hw2_model3.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "hw2_model3.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/PriyankaGona/MachineLearning-Assignments/blob/master/hw2_model3.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "metadata": {
        "id": "40t4Dfr3SxHE",
        "colab_type": "code",
        "outputId": "a612989b-3475-4028-e2e0-38d5341a8638",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "cell_type": "code",
      "source": [
        "# Required headers and libraries are imported\n",
        "import keras\n",
        "from keras.datasets import cifar10\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Dense, Dropout, Activation, Flatten, BatchNormalization\n",
        "from keras.layers import Conv2D, MaxPooling2D\n",
        "from keras.optimizers import Adam\n",
        "from keras.losses import categorical_crossentropy\n",
        "from keras.preprocessing.image import ImageDataGenerator\n",
        "from keras.utils import np_utils\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "from sklearn.utils import shuffle"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Using TensorFlow backend.\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "metadata": {
        "id": "lESZvF5RS_zR",
        "colab_type": "code",
        "outputId": "74959a49-97ed-4442-e3dd-4d3d362b6fcf",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 119
        }
      },
      "cell_type": "code",
      "source": [
        "# Required variables are declared\n",
        "batch_size = 32\n",
        "num_classes = 10\n",
        "epochs = 100\n",
        "\n",
        "#Loading and shuffling Cifar10 dataset\n",
        "(x_train, y_train), (x_test, y_test) = cifar10.load_data()\n",
        "x_train, y_train = shuffle(x_train, y_train, random_state=0)\n",
        "\n",
        "\n",
        "split_index = int(0.8 * x_train.shape[0])\n",
        "\n",
        "\n",
        "y_train = keras.utils.to_categorical(y_train, num_classes)\n",
        "y_test = keras.utils.to_categorical(y_test, num_classes)\n",
        "\n",
        "x_train = x_train.astype('float32')\n",
        "x_test = x_test.astype('float32')\n",
        "\n",
        "x_train /= 255\n",
        "x_test /= 255\n",
        "\n",
        "# Dividing data into training and validation sets\n",
        "train_x = x_train[:split_index]\n",
        "train_y = y_train[:split_index]\n",
        "\n",
        "validate_x = x_train[split_index:]\n",
        "validate_y = y_train[split_index:]\n",
        "\n",
        "print(\"\\nnumber of samples in training set: \",len(train_x))\n",
        "print(\"number of samples in validation set: \",len(validate_x))\n",
        "\n",
        "input_shape = train_x.shape[1:]\n",
        "print(\"Shape of images: \",input_shape)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading data from https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz\n",
            "170500096/170498071 [==============================] - 38s 0us/step\n",
            "\n",
            "number of samples in training set:  40000\n",
            "number of samples in validation set:  10000\n",
            "Shape of images:  (32, 32, 3)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "4hhXfF_yXVpr",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def getModel(input_shape):\n",
        "  \n",
        "  model = Sequential()\n",
        "\n",
        "  model.add(Conv2D(32, kernel_size=(3, 3), activation='relu', input_shape=input_shape))\n",
        "  model.add(BatchNormalization())\n",
        "  \n",
        "  model.add(Conv2D(32, kernel_size=(3, 3), activation='relu'))\n",
        "  model.add(BatchNormalization())\n",
        "  \n",
        "  model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "  model.add(Dropout(0.25))\n",
        "  \n",
        "  model.add(Conv2D(64, kernel_size=(3, 3), activation='relu', padding='same'))\n",
