CNN_classification.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Convolutional Neural Networks\n",
    "\n",
    "__Goals__:\n",
    "- Design and implementation of CNNs.\n",
    "- Understanding the different effects of linear and nonlinear activation functions.\n",
    "- CNN visualization.\n",
    "\n",
    "__DataSet__: the dataset consists of around 9K images (some grayscale and some RGB) belonging to 101 classes. The shape of each image is (64,64,3). Every image is labeled with one of the classes. The image file is contained in the folder named after the class name."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import os\n",
    "from matplotlib import pyplot as plt\n",
    "import random\n",
    "import cv2\n",
    "\n",
    "import tensorflow as tf\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Dense, Dropout, Activation, Flatten, Conv2D, MaxPooling2D\n",
    "from tensorflow.keras.models import model_from_json\n",
    "from tensorflow.keras.models import load_model\n",
    "from tensorflow.keras import optimizers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [],
   "source": [
    "#tf.test.is_gpu_available()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Reading the class: Faces - with label: 0\n",
      "Reading the class: Faces_easy - with label: 1\n",
      "Reading the class: Leopards - with label: 2\n",
      "Reading the class: Motorbikes - with label: 3\n",
      "Reading the class: accordion - with label: 4\n",
      "Reading the class: airplanes - with label: 5\n",
      "Reading the class: anchor - with label: 6\n",
      "Reading the class: ant - with label: 7\n",
      "Reading the class: barrel - with label: 8\n",
      "Reading the class: bass - with label: 9\n",
      "Reading the class: beaver - with label: 10\n",
      "Reading the class: binocular - with label: 11\n",
      "Reading the class: bonsai - with label: 12\n",
      "Reading the class: brain - with label: 13\n",
      "Reading the class: brontosaurus - with label: 14\n",
      "Reading the class: buddha - with label: 15\n",
      "Reading the class: butterfly - with label: 16\n",
      "Reading the class: camera - with label: 17\n",
      "Reading the class: cannon - with label: 18\n",
      "Reading the class: car_side - with label: 19\n",
      "Reading the class: ceiling_fan - with label: 20\n",
      "Reading the class: cellphone - with label: 21\n",
      "Reading the class: chair - with label: 22\n",
      "Reading the class: chandelier - with label: 23\n",
      "Reading the class: cougar_body - with label: 24\n",
      "Reading the class: cougar_face - with label: 25\n",
      "Reading the class: crab - with label: 26\n",
      "Reading the class: crayfish - with label: 27\n",
      "Reading the class: crocodile - with label: 28\n",
      "Reading the class: crocodile_head - with label: 29\n",
      "Reading the class: cup - with label: 30\n",
      "Reading the class: dalmatian - with label: 31\n",
      "Reading the class: dollar_bill - with label: 32\n",
      "Reading the class: dolphin - with label: 33\n",
      "Reading the class: dragonfly - with label: 34\n",
      "Reading the class: electric_guitar - with label: 35\n",
      "Reading the class: elephant - with label: 36\n",
      "Reading the class: emu - with label: 37\n",
      "Reading the class: euphonium - with label: 38\n",
      "Reading the class: ewer - with label: 39\n",
      "Reading the class: ferry - with label: 40\n",
      "Reading the class: flamingo - with label: 41\n",
      "Reading the class: flamingo_head - with label: 42\n",
      "Reading the class: garfield - with label: 43\n",
      "Reading the class: gerenuk - with label: 44\n",
      "Reading the class: gramophone - with label: 45\n",
      "Reading the class: grand_piano - with label: 46\n",
      "Reading the class: hawksbill - with label: 47\n",
      "Reading the class: headphone - with label: 48\n",
      "Reading the class: hedgehog - with label: 49\n",
      "Reading the class: helicopter - with label: 50\n",
      "Reading the class: ibis - with label: 51\n",
      "Reading the class: inline_skate - with label: 52\n",
      "Reading the class: joshua_tree - with label: 53\n",
      "Reading the class: kangaroo - with label: 54\n",
      "Reading the class: ketch - with label: 55\n",
      "Reading the class: lamp - with label: 56\n",
      "Reading the class: laptop - with label: 57\n",
      "Reading the class: llama - with label: 58\n",
      "Reading the class: lobster - with label: 59\n",
      "Reading the class: lotus - with label: 60\n",
      "Reading the class: mandolin - with label: 61\n",
      "Reading the class: mayfly - with label: 62\n",
      "Reading the class: menorah - with label: 63\n",
      "Reading the class: metronome - with label: 64\n",
      "Reading the class: minaret - with label: 65\n",
      "Reading the class: nautilus - with label: 66\n",
      "Reading the class: octopus - with label: 67\n",
      "Reading the class: okapi - with label: 68\n",
      "Reading the class: pagoda - with label: 69\n",
      "Reading the class: panda - with label: 70\n",
      "Reading the class: pigeon - with label: 71\n",
      "Reading the class: pizza - with label: 72\n",
      "Reading the class: platypus - with label: 73\n",
      "Reading the class: pyramid - with label: 74\n",
      "Reading the class: revolver - with label: 75\n",
      "Reading the class: rhino - with label: 76\n",
      "Reading the class: rooster - with label: 77\n",
      "Reading the class: saxophone - with label: 78\n",
      "Reading the class: schooner - with label: 79\n",
      "Reading the class: scissors - with label: 80\n",
      "Reading the class: scorpion - with label: 81\n",
      "Reading the class: sea_horse - with label: 82\n",
      "Reading the class: snoopy - with label: 83\n",
      "Reading the class: soccer_ball - with label: 84\n",
      "Reading the class: stapler - with label: 85\n",
      "Reading the class: starfish - with label: 86\n",
      "Reading the class: stegosaurus - with label: 87\n",
      "Reading the class: stop_sign - with label: 88\n",
      "Reading the class: strawberry - with label: 89\n",
      "Reading the class: sunflower - with label: 90\n",
      "Reading the class: tick - with label: 91\n",
      "Reading the class: trilobite - with label: 92\n",
      "Reading the class: umbrella - with label: 93\n",
      "Reading the class: watch - with label: 94\n",
      "Reading the class: water_lilly - with label: 95\n",
      "Reading the class: wheelchair - with label: 96\n",
      "Reading the class: wild_cat - with label: 97\n",
      "Reading the class: windsor_chair - with label: 98\n",
      "Reading the class: wrench - with label: 99\n",
      "Reading the class: yin_yang - with label: 100\n"
     ]
    }
   ],
   "source": [
    "datapath = './101_ObjectCategories_resized'\n",
    "categories = os.listdir(datapath)\n",
    "#print(categories)\n",
    "\n",
    "# READING THE DATASET INTO THE VARIABLE ALONG WITH THE ONE-HOT CLASS LABEL\n",
    "\n",
    "data = []\n",
    "for c in categories:\n",
    "    print('Reading the class:', c, '- with label:', categories.index(c))\n",
    "    classpath = os.path.join(datapath, c)\n",
    "    class_num = categories.index(c)\n",
    "    one_hot = np.zeros(101)\n",
    "    one_hot[class_num] = 1\n",
    "    for img in os.listdir(classpath):\n",
    "        img_array = cv2.imread(os.path.join(classpath, img))\n",
    "        data.append([img_array, one_hot])    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "metadata": {},
   "outputs": [],
   "source": [
    "# SHUFFLE THE DATA \n",
    "\n",
    "np.random.seed(1)\n",
    "np.random.shuffle(data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(8677, 64, 64, 3) (8677, 101)\n"
     ]
    }
   ],
   "source": [
    "# COPYING THE IMAGES TO X AND LABELS TO t\n",
    "\n",
    "X = []\n",
    "t = []\n",
    "for images, labels in data:\n",
    "    X.append(images)\n",
    "    t.append(labels)\n",
    "print(np.shape(X), np.shape(t))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "metadata": {},
   "outputs": [],
   "source": [
    "# NORMALIZING THE IMAGE DATA\n",
    "\n",
    "X = np.asarray(X)/255\n",
    "t = np.asarray(t)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# BASELINE MODEL\n",
    "\n",
    "#### Training the model using the LeNet-5 architecture\n",
    "\n",
    "#### Using the Categorical cross-entropy as the loss funtion to be minimized"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 113,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential_3\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_6 (Conv2D)            (None, 60, 60, 32)        2432      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_6 (MaxPooling2 (None, 15, 15, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_7 (Conv2D)            (None, 11, 11, 64)        51264     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_7 (MaxPooling2 (None, 2, 2, 64)          0         \n",
      "_________________________________________________________________\n",
      "flatten_3 (Flatten)          (None, 256)               0         \n",
      "_________________________________________________________________\n",
      "dense_9 (Dense)              (None, 1024)              263168    \n",
      "_________________________________________________________________\n",
      "dense_10 (Dense)             (None, 84)                86100     \n",
      "_________________________________________________________________\n",
      "dense_11 (Dense)             (None, 101)               8585      \n",
      "=================================================================\n",
      "Total params: 411,549\n",
      "Trainable params: 411,549\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "model_baseline = Sequential()\n",
    "model_baseline.add(Conv2D(filters=32, kernel_size=(5,5), strides=(1,1), activation='relu', input_shape=(64,64,3)))\n",
    "model_baseline.add(MaxPooling2D(pool_size=(4,4), strides=(4,4)))\n",
    "model_baseline.add(Conv2D(filters=64, kernel_size=(5,5), strides=(1,1), activation='relu'))\n",
    "model_baseline.add(MaxPooling2D(pool_size=(4,4), strides=(4,4)))\n",
    "model_baseline.add(Flatten())\n",
    "model_baseline.add(Dense(units=1024, activation='relu'))\n",
    "model_baseline.add(Dense(units=84, activation='relu'))\n",
    "model_baseline.add(Dense(units=101, activation='softmax'))\n",
    "\n",
    "adam = optimizers.Adam(learning_rate=0.0001)\n",
    "#sgd = optimizers.SGD(learning_rate=0.005, decay=1e-4, momentum=0.99)\n",
    "model_baseline.compile(loss='categorical_crossentropy', optimizer=adam, metrics=['accuracy'])\n",
    "\n",
    "print(model_baseline.summary())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 114,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 7809 samples, validate on 868 samples\n",
      "Epoch 1/200\n",
      "7809/7809 [==============================] - 1s 150us/sample - loss: 4.2621 - accuracy: 0.1090 - val_loss: 4.0893 - val_accuracy: 0.1786\n",
      "Epoch 2/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 3.8177 - accuracy: 0.2313 - val_loss: 3.6469 - val_accuracy: 0.2546\n",
      "Epoch 3/200\n",
      "7809/7809 [==============================] - 1s 118us/sample - loss: 3.4517 - accuracy: 0.2876 - val_loss: 3.4057 - val_accuracy: 0.2915\n",
      "Epoch 4/200\n",
      "7809/7809 [==============================] - 1s 117us/sample - loss: 3.2425 - accuracy: 0.3230 - val_loss: 3.2441 - val_accuracy: 0.3168\n",
      "Epoch 5/200\n",
      "7809/7809 [==============================] - 1s 117us/sample - loss: 3.0707 - accuracy: 0.3484 - val_loss: 3.0785 - val_accuracy: 0.3491\n",
      "Epoch 6/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 2.9214 - accuracy: 0.3694 - val_loss: 2.9870 - val_accuracy: 0.3479\n",
      "Epoch 7/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 2.7869 - accuracy: 0.3907 - val_loss: 2.8556 - val_accuracy: 0.3779\n",
      "Epoch 8/200\n",
      "7809/7809 [==============================] - 1s 116us/sample - loss: 2.6761 - accuracy: 0.4067 - val_loss: 2.7445 - val_accuracy: 0.4055\n",
      "Epoch 9/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 2.5602 - accuracy: 0.4314 - val_loss: 2.6731 - val_accuracy: 0.4205\n",
      "Epoch 10/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 2.4665 - accuracy: 0.4401 - val_loss: 2.5882 - val_accuracy: 0.4401\n",
      "Epoch 11/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 2.3852 - accuracy: 0.4543 - val_loss: 2.5892 - val_accuracy: 0.4274\n",
      "Epoch 12/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 2.3045 - accuracy: 0.4701 - val_loss: 2.4796 - val_accuracy: 0.4562\n",
      "Epoch 13/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 2.2216 - accuracy: 0.4899 - val_loss: 2.4304 - val_accuracy: 0.4574\n",
      "Epoch 14/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 2.1583 - accuracy: 0.5007 - val_loss: 2.4048 - val_accuracy: 0.4689\n",
      "Epoch 15/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 2.0973 - accuracy: 0.5180 - val_loss: 2.3794 - val_accuracy: 0.4747\n",
      "Epoch 16/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 2.0362 - accuracy: 0.5235 - val_loss: 2.3044 - val_accuracy: 0.4931\n",
      "Epoch 17/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 1.9707 - accuracy: 0.5346 - val_loss: 2.2800 - val_accuracy: 0.4942\n",
      "Epoch 18/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 1.9133 - accuracy: 0.5495 - val_loss: 2.3209 - val_accuracy: 0.4827\n",
      "Epoch 19/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 1.8614 - accuracy: 0.5551 - val_loss: 2.2361 - val_accuracy: 0.4954\n",
      "Epoch 20/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 1.8183 - accuracy: 0.5660 - val_loss: 2.2078 - val_accuracy: 0.5150\n",
      "Epoch 21/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 1.7645 - accuracy: 0.5742 - val_loss: 2.1987 - val_accuracy: 0.5069\n",
      "Epoch 22/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 1.7215 - accuracy: 0.5877 - val_loss: 2.2030 - val_accuracy: 0.5081\n",
      "Epoch 23/200\n",
      "7809/7809 [==============================] - 1s 116us/sample - loss: 1.6732 - accuracy: 0.5976 - val_loss: 2.1639 - val_accuracy: 0.5058\n",
      "Epoch 24/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 1.6261 - accuracy: 0.6080 - val_loss: 2.1414 - val_accuracy: 0.5346\n",
      "Epoch 25/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 1.5786 - accuracy: 0.6148 - val_loss: 2.1082 - val_accuracy: 0.5380\n",
      "Epoch 26/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 1.5424 - accuracy: 0.6247 - val_loss: 2.1768 - val_accuracy: 0.5230\n",
      "Epoch 27/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 1.4990 - accuracy: 0.6370 - val_loss: 2.0889 - val_accuracy: 0.5288\n",
      "Epoch 28/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 1.4620 - accuracy: 0.6384 - val_loss: 2.1061 - val_accuracy: 0.5403\n",
      "Epoch 29/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 1.4194 - accuracy: 0.6531 - val_loss: 2.0779 - val_accuracy: 0.5518\n",
      "Epoch 30/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 1.3776 - accuracy: 0.6605 - val_loss: 2.0763 - val_accuracy: 0.5449\n",
      "Epoch 31/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 1.3477 - accuracy: 0.6678 - val_loss: 2.0642 - val_accuracy: 0.5461\n",
      "Epoch 32/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 1.3060 - accuracy: 0.6738 - val_loss: 2.1539 - val_accuracy: 0.5276\n",
      "Epoch 33/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 1.2688 - accuracy: 0.6850 - val_loss: 2.0896 - val_accuracy: 0.5565\n",
      "Epoch 34/200\n",
      "7809/7809 [==============================] - 1s 116us/sample - loss: 1.2324 - accuracy: 0.6919 - val_loss: 2.0310 - val_accuracy: 0.5576\n",
      "Epoch 35/200\n",
      "7809/7809 [==============================] - 1s 116us/sample - loss: 1.2051 - accuracy: 0.6943 - val_loss: 2.0367 - val_accuracy: 0.5576\n",
      "Epoch 36/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 1.1617 - accuracy: 0.7146 - val_loss: 2.0631 - val_accuracy: 0.5576\n",
      "Epoch 37/200\n",
      "7809/7809 [==============================] - 1s 116us/sample - loss: 1.1376 - accuracy: 0.7167 - val_loss: 2.0802 - val_accuracy: 0.5553\n",
      "Epoch 38/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 1.1066 - accuracy: 0.7205 - val_loss: 2.1012 - val_accuracy: 0.5495\n",
      "Epoch 39/200\n",
      "7809/7809 [==============================] - 1s 116us/sample - loss: 1.0687 - accuracy: 0.7322 - val_loss: 2.0341 - val_accuracy: 0.5737\n",
      "Epoch 40/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 1.0263 - accuracy: 0.7444 - val_loss: 2.0654 - val_accuracy: 0.5530\n",
      "Epoch 41/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 0.9981 - accuracy: 0.7477 - val_loss: 2.0127 - val_accuracy: 0.5714\n",
      "Epoch 42/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 0.9714 - accuracy: 0.7562 - val_loss: 2.0588 - val_accuracy: 0.5599\n",
      "Epoch 43/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 0.9412 - accuracy: 0.7650 - val_loss: 2.0583 - val_accuracy: 0.5714\n",
      "Epoch 44/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 0.9007 - accuracy: 0.7758 - val_loss: 2.0772 - val_accuracy: 0.5772\n",
      "Epoch 45/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 0.8749 - accuracy: 0.7794 - val_loss: 2.1324 - val_accuracy: 0.5576\n",
      "Epoch 46/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 0.8464 - accuracy: 0.7910 - val_loss: 2.1473 - val_accuracy: 0.5714\n",
      "Epoch 47/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.8133 - accuracy: 0.7927 - val_loss: 2.1270 - val_accuracy: 0.5599\n",
      "Epoch 48/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 0.7909 - accuracy: 0.7981 - val_loss: 2.1353 - val_accuracy: 0.5565\n",
      "Epoch 49/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 0.7561 - accuracy: 0.8093 - val_loss: 2.1619 - val_accuracy: 0.5611\n",
      "Epoch 50/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 0.7349 - accuracy: 0.8106 - val_loss: 2.1761 - val_accuracy: 0.5772\n",
      "Epoch 51/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 0.6931 - accuracy: 0.8253 - val_loss: 2.1928 - val_accuracy: 0.5645\n",
      "Epoch 52/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 0.6786 - accuracy: 0.8287 - val_loss: 2.2276 - val_accuracy: 0.5634\n",
      "Epoch 53/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 0.6514 - accuracy: 0.8316 - val_loss: 2.2235 - val_accuracy: 0.5714\n",
      "Epoch 54/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.6261 - accuracy: 0.8418 - val_loss: 2.2787 - val_accuracy: 0.5645\n",
      "Epoch 55/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.5938 - accuracy: 0.8526 - val_loss: 2.2489 - val_accuracy: 0.5726\n",
      "Epoch 56/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.5789 - accuracy: 0.8498 - val_loss: 2.3031 - val_accuracy: 0.5726\n",
      "Epoch 57/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.5584 - accuracy: 0.8547 - val_loss: 2.2858 - val_accuracy: 0.5703\n",
      "Epoch 58/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.5193 - accuracy: 0.8713 - val_loss: 2.3482 - val_accuracy: 0.5749\n",
      "Epoch 59/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.5015 - accuracy: 0.8782 - val_loss: 2.3052 - val_accuracy: 0.5691\n",
      "Epoch 60/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.4750 - accuracy: 0.8818 - val_loss: 2.3652 - val_accuracy: 0.5714\n",
      "Epoch 61/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.4556 - accuracy: 0.8858 - val_loss: 2.3727 - val_accuracy: 0.5829\n",
      "Epoch 62/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.4361 - accuracy: 0.8892 - val_loss: 2.4534 - val_accuracy: 0.5611\n",
      "Epoch 63/200\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 0.4147 - accuracy: 0.8941 - val_loss: 2.4977 - val_accuracy: 0.5749\n",
      "Epoch 64/200\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 0.3891 - accuracy: 0.9056 - val_loss: 2.4487 - val_accuracy: 0.5795\n",
      "Epoch 65/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 0.3651 - accuracy: 0.9128 - val_loss: 2.5164 - val_accuracy: 0.5760\n",
      "Epoch 66/200\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 0.3473 - accuracy: 0.9159 - val_loss: 2.5966 - val_accuracy: 0.5749\n",
      "Epoch 67/200\n",
      "7809/7809 [==============================] - 1s 119us/sample - loss: 0.3245 - accuracy: 0.9265 - val_loss: 2.6246 - val_accuracy: 0.5680\n",
      "Epoch 68/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 0.3140 - accuracy: 0.9251 - val_loss: 2.6176 - val_accuracy: 0.5714\n",
      "Epoch 69/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.3030 - accuracy: 0.9287 - val_loss: 2.6260 - val_accuracy: 0.5887\n",
      "Epoch 70/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.2819 - accuracy: 0.9352 - val_loss: 2.7714 - val_accuracy: 0.5449\n",
      "Epoch 71/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.2670 - accuracy: 0.9365 - val_loss: 2.6877 - val_accuracy: 0.5611\n",
      "Epoch 72/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.2379 - accuracy: 0.9472 - val_loss: 2.7575 - val_accuracy: 0.5714\n",
      "Epoch 73/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.2329 - accuracy: 0.9494 - val_loss: 2.7784 - val_accuracy: 0.5737\n",
      "Epoch 74/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.2302 - accuracy: 0.9485 - val_loss: 2.8628 - val_accuracy: 0.5691\n",
      "Epoch 75/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.2047 - accuracy: 0.9584 - val_loss: 2.8664 - val_accuracy: 0.5703\n",
      "Epoch 76/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.2008 - accuracy: 0.9562 - val_loss: 2.9032 - val_accuracy: 0.5680\n",
      "Epoch 77/200\n",
      "7809/7809 [==============================] - 1s 117us/sample - loss: 0.1843 - accuracy: 0.9611 - val_loss: 2.8978 - val_accuracy: 0.5806\n",
      "Epoch 78/200\n",
      "7809/7809 [==============================] - 1s 117us/sample - loss: 0.1694 - accuracy: 0.9664 - val_loss: 2.9463 - val_accuracy: 0.5876\n",
      "Epoch 79/200\n",
      "7809/7809 [==============================] - 1s 116us/sample - loss: 0.1568 - accuracy: 0.9694 - val_loss: 2.9886 - val_accuracy: 0.5783\n",
      "Epoch 80/200\n",
      "7809/7809 [==============================] - 1s 117us/sample - loss: 0.1441 - accuracy: 0.9736 - val_loss: 3.1229 - val_accuracy: 0.5760\n",
      "Epoch 81/200\n",
      "7809/7809 [==============================] - 1s 116us/sample - loss: 0.1358 - accuracy: 0.9767 - val_loss: 3.0200 - val_accuracy: 0.5703\n",
      "Epoch 82/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 0.1281 - accuracy: 0.9766 - val_loss: 3.1564 - val_accuracy: 0.5726\n",
      "Epoch 83/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.1140 - accuracy: 0.9802 - val_loss: 3.1242 - val_accuracy: 0.5634\n",
