Digit_recognizer_part2.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Digit recognizer_part2.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true,
      "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/mentorbruh/digit-recognizer/blob/main/Digit_recognizer_part2.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "NUTWa3UciPFN"
      },
      "source": [
        "# Import Libraries:"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vIsoMjaqlQkA"
      },
      "source": [
        "In this second part of digit recognizer, we explore different activation functions and keras tuner functionalities and also plot our train and validation data accuracies."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RkiIB4CtiTtn"
      },
      "source": [
        "import tensorflow as tf\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Dense, Conv2D, Flatten\n",
        "from keras.utils import to_categorical\n",
        "from keras.datasets import mnist\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "%matplotlib inline"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "cOYFkpt-idQS"
      },
      "source": [
        "## Dataset:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Cuvd5X1pia-8",
        "outputId": "4f61963a-840d-452f-b567-4a923b6b65f9"
      },
      "source": [
        "data = mnist.load_data()\n",
        "(X_train, y_train), (X_test, y_test) = data"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/mnist.npz\n",
            "11493376/11490434 [==============================] - 0s 0us/step\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "BalHCZajijpR",
        "outputId": "170373ba-fe14-415a-901f-1f39f5aac683"
      },
      "source": [
        "print(X_train.shape)\n",
        "print(y_train.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(60000, 28, 28)\n",
            "(60000,)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "p95qV4noiufa"
      },
      "source": [
        "# Lets plot one number:"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Y5cXmtfvi0lX",
        "outputId": "706a4caa-df1e-4993-95c3-723e79c4d539"
      },
      "source": [
        "X_train[0]"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   3,\n",
              "         18,  18,  18, 126, 136, 175,  26, 166, 255, 247, 127,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,   0,  30,  36,  94, 154, 170,\n",
              "        253, 253, 253, 253, 253, 225, 172, 253, 242, 195,  64,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,  49, 238, 253, 253, 253, 253,\n",
              "        253, 253, 253, 253, 251,  93,  82,  82,  56,  39,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,  18, 219, 253, 253, 253, 253,\n",
              "        253, 198, 182, 247, 241,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,   0,  80, 156, 107, 253, 253,\n",
              "        205,  11,   0,  43, 154,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,  14,   1, 154, 253,\n",
              "         90,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0, 139, 253,\n",
              "        190,   2,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,  11, 190,\n",
              "        253,  70,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,  35,\n",
              "        241, 225, 160, 108,   1,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "         81, 240, 253, 253, 119,  25,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,  45, 186, 253, 253, 150,  27,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0,  16,  93, 252, 253, 187,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0,   0,   0, 249, 253, 249,  64,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,  46, 130, 183, 253, 253, 207,   2,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,  39,\n",
              "        148, 229, 253, 253, 253, 250, 182,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,  24, 114, 221,\n",
              "        253, 253, 253, 253, 201,  78,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,   0,  23,  66, 213, 253, 253,\n",
              "        253, 253, 198,  81,   2,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,  18, 171, 219, 253, 253, 253, 253,\n",
              "        195,  80,   9,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,  55, 172, 226, 253, 253, 253, 253, 244, 133,\n",
              "         11,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0, 136, 253, 253, 253, 212, 135, 132,  16,   0,\n",
              "          0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0],\n",
