NeuralNetwork_Notes/Keras_Convolutional(CNN)_NeuralNetwork_Notes.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Keras_Convolutional(CNN)_NeuralNetwork_Notes.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Dlg4Ic0Lin0W"
      },
      "source": [
        "# Convolutional Neural Network \n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PPIDqyGCB7--"
      },
      "source": [
        ""
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ImRqYPEtB8PZ"
      },
      "source": [
        ""
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "L-0xw8MpB8dC"
      },
      "source": [
        ""
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZvhSSEufB8mg"
      },
      "source": [
        ""
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OV1hbCNeIqge"
      },
      "source": [
        "import tensorflow as tf \n",
        "from tensorflow.keras import datasets, layers, models \n",
        "import matplotlib.pyplot as plt \n",
        "import numpy as np "
      ],
      "execution_count": 81,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ztVHPzzFjUKV",
        "outputId": "4785fadb-6a6a-4e69-fc16-ccb02b7adc6a"
      },
      "source": [
        "# https://www.cs.toronto.edu/~kriz/cifar.html\n",
        "(X_train, y_train), (X_test, y_test) = datasets.cifar10.load_data()\n",
        "print('X_train shape:', X_train.shape)\n",
        "print('y_train shape:', y_train.shape)\n",
        "print('X_test shape:', X_test.shape)\n",
        "print('y_test shape:', y_test.shape)"
      ],
      "execution_count": 124,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "X_train shape: (50000, 32, 32, 3)\n",
            "y_train shape: (50000, 1)\n",
            "X_test shape: (10000, 32, 32, 3)\n",
            "y_test shape: (10000, 1)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "TuuSC7V7ja9Z",
        "outputId": "72378f7f-edb7-4b14-e66b-179d45ebb433"
      },
      "source": [
        "# check a sample array\n",
        "X_train[0]"
      ],
      "execution_count": 83,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[[ 59,  62,  63],\n",
              "        [ 43,  46,  45],\n",
              "        [ 50,  48,  43],\n",
              "        ...,\n",
              "        [158, 132, 108],\n",
              "        [152, 125, 102],\n",
              "        [148, 124, 103]],\n",
              "\n",
              "       [[ 16,  20,  20],\n",
              "        [  0,   0,   0],\n",
              "        [ 18,   8,   0],\n",
              "        ...,\n",
              "        [123,  88,  55],\n",
              "        [119,  83,  50],\n",
              "        [122,  87,  57]],\n",
              "\n",
              "       [[ 25,  24,  21],\n",
              "        [ 16,   7,   0],\n",
              "        [ 49,  27,   8],\n",
              "        ...,\n",
              "        [118,  84,  50],\n",
              "        [120,  84,  50],\n",
              "        [109,  73,  42]],\n",
              "\n",
              "       ...,\n",
              "\n",
              "       [[208, 170,  96],\n",
              "        [201, 153,  34],\n",
              "        [198, 161,  26],\n",
              "        ...,\n",
              "        [160, 133,  70],\n",
              "        [ 56,  31,   7],\n",
              "        [ 53,  34,  20]],\n",
              "\n",
              "       [[180, 139,  96],\n",
              "        [173, 123,  42],\n",
              "        [186, 144,  30],\n",
              "        ...,\n",
              "        [184, 148,  94],\n",
              "        [ 97,  62,  34],\n",
              "        [ 83,  53,  34]],\n",
              "\n",
              "       [[177, 144, 116],\n",
              "        [168, 129,  94],\n",
              "        [179, 142,  87],\n",
              "        ...,\n",
              "        [216, 184, 140],\n",
              "        [151, 118,  84],\n",
              "        [123,  92,  72]]], dtype=uint8)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 83
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 159
        },
        "id": "64DCVslhj_j8",
        "outputId": "9472ff59-6b36-4abc-d43e-229d96385991"
      },
      "source": [
        "### check a sample image ###\n",
