CNN_Keras_MNIST.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "CNN_Keras_MNIST.ipynb",
      "provenance": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/young-hwanlee/CNN_Keras_MNIST/blob/master/CNN_Keras_MNIST.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "i4J3mTgsyNJC"
      },
      "source": [
        "# ** There're not much to modify since this code is written using the high-level APIs(tf.keras).**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "G_4Dt4KeS0yy",
        "outputId": "0b5d6926-6a23-4d6c-82a9-07360b489d03"
      },
      "source": [
        "!git clone https://github.com/young-hwanlee/CNN_Keras_MNIST.git"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "fatal: destination path 'CNN_Keras_MNIST' already exists and is not an empty directory.\n",
            "fatal: destination path 'CNN_Keras_MNIST' already exists and is not an empty directory.\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "v0aL26uZS8rj",
        "outputId": "83d7cba3-d4b6-4059-9477-359b614edfc5"
      },
      "source": [
        "ls -ltr"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "total 8\n",
            "drwxr-xr-x 1 root root 4096 Sep  1 19:26 \u001b[0m\u001b[01;34msample_data\u001b[0m/\n",
            "drwxr-xr-x 3 root root 4096 Sep  3 06:53 \u001b[01;34mCNN_Keras_MNIST\u001b[0m/\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2e8jH6AGHY0x",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "b0182ba5-7aca-4099-92b3-629f2cccfcb6"
      },
      "source": [
        "#%%\n",
        "import os\n",
        "os.environ['KMP_DUPLICATE_LIB_OK'] = 'True'\n",
        "\n",
        "## Import libraries and modules\n",
        "import sys\n",
        "import keras\n",
        "import tensorflow as tf\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "# from tensorflow.python.keras.models import Sequential\n",
        "# from tensorflow.python.keras.layers import Dense, Dropout, Activation, Flatten\n",
        "# from tensorflow.python.keras.layers import Convolution2D, MaxPooling2D\n",
        "# from tensorflow.python.keras import utils\n",
        "# from tensorflow.python.keras.datasets import mnist\n",
        "# from tensorflow.python.keras.callbacks import ReduceLROnPlateau\n",
        "\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Dense, Dropout, Activation, Flatten\n",
        "from keras.layers import Convolution2D, MaxPooling2D\n",
        "from keras import utils\n",
        "from keras.datasets import mnist\n",
        "from keras.callbacks import ReduceLROnPlateau\n",
        "\n",
        "## Set up the seed for reproducibility.\n",
        "seed=42\n",
        "\n",
        "# tf.reset_default_graph()           # It seems like it's only working in the old version.\n",
        "# tf.set_random_seed(seed)       # It seems like it's only working in the old version.\n",
        "\n",
        "tf.compat.v1.reset_default_graph()\n",
        "tf.compat.v1.set_random_seed(seed)\n",
        "np.random.seed(seed)\n",
        "\n",
        "print('Python version : ', sys.version)\n",
        "print('TensorFlow version : ', tf.__version__)\n",
        "print('Keras version : ', keras.__version__)"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Python version :  3.7.11 (default, Jul  3 2021, 18:01:19) \n",
            "[GCC 7.5.0]\n",
            "TensorFlow version :  2.6.0\n",
            "Keras version :  2.6.0\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "50N6LottWVbS",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "0e5f3b3b-a899-4a72-ee84-3bf341ade784"
      },
      "source": [
        "## Load pre-shuffled MNIST data into train and test sets\n",
