ADAM_OPTIMIZER_COMPARISON.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Y8vsur3VEzjH"
   },
   "source": [
    "# Learning Rate Schedule & Adaptive Gradient Descent Algorithm\n",
    "\n",
    "* A CNN model is constructed to train on CIFAR-10\n",
    "\n",
    "* The following learning rate decay schedules are used : constant learning rate, time-based decay, step decay, exponential decay\n",
    "\n",
    "* The following adaptive graident descent algorithms are used : Adagrad, Adadelta, RMSProp, Adam\n",
    "    \n",
    "* Model performance of using different learning rate schedules or adaptive gradient descent algorithms are compared"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "id": "jxTKq7FvEzjI"
   },
   "outputs": [],
   "source": [
    "from __future__ import print_function\n",
    "import numpy as np\n",
    "import math\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline  \n",
    "\n",
    "import keras\n",
    "from keras import backend as K\n",
    "from keras.datasets import cifar10\n",
    "from keras.utils import np_utils\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout, Flatten\n",
    "from keras.layers import Conv2D, MaxPooling2D\n",
    "from keras.optimizers import SGD, Adagrad, Adadelta, RMSprop, Adam\n",
    "from keras.models import model_from_json\n",
    "from keras.callbacks import LearningRateScheduler"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "VNtN3szFEzjM"
   },
   "source": [
    "# Load CIFAR-10 data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 68
    },
    "id": "EP1CfH6pEzjM",
    "outputId": "0dd27863-27e5-400b-9c1d-f20b7ebfed2e"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X_train shape: (10000, 32, 32, 3)\n",
      "10000 train samples\n",
      "2000 test samples\n"
     ]
    }
   ],
   "source": [
    "batch_size = 64\n",
    "num_classes = 2\n",
    "epochs = 100\n",
    "\n",
    "# input image dimensions\n",
    "img_rows, img_cols = 32, 32   \n",
    "\n",
    "# the data, shuffled and split between train and test sets\n",
    "(X_train, y_train), (X_test, y_test) = cifar10.load_data()   \n",
    "\n",
    "# Only look at cats [=3] and dogs [=5]\n",
    "train_picks = np.ravel(np.logical_or(y_train==3,y_train==5))  \n",
    "test_picks = np.ravel(np.logical_or(y_test==3,y_test==5))     \n",
    "\n",
    "y_train = np.array(y_train[train_picks]==5,dtype=int)\n",
    "y_test = np.array(y_test[test_picks]==5,dtype=int)\n",
    "\n",
    "X_train = X_train[train_picks]\n",
    "X_test = X_test[test_picks]\n",
    "\n",
    "if K.image_data_format() == 'channels_first':\n",
    "    X_train = X_train.reshape(X_train.shape[0], 3, img_rows, img_cols)\n",
    "    X_test = X_test.reshape(X_test.shape[0], 3, img_rows, img_cols)\n",
    "    input_shape = (3, img_rows, img_cols)\n",
    "else:\n",
    "    X_train = X_train.reshape(X_train.shape[0], img_rows, img_cols, 3)\n",
    "    X_test = X_test.reshape(X_test.shape[0], img_rows, img_cols, 3)\n",
    "    input_shape = (img_rows, img_cols, 3)\n",
    "\n",
    "X_train = X_train.astype('float32')\n",
    "X_test = X_test.astype('float32')\n",
    "X_train /= 255\n",
    "X_test /= 255\n",
    "print('X_train shape:', X_train.shape)\n",
    "print(X_train.shape[0], 'train samples')\n",
    "print(X_test.shape[0], 'test samples')\n",
    "\n",
    "# convert class vectors to binary class matrices\n",
    "y_train = keras.utils.to_categorical(np.ravel(y_train), num_classes)\n",
    "y_test = keras.utils.to_categorical(np.ravel(y_test), num_classes)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Yok5IiKrEzjO"
   },
   "source": [
    "# Define function to construct CNN model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "id": "iZ0RaruUEzjP"
   },
   "outputs": [],
   "source": [
    "def cnn_model() : \n",
    "    model = Sequential()\n",
    "    model.add(Conv2D(4, kernel_size=(3, 3),activation='relu',input_shape=input_shape))\n",
    "    model.add(Conv2D(8, (3, 3), activation='relu'))\n",
    "    model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "    model.add(Dropout(0.25))\n",
    "    model.add(Flatten())\n",
    "    model.add(Dense(16, activation='relu'))\n",
    "    model.add(Dropout(0.5))\n",
    "    model.add(Dense(2, activation='softmax'))\n",
    "    # print(model.summary())\n",
    "    return(model)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "n0hnlVODEzjk"
   },
   "source": [
    "# Adaptive gradient descent algorithms : Adagrad, Adadelta, RMSprop, Adam"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "id": "iRwReg2GEzjk",
    "outputId": "f4da5272-dba7-422c-e64c-1bd63c84b0c2",
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 10000 samples, validate on 2000 samples\n",
      "Epoch 1/100\n",
      " - 1s - loss: 0.6933 - acc: 0.4960 - val_loss: 0.6926 - val_acc: 0.5085\n",
      "Epoch 2/100\n",
      " - 1s - loss: 0.6842 - acc: 0.5560 - val_loss: 0.6666 - val_acc: 0.5830\n",
      "Epoch 3/100\n",
      " - 1s - loss: 0.6648 - acc: 0.6013 - val_loss: 0.6576 - val_acc: 0.6320\n",
      "Epoch 4/100\n",
      " - 1s - loss: 0.6506 - acc: 0.6219 - val_loss: 0.6379 - val_acc: 0.6515\n",
      "Epoch 5/100\n",
      " - 1s - loss: 0.6418 - acc: 0.6372 - val_loss: 0.6335 - val_acc: 0.6560\n",
      "Epoch 6/100\n",
      " - 1s - loss: 0.6314 - acc: 0.6477 - val_loss: 0.6242 - val_acc: 0.6540\n",
      "Epoch 7/100\n",
      " - 1s - loss: 0.6301 - acc: 0.6484 - val_loss: 0.6228 - val_acc: 0.6540\n",
      "Epoch 8/100\n",
      " - 1s - loss: 0.6241 - acc: 0.6550 - val_loss: 0.6180 - val_acc: 0.6655\n",
      "Epoch 9/100\n",
      " - 1s - loss: 0.6205 - acc: 0.6563 - val_loss: 0.6144 - val_acc: 0.6670\n",
      "Epoch 10/100\n",
      " - 1s - loss: 0.6163 - acc: 0.6603 - val_loss: 0.6114 - val_acc: 0.6745\n",
      "Epoch 11/100\n",
      " - 1s - loss: 0.6130 - acc: 0.6668 - val_loss: 0.6093 - val_acc: 0.6740\n",
      "Epoch 12/100\n",
      " - 1s - loss: 0.6115 - acc: 0.6671 - val_loss: 0.6095 - val_acc: 0.6735\n",
      "Epoch 13/100\n",
      " - 1s - loss: 0.6066 - acc: 0.6735 - val_loss: 0.6033 - val_acc: 0.6780\n",
      "Epoch 14/100\n",
      " - 1s - loss: 0.6050 - acc: 0.6719 - val_loss: 0.6051 - val_acc: 0.6790\n",
      "Epoch 15/100\n",
      " - 1s - loss: 0.5993 - acc: 0.6745 - val_loss: 0.6002 - val_acc: 0.6780\n",
      "Epoch 16/100\n",
