328.529 124,-338.529 131,-338.529\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 1599083920 -->\n",
       "<g class=\"node\" id=\"node6\"><title>1599083920</title>\n",
       "<polygon fill=\"none\" points=\"0,-219.5 0,-255.5 255,-255.5 255,-219.5 0,-219.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"127.5\" y=\"-233.8\">max_pooling1d_6: MaxPooling1D</text>\n",
       "</g>\n",
       "<!-- 1599083888&#45;&gt;1599083920 -->\n",
       "<g class=\"edge\" id=\"edge6\"><title>1599083888-&gt;1599083920</title>\n",
       "<path d=\"M127.5,-292.313C127.5,-284.289 127.5,-274.547 127.5,-265.569\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"131,-265.529 127.5,-255.529 124,-265.529 131,-265.529\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 1816345968 -->\n",
       "<g class=\"node\" id=\"node7\"><title>1816345968</title>\n",
       "<polygon fill=\"none\" points=\"57,-146.5 57,-182.5 198,-182.5 198,-146.5 57,-146.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"127.5\" y=\"-160.8\">flatten_2: Flatten</text>\n",
       "</g>\n",
       "<!-- 1599083920&#45;&gt;1816345968 -->\n",
       "<g class=\"edge\" id=\"edge7\"><title>1599083920-&gt;1816345968</title>\n",
       "<path d=\"M127.5,-219.313C127.5,-211.289 127.5,-201.547 127.5,-192.569\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"131,-192.529 127.5,-182.529 124,-192.529 131,-192.529\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 1816345456 -->\n",
       "<g class=\"node\" id=\"node8\"><title>1816345456</title>\n",
       "<polygon fill=\"none\" points=\"49,-73.5 49,-109.5 206,-109.5 206,-73.5 49,-73.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"127.5\" y=\"-87.8\">dropout_2: Dropout</text>\n",
       "</g>\n",
       "<!-- 1816345968&#45;&gt;1816345456 -->\n",
       "<g class=\"edge\" id=\"edge8\"><title>1816345968-&gt;1816345456</title>\n",
       "<path d=\"M127.5,-146.313C127.5,-138.289 127.5,-128.547 127.5,-119.569\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"131,-119.529 127.5,-109.529 124,-119.529 131,-119.529\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 1816345808 -->\n",
       "<g class=\"node\" id=\"node9\"><title>1816345808</title>\n",
       "<polygon fill=\"none\" points=\"63.5,-0.5 63.5,-36.5 191.5,-36.5 191.5,-0.5 63.5,-0.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"127.5\" y=\"-14.8\">dense_2: Dense</text>\n",
       "</g>\n",
       "<!-- 1816345456&#45;&gt;1816345808 -->\n",
       "<g class=\"edge\" id=\"edge9\"><title>1816345456-&gt;1816345808</title>\n",
       "<path d=\"M127.5,-73.3129C127.5,-65.2895 127.5,-55.5475 127.5,-46.5691\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"131,-46.5288 127.5,-36.5288 124,-46.5289 131,-46.5288\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 1817942416 -->\n",
       "<g class=\"node\" id=\"node10\"><title>1817942416</title>\n",
       "<polygon fill=\"none\" points=\"74,-657.5 74,-693.5 181,-693.5 181,-657.5 74,-657.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"127.5\" y=\"-671.8\">1817942416</text>\n",
       "</g>\n",
       "<!-- 1817942416&#45;&gt;1599006864 -->\n",
       "<g class=\"edge\" id=\"edge1\"><title>1817942416-&gt;1599006864</title>\n",
       "<path d=\"M127.5,-657.313C127.5,-649.289 127.5,-639.547 127.5,-630.569\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"131,-630.529 127.5,-620.529 124,-630.529 131,-630.529\" stroke=\"black\"/>\n",
       "</g>\n",
       "</g>\n",
       "</svg>"
      ],
      "text/plain": [
       "<IPython.core.display.SVG object>"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])\n",
    "\n",
    "from IPython.display import SVG\n",