              "       [  0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,   0,\n",
              "          0,   0]], dtype=uint8)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 5
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 282
        },
        "id": "4PzeEotRi2wL",
        "outputId": "770cf1f6-ade2-4d51-8437-ac820cfc49c9"
      },
      "source": [
        "# lets see this image:\n",
        "plt.imshow(X_train[0])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7f55402a5630>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 6
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "DahYAvcji5Jq",
        "outputId": "a3721bbd-58d8-4ddd-9418-615a35115417"
      },
      "source": [
        "# Actual label of above image:\n",
        "y_train[0]"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "5"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "qV-gA2vmi89N"
      },
      "source": [
        "# Reshape each image to 28 x 28 and convert these images to gray scale:\n",
        "\n",
        "X_train = X_train.reshape(60000, 28, 28, 1)\n",
        "X_test = X_test.reshape(10000, 28, 28, 1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "08n-kmpTjANk",
        "outputId": "bc0ad354-7438-43ae-a1a4-089458a62500"
      },
      "source": [
        "# make 10 categories for target variable y.\n",
        "\n",
        "y_train[0:10]"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([5, 0, 4, 1, 9, 2, 1, 3, 1, 4], dtype=uint8)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 9
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fl7uI941jEk8"
      },
      "source": [
        "y_train = to_categorical(y_train)\n",
        "y_test = to_categorical(y_test)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "zZxhgM4ejG1R",
        "outputId": "ea913915-2082-4dc8-a373-3de3e21c3339"
      },
      "source": [
        "y_train[0:10]"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[0., 0., 0., 0., 0., 1., 0., 0., 0., 0.],\n",
              "       [1., 0., 0., 0., 0., 0., 0., 0., 0., 0.],\n",
              "       [0., 0., 0., 0., 1., 0., 0., 0., 0., 0.],\n",
              "       [0., 1., 0., 0., 0., 0., 0., 0., 0., 0.],\n",
              "       [0., 0., 0., 0., 0., 0., 0., 0., 0., 1.],\n",
              "       [0., 0., 1., 0., 0., 0., 0., 0., 0., 0.],\n",
              "       [0., 1., 0., 0., 0., 0., 0., 0., 0., 0.],\n",
              "       [0., 0., 0., 1., 0., 0., 0., 0., 0., 0.],\n",
              "       [0., 1., 0., 0., 0., 0., 0., 0., 0., 0.],\n",
              "       [0., 0., 0., 0., 1., 0., 0., 0., 0., 0.]], dtype=float32)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 11
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Bdnwm9ErjLIn"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "OPGAp_mQjPnB"
      },
      "source": [
        "# CNN Model:\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wa85Bx2xTFjc"
      },
      "source": [
        "## Model 1:\n",
        "Activation function: Relu with 32 nodes."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "D3XeqzPRjRev",
        "outputId": "405c0b4c-4013-4cbb-b364-c1ba16f6a2aa"
      },
      "source": [
        "model_1 = Sequential()\n",
        "\n",
        "# Convolution layer:\n",
        "model_1.add(Conv2D(32, kernel_size = 3, activation = 'relu', input_shape = [28, 28, 1]))\n",
        "model_1.add(Conv2D(16, kernel_size = 3, activation = 'relu'))\n",
        "\n",
        "#Flatten:\n",
        "model_1.add(Flatten())\n",
        "\n",
        "# Output layer:\n",
        "model_1.add(Dense(units = 10, activation = 'softmax'))\n",
        "\n",
        "model_1.compile(optimizer = 'adam', loss = 'categorical_crossentropy', metrics = ['accuracy'])\n",
        "\n",
        "model_1.fit(X_train, y_train, validation_data= (X_test, y_test), epochs=10, verbose=1)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/10\n",
            "1875/1875 [==============================] - 5s 3ms/step - loss: 0.2603 - accuracy: 0.9470 - val_loss: 0.0898 - val_accuracy: 0.9725\n",
            "Epoch 2/10\n",
            "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0748 - accuracy: 0.9773 - val_loss: 0.0942 - val_accuracy: 0.9720\n",
            "Epoch 3/10\n",
            "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0513 - accuracy: 0.9838 - val_loss: 0.0975 - val_accuracy: 0.9711\n",
            "Epoch 4/10\n",
            "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0399 - accuracy: 0.9872 - val_loss: 0.1047 - val_accuracy: 0.9721\n",
            "Epoch 5/10\n",
