Fusion_Training_Vanilla.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Fushion of IMU and Audio features\n",
    "## Without doing time shifting\n",
    "## Training on Synchronized Audio and IMU data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## IMU and Audio data is concatenated at an intermediate layer and then prediction is made"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n",
      "/home/nesl/anaconda3/envs/py36/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:523: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint8 = np.dtype([(\"qint8\", np.int8, 1)])\n",
      "/home/nesl/anaconda3/envs/py36/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:524: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_quint8 = np.dtype([(\"quint8\", np.uint8, 1)])\n",
      "/home/nesl/anaconda3/envs/py36/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:525: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint16 = np.dtype([(\"qint16\", np.int16, 1)])\n",
      "/home/nesl/anaconda3/envs/py36/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:526: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_quint16 = np.dtype([(\"quint16\", np.uint16, 1)])\n",
      "/home/nesl/anaconda3/envs/py36/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:527: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  _np_qint32 = np.dtype([(\"qint32\", np.int32, 1)])\n",
      "/home/nesl/anaconda3/envs/py36/lib/python3.6/site-packages/tensorflow/python/framework/dtypes.py:532: FutureWarning: Passing (type, 1) or '1type' as a synonym of type is deprecated; in a future version of numpy, it will be understood as (type, (1,)) / '(1,)type'.\n",
      "  np_resource = np.dtype([(\"resource\", np.ubyte, 1)])\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import scipy.io\n",
    "\n",
    "from keras import optimizers\n",
    "from keras.optimizers import SGD\n",
    "from keras.optimizers import Adam\n",
    "from keras.metrics import categorical_crossentropy\n",
    "from keras.layers.normalization import BatchNormalization\n",
    "from keras.layers.convolutional import *\n",
    "from keras.callbacks import Callback\n",
    "from keras.callbacks import ModelCheckpoint\n",
    "\n",
    "\n",
    "import keras\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Conv2D, MaxPooling2D, LSTM, Dense, Dropout, Flatten, Activation\n",
    "from keras.layers.core import Permute, Reshape\n",
    "from keras import backend as K\n",
    "\n",
    "from matplotlib import pyplot as plt\n",
    "from PIL import Image as img_PIL\n",
    "\n",
    "\n",
    "# random seed.\n",
    "rand_seed = 2\n",
    "\n",
    "from numpy.random import seed\n",
    "seed(rand_seed)\n",
    "from tensorflow import set_random_seed\n",
    "set_random_seed(rand_seed)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "def one_hot_encoding(y_data):\n",
    "    Mapping=dict()\n",
    "    sub_dirs=['downstair','upstair','run','jump','walk','handwashing','exercise']\n",
    "\n",
    "    categories=10\n",
    "\n",
    "\n",
    "    count=0\n",
    "    for i in sub_dirs:\n",
    "        Mapping[i]=count\n",
    "        count=count+1\n",
    "\n",
    "    y_features2=[]\n",
    "    for i in range(len(y_data)):\n",
    "        Type=y_data[i]\n",
    "        lab=Mapping[Type]\n",
    "        y_features2.append(lab)\n",
    "\n",
    "    y_features=np.array(y_features2)\n",
    "    y_features=y_features.reshape(y_features.shape[0],1)\n",
    "    from keras.utils import to_categorical\n",
    "    y_features = to_categorical(y_features)\n",
    "\n",
    "    return y_features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "path='data/'\n",
    "\n",
    "\n",
    "def get_train_data(path=path):\n",
    "    Train_data=np.load(path+'train_5000.npz')\n",
    "    Features_imu=Train_data['arr_0']\n",
    "    Labels=Train_data['arr_1']\n",
    "    Features_sound=Train_data['arr_2']\n",
    "\n",
    "    Labels = one_hot_encoding(Labels)\n",
    "    Features_imu = Features_imu.reshape(Features_imu.shape[0],1, Features_imu.shape[1], Features_imu.shape[2]) \n",
    "\n",
    "    return Features_imu,Labels,Features_sound\n",
    "\n",
    "def get_valid_data(path=path):\n",
    "    Train_data=np.load(path+'valid_1000.npz')\n",
    "    Features_imu=Train_data['arr_0']\n",
    "    Labels=Train_data['arr_1']\n",
    "    Features_sound=Train_data['arr_2']\n",
    "\n",
    "    Labels = one_hot_encoding(Labels)\n",
    "    Features_imu = Features_imu.reshape(Features_imu.shape[0],1, Features_imu.shape[1], Features_imu.shape[2]) \n",
    "\n",
    "    return Features_imu,Labels,Features_sound\n",