      " - 1s - loss: 0.5996 - acc: 0.6782 - val_loss: 0.5992 - val_acc: 0.6850\n",
      "Epoch 17/100\n",
      " - 1s - loss: 0.5955 - acc: 0.6807 - val_loss: 0.5953 - val_acc: 0.6885\n",
      "Epoch 18/100\n",
      " - 1s - loss: 0.5913 - acc: 0.6902 - val_loss: 0.5924 - val_acc: 0.6820\n",
      "Epoch 19/100\n",
      " - 1s - loss: 0.5920 - acc: 0.6858 - val_loss: 0.5918 - val_acc: 0.6850\n",
      "Epoch 20/100\n",
      " - 1s - loss: 0.5893 - acc: 0.6899 - val_loss: 0.5868 - val_acc: 0.6875\n",
      "Epoch 21/100\n",
      " - 1s - loss: 0.5850 - acc: 0.6918 - val_loss: 0.5881 - val_acc: 0.6850\n",
      "Epoch 22/100\n",
      " - 1s - loss: 0.5819 - acc: 0.6932 - val_loss: 0.5834 - val_acc: 0.6910\n",
      "Epoch 23/100\n",
      " - 1s - loss: 0.5794 - acc: 0.6927 - val_loss: 0.5833 - val_acc: 0.6910\n",
      "Epoch 24/100\n",
      " - 1s - loss: 0.5813 - acc: 0.6952 - val_loss: 0.5847 - val_acc: 0.6975\n",
      "Epoch 25/100\n",
      " - 1s - loss: 0.5780 - acc: 0.6972 - val_loss: 0.5787 - val_acc: 0.6945\n",
      "Epoch 26/100\n",
      " - 1s - loss: 0.5715 - acc: 0.7030 - val_loss: 0.5838 - val_acc: 0.6875\n",
      "Epoch 27/100\n",
      " - 1s - loss: 0.5731 - acc: 0.7000 - val_loss: 0.5781 - val_acc: 0.7025\n",
      "Epoch 28/100\n",
      " - 1s - loss: 0.5714 - acc: 0.7024 - val_loss: 0.5758 - val_acc: 0.7075\n",
      "Epoch 29/100\n",
      " - 1s - loss: 0.5754 - acc: 0.6993 - val_loss: 0.5728 - val_acc: 0.7065\n",
      "Epoch 30/100\n",
      " - 1s - loss: 0.5691 - acc: 0.7054 - val_loss: 0.5723 - val_acc: 0.7085\n",
      "Epoch 31/100\n",
      " - 1s - loss: 0.5708 - acc: 0.6974 - val_loss: 0.5734 - val_acc: 0.7060\n",
      "Epoch 32/100\n",
      " - 1s - loss: 0.5642 - acc: 0.7103 - val_loss: 0.5750 - val_acc: 0.7000\n",
      "Epoch 33/100\n",
      " - 1s - loss: 0.5664 - acc: 0.7095 - val_loss: 0.5694 - val_acc: 0.7070\n",
      "Epoch 34/100\n",
      " - 1s - loss: 0.5623 - acc: 0.7101 - val_loss: 0.5696 - val_acc: 0.7110\n",
      "Epoch 35/100\n",
      " - 1s - loss: 0.5614 - acc: 0.7107 - val_loss: 0.5805 - val_acc: 0.6850\n",
      "Epoch 36/100\n",
      " - 1s - loss: 0.5596 - acc: 0.7148 - val_loss: 0.5679 - val_acc: 0.7100\n",
      "Epoch 37/100\n",
      " - 1s - loss: 0.5568 - acc: 0.7188 - val_loss: 0.5652 - val_acc: 0.7095\n",
      "Epoch 38/100\n",
      " - 1s - loss: 0.5572 - acc: 0.7136 - val_loss: 0.5678 - val_acc: 0.7105\n",
      "Epoch 39/100\n",
      " - 1s - loss: 0.5572 - acc: 0.7143 - val_loss: 0.5662 - val_acc: 0.7095\n",
      "Epoch 40/100\n",
      " - 1s - loss: 0.5544 - acc: 0.7193 - val_loss: 0.5646 - val_acc: 0.7105\n",
      "Epoch 41/100\n",
      " - 1s - loss: 0.5508 - acc: 0.7161 - val_loss: 0.5671 - val_acc: 0.6975\n",
      "Epoch 42/100\n",
      " - 1s - loss: 0.5521 - acc: 0.7228 - val_loss: 0.5691 - val_acc: 0.6960\n",
      "Epoch 43/100\n",
      " - 1s - loss: 0.5508 - acc: 0.7198 - val_loss: 0.5631 - val_acc: 0.7070\n",
      "Epoch 44/100\n",
      " - 1s - loss: 0.5506 - acc: 0.7163 - val_loss: 0.5626 - val_acc: 0.7045\n",
      "Epoch 45/100\n",
      " - 1s - loss: 0.5469 - acc: 0.7240 - val_loss: 0.5638 - val_acc: 0.7060\n",
      "Epoch 46/100\n",
      " - 1s - loss: 0.5485 - acc: 0.7152 - val_loss: 0.5721 - val_acc: 0.6955\n",
      "Epoch 47/100\n",
      " - 1s - loss: 0.5462 - acc: 0.7222 - val_loss: 0.5698 - val_acc: 0.6990\n",
      "Epoch 48/100\n",
      " - 1s - loss: 0.5437 - acc: 0.7187 - val_loss: 0.5607 - val_acc: 0.7090\n",
      "Epoch 49/100\n",
      " - 1s - loss: 0.5438 - acc: 0.7276 - val_loss: 0.5615 - val_acc: 0.7145\n",
      "Epoch 50/100\n",
      " - 1s - loss: 0.5376 - acc: 0.7282 - val_loss: 0.5638 - val_acc: 0.7080\n",
      "Epoch 51/100\n",
      " - 1s - loss: 0.5438 - acc: 0.7231 - val_loss: 0.5629 - val_acc: 0.7130\n",
      "Epoch 52/100\n",
      " - 1s - loss: 0.5421 - acc: 0.7296 - val_loss: 0.5589 - val_acc: 0.7085\n",
      "Epoch 53/100\n",
      " - 1s - loss: 0.5399 - acc: 0.7287 - val_loss: 0.5572 - val_acc: 0.7095\n",
      "Epoch 54/100\n",
      " - 1s - loss: 0.5398 - acc: 0.7224 - val_loss: 0.5572 - val_acc: 0.7135\n",
      "Epoch 55/100\n",
      " - 1s - loss: 0.5389 - acc: 0.7282 - val_loss: 0.5593 - val_acc: 0.7125\n",
      "Epoch 56/100\n",
      " - 1s - loss: 0.5381 - acc: 0.7315 - val_loss: 0.5575 - val_acc: 0.7145\n",
      "Epoch 57/100\n",
      " - 1s - loss: 0.5384 - acc: 0.7226 - val_loss: 0.5578 - val_acc: 0.7100\n",
      "Epoch 58/100\n",
      " - 1s - loss: 0.5345 - acc: 0.7324 - val_loss: 0.5569 - val_acc: 0.7135\n",
      "Epoch 59/100\n",
      " - 1s - loss: 0.5314 - acc: 0.7324 - val_loss: 0.5578 - val_acc: 0.7115\n",
      "Epoch 60/100\n",
      " - 1s - loss: 0.5324 - acc: 0.7268 - val_loss: 0.5559 - val_acc: 0.7135\n",
      "Epoch 61/100\n",
      " - 1s - loss: 0.5307 - acc: 0.7337 - val_loss: 0.5580 - val_acc: 0.7100\n",
      "Epoch 62/100\n",
      " - 1s - loss: 0.5302 - acc: 0.7371 - val_loss: 0.5555 - val_acc: 0.7135\n",
      "Epoch 63/100\n",
      " - 1s - loss: 0.5308 - acc: 0.7307 - val_loss: 0.5562 - val_acc: 0.7150\n",
      "Epoch 64/100\n",
      " - 1s - loss: 0.5271 - acc: 0.7335 - val_loss: 0.5597 - val_acc: 0.7070\n",
      "Epoch 65/100\n",
      " - 1s - loss: 0.5292 - acc: 0.7312 - val_loss: 0.5556 - val_acc: 0.7180\n",
      "Epoch 66/100\n",
      " - 1s - loss: 0.5309 - acc: 0.7308 - val_loss: 0.5549 - val_acc: 0.7165\n",
      "Epoch 67/100\n",
      " - 1s - loss: 0.5298 - acc: 0.7333 - val_loss: 0.5551 - val_acc: 0.7155\n",
      "Epoch 68/100\n",
      " - 1s - loss: 0.5272 - acc: 0.7332 - val_loss: 0.5535 - val_acc: 0.7125\n",
      "Epoch 69/100\n",
      " - 1s - loss: 0.5275 - acc: 0.7344 - val_loss: 0.5552 - val_acc: 0.7180\n",
      "Epoch 70/100\n",
      " - 1s - loss: 0.5250 - acc: 0.7373 - val_loss: 0.5551 - val_acc: 0.7140\n",
      "Epoch 71/100\n",
      " - 1s - loss: 0.5255 - acc: 0.7392 - val_loss: 0.5559 - val_acc: 0.7170\n",
      "Epoch 72/100\n",
      " - 1s - loss: 0.5224 - acc: 0.7384 - val_loss: 0.5567 - val_acc: 0.7105\n",
      "Epoch 73/100\n",
      " - 1s - loss: 0.5265 - acc: 0.7330 - val_loss: 0.5551 - val_acc: 0.7140\n",
      "Epoch 74/100\n",
      " - 1s - loss: 0.5233 - acc: 0.7387 - val_loss: 0.5561 - val_acc: 0.7080\n",
      "Epoch 75/100\n",
      " - 1s - loss: 0.5174 - acc: 0.7419 - val_loss: 0.5546 - val_acc: 0.7130\n",
      "Epoch 76/100\n",