      "Epoch 84/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.1137 - accuracy: 0.9795 - val_loss: 3.2753 - val_accuracy: 0.5599\n",
      "Epoch 85/200\n",
      "7809/7809 [==============================] - 1s 112us/sample - loss: 0.1105 - accuracy: 0.9795 - val_loss: 3.2619 - val_accuracy: 0.5645\n",
      "Epoch 86/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0995 - accuracy: 0.9834 - val_loss: 3.3256 - val_accuracy: 0.5680\n",
      "Epoch 87/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.1271 - accuracy: 0.9744 - val_loss: 3.3565 - val_accuracy: 0.5657\n",
      "Epoch 88/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0877 - accuracy: 0.9863 - val_loss: 3.2261 - val_accuracy: 0.5703\n",
      "Epoch 89/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0726 - accuracy: 0.9907 - val_loss: 3.3575 - val_accuracy: 0.5703\n",
      "Epoch 90/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0781 - accuracy: 0.9876 - val_loss: 3.4513 - val_accuracy: 0.5737\n",
      "Epoch 91/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0820 - accuracy: 0.9877 - val_loss: 3.3050 - val_accuracy: 0.5806\n",
      "Epoch 92/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0730 - accuracy: 0.9867 - val_loss: 3.5405 - val_accuracy: 0.5714\n",
      "Epoch 93/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0718 - accuracy: 0.9875 - val_loss: 3.4957 - val_accuracy: 0.5876\n",
      "Epoch 94/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0754 - accuracy: 0.9877 - val_loss: 3.4834 - val_accuracy: 0.5876\n",
      "Epoch 95/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0472 - accuracy: 0.9950 - val_loss: 3.5629 - val_accuracy: 0.5818\n",
      "Epoch 96/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0515 - accuracy: 0.9932 - val_loss: 3.6638 - val_accuracy: 0.5599\n",
      "Epoch 97/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0582 - accuracy: 0.9909 - val_loss: 3.6681 - val_accuracy: 0.5634\n",
      "Epoch 98/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0985 - accuracy: 0.9777 - val_loss: 3.6540 - val_accuracy: 0.5703\n",
      "Epoch 99/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0466 - accuracy: 0.9932 - val_loss: 3.8077 - val_accuracy: 0.5588\n",
      "Epoch 100/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0930 - accuracy: 0.9768 - val_loss: 3.7956 - val_accuracy: 0.5403\n",
      "Epoch 101/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0481 - accuracy: 0.9936 - val_loss: 3.6649 - val_accuracy: 0.5853\n",
      "Epoch 102/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0267 - accuracy: 0.9974 - val_loss: 3.7548 - val_accuracy: 0.5714\n",
      "Epoch 103/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0360 - accuracy: 0.9962 - val_loss: 3.7970 - val_accuracy: 0.5668\n",
      "Epoch 104/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0237 - accuracy: 0.9986 - val_loss: 3.8030 - val_accuracy: 0.5749\n",
      "Epoch 105/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0924 - accuracy: 0.9740 - val_loss: 3.8059 - val_accuracy: 0.5657\n",
      "Epoch 106/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.1351 - accuracy: 0.9630 - val_loss: 3.6468 - val_accuracy: 0.5737\n",
      "Epoch 107/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0334 - accuracy: 0.9959 - val_loss: 3.6381 - val_accuracy: 0.5899\n",
      "Epoch 108/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0233 - accuracy: 0.9980 - val_loss: 3.7690 - val_accuracy: 0.5887\n",
      "Epoch 109/200\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0189 - accuracy: 0.9990 - val_loss: 3.8140 - val_accuracy: 0.5841\n",
      "Epoch 110/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0132 - accuracy: 0.9994 - val_loss: 3.9427 - val_accuracy: 0.5668\n",
      "Epoch 111/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0167 - accuracy: 0.9987 - val_loss: 4.1155 - val_accuracy: 0.5576\n",
      "Epoch 112/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.1489 - accuracy: 0.9567 - val_loss: 3.7733 - val_accuracy: 0.5599\n",
      "Epoch 113/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0334 - accuracy: 0.9964 - val_loss: 3.8197 - val_accuracy: 0.5806\n",
      "Epoch 114/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0209 - accuracy: 0.9980 - val_loss: 3.9194 - val_accuracy: 0.5749\n",
      "Epoch 115/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0162 - accuracy: 0.9990 - val_loss: 3.8166 - val_accuracy: 0.5841\n",
      "Epoch 116/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0419 - accuracy: 0.9921 - val_loss: 4.1632 - val_accuracy: 0.5645\n",
      "Epoch 117/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0684 - accuracy: 0.9837 - val_loss: 3.9287 - val_accuracy: 0.5749\n",
      "Epoch 118/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0285 - accuracy: 0.9971 - val_loss: 3.8853 - val_accuracy: 0.5737\n",
      "Epoch 119/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0156 - accuracy: 0.9986 - val_loss: 3.8782 - val_accuracy: 0.5818\n",
      "Epoch 120/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0160 - accuracy: 0.9985 - val_loss: 3.8707 - val_accuracy: 0.5887\n",
      "Epoch 121/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0248 - accuracy: 0.9963 - val_loss: 4.2747 - val_accuracy: 0.5829\n",
      "Epoch 122/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.1476 - accuracy: 0.9572 - val_loss: 3.8943 - val_accuracy: 0.5841\n",
      "Epoch 123/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0177 - accuracy: 0.9990 - val_loss: 3.8502 - val_accuracy: 0.5772\n",
      "Epoch 124/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0127 - accuracy: 0.9992 - val_loss: 3.8764 - val_accuracy: 0.5910\n",
      "Epoch 125/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0092 - accuracy: 0.9994 - val_loss: 3.9367 - val_accuracy: 0.5829\n",
      "Epoch 126/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0100 - accuracy: 0.9990 - val_loss: 4.0616 - val_accuracy: 0.5760\n",
      "Epoch 127/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.1416 - accuracy: 0.9583 - val_loss: 4.0600 - val_accuracy: 0.5426\n",
      "Epoch 128/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0452 - accuracy: 0.9899 - val_loss: 3.8082 - val_accuracy: 0.5818\n",
      "Epoch 129/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0121 - accuracy: 0.9992 - val_loss: 3.8351 - val_accuracy: 0.5806\n",
      "Epoch 130/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0123 - accuracy: 0.9991 - val_loss: 4.0382 - val_accuracy: 0.5714\n",
      "Epoch 131/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0087 - accuracy: 0.9996 - val_loss: 3.9761 - val_accuracy: 0.5818\n",
      "Epoch 132/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0122 - accuracy: 0.9987 - val_loss: 4.0813 - val_accuracy: 0.5818\n",
      "Epoch 133/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0938 - accuracy: 0.9714 - val_loss: 4.1731 - val_accuracy: 0.5576\n",
      "Epoch 134/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0296 - accuracy: 0.9941 - val_loss: 4.0268 - val_accuracy: 0.5818\n",
      "Epoch 135/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0205 - accuracy: 0.9964 - val_loss: 4.0391 - val_accuracy: 0.5841\n",
      "Epoch 136/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0159 - accuracy: 0.9974 - val_loss: 3.9977 - val_accuracy: 0.5714\n",
      "Epoch 137/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0073 - accuracy: 0.9996 - val_loss: 4.0763 - val_accuracy: 0.5749\n",
      "Epoch 138/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0091 - accuracy: 0.9994 - val_loss: 4.1004 - val_accuracy: 0.5887\n",
      "Epoch 139/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0846 - accuracy: 0.9757 - val_loss: 4.2811 - val_accuracy: 0.5495\n",
      "Epoch 140/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0819 - accuracy: 0.9784 - val_loss: 4.0136 - val_accuracy: 0.5737\n",
      "Epoch 141/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0121 - accuracy: 0.9983 - val_loss: 4.0008 - val_accuracy: 0.5783\n",
      "Epoch 142/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0109 - accuracy: 0.9988 - val_loss: 3.9753 - val_accuracy: 0.5933\n",
      "Epoch 143/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0084 - accuracy: 0.9995 - val_loss: 4.0642 - val_accuracy: 0.5795\n",
      "Epoch 144/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0084 - accuracy: 0.9994 - val_loss: 4.0535 - val_accuracy: 0.5806\n",
      "Epoch 145/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0148 - accuracy: 0.9974 - val_loss: 4.2321 - val_accuracy: 0.5806\n",
      "Epoch 146/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.1005 - accuracy: 0.9735 - val_loss: 4.2870 - val_accuracy: 0.5611\n",
      "Epoch 147/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0815 - accuracy: 0.9777 - val_loss: 4.0632 - val_accuracy: 0.5749\n",
      "Epoch 148/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0093 - accuracy: 0.9994 - val_loss: 3.9937 - val_accuracy: 0.5806\n",
      "Epoch 149/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0076 - accuracy: 0.9991 - val_loss: 4.0516 - val_accuracy: 0.5876\n",
      "Epoch 150/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0090 - accuracy: 0.9990 - val_loss: 4.2811 - val_accuracy: 0.5726\n",
      "Epoch 151/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0132 - accuracy: 0.9985 - val_loss: 4.0864 - val_accuracy: 0.5829\n",
      "Epoch 152/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0096 - accuracy: 0.9990 - val_loss: 4.0787 - val_accuracy: 0.5887\n",
      "Epoch 153/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0113 - accuracy: 0.9988 - val_loss: 4.1714 - val_accuracy: 0.5887\n",
      "Epoch 154/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.1378 - accuracy: 0.9606 - val_loss: 4.1093 - val_accuracy: 0.5783\n",
      "Epoch 155/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0288 - accuracy: 0.9949 - val_loss: 3.9334 - val_accuracy: 0.5841\n",
      "Epoch 156/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0057 - accuracy: 0.9995 - val_loss: 4.0337 - val_accuracy: 0.5887\n",
      "Epoch 157/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0075 - accuracy: 0.9994 - val_loss: 4.0680 - val_accuracy: 0.5876\n",
      "Epoch 158/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0103 - accuracy: 0.9986 - val_loss: 4.1043 - val_accuracy: 0.5841\n",
      "Epoch 159/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0385 - accuracy: 0.9904 - val_loss: 4.1398 - val_accuracy: 0.5611\n",
      "Epoch 160/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0905 - accuracy: 0.9709 - val_loss: 4.0932 - val_accuracy: 0.5553\n",
      "Epoch 161/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0143 - accuracy: 0.9978 - val_loss: 4.0980 - val_accuracy: 0.5772\n",
      "Epoch 162/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0071 - accuracy: 0.9991 - val_loss: 4.0768 - val_accuracy: 0.5945\n",
      "Epoch 163/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0081 - accuracy: 0.9990 - val_loss: 4.0899 - val_accuracy: 0.5876\n",
      "Epoch 164/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0060 - accuracy: 0.9992 - val_loss: 4.2737 - val_accuracy: 0.5737\n",
      "Epoch 165/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0103 - accuracy: 0.9991 - val_loss: 4.1787 - val_accuracy: 0.5945\n",
      "Epoch 166/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0074 - accuracy: 0.9988 - val_loss: 4.2517 - val_accuracy: 0.5887\n",
      "Epoch 167/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0886 - accuracy: 0.9744 - val_loss: 4.4707 - val_accuracy: 0.5553\n",
      "Epoch 168/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0573 - accuracy: 0.9849 - val_loss: 4.1552 - val_accuracy: 0.5714\n",
      "Epoch 169/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0085 - accuracy: 0.9994 - val_loss: 4.1483 - val_accuracy: 0.5853\n",
      "Epoch 170/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0081 - accuracy: 0.9990 - val_loss: 4.1726 - val_accuracy: 0.5841\n",
      "Epoch 171/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0064 - accuracy: 0.9994 - val_loss: 4.1562 - val_accuracy: 0.5956\n",
      "Epoch 172/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0073 - accuracy: 0.9994 - val_loss: 4.1776 - val_accuracy: 0.5922\n",
      "Epoch 173/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0077 - accuracy: 0.9994 - val_loss: 4.1706 - val_accuracy: 0.5910\n",
      "Epoch 174/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0089 - accuracy: 0.9990 - val_loss: 4.3321 - val_accuracy: 0.5714\n",
      "Epoch 175/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0102 - accuracy: 0.9994 - val_loss: 4.2963 - val_accuracy: 0.5841\n",
      "Epoch 176/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0078 - accuracy: 0.9992 - val_loss: 4.2865 - val_accuracy: 0.5829\n",
      "Epoch 177/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0141 - accuracy: 0.9978 - val_loss: 4.9397 - val_accuracy: 0.5588\n",
      "Epoch 178/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.1824 - accuracy: 0.9483 - val_loss: 4.0888 - val_accuracy: 0.5806\n",
      "Epoch 179/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0117 - accuracy: 0.9982 - val_loss: 3.9964 - val_accuracy: 0.5933\n",
      "Epoch 180/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0060 - accuracy: 0.9995 - val_loss: 4.0156 - val_accuracy: 0.5945\n",
      "Epoch 181/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0078 - accuracy: 0.9994 - val_loss: 4.1273 - val_accuracy: 0.5853\n",
      "Epoch 182/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0074 - accuracy: 0.9990 - val_loss: 4.0998 - val_accuracy: 0.5876\n",
      "Epoch 183/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0094 - accuracy: 0.9990 - val_loss: 4.1050 - val_accuracy: 0.5864\n",
      "Epoch 184/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0060 - accuracy: 0.9992 - val_loss: 4.1556 - val_accuracy: 0.5899\n",
      "Epoch 185/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0863 - accuracy: 0.9761 - val_loss: 4.3141 - val_accuracy: 0.5565\n",
      "Epoch 186/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0504 - accuracy: 0.9863 - val_loss: 4.2744 - val_accuracy: 0.5691\n",
      "Epoch 187/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0105 - accuracy: 0.9988 - val_loss: 4.1348 - val_accuracy: 0.5910\n",
      "Epoch 188/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 0.0102 - accuracy: 0.9990 - val_loss: 4.1581 - val_accuracy: 0.5922\n",
      "Epoch 189/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 0.0066 - accuracy: 0.9994 - val_loss: 4.1500 - val_accuracy: 0.5853\n",
      "Epoch 190/200\n",
      "7809/7809 [==============================] - 1s 116us/sample - loss: 0.0091 - accuracy: 0.9988 - val_loss: 4.2088 - val_accuracy: 0.5945\n",
      "Epoch 191/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0073 - accuracy: 0.9992 - val_loss: 4.1983 - val_accuracy: 0.5876\n",
      "Epoch 192/200\n",
      "7809/7809 [==============================] - 1s 114us/sample - loss: 0.0088 - accuracy: 0.9991 - val_loss: 4.1959 - val_accuracy: 0.5853\n",
      "Epoch 193/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0051 - accuracy: 0.9995 - val_loss: 4.3397 - val_accuracy: 0.5899\n",
      "Epoch 194/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.1395 - accuracy: 0.9612 - val_loss: 4.0955 - val_accuracy: 0.5634\n",
      "Epoch 195/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0542 - accuracy: 0.9849 - val_loss: 4.0846 - val_accuracy: 0.5795\n",
      "Epoch 196/200\n",
      "7809/7809 [==============================] - 1s 115us/sample - loss: 0.0077 - accuracy: 0.9992 - val_loss: 4.0534 - val_accuracy: 0.5899\n",
      "Epoch 197/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0072 - accuracy: 0.9990 - val_loss: 4.0649 - val_accuracy: 0.5864\n",
      "Epoch 198/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0049 - accuracy: 0.9995 - val_loss: 4.1507 - val_accuracy: 0.5910\n",
      "Epoch 199/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0047 - accuracy: 0.9994 - val_loss: 4.1357 - val_accuracy: 0.5922\n",
      "Epoch 200/200\n",
      "7809/7809 [==============================] - 1s 113us/sample - loss: 0.0082 - accuracy: 0.9990 - val_loss: 4.1648 - val_accuracy: 0.5887\n"
     ]
    }
   ],
   "source": [
    "baseline = model_baseline.fit(X, t, validation_split=0.1, batch_size=30, epochs=200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\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": [
    "plt.plot(baseline.history['accuracy'])\n",
    "plt.plot(baseline.history['val_accuracy'])\n",
    "plt.title('Baseline model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'validation'], loc='upper left')\n",
    "plt.grid()\n",
    "plt.show()\n",
    "\n",
    "plt.plot(baseline.history['loss'])\n",
    "plt.plot(baseline.history['val_loss'])\n",
    "plt.title('Baseline model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'validation'], loc='upper right')\n",
    "plt.grid()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Results of the baseline model (after 100 epochs):\n",
    "\n",
    "#### Training accuracy ~ 99% (saturation)\n",
    "#### Validation accuracy ~ 60% (saturation)\n",
    "\n",
    "#### From the above plots, we can clearly see that the baseline model is overfitting to the training data. \n",
    "\n",
    "#### Hence, we apply regularization techniques (such as L2(for kernel), Dropout) to prevent the overfitting."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Regularized model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# TRAINING THE MODEL USING LENET-5 ARCHITECTURE\n",
    "# BACKPROPAGATING FROM THE CATEGORICAL CROSS-ENTROPY LOSS + L2 REGULARIZATION LOSS\n",
    "# DROPOUT REGULARIZER ALSO INCLUDED\n",
    "\n",
    "model_final = Sequential()\n",
    "model_final.add(Conv2D(filters=32, kernel_size=(5,5), strides=(1,1), activation='relu', kernel_regularizer=tf.keras.regularizers.l2(0.001), input_shape=(64,64,3)))\n",
    "model_final.add(MaxPooling2D(pool_size=(4,4), strides=(4,4)))\n",
    "model_final.add(Dropout(0.25))\n",
    "model_final.add(Conv2D(filters=64, kernel_size=(5,5), strides=(1,1), activation='relu', kernel_regularizer=tf.keras.regularizers.l2(0.001)))\n",
    "model_final.add(MaxPooling2D(pool_size=(4,4), strides=(4,4)))\n",
    "model_final.add(Dropout(0.25))\n",
    "model_final.add(Flatten())\n",
    "model_final.add(Dense(units=1024, activation='relu', kernel_regularizer=tf.keras.regularizers.l2(0.001)))\n",
    "model_final.add(Dropout(0.25))\n",
    "model_final.add(Dense(units=84, activation='relu', kernel_regularizer=tf.keras.regularizers.l2(0.001)))\n",
    "model_final.add(Dropout(0.25))\n",
    "model_final.add(Dense(units=101, activation='softmax', kernel_regularizer=tf.keras.regularizers.l2(0.001)))\n",
    "\n",
    "adam = optimizers.Adam(learning_rate=0.0001)\n",
    "#sgd = optimizers.SGD(learning_rate=0.005, decay=1e-4, momentum=0.99)\n",
    "model_final.compile(loss='categorical_crossentropy', optimizer=adam, metrics=['accuracy'])\n",
    "\n",
    "print(model_final.summary())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 7809 samples, validate on 868 samples\n",
      "Epoch 1/700\n",
      "7809/7809 [==============================] - 1s 177us/sample - loss: 5.0296 - accuracy: 0.1007 - val_loss: 4.8994 - val_accuracy: 0.1624\n",
      "Epoch 2/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 4.6773 - accuracy: 0.1731 - val_loss: 4.6192 - val_accuracy: 0.1866\n",
      "Epoch 3/700\n",
      "7809/7809 [==============================] - 1s 124us/sample - loss: 4.3789 - accuracy: 0.2199 - val_loss: 4.3921 - val_accuracy: 0.2177\n",
      "Epoch 4/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 4.1820 - accuracy: 0.2333 - val_loss: 4.2035 - val_accuracy: 0.2235\n",
      "Epoch 5/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 4.0463 - accuracy: 0.2496 - val_loss: 4.0453 - val_accuracy: 0.2500\n",
      "Epoch 6/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 3.9397 - accuracy: 0.2696 - val_loss: 3.9301 - val_accuracy: 0.2684\n",
      "Epoch 7/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 3.8325 - accuracy: 0.2889 - val_loss: 3.8368 - val_accuracy: 0.2984\n",
      "Epoch 8/700\n",
      "7809/7809 [==============================] - 1s 124us/sample - loss: 3.7318 - accuracy: 0.3078 - val_loss: 3.7444 - val_accuracy: 0.3088\n",
      "Epoch 9/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 3.6573 - accuracy: 0.3216 - val_loss: 3.6671 - val_accuracy: 0.3260\n",
      "Epoch 10/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 3.5802 - accuracy: 0.3274 - val_loss: 3.5854 - val_accuracy: 0.3306\n",
      "Epoch 11/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 3.5104 - accuracy: 0.3391 - val_loss: 3.5194 - val_accuracy: 0.3329\n",
      "Epoch 12/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 3.4392 - accuracy: 0.3446 - val_loss: 3.4165 - val_accuracy: 0.3525\n",
      "Epoch 13/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 3.3843 - accuracy: 0.3542 - val_loss: 3.3350 - val_accuracy: 0.3767\n",
      "Epoch 14/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 3.3318 - accuracy: 0.3652 - val_loss: 3.3070 - val_accuracy: 0.3790\n",
      "Epoch 15/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 3.2695 - accuracy: 0.3744 - val_loss: 3.2017 - val_accuracy: 0.3836\n",
      "Epoch 16/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 3.2203 - accuracy: 0.3783 - val_loss: 3.1736 - val_accuracy: 0.4032\n",
      "Epoch 17/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 3.1641 - accuracy: 0.3857 - val_loss: 3.1450 - val_accuracy: 0.4044\n",
      "Epoch 18/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 3.1321 - accuracy: 0.3896 - val_loss: 3.0682 - val_accuracy: 0.4101\n",
      "Epoch 19/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 3.0976 - accuracy: 0.3948 - val_loss: 3.0330 - val_accuracy: 0.4217\n",
      "Epoch 20/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 3.0671 - accuracy: 0.3992 - val_loss: 3.0014 - val_accuracy: 0.4343\n",
      "Epoch 21/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 3.0189 - accuracy: 0.4062 - val_loss: 2.9528 - val_accuracy: 0.4378\n",
      "Epoch 22/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.9984 - accuracy: 0.4024 - val_loss: 2.9291 - val_accuracy: 0.4482\n",
      "Epoch 23/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.9622 - accuracy: 0.4134 - val_loss: 2.9042 - val_accuracy: 0.4447\n",
      "Epoch 24/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.9272 - accuracy: 0.4181 - val_loss: 2.8734 - val_accuracy: 0.4482\n",
      "Epoch 25/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.9068 - accuracy: 0.4246 - val_loss: 2.8556 - val_accuracy: 0.4585\n",
      "Epoch 26/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.8782 - accuracy: 0.4248 - val_loss: 2.8068 - val_accuracy: 0.4562\n",
