hw2_model4.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "hw2_model4.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_model4.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "metadata": {
        "id": "fV7V8GJUbPYT",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 35
        },
        "outputId": "8f6c67b4-395f-44ce-a0b7-adc7ea4f897b"
      },
      "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\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": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Using TensorFlow backend.\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "metadata": {
        "id": "5J1gtNDAbvFG",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 87
        },
        "outputId": "d885666a-301e-4fc4-a033-d896faff83ae"
      },
      "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": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\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": "3vS-jRNCbv_U",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "#Defining the model with more layers.\n",
        "def getModel(input_shape):\n",
        "\tmodel = Sequential()\n",
        "\n",
        "\tmodel.add(Conv2D(32, (3, 3), padding='same', input_shape=train_x.shape[1:]))\n",
        "\tmodel.add(Activation('relu'))\n",
        "\n",
        "\tmodel.add(Conv2D(32, (3, 3)))\n",
        "\tmodel.add(Activation('relu'))\n",
        "\n",
        "\tmodel.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\tmodel.add(Dropout(0.25))\n",
        "\n",
        "\tmodel.add(Conv2D(64, (3, 3), padding='same'))\n",
        "\tmodel.add(Activation('relu'))\n",
        "\n",
        "\tmodel.add(Conv2D(64, (3, 3)))\n",
        "\tmodel.add(Activation('relu'))\n",
        "\n",
        "\tmodel.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "\tmodel.add(Dropout(0.25))\n",
        "\n",
        "\tmodel.add(Conv2D(128, (2, 2), padding='same'))\n",
        "\tmodel.add(Activation('relu'))\n",
        "\n",
        "\tmodel.add(Conv2D(128, (2, 2)))\n",
        "\tmodel.add(Activation('relu'))\n",
        "\n",
        "\tmodel.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "\tmodel.add(Dropout(0.25))\n",
        "\n",
        "\tmodel.add(Flatten())\n",
        "\n",
        "\tmodel.add(Dense(1024))\n",
        "\tmodel.add(Activation('relu'))\n",
        "\n",
        "\tmodel.add(Dropout(0.5))\n",
        "\n",
        "\tmodel.add(Dense(num_classes))\n",
        "\tmodel.add(Activation('softmax'))\n",
        "\t   \n",
        "\t \n",
        "\tmodel.compile(loss='categorical_crossentropy', optimizer=Adam(lr=0.0001, decay=1e-6), metrics=['accuracy'])\n",
        "\t\n",
        "\treturn model"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "3eQi42ObcCfE",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def train_model(model):\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": "35Ikkw10cK8t",
        "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",
        "  score = model.evaluate(validate_x, validate_y, verbose=0)\n",
        "\t\n",
        "  print('Test loss:', score[0])\n",
        "  print('Test accuracy:', score[1])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "-WoQkmjwcSVM",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "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": "7yTfu5CsccJO",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 4472
        },
        "outputId": "9c1ae76c-cea1-4789-e37c-4a764ce33b49"
      },
      "cell_type": "code",
      "source": [
        "# calling all the methods\n",
        "model = getModel(input_shape)\n",
        "History = train_model(model)\n",
        "test_model(validate_x, validate_y)\n",
        "plot_training_graph(History)"
      ],
      "execution_count": 7,