      " - 1s - loss: 0.5186 - acc: 0.7355 - val_loss: 0.5557 - val_acc: 0.7105\n",
      "Epoch 77/100\n",
      " - 1s - loss: 0.5166 - acc: 0.7386 - val_loss: 0.5539 - val_acc: 0.7130\n",
      "Epoch 78/100\n",
      " - 1s - loss: 0.5225 - acc: 0.7375 - val_loss: 0.5556 - val_acc: 0.7120\n",
      "Epoch 79/100\n",
      " - 1s - loss: 0.5170 - acc: 0.7415 - val_loss: 0.5552 - val_acc: 0.7105\n",
      "Epoch 80/100\n",
      " - 1s - loss: 0.5185 - acc: 0.7372 - val_loss: 0.5541 - val_acc: 0.7105\n",
      "Epoch 81/100\n",
      " - 1s - loss: 0.5149 - acc: 0.7432 - val_loss: 0.5533 - val_acc: 0.7100\n",
      "Epoch 82/100\n",
      " - 1s - loss: 0.5144 - acc: 0.7404 - val_loss: 0.5555 - val_acc: 0.7150\n",
      "Epoch 83/100\n",
      " - 1s - loss: 0.5163 - acc: 0.7379 - val_loss: 0.5542 - val_acc: 0.7120\n",
      "Epoch 84/100\n",
      " - 1s - loss: 0.5170 - acc: 0.7445 - val_loss: 0.5548 - val_acc: 0.7095\n",
      "Epoch 85/100\n",
      " - 1s - loss: 0.5169 - acc: 0.7395 - val_loss: 0.5535 - val_acc: 0.7145\n",
      "Epoch 86/100\n",
      " - 1s - loss: 0.5122 - acc: 0.7431 - val_loss: 0.5532 - val_acc: 0.7170\n",
      "Epoch 87/100\n",
      " - 1s - loss: 0.5157 - acc: 0.7431 - val_loss: 0.5542 - val_acc: 0.7085\n",
      "Epoch 88/100\n",
      " - 1s - loss: 0.5189 - acc: 0.7425 - val_loss: 0.5524 - val_acc: 0.7145\n",
      "Epoch 89/100\n",
      " - 1s - loss: 0.5151 - acc: 0.7409 - val_loss: 0.5540 - val_acc: 0.7105\n",
      "Epoch 90/100\n",
      " - 1s - loss: 0.5134 - acc: 0.7450 - val_loss: 0.5542 - val_acc: 0.7105\n",
      "Epoch 91/100\n",
      " - 1s - loss: 0.5161 - acc: 0.7455 - val_loss: 0.5538 - val_acc: 0.7060\n",
      "Epoch 92/100\n",
      " - 1s - loss: 0.5116 - acc: 0.7507 - val_loss: 0.5552 - val_acc: 0.7055\n",
      "Epoch 93/100\n",
      " - 1s - loss: 0.5148 - acc: 0.7495 - val_loss: 0.5518 - val_acc: 0.7120\n",
      "Epoch 94/100\n",
      " - 1s - loss: 0.5114 - acc: 0.7416 - val_loss: 0.5515 - val_acc: 0.7135\n",
      "Epoch 95/100\n",
      " - 1s - loss: 0.5162 - acc: 0.7507 - val_loss: 0.5519 - val_acc: 0.7090\n",
      "Epoch 96/100\n",
      " - 1s - loss: 0.5161 - acc: 0.7445 - val_loss: 0.5555 - val_acc: 0.7060\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 97/100\n",
      " - 1s - loss: 0.5097 - acc: 0.7514 - val_loss: 0.5509 - val_acc: 0.7155\n",
      "Epoch 98/100\n",
      " - 1s - loss: 0.5086 - acc: 0.7429 - val_loss: 0.5504 - val_acc: 0.7160\n",
      "Epoch 99/100\n",
      " - 1s - loss: 0.5100 - acc: 0.7455 - val_loss: 0.5501 - val_acc: 0.7175\n",
      "Epoch 100/100\n",
      " - 1s - loss: 0.5048 - acc: 0.7505 - val_loss: 0.5520 - val_acc: 0.7080\n",
      "Train on 10000 samples, validate on 2000 samples\n",
      "Epoch 1/100\n",
      " - 1s - loss: 0.6872 - acc: 0.5413 - val_loss: 0.6696 - val_acc: 0.5990\n",
      "Epoch 2/100\n",
      " - 1s - loss: 0.6699 - acc: 0.5874 - val_loss: 0.6682 - val_acc: 0.5895\n",
      "Epoch 3/100\n",
      " - 1s - loss: 0.6571 - acc: 0.6130 - val_loss: 0.6424 - val_acc: 0.6355\n",
      "Epoch 4/100\n",
      " - 1s - loss: 0.6438 - acc: 0.6343 - val_loss: 0.6327 - val_acc: 0.6470\n",
      "Epoch 5/100\n",
      " - 1s - loss: 0.6350 - acc: 0.6425 - val_loss: 0.6324 - val_acc: 0.6385\n",
      "Epoch 6/100\n",
      " - 1s - loss: 0.6301 - acc: 0.6500 - val_loss: 0.6116 - val_acc: 0.6655\n",
      "Epoch 7/100\n",
      " - 1s - loss: 0.6198 - acc: 0.6542 - val_loss: 0.6073 - val_acc: 0.6780\n",
      "Epoch 8/100\n",
      " - 1s - loss: 0.6159 - acc: 0.6609 - val_loss: 0.6010 - val_acc: 0.6775\n",
      "Epoch 9/100\n",
      " - 1s - loss: 0.6126 - acc: 0.6635 - val_loss: 0.6139 - val_acc: 0.6595\n",
      "Epoch 10/100\n",
      " - 1s - loss: 0.6060 - acc: 0.6698 - val_loss: 0.5943 - val_acc: 0.6860\n",
      "Epoch 11/100\n",
      " - 1s - loss: 0.6035 - acc: 0.6752 - val_loss: 0.5966 - val_acc: 0.6785\n",
      "Epoch 12/100\n",
      " - 1s - loss: 0.5985 - acc: 0.6750 - val_loss: 0.5925 - val_acc: 0.6945\n",
      "Epoch 13/100\n",
      " - 1s - loss: 0.5894 - acc: 0.6861 - val_loss: 0.5908 - val_acc: 0.6855\n",
      "Epoch 14/100\n",
      " - 1s - loss: 0.5889 - acc: 0.6845 - val_loss: 0.5869 - val_acc: 0.6835\n",
      "Epoch 15/100\n",
      " - 1s - loss: 0.5872 - acc: 0.6856 - val_loss: 0.5821 - val_acc: 0.6925\n",
      "Epoch 16/100\n",
      " - 1s - loss: 0.5871 - acc: 0.6892 - val_loss: 0.5779 - val_acc: 0.6995\n",
      "Epoch 17/100\n",
      " - 1s - loss: 0.5805 - acc: 0.6862 - val_loss: 0.5742 - val_acc: 0.7005\n",
      "Epoch 18/100\n",
      " - 1s - loss: 0.5748 - acc: 0.6997 - val_loss: 0.5724 - val_acc: 0.7005\n",
      "Epoch 19/100\n",
      " - 1s - loss: 0.5695 - acc: 0.6991 - val_loss: 0.5736 - val_acc: 0.6920\n",
      "Epoch 20/100\n",
      " - 1s - loss: 0.5697 - acc: 0.6988 - val_loss: 0.5691 - val_acc: 0.7085\n",
      "Epoch 21/100\n",
      " - 1s - loss: 0.5711 - acc: 0.7047 - val_loss: 0.5760 - val_acc: 0.6930\n",
      "Epoch 22/100\n",
      " - 1s - loss: 0.5681 - acc: 0.6995 - val_loss: 0.5683 - val_acc: 0.7025\n",
      "Epoch 23/100\n",
      " - 1s - loss: 0.5647 - acc: 0.7014 - val_loss: 0.5708 - val_acc: 0.6865\n",
      "Epoch 24/100\n",
      " - 1s - loss: 0.5595 - acc: 0.7066 - val_loss: 0.5658 - val_acc: 0.7035\n",
      "Epoch 25/100\n",
      " - 1s - loss: 0.5600 - acc: 0.7047 - val_loss: 0.5631 - val_acc: 0.7070\n",
      "Epoch 26/100\n",
      " - 1s - loss: 0.5539 - acc: 0.7156 - val_loss: 0.5627 - val_acc: 0.7015\n",
      "Epoch 27/100\n",
      " - 1s - loss: 0.5526 - acc: 0.7124 - val_loss: 0.5680 - val_acc: 0.6970\n",
      "Epoch 28/100\n",
      " - 1s - loss: 0.5519 - acc: 0.7107 - val_loss: 0.5618 - val_acc: 0.6995\n",
      "Epoch 29/100\n",
      " - 1s - loss: 0.5484 - acc: 0.7152 - val_loss: 0.5633 - val_acc: 0.6945\n",
      "Epoch 30/100\n",
      " - 1s - loss: 0.5468 - acc: 0.7176 - val_loss: 0.5649 - val_acc: 0.6900\n",
      "Epoch 31/100\n",
      " - 1s - loss: 0.5457 - acc: 0.7218 - val_loss: 0.5592 - val_acc: 0.6960\n",
      "Epoch 32/100\n",
      " - 1s - loss: 0.5478 - acc: 0.7132 - val_loss: 0.5702 - val_acc: 0.6970\n",
      "Epoch 33/100\n",
      " - 1s - loss: 0.5444 - acc: 0.7159 - val_loss: 0.5690 - val_acc: 0.6990\n",
      "Epoch 34/100\n",
      " - 1s - loss: 0.5433 - acc: 0.7238 - val_loss: 0.5555 - val_acc: 0.7040\n",
      "Epoch 35/100\n",