        "  model.add(BatchNormalization())\n",
        "  \n",
        "  model.add(Conv2D(64, kernel_size=(3, 3), activation='relu'))\n",
        "  model.add(BatchNormalization())\n",
        "  \n",
        "  model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "  model.add(Dropout(0.25))\n",
        "  \n",
        "  model.add(Flatten())\n",
        "  \n",
        "  model.add(Dense(512, activation='relu'))\n",
        "  model.add(BatchNormalization())\n",
        "  \n",
        "  model.add(Dropout(0.5))\n",
        "  \n",
        "  model.add(Dense(num_classes, activation='softmax'))\n",
        "  \n",
        "  model.compile(loss='categorical_crossentropy', optimizer=Adam(lr=0.0001, decay=1e-6), metrics=['accuracy'])\n",
        "  \n",
        "  return model"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "W0XlBq9OUW_L",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def train_model(model, train_x, train_y, validate_x, validate_y):\n",
        "  # Data Augmentation\n",
        "\tdatagen = ImageDataGenerator(rotation_range=0, width_shift_range=0.1,height_shift_range=0.1,fill_mode='nearest', horizontal_flip=True, vertical_flip=False, rescale=None)\n",
        "\n",
        "\tdatagen.fit(train_x)\n",
        "\n",
        "\tfor e in range(10):\n",
        "\t\tbatches = 0\n",
        "\t\tfor x_batch, y_batch in datagen.flow(train_x, train_y, batch_size=40000):\n",
        "\t\t\tmodel.fit(x_batch, y_batch)\n",
        "\t\t\tbatches += 1\n",
        "\t\t\tif batches >= 1:\n",
        "\t\t\t\t# we need to break the loop by hand because\n",
        "\t\t\t\t# the generator loops indefinitely\n",
        "\t\t\t\tbreak\n",
        "\t\t\t\t\n",
        "\t\t\t\t\n",
        "\treturn model.fit(train_x, train_y, batch_size=batch_size, epochs=epochs, verbose=1, validation_data=(validate_x, validate_y))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "I88ZGsKIUbMp",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# Model is tested in this model\n",
        "def test_model(validate_x, validate_y):\n",
        "\t\n",
        "\tscore = model.evaluate(validate_x, validate_y, verbose=0)\n",
        "\t\n",
        "\tprint('')\n",
        "\tprint('Test loss:', score[0])\n",
        "\tprint('Test accuracy:', score[1])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "10yhPdQvUfbT",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "#Plotted the accuracy against number of epochs\n",
        "def plot_training_graph(training_history):\n",
        "\t\n",
        "\tplotaccuracy = plt.plot(range(1,epochs+1), training_history.history['acc'],range(1,epochs+1), training_history.history['val_acc'])\n",
        "\n",
        "\tplt.xlabel('Number of epochs')\n",
        "\tplt.ylabel('Accuracy')\n",
        "\n",
        "\tplt.legend(('Train Accuracy','Test Accuracy'))\n",
        "\n",
        "\tplt.show(plotaccuracy)\n",
        "\n",
        "\tprint('\\n\\nvalidation loss:', training_history.history['val_loss'][-1])\n",
        "\tprint('validation accuracy:', training_history.history['val_acc'][-1])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "L5Q_hGHIUoQJ",
        "colab_type": "code",
        "outputId": "ce7f868d-8ccd-4d9c-b970-7874f9a9d8b3",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 4390
        }
      },
      "cell_type": "code",
      "source": [
        "# Calling all the methods\n",
        "model = getModel(input_shape)\n",
        "History = train_model(model, train_x, train_y, validate_x, validate_y)\n",
        "test_model(validate_x, validate_y)\n",
        "plot_training_graph(History)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/framework/op_def_library.py:263: colocate_with (from tensorflow.python.framework.ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Colocations handled automatically by placer.\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use tf.cast instead.\n",