    "\n",
    "def get_test_data(path=path):\n",
    "    Train_data=np.load(path+'test_1377.npz')\n",
    "    Features_imu=Train_data['arr_0']\n",
    "    Labels=Train_data['arr_1']\n",
    "    Features_sound=Train_data['arr_2']\n",
    "\n",
    "    Labels = one_hot_encoding(Labels)\n",
    "    Features_imu = Features_imu.reshape(Features_imu.shape[0],1, Features_imu.shape[1], Features_imu.shape[2]) \n",
    "\n",
    "    return Features_imu,Labels,Features_sound\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(5000, 1, 40, 12) (5000, 7) (5000, 193)\n",
      "Train Classes distribution:  [760. 994. 546. 780. 532. 732. 656.]\n",
      "(1000, 1, 40, 12) (1000, 7) (1000, 193)\n",
      "Valid Classes distribution:  [150. 188. 136. 141.  98. 157. 130.]\n",
      "(1377, 1, 40, 12) (1377, 7) (1377, 193)\n",
      "Test Classes distribution:  [219. 268. 124. 146. 143. 234. 243.]\n"
     ]
    }
   ],
   "source": [
    "Features_imu,Labels,Features_sound = get_train_data()\n",
    "print(Features_imu.shape, Labels.shape, Features_sound.shape)\n",
    "print('Train Classes distribution: ',np.sum(Labels, axis =0))\n",
    "\n",
    "Features_imu2,Labels2,Features_sound2 = get_valid_data()\n",
    "print(Features_imu2.shape, Labels2.shape, Features_sound2.shape)\n",
    "print('Valid Classes distribution: ',np.sum(Labels2, axis =0))\n",
    "\n",
    "Features_imu3,Labels3,Features_sound3 = get_test_data()\n",
    "print(Features_imu3.shape, Labels3.shape, Features_sound3.shape)\n",
    "print('Test Classes distribution: ',np.sum(Labels3, axis =0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_path = 'data/Baseline_IMU_Audio_Fusion'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Combined Model: Audio + IMU"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "rand_seed = 2\n",
    "\n",
    "from numpy.random import seed\n",
    "seed(rand_seed)\n",
    "from tensorflow import set_random_seed\n",
    "set_random_seed(rand_seed)\n",
    "\n",
    "input_audio = keras.layers.Input(shape=(193,))\n",
    "A1 = keras.layers.Dense(256, activation='relu')(input_audio)\n",
    "A2 = keras.layers.Dense(256, activation='relu')(A1)\n",
    "D1 = keras.layers.Dropout(0.3)(A2)\n",
    "A5 = keras.layers.Dense(128, activation='relu')(D1)\n",
    "\n",
    "\n",
    "\n",
    "num_feat_map=32\n",
    "_, dim, win_len, _, = Features_imu.shape\n",
    "input_imu = keras.layers.Input(shape=(dim, win_len, 12))\n",
    "\n",
    "I1 = keras.layers.Conv2D(num_feat_map, kernel_size=(1, 5),\n",
    "                 activation='relu',\n",
    "                 padding='same')(input_imu)\n",
    "I2=keras.layers.MaxPooling2D(pool_size=(1, 2))(I1)\n",
    "\n",
    "I3 = keras.layers.Conv2D(num_feat_map, kernel_size=(1, 5),\n",
    "                 activation='relu',\n",
    "                 padding='same')(I2)\n",
    "I4=keras.layers.MaxPooling2D(pool_size=(1, 2))(I3)\n",
    "I5=keras.layers.Flatten()(I4)\n",
    "\n",
    "I6= keras.layers.Dense(256, activation='relu')(I5)\n",
    "D2 = keras.layers.Dropout(0.3)(I6)\n",
    "\n",
    "I7= keras.layers.Dense(128, activation='relu')(D2)\n",
    "\n",
    "\n",
    "# equivalent to added = keras.layers.add([x1, x2])\n",
    "C1 =keras.layers.concatenate([A5, I7], axis=-1)\n",
    "\n",
    "\n",
    "C3= keras.layers.Dense(64, activation='relu')(C1)\n",
    "\n",
    "out = keras.layers.Dense(7, activation='softmax')(C3)\n",
    "\n",
    "model = keras.models.Model(inputs=[input_audio, input_imu], outputs=out)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_2 (InputLayer)            (None, 1, 40, 12)    0                                            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_1 (Conv2D)               (None, 1, 40, 32)    1952        input_2[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2D)  (None, 1, 20, 32)    0           conv2d_1[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_2 (Conv2D)               (None, 1, 20, 32)    5152        max_pooling2d_1[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "input_1 (InputLayer)            (None, 193)          0                                            \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2D)  (None, 1, 10, 32)    0           conv2d_2[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "dense_1 (Dense)                 (None, 256)          49664       input_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "flatten_1 (Flatten)             (None, 320)          0           max_pooling2d_2[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "dense_2 (Dense)                 (None, 256)          