        "(X_train, y_train), (X_test, y_test) = mnist.load_data()\n",
        "\n",
        "## Preprocess input data\n",
        "img_rows = 28\n",
        "img_cols = 28\n",
        "input_shape = (img_rows, img_cols, 1)\n",
        "\n",
        "print('Training data size before reshaping :', X_train.shape)\n",
        "print('Test data size before reshaping :', X_test.shape)\n",
        "\n",
        "X_train = X_train.reshape(X_train.shape[0], img_rows, img_cols, 1)\n",
        "X_test = X_test.reshape(X_test.shape[0], img_rows, img_cols, 1)\n",
        "# 1 for black&white image, i.e. (img_rows, img_cols, 1)\n",
        "# 3 for RGB image, i.e. (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",
        "X_train = X_train.astype('float32') / 255\n",
        "X_test = X_test.astype('float32') / 255\n",
        "\n",
        "## Reserve 10,000 samples for validation\n",
        "X_train = X_train[:-10000]\n",
        "y_train = y_train[:-10000]\n",
        "X_val = X_train[-10000:]\n",
        "y_val = y_train[-10000:]\n",
        "\n",
        "print('Training data size after reshaping :', X_train.shape)\n",
        "print('Validation data size after reshaping :', X_val.shape)\n",
        "print('Test data size after reshaping :', X_test.shape)\n",
        "\n",
        "## Preprocess class labels\n",
        "num_classes = 10\n",
        "batch_size = 32\n",
        "epochs = 10\n",
        "\n",
        "# Y_train = utils.to_categorical(y_train, num_classes)     # It seems like it's only working in the old version.\n",
        "# Y_val = utils.to_categorical(y_val, num_classes)            # It seems like it's only working in the old version.\n",
        "# Y_test = utils.to_categorical(y_test, num_classes)         # It seems like it's only working in the old version.\n",
        "\n",
        "Y_train = tf.keras.utils.to_categorical(y_train, num_classes)\n",
        "Y_val = tf.keras.utils.to_categorical(y_val, num_classes)\n",
        "Y_test = tf.keras.utils.to_categorical(y_test, num_classes)"
      ],
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Training data size before reshaping : (60000, 28, 28)\n",
            "Test data size before reshaping : (10000, 28, 28)\n",
            "Training data size after reshaping : (50000, 28, 28, 1)\n",
            "Validation data size after reshaping : (10000, 28, 28, 1)\n",
            "Test data size after reshaping : (10000, 28, 28, 1)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Cu7FbMhmWgsT",
        "outputId": "8c95c98a-5b60-4a60-d930-998c4e81917c"
      },
      "source": [
        "## Define model architecture\n",
        "model = Sequential()\n",
        "model.add(Convolution2D(filters=32, kernel_size=(3,3), strides=(3,3),\n",
        "                        activation='relu', input_shape=input_shape))\n",
        "model.add(Convolution2D(filters=32, kernel_size=(3,3), strides=(3,3),\n",
        "                        activation='relu'))\n",
        "model.add(MaxPooling2D(pool_size=(2,2)))\n",
        "model.add(Dropout(0.25))\n",
        "\n",
        "model.add(Flatten())\n",
        "model.add(Dense(128, activation='relu'))\n",
        "model.add(Dropout(0.5))\n",
        "model.add(Dense(num_classes, activation='softmax'))\n",
        "model.summary()"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d (Conv2D)              (None, 9, 9, 32)          320       \n",
            "_________________________________________________________________\n",
            "conv2d_1 (Conv2D)            (None, 3, 3, 32)          9248      \n",
            "_________________________________________________________________\n",