      " - 1s - loss: 0.5377 - acc: 0.7232 - val_loss: 0.5742 - val_acc: 0.6965\n",
      "Epoch 36/100\n",
      " - 1s - loss: 0.5337 - acc: 0.7253 - val_loss: 0.5636 - val_acc: 0.7075\n",
      "Epoch 37/100\n",
      " - 1s - loss: 0.5325 - acc: 0.7265 - val_loss: 0.5625 - val_acc: 0.7050\n",
      "Epoch 38/100\n",
      " - 1s - loss: 0.5306 - acc: 0.7313 - val_loss: 0.5543 - val_acc: 0.7110\n",
      "Epoch 39/100\n",
      " - 1s - loss: 0.5285 - acc: 0.7312 - val_loss: 0.5577 - val_acc: 0.7055\n",
      "Epoch 40/100\n",
      " - 1s - loss: 0.5286 - acc: 0.7306 - val_loss: 0.5619 - val_acc: 0.7065\n",
      "Epoch 41/100\n",
      " - 1s - loss: 0.5298 - acc: 0.7286 - val_loss: 0.5584 - val_acc: 0.6935\n",
      "Epoch 42/100\n",
      " - 1s - loss: 0.5214 - acc: 0.7351 - val_loss: 0.5553 - val_acc: 0.7155\n",
      "Epoch 43/100\n",
      " - 1s - loss: 0.5243 - acc: 0.7385 - val_loss: 0.5573 - val_acc: 0.7100\n",
      "Epoch 44/100\n",
      " - 1s - loss: 0.5228 - acc: 0.7350 - val_loss: 0.5526 - val_acc: 0.7070\n",
      "Epoch 45/100\n",
      " - 1s - loss: 0.5217 - acc: 0.7302 - val_loss: 0.5507 - val_acc: 0.7160\n",
      "Epoch 46/100\n",
      " - 1s - loss: 0.5177 - acc: 0.7405 - val_loss: 0.5540 - val_acc: 0.7125\n",
      "Epoch 47/100\n",
      " - 1s - loss: 0.5210 - acc: 0.7370 - val_loss: 0.5526 - val_acc: 0.7110\n",
      "Epoch 48/100\n",
      " - 1s - loss: 0.5192 - acc: 0.7435 - val_loss: 0.5575 - val_acc: 0.6990\n",
      "Epoch 49/100\n",
      " - 1s - loss: 0.5150 - acc: 0.7427 - val_loss: 0.5602 - val_acc: 0.7165\n",
      "Epoch 50/100\n",
      " - 1s - loss: 0.5166 - acc: 0.7341 - val_loss: 0.5483 - val_acc: 0.7100\n",
      "Epoch 51/100\n",
      " - 1s - loss: 0.5163 - acc: 0.7389 - val_loss: 0.5538 - val_acc: 0.7085\n",
      "Epoch 52/100\n",
      " - 1s - loss: 0.5143 - acc: 0.7371 - val_loss: 0.5511 - val_acc: 0.7165\n",
      "Epoch 53/100\n",
      " - 1s - loss: 0.5108 - acc: 0.7448 - val_loss: 0.5493 - val_acc: 0.7160\n",
      "Epoch 54/100\n",
      " - 1s - loss: 0.5149 - acc: 0.7373 - val_loss: 0.5522 - val_acc: 0.7165\n",
      "Epoch 55/100\n",
      " - 1s - loss: 0.5125 - acc: 0.7395 - val_loss: 0.5572 - val_acc: 0.7190\n",
      "Epoch 56/100\n",
      " - 1s - loss: 0.5120 - acc: 0.7388 - val_loss: 0.5561 - val_acc: 0.7085\n",
      "Epoch 57/100\n",
      " - 1s - loss: 0.5118 - acc: 0.7460 - val_loss: 0.5500 - val_acc: 0.7160\n",
      "Epoch 58/100\n",
      " - 1s - loss: 0.5055 - acc: 0.7466 - val_loss: 0.5481 - val_acc: 0.7080\n",
      "Epoch 59/100\n",
      " - 1s - loss: 0.5076 - acc: 0.7464 - val_loss: 0.5571 - val_acc: 0.7150\n",
      "Epoch 60/100\n",
      " - 1s - loss: 0.5004 - acc: 0.7490 - val_loss: 0.5507 - val_acc: 0.7190\n",
      "Epoch 61/100\n",
      " - 1s - loss: 0.4972 - acc: 0.7453 - val_loss: 0.5574 - val_acc: 0.7190\n",
      "Epoch 62/100\n",
      " - 1s - loss: 0.4938 - acc: 0.7543 - val_loss: 0.5541 - val_acc: 0.7190\n",
      "Epoch 63/100\n",
      " - 1s - loss: 0.4972 - acc: 0.7509 - val_loss: 0.5497 - val_acc: 0.7105\n",
      "Epoch 64/100\n",
      " - 1s - loss: 0.4983 - acc: 0.7501 - val_loss: 0.5513 - val_acc: 0.7185\n",
      "Epoch 65/100\n",
      " - 1s - loss: 0.4978 - acc: 0.7461 - val_loss: 0.5502 - val_acc: 0.7200\n",
      "Epoch 66/100\n",
      " - 1s - loss: 0.5029 - acc: 0.7519 - val_loss: 0.5554 - val_acc: 0.6985\n",
      "Epoch 67/100\n",
      " - 1s - loss: 0.4945 - acc: 0.7495 - val_loss: 0.5508 - val_acc: 0.7090\n",
      "Epoch 68/100\n",
      " - 1s - loss: 0.4970 - acc: 0.7523 - val_loss: 0.5480 - val_acc: 0.7150\n",
      "Epoch 69/100\n",
      " - 1s - loss: 0.4854 - acc: 0.7555 - val_loss: 0.5500 - val_acc: 0.7210\n",
      "Epoch 70/100\n",
      " - 1s - loss: 0.4894 - acc: 0.7570 - val_loss: 0.5523 - val_acc: 0.7180\n",
      "Epoch 71/100\n",
      " - 1s - loss: 0.4909 - acc: 0.7553 - val_loss: 0.5485 - val_acc: 0.7140\n",
      "Epoch 72/100\n",
      " - 1s - loss: 0.4887 - acc: 0.7584 - val_loss: 0.5469 - val_acc: 0.7215\n",
      "Epoch 73/100\n",
      " - 1s - loss: 0.4910 - acc: 0.7542 - val_loss: 0.5491 - val_acc: 0.7175\n",
      "Epoch 74/100\n",
      " - 1s - loss: 0.4852 - acc: 0.7594 - val_loss: 0.5579 - val_acc: 0.7175\n",
      "Epoch 75/100\n",
      " - 1s - loss: 0.4865 - acc: 0.7529 - val_loss: 0.5520 - val_acc: 0.7210\n",
      "Epoch 76/100\n",
      " - 1s - loss: 0.4837 - acc: 0.7592 - val_loss: 0.5620 - val_acc: 0.7170\n",
      "Epoch 77/100\n",
      " - 1s - loss: 0.4887 - acc: 0.7537 - val_loss: 0.5488 - val_acc: 0.7215\n",
      "Epoch 78/100\n",
      " - 1s - loss: 0.4821 - acc: 0.7584 - val_loss: 0.5566 - val_acc: 0.7155\n",
      "Epoch 79/100\n",
      " - 1s - loss: 0.4790 - acc: 0.7553 - val_loss: 0.5628 - val_acc: 0.7075\n",
      "Epoch 80/100\n",
      " - 1s - loss: 0.4764 - acc: 0.7596 - val_loss: 0.5474 - val_acc: 0.7200\n",
      "Epoch 81/100\n",
      " - 1s - loss: 0.4844 - acc: 0.7556 - val_loss: 0.5604 - val_acc: 0.7150\n",
      "Epoch 82/100\n",
      " - 1s - loss: 0.4838 - acc: 0.7560 - val_loss: 0.5459 - val_acc: 0.7150\n",
      "Epoch 83/100\n",
      " - 1s - loss: 0.4839 - acc: 0.7568 - val_loss: 0.5538 - val_acc: 0.7255\n",
      "Epoch 84/100\n",
      " - 1s - loss: 0.4749 - acc: 0.7600 - val_loss: 0.5510 - val_acc: 0.7305\n",
      "Epoch 85/100\n",
      " - 1s - loss: 0.4797 - acc: 0.7646 - val_loss: 0.5453 - val_acc: 0.7195\n",
      "Epoch 86/100\n",
      " - 1s - loss: 0.4816 - acc: 0.7594 - val_loss: 0.5525 - val_acc: 0.7250\n",
      "Epoch 87/100\n",
      " - 1s - loss: 0.4721 - acc: 0.7616 - val_loss: 0.5517 - val_acc: 0.7205\n",
      "Epoch 88/100\n",
      " - 1s - loss: 0.4720 - acc: 0.7632 - val_loss: 0.5529 - val_acc: 0.7200\n",
      "Epoch 89/100\n",
      " - 1s - loss: 0.4714 - acc: 0.7604 - val_loss: 0.5652 - val_acc: 0.7225\n",
      "Epoch 90/100\n",
      " - 1s - loss: 0.4712 - acc: 0.7607 - val_loss: 0.5505 - val_acc: 0.7205\n",
      "Epoch 91/100\n",
      " - 1s - loss: 0.4680 - acc: 0.7680 - val_loss: 0.5458 - val_acc: 0.7235\n",
      "Epoch 92/100\n",
      " - 1s - loss: 0.4739 - acc: 0.7646 - val_loss: 0.5509 - val_acc: 0.7280\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 93/100\n",
      " - 1s - loss: 0.4709 - acc: 0.7624 - val_loss: 0.5502 - val_acc: 0.7205\n",
      "Epoch 94/100\n",
      " - 1s - loss: 0.4719 - acc: 0.7590 - val_loss: 0.5622 - val_acc: 0.7135\n",