    "from keras.utils.vis_utils import model_to_dot\n",
    "\n",
    "SVG(model_to_dot(model).create(prog='dot', format='svg'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 64 samples, validate on 32 samples\n",
      "Epoch 1/30\n",
      "64/64 [==============================] - 4s 55ms/step - loss: 3.8112 - acc: 0.3750 - val_loss: 1.8127 - val_acc: 0.2813\n",
      "Epoch 2/30\n",
      "64/64 [==============================] - 0s 6ms/step - loss: 2.5595 - acc: 0.2656 - val_loss: 1.2054 - val_acc: 0.4375\n",
      "Epoch 3/30\n",
      "64/64 [==============================] - 0s 7ms/step - loss: 2.0510 - acc: 0.4375 - val_loss: 1.1341 - val_acc: 0.4375\n",
      "Epoch 4/30\n",
      "64/64 [==============================] - 0s 5ms/step - loss: 1.2303 - acc: 0.5625 - val_loss: 0.8359 - val_acc: 0.5938\n",
      "Epoch 5/30\n",
      "64/64 [==============================] - 0s 6ms/step - loss: 1.1510 - acc: 0.6094 - val_loss: 0.7354 - val_acc: 0.6875\n",
      "Epoch 6/30\n",
      "64/64 [==============================] - 0s 6ms/step - loss: 0.9110 - acc: 0.6406 - val_loss: 0.6366 - val_acc: 0.6875\n",
      "Epoch 7/30\n",
      "64/64 [==============================] - 0s 6ms/step - loss: 0.8491 - acc: 0.6563 - val_loss: 0.5721 - val_acc: 0.7812\n",
      "Epoch 8/30\n",
      "64/64 [==============================] - 0s 6ms/step - loss: 0.8586 - acc: 0.6250 - val_loss: 0.5937 - val_acc: 0.8125\n",
      "Epoch 9/30\n",
      "64/64 [==============================] - 0s 6ms/step - loss: 0.9176 - acc: 0.7344 - val_loss: 0.5591 - val_acc: 0.8437\n",
      "Epoch 10/30\n",
      "64/64 [==============================] - 0s 5ms/step - loss: 0.5991 - acc: 0.7656 - val_loss: 0.6574 - val_acc: 0.8437\n",
      "Epoch 11/30\n",
      "64/64 [==============================] - 0s 7ms/step - loss: 0.3413 - acc: 0.9062 - val_loss: 0.6925 - val_acc: 0.8750\n",
      "Epoch 12/30\n",
      "64/64 [==============================] - 0s 5ms/step - loss: 0.5921 - acc: 0.8125 - val_loss: 0.5190 - val_acc: 0.8125\n",
      "Epoch 13/30\n",
      "64/64 [==============================] - 0s 5ms/step - loss: 0.2480 - acc: 0.9375 - val_loss: 0.5591 - val_acc: 0.8750\n",
      "Epoch 14/30\n",
      "64/64 [==============================] - 0s 6ms/step - loss: 0.1699 - acc: 0.9687 - val_loss: 0.5565 - val_acc: 0.8750\n",
      "Epoch 15/30\n",
      "64/64 [==============================] - 0s 5ms/step - loss: 0.4074 - acc: 0.8750 - val_loss: 0.4261 - val_acc: 0.8750\n",
      "Epoch 16/30\n",
      "64/64 [==============================] - 0s 6ms/step - loss: 0.3726 - acc: 0.9062 - val_loss: 0.4563 - val_acc: 0.8750\n",
      "Epoch 17/30\n",
      "64/64 [==============================] - 0s 6ms/step - loss: 0.3818 - acc: 0.8750 - val_loss: 0.2726 - val_acc: 0.9687\n",
      "Epoch 18/30\n",
      "64/64 [==============================] - 0s 6ms/step - loss: 0.0913 - acc: 0.9844 - val_loss: 0.3796 - val_acc: 0.8750\n",
      "Epoch 19/30\n",
      "64/64 [==============================] - 0s 6ms/step - loss: 0.1166 - acc: 0.9375 - val_loss: 0.3274 - val_acc: 0.8750\n",
      "Epoch 20/30\n",
      "64/64 [==============================] - 0s 6ms/step - loss: 0.0526 - acc: 1.0000 - val_loss: 0.3535 - val_acc: 0.8750\n",
      "Epoch 21/30\n",
      "64/64 [==============================] - 0s 6ms/step - loss: 0.2609 - acc: 0.9531 - val_loss: 0.3542 - val_acc: 0.8750\n",
      "Epoch 22/30\n",
      "64/64 [==============================] - 0s 6ms/step - loss: 0.2408 - acc: 0.9687 - val_loss: 0.2677 - val_acc: 0.9062\n",
      "Epoch 23/30\n",
      "64/64 [==============================] - 0s 6ms/step - loss: 0.0883 - acc: 0.9687 - val_loss: 0.3227 - val_acc: 0.8750\n",