            "max_pooling2d (MaxPooling2D) (None, 1, 1, 32)          0         \n",
            "_________________________________________________________________\n",
            "dropout (Dropout)            (None, 1, 1, 32)          0         \n",
            "_________________________________________________________________\n",
            "flatten (Flatten)            (None, 32)                0         \n",
            "_________________________________________________________________\n",
            "dense (Dense)                (None, 128)               4224      \n",
            "_________________________________________________________________\n",
            "dropout_1 (Dropout)          (None, 128)               0         \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 10)                1290      \n",
            "=================================================================\n",
            "Total params: 15,082\n",
            "Trainable params: 15,082\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TB7RRrFJ3tu6"
      },
      "source": [
        "Param # = (filter size) x (# input channel(B&W)) x (# output channel) + (bias of output channel)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "aWSYlAV_4_0v"
      },
      "source": [
        "320 = (3x3) x 1 x 32 + 32"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Tj_D8Dah5D2M"
      },
      "source": [
        "9248 = (3x3) x 32 x 32 + *32*"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fhbfXpmQB6sE"
      },
      "source": [
        "4224 = (1x1) x 32 x 128 + 128"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kJu0I0n1CXdo"
      },
      "source": [
        "1290 = (1x1) x 128 x 10 + 10"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 886
        },
        "id": "IDZNWAd3eVmp",
        "outputId": "00d1c564-df55-4d55-e938-46e80e5c4ec8"
      },
      "source": [
        "from IPython.display import SVG\n",
        "from keras.utils.vis_utils import model_to_dot\n",
        "%matplotlib inline\n",
        "\n",
        "SVG(model_to_dot(model, show_shapes=True, dpi=65).create(prog='dot', format='svg'))"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<IPython.core.display.SVG object>"
            ],
            "image/svg+xml": "<svg height=\"649pt\" viewBox=\"0.00 0.00 382.00 719.00\" width=\"345pt\" xmlns=\"http://www.w3.org/2000/svg\" xmlns:xlink=\"http://www.w3.org/1999/xlink\">\n<g class=\"graph\" id=\"graph0\" transform=\"scale(.9028 .9028) rotate(0) translate(4 715)\">\n<title>G</title>\n<polygon fill=\"#ffffff\" points=\"-4,4 -4,-715 378,-715 378,4 -4,4\" stroke=\"transparent\"/>\n<!-- 139689654008464 -->\n<g class=\"node\" id=\"node1\">\n<title>139689654008464</title>\n<polygon fill=\"none\" points=\"10.5,-664.5 10.5,-710.5 363.5,-710.5 363.5,-664.5 10.5,-664.5\" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"95\" y=\"-683.8\">conv2d_input: InputLayer</text>\n<polyline fill=\"none\" points=\"179.5,-664.5 179.5,-710.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"208.5\" y=\"-695.3\">input:</text>\n<polyline fill=\"none\" points=\"179.5,-687.5 237.5,-687.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"208.5\" y=\"-672.3\">output:</text>\n<polyline fill=\"none\" points=\"237.5,-664.5 237.5,-710.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"300.5\" y=\"-695.3\">[(None, 28, 28, 1)]</text>\n<polyline fill=\"none\" points=\"237.5,-687.5 363.5,-687.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"300.5\" y=\"-672.3\">[(None, 28, 28, 1)]</text>\n</g>\n<!