            "Epoch 1/1\n",
            "40000/40000 [==============================] - 28s 691us/step - loss: 2.3377 - acc: 0.3145\n",
            "Epoch 1/1\n",
            "40000/40000 [==============================] - 24s 600us/step - loss: 1.8103 - acc: 0.4098\n",
            "Epoch 1/1\n",
            "40000/40000 [==============================] - 24s 604us/step - loss: 1.5987 - acc: 0.4541\n",
            "Epoch 1/1\n",
            "40000/40000 [==============================] - 24s 599us/step - loss: 1.4579 - acc: 0.4940\n",
            "Epoch 1/1\n",
            "40000/40000 [==============================] - 25s 621us/step - loss: 1.3310 - acc: 0.5367\n",
            "Epoch 1/1\n",
            "40000/40000 [==============================] - 24s 601us/step - loss: 1.2496 - acc: 0.5612\n",
            "Epoch 1/1\n",
            "40000/40000 [==============================] - 24s 601us/step - loss: 1.1745 - acc: 0.5824\n",
            "Epoch 1/1\n",
            "40000/40000 [==============================] - 24s 601us/step - loss: 1.1151 - acc: 0.6039\n",
            "Epoch 1/1\n",
            "40000/40000 [==============================] - 24s 598us/step - loss: 1.0636 - acc: 0.6257\n",
            "Epoch 1/1\n",
            "40000/40000 [==============================] - 24s 598us/step - loss: 1.0199 - acc: 0.6421\n",
            "Train on 40000 samples, validate on 10000 samples\n",
            "Epoch 1/100\n",
            "40000/40000 [==============================] - 25s 626us/step - loss: 0.8835 - acc: 0.6914 - val_loss: 0.8069 - val_acc: 0.7127\n",
            "Epoch 2/100\n",
            "40000/40000 [==============================] - 25s 613us/step - loss: 0.8293 - acc: 0.7093 - val_loss: 0.7442 - val_acc: 0.7350\n",
            "Epoch 3/100\n",
            "40000/40000 [==============================] - 25s 632us/step - loss: 0.7894 - acc: 0.7250 - val_loss: 0.7223 - val_acc: 0.7450\n",
            "Epoch 4/100\n",
            "40000/40000 [==============================] - 26s 649us/step - loss: 0.7651 - acc: 0.7351 - val_loss: 0.7131 - val_acc: 0.7515\n",
            "Epoch 5/100\n",
            "40000/40000 [==============================] - 26s 646us/step - loss: 0.7293 - acc: 0.7457 - val_loss: 0.6753 - val_acc: 0.7619\n",
            "Epoch 6/100\n",
            "40000/40000 [==============================] - 26s 643us/step - loss: 0.6944 - acc: 0.7583 - val_loss: 0.6901 - val_acc: 0.7556\n",
            "Epoch 7/100\n",
            "40000/40000 [==============================] - 25s 631us/step - loss: 0.6740 - acc: 0.7637 - val_loss: 0.6594 - val_acc: 0.7697\n",
            "Epoch 8/100\n",
            "40000/40000 [==============================] - 26s 641us/step - loss: 0.6501 - acc: 0.7718 - val_loss: 0.6640 - val_acc: 0.7706\n",
            "Epoch 9/100\n",
            "40000/40000 [==============================] - 28s 698us/step - loss: 0.6365 - acc: 0.7797 - val_loss: 0.6545 - val_acc: 0.7718\n",
            "Epoch 10/100\n",
            "40000/40000 [==============================] - 28s 699us/step - loss: 0.6096 - acc: 0.7878 - val_loss: 0.6170 - val_acc: 0.7814\n",
            "Epoch 11/100\n",
            "40000/40000 [==============================] - 28s 693us/step - loss: 0.5934 - acc: 0.7917 - val_loss: 0.6367 - val_acc: 0.7803\n",
            "Epoch 12/100\n",
            "40000/40000 [==============================] - 28s 694us/step - loss: 0.5769 - acc: 0.7985 - val_loss: 0.6367 - val_acc: 0.7795\n",
            "Epoch 13/100\n",
            "40000/40000 [==============================] - 28s 691us/step - loss: 0.5558 - acc: 0.8059 - val_loss: 0.6306 - val_acc: 0.7802\n",
            "Epoch 14/100\n",
            "40000/40000 [==============================] - 27s 686us/step - loss: 0.5455 - acc: 0.8085 - val_loss: 0.6173 - val_acc: 0.7888\n",