      "Epoch 95/100\n",
      " - 1s - loss: 0.4694 - acc: 0.7633 - val_loss: 0.5470 - val_acc: 0.7225\n",
      "Epoch 96/100\n",
      " - 1s - loss: 0.4673 - acc: 0.7703 - val_loss: 0.5668 - val_acc: 0.7180\n",
      "Epoch 97/100\n",
      " - 1s - loss: 0.4606 - acc: 0.7698 - val_loss: 0.5516 - val_acc: 0.7200\n",
      "Epoch 98/100\n",
      " - 1s - loss: 0.4632 - acc: 0.7727 - val_loss: 0.5550 - val_acc: 0.7225\n",
      "Epoch 99/100\n",
      " - 1s - loss: 0.4646 - acc: 0.7694 - val_loss: 0.5442 - val_acc: 0.7280\n",
      "Epoch 100/100\n",
      " - 1s - loss: 0.4669 - acc: 0.7653 - val_loss: 0.5488 - val_acc: 0.7310\n",
      "Train on 10000 samples, validate on 2000 samples\n",
      "Epoch 1/100\n",
      " - 1s - loss: 0.6951 - acc: 0.5260 - val_loss: 0.6810 - val_acc: 0.5870\n",
      "Epoch 2/100\n",
      " - 1s - loss: 0.6770 - acc: 0.5673 - val_loss: 0.6682 - val_acc: 0.6050\n",
      "Epoch 3/100\n",
      " - 1s - loss: 0.6699 - acc: 0.5724 - val_loss: 0.6565 - val_acc: 0.6215\n",
      "Epoch 4/100\n",
      " - 1s - loss: 0.6582 - acc: 0.5909 - val_loss: 0.6494 - val_acc: 0.6225\n",
      "Epoch 5/100\n",
      " - 1s - loss: 0.6487 - acc: 0.6112 - val_loss: 0.6379 - val_acc: 0.6385\n",
      "Epoch 6/100\n",
      " - 1s - loss: 0.6397 - acc: 0.6295 - val_loss: 0.6374 - val_acc: 0.6255\n",
      "Epoch 7/100\n",
      " - 1s - loss: 0.6271 - acc: 0.6490 - val_loss: 0.6134 - val_acc: 0.6620\n",
      "Epoch 8/100\n",
      " - 1s - loss: 0.6155 - acc: 0.6609 - val_loss: 0.6040 - val_acc: 0.6705\n",
      "Epoch 9/100\n",
      " - 1s - loss: 0.6044 - acc: 0.6766 - val_loss: 0.6215 - val_acc: 0.6405\n",
      "Epoch 10/100\n",
      " - 1s - loss: 0.5983 - acc: 0.6795 - val_loss: 0.5929 - val_acc: 0.6700\n",
      "Epoch 11/100\n",
      " - 1s - loss: 0.5900 - acc: 0.6843 - val_loss: 0.6338 - val_acc: 0.6370\n",
      "Epoch 12/100\n",
      " - 1s - loss: 0.5827 - acc: 0.6916 - val_loss: 0.5837 - val_acc: 0.6885\n",
      "Epoch 13/100\n",
      " - 1s - loss: 0.5728 - acc: 0.6957 - val_loss: 0.5714 - val_acc: 0.6990\n",
      "Epoch 14/100\n",
      " - 1s - loss: 0.5668 - acc: 0.7058 - val_loss: 0.6063 - val_acc: 0.6725\n",
      "Epoch 15/100\n",
      " - 1s - loss: 0.5572 - acc: 0.7093 - val_loss: 0.5895 - val_acc: 0.6785\n",
      "Epoch 16/100\n",
      " - 1s - loss: 0.5600 - acc: 0.7115 - val_loss: 0.5707 - val_acc: 0.6960\n",
      "Epoch 17/100\n",
      " - 1s - loss: 0.5494 - acc: 0.7093 - val_loss: 0.5588 - val_acc: 0.7090\n",
      "Epoch 18/100\n",
      " - 1s - loss: 0.5466 - acc: 0.7169 - val_loss: 0.5913 - val_acc: 0.6765\n",
      "Epoch 19/100\n",
      " - 1s - loss: 0.5461 - acc: 0.7166 - val_loss: 0.5611 - val_acc: 0.7045\n",
      "Epoch 20/100\n",
      " - 1s - loss: 0.5352 - acc: 0.7256 - val_loss: 0.5749 - val_acc: 0.6885\n",
      "Epoch 21/100\n",
      " - 1s - loss: 0.5341 - acc: 0.7253 - val_loss: 0.5730 - val_acc: 0.6885\n",
      "Epoch 22/100\n",
      " - 1s - loss: 0.5289 - acc: 0.7345 - val_loss: 0.5733 - val_acc: 0.6950\n",
      "Epoch 23/100\n",
      " - 1s - loss: 0.5267 - acc: 0.7326 - val_loss: 0.5497 - val_acc: 0.7125\n",
      "Epoch 24/100\n",
      " - 1s - loss: 0.5249 - acc: 0.7337 - val_loss: 0.5498 - val_acc: 0.7160\n",
      "Epoch 25/100\n",
      " - 1s - loss: 0.5170 - acc: 0.7376 - val_loss: 0.5453 - val_acc: 0.7220\n",
      "Epoch 26/100\n",
      " - 1s - loss: 0.5238 - acc: 0.7314 - val_loss: 0.5555 - val_acc: 0.7145\n",
      "Epoch 27/100\n",
      " - 1s - loss: 0.5198 - acc: 0.7365 - val_loss: 0.5732 - val_acc: 0.6940\n",
      "Epoch 28/100\n",
      " - 1s - loss: 0.5163 - acc: 0.7374 - val_loss: 0.5526 - val_acc: 0.7170\n",
      "Epoch 29/100\n",
      " - 1s - loss: 0.5115 - acc: 0.7400 - val_loss: 0.5544 - val_acc: 0.7055\n",
      "Epoch 30/100\n",
      " - 1s - loss: 0.5113 - acc: 0.7432 - val_loss: 0.5593 - val_acc: 0.7090\n",
      "Epoch 31/100\n",
      " - 1s - loss: 0.5076 - acc: 0.7440 - val_loss: 0.5684 - val_acc: 0.7135\n",
      "Epoch 32/100\n",
      " - 1s - loss: 0.5083 - acc: 0.7434 - val_loss: 0.5733 - val_acc: 0.7070\n",
      "Epoch 33/100\n",
      " - 1s - loss: 0.5025 - acc: 0.7450 - val_loss: 0.6323 - val_acc: 0.6785\n",
      "Epoch 34/100\n",
      " - 1s - loss: 0.4975 - acc: 0.7547 - val_loss: 0.5846 - val_acc: 0.7020\n",
      "Epoch 35/100\n",
      " - 1s - loss: 0.4970 - acc: 0.7541 - val_loss: 0.6136 - val_acc: 0.6945\n",
      "Epoch 36/100\n",
      " - 1s - loss: 0.5016 - acc: 0.7485 - val_loss: 0.5481 - val_acc: 0.7160\n",
      "Epoch 37/100\n",
      " - 1s - loss: 0.4960 - acc: 0.7563 - val_loss: 0.5497 - val_acc: 0.7135\n",
      "Epoch 38/100\n",
      " - 1s - loss: 0.4908 - acc: 0.7530 - val_loss: 0.5460 - val_acc: 0.7215\n",
      "Epoch 39/100\n",
      " - 1s - loss: 0.4874 - acc: 0.7560 - val_loss: 0.5563 - val_acc: 0.7165\n",
      "Epoch 40/100\n",
      " - 1s - loss: 0.4855 - acc: 0.7570 - val_loss: 0.5466 - val_acc: 0.7185\n",
      "Epoch 41/100\n",
      " - 1s - loss: 0.4908 - acc: 0.7573 - val_loss: 0.5587 - val_acc: 0.7095\n",
      "Epoch 42/100\n",
      " - 1s - loss: 0.4824 - acc: 0.7604 - val_loss: 0.5499 - val_acc: 0.7195\n",
      "Epoch 43/100\n",
      " - 1s - loss: 0.4856 - acc: 0.7574 - val_loss: 0.5555 - val_acc: 0.7050\n",
      "Epoch 44/100\n",
      " - 1s - loss: 0.4862 - acc: 0.7578 - val_loss: 0.5520 - val_acc: 0.7265\n",
      "Epoch 45/100\n",
      " - 1s - loss: 0.4784 - acc: 0.7566 - val_loss: 0.5592 - val_acc: 0.7145\n",
      "Epoch 46/100\n",
      " - 1s - loss: 0.4786 - acc: 0.7636 - val_loss: 0.5497 - val_acc: 0.7235\n",
      "Epoch 47/100\n",
      " - 1s - loss: 0.4791 - acc: 0.7596 - val_loss: 0.5536 - val_acc: 0.7165\n",
      "Epoch 48/100\n",
      " - 1s - loss: 0.4692 - acc: 0.7660 - val_loss: 0.5866 - val_acc: 0.7125\n",
      "Epoch 49/100\n",
      " - 1s - loss: 0.4701 - acc: 0.7669 - val_loss: 0.5541 - val_acc: 0.7180\n",
      "Epoch 50/100\n",
      " - 1s - loss: 0.4732 - acc: 0.7602 - val_loss: 0.6085 - val_acc: 0.7050\n",
      "Epoch 51/100\n",
      " - 1s - loss: 0.4718 - acc: 0.7654 - val_loss: 0.5775 - val_acc: 0.7120\n",
      "Epoch 52/100\n",
      " - 1s - loss: 0.4671 - acc: 0.7717 - val_loss: 0.5623 - val_acc: 0.7240\n",
      "Epoch 53/100\n",
      " - 1s - loss: 0.4725 - acc: 0.7646 - val_loss: 0.5533 - val_acc: 0.7210\n",