      "Epoch 24/30\n",
      "64/64 [==============================] - 0s 5ms/step - loss: 0.0217 - acc: 1.0000 - val_loss: 0.2342 - val_acc: 0.9375\n",
      "Epoch 25/30\n",
      "64/64 [==============================] - 0s 5ms/step - loss: 0.0463 - acc: 0.9844 - val_loss: 0.2419 - val_acc: 0.9687\n",
      "Epoch 26/30\n",
      "64/64 [==============================] - 0s 6ms/step - loss: 0.1889 - acc: 0.9531 - val_loss: 0.3109 - val_acc: 0.9062\n",
      "Epoch 27/30\n",
      "64/64 [==============================] - 0s 5ms/step - loss: 0.0757 - acc: 0.9687 - val_loss: 0.2062 - val_acc: 0.9687\n",
      "Epoch 28/30\n",
      "64/64 [==============================] - 0s 5ms/step - loss: 0.0385 - acc: 0.9844 - val_loss: 0.4542 - val_acc: 0.8750\n",
      "Epoch 29/30\n",
      "64/64 [==============================] - 0s 5ms/step - loss: 0.0351 - acc: 0.9844 - val_loss: 0.4252 - val_acc: 0.8750\n",
      "Epoch 30/30\n",
      "64/64 [==============================] - 0s 6ms/step - loss: 0.0292 - acc: 1.0000 - val_loss: 0.2277 - val_acc: 0.9375\n"
     ]
    }
   ],
   "source": [
    "EPOCHS = 30\n",
    "history = model.fit(train_data, train_labels, epochs=EPOCHS, validation_data=(test_data, test_labels), batch_size=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x5f1e1950>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAIABJREFUeJzt3Xd81fX1+PHXyc1ehISEAGGDQpBpBDe4vm5xoLhHVap26Ldqq/21dbT91trlHlSxaK2IAoqtFvdWIOwtQ0YgkE0Cmffe8/vjcwkxZufe3NzkPB+P+8i9n/u573s+XrnnvreoKsYYYwxAWLADMMYY03lYUjDGGFPLkoIxxphalhSMMcbUsqRgjDGmliUFY4wxtSwpGNMCIjJIRFREwltw7vUi8nl7yzEmGCwpmC5HRLaLSLWI9Kp3fKXvC3lQcCIzpvOzpGC6qm+BKw49EJHRQEzwwjEmNFhSMF3VS8C1dR5fB7xY9wQR6SEiL4pIvojsEJFfiUiY7zmXiPxZRApEZBtwbgOvfV5EckVkt4j8TkRcrQ1SRPqKyEIRKRKRLSJyc53nJopItoiUisg+Efmr73i0iPxTRApFpERElopI79a+tzENsaRguqqvgUQRGen7sp4O/LPeOY8DPYAhwGScJHKD77mbgfOA8UAWMK3ea2cDbmCY75z/AW5qQ5yvADlAX997/J+InOZ77lHgUVVNBIYCc33Hr/PF3R9IAW4BKtrw3sZ8jyUF05Udqi2cAWwEdh96ok6iuFdVy1R1O/AX4BrfKZcBj6jqLlUtAv5Q57W9gbOBO1T1oKrmAX8DLm9NcCLSHzgR+IWqVqrqSuC5OjHUAMNEpJeqHlDVr+scTwGGqapHVZepamlr3tuYxlhSMF3ZS8CVwPXUazoCegGRwI46x3YA/Xz3+wK76j13yEAgAsj1Nd+UAM8Caa2Mry9QpKpljcRwI3AEsNHXRHRenetaBMwRkT0i8rCIRLTyvY1pkCUF02Wp6g6cDudzgPn1ni7A+cU9sM6xARyuTeTiNM/Ufe6QXUAV0EtVk3y3RFUd1coQ9wDJIpLQUAyqullVr8BJNn8EXheROFWtUdUHVDUTOB6nmetajPEDSwqmq7sROFVVD9Y9qKoenDb634tIgogMBH7G4X6HucBPRSRDRHoC99R5bS7wLvAXEUkUkTARGSoik1sTmKruAr4E/uDrPB7ji/dlABG5WkRSVdULlPhe5hGRU0RktK8JrBQnuXla897GNMaSgunSVHWrqmY38vRPgIPANuBz4F/ALN9zf8dpolkFLOf7NY1rcZqf1gPFwOtAnzaEeAUwCKfWsAC4T1Xf8z13FrBORA7gdDpfrqqVQLrv/UqBDcAnfL8T3Zg2EdtkxxhjzCFWUzDGGFMr4EnBNwlohYj8u4HnokTkVd+kncW2/IAxxgRXR9QUbsdp92zIjUCxqg7DGef9xw6IxxhjTCMCmhREJANneYDnGjllKs7MUHA6zk4TEQlkTMYYYxoX6OV7HwF+DiQ08nw/fBOEVNUtIvtxZmoW1D1JRGYAMwDi4uKOHjFiRMACNsaYrmjZsmUFqpra3HkBSwq+2Zd5qrpMRKY0dloDx743HEpVZwIzAbKysjQ7u7ERhsYYYxoiIjuaPyuwzUcnABeIyHZgDnCqiNQfS52Db9aob9ORHkBRAGMyxhjThIAlBVW9V1UzVHUQzkJhH6rq1fVOW4iz4iM4K0R+qDZxwhhjgqbDtwQUkQeBbFVdCDwPvCQiW3BqCK1aZdIYY4x/dUhSUNWPgY99939T53glcGl7y6+pqSEnJ4fKysr2FhUyoqOjycjIICLCFsc0xvhPl9g8PCcnh4SEBAYNGkR3GNGqqhQWFpKTk8PgwYODHY4xpgvpEstcVFZWkpKS0i0SAoCIkJKS0q1qRsaYjtElkgLQbRLCId3teo0xHaPLJIXmVNR4yN1fgcfrDXYoxhjTaXWbpFDj9pJfVkVljf+TQmFhIePGjWPcuHGkp6fTr1+/2sfV1dUtKuOGG25g06ZNfo/NGGNao0t0NLdEVLiT/6rcXuKi/Ft2SkoKK1euBOD+++8nPj6eu+666zvnqCqqSlhYw3n4hRde8G9QxhjTBt2mphAZHoaIUOXuuF0Lt2zZQmZmJldddRWjRo0iNzeXGTNmkJWVxahRo3jwwQdrzz3xxBNZuXIlbrebpKQk7rnnHsaOHctxxx1HXl5eh8VsjOneulxN4YG31rF+T2mDz1XUeBAgOsLVqjIz+yZy3/mt3ZPdsXHjRmbPns0xxxwDwEMPPURycjJut5tTTjmFadOmkZmZ+Z3X7N+/n8mTJ/PQQw/xs5/9jFmzZnHPPfc0VLwxxvhVt6kpAIQJeDt4EY2hQ4fWJgSAV155hQkTJjBhwgQ2bNjA+vXrv/eamJgYzj77bACOPvpotm/f3lHhGmO6uS5XU2jqF/2+0kr2lVZyVN8ehIV1zJDOuLi42vubN2/m0UcfZcmSJSQlJXH11Vc3ONcgMjKy9r7L5cLtdndIrMYY061qCnU7m4OhtLSUhIQEEhMTyc3NZdGiRUGJwxhjGtPlagpNiQp3+hKq3B5iIlvXr+APEyZMIDMzkxEjRjBw4EBOOOGEDo/BGGOaIqG2UnVDm+xs2LCBkSNHNvtar1dZt2c/qYnRpCdGByrEDtPS6zbGGBFZpqpZzZ3XrZqPwsKEyPAwqmo6bliqMcaEkm6VFMBpQgpWn4IxxnR23S8pRIRR5fYSas1mxhjTEbpfUgh3oapUW23BGGO+J2BJQUSiRWSJiKwSkXUi8kAD51wvIvkistJ3uylQ8RwS7RuWWmlJwRhjvieQQ1KrgFNV9YCIRACfi8g7qvp1vfNeVdUfBzCO74iKODRXwQPYVpbGGFNXwJKCOo32B3wPI3y3oDfku8LCiHCFUeXHJbQLCws57bTTANi7dy8ul4vU1FQAlixZ8p0Zyk2ZNWsW55xzDunp6X6LzRhjWiOgk9dExAUsA4YBT6rq4gZOu0RETga+Af5XVXc1UM4MYAbAgAED2h1XVHiYX1dLbcnS2S0xa9YsJkyYYEnBGBM0Ae1oVlWPqo4DMoCJInJUvVPeAgap6hjgfWB2I+XMVNUsVc069Au8PaIjXFTWdMwIpNmzZzNx4kTGjRvHbbfdhtfrxe12c8011zB69GiOOuooHnvsMV599VVWrlzJ9OnTW7U5jzHG+FOHLHOhqiUi8jFwFrC2zvHCOqf9Hfhju9/snXtg75omT0n1eEl0e9FIV8v2Ok4fDWc/1OpQ1q5dy4IFC/jyyy8JDw9nxowZzJkzh6FDh1JQUMCaNU6cJSUlJCUl8fjjj/PEE08wbty4Vr+XMcb4QyBHH6WKSJLvfgxwOrCx3jl96jy8ANgQqHjqCvMlAm+Aawrvv/8+S5cuJSsri3HjxvHJJ5+wdetWhg0bxqZNm7j99ttZtGgRPXr0CGgcxhjTUoGsKfQBZvv6FcKAuar6bxF5EMhW1YXAT0XkAsANFAHXt/tdW/CL3uvxsi23lL49YuiV4Oe9OetQVX7wgx/w29/+9nvPrV69mnfeeYcnn3ySefPmMXPmzIDFYYwxLRXI0UergfENHP9Nnfv3AvcGKobGhIcJrjChMsBbc55++ulMmzaN22+/nV69elFYWMjBgweJiYkhOjqaSy+9lOHDh3PTTc70jISEBMrKygIakzHGNKVbLZ19iIh0yBpIo0eP5r777uP000/H6/USERHBM888g8vl4sYbb0RVERH++EenK+WGG27gpptuIiYmplVDWY0xxl+61dLZdeUUl1Na4Sazb6I/w+tQtnS2MaalbOnsZkSFu3B7vbg9ttyFMcYc0n2TQkRwt+Y0xpjOqMskhdY2g9UujBeiG+6EWrOfMSY0dImkEB0dTWFhYau+KCNcYYSJhGRNQVUpLCwkOjr0txQ1xnQuXWL0UUZGBjk5OeTn57fqdYWllRSHCfvjAzdXIVCio6PJyMgIdhjGmC6mSySFiIgIBg8e3OrXPTNnBdnbi/jinlMDEJUxxoSeLtF81FbD0+LZXVLBwSp3sEMxxphOoVsnhWFp8QBszT/QzJnGGNM9WFIAtuRZUjDGGOjmSWFgShzhYWJJwRhjfLp1UohwhTEwJdaSgjHG+HTrpABOE9IW61MwxhjAkgLD0uLZUVhOdQhOYjPGGH/r9klheFoCHq+yo/BgsEMxxpig6/ZJ4dAIpM3Wr2CMMQHdozlaRJaIyCoRWSciDzRwTpSIvCoiW0RksYgMClQ8jRmSGgfYsFRjjIHA1hSqgFNVdSwwDjhLRI6td86NQLGqDgP+BvwxgPE0KDYynH5JMZYUjDGGACYFdRz6po3w3eovYzoVmO27/zpwmohIoGJqzLC0eEsKxhhDgPsURMQlIiuBPOA9VV1c75R+wC4AVXUD+4GUQMbUkGFp8WwrOIDXa3sUGGO6t4AmBVX1qOo4IAOYKCJH1TuloVrB976ZRWSGiGSLSHZrl8duiWFp8VTWeNldUuH3so0xJpR0yOgjVS0BPgbOqvdUDtAfQETCgR5AUQOvn6mqWaqalZqa6vf4hteOQCrze9nGGBNKAjn6KFVEknz3Y4DTgY31TlsIXOe7Pw34UIOwz6QtjGeMMY5AbrLTB5gtIi6c5DNXVf8tIg8C2aq6EHgeeElEtuDUEC4PYDyNSoqNpFd8pCUFY0y3F7CkoKqrgfENHP9NnfuVwKWBiqE1hqbaCCRjjOn2M5oPOTQsNQitV8YY02lYUvAZlhZPaaWb/ANVwQ7FGGOCxpKCj3U2G2OMJYVaw9MSAEsKxpjuzZKCT+/EKOKjwi0pGGO6NUsKPiLCUFsDyRjTzVlSqGOYDUs1xnRzlhTqGJYWT15ZFaWVNcEOxRhjgsKSQh02AskY091ZUqjDkoIxpruzpFBH/54xRIaHWVIwxnRblhTqCHeFMaRXnCUFY0y3ZUmhHhuWaozpziwp1DMsNZ5dxeVU1niCHYoxxnQ4Swr1DEuLRxW25R8MdijGGNPhLCnUUzsCKd+akIwx3Y8lhXoG94ojTGDLPtuv2RjT/QRyj+b+IvKRiGwQkXUicnsD50wRkf0istJ3+01DZXWk6AgXg1LiWJmzP9ihGGNMhwvkHs1u4E5VXS4iCcAyEXlPVdfXO+8zVT0vgHG02jmj+/DUx1vYU1JB36SYYIdjjDEdJmA1BVXNVdXlvvtlwAagX6Dez5+mH9Mfr8Lc7F3BDsUYYzpUh/QpiMggYDywuIGnjxORVSLyjoiM6oh4mtM/OZaThvdi7tJdeLy2Z7MxpvsIeFIQkXhgHnCHqpbWe3o5MFBVxwKPA280UsYMEckWkez8/PzABuxzxcQB7NlfyaffdMz7GWNMZxDQpCAiETgJ4WVVnV//eVUtVdUDvvtvAxEi0quB82aqapaqZqWmpgYy5Fqnj+xNSlwkryzZ2SHvZ4wxnUEgRx8J8DywQVX/2sg56b7zEJGJvngKAxVTa0SGhzHt6Aw+2JhHXmllsMMxxpgOEciawgnANcCpdYacniMit4jILb5zpgFrRWQV8Bhwuap2mkb86cf0x+NVXluWE+xQjDGmQwRsSKqqfg5IM+c8ATwRqBjaa0hqPMcOSWbO0p3cOnkoYWFNXo4xxoQ8m9HcjCsmDmBXUQVfbu0UrVrGGBNQlhSaceaodJJiI3hlqXU4G2O6PksKzYiOcHHx+AzeXbeXwgNVwQ7HGGMCypJCC1wxsT81HmXecutwNsZ0bZYUWmB47wSOHtiTOUt20YkGRxljjN9ZUmihKyYOYFvBQRZ/WxTsUIwxJmAsKbTQuaP7kBAdzhyb4WyM6cIsKbRQTKSLC8f14+21eykprw52OMYYExCWFFrhiokDqHZ7WbBid7BDMcaYgLCk0AqZfRMZm9GDV5bstA5nY0yXZEmhlS6fOIBv9h1g+c6SYIdijDF+Z0mhlc4f25fYSJd1OBtjuiRLCq0UHxXO1HF9eWv1Hkora4IdjjHG+JUlhTa4/JgBVNZ4eXPlnmCHYowxfmVJoQ3GZPRgZJ9Ea0IyxnQ5lhTaQES4cmJ/1u0pZU3O/mCHY4wxfmNJoY2mju9HdEQY/7LagjGmC2lRUhCRoSIS5bs/RUR+KiJJgQ2tc0uMjuDc0X1ZuHI3B6vcwQ7HGGP8oqU1hXmAR0SGATOB/sC/mnqBiPQXkY9EZIOIrBOR2xs4R0TkMRHZIiKrRWRCq68giC4a34+D1R6W2CJ5xpguoqVJwauqbuAi4HFVvRvo08xr3MCdqjoSOBb4kYhk1jvnbGC47zYDeLrFkXcCRw/sSYRLbOVUY0yX0dKkUCMiVwDXAf/2HYto6gWqmquqy333y4ANQL96p00FXlTH10CSiDSXbDqNmEgXYzKSWPyt7d9sjOkaWpoUbgCOA36vqt+KyGDgny19ExEZBIwHFtd7qh+wq87jHL6fOBCRGSKSLSLZ+fn5LX3bDjFpcDJrcvZTXm39CsaY0NeipKCq61X1p6r6ioj0BBJU9aGWvFZE4nH6JO5Q1dL6Tzf0dg28/0xVzVLVrNTU1Ja8bYeZODgZt1dZtqM42KEYY0y7tXT00ccikigiycBy4O8i8tcWvC4CJyG8rKrzGzglB6fT+pAMIKSmCWcNSsYVJtbZbIzpElrafNTD9yv/Ypw+gEnA6U29QEQEeB7YoKqNJZCFwLW+UUjHAvtVNbeFMXUK8VHhHNU3kcXbLCkYY0JfeEvP83UAXwb8vxa+5gTgGmCNiKz0HfslMABAVZ8B3gbOAbYA5Th9FyFn0pAU/vHFdiprPERHuIIdjjHGtFlLk8KDwCLgC1VdKiJDgM1NvUBVP6fhPoO65yjwoxbG0GlNHJTMzE+3sWJnCccNTQl2OMYY02Yt7Wh+TVXHqOqtvsfbVPWSwIYWOo4ZnIwI1q9gjAl5Le1ozhCRBSKSJyL7RGSeiGQEOrhQ0SMmgpHpiTZfwRgT8lra0fwCTqdwX5x5BG/5joWO6nLY+B8I0N7Kk4Yks3xnMdVub0DKN8aYjtDSpJCqqi+oqtt3+wfQuSYMNGfdAphzJexeHpDiJw1OobLGy+oc27vZGBO6WpoUCkTkahFx+W5XA6HVVjLiHAiLgPULAlL8xMHJALYOkjEmpLU0KfwAZzjqXiAXmEaoDR+N6QlDpsD6NwPShJQcF8kRveMtKRhjQlpLRx/tVNULVDVVVdNU9UKciWyhZdSFULIT9qwISPGTBqewbHsRbo/1KxhjQlN7dl77md+i6ChHngNh4bD+jYAUP2lIMgerPazdU3+JJ2OMCQ3tSQpNTkzrlGKTYfBkWPdGQJqQDvUrLLGhqcaYENWepBCYsZ2BNupCKNkBuSubP7eV0hKiGdIrztZBMsaErCaTgoiUiUhpA7cynDkLoWfEeSAup7YQAJOGJLNkexEeb2jmTGNM99ZkUlDVBFVNbOCWoKotXTepc4lNhiGTAzYKadLgFMoq3WzItX4FY0zoaU/zUejKvBCKv4W9q/1etM1XMMaEsu6ZFALYhNQ3KYb+yTHW2WyMCUndMynEpcDgk5yhqQFqQlrybRFe61cwxoSY7pkUwGlCKtoGe9f4vehJg5MpLq9hc94Bv5dtjDGB1H2TwsjznSakAExkO3aIs9GOLaVtjAk1AUsKIjLLt//C2kaenyIi+0Vkpe/2m0DF0qC4XjDoxIBMZMvoGUOfHtHW2WyMCTmBrCn8AzirmXM+U9VxvtuDAYylYaMuhKKtsG+dX4sVESYNTmbxtiI0QPs3GGNMIAQsKajqp0Dn/qk84nyQsIA0IU0akkLBgSq2FRz0e9nGGBMowe5TOE5EVonIOyIyqrGTRGSGiGSLSHZ+fr7/3j0+FQaeEJAmpEmH5ivYkhfGmBASzKSwHBioqmOBx4FGf66r6kxVzVLVrNRUP2/4NupCKNwMeev9WuzgXnH0io+y+QrGmJAStKSgqqWqesB3/20gQkR6dXggIy9wmpD8PJFNRJg0JJnF31q/gjEmdAQtKYhIuoiI7/5EXywd/7M6Ps1pQlr/pt+LPnZwMrn7K9lVVOH3so0xJhACOST1FeAr4EgRyRGRG0XkFhG5xXfKNGCtiKwCHgMu12D9pM6cCgWbIG+DX4ud5Juv8LU1IRljQkTAVjpV1Suaef4J4IlAvX+rjLwA3r7baUJKG+m3YoelxtMzNoIl3xZxWVZ/v5VrjDGBEuzRR51DQm8YeLzfh6aGhQkTByfbzGZjTMiwpHBI5oWQvxHyNvq12EmDU9hVVMGeEutXMMZ0fpYUDsm8ABC/1xYmDTm0v4LVFowxnZ8lhUMS0mHAcX4fhTQiPZGE6HCW2DpIxpgQYEmhrlEXOpPY8r/xW5GuMGHioGSb2WyMCQmWFOoaGbgmpG0FB8krrfRrucYY42+WFOpK7AMDjvX77OZJgw/tr2C1BWNM52ZJob7MqZC3Dgo2+63IUX0TiYt08eynW3l33V7cHq/fyjbGGH+ypFDfyAucv58/Agf8syJruCuM+84fRV5pFTNeWsbxD33InxZtZGdhuV/KN8YYf5FQW6wtKytLs7OzA/sm82fA6led7TqHngqjL4UR50JUfLuKdXu8fLgxjzlLd/Hxpjy8CicO68XlE/tzRmZvosJdfroAY4z5LhFZpqpZzZ5nSaER+9bDmrmw5nXYvwsiYuHIc2DMZU6icEW0q/jc/RXMXZrD3Oxd7C6pIDkukksm9GP6MQMYlta+5GOMMfVZUvAXrxd2fQ1rXoN1C6CiGGKSYdRFToLoPwmcxV7bxONVPt9SwJwlO3lv/T7cXuW4ISk8ceV4UuKj/HghxpjuzJJCILirYesHsHoubHoH3BWQMgyufRN6ZLS7+PyyKl5flsMj73/D6H49ePnmSdakZIzxi5YmBetobo3wSDjybLj0Bbh7M1z4NJTmwsKf+GU7z9SEKG6dMpQ/XzqW7B3F3Dt/jW3QY4zpUJYU2ioqAcZdCWc8AFs/hGUv+K3o88f25Y7ThzN/+W6e+WSb38o1xpjmWFJor6wbYfBkWPQrKN7ut2JvP20454/ty8OLNrJo3V6/lWuMMU2xpNBeYWEw9Qlnn+c3f+x0TPuBiPCnaWMYk5HEHXNWsnb3fr+Ua4wxTQnkdpyzRCRPRNY28ryIyGMiskVEVovIhEDFEnBJA+DM38P2z2Dp3/1WbHSEi79fczRJsRHc/GK2rZ1kjAm4QNYU/gGc1cTzZwPDfbcZwNMBjCXwJlwLw86A9+6Dwq1+KzYtMZrnrsuipLyGm19aRmWNx29lG2NMfQFLCqr6KdDUCnBTgRfV8TWQJCJ9AhVPwInABY85I5TeuA28/vvyHtW3B49cPo7VOSXc9doqG5FkjAmYYPYp9AN21Xmc4zv2PSIyQ0SyRSQ7P98/6xEFRGJfOPthZ7Lb1/6t+Jw5Kp2fnzmCf6/O5dEP/LdYnzHG1BXMpNDQNOAGfwKr6kxVzVLVrNTU1ACH1U5jpjvLYXzwoF836wG4ZfIQLp7Qj0fe38xbq/b4tWxjjIHgJoUcoH+dxxlA6H/TicB5j0BkLLxxC3jcfixa+MPFozlmUE/uem0VK3eV+K1sY4yB4CaFhcC1vlFIxwL7VTU3iPH4T0JvOPcvsHsZfPmYX4uOCnfxzNVHk5oQxU2zs4Oy/Pb+8hoWrtrDkx9tocb2hjCmSwkPVMEi8gowBeglIjnAfUAEgKo+A7wNnANsAcqBGwIVS1CMuhjWvwkf/wGOOAt6Z/qt6JT4KGZdfwwXP/Ulk//8ERMG9OSMzN6ckdmboan+X2FVVflm3wE+3JjHRxvzWLazGI/XaekTgdumDPP7expjgsMWxAukgwXw5CSnA/rmD9u93HZ93xYc5