      "Epoch 27/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.8303 - accuracy: 0.4351 - val_loss: 2.7760 - val_accuracy: 0.4689\n",
      "Epoch 28/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.8015 - accuracy: 0.4389 - val_loss: 2.7485 - val_accuracy: 0.4747\n",
      "Epoch 29/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.7990 - accuracy: 0.4404 - val_loss: 2.7447 - val_accuracy: 0.4735\n",
      "Epoch 30/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.7798 - accuracy: 0.4469 - val_loss: 2.7392 - val_accuracy: 0.4781\n",
      "Epoch 31/700\n",
      "7809/7809 [==============================] - 1s 124us/sample - loss: 2.7560 - accuracy: 0.4520 - val_loss: 2.7002 - val_accuracy: 0.4747\n",
      "Epoch 32/700\n",
      "7809/7809 [==============================] - 1s 124us/sample - loss: 2.7266 - accuracy: 0.4527 - val_loss: 2.6587 - val_accuracy: 0.4954\n",
      "Epoch 33/700\n",
      "7809/7809 [==============================] - 1s 124us/sample - loss: 2.7092 - accuracy: 0.4554 - val_loss: 2.6619 - val_accuracy: 0.5069\n",
      "Epoch 34/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.6948 - accuracy: 0.4573 - val_loss: 2.6538 - val_accuracy: 0.4965\n",
      "Epoch 35/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.6694 - accuracy: 0.4668 - val_loss: 2.6468 - val_accuracy: 0.4885\n",
      "Epoch 36/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.6513 - accuracy: 0.4674 - val_loss: 2.5963 - val_accuracy: 0.5104\n",
      "Epoch 37/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.6266 - accuracy: 0.4725 - val_loss: 2.5901 - val_accuracy: 0.5035\n",
      "Epoch 38/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.6263 - accuracy: 0.4704 - val_loss: 2.5606 - val_accuracy: 0.5046\n",
      "Epoch 39/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.5916 - accuracy: 0.4775 - val_loss: 2.5491 - val_accuracy: 0.5092\n",
      "Epoch 40/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.5699 - accuracy: 0.4811 - val_loss: 2.5390 - val_accuracy: 0.5253\n",
      "Epoch 41/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.5631 - accuracy: 0.4830 - val_loss: 2.5127 - val_accuracy: 0.5207\n",
      "Epoch 42/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.5464 - accuracy: 0.4850 - val_loss: 2.4997 - val_accuracy: 0.5334\n",
      "Epoch 43/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.5252 - accuracy: 0.4891 - val_loss: 2.4655 - val_accuracy: 0.5392\n",
      "Epoch 44/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.5090 - accuracy: 0.4921 - val_loss: 2.4751 - val_accuracy: 0.5323\n",
      "Epoch 45/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.5274 - accuracy: 0.4814 - val_loss: 2.4526 - val_accuracy: 0.5415\n",
      "Epoch 46/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.4858 - accuracy: 0.4957 - val_loss: 2.4351 - val_accuracy: 0.5346\n",
      "Epoch 47/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.4590 - accuracy: 0.5021 - val_loss: 2.4316 - val_accuracy: 0.5495\n",
      "Epoch 48/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.4514 - accuracy: 0.5043 - val_loss: 2.4045 - val_accuracy: 0.5541\n",
      "Epoch 49/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.4212 - accuracy: 0.5168 - val_loss: 2.3965 - val_accuracy: 0.5576\n",
      "Epoch 50/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.4340 - accuracy: 0.5074 - val_loss: 2.3696 - val_accuracy: 0.5495\n",
      "Epoch 51/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.4054 - accuracy: 0.5172 - val_loss: 2.3750 - val_accuracy: 0.5541\n",
      "Epoch 52/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.3945 - accuracy: 0.5174 - val_loss: 2.3664 - val_accuracy: 0.5703\n",
      "Epoch 53/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.3859 - accuracy: 0.5113 - val_loss: 2.3453 - val_accuracy: 0.5634\n",
      "Epoch 54/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.3750 - accuracy: 0.5153 - val_loss: 2.3338 - val_accuracy: 0.5714\n",
      "Epoch 55/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.3581 - accuracy: 0.5214 - val_loss: 2.3288 - val_accuracy: 0.5657\n",
      "Epoch 56/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.3582 - accuracy: 0.5200 - val_loss: 2.3278 - val_accuracy: 0.5714\n",
      "Epoch 57/700\n",
      "7809/7809 [==============================] - 1s 124us/sample - loss: 2.3370 - accuracy: 0.5239 - val_loss: 2.3124 - val_accuracy: 0.5622\n",
      "Epoch 58/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.3143 - accuracy: 0.5359 - val_loss: 2.3087 - val_accuracy: 0.5703\n",
      "Epoch 59/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.3313 - accuracy: 0.5282 - val_loss: 2.2980 - val_accuracy: 0.5783\n",
      "Epoch 60/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.2991 - accuracy: 0.5364 - val_loss: 2.3161 - val_accuracy: 0.5818\n",
      "Epoch 61/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.2936 - accuracy: 0.5334 - val_loss: 2.2748 - val_accuracy: 0.5864\n",
      "Epoch 62/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.2997 - accuracy: 0.5359 - val_loss: 2.2755 - val_accuracy: 0.5783\n",
      "Epoch 63/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.2611 - accuracy: 0.5426 - val_loss: 2.2459 - val_accuracy: 0.5864\n",
      "Epoch 64/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.2764 - accuracy: 0.5381 - val_loss: 2.2720 - val_accuracy: 0.5945\n",
      "Epoch 65/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.2558 - accuracy: 0.5517 - val_loss: 2.2353 - val_accuracy: 0.5910\n",
      "Epoch 66/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.2563 - accuracy: 0.5450 - val_loss: 2.2429 - val_accuracy: 0.5945\n",
      "Epoch 67/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.2530 - accuracy: 0.5449 - val_loss: 2.2357 - val_accuracy: 0.5979\n",
      "Epoch 68/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.2220 - accuracy: 0.5550 - val_loss: 2.2127 - val_accuracy: 0.5991\n",
      "Epoch 69/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.2205 - accuracy: 0.5490 - val_loss: 2.2203 - val_accuracy: 0.5968\n",
      "Epoch 70/700\n",
      "7809/7809 [==============================] - 1s 124us/sample - loss: 2.1999 - accuracy: 0.5560 - val_loss: 2.2121 - val_accuracy: 0.5945\n",
      "Epoch 71/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.1786 - accuracy: 0.5603 - val_loss: 2.2215 - val_accuracy: 0.6025\n",
      "Epoch 72/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.1897 - accuracy: 0.5626 - val_loss: 2.1968 - val_accuracy: 0.6106\n",
      "Epoch 73/700\n",
      "7809/7809 [==============================] - 1s 124us/sample - loss: 2.1898 - accuracy: 0.5573 - val_loss: 2.1902 - val_accuracy: 0.6002\n",
      "Epoch 74/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.1814 - accuracy: 0.5635 - val_loss: 2.1827 - val_accuracy: 0.6187\n",
      "Epoch 75/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.1747 - accuracy: 0.5609 - val_loss: 2.1909 - val_accuracy: 0.6106\n",
      "Epoch 76/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.1586 - accuracy: 0.5604 - val_loss: 2.1880 - val_accuracy: 0.6106\n",
      "Epoch 77/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.1443 - accuracy: 0.5646 - val_loss: 2.1807 - val_accuracy: 0.6048\n",
      "Epoch 78/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.1342 - accuracy: 0.5677 - val_loss: 2.1590 - val_accuracy: 0.6118\n",
      "Epoch 79/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.1137 - accuracy: 0.5761 - val_loss: 2.1549 - val_accuracy: 0.6037\n",
      "Epoch 80/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.1323 - accuracy: 0.5731 - val_loss: 2.1456 - val_accuracy: 0.6094\n",
      "Epoch 81/700\n",
      "7809/7809 [==============================] - 1s 124us/sample - loss: 2.1248 - accuracy: 0.5768 - val_loss: 2.1406 - val_accuracy: 0.6083\n",
      "Epoch 82/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.1109 - accuracy: 0.5775 - val_loss: 2.1658 - val_accuracy: 0.6025\n",
      "Epoch 83/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.0988 - accuracy: 0.5809 - val_loss: 2.1359 - val_accuracy: 0.6152\n",
      "Epoch 84/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.1172 - accuracy: 0.5802 - val_loss: 2.1291 - val_accuracy: 0.6152\n",
      "Epoch 85/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.0833 - accuracy: 0.5884 - val_loss: 2.1339 - val_accuracy: 0.6279\n",
      "Epoch 86/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.0632 - accuracy: 0.5846 - val_loss: 2.1245 - val_accuracy: 0.6141\n",
      "Epoch 87/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.0669 - accuracy: 0.5869 - val_loss: 2.1251 - val_accuracy: 0.6233\n",
      "Epoch 88/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.0829 - accuracy: 0.5846 - val_loss: 2.1348 - val_accuracy: 0.6233\n",
      "Epoch 89/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.0502 - accuracy: 0.5912 - val_loss: 2.1119 - val_accuracy: 0.6210\n",
      "Epoch 90/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.0434 - accuracy: 0.5946 - val_loss: 2.1056 - val_accuracy: 0.6348\n",
      "Epoch 91/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.0431 - accuracy: 0.5935 - val_loss: 2.1053 - val_accuracy: 0.6210\n",
      "Epoch 92/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.0383 - accuracy: 0.5962 - val_loss: 2.0981 - val_accuracy: 0.6279\n",
      "Epoch 93/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.0345 - accuracy: 0.5926 - val_loss: 2.1149 - val_accuracy: 0.6221\n",
      "Epoch 94/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.0376 - accuracy: 0.5888 - val_loss: 2.0887 - val_accuracy: 0.6256\n",
      "Epoch 95/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 2.0177 - accuracy: 0.6024 - val_loss: 2.1011 - val_accuracy: 0.6302\n",
      "Epoch 96/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 2.0156 - accuracy: 0.6006 - val_loss: 2.0893 - val_accuracy: 0.6290\n",
      "Epoch 97/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.9988 - accuracy: 0.6017 - val_loss: 2.0803 - val_accuracy: 0.6313\n",
      "Epoch 98/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.9929 - accuracy: 0.6038 - val_loss: 2.0789 - val_accuracy: 0.6348\n",
      "Epoch 99/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.9806 - accuracy: 0.6106 - val_loss: 2.0847 - val_accuracy: 0.6336\n",
      "Epoch 100/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.9947 - accuracy: 0.6066 - val_loss: 2.0728 - val_accuracy: 0.6313\n",
      "Epoch 101/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.9897 - accuracy: 0.6052 - val_loss: 2.0771 - val_accuracy: 0.6313\n",
      "Epoch 102/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.9842 - accuracy: 0.6072 - val_loss: 2.0767 - val_accuracy: 0.6302\n",
      "Epoch 103/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.9698 - accuracy: 0.6120 - val_loss: 2.0839 - val_accuracy: 0.6267\n",
      "Epoch 104/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.9695 - accuracy: 0.6144 - val_loss: 2.0675 - val_accuracy: 0.6302\n",
      "Epoch 105/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.9519 - accuracy: 0.6112 - val_loss: 2.0592 - val_accuracy: 0.6406\n",
      "Epoch 106/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.9535 - accuracy: 0.6157 - val_loss: 2.0625 - val_accuracy: 0.6325\n",
      "Epoch 107/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.9447 - accuracy: 0.6234 - val_loss: 2.0450 - val_accuracy: 0.6394\n",
      "Epoch 108/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.9451 - accuracy: 0.6158 - val_loss: 2.0579 - val_accuracy: 0.6348\n",
      "Epoch 109/700\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.9364 - accuracy: 0.6186 - val_loss: 2.0484 - val_accuracy: 0.6371\n",
      "Epoch 110/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.9285 - accuracy: 0.6174 - val_loss: 2.0428 - val_accuracy: 0.6417\n",
      "Epoch 111/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.9394 - accuracy: 0.6245 - val_loss: 2.0499 - val_accuracy: 0.6336\n",
      "Epoch 112/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.9172 - accuracy: 0.6231 - val_loss: 2.0522 - val_accuracy: 0.6406\n",
      "Epoch 113/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.9261 - accuracy: 0.6199 - val_loss: 2.0486 - val_accuracy: 0.6440\n",
      "Epoch 114/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.9082 - accuracy: 0.6261 - val_loss: 2.0308 - val_accuracy: 0.6509\n",
      "Epoch 115/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.8929 - accuracy: 0.6336 - val_loss: 2.0368 - val_accuracy: 0.6509\n",
      "Epoch 116/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.9028 - accuracy: 0.6289 - val_loss: 2.0373 - val_accuracy: 0.6359\n",
      "Epoch 117/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.8943 - accuracy: 0.6279 - val_loss: 2.0260 - val_accuracy: 0.6440\n",
      "Epoch 118/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.8842 - accuracy: 0.6350 - val_loss: 2.0303 - val_accuracy: 0.6463\n",
      "Epoch 119/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.8893 - accuracy: 0.6362 - val_loss: 2.0320 - val_accuracy: 0.6394\n",
      "Epoch 120/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.8860 - accuracy: 0.6277 - val_loss: 2.0098 - val_accuracy: 0.6463\n",
      "Epoch 121/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.8774 - accuracy: 0.6371 - val_loss: 2.0149 - val_accuracy: 0.6417\n",
      "Epoch 122/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.8679 - accuracy: 0.6353 - val_loss: 2.0143 - val_accuracy: 0.6394\n",
      "Epoch 123/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.8809 - accuracy: 0.6363 - val_loss: 2.0120 - val_accuracy: 0.6509\n",
      "Epoch 124/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.8663 - accuracy: 0.6431 - val_loss: 2.0185 - val_accuracy: 0.6452\n",
      "Epoch 125/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.8459 - accuracy: 0.6427 - val_loss: 2.0012 - val_accuracy: 0.6452\n",
      "Epoch 126/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.8564 - accuracy: 0.6364 - val_loss: 2.0034 - val_accuracy: 0.6440\n",
      "Epoch 127/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.8481 - accuracy: 0.6431 - val_loss: 2.0115 - val_accuracy: 0.6509\n",
      "Epoch 128/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.8458 - accuracy: 0.6431 - val_loss: 2.0019 - val_accuracy: 0.6486\n",
      "Epoch 129/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.8291 - accuracy: 0.6471 - val_loss: 2.0321 - val_accuracy: 0.6313\n",
      "Epoch 130/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.8264 - accuracy: 0.6494 - val_loss: 1.9913 - val_accuracy: 0.6452\n",
      "Epoch 131/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.8298 - accuracy: 0.6480 - val_loss: 1.9904 - val_accuracy: 0.6567\n",
      "Epoch 132/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.8255 - accuracy: 0.6517 - val_loss: 2.0046 - val_accuracy: 0.6521\n",
      "Epoch 133/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.8295 - accuracy: 0.6509 - val_loss: 1.9933 - val_accuracy: 0.6452\n",
      "Epoch 134/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.8152 - accuracy: 0.6510 - val_loss: 1.9998 - val_accuracy: 0.6429\n",
      "Epoch 135/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.8026 - accuracy: 0.6545 - val_loss: 2.0164 - val_accuracy: 0.6394\n",
      "Epoch 136/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.8072 - accuracy: 0.6509 - val_loss: 1.9953 - val_accuracy: 0.6509\n",
      "Epoch 137/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.7870 - accuracy: 0.6609 - val_loss: 1.9808 - val_accuracy: 0.6509\n",
      "Epoch 138/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.8027 - accuracy: 0.6560 - val_loss: 1.9882 - val_accuracy: 0.6486\n",
      "Epoch 139/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.7786 - accuracy: 0.6596 - val_loss: 1.9911 - val_accuracy: 0.6486\n",
      "Epoch 140/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.7796 - accuracy: 0.6617 - val_loss: 1.9859 - val_accuracy: 0.6498\n",
      "Epoch 141/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.7768 - accuracy: 0.6608 - val_loss: 1.9933 - val_accuracy: 0.6521\n",
      "Epoch 142/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.7750 - accuracy: 0.6642 - val_loss: 1.9817 - val_accuracy: 0.6486\n",
      "Epoch 143/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.7672 - accuracy: 0.6638 - val_loss: 1.9822 - val_accuracy: 0.6452\n",
      "Epoch 144/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.7591 - accuracy: 0.6642 - val_loss: 1.9902 - val_accuracy: 0.6521\n",
      "Epoch 145/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.7637 - accuracy: 0.6659 - val_loss: 1.9889 - val_accuracy: 0.6440\n",
      "Epoch 146/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.7544 - accuracy: 0.6671 - val_loss: 1.9984 - val_accuracy: 0.6521\n",
      "Epoch 147/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.7488 - accuracy: 0.6713 - val_loss: 1.9902 - val_accuracy: 0.6486\n",
      "Epoch 148/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.7440 - accuracy: 0.6700 - val_loss: 1.9659 - val_accuracy: 0.6567\n",
      "Epoch 149/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.7430 - accuracy: 0.6742 - val_loss: 1.9773 - val_accuracy: 0.6532\n",
      "Epoch 150/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.7494 - accuracy: 0.6691 - val_loss: 1.9855 - val_accuracy: 0.6509\n",
      "Epoch 151/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.7575 - accuracy: 0.6687 - val_loss: 1.9788 - val_accuracy: 0.6463\n",
      "Epoch 152/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.7478 - accuracy: 0.6694 - val_loss: 1.9739 - val_accuracy: 0.6521\n",
      "Epoch 153/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.7491 - accuracy: 0.6737 - val_loss: 1.9835 - val_accuracy: 0.6475\n",
      "Epoch 154/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.7340 - accuracy: 0.6751 - val_loss: 1.9897 - val_accuracy: 0.6590\n",
      "Epoch 155/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.7275 - accuracy: 0.6783 - val_loss: 1.9600 - val_accuracy: 0.6555\n",
      "Epoch 156/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.7346 - accuracy: 0.6760 - val_loss: 1.9749 - val_accuracy: 0.6567\n",
      "Epoch 157/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.7194 - accuracy: 0.6793 - val_loss: 1.9628 - val_accuracy: 0.6521\n",
      "Epoch 158/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.7127 - accuracy: 0.6781 - val_loss: 1.9435 - val_accuracy: 0.6567\n",
      "Epoch 159/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.7100 - accuracy: 0.6827 - val_loss: 1.9706 - val_accuracy: 0.6544\n",
      "Epoch 160/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.7130 - accuracy: 0.6779 - val_loss: 1.9615 - val_accuracy: 0.6578\n",
      "Epoch 161/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.6988 - accuracy: 0.6846 - val_loss: 1.9592 - val_accuracy: 0.6647\n",
      "Epoch 162/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.6928 - accuracy: 0.6916 - val_loss: 1.9664 - val_accuracy: 0.6647\n",
      "Epoch 163/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.6917 - accuracy: 0.6863 - val_loss: 1.9679 - val_accuracy: 0.6544\n",
      "Epoch 164/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.7123 - accuracy: 0.6819 - val_loss: 1.9651 - val_accuracy: 0.6613\n",
      "Epoch 165/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.7012 - accuracy: 0.6819 - val_loss: 1.9642 - val_accuracy: 0.6636\n",
      "Epoch 166/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.6800 - accuracy: 0.6928 - val_loss: 1.9631 - val_accuracy: 0.6624\n",
      "Epoch 167/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.6951 - accuracy: 0.6895 - val_loss: 1.9677 - val_accuracy: 0.6567\n",
      "Epoch 168/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.6880 - accuracy: 0.6902 - val_loss: 1.9791 - val_accuracy: 0.6475\n",
      "Epoch 169/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.6911 - accuracy: 0.6886 - val_loss: 1.9585 - val_accuracy: 0.6578\n",
      "Epoch 170/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.6697 - accuracy: 0.6905 - val_loss: 1.9678 - val_accuracy: 0.6624\n",
      "Epoch 171/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.6662 - accuracy: 0.6911 - val_loss: 1.9790 - val_accuracy: 0.6578\n",
      "Epoch 172/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.6602 - accuracy: 0.6936 - val_loss: 1.9812 - val_accuracy: 0.6578\n",
      "Epoch 173/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.6677 - accuracy: 0.6905 - val_loss: 1.9579 - val_accuracy: 0.6590\n",
      "Epoch 174/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.6665 - accuracy: 0.6886 - val_loss: 1.9628 - val_accuracy: 0.6578\n",
      "Epoch 175/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.6705 - accuracy: 0.6897 - val_loss: 1.9667 - val_accuracy: 0.6613\n",
      "Epoch 176/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.6538 - accuracy: 0.6932 - val_loss: 1.9473 - val_accuracy: 0.6567\n",
      "Epoch 177/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.6343 - accuracy: 0.7028 - val_loss: 1.9582 - val_accuracy: 0.6578\n",
      "Epoch 178/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.6509 - accuracy: 0.6962 - val_loss: 1.9531 - val_accuracy: 0.6682\n",
      "Epoch 179/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.6362 - accuracy: 0.6998 - val_loss: 1.9514 - val_accuracy: 0.6567\n",
      "Epoch 180/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.6428 - accuracy: 0.6982 - val_loss: 1.9603 - val_accuracy: 0.6636\n",
      "Epoch 181/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.6478 - accuracy: 0.7000 - val_loss: 1.9517 - val_accuracy: 0.6624\n",
      "Epoch 182/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.6184 - accuracy: 0.7047 - val_loss: 1.9633 - val_accuracy: 0.6647\n",
      "Epoch 183/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.6309 - accuracy: 0.7038 - val_loss: 1.9627 - val_accuracy: 0.6682\n",
      "Epoch 184/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.6238 - accuracy: 0.7075 - val_loss: 1.9387 - val_accuracy: 0.6682\n",
      "Epoch 185/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.6325 - accuracy: 0.7009 - val_loss: 1.9669 - val_accuracy: 0.6659\n",
      "Epoch 186/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.6104 - accuracy: 0.7110 - val_loss: 1.9461 - val_accuracy: 0.6671\n",
      "Epoch 187/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.6311 - accuracy: 0.7042 - val_loss: 1.9414 - val_accuracy: 0.6659\n",
      "Epoch 188/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.6107 - accuracy: 0.7088 - val_loss: 1.9557 - val_accuracy: 0.6601\n",
      "Epoch 189/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.6314 - accuracy: 0.7024 - val_loss: 1.9593 - val_accuracy: 0.6601\n",
      "Epoch 190/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.6131 - accuracy: 0.7030 - val_loss: 1.9639 - val_accuracy: 0.6636\n",