65792       dense_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dense_4 (Dense)                 (None, 256)          82176       flatten_1[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dropout_1 (Dropout)             (None, 256)          0           dense_2[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dropout_2 (Dropout)             (None, 256)          0           dense_4[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dense_3 (Dense)                 (None, 128)          32896       dropout_1[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_5 (Dense)                 (None, 128)          32896       dropout_2[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_1 (Concatenate)     (None, 256)          0           dense_3[0][0]                    \n",
      "                                                                 dense_5[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dense_6 (Dense)                 (None, 64)           16448       concatenate_1[0][0]              \n",
      "__________________________________________________________________________________________________\n",
      "dense_7 (Dense)                 (None, 7)            455         dense_6[0][0]                    \n",
      "==================================================================================================\n",
      "Total params: 287,431\n",
      "Trainable params: 287,431\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "\n",
    "\n",
    "from keras.callbacks import ReduceLROnPlateau\n",
    "from keras.callbacks import Callback\n",
    "from keras.callbacks import ModelCheckpoint\n",
    "\n",
    "checkpointer = ModelCheckpoint(filepath=model_path, verbose=1,monitor='val_acc' ,save_best_only=True, save_weights_only=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1000, 193)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Features_sound2.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 5000 samples, validate on 1000 samples\n",
      "Epoch 1/100\n",
      "5000/5000 [==============================] - 33s 7ms/step - loss: 9.9775 - acc: 0.3168 - val_loss: 7.9915 - val_acc: 0.4750\n",
      "\n",
      "Epoch 00001: val_acc improved from -inf to 0.47500, saving model to data/Baseline_IMU_Audio_Fusion\n",
      "Epoch 2/100\n",
      "5000/5000 [==============================] - 18s 4ms/step - loss: 7.2574 - acc: 0.5128 - val_loss: 6.6146 - val_acc: 0.5640\n",
      "\n",
      "Epoch 00002: val_acc improved from 0.47500 to 0.56400, saving model to data/Baseline_IMU_Audio_Fusion\n",
      "Epoch 3/100\n",
      "5000/5000 [==============================] - 17s 3ms/step - loss: 3.8450 - acc: 0.6404 - val_loss: 1.0586 - val_acc: 0.8030\n",
      "\n",
      "Epoch 00003: val_acc improved from 0.56400 to 0.80300, saving model to data/Baseline_IMU_Audio_Fusion\n",
      "Epoch 4/100\n",
      "5000/5000 [==============================] - 18s 4ms/step - loss: 0.7992 - acc: 0.8120 - val_loss: 0.4143 - val_acc: 0.8750\n",
      "\n",
      "Epoch 00004: val_acc improved from 0.80300 to 0.87500, saving model to data/Baseline_IMU_Audio_Fusion\n",
      "Epoch 5/100\n",
      "5000/5000 [==============================] - 18s 4ms/step - loss: 0.3619 - acc: 0.8882 - val_loss: 0.2234 - val_acc: 0.9240\n",
      "\n",
      "Epoch 00005: val_acc improved from 0.87500 to 0.92400, saving model to data/Baseline_IMU_Audio_Fusion\n",
      "Epoch 6/100\n",
      "5000/5000 [==============================] - 18s 4ms/step - loss: 0.2374 - acc: 0.9252 - val_loss: 0.1982 - val_acc: 0.9350\n",
      "\n",
      "Epoch 00006: val_acc improved from 0.92400 to 0.93500, saving model to data/Baseline_IMU_Audio_Fusion\n",
      "Epoch 7/100\n",
      "5000/5000 [==============================] - 18s 4ms/step - loss: 0.1658 - acc: 0.9420 - val_loss: 0.2137 - val_acc: 0.9440\n",
      "\n",
      "Epoch 00007: val_acc improved from 0.93500 to 0.94400, saving model to data/Baseline_IMU_Audio_Fusion\n",
      "Epoch 8/100\n",
      "5000/5000 [==============================] - 18s 4ms/step - loss: 0.1795 - acc: 0.9390 - val_loss: 0.1624 - val_acc: 0.9390\n",
      "\n",
      "Epoch 00008: val_acc did not improve from 0.94400\n",
      "Epoch 9/100\n",
      "5000/5000 [==============================] - 18s 4ms/step - loss: 0.1374 - acc: 0.9494 - val_loss: 0.1521 - val_acc: 0.9570\n",
      "\n",
      "Epoch 00009: val_acc improved from 0.94400 to 0.95700, saving model to data/Baseline_IMU_Audio_Fusion\n",
      "Epoch 10/100\n",
      "5000/5000 [==============================] - 18s 4ms/step - loss: 0.1118 - acc: 0.9582 - val_loss: 0.1526 - val_acc: 0.9540\n",