      "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 [==============================] - 22s 548us/step - loss: 1.9747 - acc: 0.2521\n",
            "Epoch 1/1\n",
            "40000/40000 [==============================] - 18s 461us/step - loss: 1.6833 - acc: 0.3725\n",
            "Epoch 1/1\n",
            "40000/40000 [==============================] - 22s 561us/step - loss: 1.5637 - acc: 0.4203\n",
            "Epoch 1/1\n",
            "40000/40000 [==============================] - 17s 432us/step - loss: 1.4756 - acc: 0.4559\n",
            "Epoch 1/1\n",
            "40000/40000 [==============================] - 23s 568us/step - loss: 1.4104 - acc: 0.4829\n",
            "Epoch 1/1\n",
            "40000/40000 [==============================] - 18s 439us/step - loss: 1.3525 - acc: 0.5106\n",
            "Epoch 1/1\n",
            "40000/40000 [==============================] - 18s 445us/step - loss: 1.2991 - acc: 0.5284\n",
            "Epoch 1/1\n",
            "40000/40000 [==============================] - 17s 435us/step - loss: 1.2608 - acc: 0.5448\n",
            "Epoch 1/1\n",
            "40000/40000 [==============================] - 18s 445us/step - loss: 1.2249 - acc: 0.5587\n",
            "Epoch 1/1\n",
            "40000/40000 [==============================] - 20s 498us/step - loss: 1.1881 - acc: 0.5715\n",
            "Train on 40000 samples, validate on 10000 samples\n",
            "Epoch 1/100\n",
            "40000/40000 [==============================] - 23s 572us/step - loss: 1.0794 - acc: 0.6121 - val_loss: 1.0111 - val_acc: 0.6360\n",
            "Epoch 2/100\n",
            "40000/40000 [==============================] - 19s 482us/step - loss: 1.0463 - acc: 0.6268 - val_loss: 0.9677 - val_acc: 0.6577\n",
            "Epoch 3/100\n",
            "40000/40000 [==============================] - 22s 549us/step - loss: 1.0103 - acc: 0.6398 - val_loss: 0.9475 - val_acc: 0.6611\n",
            "Epoch 4/100\n",
            "40000/40000 [==============================] - 22s 551us/step - loss: 0.9869 - acc: 0.6494 - val_loss: 0.9126 - val_acc: 0.6756\n",
            "Epoch 5/100\n",
            "40000/40000 [==============================] - 22s 546us/step - loss: 0.9567 - acc: 0.6584 - val_loss: 0.9280 - val_acc: 0.6629\n",
            "Epoch 6/100\n",
            "40000/40000 [==============================] - 23s 568us/step - loss: 0.9344 - acc: 0.6666 - val_loss: 0.8744 - val_acc: 0.6873\n",
            "Epoch 7/100\n",
            "40000/40000 [==============================] - 23s 571us/step - loss: 0.9134 - acc: 0.6752 - val_loss: 0.8479 - val_acc: 0.6943\n",
            "Epoch 8/100\n",
            "40000/40000 [==============================] - 23s 570us/step - loss: 0.8841 - acc: 0.6857 - val_loss: 0.8891 - val_acc: 0.6811\n",
            "Epoch 9/100\n",
            "40000/40000 [==============================] - 23s 567us/step - loss: 0.8724 - acc: 0.6907 - val_loss: 0.8260 - val_acc: 0.7120\n",
            "Epoch 10/100\n",
            "40000/40000 [==============================] - 23s 568us/step - loss: 0.8474 - acc: 0.6994 - val_loss: 0.7879 - val_acc: 0.7224\n",
            "Epoch 11/100\n",
            "40000/40000 [==============================] - 22s 557us/step - loss: 0.8312 - acc: 0.7062 - val_loss: 0.7986 - val_acc: 0.7209\n",
            "Epoch 12/100\n",
            "40000/40000 [==============================] - 22s 560us/step - loss: 0.8110 - acc: 0.7128 - val_loss: 0.7613 - val_acc: 0.7345\n",
            "Epoch 13/100\n",
            "40000/40000 [==============================] - 22s 562us/step - loss: 0.7980 - acc: 0.7214 - val_loss: 0.7579 - val_acc: 0.7324\n",
            "Epoch 14/100\n",
            "40000/40000 [==============================] - 23s 569us/step - loss: 0.7752 - acc: 0.7251 - val_loss: 0.7646 - val_acc: 0.7326\n",
            "Epoch 15/100\n",