      "Epoch 191/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.6109 - accuracy: 0.7083 - val_loss: 1.9493 - val_accuracy: 0.6636\n",
      "Epoch 192/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.5787 - accuracy: 0.7189 - val_loss: 1.9569 - val_accuracy: 0.6624\n",
      "Epoch 193/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.6167 - accuracy: 0.7098 - val_loss: 1.9525 - val_accuracy: 0.6636\n",
      "Epoch 194/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.6078 - accuracy: 0.7153 - val_loss: 1.9458 - val_accuracy: 0.6682\n",
      "Epoch 195/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5819 - accuracy: 0.7197 - val_loss: 1.9435 - val_accuracy: 0.6671\n",
      "Epoch 196/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5972 - accuracy: 0.7111 - val_loss: 1.9515 - val_accuracy: 0.6717\n",
      "Epoch 197/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5846 - accuracy: 0.7170 - val_loss: 1.9546 - val_accuracy: 0.6682\n",
      "Epoch 198/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5832 - accuracy: 0.7184 - val_loss: 1.9450 - val_accuracy: 0.6624\n",
      "Epoch 199/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5829 - accuracy: 0.7192 - val_loss: 1.9288 - val_accuracy: 0.6717\n",
      "Epoch 200/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.5763 - accuracy: 0.7181 - val_loss: 1.9459 - val_accuracy: 0.6590\n",
      "Epoch 201/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5838 - accuracy: 0.7162 - val_loss: 1.9430 - val_accuracy: 0.6682\n",
      "Epoch 202/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5609 - accuracy: 0.7284 - val_loss: 1.9407 - val_accuracy: 0.6774\n",
      "Epoch 203/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5713 - accuracy: 0.7174 - val_loss: 1.9350 - val_accuracy: 0.6774\n",
      "Epoch 204/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5751 - accuracy: 0.7244 - val_loss: 1.9619 - val_accuracy: 0.6728\n",
      "Epoch 205/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.5691 - accuracy: 0.7247 - val_loss: 1.9395 - val_accuracy: 0.6705\n",
      "Epoch 206/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5646 - accuracy: 0.7222 - val_loss: 1.9396 - val_accuracy: 0.6705\n",
      "Epoch 207/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.5685 - accuracy: 0.7247 - val_loss: 1.9407 - val_accuracy: 0.6717\n",
      "Epoch 208/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.5603 - accuracy: 0.7292 - val_loss: 1.9382 - val_accuracy: 0.6694\n",
      "Epoch 209/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.5672 - accuracy: 0.7256 - val_loss: 1.9465 - val_accuracy: 0.6751\n",
      "Epoch 210/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5424 - accuracy: 0.7372 - val_loss: 1.9474 - val_accuracy: 0.6809\n",
      "Epoch 211/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5537 - accuracy: 0.7258 - val_loss: 1.9318 - val_accuracy: 0.6786\n",
      "Epoch 212/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5443 - accuracy: 0.7320 - val_loss: 1.9331 - val_accuracy: 0.6728\n",
      "Epoch 213/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.5445 - accuracy: 0.7321 - val_loss: 1.9232 - val_accuracy: 0.6694\n",
      "Epoch 214/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5464 - accuracy: 0.7265 - val_loss: 1.9253 - val_accuracy: 0.6671\n",
      "Epoch 215/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5397 - accuracy: 0.7327 - val_loss: 1.9528 - val_accuracy: 0.6728\n",
      "Epoch 216/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5305 - accuracy: 0.7386 - val_loss: 1.9528 - val_accuracy: 0.6728\n",
      "Epoch 217/700\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5272 - accuracy: 0.7409 - val_loss: 1.9320 - val_accuracy: 0.6763\n",
      "Epoch 218/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5374 - accuracy: 0.7353 - val_loss: 1.9471 - val_accuracy: 0.6774\n",
      "Epoch 219/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5275 - accuracy: 0.7361 - val_loss: 1.9611 - val_accuracy: 0.6763\n",
      "Epoch 220/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5273 - accuracy: 0.7336 - val_loss: 1.9453 - val_accuracy: 0.6682\n",
      "Epoch 221/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5256 - accuracy: 0.7394 - val_loss: 1.9430 - val_accuracy: 0.6740\n",
      "Epoch 222/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5161 - accuracy: 0.7417 - val_loss: 1.9430 - val_accuracy: 0.6694\n",
      "Epoch 223/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5178 - accuracy: 0.7413 - val_loss: 1.9429 - val_accuracy: 0.6740\n",
      "Epoch 224/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.5162 - accuracy: 0.7420 - val_loss: 1.9619 - val_accuracy: 0.6659\n",
      "Epoch 225/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5029 - accuracy: 0.7422 - val_loss: 1.9447 - val_accuracy: 0.6717\n",
      "Epoch 226/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5139 - accuracy: 0.7398 - val_loss: 1.9386 - val_accuracy: 0.6682\n",
      "Epoch 227/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.4901 - accuracy: 0.7464 - val_loss: 1.9437 - val_accuracy: 0.6728\n",
      "Epoch 228/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.5060 - accuracy: 0.7390 - val_loss: 1.9456 - val_accuracy: 0.6717\n",
      "Epoch 229/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4955 - accuracy: 0.7473 - val_loss: 1.9512 - val_accuracy: 0.6728\n",
      "Epoch 230/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.5151 - accuracy: 0.7404 - val_loss: 1.9561 - val_accuracy: 0.6694\n",
      "Epoch 231/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.5171 - accuracy: 0.7417 - val_loss: 1.9481 - val_accuracy: 0.6728\n",
      "Epoch 232/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4855 - accuracy: 0.7511 - val_loss: 1.9520 - val_accuracy: 0.6740\n",
      "Epoch 233/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.5095 - accuracy: 0.7395 - val_loss: 1.9389 - val_accuracy: 0.6705\n",
      "Epoch 234/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.5105 - accuracy: 0.7398 - val_loss: 1.9501 - val_accuracy: 0.6705\n",
      "Epoch 235/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4808 - accuracy: 0.7498 - val_loss: 1.9409 - val_accuracy: 0.6728\n",
      "Epoch 236/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4983 - accuracy: 0.7450 - val_loss: 1.9547 - val_accuracy: 0.6797\n",
      "Epoch 237/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4643 - accuracy: 0.7543 - val_loss: 1.9591 - val_accuracy: 0.6717\n",
      "Epoch 238/700\n",
      "7809/7809 [==============================] - 1s 124us/sample - loss: 1.4858 - accuracy: 0.7508 - val_loss: 1.9521 - val_accuracy: 0.6774\n",
      "Epoch 239/700\n",
      "7809/7809 [==============================] - 1s 124us/sample - loss: 1.4845 - accuracy: 0.7464 - val_loss: 1.9386 - val_accuracy: 0.6717\n",
      "Epoch 240/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4832 - accuracy: 0.7473 - val_loss: 1.9518 - val_accuracy: 0.6717\n",
      "Epoch 241/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4812 - accuracy: 0.7503 - val_loss: 1.9674 - val_accuracy: 0.6717\n",
      "Epoch 242/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.4865 - accuracy: 0.7531 - val_loss: 1.9360 - val_accuracy: 0.6694\n",
      "Epoch 243/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4651 - accuracy: 0.7505 - val_loss: 1.9573 - val_accuracy: 0.6809\n",
      "Epoch 244/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.4536 - accuracy: 0.7589 - val_loss: 1.9460 - val_accuracy: 0.6763\n",
      "Epoch 245/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4746 - accuracy: 0.7540 - val_loss: 1.9460 - val_accuracy: 0.6797\n",
      "Epoch 246/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4716 - accuracy: 0.7578 - val_loss: 1.9386 - val_accuracy: 0.6751\n",
      "Epoch 247/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.4630 - accuracy: 0.7571 - val_loss: 1.9606 - val_accuracy: 0.6682\n",
      "Epoch 248/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4689 - accuracy: 0.7567 - val_loss: 1.9597 - val_accuracy: 0.6786\n",
      "Epoch 249/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4467 - accuracy: 0.7610 - val_loss: 1.9522 - val_accuracy: 0.6751\n",
      "Epoch 250/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4505 - accuracy: 0.7613 - val_loss: 1.9532 - val_accuracy: 0.6786\n",
      "Epoch 251/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4335 - accuracy: 0.7637 - val_loss: 1.9486 - val_accuracy: 0.6763\n",
      "Epoch 252/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4429 - accuracy: 0.7682 - val_loss: 1.9338 - val_accuracy: 0.6843\n",
      "Epoch 253/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4613 - accuracy: 0.7594 - val_loss: 1.9497 - val_accuracy: 0.6786\n",
      "Epoch 254/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.4568 - accuracy: 0.7593 - val_loss: 1.9424 - val_accuracy: 0.6774\n",
      "Epoch 255/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.4508 - accuracy: 0.7569 - val_loss: 1.9668 - val_accuracy: 0.6809\n",
      "Epoch 256/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4437 - accuracy: 0.7664 - val_loss: 1.9633 - val_accuracy: 0.6763\n",
      "Epoch 257/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.4364 - accuracy: 0.7669 - val_loss: 1.9603 - val_accuracy: 0.6786\n",
      "Epoch 258/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.4465 - accuracy: 0.7616 - val_loss: 1.9692 - val_accuracy: 0.6832\n",
      "Epoch 259/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4400 - accuracy: 0.7650 - val_loss: 1.9549 - val_accuracy: 0.6820\n",
      "Epoch 260/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4214 - accuracy: 0.7659 - val_loss: 1.9501 - val_accuracy: 0.6809\n",
      "Epoch 261/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4389 - accuracy: 0.7682 - val_loss: 1.9497 - val_accuracy: 0.6889\n",
      "Epoch 262/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4105 - accuracy: 0.7718 - val_loss: 1.9614 - val_accuracy: 0.6786\n",
      "Epoch 263/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4350 - accuracy: 0.7683 - val_loss: 1.9603 - val_accuracy: 0.6866\n",
      "Epoch 264/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4311 - accuracy: 0.7672 - val_loss: 1.9511 - val_accuracy: 0.6878\n",
      "Epoch 265/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4328 - accuracy: 0.7650 - val_loss: 1.9665 - val_accuracy: 0.6843\n",
      "Epoch 266/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4264 - accuracy: 0.7678 - val_loss: 1.9499 - val_accuracy: 0.6855\n",
      "Epoch 267/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.4208 - accuracy: 0.7695 - val_loss: 1.9562 - val_accuracy: 0.6901\n",
      "Epoch 268/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.4231 - accuracy: 0.7680 - val_loss: 1.9594 - val_accuracy: 0.6855\n",
      "Epoch 269/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4296 - accuracy: 0.7680 - val_loss: 1.9539 - val_accuracy: 0.6855\n",
      "Epoch 270/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.4069 - accuracy: 0.7758 - val_loss: 1.9637 - val_accuracy: 0.6820\n",
      "Epoch 271/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4084 - accuracy: 0.7726 - val_loss: 1.9675 - val_accuracy: 0.6889\n",
      "Epoch 272/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.4041 - accuracy: 0.7787 - val_loss: 1.9652 - val_accuracy: 0.6843\n",
      "Epoch 273/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4096 - accuracy: 0.7760 - val_loss: 1.9499 - val_accuracy: 0.6809\n",
      "Epoch 274/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.4082 - accuracy: 0.7742 - val_loss: 1.9586 - val_accuracy: 0.6855\n",
      "Epoch 275/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.4027 - accuracy: 0.7728 - val_loss: 1.9559 - val_accuracy: 0.6843\n",
      "Epoch 276/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3999 - accuracy: 0.7781 - val_loss: 1.9672 - val_accuracy: 0.6809\n",
      "Epoch 277/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.4183 - accuracy: 0.7710 - val_loss: 1.9555 - val_accuracy: 0.6889\n",
      "Epoch 278/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3896 - accuracy: 0.7805 - val_loss: 1.9643 - val_accuracy: 0.6912\n",
      "Epoch 279/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.4021 - accuracy: 0.7806 - val_loss: 1.9463 - val_accuracy: 0.6959\n",
      "Epoch 280/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3858 - accuracy: 0.7831 - val_loss: 1.9579 - val_accuracy: 0.6924\n",
      "Epoch 281/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3929 - accuracy: 0.7794 - val_loss: 1.9508 - val_accuracy: 0.6786\n",
      "Epoch 282/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3993 - accuracy: 0.7811 - val_loss: 1.9608 - val_accuracy: 0.6866\n",
      "Epoch 283/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3902 - accuracy: 0.7808 - val_loss: 1.9580 - val_accuracy: 0.6935\n",
      "Epoch 284/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3897 - accuracy: 0.7836 - val_loss: 1.9609 - val_accuracy: 0.6786\n",
      "Epoch 285/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.4087 - accuracy: 0.7783 - val_loss: 1.9627 - val_accuracy: 0.6786\n",
      "Epoch 286/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3838 - accuracy: 0.7860 - val_loss: 1.9637 - val_accuracy: 0.6855\n",
      "Epoch 287/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3947 - accuracy: 0.7787 - val_loss: 1.9643 - val_accuracy: 0.6912\n",
      "Epoch 288/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3827 - accuracy: 0.7859 - val_loss: 1.9602 - val_accuracy: 0.6901\n",
      "Epoch 289/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3794 - accuracy: 0.7852 - val_loss: 1.9723 - val_accuracy: 0.6878\n",
      "Epoch 290/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3753 - accuracy: 0.7887 - val_loss: 1.9743 - val_accuracy: 0.6855\n",
      "Epoch 291/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3842 - accuracy: 0.7837 - val_loss: 1.9592 - val_accuracy: 0.6912\n",
      "Epoch 292/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3822 - accuracy: 0.7856 - val_loss: 1.9708 - val_accuracy: 0.6982\n",
      "Epoch 293/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3774 - accuracy: 0.7849 - val_loss: 1.9582 - val_accuracy: 0.6889\n",
      "Epoch 294/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3768 - accuracy: 0.7883 - val_loss: 1.9661 - val_accuracy: 0.6843\n",
      "Epoch 295/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.3839 - accuracy: 0.7858 - val_loss: 1.9508 - val_accuracy: 0.6912\n",
      "Epoch 296/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3931 - accuracy: 0.7795 - val_loss: 1.9708 - val_accuracy: 0.6866\n",
      "Epoch 297/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3590 - accuracy: 0.7945 - val_loss: 1.9692 - val_accuracy: 0.6866\n",
      "Epoch 298/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3592 - accuracy: 0.7924 - val_loss: 1.9673 - val_accuracy: 0.6820\n",
      "Epoch 299/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3695 - accuracy: 0.7869 - val_loss: 1.9757 - val_accuracy: 0.6855\n",
      "Epoch 300/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3688 - accuracy: 0.7951 - val_loss: 1.9597 - val_accuracy: 0.6878\n",
      "Epoch 301/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3654 - accuracy: 0.7934 - val_loss: 1.9616 - val_accuracy: 0.6901\n",
      "Epoch 302/700\n",
      "7809/7809 [==============================] - 1s 124us/sample - loss: 1.3707 - accuracy: 0.7955 - val_loss: 1.9625 - val_accuracy: 0.6889\n",
      "Epoch 303/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3658 - accuracy: 0.7922 - val_loss: 1.9856 - val_accuracy: 0.6924\n",
      "Epoch 304/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3546 - accuracy: 0.7933 - val_loss: 1.9719 - val_accuracy: 0.6959\n",
      "Epoch 305/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3548 - accuracy: 0.7977 - val_loss: 1.9702 - val_accuracy: 0.6786\n",
      "Epoch 306/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3536 - accuracy: 0.7959 - val_loss: 1.9821 - val_accuracy: 0.6912\n",
      "Epoch 307/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3546 - accuracy: 0.7960 - val_loss: 1.9764 - val_accuracy: 0.6924\n",
      "Epoch 308/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3590 - accuracy: 0.7909 - val_loss: 1.9804 - val_accuracy: 0.6935\n",
      "Epoch 309/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3586 - accuracy: 0.7932 - val_loss: 1.9873 - val_accuracy: 0.6878\n",
      "Epoch 310/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3401 - accuracy: 0.7987 - val_loss: 1.9704 - val_accuracy: 0.6959\n",
      "Epoch 311/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3595 - accuracy: 0.7906 - val_loss: 1.9707 - val_accuracy: 0.6878\n",
      "Epoch 312/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3401 - accuracy: 0.8016 - val_loss: 1.9728 - val_accuracy: 0.6912\n",
      "Epoch 313/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3514 - accuracy: 0.7963 - val_loss: 1.9601 - val_accuracy: 0.6959\n",
      "Epoch 314/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3501 - accuracy: 0.8004 - val_loss: 1.9733 - val_accuracy: 0.6947\n",
      "Epoch 315/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3442 - accuracy: 0.8018 - val_loss: 1.9910 - val_accuracy: 0.6832\n",
      "Epoch 316/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3560 - accuracy: 0.7969 - val_loss: 1.9758 - val_accuracy: 0.6935\n",
      "Epoch 317/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3533 - accuracy: 0.7960 - val_loss: 1.9808 - val_accuracy: 0.6889\n",
      "Epoch 318/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3491 - accuracy: 0.7979 - val_loss: 1.9699 - val_accuracy: 0.6924\n",
      "Epoch 319/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.3415 - accuracy: 0.8007 - val_loss: 1.9820 - val_accuracy: 0.6959\n",
      "Epoch 320/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3530 - accuracy: 0.7983 - val_loss: 1.9667 - val_accuracy: 0.6866\n",
      "Epoch 321/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3421 - accuracy: 0.7998 - val_loss: 1.9741 - val_accuracy: 0.6935\n",
      "Epoch 322/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3376 - accuracy: 0.8027 - val_loss: 1.9650 - val_accuracy: 0.6866\n",
      "Epoch 323/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3238 - accuracy: 0.8011 - val_loss: 1.9713 - val_accuracy: 0.6878\n",
      "Epoch 324/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3252 - accuracy: 0.8083 - val_loss: 1.9963 - val_accuracy: 0.6855\n",
      "Epoch 325/700\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3338 - accuracy: 0.8054 - val_loss: 1.9918 - val_accuracy: 0.6959\n",
      "Epoch 326/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3193 - accuracy: 0.8059 - val_loss: 1.9857 - val_accuracy: 0.6901\n",
      "Epoch 327/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3335 - accuracy: 0.8038 - val_loss: 1.9792 - val_accuracy: 0.6935\n",
      "Epoch 328/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3341 - accuracy: 0.8078 - val_loss: 1.9693 - val_accuracy: 0.6901\n",
      "Epoch 329/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3265 - accuracy: 0.8092 - val_loss: 1.9992 - val_accuracy: 0.6855\n",
      "Epoch 330/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3238 - accuracy: 0.8052 - val_loss: 1.9895 - val_accuracy: 0.6959\n",
      "Epoch 331/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3239 - accuracy: 0.8091 - val_loss: 1.9814 - val_accuracy: 0.6878\n",
      "Epoch 332/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3269 - accuracy: 0.8078 - val_loss: 1.9859 - val_accuracy: 0.6889\n",
      "Epoch 333/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3269 - accuracy: 0.8080 - val_loss: 1.9773 - val_accuracy: 0.6947\n",
      "Epoch 334/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3206 - accuracy: 0.8083 - val_loss: 1.9855 - val_accuracy: 0.6912\n",
      "Epoch 335/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3210 - accuracy: 0.8032 - val_loss: 1.9844 - val_accuracy: 0.6878\n",
      "Epoch 336/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3117 - accuracy: 0.8111 - val_loss: 1.9851 - val_accuracy: 0.6959\n",
      "Epoch 337/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3177 - accuracy: 0.8065 - val_loss: 1.9966 - val_accuracy: 0.6901\n",
      "Epoch 338/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3291 - accuracy: 0.8074 - val_loss: 1.9945 - val_accuracy: 0.6832\n",
      "Epoch 339/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3241 - accuracy: 0.8070 - val_loss: 1.9800 - val_accuracy: 0.6901\n",
      "Epoch 340/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3155 - accuracy: 0.8115 - val_loss: 1.9769 - val_accuracy: 0.6947\n",
      "Epoch 341/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3178 - accuracy: 0.8128 - val_loss: 1.9717 - val_accuracy: 0.6901\n",
      "Epoch 342/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3019 - accuracy: 0.8174 - val_loss: 1.9906 - val_accuracy: 0.6889\n",
      "Epoch 343/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3156 - accuracy: 0.8110 - val_loss: 1.9795 - val_accuracy: 0.6901\n",
      "Epoch 344/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3313 - accuracy: 0.8070 - val_loss: 1.9762 - val_accuracy: 0.6912\n",
      "Epoch 345/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3106 - accuracy: 0.8080 - val_loss: 1.9734 - val_accuracy: 0.6947\n",
      "Epoch 346/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3090 - accuracy: 0.8164 - val_loss: 1.9836 - val_accuracy: 0.6947\n",
      "Epoch 347/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3052 - accuracy: 0.8174 - val_loss: 1.9916 - val_accuracy: 0.6947\n",
      "Epoch 348/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.2926 - accuracy: 0.8198 - val_loss: 2.0029 - val_accuracy: 0.6924\n",
      "Epoch 349/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3079 - accuracy: 0.8141 - val_loss: 2.0031 - val_accuracy: 0.6901\n",
      "Epoch 350/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3048 - accuracy: 0.8173 - val_loss: 1.9757 - val_accuracy: 0.6924\n",
      "Epoch 351/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3029 - accuracy: 0.8133 - val_loss: 1.9606 - val_accuracy: 0.7016\n",
      "Epoch 352/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.2911 - accuracy: 0.8234 - val_loss: 1.9887 - val_accuracy: 0.6970\n",
      "Epoch 353/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.2892 - accuracy: 0.8229 - val_loss: 1.9849 - val_accuracy: 0.6993\n",
      "Epoch 354/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3083 - accuracy: 0.8144 - val_loss: 1.9879 - val_accuracy: 0.6889\n",
      "Epoch 355/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3087 - accuracy: 0.8175 - val_loss: 1.9877 - val_accuracy: 0.6993\n",