        "plt.figure(figsize = (15,2))\n",
        "plt.imshow(X_train[0])\n",
        "plt.show()"
      ],
      "execution_count": 88,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x144 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "VPcmqXu9JUFB",
        "outputId": "2446f84e-bc47-4db3-b98d-65f907a986fc"
      },
      "source": [
        "### reshape the y_train data ### \n",
        "print('--- y_train ---')\n",
        "print(y_train[:5])\n",
        "print('--- y_train reshaped ---')\n",
        "y_train = y_train.reshape(-1,)\n",
        "y_train[:5]"
      ],
      "execution_count": 125,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- y_train ---\n",
            "[[6]\n",
            " [9]\n",
            " [9]\n",
            " [4]\n",
            " [1]]\n",
            "--- y_train reshaped ---\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([6, 9, 9, 4, 1], dtype=uint8)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 125
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "c5nNd1VMlo2J",
        "outputId": "9b957b6e-624c-4656-ae75-902a5585cf02"
      },
      "source": [
        "### set the classes for the categories ###\n",
        "classes = ['airplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck']\n",
        "print('class test #9:', classes[9])"
      ],
      "execution_count": 126,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "class test #9: truck\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kbNokIquj7IX"
      },
      "source": [
        "### a function to show the sample image ###\n",
        "def plot_sample(X, y, index):\n",
        "  plt.figure(figsize = (15,2))\n",
        "  plt.imshow(X[index])\n",
        "  plt.xlabel(classes[y[index]])"
      ],
      "execution_count": 127,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 190
        },
        "id": "ovDP3ZKdl9Aa",
        "outputId": "dd82a501-ffbd-4efa-f39d-41ee9117309b"
      },
      "source": [
        "print('--- sample image ---')\n",
        "plot_sample(X_train, y_train, 3)"
      ],
      "execution_count": 96,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- sample image ---\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x144 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2bkO5WC9kT1O",
        "outputId": "0c227d6c-ab5e-4020-8efd-e385f9dd5c37"
      },
      "source": [
        "### normalize all the data ### \n",
        "print('X_train shape:', X_train[0])\n",
        "X_train = X_train/255\n",
        "print('X_train normalized shape:', X_train[0])\n",
        "print('X_test shape:', X_test[0])\n",
        "X_test = X_test/255\n",
        "print('X_test normalized shape:', X_test[0])"
      ],
      "execution_count": 129,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "X_train shape: [[[ 59  62  63]\n",
            "  [ 43  46  45]\n",
            "  [ 50  48  43]\n",
            "  ...\n",
            "  [158 132 108]\n",
            "  [152 125 102]\n",
            "  [148 124 103]]\n",
            "\n",
            " [[ 16  20  20]\n",
            "  [  0   0   0]\n",
            "  [ 18   8   0]\n",
            "  ...\n",
            "  [123  88  55]\n",
            "  [119  83  50]\n",
            "  [122  87  57]]\n",
            "\n",
            " [[ 25  24  21]\n",
            "  [ 16   7   0]\n",
            "  [ 49  27   8]\n",
            "  ...\n",
            "  [118  84  50]\n",
            "  [120  84  50]\n",
            "  [109  73  42]]\n",
            "\n",
            " ...\n",
            "\n",
            " [[208 170  96]\n",
            "  [201 153  34]\n",
            "  [198 161  26]\n",
            "  ...\n",
            "  [160 133  70]\n",
            "  [ 56  31   7]\n",
            "  [ 53  34  20]]\n",
            "\n",
            " [[180 139  96]\n",
            "  [173 123  42]\n",
            "  [186 144  30]\n",
            "  ...\n",
            "  [184 148  94]\n",
            "  [ 97  62  34]\n",
            "  [ 83  53  34]]\n",
            "\n",
            " [[177 144 116]\n",
            "  [168 129  94]\n",
            "  [179 142  87]\n",
            "  ...\n",
            "  [216 184 140]\n",
            "  [151 118  84]\n",
            "  [123  92  72]]]\n",