-- 139689549481424 -->\n<g class=\"node\" id=\"node2\">\n<title>139689549481424</title>\n<polygon fill=\"none\" points=\"40.5,-581.5 40.5,-627.5 333.5,-627.5 333.5,-581.5 40.5,-581.5\" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"99.5\" y=\"-600.8\">conv2d: Conv2D</text>\n<polyline fill=\"none\" points=\"158.5,-581.5 158.5,-627.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"187.5\" y=\"-612.3\">input:</text>\n<polyline fill=\"none\" points=\"158.5,-604.5 216.5,-604.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"187.5\" y=\"-589.3\">output:</text>\n<polyline fill=\"none\" points=\"216.5,-581.5 216.5,-627.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"275\" y=\"-612.3\">(None, 28, 28, 1)</text>\n<polyline fill=\"none\" points=\"216.5,-604.5 333.5,-604.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"275\" y=\"-589.3\">(None, 9, 9, 32)</text>\n</g>\n<!-- 139689654008464&#45;&gt;139689549481424 -->\n<g class=\"edge\" id=\"edge1\">\n<title>139689654008464-&gt;139689549481424</title>\n<path d=\"M187,-664.3799C187,-656.1745 187,-646.7679 187,-637.8786\" fill=\"none\" stroke=\"#000000\"/>\n<polygon fill=\"#000000\" points=\"190.5001,-637.784 187,-627.784 183.5001,-637.784 190.5001,-637.784\" stroke=\"#000000\"/>\n</g>\n<!-- 139689553346448 -->\n<g class=\"node\" id=\"node3\">\n<title>139689553346448</title>\n<polygon fill=\"none\" points=\"36.5,-498.5 36.5,-544.5 337.5,-544.5 337.5,-498.5 36.5,-498.5\" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"103\" y=\"-517.8\">conv2d_1: Conv2D</text>\n<polyline fill=\"none\" points=\"169.5,-498.5 169.5,-544.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"198.5\" y=\"-529.3\">input:</text>\n<polyline fill=\"none\" points=\"169.5,-521.5 227.5,-521.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"198.5\" y=\"-506.3\">output:</text>\n<polyline fill=\"none\" points=\"227.5,-498.5 227.5,-544.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"282.5\" y=\"-529.3\">(None, 9, 9, 32)</text>\n<polyline fill=\"none\" points=\"227.5,-521.5 337.5,-521.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"282.5\" y=\"-506.3\">(None, 3, 3, 32)</text>\n</g>\n<!-- 139689549481424&#45;&gt;139689553346448 -->\n<g class=\"edge\" id=\"edge2\">\n<title>139689549481424-&gt;139689553346448</title>\n<path d=\"M187,-581.3799C187,-573.1745 187,-563.7679 187,-554.8786\" fill=\"none\" stroke=\"#000000\"/>\n<polygon fill=\"#000000\" points=\"190.5001,-554.784 187,-544.784 183.5001,-554.784 190.5001,-554.784\" stroke=\"#000000\"/>\n</g>\n<!-- 139689489068240 -->\n<g class=\"node\" id=\"node4\">\n<title>139689489068240</title>\n<polygon fill=\"none\" points=\"0,-415.5 0,-461.5 374,-461.5 374,-415.5 0,-415.5\" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"103\" y=\"-434.8\">max_pooling2d: MaxPooling2D</text>\n<polyline fill=\"none\" points=\"206,-415.5 206,-461.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"235\" y=\"-446.3\">input:</text>\n<polyline fill=\"none\" points=\"206,-438.5 264,-438.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"235\" y=\"-423.3\">output:</text>\n<polyline fill=\"none\" points=\"264,-415.5 264,-461.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"319\" y=\"-446.3\">(None, 3, 3, 32)</text>\n<polyline fill=\"none\" points=\"264,-438.5 374,-438.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"319\" y=\"-423.3\">(None, 1, 1, 32)</text>\n</g>\n<!-- 139689553346448&#45;&gt;139689489068240 -->\n<g class=\"edge\" id=\"edge3\">\n<title>139689553346448-&gt;139689489068240</title>\n<path d=\"M187,-498.3799C187,-490.1745 187,-480.7679 187,-471.8786\" fill=\"none\" stroke=\"#000000\"/>\n<polygon fill=\"#000000\" points=\"190.5001,-471.784 187,-461.784 183.5001,-471.784 190.5001,-471.784\" stroke=\"#000000\"/>\n</g>\n<!