      "Epoch 356/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3032 - accuracy: 0.8189 - val_loss: 1.9955 - val_accuracy: 0.6912\n",
      "Epoch 357/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.2791 - accuracy: 0.8257 - val_loss: 1.9934 - val_accuracy: 0.6832\n",
      "Epoch 358/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2998 - accuracy: 0.8147 - val_loss: 1.9901 - val_accuracy: 0.6878\n",
      "Epoch 359/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.3036 - accuracy: 0.8157 - val_loss: 1.9911 - val_accuracy: 0.6924\n",
      "Epoch 360/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.2962 - accuracy: 0.8223 - val_loss: 2.0099 - val_accuracy: 0.6832\n",
      "Epoch 361/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3035 - accuracy: 0.8155 - val_loss: 1.9877 - val_accuracy: 0.6982\n",
      "Epoch 362/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.3031 - accuracy: 0.8167 - val_loss: 2.0066 - val_accuracy: 0.6866\n",
      "Epoch 363/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.2989 - accuracy: 0.8153 - val_loss: 1.9739 - val_accuracy: 0.6970\n",
      "Epoch 364/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.2882 - accuracy: 0.8242 - val_loss: 1.9856 - val_accuracy: 0.7039\n",
      "Epoch 365/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2934 - accuracy: 0.8198 - val_loss: 2.0047 - val_accuracy: 0.6982\n",
      "Epoch 366/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2961 - accuracy: 0.8184 - val_loss: 1.9934 - val_accuracy: 0.6912\n",
      "Epoch 367/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2872 - accuracy: 0.8183 - val_loss: 1.9775 - val_accuracy: 0.6959\n",
      "Epoch 368/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.2713 - accuracy: 0.8305 - val_loss: 1.9955 - val_accuracy: 0.6947\n",
      "Epoch 369/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.2953 - accuracy: 0.8211 - val_loss: 1.9900 - val_accuracy: 0.6843\n",
      "Epoch 370/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2723 - accuracy: 0.8247 - val_loss: 1.9858 - val_accuracy: 0.6935\n",
      "Epoch 371/700\n",
      "7809/7809 [==============================] - 1s 124us/sample - loss: 1.2755 - accuracy: 0.8224 - val_loss: 1.9817 - val_accuracy: 0.6970\n",
      "Epoch 372/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2777 - accuracy: 0.8200 - val_loss: 1.9798 - val_accuracy: 0.6993\n",
      "Epoch 373/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2740 - accuracy: 0.8247 - val_loss: 2.0014 - val_accuracy: 0.6935\n",
      "Epoch 374/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2891 - accuracy: 0.8223 - val_loss: 1.9915 - val_accuracy: 0.6889\n",
      "Epoch 375/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2698 - accuracy: 0.8308 - val_loss: 1.9855 - val_accuracy: 0.7039\n",
      "Epoch 376/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2598 - accuracy: 0.8311 - val_loss: 1.9881 - val_accuracy: 0.6912\n",
      "Epoch 377/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2745 - accuracy: 0.8271 - val_loss: 1.9901 - val_accuracy: 0.6855\n",
      "Epoch 378/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2765 - accuracy: 0.8238 - val_loss: 1.9900 - val_accuracy: 0.7005\n",
      "Epoch 379/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2819 - accuracy: 0.8223 - val_loss: 1.9945 - val_accuracy: 0.7005\n",
      "Epoch 380/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2667 - accuracy: 0.8297 - val_loss: 1.9982 - val_accuracy: 0.6970\n",
      "Epoch 381/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2722 - accuracy: 0.8182 - val_loss: 1.9908 - val_accuracy: 0.6993\n",
      "Epoch 382/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2686 - accuracy: 0.8328 - val_loss: 1.9916 - val_accuracy: 0.6959\n",
      "Epoch 383/700\n",
      "7809/7809 [==============================] - 1s 120us/sample - loss: 1.2610 - accuracy: 0.8305 - val_loss: 1.9915 - val_accuracy: 0.6947\n",
      "Epoch 384/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2627 - accuracy: 0.8288 - val_loss: 1.9955 - val_accuracy: 0.7005\n",
      "Epoch 385/700\n",
      "7809/7809 [==============================] - 1s 120us/sample - loss: 1.2579 - accuracy: 0.8337 - val_loss: 1.9885 - val_accuracy: 0.7028\n",
      "Epoch 386/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2572 - accuracy: 0.8307 - val_loss: 1.9906 - val_accuracy: 0.6959\n",
      "Epoch 387/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2627 - accuracy: 0.8315 - val_loss: 2.0087 - val_accuracy: 0.7085\n",
      "Epoch 388/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2635 - accuracy: 0.8301 - val_loss: 1.9917 - val_accuracy: 0.7005\n",
      "Epoch 389/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2574 - accuracy: 0.8317 - val_loss: 2.0141 - val_accuracy: 0.6935\n",
      "Epoch 390/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2724 - accuracy: 0.8337 - val_loss: 2.0029 - val_accuracy: 0.6959\n",
      "Epoch 391/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2475 - accuracy: 0.8324 - val_loss: 2.0059 - val_accuracy: 0.6982\n",
      "Epoch 392/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2624 - accuracy: 0.8283 - val_loss: 2.0267 - val_accuracy: 0.6947\n",
      "Epoch 393/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2631 - accuracy: 0.8287 - val_loss: 2.0045 - val_accuracy: 0.6959\n",
      "Epoch 394/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2505 - accuracy: 0.8366 - val_loss: 1.9936 - val_accuracy: 0.6982\n",
      "Epoch 395/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2519 - accuracy: 0.8329 - val_loss: 2.0079 - val_accuracy: 0.6912\n",
      "Epoch 396/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2481 - accuracy: 0.8357 - val_loss: 2.0016 - val_accuracy: 0.6959\n",
      "Epoch 397/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2606 - accuracy: 0.8316 - val_loss: 2.0085 - val_accuracy: 0.7005\n",
      "Epoch 398/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2452 - accuracy: 0.8363 - val_loss: 2.0003 - val_accuracy: 0.6889\n",
      "Epoch 399/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2579 - accuracy: 0.8299 - val_loss: 1.9820 - val_accuracy: 0.6970\n",
      "Epoch 400/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2662 - accuracy: 0.8302 - val_loss: 1.9922 - val_accuracy: 0.6947\n",
      "Epoch 401/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2649 - accuracy: 0.8331 - val_loss: 1.9985 - val_accuracy: 0.6993\n",
      "Epoch 402/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2552 - accuracy: 0.8331 - val_loss: 2.0002 - val_accuracy: 0.6889\n",
      "Epoch 403/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2448 - accuracy: 0.8375 - val_loss: 2.0016 - val_accuracy: 0.6959\n",
      "Epoch 404/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2513 - accuracy: 0.8362 - val_loss: 1.9871 - val_accuracy: 0.6959\n",
      "Epoch 405/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2501 - accuracy: 0.8374 - val_loss: 2.0027 - val_accuracy: 0.6993\n",
      "Epoch 406/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2491 - accuracy: 0.8413 - val_loss: 2.0123 - val_accuracy: 0.7051\n",
      "Epoch 407/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2441 - accuracy: 0.8369 - val_loss: 1.9990 - val_accuracy: 0.7005\n",
      "Epoch 408/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2596 - accuracy: 0.8329 - val_loss: 2.0119 - val_accuracy: 0.6912\n",
      "Epoch 409/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2161 - accuracy: 0.8453 - val_loss: 2.0073 - val_accuracy: 0.7016\n",
      "Epoch 410/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2420 - accuracy: 0.8436 - val_loss: 2.0017 - val_accuracy: 0.6970\n",
      "Epoch 411/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2522 - accuracy: 0.8383 - val_loss: 2.0257 - val_accuracy: 0.6959\n",
      "Epoch 412/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2288 - accuracy: 0.8421 - val_loss: 2.0079 - val_accuracy: 0.7028\n",
      "Epoch 413/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2474 - accuracy: 0.8379 - val_loss: 2.0251 - val_accuracy: 0.6912\n",
      "Epoch 414/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2427 - accuracy: 0.8383 - val_loss: 2.0114 - val_accuracy: 0.6993\n",
      "Epoch 415/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2446 - accuracy: 0.8398 - val_loss: 2.0095 - val_accuracy: 0.6993\n",
      "Epoch 416/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2438 - accuracy: 0.8395 - val_loss: 2.0004 - val_accuracy: 0.7016\n",
      "Epoch 417/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2484 - accuracy: 0.8379 - val_loss: 2.0146 - val_accuracy: 0.6970\n",
      "Epoch 418/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2411 - accuracy: 0.8389 - val_loss: 2.0195 - val_accuracy: 0.6935\n",
      "Epoch 419/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2276 - accuracy: 0.8422 - val_loss: 2.0313 - val_accuracy: 0.6935\n",
      "Epoch 420/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2314 - accuracy: 0.8420 - val_loss: 2.0242 - val_accuracy: 0.6970\n",
      "Epoch 421/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2482 - accuracy: 0.8376 - val_loss: 2.0186 - val_accuracy: 0.6993\n",
      "Epoch 422/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2491 - accuracy: 0.8369 - val_loss: 2.0138 - val_accuracy: 0.7005\n",
      "Epoch 423/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2540 - accuracy: 0.8361 - val_loss: 2.0249 - val_accuracy: 0.6889\n",
      "Epoch 424/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2331 - accuracy: 0.8413 - val_loss: 2.0124 - val_accuracy: 0.7039\n",
      "Epoch 425/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2347 - accuracy: 0.8447 - val_loss: 1.9983 - val_accuracy: 0.6924\n",
      "Epoch 426/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2162 - accuracy: 0.8451 - val_loss: 2.0017 - val_accuracy: 0.7028\n",
      "Epoch 427/700\n",
      "7809/7809 [==============================] - 1s 120us/sample - loss: 1.2303 - accuracy: 0.8447 - val_loss: 1.9968 - val_accuracy: 0.6982\n",
      "Epoch 428/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2329 - accuracy: 0.8431 - val_loss: 2.0031 - val_accuracy: 0.6993\n",
      "Epoch 429/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2231 - accuracy: 0.8431 - val_loss: 2.0191 - val_accuracy: 0.7016\n",
      "Epoch 430/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2256 - accuracy: 0.8489 - val_loss: 1.9983 - val_accuracy: 0.6947\n",
      "Epoch 431/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2349 - accuracy: 0.8415 - val_loss: 2.0163 - val_accuracy: 0.6901\n",
      "Epoch 432/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2068 - accuracy: 0.8531 - val_loss: 2.0256 - val_accuracy: 0.7039\n",
      "Epoch 433/700\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2362 - accuracy: 0.8415 - val_loss: 2.0204 - val_accuracy: 0.6912\n",
      "Epoch 434/700\n",
      "7809/7809 [==============================] - 1s 120us/sample - loss: 1.2376 - accuracy: 0.8440 - val_loss: 2.0148 - val_accuracy: 0.6912\n",
      "Epoch 435/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2319 - accuracy: 0.8418 - val_loss: 2.0293 - val_accuracy: 0.6970\n",
      "Epoch 436/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2058 - accuracy: 0.8539 - val_loss: 2.0164 - val_accuracy: 0.7028\n",
      "Epoch 437/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2315 - accuracy: 0.8399 - val_loss: 2.0276 - val_accuracy: 0.7097\n",
      "Epoch 438/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2292 - accuracy: 0.8457 - val_loss: 2.0150 - val_accuracy: 0.7005\n",
      "Epoch 439/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2058 - accuracy: 0.8485 - val_loss: 2.0217 - val_accuracy: 0.7051\n",
      "Epoch 440/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2255 - accuracy: 0.8399 - val_loss: 2.0218 - val_accuracy: 0.6993\n",
      "Epoch 441/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2201 - accuracy: 0.8456 - val_loss: 2.0263 - val_accuracy: 0.6947\n",
      "Epoch 442/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2214 - accuracy: 0.8458 - val_loss: 2.0254 - val_accuracy: 0.6959\n",
      "Epoch 443/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2209 - accuracy: 0.8502 - val_loss: 2.0354 - val_accuracy: 0.7016\n",
      "Epoch 444/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2199 - accuracy: 0.8477 - val_loss: 2.0259 - val_accuracy: 0.6901\n",
      "Epoch 445/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2199 - accuracy: 0.8452 - val_loss: 2.0228 - val_accuracy: 0.6901\n",
      "Epoch 446/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.2192 - accuracy: 0.8497 - val_loss: 2.0192 - val_accuracy: 0.6947\n",
      "Epoch 447/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1978 - accuracy: 0.8509 - val_loss: 2.0403 - val_accuracy: 0.6935\n",
      "Epoch 448/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2190 - accuracy: 0.8445 - val_loss: 2.0480 - val_accuracy: 0.6947\n",
      "Epoch 449/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2160 - accuracy: 0.8436 - val_loss: 2.0165 - val_accuracy: 0.7062\n",
      "Epoch 450/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2205 - accuracy: 0.8445 - val_loss: 2.0297 - val_accuracy: 0.6959\n",
      "Epoch 451/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2106 - accuracy: 0.8497 - val_loss: 2.0147 - val_accuracy: 0.6982\n",
      "Epoch 452/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2189 - accuracy: 0.8493 - val_loss: 2.0084 - val_accuracy: 0.7062\n",
      "Epoch 453/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1862 - accuracy: 0.8621 - val_loss: 2.0268 - val_accuracy: 0.6993\n",
      "Epoch 454/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2015 - accuracy: 0.8577 - val_loss: 2.0334 - val_accuracy: 0.6982\n",
      "Epoch 455/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2029 - accuracy: 0.8536 - val_loss: 2.0427 - val_accuracy: 0.7051\n",
      "Epoch 456/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2128 - accuracy: 0.8484 - val_loss: 2.0182 - val_accuracy: 0.7005\n",
      "Epoch 457/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2077 - accuracy: 0.8548 - val_loss: 2.0339 - val_accuracy: 0.6889\n",
      "Epoch 458/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2106 - accuracy: 0.8538 - val_loss: 2.0237 - val_accuracy: 0.7039\n",
      "Epoch 459/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2079 - accuracy: 0.8531 - val_loss: 2.0147 - val_accuracy: 0.6982\n",
      "Epoch 460/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2098 - accuracy: 0.8564 - val_loss: 2.0406 - val_accuracy: 0.7074\n",
      "Epoch 461/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2016 - accuracy: 0.8534 - val_loss: 2.0268 - val_accuracy: 0.7016\n",
      "Epoch 462/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2121 - accuracy: 0.8486 - val_loss: 2.0110 - val_accuracy: 0.7005\n",
      "Epoch 463/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.1971 - accuracy: 0.8563 - val_loss: 2.0211 - val_accuracy: 0.6982\n",
      "Epoch 464/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1984 - accuracy: 0.8572 - val_loss: 2.0186 - val_accuracy: 0.6993\n",
      "Epoch 465/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2013 - accuracy: 0.8586 - val_loss: 2.0270 - val_accuracy: 0.6947\n",
      "Epoch 466/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2160 - accuracy: 0.8521 - val_loss: 2.0267 - val_accuracy: 0.6982\n",
      "Epoch 467/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2035 - accuracy: 0.8527 - val_loss: 2.0288 - val_accuracy: 0.6982\n",
      "Epoch 468/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1925 - accuracy: 0.8580 - val_loss: 2.0256 - val_accuracy: 0.7097\n",
      "Epoch 469/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2277 - accuracy: 0.8503 - val_loss: 2.0209 - val_accuracy: 0.7062\n",
      "Epoch 470/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2057 - accuracy: 0.8543 - val_loss: 2.0144 - val_accuracy: 0.7016\n",
      "Epoch 471/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2064 - accuracy: 0.8545 - val_loss: 2.0455 - val_accuracy: 0.6924\n",
      "Epoch 472/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2117 - accuracy: 0.8489 - val_loss: 2.0051 - val_accuracy: 0.7039\n",
      "Epoch 473/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1978 - accuracy: 0.8559 - val_loss: 2.0136 - val_accuracy: 0.6970\n",
      "Epoch 474/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2117 - accuracy: 0.8523 - val_loss: 2.0185 - val_accuracy: 0.6970\n",
      "Epoch 475/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1910 - accuracy: 0.8535 - val_loss: 2.0266 - val_accuracy: 0.6970\n",
      "Epoch 476/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.2043 - accuracy: 0.8532 - val_loss: 2.0371 - val_accuracy: 0.7016\n",
      "Epoch 477/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1882 - accuracy: 0.8585 - val_loss: 2.0216 - val_accuracy: 0.7016\n",
      "Epoch 478/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1908 - accuracy: 0.8598 - val_loss: 2.0146 - val_accuracy: 0.6970\n",
      "Epoch 479/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1905 - accuracy: 0.8570 - val_loss: 2.0127 - val_accuracy: 0.7005\n",
      "Epoch 480/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2143 - accuracy: 0.8558 - val_loss: 2.0290 - val_accuracy: 0.6947\n",
      "Epoch 481/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1767 - accuracy: 0.8590 - val_loss: 2.0317 - val_accuracy: 0.6993\n",
      "Epoch 482/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1878 - accuracy: 0.8585 - val_loss: 2.0178 - val_accuracy: 0.7005\n",
      "Epoch 483/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1907 - accuracy: 0.8608 - val_loss: 2.0400 - val_accuracy: 0.6993\n",
      "Epoch 484/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2068 - accuracy: 0.8567 - val_loss: 2.0214 - val_accuracy: 0.7016\n",
      "Epoch 485/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1996 - accuracy: 0.8536 - val_loss: 2.0490 - val_accuracy: 0.7051\n",
      "Epoch 486/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.2030 - accuracy: 0.8576 - val_loss: 2.0387 - val_accuracy: 0.7028\n",
      "Epoch 487/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1799 - accuracy: 0.8630 - val_loss: 2.0356 - val_accuracy: 0.6970\n",
      "Epoch 488/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1749 - accuracy: 0.8603 - val_loss: 2.0385 - val_accuracy: 0.6970\n",
      "Epoch 489/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1878 - accuracy: 0.8584 - val_loss: 2.0462 - val_accuracy: 0.6993\n",
      "Epoch 490/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1906 - accuracy: 0.8594 - val_loss: 2.0372 - val_accuracy: 0.7016\n",
      "Epoch 491/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1961 - accuracy: 0.8559 - val_loss: 2.0334 - val_accuracy: 0.6935\n",
      "Epoch 492/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1860 - accuracy: 0.8605 - val_loss: 2.0257 - val_accuracy: 0.7051\n",
      "Epoch 493/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1794 - accuracy: 0.8635 - val_loss: 2.0338 - val_accuracy: 0.7028\n",
      "Epoch 494/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1967 - accuracy: 0.8590 - val_loss: 2.0385 - val_accuracy: 0.7016\n",
      "Epoch 495/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1900 - accuracy: 0.8590 - val_loss: 2.0534 - val_accuracy: 0.7016\n",
      "Epoch 496/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1785 - accuracy: 0.8657 - val_loss: 2.0377 - val_accuracy: 0.7028\n",
      "Epoch 497/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1698 - accuracy: 0.8686 - val_loss: 2.0356 - val_accuracy: 0.7016\n",
      "Epoch 498/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1866 - accuracy: 0.8603 - val_loss: 2.0410 - val_accuracy: 0.6970\n",
      "Epoch 499/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1904 - accuracy: 0.8616 - val_loss: 2.0522 - val_accuracy: 0.7005\n",
      "Epoch 500/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1801 - accuracy: 0.8623 - val_loss: 2.0352 - val_accuracy: 0.7039\n",
      "Epoch 501/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1887 - accuracy: 0.8532 - val_loss: 2.0374 - val_accuracy: 0.6982\n",
      "Epoch 502/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1913 - accuracy: 0.8586 - val_loss: 2.0230 - val_accuracy: 0.7131\n",
      "Epoch 503/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1714 - accuracy: 0.8648 - val_loss: 2.0428 - val_accuracy: 0.7028\n",
      "Epoch 504/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1838 - accuracy: 0.8604 - val_loss: 2.0271 - val_accuracy: 0.7028\n",
      "Epoch 505/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1764 - accuracy: 0.8653 - val_loss: 2.0348 - val_accuracy: 0.6982\n",
      "Epoch 506/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1849 - accuracy: 0.8614 - val_loss: 2.0539 - val_accuracy: 0.6982\n",
      "Epoch 507/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1766 - accuracy: 0.8667 - val_loss: 2.0385 - val_accuracy: 0.7005\n",
      "Epoch 508/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1777 - accuracy: 0.8680 - val_loss: 2.0367 - val_accuracy: 0.7028\n",
      "Epoch 509/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1788 - accuracy: 0.8641 - val_loss: 2.0553 - val_accuracy: 0.6970\n",
      "Epoch 510/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1867 - accuracy: 0.8622 - val_loss: 2.0693 - val_accuracy: 0.6970\n",
      "Epoch 511/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1672 - accuracy: 0.8684 - val_loss: 2.0401 - val_accuracy: 0.7016\n",
      "Epoch 512/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1984 - accuracy: 0.8563 - val_loss: 2.0337 - val_accuracy: 0.6959\n",
      "Epoch 513/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1877 - accuracy: 0.8652 - val_loss: 2.0592 - val_accuracy: 0.6970\n",
      "Epoch 514/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1770 - accuracy: 0.8620 - val_loss: 2.0333 - val_accuracy: 0.7074\n",
      "Epoch 515/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1694 - accuracy: 0.8669 - val_loss: 2.0557 - val_accuracy: 0.7062\n",
      "Epoch 516/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1744 - accuracy: 0.8648 - val_loss: 2.0380 - val_accuracy: 0.7039\n",
      "Epoch 517/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1764 - accuracy: 0.8612 - val_loss: 2.0320 - val_accuracy: 0.7039\n",
      "Epoch 518/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1639 - accuracy: 0.8673 - val_loss: 2.0481 - val_accuracy: 0.6993\n",
      "Epoch 519/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1795 - accuracy: 0.8641 - val_loss: 2.0398 - val_accuracy: 0.7005\n",