            "X_train normalized shape: [[[0.23137255 0.24313725 0.24705882]\n",
            "  [0.16862745 0.18039216 0.17647059]\n",
            "  [0.19607843 0.18823529 0.16862745]\n",
            "  ...\n",
            "  [0.61960784 0.51764706 0.42352941]\n",
            "  [0.59607843 0.49019608 0.4       ]\n",
            "  [0.58039216 0.48627451 0.40392157]]\n",
            "\n",
            " [[0.0627451  0.07843137 0.07843137]\n",
            "  [0.         0.         0.        ]\n",
            "  [0.07058824 0.03137255 0.        ]\n",
            "  ...\n",
            "  [0.48235294 0.34509804 0.21568627]\n",
            "  [0.46666667 0.3254902  0.19607843]\n",
            "  [0.47843137 0.34117647 0.22352941]]\n",
            "\n",
            " [[0.09803922 0.09411765 0.08235294]\n",
            "  [0.0627451  0.02745098 0.        ]\n",
            "  [0.19215686 0.10588235 0.03137255]\n",
            "  ...\n",
            "  [0.4627451  0.32941176 0.19607843]\n",
            "  [0.47058824 0.32941176 0.19607843]\n",
            "  [0.42745098 0.28627451 0.16470588]]\n",
            "\n",
            " ...\n",
            "\n",
            " [[0.81568627 0.66666667 0.37647059]\n",
            "  [0.78823529 0.6        0.13333333]\n",
            "  [0.77647059 0.63137255 0.10196078]\n",
            "  ...\n",
            "  [0.62745098 0.52156863 0.2745098 ]\n",
            "  [0.21960784 0.12156863 0.02745098]\n",
            "  [0.20784314 0.13333333 0.07843137]]\n",
            "\n",
            " [[0.70588235 0.54509804 0.37647059]\n",
            "  [0.67843137 0.48235294 0.16470588]\n",
            "  [0.72941176 0.56470588 0.11764706]\n",
            "  ...\n",
            "  [0.72156863 0.58039216 0.36862745]\n",
            "  [0.38039216 0.24313725 0.13333333]\n",
            "  [0.3254902  0.20784314 0.13333333]]\n",
            "\n",
            " [[0.69411765 0.56470588 0.45490196]\n",
            "  [0.65882353 0.50588235 0.36862745]\n",
            "  [0.70196078 0.55686275 0.34117647]\n",
            "  ...\n",
            "  [0.84705882 0.72156863 0.54901961]\n",
            "  [0.59215686 0.4627451  0.32941176]\n",
            "  [0.48235294 0.36078431 0.28235294]]]\n",
            "X_test shape: [[[158 112  49]\n",
            "  [159 111  47]\n",
            "  [165 116  51]\n",
            "  ...\n",
            "  [137  95  36]\n",
            "  [126  91  36]\n",
            "  [116  85  33]]\n",
            "\n",
            " [[152 112  51]\n",
            "  [151 110  40]\n",
            "  [159 114  45]\n",
            "  ...\n",
            "  [136  95  31]\n",
            "  [125  91  32]\n",
            "  [119  88  34]]\n",
            "\n",
            " [[151 110  47]\n",
            "  [151 109  33]\n",
            "  [158 111  36]\n",
            "  ...\n",
            "  [139  98  34]\n",
            "  [130  95  34]\n",
            "  [120  89  33]]\n",
            "\n",
            " ...\n",
            "\n",
            " [[ 68 124 177]\n",
            "  [ 42 100 148]\n",
            "  [ 31  88 137]\n",
            "  ...\n",
            "  [ 38  97 146]\n",
            "  [ 13  64 108]\n",
            "  [ 40  85 127]]\n",
            "\n",
            " [[ 61 116 168]\n",
            "  [ 49 102 148]\n",
            "  [ 35  85 132]\n",
            "  ...\n",
            "  [ 26  82 130]\n",
            "  [ 29  82 126]\n",
            "  [ 20  64 107]]\n",
            "\n",
            " [[ 54 107 160]\n",
            "  [ 56 105 149]\n",
            "  [ 45  89 132]\n",
            "  ...\n",
            "  [ 24  77 124]\n",
            "  [ 34  84 129]\n",
            "  [ 21  67 110]]]\n",
            "X_test normalized shape: [[[0.61960784 0.43921569 0.19215686]\n",
            "  [0.62352941 0.43529412 0.18431373]\n",
            "  [0.64705882 0.45490196 0.2       ]\n",
            "  ...\n",
            "  [0.5372549  0.37254902 0.14117647]\n",
            "  [0.49411765 0.35686275 0.14117647]\n",
            "  [0.45490196 0.33333333 0.12941176]]\n",
            "\n",
            " [[0.59607843 0.43921569 0.2       ]\n",
            "  [0.59215686 0.43137255 0.15686275]\n",
            "  [0.62352941 0.44705882 0.17647059]\n",