            "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0343 - accuracy: 0.9893 - val_loss: 0.1148 - val_accuracy: 0.9749\n",
            "Epoch 6/10\n",
            "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0266 - accuracy: 0.9920 - val_loss: 0.1087 - val_accuracy: 0.9768\n",
            "Epoch 7/10\n",
            "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0232 - accuracy: 0.9929 - val_loss: 0.1097 - val_accuracy: 0.9797\n",
            "Epoch 8/10\n",
            "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0201 - accuracy: 0.9942 - val_loss: 0.1382 - val_accuracy: 0.9785\n",
            "Epoch 9/10\n",
            "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0173 - accuracy: 0.9952 - val_loss: 0.1908 - val_accuracy: 0.9746\n",
            "Epoch 10/10\n",
            "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0189 - accuracy: 0.9951 - val_loss: 0.1616 - val_accuracy: 0.9802\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tensorflow.python.keras.callbacks.History at 0x7f553029a748>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 12
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "9sHe1hKtnM9W"
      },
      "source": [
        "The predicton accuracy of this model is 99.54%"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nt5XaBpVn1FV"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Z-InQpKVnp3x"
      },
      "source": [
        "Now we increase the number of nodes to 64 and then check the accuracy of this model."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "7Od0kfGZn4LI",
        "outputId": "dc497250-fefc-418e-c55a-485043777005"
      },
      "source": [
        "model_1 = Sequential()\n",
        "\n",
        "# Convolution layer:\n",
        "model_1.add(Conv2D(64, kernel_size = 3, activation = 'relu', input_shape = [28, 28, 1]))\n",
        "model_1.add(Conv2D(32, kernel_size = 3, activation = 'relu'))\n",
        "\n",
        "#Flatten:\n",
        "model_1.add(Flatten())\n",
        "\n",
        "# Output layer:\n",
        "model_1.add(Dense(units = 10, activation = 'softmax'))\n",
        "\n",
        "model_1.compile(optimizer = 'adam', loss = 'categorical_crossentropy', metrics = ['accuracy'])\n",
        "\n",
        "model_1.fit(X_train, y_train, validation_data= (X_test, y_test), epochs=10, verbose=1)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.2165 - accuracy: 0.9501 - val_loss: 0.1072 - val_accuracy: 0.9687\n",
            "Epoch 2/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0680 - accuracy: 0.9795 - val_loss: 0.0923 - val_accuracy: 0.9746\n",
            "Epoch 3/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0487 - accuracy: 0.9849 - val_loss: 0.0959 - val_accuracy: 0.9745\n",
            "Epoch 4/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0367 - accuracy: 0.9883 - val_loss: 0.1151 - val_accuracy: 0.9770\n",
            "Epoch 5/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0296 - accuracy: 0.9905 - val_loss: 0.0991 - val_accuracy: 0.9789\n",
            "Epoch 6/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0259 - accuracy: 0.9924 - val_loss: 0.1144 - val_accuracy: 0.9771\n",
            "Epoch 7/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0227 - accuracy: 0.9937 - val_loss: 0.1273 - val_accuracy: 0.9744\n",
            "Epoch 8/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0184 - accuracy: 0.9948 - val_loss: 0.1418 - val_accuracy: 0.9794\n",
            "Epoch 9/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0208 - accuracy: 0.9947 - val_loss: 0.1505 - val_accuracy: 0.9778\n",
            "Epoch 10/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0137 - accuracy: 0.9961 - val_loss: 0.2032 - val_accuracy: 0.9747\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tensorflow.python.keras.callbacks.History at 0x7f55300cd588>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 13
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1fdi_BH9ywKx"
      },
      "source": [
        "The prediction acuuracy of this model is 99.54% which is same as previous model."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "a0xVxzD2j0lp"
      },
      "source": [
        "## Model 2:\n",
        "Activation function: Leaky Relu"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "IFi7STr-jnby"
      },
      "source": [
        "lrelu = lambda x: tf.keras.activations.relu(x, alpha=0.1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fS-rgN_ikLql"
      },
      "source": [
        "[Referance](https://stackoverflow.com/questions/48828478/how-do-you-use-keras-leakyrelu-in-python/48828561)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "evexe24bkKdl",