      "Epoch 520/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.1622 - accuracy: 0.8735 - val_loss: 2.0309 - val_accuracy: 0.6970\n",
      "Epoch 521/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1836 - accuracy: 0.8600 - val_loss: 2.0332 - val_accuracy: 0.6901\n",
      "Epoch 522/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1695 - accuracy: 0.8669 - val_loss: 2.0233 - val_accuracy: 0.7085\n",
      "Epoch 523/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1798 - accuracy: 0.8625 - val_loss: 2.0223 - val_accuracy: 0.7005\n",
      "Epoch 524/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1669 - accuracy: 0.8678 - val_loss: 2.0368 - val_accuracy: 0.6924\n",
      "Epoch 525/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1755 - accuracy: 0.8646 - val_loss: 2.0453 - val_accuracy: 0.6924\n",
      "Epoch 526/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1749 - accuracy: 0.8676 - val_loss: 2.0387 - val_accuracy: 0.6993\n",
      "Epoch 527/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1779 - accuracy: 0.8643 - val_loss: 2.0265 - val_accuracy: 0.6970\n",
      "Epoch 528/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1752 - accuracy: 0.8675 - val_loss: 2.0336 - val_accuracy: 0.6912\n",
      "Epoch 529/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1649 - accuracy: 0.8686 - val_loss: 2.0412 - val_accuracy: 0.6993\n",
      "Epoch 530/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1651 - accuracy: 0.8684 - val_loss: 2.0443 - val_accuracy: 0.6959\n",
      "Epoch 531/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1709 - accuracy: 0.8608 - val_loss: 2.0388 - val_accuracy: 0.7016\n",
      "Epoch 532/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1808 - accuracy: 0.8613 - val_loss: 2.0343 - val_accuracy: 0.7039\n",
      "Epoch 533/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1884 - accuracy: 0.8608 - val_loss: 2.0276 - val_accuracy: 0.6912\n",
      "Epoch 534/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1759 - accuracy: 0.8652 - val_loss: 2.0308 - val_accuracy: 0.7028\n",
      "Epoch 535/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1589 - accuracy: 0.8745 - val_loss: 2.0439 - val_accuracy: 0.6924\n",
      "Epoch 536/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1713 - accuracy: 0.8661 - val_loss: 2.0228 - val_accuracy: 0.7062\n",
      "Epoch 537/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1799 - accuracy: 0.8625 - val_loss: 2.0351 - val_accuracy: 0.7005\n",
      "Epoch 538/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1668 - accuracy: 0.8708 - val_loss: 2.0302 - val_accuracy: 0.7039\n",
      "Epoch 539/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1576 - accuracy: 0.8686 - val_loss: 2.0367 - val_accuracy: 0.6993\n",
      "Epoch 540/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1666 - accuracy: 0.8685 - val_loss: 2.0331 - val_accuracy: 0.6959\n",
      "Epoch 541/700\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1743 - accuracy: 0.8663 - val_loss: 2.0426 - val_accuracy: 0.6993\n",
      "Epoch 542/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1708 - accuracy: 0.8680 - val_loss: 2.0328 - val_accuracy: 0.6959\n",
      "Epoch 543/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1726 - accuracy: 0.8661 - val_loss: 2.0228 - val_accuracy: 0.7039\n",
      "Epoch 544/700\n",
      "7809/7809 [==============================] - 1s 120us/sample - loss: 1.1716 - accuracy: 0.8707 - val_loss: 2.0145 - val_accuracy: 0.7051\n",
      "Epoch 545/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1607 - accuracy: 0.8698 - val_loss: 2.0263 - val_accuracy: 0.7028\n",
      "Epoch 546/700\n",
      "7809/7809 [==============================] - 1s 120us/sample - loss: 1.1863 - accuracy: 0.8669 - val_loss: 2.0352 - val_accuracy: 0.6970\n",
      "Epoch 547/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1580 - accuracy: 0.8750 - val_loss: 2.0399 - val_accuracy: 0.7028\n",
      "Epoch 548/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1563 - accuracy: 0.8718 - val_loss: 2.0381 - val_accuracy: 0.7074\n",
      "Epoch 549/700\n",
      "7809/7809 [==============================] - 1s 120us/sample - loss: 1.1639 - accuracy: 0.8708 - val_loss: 2.0242 - val_accuracy: 0.7051\n",
      "Epoch 550/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1561 - accuracy: 0.8712 - val_loss: 2.0250 - val_accuracy: 0.6982\n",
      "Epoch 551/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1733 - accuracy: 0.8668 - val_loss: 2.0420 - val_accuracy: 0.6970\n",
      "Epoch 552/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1704 - accuracy: 0.8646 - val_loss: 2.0377 - val_accuracy: 0.7051\n",
      "Epoch 553/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1775 - accuracy: 0.8641 - val_loss: 2.0385 - val_accuracy: 0.7005\n",
      "Epoch 554/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1495 - accuracy: 0.8782 - val_loss: 2.0455 - val_accuracy: 0.7016\n",
      "Epoch 555/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1606 - accuracy: 0.8719 - val_loss: 2.0224 - val_accuracy: 0.6993\n",
      "Epoch 556/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1427 - accuracy: 0.8759 - val_loss: 2.0312 - val_accuracy: 0.6959\n",
      "Epoch 557/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1687 - accuracy: 0.8714 - val_loss: 2.0250 - val_accuracy: 0.6993\n",
      "Epoch 558/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1781 - accuracy: 0.8662 - val_loss: 2.0251 - val_accuracy: 0.7028\n",
      "Epoch 559/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1454 - accuracy: 0.8735 - val_loss: 2.0350 - val_accuracy: 0.6970\n",
      "Epoch 560/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.1576 - accuracy: 0.8709 - val_loss: 2.0489 - val_accuracy: 0.6993\n",
      "Epoch 561/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1587 - accuracy: 0.8694 - val_loss: 2.0413 - val_accuracy: 0.7016\n",
      "Epoch 562/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1578 - accuracy: 0.8722 - val_loss: 2.0476 - val_accuracy: 0.7085\n",
      "Epoch 563/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1577 - accuracy: 0.8705 - val_loss: 2.0310 - val_accuracy: 0.7028\n",
      "Epoch 564/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1478 - accuracy: 0.8746 - val_loss: 2.0347 - val_accuracy: 0.6924\n",
      "Epoch 565/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1561 - accuracy: 0.8741 - val_loss: 2.0298 - val_accuracy: 0.6970\n",
      "Epoch 566/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1686 - accuracy: 0.8693 - val_loss: 2.0504 - val_accuracy: 0.7005\n",
      "Epoch 567/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1548 - accuracy: 0.8717 - val_loss: 2.0314 - val_accuracy: 0.6959\n",
      "Epoch 568/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1551 - accuracy: 0.8725 - val_loss: 2.0609 - val_accuracy: 0.6924\n",
      "Epoch 569/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1425 - accuracy: 0.8717 - val_loss: 2.0411 - val_accuracy: 0.7005\n",
      "Epoch 570/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1579 - accuracy: 0.8727 - val_loss: 2.0392 - val_accuracy: 0.6947\n",
      "Epoch 571/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1553 - accuracy: 0.8714 - val_loss: 2.0337 - val_accuracy: 0.6993\n",
      "Epoch 572/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1561 - accuracy: 0.8722 - val_loss: 2.0461 - val_accuracy: 0.6993\n",
      "Epoch 573/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1543 - accuracy: 0.8744 - val_loss: 2.0519 - val_accuracy: 0.6947\n",
      "Epoch 574/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1553 - accuracy: 0.8762 - val_loss: 2.0647 - val_accuracy: 0.6982\n",
      "Epoch 575/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.1580 - accuracy: 0.8767 - val_loss: 2.0588 - val_accuracy: 0.6959\n",
      "Epoch 576/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1536 - accuracy: 0.8722 - val_loss: 2.0596 - val_accuracy: 0.7005\n",
      "Epoch 577/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1528 - accuracy: 0.8751 - val_loss: 2.0523 - val_accuracy: 0.6935\n",
      "Epoch 578/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1329 - accuracy: 0.8810 - val_loss: 2.0504 - val_accuracy: 0.6947\n",
      "Epoch 579/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1517 - accuracy: 0.8707 - val_loss: 2.0372 - val_accuracy: 0.7074\n",
      "Epoch 580/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1456 - accuracy: 0.8790 - val_loss: 2.0290 - val_accuracy: 0.7005\n",
      "Epoch 581/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1500 - accuracy: 0.8730 - val_loss: 2.0364 - val_accuracy: 0.7028\n",
      "Epoch 582/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1451 - accuracy: 0.8774 - val_loss: 2.0300 - val_accuracy: 0.6993\n",
      "Epoch 583/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1546 - accuracy: 0.8741 - val_loss: 2.0520 - val_accuracy: 0.6970\n",
      "Epoch 584/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1319 - accuracy: 0.8782 - val_loss: 2.0372 - val_accuracy: 0.7005\n",
      "Epoch 585/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1555 - accuracy: 0.8680 - val_loss: 2.0391 - val_accuracy: 0.6820\n",
      "Epoch 586/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1487 - accuracy: 0.8751 - val_loss: 2.0459 - val_accuracy: 0.7028\n",
      "Epoch 587/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.1400 - accuracy: 0.8755 - val_loss: 2.0471 - val_accuracy: 0.7028\n",
      "Epoch 588/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1623 - accuracy: 0.8704 - val_loss: 2.0242 - val_accuracy: 0.7039\n",
      "Epoch 589/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1601 - accuracy: 0.8707 - val_loss: 2.0393 - val_accuracy: 0.7051\n",
      "Epoch 590/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1504 - accuracy: 0.8736 - val_loss: 2.0362 - val_accuracy: 0.7039\n",
      "Epoch 591/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1653 - accuracy: 0.8675 - val_loss: 2.0470 - val_accuracy: 0.7051\n",
      "Epoch 592/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1624 - accuracy: 0.8691 - val_loss: 2.0323 - val_accuracy: 0.7005\n",
      "Epoch 593/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1414 - accuracy: 0.8809 - val_loss: 2.0374 - val_accuracy: 0.6935\n",
      "Epoch 594/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1442 - accuracy: 0.8771 - val_loss: 2.0463 - val_accuracy: 0.6970\n",
      "Epoch 595/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1584 - accuracy: 0.8771 - val_loss: 2.0396 - val_accuracy: 0.7005\n",
      "Epoch 596/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1378 - accuracy: 0.8763 - val_loss: 2.0200 - val_accuracy: 0.7097\n",
      "Epoch 597/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1348 - accuracy: 0.8786 - val_loss: 2.0571 - val_accuracy: 0.6924\n",
      "Epoch 598/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1430 - accuracy: 0.8783 - val_loss: 2.0400 - val_accuracy: 0.7005\n",
      "Epoch 599/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1561 - accuracy: 0.8727 - val_loss: 2.0350 - val_accuracy: 0.7039\n",
      "Epoch 600/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1448 - accuracy: 0.8742 - val_loss: 2.0489 - val_accuracy: 0.7074\n",
      "Epoch 601/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1389 - accuracy: 0.8737 - val_loss: 2.0289 - val_accuracy: 0.7085\n",
      "Epoch 602/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1361 - accuracy: 0.8732 - val_loss: 2.0408 - val_accuracy: 0.7016\n",
      "Epoch 603/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1329 - accuracy: 0.8774 - val_loss: 2.0456 - val_accuracy: 0.7062\n",
      "Epoch 604/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1391 - accuracy: 0.8776 - val_loss: 2.0737 - val_accuracy: 0.6982\n",
      "Epoch 605/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1349 - accuracy: 0.8812 - val_loss: 2.0506 - val_accuracy: 0.6959\n",
      "Epoch 606/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1442 - accuracy: 0.8759 - val_loss: 2.0381 - val_accuracy: 0.7074\n",
      "Epoch 607/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1383 - accuracy: 0.8780 - val_loss: 2.0577 - val_accuracy: 0.6924\n",
      "Epoch 608/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.1449 - accuracy: 0.8771 - val_loss: 2.0536 - val_accuracy: 0.7005\n",
      "Epoch 609/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1382 - accuracy: 0.8795 - val_loss: 2.0381 - val_accuracy: 0.7028\n",
      "Epoch 610/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1446 - accuracy: 0.8734 - val_loss: 2.0437 - val_accuracy: 0.7016\n",
      "Epoch 611/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1455 - accuracy: 0.8776 - val_loss: 2.0374 - val_accuracy: 0.6982\n",
      "Epoch 612/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1356 - accuracy: 0.8766 - val_loss: 2.0660 - val_accuracy: 0.6912\n",
      "Epoch 613/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1333 - accuracy: 0.8844 - val_loss: 2.0490 - val_accuracy: 0.6901\n",
      "Epoch 614/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1413 - accuracy: 0.8776 - val_loss: 2.0634 - val_accuracy: 0.7016\n",
      "Epoch 615/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1217 - accuracy: 0.8823 - val_loss: 2.0450 - val_accuracy: 0.7005\n",
      "Epoch 616/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1371 - accuracy: 0.8796 - val_loss: 2.0345 - val_accuracy: 0.7074\n",
      "Epoch 617/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1207 - accuracy: 0.8842 - val_loss: 2.0525 - val_accuracy: 0.6924\n",
      "Epoch 618/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1296 - accuracy: 0.8807 - val_loss: 2.0606 - val_accuracy: 0.6947\n",
      "Epoch 619/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1487 - accuracy: 0.8757 - val_loss: 2.0365 - val_accuracy: 0.6993\n",
      "Epoch 620/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1363 - accuracy: 0.8782 - val_loss: 2.0385 - val_accuracy: 0.7062\n",
      "Epoch 621/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1406 - accuracy: 0.8782 - val_loss: 2.0241 - val_accuracy: 0.6982\n",
      "Epoch 622/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1396 - accuracy: 0.8839 - val_loss: 2.0310 - val_accuracy: 0.6982\n",
      "Epoch 623/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.1332 - accuracy: 0.8800 - val_loss: 2.0489 - val_accuracy: 0.7005\n",
      "Epoch 624/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1410 - accuracy: 0.8804 - val_loss: 2.0321 - val_accuracy: 0.7005\n",
      "Epoch 625/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1604 - accuracy: 0.8672 - val_loss: 2.0276 - val_accuracy: 0.7085\n",
      "Epoch 626/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1423 - accuracy: 0.8768 - val_loss: 2.0261 - val_accuracy: 0.7028\n",
      "Epoch 627/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1275 - accuracy: 0.8859 - val_loss: 2.0284 - val_accuracy: 0.7074\n",
      "Epoch 628/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1322 - accuracy: 0.8822 - val_loss: 2.0325 - val_accuracy: 0.7051\n",
      "Epoch 629/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1490 - accuracy: 0.8757 - val_loss: 2.0348 - val_accuracy: 0.7016\n",
      "Epoch 630/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1457 - accuracy: 0.8796 - val_loss: 2.0398 - val_accuracy: 0.6970\n",
      "Epoch 631/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1470 - accuracy: 0.8731 - val_loss: 2.0461 - val_accuracy: 0.6924\n",
      "Epoch 632/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1215 - accuracy: 0.8865 - val_loss: 2.0537 - val_accuracy: 0.6970\n",
      "Epoch 633/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1267 - accuracy: 0.8850 - val_loss: 2.0590 - val_accuracy: 0.7028\n",
      "Epoch 634/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1473 - accuracy: 0.8727 - val_loss: 2.0537 - val_accuracy: 0.6959\n",
      "Epoch 635/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1304 - accuracy: 0.8839 - val_loss: 2.0529 - val_accuracy: 0.7016\n",
      "Epoch 636/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1229 - accuracy: 0.8869 - val_loss: 2.0300 - val_accuracy: 0.7028\n",
      "Epoch 637/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1393 - accuracy: 0.8773 - val_loss: 2.0326 - val_accuracy: 0.7062\n",
      "Epoch 638/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1137 - accuracy: 0.8882 - val_loss: 2.0162 - val_accuracy: 0.6970\n",
      "Epoch 639/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1369 - accuracy: 0.8812 - val_loss: 2.0219 - val_accuracy: 0.7005\n",
      "Epoch 640/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1390 - accuracy: 0.8826 - val_loss: 2.0314 - val_accuracy: 0.7016\n",
      "Epoch 641/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1297 - accuracy: 0.8818 - val_loss: 2.0393 - val_accuracy: 0.7016\n",
      "Epoch 642/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1383 - accuracy: 0.8760 - val_loss: 2.0359 - val_accuracy: 0.7039\n",
      "Epoch 643/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1420 - accuracy: 0.8780 - val_loss: 2.0273 - val_accuracy: 0.6959\n",
      "Epoch 644/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1306 - accuracy: 0.8826 - val_loss: 2.0517 - val_accuracy: 0.7028\n",
      "Epoch 645/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1343 - accuracy: 0.8810 - val_loss: 2.0350 - val_accuracy: 0.6993\n",
      "Epoch 646/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1200 - accuracy: 0.8836 - val_loss: 2.0286 - val_accuracy: 0.6982\n",
      "Epoch 647/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1375 - accuracy: 0.8815 - val_loss: 2.0286 - val_accuracy: 0.7005\n",
      "Epoch 648/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1354 - accuracy: 0.8751 - val_loss: 2.0365 - val_accuracy: 0.6935\n",
      "Epoch 649/700\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1270 - accuracy: 0.8837 - val_loss: 2.0484 - val_accuracy: 0.6912\n",
      "Epoch 650/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1321 - accuracy: 0.8830 - val_loss: 2.0435 - val_accuracy: 0.7039\n",
      "Epoch 651/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1109 - accuracy: 0.8890 - val_loss: 2.0413 - val_accuracy: 0.6970\n",
      "Epoch 652/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1317 - accuracy: 0.8815 - val_loss: 2.0433 - val_accuracy: 0.6935\n",
      "Epoch 653/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1209 - accuracy: 0.8865 - val_loss: 2.0488 - val_accuracy: 0.7005\n",
      "Epoch 654/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1148 - accuracy: 0.8844 - val_loss: 2.0319 - val_accuracy: 0.7028\n",
      "Epoch 655/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1272 - accuracy: 0.8809 - val_loss: 2.0404 - val_accuracy: 0.7062\n",
      "Epoch 656/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1289 - accuracy: 0.8859 - val_loss: 2.0253 - val_accuracy: 0.6993\n",
      "Epoch 657/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1278 - accuracy: 0.8836 - val_loss: 2.0434 - val_accuracy: 0.6982\n",
      "Epoch 658/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1499 - accuracy: 0.8780 - val_loss: 2.0436 - val_accuracy: 0.6924\n",
      "Epoch 659/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1205 - accuracy: 0.8881 - val_loss: 2.0402 - val_accuracy: 0.6982\n",
      "Epoch 660/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1280 - accuracy: 0.8831 - val_loss: 2.0533 - val_accuracy: 0.6924\n",
      "Epoch 661/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1407 - accuracy: 0.8787 - val_loss: 2.0595 - val_accuracy: 0.6924\n",
      "Epoch 662/700\n",
      "7809/7809 [==============================] - 1s 123us/sample - loss: 1.1328 - accuracy: 0.8790 - val_loss: 2.0734 - val_accuracy: 0.6866\n",
      "Epoch 663/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1287 - accuracy: 0.8846 - val_loss: 2.0424 - val_accuracy: 0.6959\n",
      "Epoch 664/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1218 - accuracy: 0.8883 - val_loss: 2.0433 - val_accuracy: 0.6993\n",
      "Epoch 665/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1198 - accuracy: 0.8877 - val_loss: 2.0439 - val_accuracy: 0.6982\n",
      "Epoch 666/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1217 - accuracy: 0.8850 - val_loss: 2.0445 - val_accuracy: 0.6993\n",
      "Epoch 667/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1334 - accuracy: 0.8814 - val_loss: 2.0448 - val_accuracy: 0.6970\n",
      "Epoch 668/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1370 - accuracy: 0.8862 - val_loss: 2.0525 - val_accuracy: 0.7085\n",
      "Epoch 669/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1277 - accuracy: 0.8844 - val_loss: 2.0319 - val_accuracy: 0.6970\n",
      "Epoch 670/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1207 - accuracy: 0.8887 - val_loss: 2.0434 - val_accuracy: 0.7028\n",
      "Epoch 671/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1198 - accuracy: 0.8878 - val_loss: 2.0296 - val_accuracy: 0.7085\n",
      "Epoch 672/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1263 - accuracy: 0.8917 - val_loss: 2.0289 - val_accuracy: 0.7028\n",
      "Epoch 673/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1129 - accuracy: 0.8831 - val_loss: 2.0574 - val_accuracy: 0.7039\n",
      "Epoch 674/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1278 - accuracy: 0.8828 - val_loss: 2.0668 - val_accuracy: 0.7028\n",
      "Epoch 675/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1219 - accuracy: 0.8865 - val_loss: 2.0474 - val_accuracy: 0.7051\n",
      "Epoch 676/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1378 - accuracy: 0.8776 - val_loss: 2.0505 - val_accuracy: 0.6982\n",
      "Epoch 677/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1242 - accuracy: 0.8840 - val_loss: 2.0435 - val_accuracy: 0.7051\n",
      "Epoch 678/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1327 - accuracy: 0.8840 - val_loss: 2.0449 - val_accuracy: 0.7062\n",
      "Epoch 679/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1232 - accuracy: 0.8863 - val_loss: 2.0661 - val_accuracy: 0.6993\n",
      "Epoch 680/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1335 - accuracy: 0.8845 - val_loss: 2.0506 - val_accuracy: 0.7039\n",
      "Epoch 681/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1220 - accuracy: 0.8878 - val_loss: 2.0690 - val_accuracy: 0.7039\n",
      "Epoch 682/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1215 - accuracy: 0.8804 - val_loss: 2.0440 - val_accuracy: 0.7016\n",
      "Epoch 683/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1216 - accuracy: 0.8856 - val_loss: 2.0797 - val_accuracy: 0.6993\n",
      "Epoch 684/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1220 - accuracy: 0.8849 - val_loss: 2.0400 - val_accuracy: 0.7005\n",