            "  ...\n",
            "  [0.53333333 0.37254902 0.12156863]\n",
            "  [0.49019608 0.35686275 0.1254902 ]\n",
            "  [0.46666667 0.34509804 0.13333333]]\n",
            "\n",
            " [[0.59215686 0.43137255 0.18431373]\n",
            "  [0.59215686 0.42745098 0.12941176]\n",
            "  [0.61960784 0.43529412 0.14117647]\n",
            "  ...\n",
            "  [0.54509804 0.38431373 0.13333333]\n",
            "  [0.50980392 0.37254902 0.13333333]\n",
            "  [0.47058824 0.34901961 0.12941176]]\n",
            "\n",
            " ...\n",
            "\n",
            " [[0.26666667 0.48627451 0.69411765]\n",
            "  [0.16470588 0.39215686 0.58039216]\n",
            "  [0.12156863 0.34509804 0.5372549 ]\n",
            "  ...\n",
            "  [0.14901961 0.38039216 0.57254902]\n",
            "  [0.05098039 0.25098039 0.42352941]\n",
            "  [0.15686275 0.33333333 0.49803922]]\n",
            "\n",
            " [[0.23921569 0.45490196 0.65882353]\n",
            "  [0.19215686 0.4        0.58039216]\n",
            "  [0.1372549  0.33333333 0.51764706]\n",
            "  ...\n",
            "  [0.10196078 0.32156863 0.50980392]\n",
            "  [0.11372549 0.32156863 0.49411765]\n",
            "  [0.07843137 0.25098039 0.41960784]]\n",
            "\n",
            " [[0.21176471 0.41960784 0.62745098]\n",
            "  [0.21960784 0.41176471 0.58431373]\n",
            "  [0.17647059 0.34901961 0.51764706]\n",
            "  ...\n",
            "  [0.09411765 0.30196078 0.48627451]\n",
            "  [0.13333333 0.32941176 0.50588235]\n",
            "  [0.08235294 0.2627451  0.43137255]]]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "o8-LsGIX0A87"
      },
      "source": [
        "### Basic Artifical Neural Network"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "KwxTtlZtmLB3",
        "outputId": "73a8922d-6779-4f2d-db90-352988e5fb9c"
      },
      "source": [
        "from tensorflow import keras \n",
        "from tensorflow.keras import datasets, layers, models \n",
        "### artificial neural network ### \n",
        "ann_model = keras.Sequential([\n",
        "                           keras.layers.Flatten(input_shape=(32,32,3)), # input layer, flatten, set shape \n",
        "                           keras.layers.Dense(3000, activation='relu'), # deep layers, neurons, aqctivation \n",
        "                           keras.layers.Dense(1000, activation='relu'), # deep layers, neurons, aqctivation \n",
        "                           keras.layers.Dense(10, activation='sigmoid') # output layer, 10 neurons for 10 categories in the data, activation \n",
        "                          ])\n",
        "ann_model.compile(\n",
        "              optimizer='SGD', # Stochastic Gradient Descent\n",
        "              loss='sparse_categorical_crossentropy', # directly using the value, was not encoded\n",
        "              metrics=['accuracy'] # measure with accuracy \n",
        "              )\n",
        "### show summary ### \n",
        "ann_model.summary()"
      ],
      "execution_count": 130,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_6\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "flatten_6 (Flatten)          (None, 3072)              0         \n",
            "_________________________________________________________________\n",
            "dense_16 (Dense)             (None, 3000)              9219000   \n",
            "_________________________________________________________________\n",
            "dense_17 (Dense)             (None, 1000)              3001000   \n",
            "_________________________________________________________________\n",
            "dense_18 (Dense)             (None, 10)                10010     \n",
            "=================================================================\n",
            "Total params: 12,230,010\n",
            "Trainable params: 12,230,010\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "SY2U4FMQ8iIQ",
        "outputId": "254b3f3c-6ed8-49ec-ed63-af77314693c0"
      },
      "source": [
        "from sklearn import metrics\n",
        "### fit the model, get metrics ###  \n",