      "\n",
      "Epoch 00010: val_acc did not improve from 0.95700\n",
      "Epoch 11/100\n",
      "5000/5000 [==============================] - 18s 4ms/step - loss: 0.1129 - acc: 0.9566 - val_loss: 0.1498 - val_acc: 0.9500\n",
      "\n",
      "Epoch 00011: val_acc did not improve from 0.95700\n",
      "Epoch 12/100\n",
      "5000/5000 [==============================] - 18s 4ms/step - loss: 0.0929 - acc: 0.9632 - val_loss: 0.1314 - val_acc: 0.9660\n",
      "\n",
      "Epoch 00012: val_acc improved from 0.95700 to 0.96600, saving model to data/Baseline_IMU_Audio_Fusion\n",
      "Epoch 13/100\n",
      "5000/5000 [==============================] - 13s 3ms/step - loss: 0.0798 - acc: 0.9708 - val_loss: 0.1270 - val_acc: 0.9640\n",
      "\n",
      "Epoch 00013: val_acc did not improve from 0.96600\n",
      "Epoch 14/100\n",
      "5000/5000 [==============================] - 13s 3ms/step - loss: 0.0710 - acc: 0.9712 - val_loss: 0.1460 - val_acc: 0.9480\n",
      "\n",
      "Epoch 00014: val_acc did not improve from 0.96600\n",
      "Epoch 15/100\n",
      "5000/5000 [==============================] - 12s 2ms/step - loss: 0.0684 - acc: 0.9708 - val_loss: 0.1277 - val_acc: 0.9610\n",
      "\n",
      "Epoch 00015: val_acc did not improve from 0.96600\n",
      "Epoch 16/100\n",
      "5000/5000 [==============================] - 12s 2ms/step - loss: 0.0466 - acc: 0.9806 - val_loss: 0.1306 - val_acc: 0.9600\n",
      "\n",
      "Epoch 00016: val_acc did not improve from 0.96600\n",
      "Epoch 17/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0444 - acc: 0.9828 - val_loss: 0.1264 - val_acc: 0.9650\n",
      "\n",
      "Epoch 00017: val_acc did not improve from 0.96600\n",
      "Epoch 18/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0411 - acc: 0.9840 - val_loss: 0.1259 - val_acc: 0.9620\n",
      "\n",
      "Epoch 00018: val_acc did not improve from 0.96600\n",
      "Epoch 19/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0366 - acc: 0.9838 - val_loss: 0.1234 - val_acc: 0.9690\n",
      "\n",
      "Epoch 00019: val_acc improved from 0.96600 to 0.96900, saving model to data/Baseline_IMU_Audio_Fusion\n",
      "Epoch 20/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0416 - acc: 0.9832 - val_loss: 0.1142 - val_acc: 0.9630\n",
      "\n",
      "Epoch 00020: val_acc did not improve from 0.96900\n",
      "Epoch 21/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0410 - acc: 0.9848 - val_loss: 0.1170 - val_acc: 0.9620\n",
      "\n",
      "Epoch 00021: val_acc did not improve from 0.96900\n",
      "Epoch 22/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0337 - acc: 0.9860 - val_loss: 0.1797 - val_acc: 0.9570\n",
      "\n",
      "Epoch 00022: val_acc did not improve from 0.96900\n",
      "Epoch 23/100\n",
      "5000/5000 [==============================] - 12s 2ms/step - loss: 0.0378 - acc: 0.9856 - val_loss: 0.1171 - val_acc: 0.9730\n",
      "\n",
      "Epoch 00023: val_acc improved from 0.96900 to 0.97300, saving model to data/Baseline_IMU_Audio_Fusion\n",
      "Epoch 24/100\n",
      "5000/5000 [==============================] - 12s 2ms/step - loss: 0.0270 - acc: 0.9898 - val_loss: 0.1150 - val_acc: 0.9700\n",
      "\n",
      "Epoch 00024: val_acc did not improve from 0.97300\n",
      "Epoch 25/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0339 - acc: 0.9864 - val_loss: 0.1182 - val_acc: 0.9630\n",
      "\n",
      "Epoch 00025: val_acc did not improve from 0.97300\n",
      "Epoch 26/100\n",
      "5000/5000 [==============================] - 12s 2ms/step - loss: 0.0291 - acc: 0.9876 - val_loss: 0.1235 - val_acc: 0.9750\n",
      "\n",
      "Epoch 00026: val_acc improved from 0.97300 to 0.97500, saving model to data/Baseline_IMU_Audio_Fusion\n",
      "Epoch 27/100\n",
      "5000/5000 [==============================] - 12s 2ms/step - loss: 0.0270 - acc: 0.9898 - val_loss: 0.1179 - val_acc: 0.9690\n",
      "\n",
      "Epoch 00027: val_acc did not improve from 0.97500\n",
      "Epoch 28/100\n",
      "5000/5000 [==============================] - 12s 2ms/step - loss: 0.0410 - acc: 0.9880 - val_loss: 0.1509 - val_acc: 0.9730\n",
      "\n",
      "Epoch 00028: val_acc did not improve from 0.97500\n",
      "Epoch 29/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0466 - acc: 0.9832 - val_loss: 0.1909 - val_acc: 0.9500\n",
      "\n",
      "Epoch 00029: val_acc did not improve from 0.97500\n",
      "Epoch 30/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0366 - acc: 0.9862 - val_loss: 0.1332 - val_acc: 0.9690\n",
      "\n",
      "Epoch 00030: val_acc did not improve from 0.97500\n",