      "Epoch 54/100\n",
      " - 1s - loss: 0.4688 - acc: 0.7675 - val_loss: 0.5589 - val_acc: 0.7165\n",
      "Epoch 55/100\n",
      " - 1s - loss: 0.4637 - acc: 0.7691 - val_loss: 0.5772 - val_acc: 0.7125\n",
      "Epoch 56/100\n",
      " - 1s - loss: 0.4665 - acc: 0.7627 - val_loss: 0.5542 - val_acc: 0.7155\n",
      "Epoch 57/100\n",
      " - 1s - loss: 0.4648 - acc: 0.7678 - val_loss: 0.5767 - val_acc: 0.7175\n",
      "Epoch 58/100\n",
      " - 1s - loss: 0.4628 - acc: 0.7675 - val_loss: 0.5839 - val_acc: 0.6945\n",
      "Epoch 59/100\n",
      " - 1s - loss: 0.4611 - acc: 0.7702 - val_loss: 0.5596 - val_acc: 0.7195\n",
      "Epoch 60/100\n",
      " - 1s - loss: 0.4617 - acc: 0.7667 - val_loss: 0.5548 - val_acc: 0.7180\n",
      "Epoch 61/100\n",
      " - 1s - loss: 0.4613 - acc: 0.7701 - val_loss: 0.5755 - val_acc: 0.7220\n",
      "Epoch 62/100\n",
      " - 1s - loss: 0.4584 - acc: 0.7714 - val_loss: 0.5600 - val_acc: 0.7160\n",
      "Epoch 63/100\n",
      " - 1s - loss: 0.4507 - acc: 0.7722 - val_loss: 0.5750 - val_acc: 0.7110\n",
      "Epoch 64/100\n",
      " - 1s - loss: 0.4551 - acc: 0.7732 - val_loss: 0.5844 - val_acc: 0.7100\n",
      "Epoch 65/100\n",
      " - 1s - loss: 0.4529 - acc: 0.7765 - val_loss: 0.5498 - val_acc: 0.7295\n",
      "Epoch 66/100\n",
      " - 1s - loss: 0.4608 - acc: 0.7670 - val_loss: 0.5882 - val_acc: 0.7110\n",
      "Epoch 67/100\n",
      " - 1s - loss: 0.4527 - acc: 0.7763 - val_loss: 0.5874 - val_acc: 0.7115\n",
      "Epoch 68/100\n",
      " - 1s - loss: 0.4515 - acc: 0.7721 - val_loss: 0.5833 - val_acc: 0.6960\n",
      "Epoch 69/100\n",
      " - 1s - loss: 0.4541 - acc: 0.7687 - val_loss: 0.5696 - val_acc: 0.7185\n",
      "Epoch 70/100\n",
      " - 1s - loss: 0.4604 - acc: 0.7705 - val_loss: 0.5952 - val_acc: 0.7175\n",
      "Epoch 71/100\n",
      " - 1s - loss: 0.4506 - acc: 0.7757 - val_loss: 0.5634 - val_acc: 0.7085\n",
      "Epoch 72/100\n",
      " - 1s - loss: 0.4580 - acc: 0.7694 - val_loss: 0.6142 - val_acc: 0.7170\n",
      "Epoch 73/100\n",
      " - 1s - loss: 0.4523 - acc: 0.7758 - val_loss: 0.5748 - val_acc: 0.7095\n",
      "Epoch 74/100\n",
      " - 1s - loss: 0.4580 - acc: 0.7684 - val_loss: 0.5712 - val_acc: 0.7035\n",
      "Epoch 75/100\n",
      " - 1s - loss: 0.4517 - acc: 0.7756 - val_loss: 0.5681 - val_acc: 0.7035\n",
      "Epoch 76/100\n",
      " - 1s - loss: 0.4475 - acc: 0.7775 - val_loss: 0.5943 - val_acc: 0.6955\n",
      "Epoch 77/100\n",
      " - 1s - loss: 0.4532 - acc: 0.7733 - val_loss: 0.5615 - val_acc: 0.7185\n",
      "Epoch 78/100\n",
      " - 1s - loss: 0.4529 - acc: 0.7670 - val_loss: 0.5708 - val_acc: 0.7090\n",
      "Epoch 79/100\n",
      " - 1s - loss: 0.4584 - acc: 0.7722 - val_loss: 0.5575 - val_acc: 0.7235\n",
      "Epoch 80/100\n",
      " - 1s - loss: 0.4543 - acc: 0.7725 - val_loss: 0.5593 - val_acc: 0.7115\n",
      "Epoch 81/100\n",
      " - 1s - loss: 0.4552 - acc: 0.7719 - val_loss: 0.5701 - val_acc: 0.7110\n",
      "Epoch 82/100\n",
      " - 1s - loss: 0.4528 - acc: 0.7676 - val_loss: 0.5613 - val_acc: 0.7270\n",
      "Epoch 83/100\n",
      " - 1s - loss: 0.4508 - acc: 0.7737 - val_loss: 0.5776 - val_acc: 0.7155\n",
      "Epoch 84/100\n",
      " - 1s - loss: 0.4481 - acc: 0.7749 - val_loss: 0.5578 - val_acc: 0.7220\n",
      "Epoch 85/100\n",
      " - 1s - loss: 0.4529 - acc: 0.7775 - val_loss: 0.5578 - val_acc: 0.7075\n",
      "Epoch 86/100\n",
      " - 1s - loss: 0.4447 - acc: 0.7786 - val_loss: 0.5793 - val_acc: 0.7120\n",
      "Epoch 87/100\n",
      " - 1s - loss: 0.4438 - acc: 0.7816 - val_loss: 0.5576 - val_acc: 0.7005\n",
      "Epoch 88/100\n",
      " - 1s - loss: 0.4423 - acc: 0.7740 - val_loss: 0.5607 - val_acc: 0.7015\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 89/100\n",
      " - 1s - loss: 0.4477 - acc: 0.7707 - val_loss: 0.6062 - val_acc: 0.7145\n",
      "Epoch 90/100\n",
      " - 1s - loss: 0.4505 - acc: 0.7792 - val_loss: 0.5890 - val_acc: 0.7185\n",
      "Epoch 91/100\n",
      " - 1s - loss: 0.4482 - acc: 0.7802 - val_loss: 0.6205 - val_acc: 0.7140\n",
      "Epoch 92/100\n",
      " - 1s - loss: 0.4517 - acc: 0.7699 - val_loss: 0.5777 - val_acc: 0.7170\n",
      "Epoch 93/100\n",
      " - 1s - loss: 0.4422 - acc: 0.7819 - val_loss: 0.5744 - val_acc: 0.7070\n",
      "Epoch 94/100\n",
      " - 1s - loss: 0.4446 - acc: 0.7797 - val_loss: 0.5812 - val_acc: 0.7125\n",
      "Epoch 95/100\n",
      " - 1s - loss: 0.4398 - acc: 0.7801 - val_loss: 0.5730 - val_acc: 0.7185\n",
      "Epoch 96/100\n",
      " - 1s - loss: 0.4396 - acc: 0.7735 - val_loss: 0.5920 - val_acc: 0.7170\n",
      "Epoch 97/100\n",
      " - 1s - loss: 0.4471 - acc: 0.7781 - val_loss: 0.5596 - val_acc: 0.7155\n",
      "Epoch 98/100\n",
      " - 1s - loss: 0.4440 - acc: 0.7773 - val_loss: 0.6201 - val_acc: 0.7235\n",
      "Epoch 99/100\n",
      " - 1s - loss: 0.4491 - acc: 0.7762 - val_loss: 0.5695 - val_acc: 0.7060\n",
      "Epoch 100/100\n",
      " - 1s - loss: 0.4475 - acc: 0.7808 - val_loss: 0.5688 - val_acc: 0.7205\n",
      "Train on 10000 samples, validate on 2000 samples\n",
      "Epoch 1/100\n",
      " - 1s - loss: 0.6863 - acc: 0.5486 - val_loss: 0.6801 - val_acc: 0.5960\n",
      "Epoch 2/100\n",
      " - 1s - loss: 0.6669 - acc: 0.5958 - val_loss: 0.6529 - val_acc: 0.6090\n",
      "Epoch 3/100\n",
      " - 1s - loss: 0.6473 - acc: 0.6232 - val_loss: 0.6326 - val_acc: 0.6410\n",
      "Epoch 4/100\n",
      " - 1s - loss: 0.6319 - acc: 0.6432 - val_loss: 0.6224 - val_acc: 0.6485\n",
      "Epoch 5/100\n",
      " - 1s - loss: 0.6256 - acc: 0.6499 - val_loss: 0.6100 - val_acc: 0.6565\n",
      "Epoch 6/100\n",
      " - 1s - loss: 0.6132 - acc: 0.6603 - val_loss: 0.6021 - val_acc: 0.6635\n",
      "Epoch 7/100\n",
      " - 1s - loss: 0.6047 - acc: 0.6707 - val_loss: 0.6006 - val_acc: 0.6840\n",
      "Epoch 8/100\n",
      " - 1s - loss: 0.5954 - acc: 0.6774 - val_loss: 0.5957 - val_acc: 0.6795\n",
      "Epoch 9/100\n",
      " - 1s - loss: 0.5883 - acc: 0.6846 - val_loss: 0.5883 - val_acc: 0.6735\n",
      "Epoch 10/100\n",
      " - 1s - loss: 0.5784 - acc: 0.6820 - val_loss: 0.5759 - val_acc: 0.7015\n",
      "Epoch 11/100\n",
      " - 1s - loss: 0.5771 - acc: 0.6966 - val_loss: 0.5781 - val_acc: 0.6980\n",
      "Epoch 12/100\n",
      " - 1s - loss: 0.5687 - acc: 0.7017 - val_loss: 0.5748 - val_acc: 0.7000\n",