            "40000/40000 [==============================] - 23s 576us/step - loss: 0.7635 - acc: 0.7303 - val_loss: 0.7335 - val_acc: 0.7419\n",
            "Epoch 16/100\n",
            "40000/40000 [==============================] - 24s 597us/step - loss: 0.7446 - acc: 0.7380 - val_loss: 0.7142 - val_acc: 0.7518\n",
            "Epoch 17/100\n",
            "40000/40000 [==============================] - 24s 597us/step - loss: 0.7371 - acc: 0.7403 - val_loss: 0.6991 - val_acc: 0.7573\n",
            "Epoch 18/100\n",
            "40000/40000 [==============================] - 24s 591us/step - loss: 0.7214 - acc: 0.7481 - val_loss: 0.7198 - val_acc: 0.7456\n",
            "Epoch 19/100\n",
            "40000/40000 [==============================] - 24s 589us/step - loss: 0.7112 - acc: 0.7476 - val_loss: 0.7082 - val_acc: 0.7519\n",
            "Epoch 20/100\n",
            "40000/40000 [==============================] - 23s 585us/step - loss: 0.7015 - acc: 0.7524 - val_loss: 0.7150 - val_acc: 0.7500\n",
            "Epoch 21/100\n",
            "40000/40000 [==============================] - 23s 584us/step - loss: 0.6859 - acc: 0.7570 - val_loss: 0.6676 - val_acc: 0.7657\n",
            "Epoch 22/100\n",
            "40000/40000 [==============================] - 23s 584us/step - loss: 0.6709 - acc: 0.7665 - val_loss: 0.6841 - val_acc: 0.7580\n",
            "Epoch 23/100\n",
            "40000/40000 [==============================] - 23s 583us/step - loss: 0.6619 - acc: 0.7668 - val_loss: 0.6584 - val_acc: 0.7694\n",
            "Epoch 24/100\n",
            "40000/40000 [==============================] - 23s 579us/step - loss: 0.6535 - acc: 0.7703 - val_loss: 0.6468 - val_acc: 0.7759\n",
            "Epoch 25/100\n",
            "40000/40000 [==============================] - 23s 584us/step - loss: 0.6386 - acc: 0.7749 - val_loss: 0.6733 - val_acc: 0.7669\n",
            "Epoch 26/100\n",
            "40000/40000 [==============================] - 24s 588us/step - loss: 0.6334 - acc: 0.7752 - val_loss: 0.6392 - val_acc: 0.7779\n",
            "Epoch 27/100\n",
            "40000/40000 [==============================] - 23s 582us/step - loss: 0.6212 - acc: 0.7845 - val_loss: 0.6416 - val_acc: 0.7814\n",
            "Epoch 28/100\n",
            "40000/40000 [==============================] - 23s 581us/step - loss: 0.6153 - acc: 0.7829 - val_loss: 0.6332 - val_acc: 0.7797\n",
            "Epoch 29/100\n",
            "40000/40000 [==============================] - 24s 597us/step - loss: 0.6078 - acc: 0.7864 - val_loss: 0.6336 - val_acc: 0.7808\n",
            "Epoch 30/100\n",
            "40000/40000 [==============================] - 23s 580us/step - loss: 0.5958 - acc: 0.7921 - val_loss: 0.6202 - val_acc: 0.7846\n",
            "Epoch 31/100\n",
            "40000/40000 [==============================] - 24s 595us/step - loss: 0.5884 - acc: 0.7936 - val_loss: 0.6166 - val_acc: 0.7884\n",
            "Epoch 32/100\n",
            "40000/40000 [==============================] - 24s 599us/step - loss: 0.5821 - acc: 0.7945 - val_loss: 0.6129 - val_acc: 0.7893\n",
            "Epoch 33/100\n",
            "40000/40000 [==============================] - 24s 597us/step - loss: 0.5737 - acc: 0.7979 - val_loss: 0.6125 - val_acc: 0.7882\n",
            "Epoch 34/100\n",
            "40000/40000 [==============================] - 25s 628us/step - loss: 0.5696 - acc: 0.7980 - val_loss: 0.6156 - val_acc: 0.7880\n",
            "Epoch 35/100\n",
            "40000/40000 [==============================] - 25s 619us/step - loss: 0.5587 - acc: 0.8014 - val_loss: 0.5955 - val_acc: 0.7947\n",
            "Epoch 36/100\n",
            "40000/40000 [==============================] - 25s 616us/step - loss: 0.5479 - acc: 0.8057 - val_loss: 0.5902 - val_acc: 0.7962\n",
            "Epoch 37/100\n",