      "Epoch 31/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0259 - acc: 0.9896 - val_loss: 0.1319 - val_acc: 0.9650\n",
      "\n",
      "Epoch 00031: val_acc did not improve from 0.97500\n",
      "Epoch 32/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0522 - acc: 0.9826 - val_loss: 0.1734 - val_acc: 0.9580\n",
      "\n",
      "Epoch 00032: val_acc did not improve from 0.97500\n",
      "Epoch 33/100\n",
      "5000/5000 [==============================] - 12s 2ms/step - loss: 0.0390 - acc: 0.9888 - val_loss: 0.1738 - val_acc: 0.9620\n",
      "\n",
      "Epoch 00033: val_acc did not improve from 0.97500\n",
      "Epoch 34/100\n",
      "5000/5000 [==============================] - 12s 2ms/step - loss: 0.2220 - acc: 0.9498 - val_loss: 0.1706 - val_acc: 0.9560\n",
      "\n",
      "Epoch 00034: val_acc did not improve from 0.97500\n",
      "Epoch 35/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0773 - acc: 0.9730 - val_loss: 0.1281 - val_acc: 0.9640\n",
      "\n",
      "Epoch 00035: val_acc did not improve from 0.97500\n",
      "Epoch 36/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0397 - acc: 0.9870 - val_loss: 0.1301 - val_acc: 0.9630\n",
      "\n",
      "Epoch 00036: val_acc did not improve from 0.97500\n",
      "Epoch 37/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0344 - acc: 0.9868 - val_loss: 0.1229 - val_acc: 0.9710\n",
      "\n",
      "Epoch 00037: val_acc did not improve from 0.97500\n",
      "Epoch 38/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0207 - acc: 0.9922 - val_loss: 0.1064 - val_acc: 0.9760\n",
      "\n",
      "Epoch 00038: val_acc improved from 0.97500 to 0.97600, saving model to data/Baseline_IMU_Audio_Fusion\n",
      "Epoch 39/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0268 - acc: 0.9896 - val_loss: 0.1082 - val_acc: 0.9670\n",
      "\n",
      "Epoch 00039: val_acc did not improve from 0.97600\n",
      "Epoch 40/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0307 - acc: 0.9902 - val_loss: 0.1217 - val_acc: 0.9690\n",
      "\n",
      "Epoch 00040: val_acc did not improve from 0.97600\n",
      "Epoch 41/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0350 - acc: 0.9872 - val_loss: 0.1432 - val_acc: 0.9610\n",
      "\n",
      "Epoch 00041: val_acc did not improve from 0.97600\n",
      "Epoch 42/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0361 - acc: 0.9874 - val_loss: 0.1226 - val_acc: 0.9680\n",
      "\n",
      "Epoch 00042: val_acc did not improve from 0.97600\n",
      "Epoch 43/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0202 - acc: 0.9942 - val_loss: 0.1276 - val_acc: 0.9720\n",
      "\n",
      "Epoch 00043: val_acc did not improve from 0.97600\n",
      "Epoch 44/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0169 - acc: 0.9934 - val_loss: 0.1366 - val_acc: 0.9730\n",
      "\n",
      "Epoch 00044: val_acc did not improve from 0.97600\n",
      "Epoch 45/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0223 - acc: 0.9936 - val_loss: 0.1459 - val_acc: 0.9680\n",
      "\n",
      "Epoch 00045: val_acc did not improve from 0.97600\n",
      "Epoch 46/100\n",
      "5000/5000 [==============================] - 12s 2ms/step - loss: 0.0190 - acc: 0.9934 - val_loss: 0.1133 - val_acc: 0.9710\n",
      "\n",
      "Epoch 00046: val_acc did not improve from 0.97600\n",
      "Epoch 47/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0150 - acc: 0.9950 - val_loss: 0.1054 - val_acc: 0.9740\n",
      "\n",
      "Epoch 00047: val_acc did not improve from 0.97600\n",
      "Epoch 48/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0151 - acc: 0.9960 - val_loss: 0.0893 - val_acc: 0.9760\n",
      "\n",
      "Epoch 00048: val_acc improved from 0.97600 to 0.97600, saving model to data/Baseline_IMU_Audio_Fusion\n",
      "Epoch 49/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0070 - acc: 0.9980 - val_loss: 0.1007 - val_acc: 0.9750\n",
      "\n",
      "Epoch 00049: val_acc did not improve from 0.97600\n",
      "Epoch 50/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0078 - acc: 0.9972 - val_loss: 0.0901 - val_acc: 0.9800\n",
      "\n",
      "Epoch 00050: val_acc improved from 0.97600 to 0.98000, saving model to data/Baseline_IMU_Audio_Fusion\n",
      "Epoch 51/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0070 - acc: 0.9972 - val_loss: 0.1088 - val_acc: 0.9740\n",
      "\n",
      "Epoch 00051: val_acc did not improve from 0.98000\n",
      "Epoch 52/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0143 - acc: 0.9956 - val_loss: 0.0910 - val_acc: 0.9780\n",
      "\n",
      "Epoch 00052: val_acc did not improve from 0.98000\n",