            "Epoch 15/100\n",
            "40000/40000 [==============================] - 27s 683us/step - loss: 0.5282 - acc: 0.8156 - val_loss: 0.6107 - val_acc: 0.7885\n",
            "Epoch 16/100\n",
            "40000/40000 [==============================] - 26s 653us/step - loss: 0.5099 - acc: 0.8209 - val_loss: 0.5806 - val_acc: 0.7982\n",
            "Epoch 17/100\n",
            "40000/40000 [==============================] - 27s 665us/step - loss: 0.4996 - acc: 0.8249 - val_loss: 0.5880 - val_acc: 0.7974\n",
            "Epoch 18/100\n",
            "40000/40000 [==============================] - 26s 662us/step - loss: 0.4854 - acc: 0.8275 - val_loss: 0.5977 - val_acc: 0.7924\n",
            "Epoch 19/100\n",
            "40000/40000 [==============================] - 26s 651us/step - loss: 0.4715 - acc: 0.8330 - val_loss: 0.5762 - val_acc: 0.8049\n",
            "Epoch 20/100\n",
            "40000/40000 [==============================] - 29s 726us/step - loss: 0.4610 - acc: 0.8387 - val_loss: 0.6062 - val_acc: 0.7933\n",
            "Epoch 21/100\n",
            "40000/40000 [==============================] - 29s 723us/step - loss: 0.4495 - acc: 0.8430 - val_loss: 0.5950 - val_acc: 0.7955\n",
            "Epoch 22/100\n",
            "40000/40000 [==============================] - 29s 723us/step - loss: 0.4331 - acc: 0.8489 - val_loss: 0.5994 - val_acc: 0.7929\n",
            "Epoch 23/100\n",
            "40000/40000 [==============================] - 29s 722us/step - loss: 0.4243 - acc: 0.8494 - val_loss: 0.5830 - val_acc: 0.8023\n",
            "Epoch 24/100\n",
            "40000/40000 [==============================] - 28s 696us/step - loss: 0.4220 - acc: 0.8512 - val_loss: 0.5797 - val_acc: 0.8001\n",
            "Epoch 25/100\n",
            "40000/40000 [==============================] - 28s 697us/step - loss: 0.4123 - acc: 0.8532 - val_loss: 0.5839 - val_acc: 0.8001\n",
            "Epoch 26/100\n",
            "40000/40000 [==============================] - 28s 694us/step - loss: 0.4024 - acc: 0.8575 - val_loss: 0.5784 - val_acc: 0.8037\n",
            "Epoch 27/100\n",
            "40000/40000 [==============================] - 28s 690us/step - loss: 0.3893 - acc: 0.8605 - val_loss: 0.6078 - val_acc: 0.7979\n",
            "Epoch 28/100\n",
            "40000/40000 [==============================] - 28s 690us/step - loss: 0.3804 - acc: 0.8687 - val_loss: 0.5617 - val_acc: 0.8088\n",
            "Epoch 29/100\n",
            "40000/40000 [==============================] - 27s 687us/step - loss: 0.3723 - acc: 0.8690 - val_loss: 0.5689 - val_acc: 0.8075\n",
            "Epoch 30/100\n",
            "40000/40000 [==============================] - 27s 683us/step - loss: 0.3644 - acc: 0.8715 - val_loss: 0.5711 - val_acc: 0.8069\n",
            "Epoch 31/100\n",
            "40000/40000 [==============================] - 27s 680us/step - loss: 0.3534 - acc: 0.8748 - val_loss: 0.5882 - val_acc: 0.8024\n",
            "Epoch 32/100\n",
            "40000/40000 [==============================] - 28s 694us/step - loss: 0.3469 - acc: 0.8780 - val_loss: 0.5962 - val_acc: 0.8059\n",
            "Epoch 33/100\n",
            "40000/40000 [==============================] - 27s 682us/step - loss: 0.3396 - acc: 0.8806 - val_loss: 0.5647 - val_acc: 0.8108\n",
            "Epoch 34/100\n",
            "40000/40000 [==============================] - 27s 680us/step - loss: 0.3341 - acc: 0.8810 - val_loss: 0.5596 - val_acc: 0.8156\n",
            "Epoch 35/100\n",
            "40000/40000 [==============================] - 27s 681us/step - loss: 0.3318 - acc: 0.8816 - val_loss: 0.5641 - val_acc: 0.8146\n",
            "Epoch 36/100\n",
            "40000/40000 [==============================] - 27s 680us/step - loss: 0.3225 - acc: 0.8875 - val_loss: 0.5628 - val_acc: 0.8109\n",
            "Epoch 37/100\n",