            "40000/40000 [==============================] - 25s 613us/step - loss: 0.5517 - acc: 0.8051 - val_loss: 0.5828 - val_acc: 0.7983\n",
            "Epoch 38/100\n",
            "40000/40000 [==============================] - 24s 590us/step - loss: 0.5371 - acc: 0.8098 - val_loss: 0.6086 - val_acc: 0.7927\n",
            "Epoch 39/100\n",
            "40000/40000 [==============================] - 23s 587us/step - loss: 0.5291 - acc: 0.8124 - val_loss: 0.5856 - val_acc: 0.7986\n",
            "Epoch 40/100\n",
            "40000/40000 [==============================] - 24s 592us/step - loss: 0.5214 - acc: 0.8151 - val_loss: 0.5724 - val_acc: 0.8073\n",
            "Epoch 41/100\n",
            "40000/40000 [==============================] - 23s 580us/step - loss: 0.5199 - acc: 0.8148 - val_loss: 0.5654 - val_acc: 0.8057\n",
            "Epoch 42/100\n",
            "40000/40000 [==============================] - 24s 590us/step - loss: 0.5095 - acc: 0.8193 - val_loss: 0.5638 - val_acc: 0.8110\n",
            "Epoch 43/100\n",
            "40000/40000 [==============================] - 23s 587us/step - loss: 0.5055 - acc: 0.8204 - val_loss: 0.5670 - val_acc: 0.8068\n",
            "Epoch 44/100\n",
            "40000/40000 [==============================] - 24s 596us/step - loss: 0.4981 - acc: 0.8231 - val_loss: 0.5867 - val_acc: 0.7968\n",
            "Epoch 45/100\n",
            "40000/40000 [==============================] - 24s 603us/step - loss: 0.4885 - acc: 0.8269 - val_loss: 0.5887 - val_acc: 0.8012\n",
            "Epoch 46/100\n",
            "40000/40000 [==============================] - 25s 624us/step - loss: 0.4938 - acc: 0.8262 - val_loss: 0.5862 - val_acc: 0.8000\n",
            "Epoch 47/100\n",
            "40000/40000 [==============================] - 24s 610us/step - loss: 0.4850 - acc: 0.8271 - val_loss: 0.5539 - val_acc: 0.8111\n",
            "Epoch 48/100\n",
            "40000/40000 [==============================] - 25s 617us/step - loss: 0.4785 - acc: 0.8307 - val_loss: 0.5587 - val_acc: 0.8126\n",
            "Epoch 49/100\n",
            "40000/40000 [==============================] - 25s 621us/step - loss: 0.4714 - acc: 0.8347 - val_loss: 0.5548 - val_acc: 0.8113\n",
            "Epoch 50/100\n",
            "40000/40000 [==============================] - 25s 614us/step - loss: 0.4694 - acc: 0.8337 - val_loss: 0.5525 - val_acc: 0.8135\n",
            "Epoch 51/100\n",
            "40000/40000 [==============================] - 25s 622us/step - loss: 0.4618 - acc: 0.8345 - val_loss: 0.5477 - val_acc: 0.8142\n",
            "Epoch 52/100\n",
            "40000/40000 [==============================] - 25s 616us/step - loss: 0.4581 - acc: 0.8364 - val_loss: 0.5381 - val_acc: 0.8188\n",
            "Epoch 53/100\n",
            "40000/40000 [==============================] - 25s 626us/step - loss: 0.4503 - acc: 0.8396 - val_loss: 0.5621 - val_acc: 0.8081\n",
            "Epoch 54/100\n",
            "40000/40000 [==============================] - 25s 622us/step - loss: 0.4473 - acc: 0.8394 - val_loss: 0.5568 - val_acc: 0.8123\n",
            "Epoch 55/100\n",
            "40000/40000 [==============================] - 25s 621us/step - loss: 0.4432 - acc: 0.8430 - val_loss: 0.5527 - val_acc: 0.8149\n",
            "Epoch 56/100\n",
            "40000/40000 [==============================] - 25s 621us/step - loss: 0.4380 - acc: 0.8437 - val_loss: 0.5554 - val_acc: 0.8134\n",
            "Epoch 57/100\n",
            "40000/40000 [==============================] - 25s 621us/step - loss: 0.4360 - acc: 0.8450 - val_loss: 0.5524 - val_acc: 0.8154\n",
            "Epoch 58/100\n",
            "40000/40000 [==============================] - 25s 619us/step - loss: 0.4286 - acc: 0.8465 - val_loss: 0.5578 - val_acc: 0.8096\n",
            "Epoch 59/100\n",