      "Epoch 685/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1208 - accuracy: 0.8868 - val_loss: 2.0454 - val_accuracy: 0.7005\n",
      "Epoch 686/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1177 - accuracy: 0.8885 - val_loss: 2.0189 - val_accuracy: 0.7028\n",
      "Epoch 687/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1319 - accuracy: 0.8847 - val_loss: 2.0291 - val_accuracy: 0.7097\n",
      "Epoch 688/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1253 - accuracy: 0.8831 - val_loss: 2.0408 - val_accuracy: 0.6993\n",
      "Epoch 689/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1148 - accuracy: 0.8878 - val_loss: 2.0455 - val_accuracy: 0.7028\n",
      "Epoch 690/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1225 - accuracy: 0.8854 - val_loss: 2.0419 - val_accuracy: 0.7016\n",
      "Epoch 691/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1173 - accuracy: 0.8868 - val_loss: 2.0469 - val_accuracy: 0.7039\n",
      "Epoch 692/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1105 - accuracy: 0.8853 - val_loss: 2.0420 - val_accuracy: 0.7051\n",
      "Epoch 693/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1167 - accuracy: 0.8872 - val_loss: 2.0585 - val_accuracy: 0.6959\n",
      "Epoch 694/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1122 - accuracy: 0.8874 - val_loss: 2.0501 - val_accuracy: 0.6982\n",
      "Epoch 695/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1118 - accuracy: 0.8886 - val_loss: 2.0533 - val_accuracy: 0.7005\n",
      "Epoch 696/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1410 - accuracy: 0.8764 - val_loss: 2.0216 - val_accuracy: 0.7051\n",
      "Epoch 697/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1143 - accuracy: 0.8900 - val_loss: 2.0330 - val_accuracy: 0.7062\n",
      "Epoch 698/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1110 - accuracy: 0.8915 - val_loss: 2.0299 - val_accuracy: 0.7051\n",
      "Epoch 699/700\n",
      "7809/7809 [==============================] - 1s 121us/sample - loss: 1.1151 - accuracy: 0.8869 - val_loss: 2.0413 - val_accuracy: 0.7005\n",
      "Epoch 700/700\n",
      "7809/7809 [==============================] - 1s 122us/sample - loss: 1.1130 - accuracy: 0.8924 - val_loss: 2.0392 - val_accuracy: 0.6993\n"
     ]
    }
   ],
   "source": [
    "regularized = model_final.fit(X, t, validation_split=0.1, batch_size=30, epochs=700)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\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": [
    "plt.plot(regularized.history['accuracy'])\n",
    "plt.plot(regularized.history['val_accuracy'])\n",
    "plt.title('Regularized model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'validation'], loc='upper left')\n",
    "plt.grid()\n",
    "plt.show()\n",
    "\n",
    "plt.plot(regularized.history['loss'])\n",
    "plt.plot(regularized.history['val_loss'])\n",
    "plt.title('Regularized model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'validation'], loc='upper right')\n",
    "plt.grid()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Results of the baseline model (after 100 epochs):\n",
    "\n",
    "#### Training accuracy ~ 90%\n",
    "#### Validation accuracy ~ 70%\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Comparing the baseline model to the model with linear activation functions\n",
    "\n",
    "Here we take the baseline model and change all the activation functions to linear, in order to compare with the baseline model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_linear = Sequential()\n",
    "model_linear.add(Conv2D(filters=32, kernel_size=(5,5), strides=(1,1), activation='linear', input_shape=(64,64,3)))\n",
    "model_linear.add(MaxPooling2D(pool_size=(4,4), strides=(4,4)))\n",
    "model_linear.add(Conv2D(filters=64, kernel_size=(5,5), strides=(1,1), activation='linear'))\n",
    "model_linear.add(MaxPooling2D(pool_size=(4,4), strides=(4,4)))\n",
    "model_linear.add(Flatten())\n",
    "model_linear.add(Dense(units=1024, activation='linear'))\n",
    "model_linear.add(Dense(units=84, activation='linear'))\n",
    "model_linear.add(Dense(units=101, activation='softmax'))\n",
    "\n",
    "adam = optimizers.Adam(learning_rate=0.0001)\n",
    "#sgd = optimizers.SGD(learning_rate=0.005, decay=1e-4, momentum=0.99)\n",
    "model_linear.compile(loss='categorical_crossentropy', optimizer=adam, metrics=['accuracy'])\n",
    "\n",
    "print(model_linear.summary())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 7809 samples, validate on 868 samples\n",
      "Epoch 1/200\n",
      "7809/7809 [==============================] - 1s 141us/sample - loss: 3.9233 - accuracy: 0.2090 - val_loss: 3.5920 - val_accuracy: 0.2638\n",
      "Epoch 2/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 3.2622 - accuracy: 0.3201 - val_loss: 3.1668 - val_accuracy: 0.3306\n",
      "Epoch 3/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 2.9023 - accuracy: 0.3755 - val_loss: 2.9284 - val_accuracy: 0.3744\n",
      "Epoch 4/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 2.6619 - accuracy: 0.4100 - val_loss: 2.7399 - val_accuracy: 0.4055\n",
      "Epoch 5/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 2.4782 - accuracy: 0.4488 - val_loss: 2.5996 - val_accuracy: 0.4412\n",
      "Epoch 6/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 2.3300 - accuracy: 0.4700 - val_loss: 2.4766 - val_accuracy: 0.4493\n",
      "Epoch 7/200\n",
      "7809/7809 [==============================] - 1s 110us/sample - loss: 2.2021 - accuracy: 0.4987 - val_loss: 2.3895 - val_accuracy: 0.4689\n",
      "Epoch 8/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 2.0937 - accuracy: 0.5190 - val_loss: 2.2912 - val_accuracy: 0.4885\n",
      "Epoch 9/200\n",
      "7809/7809 [==============================] - 1s 109us/sample - loss: 1.9956 - accuracy: 0.5322 - val_loss: 2.2566 - val_accuracy: 0.4827\n",
      "Epoch 10/200\n",
      "7809/7809 [==============================] - 1s 109us/sample - loss: 1.9140 - accuracy: 0.5532 - val_loss: 2.1779 - val_accuracy: 0.5150\n",
      "Epoch 11/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 1.8391 - accuracy: 0.5716 - val_loss: 2.1332 - val_accuracy: 0.5150\n",
      "Epoch 12/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 1.7648 - accuracy: 0.5845 - val_loss: 2.0623 - val_accuracy: 0.5346\n",
      "Epoch 13/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 1.6916 - accuracy: 0.5980 - val_loss: 2.0497 - val_accuracy: 0.5403\n",
      "Epoch 14/200\n",
      "7809/7809 [==============================] - 1s 109us/sample - loss: 1.6341 - accuracy: 0.6099 - val_loss: 2.0514 - val_accuracy: 0.5530\n",
      "Epoch 15/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 1.5800 - accuracy: 0.6234 - val_loss: 1.9783 - val_accuracy: 0.5403\n",
      "Epoch 16/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 1.5294 - accuracy: 0.6320 - val_loss: 2.0122 - val_accuracy: 0.5403\n",
      "Epoch 17/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 1.4732 - accuracy: 0.6395 - val_loss: 1.9536 - val_accuracy: 0.5668\n",
      "Epoch 18/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 1.4293 - accuracy: 0.6549 - val_loss: 1.9309 - val_accuracy: 0.5645\n",
      "Epoch 19/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 1.3861 - accuracy: 0.6638 - val_loss: 1.9474 - val_accuracy: 0.5703\n",
      "Epoch 20/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 1.3455 - accuracy: 0.6719 - val_loss: 1.9196 - val_accuracy: 0.5703\n",
      "Epoch 21/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 1.2984 - accuracy: 0.6843 - val_loss: 1.8927 - val_accuracy: 0.5691\n",
      "Epoch 22/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 1.2630 - accuracy: 0.6914 - val_loss: 1.9067 - val_accuracy: 0.5806\n",
      "Epoch 23/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 1.2330 - accuracy: 0.6996 - val_loss: 1.8729 - val_accuracy: 0.5864\n",
      "Epoch 24/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 1.1885 - accuracy: 0.7092 - val_loss: 1.8678 - val_accuracy: 0.5783\n",
      "Epoch 25/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 1.1539 - accuracy: 0.7132 - val_loss: 1.8486 - val_accuracy: 0.5899\n",
      "Epoch 26/200\n",
      "7809/7809 [==============================] - 1s 109us/sample - loss: 1.1176 - accuracy: 0.7233 - val_loss: 1.8907 - val_accuracy: 0.5841\n",
      "Epoch 27/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 1.0865 - accuracy: 0.7308 - val_loss: 1.8725 - val_accuracy: 0.5760\n",
      "Epoch 28/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 1.0544 - accuracy: 0.7365 - val_loss: 1.8515 - val_accuracy: 0.5922\n",
      "Epoch 29/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 1.0235 - accuracy: 0.7422 - val_loss: 1.8596 - val_accuracy: 0.5945\n",
      "Epoch 30/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.9847 - accuracy: 0.7535 - val_loss: 1.8454 - val_accuracy: 0.5945\n",
      "Epoch 31/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.9626 - accuracy: 0.7596 - val_loss: 1.8580 - val_accuracy: 0.5979\n",
      "Epoch 32/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.9375 - accuracy: 0.7622 - val_loss: 1.8651 - val_accuracy: 0.5991\n",
      "Epoch 33/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.9135 - accuracy: 0.7692 - val_loss: 1.8550 - val_accuracy: 0.6014\n",
      "Epoch 34/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.8748 - accuracy: 0.7800 - val_loss: 1.8801 - val_accuracy: 0.5968\n",
      "Epoch 35/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.8515 - accuracy: 0.7856 - val_loss: 1.8882 - val_accuracy: 0.5887\n",
      "Epoch 36/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.8253 - accuracy: 0.7928 - val_loss: 1.9000 - val_accuracy: 0.5922\n",
      "Epoch 37/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.7966 - accuracy: 0.8010 - val_loss: 1.8923 - val_accuracy: 0.6025\n",
      "Epoch 38/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.7822 - accuracy: 0.8037 - val_loss: 1.9201 - val_accuracy: 0.5899\n",
      "Epoch 39/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.7544 - accuracy: 0.8064 - val_loss: 1.9340 - val_accuracy: 0.6152\n",
      "Epoch 40/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.7291 - accuracy: 0.8192 - val_loss: 1.9959 - val_accuracy: 0.5922\n",
      "Epoch 41/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.7060 - accuracy: 0.8253 - val_loss: 1.9306 - val_accuracy: 0.6094\n",
      "Epoch 42/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.6869 - accuracy: 0.8278 - val_loss: 1.9690 - val_accuracy: 0.6187\n",
      "Epoch 43/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.6575 - accuracy: 0.8349 - val_loss: 2.0621 - val_accuracy: 0.5899\n",
      "Epoch 44/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.6439 - accuracy: 0.8376 - val_loss: 1.9937 - val_accuracy: 0.5945\n",
      "Epoch 45/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.6236 - accuracy: 0.8458 - val_loss: 2.0162 - val_accuracy: 0.6025\n",
      "Epoch 46/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.5914 - accuracy: 0.8527 - val_loss: 2.0401 - val_accuracy: 0.5991\n",
      "Epoch 47/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.5775 - accuracy: 0.8550 - val_loss: 2.0697 - val_accuracy: 0.6025\n",
      "Epoch 48/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.5573 - accuracy: 0.8585 - val_loss: 2.0419 - val_accuracy: 0.6060\n",
      "Epoch 49/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.5389 - accuracy: 0.8643 - val_loss: 2.0923 - val_accuracy: 0.6002\n",
      "Epoch 50/200\n",
      "7809/7809 [==============================] - 1s 106us/sample - loss: 0.5243 - accuracy: 0.8671 - val_loss: 2.0836 - val_accuracy: 0.6083\n",
      "Epoch 51/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.5044 - accuracy: 0.8734 - val_loss: 2.1059 - val_accuracy: 0.5991\n",
      "Epoch 52/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.4856 - accuracy: 0.8798 - val_loss: 2.1243 - val_accuracy: 0.6014\n",
      "Epoch 53/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.4654 - accuracy: 0.8836 - val_loss: 2.1522 - val_accuracy: 0.5991\n",
      "Epoch 54/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.4596 - accuracy: 0.8844 - val_loss: 2.1701 - val_accuracy: 0.5979\n",
      "Epoch 55/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.4337 - accuracy: 0.8918 - val_loss: 2.2156 - val_accuracy: 0.6025\n",
      "Epoch 56/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.4191 - accuracy: 0.8959 - val_loss: 2.2387 - val_accuracy: 0.6025\n",
      "Epoch 57/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.3971 - accuracy: 0.9011 - val_loss: 2.2587 - val_accuracy: 0.6106\n",
      "Epoch 58/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.3835 - accuracy: 0.9068 - val_loss: 2.3018 - val_accuracy: 0.5991\n",
      "Epoch 59/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.3750 - accuracy: 0.9079 - val_loss: 2.3077 - val_accuracy: 0.5968\n",
      "Epoch 60/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.3577 - accuracy: 0.9125 - val_loss: 2.4087 - val_accuracy: 0.5910\n",
      "Epoch 61/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.3488 - accuracy: 0.9138 - val_loss: 2.4613 - val_accuracy: 0.5864\n",
      "Epoch 62/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.3311 - accuracy: 0.9175 - val_loss: 2.4339 - val_accuracy: 0.5956\n",
      "Epoch 63/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.3140 - accuracy: 0.9225 - val_loss: 2.4553 - val_accuracy: 0.5945\n",
      "Epoch 64/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.3084 - accuracy: 0.9237 - val_loss: 2.5462 - val_accuracy: 0.5829\n",
      "Epoch 65/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.2972 - accuracy: 0.9276 - val_loss: 2.5085 - val_accuracy: 0.5887\n",
      "Epoch 66/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.2812 - accuracy: 0.9305 - val_loss: 2.5972 - val_accuracy: 0.5772\n",
      "Epoch 67/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.2695 - accuracy: 0.9360 - val_loss: 2.6441 - val_accuracy: 0.5772\n",
      "Epoch 68/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.2549 - accuracy: 0.9411 - val_loss: 2.6838 - val_accuracy: 0.5795\n",
      "Epoch 69/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.2485 - accuracy: 0.9421 - val_loss: 2.7316 - val_accuracy: 0.5737\n",
      "Epoch 70/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.2336 - accuracy: 0.9442 - val_loss: 2.6863 - val_accuracy: 0.5887\n",
      "Epoch 71/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.2465 - accuracy: 0.9388 - val_loss: 2.7337 - val_accuracy: 0.5979\n",
      "Epoch 72/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.2109 - accuracy: 0.9531 - val_loss: 2.7609 - val_accuracy: 0.5795\n",
      "Epoch 73/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.2008 - accuracy: 0.9553 - val_loss: 2.7972 - val_accuracy: 0.5876\n",
      "Epoch 74/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.1948 - accuracy: 0.9583 - val_loss: 2.8744 - val_accuracy: 0.5783\n",
      "Epoch 75/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.1862 - accuracy: 0.9603 - val_loss: 2.8734 - val_accuracy: 0.5933\n",
      "Epoch 76/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.1705 - accuracy: 0.9654 - val_loss: 2.9339 - val_accuracy: 0.5841\n",
      "Epoch 77/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.1665 - accuracy: 0.9648 - val_loss: 2.9801 - val_accuracy: 0.5876\n",
      "Epoch 78/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.1642 - accuracy: 0.9653 - val_loss: 3.0447 - val_accuracy: 0.5783\n",
      "Epoch 79/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.1527 - accuracy: 0.9661 - val_loss: 3.0361 - val_accuracy: 0.5956\n",
      "Epoch 80/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.1497 - accuracy: 0.9680 - val_loss: 3.0847 - val_accuracy: 0.5910\n",
      "Epoch 81/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.1438 - accuracy: 0.9686 - val_loss: 3.1723 - val_accuracy: 0.5772\n",
      "Epoch 82/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.1503 - accuracy: 0.9657 - val_loss: 3.2100 - val_accuracy: 0.5714\n",
      "Epoch 83/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.1357 - accuracy: 0.9721 - val_loss: 3.1792 - val_accuracy: 0.5887\n",
      "Epoch 84/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.1157 - accuracy: 0.9784 - val_loss: 3.2343 - val_accuracy: 0.5749\n",
      "Epoch 85/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.1173 - accuracy: 0.9757 - val_loss: 3.2436 - val_accuracy: 0.5864\n",
      "Epoch 86/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.1026 - accuracy: 0.9826 - val_loss: 3.2861 - val_accuracy: 0.5864\n",
      "Epoch 87/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.1096 - accuracy: 0.9775 - val_loss: 3.3352 - val_accuracy: 0.5922\n",
      "Epoch 88/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0998 - accuracy: 0.9808 - val_loss: 3.3779 - val_accuracy: 0.5829\n",
      "Epoch 89/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.1080 - accuracy: 0.9768 - val_loss: 3.4484 - val_accuracy: 0.5853\n",
      "Epoch 90/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0949 - accuracy: 0.9834 - val_loss: 3.5189 - val_accuracy: 0.5853\n",
      "Epoch 91/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0920 - accuracy: 0.9816 - val_loss: 3.5333 - val_accuracy: 0.5899\n",
      "Epoch 92/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0902 - accuracy: 0.9839 - val_loss: 3.4826 - val_accuracy: 0.5876\n",
      "Epoch 93/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0799 - accuracy: 0.9850 - val_loss: 3.5606 - val_accuracy: 0.5795\n",
      "Epoch 94/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0671 - accuracy: 0.9894 - val_loss: 3.6252 - val_accuracy: 0.5910\n",
      "Epoch 95/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0651 - accuracy: 0.9887 - val_loss: 3.7544 - val_accuracy: 0.5749\n",
      "Epoch 96/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.1000 - accuracy: 0.9772 - val_loss: 3.7527 - val_accuracy: 0.5737\n",
      "Epoch 97/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0828 - accuracy: 0.9834 - val_loss: 3.7735 - val_accuracy: 0.5806\n",
      "Epoch 98/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0691 - accuracy: 0.9866 - val_loss: 3.8119 - val_accuracy: 0.5783\n",
      "Epoch 99/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0706 - accuracy: 0.9860 - val_loss: 3.7140 - val_accuracy: 0.5841\n",
      "Epoch 100/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0674 - accuracy: 0.9891 - val_loss: 3.7610 - val_accuracy: 0.5899\n",
      "Epoch 101/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0497 - accuracy: 0.9923 - val_loss: 3.8426 - val_accuracy: 0.5887\n",
      "Epoch 102/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0893 - accuracy: 0.9761 - val_loss: 3.9554 - val_accuracy: 0.5795\n",
      "Epoch 103/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0995 - accuracy: 0.9744 - val_loss: 3.9538 - val_accuracy: 0.5760\n",
      "Epoch 104/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0497 - accuracy: 0.9926 - val_loss: 3.8229 - val_accuracy: 0.5922\n",
      "Epoch 105/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0519 - accuracy: 0.9913 - val_loss: 3.8491 - val_accuracy: 0.6025\n",
      "Epoch 106/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0418 - accuracy: 0.9942 - val_loss: 3.9922 - val_accuracy: 0.5910\n",
      "Epoch 107/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0348 - accuracy: 0.9962 - val_loss: 3.9566 - val_accuracy: 0.5887\n",
      "Epoch 108/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0395 - accuracy: 0.9939 - val_loss: 3.9855 - val_accuracy: 0.5887\n",
      "Epoch 109/200\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0613 - accuracy: 0.9857 - val_loss: 4.1655 - val_accuracy: 0.5760\n",
      "Epoch 110/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0956 - accuracy: 0.9717 - val_loss: 4.2551 - val_accuracy: 0.5956\n",
      "Epoch 111/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0900 - accuracy: 0.9769 - val_loss: 4.0333 - val_accuracy: 0.5795\n",
      "Epoch 112/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0282 - accuracy: 0.9974 - val_loss: 3.9759 - val_accuracy: 0.5945\n",
      "Epoch 113/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0233 - accuracy: 0.9982 - val_loss: 4.0386 - val_accuracy: 0.5876\n",
      "Epoch 114/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0306 - accuracy: 0.9964 - val_loss: 4.0597 - val_accuracy: 0.5945\n",
      "Epoch 115/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0184 - accuracy: 0.9990 - val_loss: 4.1026 - val_accuracy: 0.5853\n",
      "Epoch 116/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0523 - accuracy: 0.9868 - val_loss: 4.4342 - val_accuracy: 0.5956\n",
      "Epoch 117/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.1207 - accuracy: 0.9621 - val_loss: 4.2173 - val_accuracy: 0.5853\n",
      "Epoch 118/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0557 - accuracy: 0.9875 - val_loss: 4.2584 - val_accuracy: 0.5864\n",
      "Epoch 119/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0265 - accuracy: 0.9965 - val_loss: 4.1751 - val_accuracy: 0.5829\n",
      "Epoch 120/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0192 - accuracy: 0.9985 - val_loss: 4.2228 - val_accuracy: 0.5933\n",
      "Epoch 121/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0214 - accuracy: 0.9978 - val_loss: 4.2347 - val_accuracy: 0.5887\n",
      "Epoch 122/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0121 - accuracy: 0.9994 - val_loss: 4.2475 - val_accuracy: 0.5956\n",
      "Epoch 123/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0157 - accuracy: 0.9988 - val_loss: 4.3199 - val_accuracy: 0.5945\n",
      "Epoch 124/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.1889 - accuracy: 0.9411 - val_loss: 4.4742 - val_accuracy: 0.5841\n",
      "Epoch 125/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0642 - accuracy: 0.9822 - val_loss: 4.4016 - val_accuracy: 0.5818\n",
      "Epoch 126/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0219 - accuracy: 0.9977 - val_loss: 4.2239 - val_accuracy: 0.6025\n",
      "Epoch 127/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0123 - accuracy: 0.9994 - val_loss: 4.2568 - val_accuracy: 0.6025\n",
      "Epoch 128/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0155 - accuracy: 0.9986 - val_loss: 4.2584 - val_accuracy: 0.6037\n",
      "Epoch 129/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0113 - accuracy: 0.9994 - val_loss: 4.2980 - val_accuracy: 0.6037\n",
      "Epoch 130/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0130 - accuracy: 0.9991 - val_loss: 4.3221 - val_accuracy: 0.6048\n",
      "Epoch 131/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0102 - accuracy: 0.9995 - val_loss: 4.3501 - val_accuracy: 0.6002\n",
      "Epoch 132/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0259 - accuracy: 0.9955 - val_loss: 4.7933 - val_accuracy: 0.5691\n",