        "outputId": "210f2669-53b6-488f-fb42-32e3bb24a26f"
      },
      "source": [
        "model_2 = Sequential()\n",
        "\n",
        "# Convolution layer:\n",
        "model_2.add(Conv2D(32, kernel_size = 3, activation= lrelu , input_shape = [28, 28, 1]))\n",
        "model_2.add(Conv2D(16, kernel_size = 3, activation= lrelu))\n",
        "\n",
        "#Flatten:\n",
        "model_2.add(Flatten())\n",
        "\n",
        "# Output layer:\n",
        "model_2.add(Dense(units = 10, activation = 'softmax'))\n",
        "\n",
        "model_2.compile(optimizer = 'adam', loss = 'categorical_crossentropy', metrics = ['accuracy'])\n",
        "\n",
        "model_2.fit(X_train, y_train, validation_data= (X_test, y_test), epochs=10, verbose=1)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/10\n",
            "1875/1875 [==============================] - 5s 3ms/step - loss: 0.4088 - accuracy: 0.9348 - val_loss: 0.1061 - val_accuracy: 0.9679\n",
            "Epoch 2/10\n",
            "1875/1875 [==============================] - 5s 3ms/step - loss: 0.1151 - accuracy: 0.9696 - val_loss: 0.1128 - val_accuracy: 0.9721\n",
            "Epoch 3/10\n",
            "1875/1875 [==============================] - 5s 3ms/step - loss: 0.1020 - accuracy: 0.9743 - val_loss: 0.1482 - val_accuracy: 0.9645\n",
            "Epoch 4/10\n",
            "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0966 - accuracy: 0.9768 - val_loss: 0.1308 - val_accuracy: 0.9720\n",
            "Epoch 5/10\n",
            "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0864 - accuracy: 0.9805 - val_loss: 0.1548 - val_accuracy: 0.9740\n",
            "Epoch 6/10\n",
            "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0941 - accuracy: 0.9812 - val_loss: 0.1889 - val_accuracy: 0.9725\n",
            "Epoch 7/10\n",
            "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0753 - accuracy: 0.9850 - val_loss: 0.2518 - val_accuracy: 0.9691\n",
            "Epoch 8/10\n",
            "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0878 - accuracy: 0.9850 - val_loss: 0.2116 - val_accuracy: 0.9751\n",
            "Epoch 9/10\n",
            "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0860 - accuracy: 0.9871 - val_loss: 0.2970 - val_accuracy: 0.9739\n",
            "Epoch 10/10\n",
            "1875/1875 [==============================] - 5s 3ms/step - loss: 0.0886 - accuracy: 0.9877 - val_loss: 0.2820 - val_accuracy: 0.9759\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tensorflow.python.keras.callbacks.History at 0x7f54cc769630>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 15
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "KeArnV3ozU-r"
      },
      "source": [
        "The prediction accuracy of this model is 98.72% which is less than the 1st model."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qHyVQgaSzf7Y"
      },
      "source": [
        "Now we increase the nodes and check whether we get better accuracy or not."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "oeqf8xaVzJ8Z",
        "outputId": "d52e5267-d19d-4d92-aa32-a35b3fca5bb5"
      },
      "source": [
        "model_2 = Sequential()\n",
        "\n",
        "# Convolution layer:\n",
        "model_2.add(Conv2D(64, kernel_size = 3, activation= lrelu , input_shape = [28, 28, 1]))\n",
        "model_2.add(Conv2D(32, kernel_size = 3, activation= lrelu))\n",
        "\n",
        "#Flatten:\n",
        "model_2.add(Flatten())\n",
        "\n",
        "# Output layer:\n",
        "model_2.add(Dense(units = 10, activation = 'softmax'))\n",
        "\n",
        "model_2.compile(optimizer = 'adam', loss = 'categorical_crossentropy', metrics = ['accuracy'])\n",
        "\n",
        "model_2.fit(X_train, y_train, validation_data= (X_test, y_test), epochs=10, verbose=1)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.3069 - accuracy: 0.9433 - val_loss: 0.1229 - val_accuracy: 0.9725\n",
            "Epoch 2/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.1516 - accuracy: 0.9667 - val_loss: 0.1723 - val_accuracy: 0.9612\n",
            "Epoch 3/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.1913 - accuracy: 0.9684 - val_loss: 0.2422 - val_accuracy: 0.9577\n",
            "Epoch 4/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.1677 - accuracy: 0.9750 - val_loss: 0.2291 - val_accuracy: 0.9726\n",
            "Epoch 5/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.1776 - accuracy: 0.9785 - val_loss: 0.4960 - val_accuracy: 0.9626\n",