I0Vu3lv/T7W55YCMCQ1zkkQI3szfkBPXGFtW8G1otrDV9sKfIkgn90lFQBk9knk1BFpnDIijWc/2con3+Tz7v+ezMCUOL9dlzHG/2yV1M5iw1vw6tUw5Zcw5RcBe5vdJRW8v34f763fx9fbCnF7lZS4SE4bmcYZmekcPzQFBcqr3Bys9nCwys3BKjfl1R4OVrspr3L+Hqh0s3xnMV9uLaTK7SU20sUJw3o5ieDINNJ7RNe+5979lZz+108Y1z+Jl26ciLRjCXFjTGBZUuhM5t0M6+bDje9Bv8BP3C6trOHjTfm8t34fH2/Mo6yqdZ3dA1NiOeXINE4dkcakIclNLt/90lfb+fWb6/jrZWO5eEL7lw83xgSGJYXOpLwInj7eaU4afzWc9DNnaYwOUO32suTbIlbsLCYqIozYyHDiolzO38hwYqNcxEeFExvpqn3cmj0cvF5l2jNf8m3BQT64cwrJcZEBvBpjTFtZUuhsSvfAZ3+B5S86ey+MvwpOurPDkkMgbdpbxrmPfcbUcf34y2Vjgx2OMaYBtslOZ5PY1xmm+tOVcPT1sPJf8NgEeOt2KNkZ7Oja5cj0BH44eQjzlufwxZaCYIdjjGkHSwodrUc/OPfPXS45/OTU4QxKieWXC9bYon3GhDBLCsHSYHIYH7LJITrCxf9dNJodheU8/qGtzWRMqLKkEGzfSQ43HE4Ob/4YikJrK87jh/XikgkZPPvJNjbuLQ12OMaYNrCk0FnUTQ5ZP4DVc+HxLFhwKxRsCXZ0Lfb/zh1JYkwE985fg9cbWoMYjDGWFDqfHv3gnD/BHath0i2wbgE8eQzMuwnyNgY7umYlx0Xy6/NGsmJnCS8v3tH0yR63c01r58OHv4NXroQXL4T8TR0TrDHme2xIamd3IA++fByWPg815TDqQjj5bug9KtiRNUpVuXbWElbsLOH9n00mPTEK9udA3gbIWwf71jv3CzaBp9p5kbggZSiUF4LXDdP/CYNPDu6FGNOF2DyFruZgIXz9JCyeCdVlMOI8mPxz6NM55wXsLCznvEfe4xe9l3FVzQLYX6fzPDED0kY6iwSm+W69joCIaCjeAS9f6vSnXPAYjLsyeBdhTBdiSaGrKi+Cxc86O7tV7YcBx0PmBU6SSOrf/Os7QnU5LJ/NgQ//Qnx1PsUp4+l57NWQNspJBjFJTb++ogTmXgPffgqTfwFT7nWWYzXGtJklha6ucj8sfQ7WzHOaZAD6ToCR58PIC6BXEJazrjqVzYnfAAAa2UlEQVQA2c87zV0H8/EOPIFfFpzNx9Ujee/OySREt2KVWHc1/PsOWPkyjL4Mpj4B4VGBi92YLs6SQndSuBU2LHRWZN29zDmWOtKXIM6H9NHf/6Xt9cDBfCjd7SzBUZrr3C/LhahE5xd9WiakjYCYnk2/f+V+WDITvnoKKopgyClO09bA41m5q4SLnvqC6Vn9eeiSMa27LlX47M9OJ/TAE5x+hthkAF5flsOwtHjG9W+m1mGMASwpdF/7c2Djf5wEseMLUC8kDXQ6batKDyeAslzQejOPwyIgId1pvqkuO3w8oa8vSRxKFCMh9UhwV8HiZ5xb5X4YfqbTCd7/mO8U+/B/N/LUx1t5eNoYLstqQxPXmtfhjVuddaKueo3XtkVw9+urSU+M5oM7JxMXFbC9oozpMiwpGGdV1k1vOwkiZynEpUJCH0jsB4l9nPWYEvo6fxP7QmwvCAtzfqHXjhZaf/hvwTfgrvQVLuCKBE+V059x8t3Qd1yDYbg9Xq57YQnZ24uZd+vxHNWvR+uvZcdXMOcK3CpcdeB29veawMa9ZfzolKHcfeaIw+epOjWePSshd6Xzt3CLU1MSF4SFQ5gLJKzO/TrHh/+PMxTYZYnGdC2dIimIyFnAo4ALeE5VH6r3/PXAn4DdvkNPqOpzTZVpSSGIvB4o3n44UZQXwYRrWjQ8tvBAFec9/jnhLuGtH59IUmzrl9gu2rmBg7MuIo0Cas5/il9vHs7yNWt5bWoMaWUbDieBct+ifOI6XKuRMGeoq9fj3NRz+LH6jlWWwr410Gec04eRPrrVMRrTWQU9KYiIC/gGOAPIAZYCV6jq+jrnXA9kqeqPW1quJYXQtWJnMZc9+xUnDOvFrOuOIawVW4VWu71c/dxidu7eyQd9ZxK3LxtvTAphFYXOCYcSQJ9xTo2l73gnWUXEtDxAVVj/Brx9N1QUw4n/69SArIPbdAEtTQqBrCNPBLao6jZfQHOAqcD6Jl9luqzxA3py3/mj+NUba3nsw83ccfoRLX7t7/6zniXbi3j08hOJGzUVPvo9YRVFfHKgH39bG8ed11zMSZntHJIrAqMugsGTYdEv4dM/wfo34YInYMCk9pVtTIgI5DIX/YBddR7n+I7Vd4mIrBaR10WkwX/VIjJDRLJFJDs/Pz8QsZoOctWkAVwyIYNHP9jMRxvzWvSauUt38eJXO5hx8hCmjuvnTHL7n9/C1Cc5dvovKE4ew/3vbKXG4/VPkLHJcNEzcNU8qKmAWWfC2z93htwa08UFMik01DZQv63qLWCQqo4B3gdmN1SQqs5U1SxVzUpNTfVzmKYjiQi/v+goRqYncserK9lVVN7k+St2FvOrN9Zy4rBe/PzMI7/3fFS4i9+cl8nW/IPM/nK7f4Mdfjrc9hVMvNkZcvvUcbDl/eZf565y1nTa8G9nzsbetf6Ny3R+u5Y4owDzNjr/P4SQQPYpHAfcr6pn+h7fC6Cqf2jkfBdQpKpNDk2xPoWuYWdhOec9/hn9k2OZd+vxREd8f1/ovLJKzn/8cyLDw1j4oxPp2cj+z6rK9S8sZfmOYj68awqpCQHoA9j5tbOceeFmGHslnPl7pxZRuNkZ3VSwxflbuNnZD0Pr1FoiYuGyl5wkY7q+jf+BOXWXZxHo0R9ShkDyUGeNr0N/kwZCeMfsa94ZOprDcTqaT8MZXbQUuFJV19U5p4+q5vruXwT8QlWPbapcSwpdx0cb87jhH0uZdnQGf5o2Bqkzwa7a7eXKv3/Nuj2lzL/teEb2SWyyrK35Bzjzb59y8YR+PDwtQOtB1VQ6/Qyf/+37czwiYp1/5CnDoddwSBnm3KJ7wGvXOyO2LnwaxlwWmNhM55C3EZ47zfl/4OyHndF6hVuhaOvhv5X7D58vLmf9sstmB3y/9qB3NKuqW0R+DCzCGZI6S1XXiciDQLaqLgR+KiIXAG6gCLg+UPGYzueUEWn89LThPPbBZiYM6MmVkw7/o3jgrXVk7yjmiSvHN5sQAIamxvODEwfz98+2cdWkgYwNxEzniGg47dd8EHYsuz9/hcrYdMJSj6BHxkgyBgxlRJ/Ehmsz1/t+Oc6/2VkF9thb/R+bCb6KYphzhfMDYfrLzjL4/Sd+9xxVZyj3oSRRuAWW/h1eOBeufwt6DgpK6HXZ5DUTVB6v8oN/LOWrrYXMveU4xvVP4pUlO7l3/hpumTyUe84e0XwhPmWVNZzy50/onxzDvFuOb9WQ15b6cmsB1z6/hCN6JxAfHc6mvWXsr6ipfT41IYojeydwRO8ERqQncER6AjERLirKD5Dx0e302rWITcNvZtnQn1Be4+FglYfyGjflVR6S4yL50SnDiAxvoqvPXQVr5zmT7Qad5ExCNMHnccO/LoVvP4Pr/w0Dmmzw+K49K+HFqRAZD9ctdGqcARD05qNAsaTQ9ZSUV3Pe