      "Epoch 53/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0109 - acc: 0.9966 - val_loss: 0.1278 - val_acc: 0.9730\n",
      "\n",
      "Epoch 00053: val_acc did not improve from 0.98000\n",
      "Epoch 54/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0078 - acc: 0.9978 - val_loss: 0.1128 - val_acc: 0.9690\n",
      "\n",
      "Epoch 00054: val_acc did not improve from 0.98000\n",
      "Epoch 55/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0148 - acc: 0.9938 - val_loss: 0.1963 - val_acc: 0.9600\n",
      "\n",
      "Epoch 00055: val_acc did not improve from 0.98000\n",
      "Epoch 56/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0733 - acc: 0.9804 - val_loss: 0.1521 - val_acc: 0.9690\n",
      "\n",
      "Epoch 00056: val_acc did not improve from 0.98000\n",
      "Epoch 57/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0408 - acc: 0.9862 - val_loss: 0.1286 - val_acc: 0.9730\n",
      "\n",
      "Epoch 00057: val_acc did not improve from 0.98000\n",
      "Epoch 58/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0169 - acc: 0.9938 - val_loss: 0.1154 - val_acc: 0.9790\n",
      "\n",
      "Epoch 00058: val_acc did not improve from 0.98000\n",
      "Epoch 59/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0165 - acc: 0.9954 - val_loss: 0.1204 - val_acc: 0.9700\n",
      "\n",
      "Epoch 00059: val_acc did not improve from 0.98000\n",
      "Epoch 60/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0200 - acc: 0.9938 - val_loss: 0.1273 - val_acc: 0.9730\n",
      "\n",
      "Epoch 00060: val_acc did not improve from 0.98000\n",
      "Epoch 61/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0163 - acc: 0.9954 - val_loss: 0.1561 - val_acc: 0.9710\n",
      "\n",
      "Epoch 00061: val_acc did not improve from 0.98000\n",
      "Epoch 62/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0352 - acc: 0.9904 - val_loss: 0.1332 - val_acc: 0.9650\n",
      "\n",
      "Epoch 00062: val_acc did not improve from 0.98000\n",
      "Epoch 63/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0154 - acc: 0.9952 - val_loss: 0.0700 - val_acc: 0.9780\n",
      "\n",
      "Epoch 00063: val_acc did not improve from 0.98000\n",
      "Epoch 64/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0068 - acc: 0.9976 - val_loss: 0.0981 - val_acc: 0.9730\n",
      "\n",
      "Epoch 00064: val_acc did not improve from 0.98000\n",
      "Epoch 65/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0075 - acc: 0.9980 - val_loss: 0.0884 - val_acc: 0.9800\n",
      "\n",
      "Epoch 00065: val_acc improved from 0.98000 to 0.98000, saving model to data/Baseline_IMU_Audio_Fusion\n",
      "Epoch 66/100\n",
      "5000/5000 [==============================] - 12s 2ms/step - loss: 0.0124 - acc: 0.9958 - val_loss: 0.1201 - val_acc: 0.9780\n",
      "\n",
      "Epoch 00066: val_acc did not improve from 0.98000\n",
      "Epoch 67/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0130 - acc: 0.9962 - val_loss: 0.1664 - val_acc: 0.9750\n",
      "\n",
      "Epoch 00067: val_acc did not improve from 0.98000\n",
      "Epoch 68/100\n",
      "5000/5000 [==============================] - 12s 2ms/step - loss: 0.0059 - acc: 0.9982 - val_loss: 0.1223 - val_acc: 0.9760\n",
      "\n",
      "Epoch 00068: val_acc did not improve from 0.98000\n",
      "Epoch 69/100\n",
      "5000/5000 [==============================] - 12s 2ms/step - loss: 0.0057 - acc: 0.9978 - val_loss: 0.1616 - val_acc: 0.9730\n",
      "\n",
      "Epoch 00069: val_acc did not improve from 0.98000\n",
      "Epoch 70/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0054 - acc: 0.9982 - val_loss: 0.1215 - val_acc: 0.9790\n",
      "\n",
      "Epoch 00070: val_acc did not improve from 0.98000\n",
      "Epoch 71/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0029 - acc: 0.9984 - val_loss: 0.1304 - val_acc: 0.9730\n",
      "\n",
      "Epoch 00071: val_acc did not improve from 0.98000\n",
      "Epoch 72/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0044 - acc: 0.9978 - val_loss: 0.1390 - val_acc: 0.9780\n",
      "\n",
      "Epoch 00072: val_acc did not improve from 0.98000\n",
      "Epoch 73/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0065 - acc: 0.9980 - val_loss: 0.1406 - val_acc: 0.9730\n",
      "\n",
      "Epoch 00073: val_acc did not improve from 0.98000\n",
      "Epoch 74/100\n",
      "5000/5000 [==============================] - 12s 2ms/step - loss: 0.0202 - acc: 0.9954 - val_loss: 0.2111 - val_acc: 0.9640\n",
      "\n",
      "Epoch 00074: val_acc did not improve from 0.98000\n",