            "40000/40000 [==============================] - 27s 684us/step - loss: 0.3173 - acc: 0.8876 - val_loss: 0.5663 - val_acc: 0.8165\n",
            "Epoch 38/100\n",
            "40000/40000 [==============================] - 27s 684us/step - loss: 0.3133 - acc: 0.8881 - val_loss: 0.5970 - val_acc: 0.8063\n",
            "Epoch 39/100\n",
            "40000/40000 [==============================] - 27s 683us/step - loss: 0.3056 - acc: 0.8925 - val_loss: 0.5630 - val_acc: 0.8174\n",
            "Epoch 40/100\n",
            "40000/40000 [==============================] - 27s 680us/step - loss: 0.3014 - acc: 0.8945 - val_loss: 0.5719 - val_acc: 0.8110\n",
            "Epoch 41/100\n",
            "40000/40000 [==============================] - 27s 680us/step - loss: 0.2942 - acc: 0.8962 - val_loss: 0.5814 - val_acc: 0.8096\n",
            "Epoch 42/100\n",
            "40000/40000 [==============================] - 27s 681us/step - loss: 0.2888 - acc: 0.8963 - val_loss: 0.5840 - val_acc: 0.8118\n",
            "Epoch 43/100\n",
            "40000/40000 [==============================] - 27s 683us/step - loss: 0.2852 - acc: 0.8993 - val_loss: 0.5680 - val_acc: 0.8174\n",
            "Epoch 44/100\n",
            "40000/40000 [==============================] - 27s 682us/step - loss: 0.2744 - acc: 0.9023 - val_loss: 0.5733 - val_acc: 0.8134\n",
            "Epoch 45/100\n",
            "40000/40000 [==============================] - 27s 679us/step - loss: 0.2753 - acc: 0.9030 - val_loss: 0.5756 - val_acc: 0.8152\n",
            "Epoch 46/100\n",
            "40000/40000 [==============================] - 28s 695us/step - loss: 0.2724 - acc: 0.9034 - val_loss: 0.5666 - val_acc: 0.8163\n",
            "Epoch 47/100\n",
            "40000/40000 [==============================] - 27s 683us/step - loss: 0.2745 - acc: 0.9029 - val_loss: 0.6241 - val_acc: 0.8078\n",
            "Epoch 48/100\n",
            "40000/40000 [==============================] - 27s 684us/step - loss: 0.2672 - acc: 0.9057 - val_loss: 0.5886 - val_acc: 0.8131\n",
            "Epoch 49/100\n",
            "40000/40000 [==============================] - 27s 685us/step - loss: 0.2588 - acc: 0.9065 - val_loss: 0.5635 - val_acc: 0.8219\n",
            "Epoch 50/100\n",
            "40000/40000 [==============================] - 27s 681us/step - loss: 0.2569 - acc: 0.9100 - val_loss: 0.5769 - val_acc: 0.8161\n",
            "Epoch 51/100\n",
            "40000/40000 [==============================] - 27s 681us/step - loss: 0.2578 - acc: 0.9090 - val_loss: 0.5933 - val_acc: 0.8129\n",
            "Epoch 52/100\n",
            "40000/40000 [==============================] - 27s 679us/step - loss: 0.2479 - acc: 0.9118 - val_loss: 0.5771 - val_acc: 0.8167\n",
            "Epoch 53/100\n",
            "40000/40000 [==============================] - 27s 670us/step - loss: 0.2412 - acc: 0.9138 - val_loss: 0.5994 - val_acc: 0.8098\n",
            "Epoch 54/100\n",
            "40000/40000 [==============================] - 26s 656us/step - loss: 0.2388 - acc: 0.9156 - val_loss: 0.5862 - val_acc: 0.8176\n",
            "Epoch 55/100\n",
            "40000/40000 [==============================] - 26s 651us/step - loss: 0.2349 - acc: 0.9167 - val_loss: 0.6019 - val_acc: 0.8120\n",
            "Epoch 56/100\n",
            "40000/40000 [==============================] - 26s 654us/step - loss: 0.2312 - acc: 0.9176 - val_loss: 0.5857 - val_acc: 0.8176\n",
            "Epoch 57/100\n",
            "40000/40000 [==============================] - 26s 653us/step - loss: 0.2301 - acc: 0.9185 - val_loss: 0.5806 - val_acc: 0.8188\n",
            "Epoch 58/100\n",
            "40000/40000 [==============================] - 26s 652us/step - loss: 0.2262 - acc: 0.9219 - val_loss: 0.5864 - val_acc: 0.8204\n",
            "Epoch 59/100\n",