      "Epoch 133/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.2623 - accuracy: 0.9269 - val_loss: 4.3973 - val_accuracy: 0.5968\n",
      "Epoch 134/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0380 - accuracy: 0.9928 - val_loss: 4.3625 - val_accuracy: 0.6014\n",
      "Epoch 135/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0169 - accuracy: 0.9982 - val_loss: 4.3346 - val_accuracy: 0.6060\n",
      "Epoch 136/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0189 - accuracy: 0.9978 - val_loss: 4.3480 - val_accuracy: 0.6094\n",
      "Epoch 137/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0113 - accuracy: 0.9991 - val_loss: 4.3668 - val_accuracy: 0.6060\n",
      "Epoch 138/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0076 - accuracy: 0.9997 - val_loss: 4.3836 - val_accuracy: 0.6094\n",
      "Epoch 139/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0088 - accuracy: 0.9995 - val_loss: 4.4148 - val_accuracy: 0.6037\n",
      "Epoch 140/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0092 - accuracy: 0.9995 - val_loss: 4.4141 - val_accuracy: 0.6037\n",
      "Epoch 141/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0103 - accuracy: 0.9994 - val_loss: 4.4937 - val_accuracy: 0.5991\n",
      "Epoch 142/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0129 - accuracy: 0.9992 - val_loss: 4.4943 - val_accuracy: 0.5979\n",
      "Epoch 143/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0271 - accuracy: 0.9954 - val_loss: 4.9045 - val_accuracy: 0.5680\n",
      "Epoch 144/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.2460 - accuracy: 0.9271 - val_loss: 4.5879 - val_accuracy: 0.5956\n",
      "Epoch 145/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0481 - accuracy: 0.9872 - val_loss: 4.4880 - val_accuracy: 0.6002\n",
      "Epoch 146/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0155 - accuracy: 0.9986 - val_loss: 4.3941 - val_accuracy: 0.6048\n",
      "Epoch 147/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0106 - accuracy: 0.9992 - val_loss: 4.3933 - val_accuracy: 0.6014\n",
      "Epoch 148/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0093 - accuracy: 0.9992 - val_loss: 4.4155 - val_accuracy: 0.6037\n",
      "Epoch 149/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0088 - accuracy: 0.9994 - val_loss: 4.4491 - val_accuracy: 0.5991\n",
      "Epoch 150/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0082 - accuracy: 0.9992 - val_loss: 4.4611 - val_accuracy: 0.5933\n",
      "Epoch 151/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0107 - accuracy: 0.9992 - val_loss: 4.4785 - val_accuracy: 0.5956\n",
      "Epoch 152/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0076 - accuracy: 0.9994 - val_loss: 4.5046 - val_accuracy: 0.5956\n",
      "Epoch 153/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0618 - accuracy: 0.9831 - val_loss: 5.3306 - val_accuracy: 0.5680\n",
      "Epoch 154/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.1869 - accuracy: 0.9443 - val_loss: 4.6439 - val_accuracy: 0.5853\n",
      "Epoch 155/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0268 - accuracy: 0.9954 - val_loss: 4.5714 - val_accuracy: 0.6025\n",
      "Epoch 156/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0120 - accuracy: 0.9991 - val_loss: 4.5123 - val_accuracy: 0.6060\n",
      "Epoch 157/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0084 - accuracy: 0.9994 - val_loss: 4.5165 - val_accuracy: 0.6071\n",
      "Epoch 158/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0073 - accuracy: 0.9995 - val_loss: 4.5230 - val_accuracy: 0.6083\n",
      "Epoch 159/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0072 - accuracy: 0.9995 - val_loss: 4.5488 - val_accuracy: 0.6048\n",
      "Epoch 160/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0095 - accuracy: 0.9990 - val_loss: 4.5708 - val_accuracy: 0.5991\n",
      "Epoch 161/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0083 - accuracy: 0.9992 - val_loss: 4.6060 - val_accuracy: 0.6025\n",
      "Epoch 162/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0094 - accuracy: 0.9988 - val_loss: 4.6322 - val_accuracy: 0.6071\n",
      "Epoch 163/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0114 - accuracy: 0.9986 - val_loss: 4.7533 - val_accuracy: 0.5956\n",
      "Epoch 164/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.2221 - accuracy: 0.9356 - val_loss: 4.7644 - val_accuracy: 0.5945\n",
      "Epoch 165/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0505 - accuracy: 0.9871 - val_loss: 4.5873 - val_accuracy: 0.5887\n",
      "Epoch 166/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0167 - accuracy: 0.9977 - val_loss: 4.5517 - val_accuracy: 0.5979\n",
      "Epoch 167/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0087 - accuracy: 0.9994 - val_loss: 4.5704 - val_accuracy: 0.5979\n",
      "Epoch 168/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0071 - accuracy: 0.9995 - val_loss: 4.5568 - val_accuracy: 0.5956\n",
      "Epoch 169/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0074 - accuracy: 0.9992 - val_loss: 4.5879 - val_accuracy: 0.6014\n",
      "Epoch 170/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0077 - accuracy: 0.9992 - val_loss: 4.5837 - val_accuracy: 0.6071\n",
      "Epoch 171/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0064 - accuracy: 0.9995 - val_loss: 4.5914 - val_accuracy: 0.6083\n",
      "Epoch 172/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0066 - accuracy: 0.9994 - val_loss: 4.5976 - val_accuracy: 0.6014\n",
      "Epoch 173/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0165 - accuracy: 0.9967 - val_loss: 4.9281 - val_accuracy: 0.5887\n",
      "Epoch 174/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.1861 - accuracy: 0.9463 - val_loss: 4.8575 - val_accuracy: 0.5922\n",
      "Epoch 175/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0429 - accuracy: 0.9883 - val_loss: 4.7758 - val_accuracy: 0.6002\n",
      "Epoch 176/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0205 - accuracy: 0.9963 - val_loss: 4.7272 - val_accuracy: 0.5956\n",
      "Epoch 177/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0103 - accuracy: 0.9986 - val_loss: 4.6424 - val_accuracy: 0.6083\n",
      "Epoch 178/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0063 - accuracy: 0.9994 - val_loss: 4.6389 - val_accuracy: 0.6094\n",
      "Epoch 179/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0067 - accuracy: 0.9994 - val_loss: 4.6675 - val_accuracy: 0.6037\n",
      "Epoch 180/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0065 - accuracy: 0.9994 - val_loss: 4.6574 - val_accuracy: 0.6094\n",
      "Epoch 181/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0073 - accuracy: 0.9991 - val_loss: 4.7986 - val_accuracy: 0.6014\n",
      "Epoch 182/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0050 - accuracy: 0.9996 - val_loss: 4.7244 - val_accuracy: 0.6083\n",
      "Epoch 183/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0325 - accuracy: 0.9931 - val_loss: 5.2185 - val_accuracy: 0.5887\n",
      "Epoch 184/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.1784 - accuracy: 0.9451 - val_loss: 4.9046 - val_accuracy: 0.5818\n",
      "Epoch 185/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0357 - accuracy: 0.9907 - val_loss: 4.8186 - val_accuracy: 0.6002\n",
      "Epoch 186/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0118 - accuracy: 0.9986 - val_loss: 4.7604 - val_accuracy: 0.5979\n",
      "Epoch 187/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0077 - accuracy: 0.9991 - val_loss: 4.7344 - val_accuracy: 0.5991\n",
      "Epoch 188/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0064 - accuracy: 0.9994 - val_loss: 4.7397 - val_accuracy: 0.6037\n",
      "Epoch 189/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0066 - accuracy: 0.9991 - val_loss: 4.7737 - val_accuracy: 0.5933\n",
      "Epoch 190/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0065 - accuracy: 0.9992 - val_loss: 4.7788 - val_accuracy: 0.5979\n",
      "Epoch 191/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0062 - accuracy: 0.9994 - val_loss: 4.7801 - val_accuracy: 0.6071\n",
      "Epoch 192/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0066 - accuracy: 0.9991 - val_loss: 4.8011 - val_accuracy: 0.6037\n",
      "Epoch 193/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0045 - accuracy: 0.9995 - val_loss: 4.8302 - val_accuracy: 0.5968\n",
      "Epoch 194/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0081 - accuracy: 0.9991 - val_loss: 4.8546 - val_accuracy: 0.6071\n",
      "Epoch 195/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.1159 - accuracy: 0.9703 - val_loss: 5.6969 - val_accuracy: 0.5726\n",
      "Epoch 196/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.1371 - accuracy: 0.9589 - val_loss: 4.9358 - val_accuracy: 0.5968\n",
      "Epoch 197/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0206 - accuracy: 0.9955 - val_loss: 4.9297 - val_accuracy: 0.6014\n",
      "Epoch 198/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0083 - accuracy: 0.9991 - val_loss: 4.8698 - val_accuracy: 0.6002\n",
      "Epoch 199/200\n",
      "7809/7809 [==============================] - 1s 108us/sample - loss: 0.0061 - accuracy: 0.9992 - val_loss: 4.8584 - val_accuracy: 0.6083\n",
      "Epoch 200/200\n",
      "7809/7809 [==============================] - 1s 107us/sample - loss: 0.0066 - accuracy: 0.9991 - val_loss: 4.8574 - val_accuracy: 0.6083\n"
     ]
    }
   ],
   "source": [
    "baseline_linear = model_linear.fit(X, t, validation_split=0.1, batch_size=30, epochs=200)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\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": [
    "# plt.plot(baseline_linear.history['accuracy'])\n",
    "# plt.plot(baseline_linear.history['val_accuracy'])\n",
    "# plt.title('Baseline linear model accuracy')\n",
    "# plt.ylabel('accuracy')\n",
    "# plt.xlabel('epoch')\n",
    "# plt.legend(['train', 'val'], loc='upper left')\n",
    "# plt.grid()\n",
    "# plt.show()\n",
    "\n",
    "plt.plot(baseline.history['loss'])\n",
    "#plt.plot(baseline.history['val_loss'])\n",
    "plt.plot(baseline_linear.history['loss'])\n",
    "#plt.plot(baseline_linear.history['val_loss'])\n",
    "plt.title('Model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train (baseline)', 'train (linear model)'], loc='upper center')\n",
    "plt.grid()\n",
    "plt.show()\n",
    "\n",
    "#plt.plot(baseline.history['loss'])\n",
    "plt.plot(baseline.history['val_loss'])\n",
    "#plt.plot(baseline_linear.history['loss'])\n",
    "plt.plot(baseline_linear.history['val_loss'])\n",
    "plt.title('Model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['val (baseline)', 'val (linear model)'], loc='upper center')\n",
    "plt.grid()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "When the layer activation functions are changed to linear, the model loses its ability to fit to wide variety data mappings. This is due to the fact that all the layers with linear activation functions would be equivalent to one layer (since the composition of multiple linear functions is also a linear function), theryby nullifying the effect of having multiple layers in the architecture. Therefore the model performance is supposed to be decreased.\n",
    "\n",
    "When comparing with the baseline model, which had ReLU activation functions, the linear model has almost the similar plots for training and validation losses as shown in the above plots. Clearly, both the models are overfit to the training data and perform poorly on the validation data. The reason for similar performance of 2 models, as opposed to the expected decrease in performance in case of linear model, could be due to the fact that the original non-linearity (ReLU) is linear in half of the universe, because of which the improvements are not significant when we replace it with fully linear activation functions.\n",
    "\n",
    "We cannot compensate for the effect of removing the non-linearity by adding more convolution layers with linear activation functions, since all the layers (irrespective of the number) with linear activation functions are equivalent to one layer with linear activation function (since the composition of multiple linear functions is also linear). "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Visualizing the weights of first layer\n",
    "\n",
    "Here we are visualizing the weights of the first layer (which is the convolution layer) of the regularized non-linear model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [],
   "source": [
    "layer = model_final.layers[0]\n",
    "filters, biases = layer.get_weights()\n",
    "f = np.transpose(filters,(3,0,1,2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/local-scratch/mmallya/mconda/envs/env_cmpt980_a2/lib/python3.7/site-packages/ipykernel_launcher.py:3: RuntimeWarning: More than 20 figures have been opened. Figures created through the pyplot interface (`matplotlib.pyplot.figure`) are retained until explicitly closed and may consume too much memory. (To control this warning, see the rcParam `figure.max_open_warning`).\n",
      "  This is separate from the ipykernel package so we can avoid doing imports until\n"
     ]
    },
    {
     "data": {
      "image/png": "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\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"
    },
    {
     "data": {
      "image/png": "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\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"
    },
    {
     "data": {
      "image/png": "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\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"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAPUAAAD4CAYAAAA0L6C7AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjMsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+AADFEAAAJlklEQVR4nO3dzYtdhR3G8efJ5MXY2CbRWDWTNhZEGoQqDEHILriIL+hWQVdCNrVEEESX/gPixk3QYEFRBF2IWCSgIoJVR41iGoUgiiG2SYxpTG0TJ/N0MbNIbSZz7s0998z58f3AwNzc4dyHMN85M3eGc51EAOpY1vUAAKNF1EAxRA0UQ9RAMUQNFLO8jYNOrF2XiauuaePQozd7tusFA1m2vF9fh8/O9mfvzOx/up7Q3D+OKP886fPd1U7UV12jq/Y838ahR27mxxNdTxjIpWsv63rCQP714+quJzT2939/0fWE5v700IJ39efLKIBGiBoohqiBYogaKIaogWKIGiiGqIFiiBoohqiBYogaKIaogWKIGiiGqIFiiBoohqiBYogaKIaogWIaRW17h+0vbB+0/UjbowAMb9GobU9IelLSrZK2SLrH9pa2hwEYTpMz9VZJB5N8meSMpBck3dXuLADDahL1RknfnHP70Py//Q/bO21P256ePfH9qPYBGFCTqM93GdL/e1W9JLuTTCWZWrZ23cUvAzCUJlEfkrTpnNuTkg63MwfAxWoS9QeSrrN9re2Vku6W9Eq7swAMa9GL+SeZsf2ApNclTUjak2R/68sADKXRK3QkeU3Say1vATAC/EUZUAxRA8UQNVAMUQPFEDVQDFEDxRA1UAxRA8UQNVAMUQPFEDVQDFEDxRA1UAxRA8UQNVAMUQPFNLpIwsAHnVimdb+8tI1Dj9z6K9Z3PWEgx9ds6HrCQH799dmuJzTnK7te0NixZWsWvI8zNVAMUQPFEDVQDFEDxRA1UAxRA8UQNVAMUQPFEDVQDFEDxRA1UAxRA8UQNVAMUQPFEDVQDFEDxRA1UMyiUdveY/uI7c/GMQjAxWlypn5G0o6WdwAYkUWjTvK2pONj2AJgBPiZGihmZFHb3ml72vb0zPec2IGujCzqJLuTTCWZWr6uX5fdBSrh22+gmCa/0npe0ruSrrd9yPb97c8CMKxFX6EjyT3jGAJgNPj2GyiGqIFiiBoohqiBYogaKIaogWKIGiiGqIFiiBoohqiBYogaKIaogWKIGiiGqIFiiBoohqiBYha9SMIwZies079a1cahR+/E6q4XDMSXH+16wkBWXt3Kp1grlh3rzwUzvWxmwfs4UwPFEDVQDFEDxRA1UAxRA8UQNVAMUQPFEDVQDFEDxRA1UAxRA8UQNVAMUQPFEDVQDFEDxRA1UAxRA8UQNVDMolHb3mT7TdsHbO+3vWscwwAMp8kFpGYkPZTkI9uXSfrQ9t4kf2t5G4AhLHqmTvJtko/m3/9B0gFJG9seBmA4A/1MbXuzpJskvXee+3banrY9ffZ4f67KCFTTOGrbayS9JOnBJCd/fn+S3UmmkkxNrF8/yo0ABtAoatsrNBf0c0lebncSgIvR5NlvS3pa0oEkj7c/CcDFaHKm3ibpPknbbe+bf7ut5V0AhrTor7SSvCPJY9gCYAT4izKgGKIGiiFqoBiiBoohaqAYogaKIWqgGKIGiiFqoBiiBoohaqAYogaKIWqgGKIGiiFqoBiiBoppct3vwQ86E11+9Kc2Dj1y323o10USJ06d6XrCQHLmUNcTGjtz5kTXExrL7MyC93GmBoohaqAYogaKIWqgGKIGiiFqoBiiBoohaqAYogaKIWqgGKIGiiFqoBiiBoohaqAYogaKIWqgGKIGilk0atuX2H7f9ie299t+bBzDAAynyeWMTkvanuSU7RWS3rH9lyR/bXkbgCEsGnWSSDo1f3PF/FvaHAVgeI1+prY9YXufpCOS9iZ5r91ZAIbVKOokZ5PcKGlS0lbbN/z8Y2zvtD1te/qn74+PeieAhgZ69jvJCUlvSdpxnvt2J5lKMrViXb8uuwtU0uTZ7w22186/v1rSLZI+b3sYgOE0efb7akl/tj2huS8CLyZ5td1ZAIbV5NnvTyXdNIYtAEaAvygDiiFqoBiiBoohaqAYogaKIWqgGKIGiiFqoBiiBoohaqAYogaKIWqgGKIGiiFqoBiiBoohaqCYJlc+GdjK5av0myt/18ahRy6rDnc9YSDf/bC26wkDOfzVpV1PaOwXx/qz9eSZhc/HnKmBYogaKIaogWKIGiiGqIFiiBoohqiBYogaKIaogWKIGiiGqIFiiBoohqiBYogaKIaogWKIGiiGqIFiiBoopnHUtidsf2z71TYHAbg4g5ypd0k60NYQAKPRKGrbk5Jul/RUu3MAXKymZ+onJD0saXahD7C90/a07enTx4+OZByAwS0ate07JB1J8uGFPi7J7iRTSaZWrd8wsoEABtPkTL1N0p22v5L0gqTttp9tdRWAoS0adZJHk0wm2SzpbklvJLm39WUAhsLvqYFiBnrZnSRvSXqrlSUARoIzNVAMUQPFEDVQDFEDxRA1UAxRA8UQNVAMUQPFEDVQDFEDxRA1UAxRA8UQNVAMUQPFEDVQDFEDxTjJ6A9qH5X09YgPe4WkYyM+Zpv6tLdPW6V+7W1r62+TnPcKn61E3Qbb00mmut7RVJ/29mmr1K+9XWzl22+gGKIGiulT1Lu7HjCgPu3t01apX3vHvrU3P1MDaKZPZ2oADRA1UEwvora9w/YXtg/afqTrPRdie4/tI7Y/63rLYmxvsv2m7QO299ve1fWmhdi+xPb7tj+Z3/pY15uasD1h+2Pbr47rMZd81LYnJD0p6VZJWyTdY3tLt6su6BlJO7oe0dCMpIeS/F7SzZL+uIT/b09L2p7kD5JulLTD9s0db2pil6QD43zAJR+1pK2SDib5MskZzb3y5l0db1pQkrclHe96RxNJvk3y0fz7P2juk29jt6vOL3NOzd9cMf+2pJ/ltT0p6XZJT43zcfsQ9UZJ35xz+5CW6Cden9neLOkmSe91u2Rh89/K7pN0RNLeJEt267wnJD0saXacD9qHqH2ef1vSX6H7xvYaSS9JejDJya73LCTJ2SQ3SpqUtNX2DV1vWojtOyQdSfLhuB+7D1EfkrTpnNuTkg53tKUc2ys0F/RzSV7uek8TSU5o7tVXl/JzF9sk3Wn7K839yLjd9rPjeOA+RP2BpOtsX2t7peZe+P6VjjeVYNuSnpZ0IMnjXe+5ENsbbK+df3+1pFskfd7tqoUleTTJZJLNmvucfSPJveN47CUfdZIZSQ9Iel1zT+S8mGR/t6sWZvt5Se9Kut72Idv3d73pArZJuk9zZ5F982+3dT1qAVdLetP2p5r7Qr83ydh+TdQn/JkoUMySP1MDGAxRA8UQNVAMUQPFEDVQDFEDxRA1UMx/AUZ7BrRxFQZXAAAAAElFTkSuQmCC\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"
    },
    {
     "data": {
      "image/png": "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\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"
    },
    {
     "data": {
      "image/png": "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\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"
    },
    {
     "data": {
      "image/png": "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\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"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAPUAAAD4CAYAAAA0L6C7AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjMsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+AADFEAAAJn0lEQVR4nO3dzWvcBR7H8c/HGJ+2Lm2zXdCku/UgskVYhVCE3oqH+oBeFfQk9LJCBUH06D8gwuKlqLi7iiLoQcRFClZEcKtRq9itQvEBiy7RtNIHH9o0nz0kh66bNL+Zzm9+mS/vFwQynfCbDyXv/GYmYcZJBKCOi7oeAGCwiBoohqiBYogaKIaogWIubuOgGzasz+Tk1W0ceuB+OT3f9YSezM7Ndj2hJ6dOnOp6QmPjY2NdT2jszJkzmp8/6+WuayXqycmr9eLL/2jj0AP31RdzXU/oyV///kTXE3qyf987XU9o7Or1G7ue0NjhL75c8TrufgPFEDVQDFEDxRA1UAxRA8UQNVAMUQPFEDVQDFEDxRA1UAxRA8UQNVAMUQPFEDVQDFEDxRA1UAxRA8U0itr2Ttuf2T5s++G2RwHo36pR2x6T9ISkWyRtlXS37a1tDwPQnyZn6m2SDif5PMlpSS9IurPdWQD61STqSUlfn3P5yNK//Q/bu2zP2J45euzYoPYB6FGTqJd7GdL/e1e9JHuSTCeZ3rhhw4UvA9CXJlEfkbT5nMtTkr5pZw6AC9Uk6vckXWv7GtuXSLpL0ivtzgLQr1VfzD/JvO37Jb0uaUzS00kOtr4MQF8avUNHktckvdbyFgADwF+UAcUQNVAMUQPFEDVQDFEDxRA1UAxRA8UQNVAMUQPFEDVQDFEDxRA1UAxRA8UQNVAMUQPFEDVQTKMXSejV2NjFWn/l79s49MD9NPlT1xN68vPcya4n9OTHY6e6ntDYj1eMxvesJC1kudcDXcSZGiiGqIFiiBoohqiBYogaKIaogWKIGiiGqIFiiBoohqiBYogaKIaogWKIGiiGqIFiiBoohqiBYogaKGbVqG0/bXvW9ifDGATgwjQ5Uz8jaWfLOwAMyKpRJ3lL0tEhbAEwADymBooZWNS2d9mesT0zN8eJHejKwKJOsifJdJLpiYmNgzosgB5x9xsopsmvtJ6X9I6k62wfsX1f+7MA9GvVd+hIcvcwhgAYDO5+A8UQNVAMUQPFEDVQDFEDxRA1UAxRA8UQNVAMUQPFEDVQDFEDxRA1UAxRA8UQNVAMUQPFEDVQzKovktAPj1uXXjXexqEHbv70aP1cO3Hyx64n9OTn+dNdT+jBya4H9GBhxWtG6zsawKqIGiiGqIFiiBoohqiBYogaKIaogWKIGiiGqIFiiBoohqiBYogaKIaogWKIGiiGqIFiiBoohqiBYogaKGbVqG1vtr3P9iHbB23vHsYwAP1p8hpl85IeTPKB7SslvW97b5J/t7wNQB9WPVMn+TbJB0ufn5B0SNJk28MA9Kenx9S2t0i6UdL+Za7bZXvG9szcd3ODWQegZ42jtr1O0kuSHkhy/NfXJ9mTZDrJ9MSmiUFuBNCDRlHbHtdi0M8lebndSQAuRJNnvy3pKUmHkjzW/iQAF6LJmXq7pHsl7bB9YOnj1pZ3AejTqr/SSvK2JA9hC4AB4C/KgGKIGiiGqIFiiBoohqiBYogaKIaogWKIGiiGqIFiiBoohqiBYogaKIaogWKIGiiGqIFiiBoopsnrfvdsQQs6tXCqjUMP3MJ/Ruv1H+Z1pusJPbn40iu6ntDYut/8tusJjY1dNLvidZypgWKIGiiGqIFiiBoohqiBYogaKIaogWKIGiiGqIFiiBoohqiBYogaKIaogWKIGiiGqIFiiBoohqiBYlaN2vZltt+1/ZHtg7YfHcYwAP1p8nJGv0jakeSk7XFJb9v+Z5J/tbwNQB9WjTpJJJ1cuji+9JE2RwHoX6PH1LbHbB+QNCtpb5L97c4C0K9GUSc5m+QGSVOSttm+/tdfY3uX7RnbM0e/OzronQAa6unZ7yQ/SHpT0s5lrtuTZDrJ9MZNGwc0D0Cvmjz7vcn2+qXPL5d0s6RP2x4GoD9Nnv2+StLfbI9p8YfAi0lebXcWgH41efb7Y0k3DmELgAHgL8qAYogaKIaogWKIGiiGqIFiiBoohqiBYogaKIaogWKIGiiGqIFiiBoohqiBYogaKIaogWKIGiimySuf9G7+rBa+P97KoQft27NHup7QkyvWTXQ9oSdX/mG26wmNnZ64rOsJjeULr3gdZ2qgGKIGiiFqoBiiBoohaqAYogaKIWqgGKIGiiFqoBiiBoohaqAYogaKIWqgGKIGiiFqoBiiBoohaqAYogaKaRy17THbH9p+tc1BAC5ML2fq3ZIOtTUEwGA0itr2lKTbJD3Z7hwAF6rpmfpxSQ9JWljpC2zvsj1je2Zu7thAxgHo3apR275d0myS98/3dUn2JJlOMj0xsWFgAwH0psmZerukO2x/KekFSTtsP9vqKgB9WzXqJI8kmUqyRdJdkt5Ick/rywD0hd9TA8X09LY7Sd6U9GYrSwAMBGdqoBiiBoohaqAYogaKIWqgGKIGiiFqoBiiBoohaqAYogaKIWqgGKIGiiFqoBiiBoohaqAYogaKcZLBH9T+TtJXAz7s7yR9P+BjtmmU9o7SVmm09ra19Y9JNi13RStRt8H2TJLprnc0NUp7R2mrNFp7u9jK3W+gGKIGihmlqPd0PaBHo7R3lLZKo7V36FtH5jE1gGZG6UwNoAGiBooZiaht77T9me3Dth/ues/52H7a9qztT7reshrbm23vs33I9kHbu7vetBLbl9l+1/ZHS1sf7XpTE7bHbH9o+9Vh3eaaj9r2mKQnJN0iaauku21v7XbVeT0jaWfXIxqal/Rgkj9JuknSX9bw/+0vknYk+bOkGyTttH1Tx5ua2C3p0DBvcM1HLWmbpMNJPk9yWovvvHlnx5tWlOQtSUe73tFEkm+TfLD0+QktfvNNdrtqeVl0cuni+NLHmn6W1/aUpNskPTnM2x2FqCclfX3O5SNao994o8z2Fkk3Strf7ZKVLd2VPSBpVtLeJGt265LHJT0kaWGYNzoKUXuZf1vTP6FHje11kl6S9ECS413vWUmSs0lukDQlaZvt67vetBLbt0uaTfL+sG97FKI+ImnzOZenJH3T0ZZybI9rMejnkrzc9Z4mkvygxXdfXcvPXWyXdIftL7X4kHGH7WeHccOjEPV7kq61fY3tS7T4xvevdLypBNuW9JSkQ0ke63rP+djeZHv90ueXS7pZ0qfdrlpZkkeSTCXZosXv2TeS3DOM217zUSeZl3S/pNe1+ETOi0kOdrtqZbafl/SOpOtsH7F9X9ebzmO7pHu1eBY5sPRxa9ejVnCVpH22P9biD/q9SYb2a6JRwp+JAsWs+TM1gN4QNVAMUQPFEDVQDFEDxRA1UAxRA8X8F+OoEd4nt7hyAAAAAElFTkSuQmCC\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"
    },
    {
     "data": {
      "image/png": "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\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"
    },
    {
     "data": {
      "image/png": "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\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"
    },
    {
     "data": {
      "image/png": "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\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"
    },
    {
     "data": {
      "image/png": "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\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"
    },
    {
     "data": {
      "image/png": "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\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"
    },
    {
     "data": {
      "image/png": "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\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"
    },
    {
     "data": {
      "image/png": "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\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"
    },
    {
     "data": {
      "image/png": "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\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": [
    "for i in range(32):\n",
    "    img = (f[i]-np.min(f[i]))/(np.max(f[i])-np.min(f[i]))\n",
    "    plt.figure()\n",
    "    plt.imshow(img)\n",
    "    "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The first layer convolution filters after sufficient training are as shown above. These learned are supposed to detect some low level features from the input image."
   ]
  },
  {
   "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": "3.7.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}