            "Epoch 6/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.2259 - accuracy: 0.9799 - val_loss: 0.3584 - val_accuracy: 0.9791\n",
            "Epoch 7/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.2168 - accuracy: 0.9829 - val_loss: 0.3735 - val_accuracy: 0.9768\n",
            "Epoch 8/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.1770 - accuracy: 0.9861 - val_loss: 0.7790 - val_accuracy: 0.9711\n",
            "Epoch 9/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.2499 - accuracy: 0.9863 - val_loss: 0.6801 - val_accuracy: 0.9796\n",
            "Epoch 10/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.2661 - accuracy: 0.9882 - val_loss: 0.6531 - val_accuracy: 0.9792\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tensorflow.python.keras.callbacks.History at 0x7f54cc5dd588>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 16
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZRsepPmjzzWw"
      },
      "source": [
        "The prediction accuracy of this model increase slightly after increment in nodes."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "rFLCv0pOkbjA"
      },
      "source": [
        "## Model 3:\n",
        "Activation function: Elu\n",
        "\n",
        "[Referance](https://stackoverflow.com/questions/56184666/how-to-set-alpha-parameter-in-activationelu-layer)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CquLdUtrkkjG"
      },
      "source": [
        "from keras.activations import elu\n",
        "\n",
        "ELU = lambda x: elu(x, alpha=0.4)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "L4631M0hkrCZ",
        "outputId": "10a55296-959e-4954-807f-669fa806d443"
      },
      "source": [
        "model_3 = Sequential()\n",
        "\n",
        "# Convolution layer:\n",
        "model_3.add(Conv2D(32, kernel_size = 3, activation= ELU , input_shape = [28, 28, 1]))\n",
        "model_3.add(Conv2D(16, kernel_size = 3, activation= ELU))\n",
        "\n",
        "#Flatten:\n",
        "model_3.add(Flatten())\n",
        "\n",
        "# Output layer:\n",
        "model_3.add(Dense(units = 10, activation = 'softmax'))\n",
        "\n",
        "model_3.compile(optimizer = 'adam', loss = 'categorical_crossentropy', metrics = ['accuracy'])\n",
        "\n",
        "model_3.fit(X_train, y_train, validation_data= (X_test, y_test), epochs=10, verbose=1)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.2398 - accuracy: 0.9441 - val_loss: 0.1167 - val_accuracy: 0.9705\n",
            "Epoch 2/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0941 - accuracy: 0.9743 - val_loss: 0.1031 - val_accuracy: 0.9752\n",
            "Epoch 3/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0774 - accuracy: 0.9805 - val_loss: 0.1275 - val_accuracy: 0.9763\n",
            "Epoch 4/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0659 - accuracy: 0.9847 - val_loss: 0.1495 - val_accuracy: 0.9747\n",
            "Epoch 5/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0629 - accuracy: 0.9872 - val_loss: 0.1688 - val_accuracy: 0.9781\n",
            "Epoch 6/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0509 - accuracy: 0.9905 - val_loss: 0.2361 - val_accuracy: 0.9743\n",
            "Epoch 7/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0550 - accuracy: 0.9908 - val_loss: 0.2190 - val_accuracy: 0.9775\n",
            "Epoch 8/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0483 - accuracy: 0.9924 - val_loss: 0.2677 - val_accuracy: 0.9767\n",
            "Epoch 9/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0408 - accuracy: 0.9938 - val_loss: 0.2391 - val_accuracy: 0.9835\n",
            "Epoch 10/10\n",
            "1875/1875 [==============================] - 6s 3ms/step - loss: 0.0406 - accuracy: 0.9943 - val_loss: 0.2816 - val_accuracy: 0.9800\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tensorflow.python.keras.callbacks.History at 0x7f54cc4b0710>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 18
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AKUskld-0DEv"
      },
      "source": [
        "The prediction accuracy of this model is 99.30%"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "nA_BBGnNkuHA",
        "outputId": "c47af5bd-9df6-4d73-9ec8-d3d1de268a71"
      },
      "source": [
        "model_3 = Sequential()\n",
        "\n",
        "# Convolution layer:\n",
        "model_3.add(Conv2D(64, kernel_size = 3, activation= ELU , input_shape = [28, 28, 1]))\n",
        "model_3.add(Conv2D(32, kernel_size = 3, activation= ELU))\n",
        "\n",