            "40000/40000 [==============================] - 26s 649us/step - loss: 0.2271 - acc: 0.9198 - val_loss: 0.5928 - val_acc: 0.8164\n",
            "Epoch 60/100\n",
            "40000/40000 [==============================] - 26s 659us/step - loss: 0.2236 - acc: 0.9207 - val_loss: 0.6175 - val_acc: 0.8107\n",
            "Epoch 61/100\n",
            "40000/40000 [==============================] - 26s 651us/step - loss: 0.2147 - acc: 0.9250 - val_loss: 0.5975 - val_acc: 0.8200\n",
            "Epoch 62/100\n",
            "40000/40000 [==============================] - 26s 649us/step - loss: 0.2190 - acc: 0.9214 - val_loss: 0.6021 - val_acc: 0.8165\n",
            "Epoch 63/100\n",
            "40000/40000 [==============================] - 26s 650us/step - loss: 0.2128 - acc: 0.9254 - val_loss: 0.5947 - val_acc: 0.8156\n",
            "Epoch 64/100\n",
            "40000/40000 [==============================] - 26s 649us/step - loss: 0.2091 - acc: 0.9260 - val_loss: 0.5887 - val_acc: 0.8169\n",
            "Epoch 65/100\n",
            "40000/40000 [==============================] - 26s 651us/step - loss: 0.2089 - acc: 0.9256 - val_loss: 0.5849 - val_acc: 0.8228\n",
            "Epoch 66/100\n",
            "40000/40000 [==============================] - 26s 651us/step - loss: 0.2102 - acc: 0.9255 - val_loss: 0.5874 - val_acc: 0.8213\n",
            "Epoch 67/100\n",
            "40000/40000 [==============================] - 26s 649us/step - loss: 0.2047 - acc: 0.9277 - val_loss: 0.5982 - val_acc: 0.8159\n",
            "Epoch 68/100\n",
            "40000/40000 [==============================] - 28s 704us/step - loss: 0.2011 - acc: 0.9289 - val_loss: 0.5903 - val_acc: 0.8223\n",
            "Epoch 69/100\n",
            "40000/40000 [==============================] - 26s 656us/step - loss: 0.2052 - acc: 0.9278 - val_loss: 0.6156 - val_acc: 0.8193\n",
            "Epoch 70/100\n",
            "40000/40000 [==============================] - 26s 653us/step - loss: 0.2037 - acc: 0.9297 - val_loss: 0.5930 - val_acc: 0.8223\n",
            "Epoch 71/100\n",
            "40000/40000 [==============================] - 26s 651us/step - loss: 0.2042 - acc: 0.9276 - val_loss: 0.5946 - val_acc: 0.8212\n",
            "Epoch 72/100\n",
            "40000/40000 [==============================] - 32s 805us/step - loss: 0.1916 - acc: 0.9339 - val_loss: 0.5935 - val_acc: 0.8230\n",
            "Epoch 73/100\n",
            "40000/40000 [==============================] - 28s 688us/step - loss: 0.1933 - acc: 0.9321 - val_loss: 0.5947 - val_acc: 0.8201\n",
            "Epoch 74/100\n",
            "40000/40000 [==============================] - 28s 697us/step - loss: 0.1905 - acc: 0.9331 - val_loss: 0.6012 - val_acc: 0.8209\n",
            "Epoch 75/100\n",
            "40000/40000 [==============================] - 28s 693us/step - loss: 0.1888 - acc: 0.9340 - val_loss: 0.6043 - val_acc: 0.8189\n",
            "Epoch 76/100\n",
            "40000/40000 [==============================] - 28s 690us/step - loss: 0.1901 - acc: 0.9333 - val_loss: 0.6128 - val_acc: 0.8214\n",
            "Epoch 77/100\n",
            "40000/40000 [==============================] - 28s 689us/step - loss: 0.1840 - acc: 0.9348 - val_loss: 0.6107 - val_acc: 0.8192\n",
            "Epoch 78/100\n",
            "40000/40000 [==============================] - 27s 682us/step - loss: 0.1829 - acc: 0.9344 - val_loss: 0.6043 - val_acc: 0.8210\n",
            "Epoch 79/100\n",
            "40000/40000 [==============================] - 27s 684us/step - loss: 0.1833 - acc: 0.9351 - val_loss: 0.6156 - val_acc: 0.8192\n",
            "Epoch 80/100\n",
            "40000/40000 [==============================] - 27s 684us/step - loss: 0.1783 - acc: 0.9370 - val_loss: 0.6101 - val_acc: 0.8212\n",
            "Epoch 81/100\n",
            "40000/40000 [==============================] - 28s 688us/step - loss: 0.1770 - acc: 0.9368 - val_loss: 0.6112 - val_acc: 0.8213\n",