        "print('--- model runtime ---')\n",
        "%time history = ann_model.fit(X_train, y_train, epochs=5) # fit the model on the data \n",
        "### make the prediction on the data ###\n",
        "y_pred = ann_model.predict(X_test)\n",
        "y_classes = [np.argmax(element) for element in y_pred] # find the largest value in each element in the prediction\n",
        "print('--- model metrics ---')\n",
        "scores = ann_model.evaluate(X_test, y_test) # get the model score from evaluation\n",
        "print(f'model evaluation score: {scores[1]*100}') # show the models accuracy score \n",
        "print('model accuracy score:', metrics.accuracy_score(y_test, y_classes)) # model accuracy\n",
        "print('--- classification report ---') \n",
        "print(metrics.classification_report(y_test,y_classes)) # classificatin report"
      ],
      "execution_count": 131,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- model runtime ---\n",
            "Epoch 1/5\n",
            "1563/1563 [==============================] - 90s 57ms/step - loss: 1.9301 - accuracy: 0.3048\n",
            "Epoch 2/5\n",
            "1563/1563 [==============================] - 91s 58ms/step - loss: 1.6472 - accuracy: 0.4204\n",
            "Epoch 3/5\n",
            "1563/1563 [==============================] - 90s 58ms/step - loss: 1.5485 - accuracy: 0.4518\n",
            "Epoch 4/5\n",
            "1563/1563 [==============================] - 93s 59ms/step - loss: 1.4901 - accuracy: 0.4772\n",
            "Epoch 5/5\n",
            "1563/1563 [==============================] - 92s 59ms/step - loss: 1.4450 - accuracy: 0.4934\n",
            "CPU times: user 12min 56s, sys: 17 s, total: 13min 13s\n",
            "Wall time: 7min 35s\n",
            "--- model metrics ---\n",
            "313/313 [==============================] - 6s 19ms/step - loss: 1.4834 - accuracy: 0.4710\n",
            "model evaluation score: 47.09999859333038\n",
            "model accuracy score: 0.471\n",
            "--- classification report ---\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.56      0.56      0.56      1000\n",
            "           1       0.52      0.74      0.61      1000\n",
            "           2       0.38      0.32      0.35      1000\n",
            "           3       0.33      0.31      0.32      1000\n",
            "           4       0.65      0.15      0.25      1000\n",
            "           5       0.28      0.58      0.38      1000\n",
            "           6       0.52      0.52      0.52      1000\n",
            "           7       0.53      0.56      0.54      1000\n",
            "           8       0.67      0.54      0.60      1000\n",
            "           9       0.65      0.42      0.51      1000\n",
            "\n",
            "    accuracy                           0.47     10000\n",
            "   macro avg       0.51      0.47      0.46     10000\n",
            "weighted avg       0.51      0.47      0.46     10000\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 445
        },
        "id": "f-vY5zmSqnk2",
        "outputId": "0c16a33e-6cbd-4cec-c337-a3778c258d0b"
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "### plot the model loss ###\n",
        "### configure the plot ###\n",
        "print('--- model loss-MAE --- ')\n",
        "f, ax = plt.subplots(1, 1, figsize = (10, 7))\n",
        "ax1 = plt.plot(history.history['loss'], color=\"r\", label=\"loss-MSE\")\n",
        "plt.ylabel('Loss')\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": 132,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- model loss-MAE --- \n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x504 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7FA_Fnlm3vKp"
      },
      "source": [
        "### Convolutional Neural Network"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8DM2trxA2JNB",
        "outputId": "d6d4d134-8583-4d0f-f511-53ffa9fb4b23"
      },
      "source": [
        "from tensorflow import keras \n",
        "### convolutional neural network ### \n",
        "cnn_model = keras.Sequential([\n",
        "                          keras.layers.Conv2D(filters= 32, kernel_size=(3,3), activation = 'relu', input_shape=(32,32,3)), # conv layer \n",