        "#Flatten:\n",
        "model_3.add(Flatten())\n",
        "\n",
        "# Output layer:\n",
        "model_3.add(Dense(units = 10, activation = 'softmax'))\n",
        "\n",
        "model_3.compile(optimizer = 'adam', loss = 'categorical_crossentropy', metrics = ['accuracy'])\n",
        "\n",
        "model_3.fit(X_train, y_train, validation_data= (X_test, y_test), epochs=10, verbose=1)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/10\n",
            "1875/1875 [==============================] - 8s 4ms/step - loss: 0.2779 - accuracy: 0.9418 - val_loss: 0.1378 - val_accuracy: 0.9671\n",
            "Epoch 2/10\n",
            "1875/1875 [==============================] - 7s 4ms/step - loss: 0.1394 - accuracy: 0.9710 - val_loss: 0.2218 - val_accuracy: 0.9611\n",
            "Epoch 3/10\n",
            "1875/1875 [==============================] - 7s 4ms/step - loss: 0.1264 - accuracy: 0.9788 - val_loss: 0.2339 - val_accuracy: 0.9654\n",
            "Epoch 4/10\n",
            "1875/1875 [==============================] - 7s 4ms/step - loss: 0.1249 - accuracy: 0.9832 - val_loss: 0.2362 - val_accuracy: 0.9772\n",
            "Epoch 5/10\n",
            "1875/1875 [==============================] - 7s 4ms/step - loss: 0.1023 - accuracy: 0.9877 - val_loss: 0.3062 - val_accuracy: 0.9795\n",
            "Epoch 6/10\n",
            "1875/1875 [==============================] - 7s 4ms/step - loss: 0.1103 - accuracy: 0.9890 - val_loss: 0.5387 - val_accuracy: 0.9723\n",
            "Epoch 7/10\n",
            "1875/1875 [==============================] - 7s 4ms/step - loss: 0.0944 - accuracy: 0.9916 - val_loss: 0.4826 - val_accuracy: 0.9785\n",
            "Epoch 8/10\n",
            "1875/1875 [==============================] - 7s 4ms/step - loss: 0.1014 - accuracy: 0.9922 - val_loss: 0.4421 - val_accuracy: 0.9804\n",
            "Epoch 9/10\n",
            "1875/1875 [==============================] - 7s 4ms/step - loss: 0.0913 - accuracy: 0.9935 - val_loss: 0.4831 - val_accuracy: 0.9797\n",
            "Epoch 10/10\n",
            "1875/1875 [==============================] - 7s 4ms/step - loss: 0.0860 - accuracy: 0.9943 - val_loss: 0.6276 - val_accuracy: 0.9810\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tensorflow.python.keras.callbacks.History at 0x7f54cc3a0978>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 19
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sw8LhOH51UZg"
      },
      "source": [
        "The prediction accuracy of this model is 99.38%"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gveRsaURk2Gf"
      },
      "source": [
        "# Keras Tuner:\n",
        "[Referance](https://keras-team.github.io/keras-tuner/)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2VbTI7rmPuwJ"
      },
      "source": [
        "###  1. Using Relu actication function"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DkZ1OSiLk5mM"
      },
      "source": [
        "import keras\n",
        "\n",
        "def build_model(hp):\n",
        "  model = keras.Sequential([\n",
        "                            keras.layers.Conv2D(\n",
        "                                filters = hp.Int('filter_1', min_value = 32, max_value = 100, step = 10),\n",
        "                                kernel_size = hp.Choice('kernel_1', values = [3, 5]),\n",
        "                                activation = 'relu',\n",
        "                                input_shape = (28, 28, 1)\n",
        "                            ),\n",
        "                            keras.layers.Conv2D(\n",
        "                                filters = hp.Int('filter_2', min_value = 32, max_value = 64, step = 10),\n",
        "                                kernel_size = hp.Choice('kernel_2', values = [3, 5]),\n",
        "                                activation = 'relu'\n",
        "                            ),\n",
        "                            keras.layers.Flatten(),\n",
        "                            keras.layers.Dense(\n",
        "                                units = hp.Int('dense_1', min_value = 32, max_value = 100, step = 10),\n",
        "                                activation = 'relu'\n",
        "                            ),\n",
        "                            keras.layers.Dense(10, activation='softmax')\n",
        "  ])\n",
        "\n",
        "  model.compile(optimizer = keras.optimizers.Adam(hp.Choice('learning_rate', values = [1e-2, 1e-3])),\n",
        "                loss = 'categorical_crossentropy',\n",
        "                metrics = ['accuracy'])\n",