      "Epoch 75/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0376 - acc: 0.9902 - val_loss: 0.2268 - val_acc: 0.9550\n",
      "\n",
      "Epoch 00075: val_acc did not improve from 0.98000\n",
      "Epoch 76/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0426 - acc: 0.9880 - val_loss: 0.2034 - val_acc: 0.9630\n",
      "\n",
      "Epoch 00076: val_acc did not improve from 0.98000\n",
      "Epoch 77/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.2110 - acc: 0.9586 - val_loss: 0.2489 - val_acc: 0.9490\n",
      "\n",
      "Epoch 00077: val_acc did not improve from 0.98000\n",
      "Epoch 78/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.3014 - acc: 0.9514 - val_loss: 0.1553 - val_acc: 0.9730\n",
      "\n",
      "Epoch 00078: val_acc did not improve from 0.98000\n",
      "Epoch 79/100\n",
      "5000/5000 [==============================] - 11s 2ms/step - loss: 0.0908 - acc: 0.9818 - val_loss: 0.2140 - val_acc: 0.9660\n",
      "\n",
      "Epoch 00079: val_acc did not improve from 0.98000\n",
      "Epoch 80/100\n",
      "5000/5000 [==============================] - 6s 1ms/step - loss: 0.0699 - acc: 0.9838 - val_loss: 0.2070 - val_acc: 0.9630\n",
      "\n",
      "Epoch 00080: val_acc did not improve from 0.98000\n",
      "Epoch 81/100\n",
      "5000/5000 [==============================] - 5s 925us/step - loss: 0.1504 - acc: 0.9760 - val_loss: 0.1549 - val_acc: 0.9660\n",
      "\n",
      "Epoch 00081: val_acc did not improve from 0.98000\n",
      "Epoch 82/100\n",
      "5000/5000 [==============================] - 5s 921us/step - loss: 0.0318 - acc: 0.9918 - val_loss: 0.1412 - val_acc: 0.9780\n",
      "\n",
      "Epoch 00082: val_acc did not improve from 0.98000\n",
      "Epoch 83/100\n",
      "5000/5000 [==============================] - 5s 927us/step - loss: 0.0252 - acc: 0.9926 - val_loss: 0.1177 - val_acc: 0.9790\n",
      "\n",
      "Epoch 00083: val_acc did not improve from 0.98000\n",
      "Epoch 84/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5000/5000 [==============================] - 5s 915us/step - loss: 0.0217 - acc: 0.9920 - val_loss: 0.1030 - val_acc: 0.9770\n",
      "\n",
      "Epoch 00084: val_acc did not improve from 0.98000\n",
      "Epoch 85/100\n",
      "5000/5000 [==============================] - 5s 915us/step - loss: 0.0373 - acc: 0.9910 - val_loss: 0.1293 - val_acc: 0.9730\n",
      "\n",
      "Epoch 00085: val_acc did not improve from 0.98000\n",
      "Epoch 86/100\n",
      "5000/5000 [==============================] - 5s 942us/step - loss: 0.0229 - acc: 0.9932 - val_loss: 0.1221 - val_acc: 0.9780\n",
      "\n",
      "Epoch 00086: val_acc did not improve from 0.98000\n",
      "Epoch 87/100\n",
      "5000/5000 [==============================] - 6s 1ms/step - loss: 0.0066 - acc: 0.9976 - val_loss: 0.1174 - val_acc: 0.9780\n",
      "\n",
      "Epoch 00087: val_acc did not improve from 0.98000\n",
      "Epoch 88/100\n",
      "5000/5000 [==============================] - 8s 2ms/step - loss: 0.0077 - acc: 0.9968 - val_loss: 0.1659 - val_acc: 0.9740\n",
      "\n",
      "Epoch 00088: val_acc did not improve from 0.98000\n",
      "Epoch 89/100\n",
      "5000/5000 [==============================] - 8s 2ms/step - loss: 0.0088 - acc: 0.9972 - val_loss: 0.1212 - val_acc: 0.9730\n",
      "\n",
      "Epoch 00089: val_acc did not improve from 0.98000\n",
      "Epoch 90/100\n",
      "5000/5000 [==============================] - 8s 2ms/step - loss: 0.0059 - acc: 0.9990 - val_loss: 0.1051 - val_acc: 0.9800\n",
      "\n",
      "Epoch 00090: val_acc did not improve from 0.98000\n",
      "Epoch 91/100\n",
      "5000/5000 [==============================] - 8s 2ms/step - loss: 0.0022 - acc: 0.9992 - val_loss: 0.1145 - val_acc: 0.9800\n",
      "\n",
      "Epoch 00091: val_acc did not improve from 0.98000\n",
      "Epoch 92/100\n",
      "5000/5000 [==============================] - 8s 2ms/step - loss: 0.0027 - acc: 0.9994 - val_loss: 0.1379 - val_acc: 0.9750\n",
      "\n",
      "Epoch 00092: val_acc did not improve from 0.98000\n",
      "Epoch 93/100\n",
      "5000/5000 [==============================] - 8s 2ms/step - loss: 0.0020 - acc: 0.9994 - val_loss: 0.1311 - val_acc: 0.9780\n",
      "\n",
      "Epoch 00093: val_acc did not improve from 0.98000\n",
      "Epoch 94/100\n",
      "5000/5000 [==============================] - 8s 2ms/step - loss: 0.0042 - acc: 0.9988 - val_loss: 0.1499 - val_acc: 0.9770\n",
      "\n",
      "Epoch 00094: val_acc did not improve from 0.98000\n",
      "Epoch 95/100\n",
      "5000/5000 [==============================] - 8s 2ms/step - loss: 0.0019 - acc: 0.9996 - val_loss: 0.1388 - val_acc: 0.9800\n",