            "40000/40000 [==============================] - 25s 613us/step - loss: 0.4281 - acc: 0.8477 - val_loss: 0.5539 - val_acc: 0.8125\n",
            "Epoch 60/100\n",
            "40000/40000 [==============================] - 25s 619us/step - loss: 0.4167 - acc: 0.8521 - val_loss: 0.5349 - val_acc: 0.8182\n",
            "Epoch 61/100\n",
            "40000/40000 [==============================] - 25s 616us/step - loss: 0.4125 - acc: 0.8551 - val_loss: 0.5295 - val_acc: 0.8237\n",
            "Epoch 62/100\n",
            "40000/40000 [==============================] - 24s 609us/step - loss: 0.4152 - acc: 0.8512 - val_loss: 0.5316 - val_acc: 0.8191\n",
            "Epoch 63/100\n",
            "40000/40000 [==============================] - 24s 610us/step - loss: 0.4002 - acc: 0.8571 - val_loss: 0.5192 - val_acc: 0.8260\n",
            "Epoch 64/100\n",
            "40000/40000 [==============================] - 24s 610us/step - loss: 0.4015 - acc: 0.8575 - val_loss: 0.5261 - val_acc: 0.8243\n",
            "Epoch 65/100\n",
            "40000/40000 [==============================] - 25s 623us/step - loss: 0.3968 - acc: 0.8584 - val_loss: 0.5361 - val_acc: 0.8220\n",
            "Epoch 66/100\n",
            "40000/40000 [==============================] - 25s 613us/step - loss: 0.3911 - acc: 0.8600 - val_loss: 0.5489 - val_acc: 0.8139\n",
            "Epoch 67/100\n",
            "40000/40000 [==============================] - 25s 620us/step - loss: 0.3906 - acc: 0.8599 - val_loss: 0.5235 - val_acc: 0.8264\n",
            "Epoch 68/100\n",
            "40000/40000 [==============================] - 25s 621us/step - loss: 0.3899 - acc: 0.8621 - val_loss: 0.5300 - val_acc: 0.8234\n",
            "Epoch 69/100\n",
            "40000/40000 [==============================] - 25s 614us/step - loss: 0.3853 - acc: 0.8627 - val_loss: 0.5407 - val_acc: 0.8184\n",
            "Epoch 70/100\n",
            "40000/40000 [==============================] - 25s 615us/step - loss: 0.3827 - acc: 0.8638 - val_loss: 0.5340 - val_acc: 0.8205\n",
            "Epoch 71/100\n",
            "40000/40000 [==============================] - 25s 614us/step - loss: 0.3758 - acc: 0.8664 - val_loss: 0.5249 - val_acc: 0.8250\n",
            "Epoch 72/100\n",
            "40000/40000 [==============================] - 24s 608us/step - loss: 0.3726 - acc: 0.8662 - val_loss: 0.5351 - val_acc: 0.8229\n",
            "Epoch 73/100\n",
            "40000/40000 [==============================] - 25s 617us/step - loss: 0.3709 - acc: 0.8669 - val_loss: 0.5282 - val_acc: 0.8251\n",
            "Epoch 74/100\n",
            "40000/40000 [==============================] - 25s 618us/step - loss: 0.3658 - acc: 0.8671 - val_loss: 0.5270 - val_acc: 0.8246\n",
            "Epoch 75/100\n",
            "40000/40000 [==============================] - 25s 615us/step - loss: 0.3596 - acc: 0.8709 - val_loss: 0.5274 - val_acc: 0.8253\n",
            "Epoch 76/100\n",
            "40000/40000 [==============================] - 25s 615us/step - loss: 0.3534 - acc: 0.8737 - val_loss: 0.5299 - val_acc: 0.8254\n",
            "Epoch 77/100\n",
            "40000/40000 [==============================] - 25s 613us/step - loss: 0.3584 - acc: 0.8744 - val_loss: 0.5257 - val_acc: 0.8261\n",
            "Epoch 78/100\n",
            "40000/40000 [==============================] - 24s 610us/step - loss: 0.3498 - acc: 0.8761 - val_loss: 0.5241 - val_acc: 0.8284\n",
            "Epoch 79/100\n",
            "40000/40000 [==============================] - 24s 608us/step - loss: 0.3478 - acc: 0.8761 - val_loss: 0.5341 - val_acc: 0.8224\n",
            "Epoch 80/100\n",
            "40000/40000 [==============================] - 25s 622us/step - loss: 0.3519 - acc: 0.8741 - val_loss: 0.5329 - val_acc: 0.8228\n",
            "Epoch 81/100\n",