        "  \n",
        "  return model\n",
        "  "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "NEueP9OHlimM",
        "outputId": "35a7be25-8445-4e79-88d6-a8588f3937fa"
      },
      "source": [
        "pip install -U keras-tuner"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Collecting keras-tuner\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/20/ec/1ef246787174b1e2bb591c95f29d3c1310070cad877824f907faba3dade9/keras-tuner-1.0.2.tar.gz (62kB)\n",
            "\r\u001b[K     |█████▏                          | 10kB 20.0MB/s eta 0:00:01\r\u001b[K     |██████████▍                     | 20kB 16.9MB/s eta 0:00:01\r\u001b[K     |███████████████▋                | 30kB 13.7MB/s eta 0:00:01\r\u001b[K     |████████████████████▉           | 40kB 12.5MB/s eta 0:00:01\r\u001b[K     |██████████████████████████      | 51kB 8.5MB/s eta 0:00:01\r\u001b[K     |███████████████████████████████▎| 61kB 9.1MB/s eta 0:00:01\r\u001b[K     |████████████████████████████████| 71kB 6.1MB/s \n",
            "\u001b[?25hRequirement already satisfied, skipping upgrade: packaging in /usr/local/lib/python3.6/dist-packages (from keras-tuner) (20.4)\n",
            "Requirement already satisfied, skipping upgrade: future in /usr/local/lib/python3.6/dist-packages (from keras-tuner) (0.16.0)\n",
            "Requirement already satisfied, skipping upgrade: numpy in /usr/local/lib/python3.6/dist-packages (from keras-tuner) (1.18.5)\n",
            "Requirement already satisfied, skipping upgrade: tabulate in /usr/local/lib/python3.6/dist-packages (from keras-tuner) (0.8.7)\n",
            "Collecting terminaltables\n",
            "  Downloading https://files.pythonhosted.org/packages/9b/c4/4a21174f32f8a7e1104798c445dacdc1d4df86f2f26722767034e4de4bff/terminaltables-3.1.0.tar.gz\n",
            "Collecting colorama\n",
            "  Downloading https://files.pythonhosted.org/packages/44/98/5b86278fbbf250d239ae0ecb724f8572af1c91f4a11edf4d36a206189440/colorama-0.4.4-py2.py3-none-any.whl\n",
            "Requirement already satisfied, skipping upgrade: tqdm in /usr/local/lib/python3.6/dist-packages (from keras-tuner) (4.41.1)\n",
            "Requirement already satisfied, skipping upgrade: requests in /usr/local/lib/python3.6/dist-packages (from keras-tuner) (2.23.0)\n",
            "Requirement already satisfied, skipping upgrade: scipy in /usr/local/lib/python3.6/dist-packages (from keras-tuner) (1.4.1)\n",
            "Requirement already satisfied, skipping upgrade: scikit-learn in /usr/local/lib/python3.6/dist-packages (from keras-tuner) (0.22.2.post1)\n",
            "Requirement already satisfied, skipping upgrade: pyparsing>=2.0.2 in /usr/local/lib/python3.6/dist-packages (from packaging->keras-tuner) (2.4.7)\n",
            "Requirement already satisfied, skipping upgrade: six in /usr/local/lib/python3.6/dist-packages (from packaging->keras-tuner) (1.15.0)\n",
            "Requirement already satisfied, skipping upgrade: chardet<4,>=3.0.2 in /usr/local/lib/python3.6/dist-packages (from requests->keras-tuner) (3.0.4)\n",
            "Requirement already satisfied, skipping upgrade: idna<3,>=2.5 in /usr/local/lib/python3.6/dist-packages (from requests->keras-tuner) (2.10)\n",
            "Requirement already satisfied, skipping upgrade: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.6/dist-packages (from requests->keras-tuner) (1.24.3)\n",
            "Requirement already satisfied, skipping upgrade: certifi>=2017.4.17 in /usr/local/lib/python3.6/dist-packages (from requests->keras-tuner) (2020.11.8)\n",
            "Requirement already satisfied, skipping upgrade: joblib>=0.11 in /usr/local/lib/python3.6/dist-packages (from scikit-learn->keras-tuner) (0.17.0)\n",
            "Building wheels for collected packages: keras-tuner, terminaltables\n",
            "  Building wheel for keras-tuner (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for keras-tuner: filename=keras_tuner-1.0.2-cp36-none-any.whl size=78937 sha256=51d9ad64798d110fda350775a64fd22220002e25f66b12ae521ecffd65fcc63f\n",
            "  Stored in directory: /root/.cache/pip/wheels/bb/a1/8a/7c3de0efb3707a1701b36ebbfdbc4e67aedf6d4943a1f463d6\n",
            "  Building wheel for terminaltables (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for terminaltables: filename=terminaltables-3.1.0-cp36-none-any.whl size=15356 sha256=1bd9aff4ac8ca364bf47822ed5163bd0b6931a195fb404bbe1a41ada6e54801e\n",
            "  Stored in directory: /root/.cache/pip/wheels/30/6b/50/6c75775b681fb36cdfac7f19799888ef9d8813aff9e379663e\n",
            "Successfully built keras-tuner terminaltables\n",