        "                          keras.layers.MaxPooling2D((2,2)), # pool layer \n",
        "                          keras.layers.Conv2D(filters= 64, kernel_size=(3,3), activation = 'relu'), \n",
        "                          keras.layers.MaxPooling2D((2,2)), \n",
        "\n",
        "                           # dense  \n",
        "                           keras.layers.Flatten(), # input 'dense' layer, flatten\n",
        "                           keras.layers.Dense(64, activation='relu'), # deep layers, neurons, aqctivation\n",
        "                           keras.layers.Dense(10, activation='softmax') # output layer, 10 neurons for 10 categories in the data, activation, softmax normalize probability to 1\n",
        "                              ])\n",
        "cnn_model.compile(\n",
        "              optimizer='adam', # popular, gives good accuracy\n",
        "              loss='sparse_categorical_crossentropy', # directly using the value, was not encoded\n",
        "              metrics=['accuracy'] # measure with accuracy \n",
        "              )\n",
        "### show summary ### \n",
        "cnn_model.summary()"
      ],
      "execution_count": 133,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_7\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d_5 (Conv2D)            (None, 30, 30, 32)        896       \n",
            "_________________________________________________________________\n",
            "max_pooling2d_5 (MaxPooling2 (None, 15, 15, 32)        0         \n",
            "_________________________________________________________________\n",
            "conv2d_6 (Conv2D)            (None, 13, 13, 64)        18496     \n",
            "_________________________________________________________________\n",
            "max_pooling2d_6 (MaxPooling2 (None, 6, 6, 64)          0         \n",
            "_________________________________________________________________\n",
            "flatten_7 (Flatten)          (None, 2304)              0         \n",
            "_________________________________________________________________\n",
            "dense_19 (Dense)             (None, 64)                147520    \n",
            "_________________________________________________________________\n",
            "dense_20 (Dense)             (None, 10)                650       \n",
            "=================================================================\n",
            "Total params: 167,562\n",
            "Trainable params: 167,562\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "oJ2ViPfT5pjf",
        "outputId": "e3754876-43aa-4696-d6e6-5e04f7e86ac5"
      },
      "source": [
        "from sklearn import metrics \n",
        "### fit the model, get metrics ###  \n",
        "print('--- model runtime ---')\n",
        "%time history = cnn_model.fit(X_train, y_train, epochs=5) # fit the model on the data \n",
        "### make the prediction on the data ###\n",
        "y_pred = cnn_model.predict(X_test)\n",
        "y_classes = [np.argmax(element) for element in y_pred] # find the largest value in each element in the prediction\n",
        "print('--- model metrics ---')\n",
        "scores = cnn_model.evaluate(X_test, y_test) # get the model score from evaluation\n",
        "print(f'model evaluation score: {scores[1]*100}') # show the models accuracy score \n",
        "print('model accuracy score:', metrics.accuracy_score(y_test, y_classes)) # model accuracy\n",
        "print('--- classification report ---') \n",
        "print(metrics.classification_report(y_test,y_classes)) # classificatin report"
      ],
      "execution_count": 134,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- model runtime ---\n",
            "Epoch 1/5\n",
            "1563/1563 [==============================] - 54s 34ms/step - loss: 1.7223 - accuracy: 0.3716\n",
            "Epoch 2/5\n",
            "1563/1563 [==============================] - 52s 33ms/step - loss: 1.1757 - accuracy: 0.5830\n",
            "Epoch 3/5\n",
            "1563/1563 [==============================] - 53s 34ms/step - loss: 1.0245 - accuracy: 0.6393\n",
            "Epoch 4/5\n",
            "1563/1563 [==============================] - 52s 33ms/step - loss: 0.9459 - accuracy: 0.6671\n",