            "40000/40000 [==============================] - 25s 617us/step - loss: 0.3460 - acc: 0.8760 - val_loss: 0.5349 - val_acc: 0.8227\n",
            "Epoch 82/100\n",
            "40000/40000 [==============================] - 25s 614us/step - loss: 0.3401 - acc: 0.8770 - val_loss: 0.5248 - val_acc: 0.8248\n",
            "Epoch 83/100\n",
            "40000/40000 [==============================] - 25s 614us/step - loss: 0.3344 - acc: 0.8800 - val_loss: 0.5229 - val_acc: 0.8274\n",
            "Epoch 84/100\n",
            "40000/40000 [==============================] - 25s 618us/step - loss: 0.3346 - acc: 0.8800 - val_loss: 0.5198 - val_acc: 0.8290\n",
            "Epoch 85/100\n",
            "40000/40000 [==============================] - 25s 617us/step - loss: 0.3319 - acc: 0.8831 - val_loss: 0.5239 - val_acc: 0.8292\n",
            "Epoch 86/100\n",
            "40000/40000 [==============================] - 24s 608us/step - loss: 0.3294 - acc: 0.8821 - val_loss: 0.5335 - val_acc: 0.8232\n",
            "Epoch 87/100\n",
            "40000/40000 [==============================] - 24s 606us/step - loss: 0.3283 - acc: 0.8845 - val_loss: 0.5340 - val_acc: 0.8235\n",
            "Epoch 88/100\n",
            "40000/40000 [==============================] - 24s 604us/step - loss: 0.3273 - acc: 0.8847 - val_loss: 0.5320 - val_acc: 0.8310\n",
            "Epoch 89/100\n",
            "40000/40000 [==============================] - 25s 615us/step - loss: 0.3231 - acc: 0.8851 - val_loss: 0.5240 - val_acc: 0.8284\n",
            "Epoch 90/100\n",
            "40000/40000 [==============================] - 24s 604us/step - loss: 0.3186 - acc: 0.8858 - val_loss: 0.5188 - val_acc: 0.8276\n",
            "Epoch 91/100\n",
            "40000/40000 [==============================] - 24s 604us/step - loss: 0.3183 - acc: 0.8876 - val_loss: 0.5178 - val_acc: 0.8284\n",
            "Epoch 92/100\n",
            "40000/40000 [==============================] - 24s 602us/step - loss: 0.3080 - acc: 0.8876 - val_loss: 0.5221 - val_acc: 0.8325\n",
            "Epoch 93/100\n",
            "40000/40000 [==============================] - 24s 604us/step - loss: 0.3128 - acc: 0.8864 - val_loss: 0.5157 - val_acc: 0.8313\n",
            "Epoch 94/100\n",
            "40000/40000 [==============================] - 24s 602us/step - loss: 0.3099 - acc: 0.8888 - val_loss: 0.5303 - val_acc: 0.8284\n",
            "Epoch 95/100\n",
            "40000/40000 [==============================] - 24s 598us/step - loss: 0.3075 - acc: 0.8908 - val_loss: 0.5355 - val_acc: 0.8281\n",
            "Epoch 96/100\n",
            "40000/40000 [==============================] - 24s 609us/step - loss: 0.2983 - acc: 0.8916 - val_loss: 0.5206 - val_acc: 0.8301\n",
            "Epoch 97/100\n",
            "40000/40000 [==============================] - 24s 595us/step - loss: 0.3030 - acc: 0.8902 - val_loss: 0.5373 - val_acc: 0.8248\n",
            "Epoch 98/100\n",
            "40000/40000 [==============================] - 24s 591us/step - loss: 0.2994 - acc: 0.8920 - val_loss: 0.5311 - val_acc: 0.8276\n",
            "Epoch 99/100\n",
            "40000/40000 [==============================] - 24s 592us/step - loss: 0.2930 - acc: 0.8938 - val_loss: 0.5240 - val_acc: 0.8298\n",
            "Epoch 100/100\n",
            "40000/40000 [==============================] - 24s 590us/step - loss: 0.2916 - acc: 0.8955 - val_loss: 0.5339 - val_acc: 0.8293\n",
            "Test loss: 0.5339189003705979\n",
            "Test accuracy: 0.8293\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfUAAAFYCAYAAABKymUhAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 576x396 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n",
            "\n",
            "validation loss: 0.5339189003705979\n",
            "validation accuracy: 0.8293\n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}