-- 139689489071504 -->\n<g class=\"node\" id=\"node5\">\n<title>139689489071504</title>\n<polygon fill=\"none\" points=\"43.5,-332.5 43.5,-378.5 330.5,-378.5 330.5,-332.5 43.5,-332.5\" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"103\" y=\"-351.8\">dropout: Dropout</text>\n<polyline fill=\"none\" points=\"162.5,-332.5 162.5,-378.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"191.5\" y=\"-363.3\">input:</text>\n<polyline fill=\"none\" points=\"162.5,-355.5 220.5,-355.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"191.5\" y=\"-340.3\">output:</text>\n<polyline fill=\"none\" points=\"220.5,-332.5 220.5,-378.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"275.5\" y=\"-363.3\">(None, 1, 1, 32)</text>\n<polyline fill=\"none\" points=\"220.5,-355.5 330.5,-355.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"275.5\" y=\"-340.3\">(None, 1, 1, 32)</text>\n</g>\n<!-- 139689489068240&#45;&gt;139689489071504 -->\n<g class=\"edge\" id=\"edge4\">\n<title>139689489068240-&gt;139689489071504</title>\n<path d=\"M187,-415.3799C187,-407.1745 187,-397.7679 187,-388.8786\" fill=\"none\" stroke=\"#000000\"/>\n<polygon fill=\"#000000\" points=\"190.5001,-388.784 187,-378.784 183.5001,-388.784 190.5001,-388.784\" stroke=\"#000000\"/>\n</g>\n<!-- 139689488422864 -->\n<g class=\"node\" id=\"node6\">\n<title>139689488422864</title>\n<polygon fill=\"none\" points=\"54,-249.5 54,-295.5 320,-295.5 320,-249.5 54,-249.5\" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"103\" y=\"-268.8\">flatten: Flatten</text>\n<polyline fill=\"none\" points=\"152,-249.5 152,-295.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"181\" y=\"-280.3\">input:</text>\n<polyline fill=\"none\" points=\"152,-272.5 210,-272.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"181\" y=\"-257.3\">output:</text>\n<polyline fill=\"none\" points=\"210,-249.5 210,-295.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"265\" y=\"-280.3\">(None, 1, 1, 32)</text>\n<polyline fill=\"none\" points=\"210,-272.5 320,-272.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"265\" y=\"-257.3\">(None, 32)</text>\n</g>\n<!-- 139689489071504&#45;&gt;139689488422864 -->\n<g class=\"edge\" id=\"edge5\">\n<title>139689489071504-&gt;139689488422864</title>\n<path d=\"M187,-332.3799C187,-324.1745 187,-314.7679 187,-305.8786\" fill=\"none\" stroke=\"#000000\"/>\n<polygon fill=\"#000000\" points=\"190.5001,-305.784 187,-295.784 183.5001,-305.784 190.5001,-305.784\" stroke=\"#000000\"/>\n</g>\n<!-- 139689489068176 -->\n<g class=\"node\" id=\"node7\">\n<title>139689489068176</title>\n<polygon fill=\"none\" points=\"68.5,-166.5 68.5,-212.5 305.5,-212.5 305.5,-166.5 68.5,-166.5\" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"114.5\" y=\"-185.8\">dense: Dense</text>\n<polyline fill=\"none\" points=\"160.5,-166.5 160.5,-212.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"189.5\" y=\"-197.3\">input:</text>\n<polyline fill=\"none\" points=\"160.5,-189.5 218.5,-189.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"189.5\" y=\"-174.3\">output:</text>\n<polyline fill=\"none\" points=\"218.5,-166.5 218.5,-212.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"262\" y=\"-197.3\">(None, 32)</text>\n<polyline fill=\"none\" points=\"218.5,-189.5 305.5,-189.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"262\" y=\"-174.3\">(None, 128)</text>\n</g>\n<!-- 139689488422864&#45;&gt;139689489068176 -->\n<g class=\"edge\" id=\"edge6\">\n<title>139689488422864-&gt;139689489068176</title>\n<path d=\"M187,-249.3799C187,-241.1745 187,-231.7679 187,-222.8786\" fill=\"none\" stroke=\"#000000\"/>\n<polygon fill=\"#000000\" points=\"190.5001,-222.784 187,-212.784 183.5001,-222.784 190.5001,-222.784\" stroke=\"#000000\"/>\n</g>\n<!