      "\n",
      "Epoch 00095: val_acc did not improve from 0.98000\n",
      "Epoch 96/100\n",
      "5000/5000 [==============================] - 8s 2ms/step - loss: 0.0020 - acc: 0.9996 - val_loss: 0.1264 - val_acc: 0.9790\n",
      "\n",
      "Epoch 00096: val_acc did not improve from 0.98000\n",
      "Epoch 97/100\n",
      "5000/5000 [==============================] - 8s 2ms/step - loss: 0.0036 - acc: 0.9992 - val_loss: 0.1272 - val_acc: 0.9780\n",
      "\n",
      "Epoch 00097: val_acc did not improve from 0.98000\n",
      "Epoch 98/100\n",
      "5000/5000 [==============================] - 8s 2ms/step - loss: 0.0019 - acc: 0.9996 - val_loss: 0.1191 - val_acc: 0.9800\n",
      "\n",
      "Epoch 00098: val_acc did not improve from 0.98000\n",
      "Epoch 99/100\n",
      "5000/5000 [==============================] - 8s 2ms/step - loss: 0.0026 - acc: 0.9996 - val_loss: 0.1090 - val_acc: 0.9790\n",
      "\n",
      "Epoch 00099: val_acc did not improve from 0.98000\n",
      "Epoch 100/100\n",
      "5000/5000 [==============================] - 8s 2ms/step - loss: 0.0012 - acc: 0.9994 - val_loss: 0.1052 - val_acc: 0.9790\n",
      "\n",
      "Epoch 00100: val_acc did not improve from 0.98000\n",
      "---Time 1147.1418325901031 seconds ---\n"
     ]
    }
   ],
   "source": [
    "batch_size = 128\n",
    "epochs = 100\n",
    "from keras import optimizers\n",
    "from keras.optimizers import Adam\n",
    "\n",
    "adam = optimizers.Adam(lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=None, decay=0.0, amsgrad=False)\n",
    "model.compile(loss=keras.losses.categorical_crossentropy,\n",
    "              optimizer= adam,\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "import time\n",
    "start_time = time.time()\n",
    "\n",
    "learning_hist = model.fit([Features_sound,Features_imu], Labels,\n",
    "                    verbose=1,\n",
    "                    epochs=epochs,\n",
    "                        shuffle=True,\n",
    "                    batch_size=batch_size,\n",
    "                        callbacks=[checkpointer],\n",
    "                    validation_data=([Features_sound2,Features_imu2], Labels2)\n",
    "                         \n",
    "                         )\n",
    "\n",
    "end_time = time.time()\n",
    "\n",
    "print(\"---Time %s seconds ---\" % (end_time - start_time))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Accuracies"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import load_model\n",
    "\n",
    "model = load_model(model_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5000/5000 [==============================] - 8s 2ms/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.0015213509622852143, 0.9998]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Train Accuracy\n",
    "model.evaluate([Features_sound,Features_imu],Labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1000/1000 [==============================] - 2s 2ms/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.08836310338915064, 0.98]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Valid Accuracy\n",
    "model.evaluate([Features_sound2,Features_imu2],Labels2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1377/1377 [==============================] - 2s 2ms/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.15336600019663232, 0.9607843137254902]"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## Test Accuracy\n",
    "model.evaluate([Features_sound3,Features_imu3],Labels3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Plotting"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys(['val_loss', 'val_acc', 'loss', 'acc'])\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from matplotlib import pyplot as plt\n",
    "\n",
    "print(learning_hist.history.keys())\n",
    "plt.rcParams.update({'font.size': 18})\n",
    "\n",
    "plt.figure(figsize=(10,5))\n",
    "\n",
    "plt.plot(learning_hist.history['acc'])\n",
    "plt.plot(learning_hist.history['val_acc'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'Valid'], loc='upper right')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# summarize history for loss\n",
    "\n",
    "from matplotlib import pyplot as plt\n",
    "\n",
    "plt.figure(figsize=(10,5))\n",
    "plt.rcParams.update({'font.size': 18})\n",
    "\n",
    "\n",
    "plt.plot(learning_hist.history['loss'])\n",
    "plt.plot(learning_hist.history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'Valid'], loc='upper right')\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": "3.6.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}