            "Epoch 82/100\n",
            "40000/40000 [==============================] - 27s 682us/step - loss: 0.1765 - acc: 0.9369 - val_loss: 0.6030 - val_acc: 0.8226\n",
            "Epoch 83/100\n",
            "40000/40000 [==============================] - 27s 682us/step - loss: 0.1776 - acc: 0.9369 - val_loss: 0.6032 - val_acc: 0.8243\n",
            "Epoch 84/100\n",
            "40000/40000 [==============================] - 27s 684us/step - loss: 0.1723 - acc: 0.9395 - val_loss: 0.6126 - val_acc: 0.8212\n",
            "Epoch 85/100\n",
            "40000/40000 [==============================] - 27s 683us/step - loss: 0.1773 - acc: 0.9377 - val_loss: 0.6063 - val_acc: 0.8197\n",
            "Epoch 86/100\n",
            "40000/40000 [==============================] - 27s 684us/step - loss: 0.1730 - acc: 0.9394 - val_loss: 0.6115 - val_acc: 0.8257\n",
            "Epoch 87/100\n",
            "40000/40000 [==============================] - 27s 678us/step - loss: 0.1687 - acc: 0.9403 - val_loss: 0.6219 - val_acc: 0.8156\n",
            "Epoch 88/100\n",
            "40000/40000 [==============================] - 28s 689us/step - loss: 0.1759 - acc: 0.9361 - val_loss: 0.6207 - val_acc: 0.8223\n",
            "Epoch 89/100\n",
            "40000/40000 [==============================] - 27s 685us/step - loss: 0.1678 - acc: 0.9410 - val_loss: 0.6106 - val_acc: 0.8235\n",
            "Epoch 90/100\n",
            "40000/40000 [==============================] - 27s 686us/step - loss: 0.1636 - acc: 0.9435 - val_loss: 0.6295 - val_acc: 0.8198\n",
            "Epoch 91/100\n",
            "40000/40000 [==============================] - 27s 682us/step - loss: 0.1622 - acc: 0.9426 - val_loss: 0.6215 - val_acc: 0.8217\n",
            "Epoch 92/100\n",
            "40000/40000 [==============================] - 27s 681us/step - loss: 0.1627 - acc: 0.9428 - val_loss: 0.5945 - val_acc: 0.8230\n",
            "Epoch 93/100\n",
            "40000/40000 [==============================] - 27s 683us/step - loss: 0.1647 - acc: 0.9422 - val_loss: 0.6085 - val_acc: 0.8200\n",
            "Epoch 94/100\n",
            "40000/40000 [==============================] - 27s 684us/step - loss: 0.1653 - acc: 0.9414 - val_loss: 0.6391 - val_acc: 0.8175\n",
            "Epoch 95/100\n",
            "40000/40000 [==============================] - 27s 684us/step - loss: 0.1621 - acc: 0.9434 - val_loss: 0.6279 - val_acc: 0.8180\n",
            "Epoch 96/100\n",
            "40000/40000 [==============================] - 27s 684us/step - loss: 0.1515 - acc: 0.9451 - val_loss: 0.6254 - val_acc: 0.8217\n",
            "Epoch 97/100\n",
            "40000/40000 [==============================] - 26s 649us/step - loss: 0.1629 - acc: 0.9432 - val_loss: 0.6405 - val_acc: 0.8185\n",
            "Epoch 98/100\n",
            "40000/40000 [==============================] - 25s 614us/step - loss: 0.1584 - acc: 0.9450 - val_loss: 0.6251 - val_acc: 0.8210\n",
            "Epoch 99/100\n",
            "40000/40000 [==============================] - 24s 607us/step - loss: 0.1581 - acc: 0.9441 - val_loss: 0.6246 - val_acc: 0.8230\n",
            "Epoch 100/100\n",
            "40000/40000 [==============================] - 24s 610us/step - loss: 0.1546 - acc: 0.9461 - val_loss: 0.6192 - val_acc: 0.8221\n",
            "\n",
            "Test loss: 0.6192357271432877\n",
            "Test accuracy: 0.8221\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfUAAAFYCAYAAABKymUhAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nQmCC\n",
            "text/plain": [
              "<Figure size 576x396 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n",
            "\n",
            "validation loss: 0.6192357271432877\n",
            "validation accuracy: 0.8221\n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}