      "Epoch 13/100\n",
      " - 1s - loss: 0.5569 - acc: 0.7034 - val_loss: 0.5661 - val_acc: 0.7155\n",
      "Epoch 14/100\n",
      " - 1s - loss: 0.5528 - acc: 0.7074 - val_loss: 0.5812 - val_acc: 0.6940\n",
      "Epoch 15/100\n",
      " - 1s - loss: 0.5485 - acc: 0.7080 - val_loss: 0.5871 - val_acc: 0.6865\n",
      "Epoch 16/100\n",
      " - 1s - loss: 0.5463 - acc: 0.7130 - val_loss: 0.5710 - val_acc: 0.7105\n",
      "Epoch 17/100\n",
      " - 1s - loss: 0.5375 - acc: 0.7210 - val_loss: 0.5756 - val_acc: 0.6910\n",
      "Epoch 18/100\n",
      " - 1s - loss: 0.5354 - acc: 0.7180 - val_loss: 0.5711 - val_acc: 0.7010\n",
      "Epoch 19/100\n",
      " - 1s - loss: 0.5343 - acc: 0.7238 - val_loss: 0.5704 - val_acc: 0.7015\n",
      "Epoch 20/100\n",
      " - 1s - loss: 0.5227 - acc: 0.7265 - val_loss: 0.5681 - val_acc: 0.7100\n",
      "Epoch 21/100\n",
      " - 1s - loss: 0.5199 - acc: 0.7337 - val_loss: 0.5699 - val_acc: 0.7020\n",
      "Epoch 22/100\n",
      " - 1s - loss: 0.5178 - acc: 0.7327 - val_loss: 0.5670 - val_acc: 0.7045\n",
      "Epoch 23/100\n",
      " - 1s - loss: 0.5210 - acc: 0.7321 - val_loss: 0.5764 - val_acc: 0.7015\n",
      "Epoch 24/100\n",
      " - 1s - loss: 0.5125 - acc: 0.7400 - val_loss: 0.5693 - val_acc: 0.7060\n",
      "Epoch 25/100\n",
      " - 1s - loss: 0.5123 - acc: 0.7354 - val_loss: 0.5702 - val_acc: 0.7080\n",
      "Epoch 26/100\n",
      " - 1s - loss: 0.5081 - acc: 0.7339 - val_loss: 0.5661 - val_acc: 0.7085\n",
      "Epoch 27/100\n",
      " - 1s - loss: 0.4981 - acc: 0.7425 - val_loss: 0.5705 - val_acc: 0.6955\n",
      "Epoch 28/100\n",
      " - 1s - loss: 0.4964 - acc: 0.7445 - val_loss: 0.5700 - val_acc: 0.7045\n",
      "Epoch 29/100\n",
      " - 1s - loss: 0.4873 - acc: 0.7497 - val_loss: 0.5707 - val_acc: 0.7055\n",
      "Epoch 30/100\n",
      " - 1s - loss: 0.4994 - acc: 0.7438 - val_loss: 0.5756 - val_acc: 0.7020\n",
      "Epoch 31/100\n",
      " - 1s - loss: 0.4998 - acc: 0.7399 - val_loss: 0.5700 - val_acc: 0.7090\n",
      "Epoch 32/100\n",
      " - 1s - loss: 0.4865 - acc: 0.7474 - val_loss: 0.5786 - val_acc: 0.6940\n",
      "Epoch 33/100\n",
      " - 1s - loss: 0.4747 - acc: 0.7560 - val_loss: 0.5770 - val_acc: 0.6970\n",
      "Epoch 34/100\n",
      " - 1s - loss: 0.4841 - acc: 0.7566 - val_loss: 0.5821 - val_acc: 0.6940\n",
      "Epoch 35/100\n",
      " - 1s - loss: 0.4773 - acc: 0.7517 - val_loss: 0.5853 - val_acc: 0.6990\n",
      "Epoch 36/100\n",
      " - 1s - loss: 0.4772 - acc: 0.7565 - val_loss: 0.5773 - val_acc: 0.6945\n",
      "Epoch 37/100\n",
      " - 1s - loss: 0.4781 - acc: 0.7556 - val_loss: 0.5823 - val_acc: 0.7045\n",
      "Epoch 38/100\n",
      " - 1s - loss: 0.4714 - acc: 0.7556 - val_loss: 0.5782 - val_acc: 0.7050\n",
      "Epoch 39/100\n",
      " - 1s - loss: 0.4652 - acc: 0.7594 - val_loss: 0.5750 - val_acc: 0.7100\n",
      "Epoch 40/100\n",
      " - 1s - loss: 0.4685 - acc: 0.7569 - val_loss: 0.5765 - val_acc: 0.7045\n",
      "Epoch 41/100\n",
      " - 1s - loss: 0.4624 - acc: 0.7583 - val_loss: 0.5784 - val_acc: 0.7010\n",
      "Epoch 42/100\n",
      " - 1s - loss: 0.4659 - acc: 0.7613 - val_loss: 0.5870 - val_acc: 0.6995\n",
      "Epoch 43/100\n",
      " - 1s - loss: 0.4604 - acc: 0.7609 - val_loss: 0.5936 - val_acc: 0.7050\n",
      "Epoch 44/100\n",
      " - 1s - loss: 0.4558 - acc: 0.7665 - val_loss: 0.5816 - val_acc: 0.7025\n",
      "Epoch 45/100\n",
      " - 1s - loss: 0.4547 - acc: 0.7614 - val_loss: 0.5872 - val_acc: 0.7020\n",
      "Epoch 46/100\n",
      " - 1s - loss: 0.4496 - acc: 0.7679 - val_loss: 0.5966 - val_acc: 0.7040\n",
      "Epoch 47/100\n",
      " - 1s - loss: 0.4536 - acc: 0.7615 - val_loss: 0.5913 - val_acc: 0.7010\n",
      "Epoch 48/100\n",
      " - 1s - loss: 0.4567 - acc: 0.7553 - val_loss: 0.5949 - val_acc: 0.6870\n",
      "Epoch 49/100\n",
      " - 1s - loss: 0.4608 - acc: 0.7595 - val_loss: 0.5853 - val_acc: 0.7010\n",
      "Epoch 50/100\n",
      " - 1s - loss: 0.4470 - acc: 0.7686 - val_loss: 0.5941 - val_acc: 0.6980\n",
      "Epoch 51/100\n",
      " - 1s - loss: 0.4476 - acc: 0.7640 - val_loss: 0.5957 - val_acc: 0.7065\n",
      "Epoch 52/100\n",
      " - 1s - loss: 0.4400 - acc: 0.7733 - val_loss: 0.5903 - val_acc: 0.7015\n",
      "Epoch 53/100\n",
      " - 1s - loss: 0.4441 - acc: 0.7680 - val_loss: 0.5898 - val_acc: 0.7020\n",
      "Epoch 54/100\n",
      " - 1s - loss: 0.4443 - acc: 0.7653 - val_loss: 0.5968 - val_acc: 0.7045\n",
      "Epoch 55/100\n",
      " - 1s - loss: 0.4419 - acc: 0.7705 - val_loss: 0.5985 - val_acc: 0.7080\n",
      "Epoch 56/100\n",
      " - 1s - loss: 0.4442 - acc: 0.7645 - val_loss: 0.5883 - val_acc: 0.6985\n",
      "Epoch 57/100\n",
      " - 1s - loss: 0.4368 - acc: 0.7698 - val_loss: 0.5882 - val_acc: 0.7050\n",
      "Epoch 58/100\n",
      " - 1s - loss: 0.4368 - acc: 0.7731 - val_loss: 0.5924 - val_acc: 0.7000\n",
      "Epoch 59/100\n",
      " - 1s - loss: 0.4356 - acc: 0.7718 - val_loss: 0.5966 - val_acc: 0.7115\n",
      "Epoch 60/100\n",
      " - 1s - loss: 0.4356 - acc: 0.7733 - val_loss: 0.5948 - val_acc: 0.6990\n",
      "Epoch 61/100\n",
      " - 1s - loss: 0.4328 - acc: 0.7757 - val_loss: 0.5889 - val_acc: 0.7165\n",
      "Epoch 62/100\n",
      " - 1s - loss: 0.4287 - acc: 0.7767 - val_loss: 0.6138 - val_acc: 0.7015\n",
      "Epoch 63/100\n",
      " - 1s - loss: 0.4313 - acc: 0.7714 - val_loss: 0.5952 - val_acc: 0.7035\n",
      "Epoch 64/100\n",
      " - 1s - loss: 0.4240 - acc: 0.7868 - val_loss: 0.5968 - val_acc: 0.6930\n",
      "Epoch 65/100\n",
      " - 1s - loss: 0.4371 - acc: 0.7707 - val_loss: 0.5914 - val_acc: 0.7000\n",
      "Epoch 66/100\n",
      " - 1s - loss: 0.4297 - acc: 0.7783 - val_loss: 0.5967 - val_acc: 0.7085\n",
      "Epoch 67/100\n",
      " - 1s - loss: 0.4343 - acc: 0.7740 - val_loss: 0.5884 - val_acc: 0.7120\n",
      "Epoch 68/100\n",
      " - 1s - loss: 0.4306 - acc: 0.7797 - val_loss: 0.6014 - val_acc: 0.7030\n",
      "Epoch 69/100\n",
      " - 1s - loss: 0.4219 - acc: 0.7778 - val_loss: 0.5954 - val_acc: 0.7055\n",
      "Epoch 70/100\n",
      " - 1s - loss: 0.4236 - acc: 0.7811 - val_loss: 0.6021 - val_acc: 0.6920\n",
      "Epoch 71/100\n",
      " - 1s - loss: 0.4210 - acc: 0.7839 - val_loss: 0.6026 - val_acc: 0.7080\n",
      "Epoch 72/100\n",
      " - 1s - loss: 0.4151 - acc: 0.7810 - val_loss: 0.5950 - val_acc: 0.7045\n",