45/j9SoPTj2KW19exnFDe/HC9cfgauUX+2vZu7j79dX85dKxXHJ0hl/j3JZ/gIue+pLUhCjm3Xo8PWIiUFXyyqrYtLfMue0r4xvfrbLmu6OhwvDyu/BZXBn+IXPcU/h/7hvx4CIqPIzYSBfF5TX85NRh3Pk/3+9Qx+uBNa/BR7//7h7evY50ljAZfDIMOrF2bahW8XrhwD6oPgio82sW6tyv9zc6MeBNHSHn3V85gwrOf9TZc7219q6B2Rc4c2Ku+3dAFrS0pGBCytrd+7n46S+pdnsZkBzLwh+f0KZZz16vctHTX7KnpIKP7ppCvJ/WRSo+WM1FT31BaaWbN247gQEpsU2e7/Equ4rK+WZfGTUeJTbKRWyEi7hIF/1W/o2e2Y9QM/wcZNrzhEc5Zd05dxULVuTw+q3HM2GAbxFCVfhmEXzwoLMabvoYOO0+iE+FbZ84y4vv+BJqDgICfcb4ksQU59dqVLxTRkUxlOxw9qso3n74fskOKNnlLFfSGoNOgkk/hCPPcZYH6c5Wz3WaBo+5Cc79S9vL2bfOSQxhLicxpLZ8Hk9LWFIwIWfBihz+9t5mZl57NCPSm+9HaMyKncVc9NSX/HDyEO49e2S746p2e7nm+cWs2FnCv26eRNagNvwar2/xs/DOz2HgiXDFvyC6B6WVNZz9yGdEuIS3bz+J2Nyl8P79sOtrSB4Cp/4KMi9y1qeqy10Ne5Y7CWLbJ5CzxNnRLizceV3ZXmcxxLpiejojX3oOPPw3KhGQwyvq1nb8Hzrm+1u4FbJnwf5d0GOAM2R3wjXNr6bbkIOFsO0jZ6HG0dOcNbhCye7l8MLZ0O9ouPZNcLViefiG5G2E2ecDCte95czS9xNLCqZbu+u1Vby5cjeL7jiZIanxbS5HVfn566t5bVkOj0wfx4XjG5p/2UZrXocFt0DqCLh6HiT05utthTzw3Fz+lrKQEWVfQnw6TPkFjL+m5V841eVOIvn2U8j/xunwrJ8AotuwKGFdHrez2OLiZ2HH507n6pjpTu2hqS8ydzXsWgxbP3Ruuauonb4UFg6ZF8JxtzlfsoGk6owIWzsPNv3XWSJ+yr3OqKGWOpAHM6c462rN+BjievkntvxvnMTgrYFrF0L6UX4p1pKC6dbyyio59c+fMHFwMrOuP6b5FzTi6Y+38sf/buSnpw3nZ2f4tzoPwJYP4NVrnOag8x+DFf9E17xGqcZQMO42hp57J0Q23VQVdHvXOMlhzWvOKrqDJzsrzR5xpvOFWbD5cBLY/rnT1BUWDhkTYeipMPQUp5ax9HlY/qKzbHv/Y53kcOS5/l2xtmALrJvvJIP8jc6Q0AHHOp297gpn+9fJv2i+z8Rd7Xxx566CGxf5f1vcwq1O+TXlTg3ED+VbUjDd3sxPt/J/b2/kheuP4ZQRaa1+/X/X5nLLP5dz/ti+PHb5uO/Mo/CrnGxnX+qKIgiPwT3xh0xfO4mdFZG8e8fJjU7aC4bKGg/LdhSTGB1B78QoUuKjDg8GOFgIy2c7OwKW7na+WL1eKM1xnk8e6ksCpzqd4tENNBFWljq77X39tNPf0WOAU/uYcE3bazfFO2DdAicR7F0NiLNp01EXQ+ZU5xf+gXz4/K9OYkKdzuKT7oKE3t8vTxXeut251mmz4KhL2hZXc4q+dRJDVSlc8wb0m9Cu4iwpmG6v2u3lrEc+xaPKHy4ezbGDU1o8THV1TgmXPfsVI/sk8srNxzY449qvCjbD2vkw4VpI7MO6Pfu58MkvOCOzN09eOSFwCakVKqo93PCPJXy9raj2WJhAr/goeidGk5YQRVpiNL3jXRxd/iWZe98gMbEHEUec7tQGWjMG3+uBTe/A1085m0VFxsP4q50EkTzE+WL2VDuzymsqnF/UNRVOTeXQ/cKtTq0gZ6lTZr8s5wt81IWN913sz3EmKC5/ydkvZNIP4YTbvzuqa+lz8J87nVV0T7+/tf8ZW6d4B8w+Dyr2wzXzIaPZ7/RGWVIwBvh6WyE3v5hNWaWbgSmxXJbVn2lHZ9A7MbrR1+Tur2DqE18Q4QrjjR+dEJhlM1rgqY+38PB/N/G36WO5aLx/h9e2VpXbw02zs/l8SwH3nz+K9B7R5JVVkVdaSV5pFfvKKtlXWkV+WSUFB6prX5fRM4Ynr5zQvsmEe1Y6NYe185w9MCLjnC9+bcECiL1HOzWCoy5uXVIq3AofP+Q0iUUlwPE/cSYd5q6GFy9wajtXzOmYkVclu5waw8ECp++pjSv2WlIwxqei2sN/1+Xy6tJdfL2tiDCBU45M47Jj+nPqiDQiXIdH8xyscnPpM1+xs6icebcez5HpCUGL2+NVpj/7FZv2lbHojpPpm9SKvSH8qNrt5baXl/H+hrwWbala4/FScKCKjXvL+NWCteSVVfKrczO59riB7avxlO2FFS8528WGRzt7ZUTE+v7GfPdxeIzTLNRzYNvfD2DfemduyMZ/Q2yKU0OJTYGbP2h/Z31r7N/tJIaxlzv7n7eBJQVjGrC94CBzs3fx+rIc8sqq6BUfxSVH92N6Vn8GpsTxw5ey+XBjHrOuP4YpR7a+H8LfdhaWc/ajnzK2fxL/vHFSQGZpN8Xt8fLTOSt4e81efjt1FNccN6hVry8pr+bOuav4YGMe547uw0OXjCYhup3DNv2oxuPl5a93MDQtnpOGN7EC8+5l8OHvnJrCD/7bulFK/lJZ6tRa2phYLSkY0wS3x8sn3+QzZ+kuPtyYh8erDEyJZUdhOQ9cMIrrjh8U7BBrzVmyk3vmr+HX52Vy44mDO+x9PV7lrtdWsWDFbn517khuOmlIm8rxepWZn23jT4s20b9nDE9ddTSZfds+D8Vfvi04yB1zVrAqx1mgbsqRqfy/c0YyvHcTtUPVNn8pB5slBWNaKK+skvnLd7Ng+W6mHJnKvef4b8KQP6gqN7+YzaebC/jPT05s+kvLT7xe5ZcL1jBn6S7uPvNIfnRK+5ddWLq9iB//azkl5TU8cMEoph/TPygd6KrKq0t38cBb64kMD+O3Fx5FXmklj36wmfJqD1dM7M8dpx9Br/iutQ2rJQVjupD8sirOfORT+iZFM//WE5peNK+dVJX7F65j9lc7Gl+LqY0KDlTxv6+u5LPNBVw8vh+/u+goYiMDuSvwdxUfrOae+atZtG4fxw9N4S+XjaVPD6evpuhgNY99sJmXvt5BbISLH506jOuPHxT4kWcdxJKCMV3MonV7+eFLyxr8oi6trCGnqIKc4nJ2FTt/c4orqKj2cPIRvThzVDoDU+KafQ9V5Q/vbGTmp9uYcfIQ7j17hN9/zXu8yhMfbuGRD75hWGo8T101oUNqP59tzufOuasoLq/m7jOP5KYThzTYR7Ml7wAPvbOB9zfkkdEzhnvOHsG5o/t0imHB7WFJwZgu6K7XVjF/eQ5XThpAXmkVOb4EUFrp/s55sZEu+vd0ZkJv2lcGwIj0BM4clc5ZR6UzIj2hwS+5v767icc+3MJ1xw3k/gtGBfSL8IstBdw+ZwUHqzw8cMEoLhzfLyA1oMoaD39atInnP/+WYWnxPHr5OEb1bX7k0BdbCvjdfzawIbeUCQOS+PV5mYwf0Ib1nToJSwrGdEFllTVMfeILcvdX0j85hoyesWT0jPHdYunve5wUG1H7hb6rqJxF6/ayaN1esncUowoDkmM566h0zhzVm/H9exIWJjz50Rb