            "Epoch 5/5\n",
            "1563/1563 [==============================] - 52s 33ms/step - loss: 0.8816 - accuracy: 0.6902\n",
            "CPU times: user 7min 22s, sys: 20.3 s, total: 7min 42s\n",
            "Wall time: 4min 23s\n",
            "--- model metrics ---\n",
            "313/313 [==============================] - 3s 10ms/step - loss: 0.9833 - accuracy: 0.6550\n",
            "model evaluation score: 65.49999713897705\n",
            "model accuracy score: 0.655\n",
            "--- classification report ---\n",
            "              precision    recall  f1-score   support\n",
            "\n",
            "           0       0.75      0.68      0.71      1000\n",
            "           1       0.87      0.69      0.77      1000\n",
            "           2       0.72      0.43      0.53      1000\n",
            "           3       0.38      0.68      0.49      1000\n",
            "           4       0.67      0.50      0.57      1000\n",
            "           5       0.63      0.43      0.51      1000\n",
            "           6       0.74      0.75      0.74      1000\n",
            "           7       0.67      0.76      0.71      1000\n",
            "           8       0.66      0.86      0.75      1000\n",
            "           9       0.73      0.78      0.75      1000\n",
            "\n",
            "    accuracy                           0.66     10000\n",
            "   macro avg       0.68      0.66      0.66     10000\n",
            "weighted avg       0.68      0.66      0.66     10000\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 445
        },
        "id": "eQlLSRvB8QM1",
        "outputId": "5ca32b8a-44d6-4d96-d55b-ac9ad72754cc"
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "### plot the model loss ###\n",
        "### configure the plot ###\n",
        "print('--- model loss-MAE --- ')\n",
        "f, ax = plt.subplots(1, 1, figsize = (10, 7))\n",
        "ax1 = plt.plot(history.history['loss'], color=\"r\", label=\"loss-MSE\")\n",
        "plt.ylabel('Loss')\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": 135,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- model loss-MAE --- \n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x504 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qnRKNW5oIcLW"
      },
      "source": [
        "### Show the predictions "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "y8scKhpA-OhW",
        "outputId": "b44dd619-0ea4-42cf-9d92-74d8a0a836c6"
      },
      "source": [
        "### reshape the y_test data ### \n",
        "print('--- y_test ---')\n",
        "print(y_test[:5])\n",
        "print('--- y_test reshaped ---')\n",
        "y_test = y_test.reshape(-1,)\n",
        "y_test[:5]"
      ],
      "execution_count": 136,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "--- y_test ---\n",
            "[[3]\n",
            " [8]\n",
            " [8]\n",
            " [0]\n",
            " [6]]\n",
            "--- y_test reshaped ---\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([3, 8, 8, 0, 6], dtype=uint8)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 136
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BioFbXTX-fs_"
      },
      "source": [
        "y_pred = cnn_model.predict(X_test)"
      ],
      "execution_count": 122,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 207
        },
        "id": "o684bJ19BYU_",
        "outputId": "f4efbc64-cb3b-4b3a-d664-5fc8090e8786"
      },
      "source": [
        "### make the prediction on the data ###\n",
        "y_pred = cnn_model.predict(X_test)\n",
        "### find the largest value in each element in the prediction ### \n",
        "y_classes = [np.argmax(element) for element in y_pred]\n",
        "### show a sample vs the prediction ### \n",
        "print('prediction:', classes[y_classes[5]])\n",
        "print('--- show sample image ---')\n",
        "plot_sample(X_test, y_test, 5)"
      ],
      "execution_count": 138,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "prediction: frog\n",
            "--- show sample image ---\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 1080x144 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}