-- 139689488244624 -->\n<g class=\"node\" id=\"node8\">\n<title>139689488244624</title>\n<polygon fill=\"none\" points=\"47.5,-83.5 47.5,-129.5 326.5,-129.5 326.5,-83.5 47.5,-83.5\" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"114.5\" y=\"-102.8\">dropout_1: Dropout</text>\n<polyline fill=\"none\" points=\"181.5,-83.5 181.5,-129.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"210.5\" y=\"-114.3\">input:</text>\n<polyline fill=\"none\" points=\"181.5,-106.5 239.5,-106.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"210.5\" y=\"-91.3\">output:</text>\n<polyline fill=\"none\" points=\"239.5,-83.5 239.5,-129.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"283\" y=\"-114.3\">(None, 128)</text>\n<polyline fill=\"none\" points=\"239.5,-106.5 326.5,-106.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"283\" y=\"-91.3\">(None, 128)</text>\n</g>\n<!-- 139689489068176&#45;&gt;139689488244624 -->\n<g class=\"edge\" id=\"edge7\">\n<title>139689489068176-&gt;139689488244624</title>\n<path d=\"M187,-166.3799C187,-158.1745 187,-148.7679 187,-139.8786\" fill=\"none\" stroke=\"#000000\"/>\n<polygon fill=\"#000000\" points=\"190.5001,-139.784 187,-129.784 183.5001,-139.784 190.5001,-139.784\" stroke=\"#000000\"/>\n</g>\n<!-- 139689488163920 -->\n<g class=\"node\" id=\"node9\">\n<title>139689488163920</title>\n<polygon fill=\"none\" points=\"61,-.5 61,-46.5 313,-46.5 313,-.5 61,-.5\" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"114.5\" y=\"-19.8\">dense_1: Dense</text>\n<polyline fill=\"none\" points=\"168,-.5 168,-46.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"197\" y=\"-31.3\">input:</text>\n<polyline fill=\"none\" points=\"168,-23.5 226,-23.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"197\" y=\"-8.3\">output:</text>\n<polyline fill=\"none\" points=\"226,-.5 226,-46.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"269.5\" y=\"-31.3\">(None, 128)</text>\n<polyline fill=\"none\" points=\"226,-23.5 313,-23.5 \" stroke=\"#000000\"/>\n<text fill=\"#000000\" font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"269.5\" y=\"-8.3\">(None, 10)</text>\n</g>\n<!-- 139689488244624&#45;&gt;139689488163920 -->\n<g class=\"edge\" id=\"edge8\">\n<title>139689488244624-&gt;139689488163920</title>\n<path d=\"M187,-83.3799C187,-75.1745 187,-65.7679 187,-56.8786\" fill=\"none\" stroke=\"#000000\"/>\n<polygon fill=\"#000000\" points=\"190.5001,-56.784 187,-46.784 183.5001,-56.784 190.5001,-56.784\" stroke=\"#000000\"/>\n</g>\n</g>\n</svg>"
          },
          "metadata": {},
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "NDueMMdOdx83",
        "outputId": "7e6be115-e97f-457e-c6a6-c84d595b6d1c"
      },
      "source": [
        "## Compile model\n",
        "model.compile(loss='categorical_crossentropy',\n",
        "              optimizer='adam',\n",
        "              metrics=['accuracy'])\n",
        "\n",
        "## Fit model on training data\n",
        "# history = model.fit(X_train, Y_train, validation_data=(X_val, Y_val),\n",
        "#                  batch_size=batch_size, nb_epoch=epochs, verbose=1)           # It seems like it's only working in the old version.\n",
        "history = model.fit(X_train, Y_train, validation_data=(X_val, Y_val),\n",
        "                 batch_size=batch_size, epochs=epochs, verbose=1)\n",
        "\n",
        "## Evaluate model on test data\n",
        "score = model.evaluate(X_test, Y_test, verbose=0)"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/10\n",
            "1563/1563 [==============================] - 9s 5ms/step - loss: 0.9647 - accuracy: 0.6926 - val_loss: 0.4810 - val_accuracy: 0.8527\n",
            "Epoch 2/10\n",