+tGgTlx/Tn/+7aHSHjHTaV1rJT15ZwZJvi0iIDueMzN6cO7oPJw7vRVR4+5ttNu0t4/Y5K9i4t4xrjxvIvWePJCay5eV6vMq8ZTn86d1N5JdVcUZmbyYNTmZkn0RG9kkkuRPNNm+OJQVjuiivVxGhTb/i88uqeG/9Phat28uXWwuo8ShpCVGMyUji/Q37uGh8P/586dhW72PRHh6v8sk3eby9Zi/vrttLaaXbSRAje3OOL0G0tl2/ssbDK0t28od3NpIYHc7D08Zw6ogGlqxooYNVbp79ZCtzlu4ir+zwMuPpidFk9k1kZJ8EMvv0YGSfBAalxHX40OGW6BRJQUTOAh7F2Y7zOVV9qN7zUcCLwNFAITBdVbc3VaYlBWP8o7Syho825rFo3V4+2pjPaSPTeGT6OMJdgevEbk6128sXWwt4e3Uu767fx/6KGuKjwjl9ZBrnjO7DyUekEh3h4kCVm93FFewucfpOdhdXkFNSUXu/4IDzxX3qiDT+eMkYv85KLzhQxYbcUt+tjPV7StmSfwCP1/kujYlwcWR6AilxkbjChHCXEB4WRniYfOexK0wIDxPCXWFEhocR9Z2bq/ZYZJ3HfZOiyejZtgUSg54URMQFfAOcAeQAS4ErVHV9nXNuA8ao6i0icjlwkapOb6pcSwrG+J/Hq4S1sfYRKDUeL19sKeCdNXtZtH4vJeU1xEW6iAgPo6S85jvnRroOf2H2S4qhX88YRqQncEZm7w65piq3h837DrDelyw25pZRVlWD26N4vIrbq7i9XjyeQ/cVt8eLx6vUeJRqTwuW7IBWb1FbV2dICscB96vqmb7H9wKo6h/qnLPId85XIhIO7AVStYmgLCkY0/3UeLx8tbWQ99bvQ1H6JcXS71BfSlIMveKjOmWTTUt5vU5iqPZ4qao59NdT+7jK7aXa7aVvUnSb9wdpaVII5ADhfsCuOo9zgPorOdWeo6puEdkPpAAFdU8SkRnADIABA2xvWGO6mwhXGCcfkcrJRzSxFEUICwsTosNcTt9J42s1dkwsASy7obRdvwbQknNQ1ZmqmqWqWampXfN/CmOM6QwCmRRygLrLKWYAexo7x9d81AMowhhjTFAEMiksBYaLyGARiQQuBxbWO2chcJ3v/jTgw6b6E4wxxgRWwPoUfH0EPwYW4QxJnaWq60TkQSBbVRcCzwMvicgWnBrC5YGKxxhjTPMCuhKVqr4NvF3v2G/q3K8ELg1kDMYYY1oueLNUjDHGdDqWFIwxxtSypGCMMaZWyC2IJyL5wI42vrwX9SbGdQFd7Zq62vVA17umrnY90PWuqaHrGaiqzU70Crmk0B4ikt2Sad6hpKtdU1e7Huh619TVrge63jW153qs+cgYY0wtSwrGGGNqdbekMDPYAQRAV7umrnY90PWuqatdD3S9a2rz9XSrPgVjjDFN6241BWOMMU2wpGCMMaZWt0kKInKWiGwSkS0ick+w4/EHEdkuImtEZKWIhNx2dCIyS0TyRGRtnWPJIvKeiGz2/e0ZzBhbq5Frul9Edvs+p5Uick4wY2wNEekvIh+JyAYRWScit/uOh+Tn1MT1hPJnFC0iS0Rkle+aHvAdHywii32f0au+1aqbL6879Cm0ZL/oUCQi24EsVQ3JSTcicjJwAHhRVY/yHXsYKFLVh3zJu6eq/iKYcbZGI9d0P3BAVf8czNjaQkT6AH1UdbmIJADLgAuB6wnBz6mJ67mM0P2MBIhT1QMiEgF8DtwO/AyYr6pzROQZYJWqPt1ced2lpjAR2KKq21S1GpgDTA1yTN2eqn7K9zdVmgrM9t2fjfMPNmQ0ck0hS1VzVXW5734ZsAFnG92Q/JyauJ6QpY4DvocRvpsCpwKv+463+DPqLkmhof2iQ/p/BB8F3hWRZb59rLuC3qqaC84/YCAtyPH4y49FZLWveSkkmlrqE5FBwHhgMV3gc6p3PRDCn5GIuERkJZAHvAdsBUpU1e07pcXfed0lKbRoL+gQdIKqTgDOBn7ka7ownc/TwFBgHJAL/CW44bSeiMQD84A7VLU02PG0VwPXE9Kfkap6VHUczrbHE4GRDZ3WkrK6S1JoyX7RIUdV9/j+5gELcP5nCHX7fO2+h9p/84IcT7up6j7fP1ov8HdC7HPytVPPA15W1fm+wyH7OTV0PaH+GR2iqiXAx8CxQJKIHNpIrcXfed0lKbRkv+iQIiJxvo4yRCQO+B9gbdOvCgl19+2+DngziLH4xaEvT5+LCKHPydeJ+TywQVX/WuepkPycGrueEP+MUkUkyXc/Bjgdp6/kI2Ca77QWf0bdYvQRgG+I2SMc3i/690EOqV1EZAhO7QCcbVX/FWrXJCKvAFNwlvndB9wHvAHMBQYAO4FLVTVkOm4buaYpOM0SCmwHfnioPb6zE5ETgc+ANYDXd/iXOO3wIfc5NXE9VxC6n9EYnI5kF84P/bmq+qDvO2IOkAysAK5W1apmy+suScEYY0zzukvzkTHGmBawpGCMMaaWJQVjjDG1LCkYY4ypZUnBGGNMLUsKxtQjIp46q2Wu9OequiIyqO4KqsZ0NuHNn2JMt1PhWzLAmG7HagrGtJBv/4qHfXtYLBGRYb7jA0XkA99iah+IyADf8d4issC3zv0qETneV5RLRP7uW/v+Xd8sVGM6BUsKxnxfTL3mo+l1ntuvqqOBJ3BmyOO7/6KqjgFeBh7zHX8M+ERVxwITgHW+48OBJ1V1FFACXBLg6zGmxWxGszH1iMgBVY1v4Ph24FRV3eZbVG2vqqaISAHOxi01vuO5qtpLRPKBjLpLC/iWa35PVYf7Hv8CiFDV3wX+yoxpntUUjGkdbeR+Y+c0pO76Mx6sb890IpYUjGmd6XX+fuW7/yXOyrsAV+FshwjwAXAr1G6CkthRQRrTVvYLxZjvi/HtYnXIf1X10LDUniKyGufX/hW+Yz8FZonI3UA+cIPv+O3ATBG5EadGcCvOBi7GdFrWp2BMC/n6FLJUtSDYsRgTKNZ8ZIwxppbVFIwxxtSymoIxxphalhSMMcbUsqRgjDGmliUFY4wxtSwpGGOMqfX/AeFKn9XrSWiFAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x5f171e70>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot training & validation accuracy values\n",
    "plt.plot(history.history['acc'])\n",
    "plt.plot(history.history['val_acc'])\n",
    "\n",
    "plt.title('Model accuracy')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()\n",
    "\n",
    "# Plot training & validation loss values\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('Model loss')\n",
    "plt.ylabel('Loss')\n",
    "plt.xlabel('Epoch')\n",
    "plt.legend(['Train', 'Test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}