      "Epoch 73/100\n",
      " - 1s - loss: 0.4260 - acc: 0.7818 - val_loss: 0.6011 - val_acc: 0.6960\n",
      "Epoch 74/100\n",
      " - 1s - loss: 0.4165 - acc: 0.7806 - val_loss: 0.6039 - val_acc: 0.7000\n",
      "Epoch 75/100\n",
      " - 1s - loss: 0.4211 - acc: 0.7803 - val_loss: 0.6016 - val_acc: 0.6930\n",
      "Epoch 76/100\n",
      " - 1s - loss: 0.4276 - acc: 0.7766 - val_loss: 0.6061 - val_acc: 0.6955\n",
      "Epoch 77/100\n",
      " - 1s - loss: 0.4140 - acc: 0.7852 - val_loss: 0.5980 - val_acc: 0.7015\n",
      "Epoch 78/100\n",
      " - 1s - loss: 0.4221 - acc: 0.7778 - val_loss: 0.5956 - val_acc: 0.7075\n",
      "Epoch 79/100\n",
      " - 1s - loss: 0.4197 - acc: 0.7806 - val_loss: 0.6045 - val_acc: 0.7060\n",
      "Epoch 80/100\n",
      " - 1s - loss: 0.4185 - acc: 0.7815 - val_loss: 0.6041 - val_acc: 0.6995\n",
      "Epoch 81/100\n",
      " - 1s - loss: 0.4169 - acc: 0.7847 - val_loss: 0.6087 - val_acc: 0.7075\n",
      "Epoch 82/100\n",
      " - 1s - loss: 0.4203 - acc: 0.7793 - val_loss: 0.5971 - val_acc: 0.6875\n",
      "Epoch 83/100\n",
      " - 1s - loss: 0.4199 - acc: 0.7782 - val_loss: 0.5904 - val_acc: 0.7035\n",
      "Epoch 84/100\n",
      " - 1s - loss: 0.4261 - acc: 0.7747 - val_loss: 0.5922 - val_acc: 0.7105\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 85/100\n",
      " - 1s - loss: 0.4080 - acc: 0.7869 - val_loss: 0.6103 - val_acc: 0.7085\n",
      "Epoch 86/100\n",
      " - 1s - loss: 0.4174 - acc: 0.7777 - val_loss: 0.6072 - val_acc: 0.6985\n",
      "Epoch 87/100\n",
      " - 1s - loss: 0.4142 - acc: 0.7850 - val_loss: 0.6096 - val_acc: 0.7070\n",
      "Epoch 88/100\n",
      " - 1s - loss: 0.4093 - acc: 0.7841 - val_loss: 0.6123 - val_acc: 0.6980\n",
      "Epoch 89/100\n",
      " - 1s - loss: 0.4128 - acc: 0.7788 - val_loss: 0.6170 - val_acc: 0.6990\n",
      "Epoch 90/100\n",
      " - 1s - loss: 0.4089 - acc: 0.7866 - val_loss: 0.6074 - val_acc: 0.7045\n",
      "Epoch 91/100\n",
      " - 1s - loss: 0.4078 - acc: 0.7886 - val_loss: 0.6194 - val_acc: 0.7090\n",
      "Epoch 92/100\n",
      " - 1s - loss: 0.4064 - acc: 0.7912 - val_loss: 0.6020 - val_acc: 0.7090\n",
      "Epoch 93/100\n",
      " - 1s - loss: 0.4081 - acc: 0.7843 - val_loss: 0.6279 - val_acc: 0.7055\n",
      "Epoch 94/100\n",
      " - 1s - loss: 0.4045 - acc: 0.7882 - val_loss: 0.6201 - val_acc: 0.7060\n",
      "Epoch 95/100\n",
      " - 1s - loss: 0.3984 - acc: 0.7853 - val_loss: 0.6234 - val_acc: 0.7050\n",
      "Epoch 96/100\n",
      " - 1s - loss: 0.4037 - acc: 0.7854 - val_loss: 0.6345 - val_acc: 0.6905\n",
      "Epoch 97/100\n",
      " - 1s - loss: 0.4071 - acc: 0.7875 - val_loss: 0.6091 - val_acc: 0.7080\n",
      "Epoch 98/100\n",
      " - 1s - loss: 0.3962 - acc: 0.7927 - val_loss: 0.6189 - val_acc: 0.6995\n",
      "Epoch 99/100\n",
      " - 1s - loss: 0.4071 - acc: 0.7878 - val_loss: 0.6143 - val_acc: 0.6975\n",
      "Epoch 100/100\n",
      " - 1s - loss: 0.4015 - acc: 0.7903 - val_loss: 0.6161 - val_acc: 0.7035\n"
     ]
    }
   ],
   "source": [
    "# fit CNN model using Adagrad optimizer\n",
    "model1 = cnn_model()\n",
    "model1.compile(loss=keras.losses.categorical_crossentropy,\n",
    "              optimizer=Adagrad(lr=0.01, epsilon=1e-08, decay=0.0),\n",
    "              metrics=['accuracy'])\n",
    "history1 = model1.fit(X_train, y_train, \n",
    "                     validation_data=(X_test, y_test), \n",
    "                     epochs=epochs, \n",
    "                     batch_size=batch_size,\n",
    "                     verbose=2)\n",
    "\n",
    "# fit CNN model using Adadelta optimizer\n",
    "model2 = cnn_model()\n",
    "model2.compile(loss=keras.losses.categorical_crossentropy,\n",
    "              optimizer=Adadelta(lr=1.0, rho=0.95, epsilon=1e-08, decay=0.0),\n",
    "              metrics=['accuracy'])\n",
    "history2 = model2.fit(X_train, y_train, \n",
    "                     validation_data=(X_test, y_test), \n",
    "                     epochs=epochs, \n",
    "                     batch_size=batch_size,\n",
    "                     verbose=2)\n",
    "\n",
    "# fit CNN model using RMSprop optimizer\n",
    "model3 = cnn_model()\n",
    "model3.compile(loss=keras.losses.categorical_crossentropy,\n",
    "              optimizer=RMSprop(lr=0.001, rho=0.9, epsilon=1e-08, decay=0.0),\n",
    "              metrics=['accuracy'])\n",
    "history3 = model3.fit(X_train, y_train, \n",
    "                     validation_data=(X_test, y_test), \n",
    "                     epochs=epochs, \n",
    "                     batch_size=batch_size,\n",
    "                     verbose=2)\n",
    "\n",
    "# fit CNN model using Adam optimizer\n",
    "model4 = cnn_model()\n",
    "model4.compile(loss=keras.losses.categorical_crossentropy,\n",
    "              optimizer=Adam(lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-08, decay=0.0),\n",
    "              metrics=['accuracy'])\n",
    "history4 = model4.fit(X_train, y_train, \n",
    "                     validation_data=(X_test, y_test), \n",
    "                     epochs=epochs, \n",
    "                     batch_size=batch_size,\n",
    "                     verbose=2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "bESEUvEEEzjm"
   },
   "source": [
    "# Compare model accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "id": "H_0H6uDmEzjm",
    "outputId": "16859701-e5b9-4feb-c1d7-f58d171c71a8"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize=(12,4))\n",
    "plt.plot(range(epochs),history1.history['val_acc'],label='Adagrad')\n",
    "plt.plot(range(epochs),history2.history['val_acc'],label='Adadelta')\n",
    "plt.plot(range(epochs),history3.history['val_acc'],label='RMSprop')\n",
    "plt.plot(range(epochs),history4.history['val_acc'],label='Adam')\n",
    "plt.legend(loc=0)\n",
    "plt.xlabel('epochs')\n",
    "plt.xlim([0,epochs])\n",
    "plt.ylabel('accuracy om validation set')\n",
    "plt.grid(True)\n",
    "plt.title(\"Comparing Model Accuracy\")\n",
    "plt.show()\n",
    "fig.savefig('img/compare-accuracy.jpg')\n",
    "plt.close(fig)"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "name": "CIFAR10-lrate.ipynb",
   "provenance": []
  },
  "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": "3.7.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}