            "1563/1563 [==============================] - 8s 5ms/step - loss: 0.6097 - accuracy: 0.8154 - val_loss: 0.3721 - val_accuracy: 0.8856\n",
            "Epoch 3/10\n",
            "1563/1563 [==============================] - 8s 5ms/step - loss: 0.5265 - accuracy: 0.8392 - val_loss: 0.3186 - val_accuracy: 0.9032\n",
            "Epoch 4/10\n",
            "1563/1563 [==============================] - 8s 5ms/step - loss: 0.4846 - accuracy: 0.8515 - val_loss: 0.2978 - val_accuracy: 0.9085\n",
            "Epoch 5/10\n",
            "1563/1563 [==============================] - 8s 5ms/step - loss: 0.4529 - accuracy: 0.8630 - val_loss: 0.2723 - val_accuracy: 0.9194\n",
            "Epoch 6/10\n",
            "1563/1563 [==============================] - 8s 5ms/step - loss: 0.4390 - accuracy: 0.8685 - val_loss: 0.2595 - val_accuracy: 0.9207\n",
            "Epoch 7/10\n",
            "1563/1563 [==============================] - 8s 5ms/step - loss: 0.4210 - accuracy: 0.8741 - val_loss: 0.2495 - val_accuracy: 0.9239\n",
            "Epoch 8/10\n",
            "1563/1563 [==============================] - 8s 5ms/step - loss: 0.4143 - accuracy: 0.8759 - val_loss: 0.2443 - val_accuracy: 0.9252\n",
            "Epoch 9/10\n",
            "1563/1563 [==============================] - 8s 5ms/step - loss: 0.4011 - accuracy: 0.8795 - val_loss: 0.2310 - val_accuracy: 0.9305\n",
            "Epoch 10/10\n",
            "1563/1563 [==============================] - 8s 5ms/step - loss: 0.3923 - accuracy: 0.8816 - val_loss: 0.2341 - val_accuracy: 0.9281\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 283
        },
        "id": "eQWEb7qgWufp",
        "outputId": "7fb4cb5b-f889-445f-ba23-084cc9eb0ff4"
      },
      "source": [
        "#%%\n",
        "## Check the results\n",
        "# Plots the loss\n",
        "plt.figure()\n",
        "plt.plot(history.history['loss'])\n",
        "plt.plot(history.history['val_loss'])\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'validation'], loc='upper right')\n",
        "plt.show()"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "id": "VAcprsz8XuXM",
        "outputId": "0b58c9a1-30f0-40ee-f393-6564e8c18f9c"
      },
      "source": [
        "# Plots the accuracy\n",
        "plt.figure()\n",
        "# plt.plot(history.history['acc'])                  # It seems like it's only working in the old version.\n",
        "# plt.plot(history.history['val_acc'])          # It seems like it's only working in the old version.\n",
        "plt.plot(history.history['accuracy'])\n",
        "plt.plot(history.history['val_accuracy'])\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'validation'], loc='upper left')\n",
        "plt.show()"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 372
        },
        "id": "5vVUhq-cXwaI",
        "outputId": "bf644640-6358-46f7-e7fd-42e32dd28db1"
      },
      "source": [
        "## Predict using the test dataset\n",
        "Y_pred = model.predict(X_test)\n",
        "X_test__ = X_test.reshape(X_test.shape[0], img_rows, img_cols)\n",
        "\n",
        "# fig, axis = plt.subplots(2, 5, figsize=(12, 14))          # The gap between the 1st row and the 2nd row is large.\n",
        "fig, axis = plt.subplots(2, 5, figsize=(12,6))\n",
        "for i, ax in enumerate(axis.flat):\n",
        "    ax.imshow(X_test__[i], cmap='binary')\n",
        "    ax.set(title = f\"Real Number is {Y_test[i].argmax()}\\n\"\n",
        "                   f\"Predict Number is {Y_pred[i].argmax()}\")"
      ],
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 864x432 with 10 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}