            "Installing collected packages: terminaltables, colorama, keras-tuner\n",
            "Successfully installed colorama-0.4.4 keras-tuner-1.0.2 terminaltables-3.1.0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "I6h-94ALlJr-"
      },
      "source": [
        "from kerastuner import RandomSearch\n",
        "from kerastuner.engine.hyperparameters import HyperParameters"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "mztRdAtglPOK",
        "outputId": "bf88ea45-016d-493d-a14a-594aeb7ff7b2"
      },
      "source": [
        "tuner = RandomSearch(build_model, objective='val_accuracy', max_trials = 10, \n",
        "                     directory = 'output', project_name = 'DR')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Reloading Oracle from existing project output/DR/oracle.json\n",
            "INFO:tensorflow:Reloading Tuner from output/DR/tuner0.json\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zqMWpQliRIV-",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "ceb52379-779a-4cb7-f29d-a881db8c84b5"
      },
      "source": [
        "tuner.search_space_summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Search space summary\n",
            "Default search space size: 6\n",
            "filter_1 (Int)\n",
            "{'default': None, 'conditions': [], 'min_value': 32, 'max_value': 100, 'step': 10, 'sampling': None}\n",
            "kernel_1 (Choice)\n",
            "{'default': 3, 'conditions': [], 'values': [3, 5], 'ordered': True}\n",
            "filter_2 (Int)\n",
            "{'default': None, 'conditions': [], 'min_value': 32, 'max_value': 64, 'step': 10, 'sampling': None}\n",
            "kernel_2 (Choice)\n",
            "{'default': 3, 'conditions': [], 'values': [3, 5], 'ordered': True}\n",
            "dense_1 (Int)\n",
            "{'default': None, 'conditions': [], 'min_value': 32, 'max_value': 100, 'step': 10, 'sampling': None}\n",
            "lerning_rate (Choice)\n",
            "{'default': 0.01, 'conditions': [], 'values': [0.01, 0.001], 'ordered': True}\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Ims_UNJrmaHM",
        "outputId": "7b556199-636d-4e72-a30e-044e4a92edde"
      },
      "source": [
        "tuner.search(X_train, y_train, epochs = 30, validation_split = 0.2)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Trial 5 Complete [00h 03m 00s]\n",
            "val_accuracy: 0.1080833300948143\n",
            "\n",
            "Best val_accuracy So Far: 0.9835000038146973\n",
            "Total elapsed time: 00h 14m 21s\n",
            "INFO:tensorflow:Oracle triggered exit\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AUCqLngO_Ba4"
      },
      "source": [
        "best_model_1 = tuner.get_best_models(num_models=1)[0]"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "BvNLf-JiA2BK",
        "outputId": "fa9aaefd-b78d-40e4-ce50-9a68c4091a39"
      },
      "source": [
        "best_model_1.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d (Conv2D)              (None, 24, 24, 82)        2132      \n",
            "_________________________________________________________________\n",
            "conv2d_1 (Conv2D)            (None, 22, 22, 52)        38428     \n",
            "_________________________________________________________________\n",
            "flatten (Flatten)            (None, 25168)             0         \n",
            "_________________________________________________________________\n",
            "dense (Dense)                (None, 82)                2063858   \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 10)                830       \n",
            "=================================================================\n",
            "Total params: 2,105,248\n",
            "Trainable params: 2,105,248\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "V7YO8Xn3Brw-"
      },
      "source": [
        " **Now we apply this model to our train data and then see the accuracy of this best model**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0WkkNxmLA_ZF",
        "outputId": "2f706461-6d37-4425-b72e-1c397ebc7856"
      },
      "source": [
        "history_1 = best_model_1.fit(X_train, y_train, epochs = 10, validation_split=0.1, initial_epoch=3)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 4/10\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0241 - accuracy: 0.9963 - val_loss: 0.2860 - val_accuracy: 0.9845\n",
            "Epoch 5/10\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0241 - accuracy: 0.9964 - val_loss: 0.2689 - val_accuracy: 0.9813\n",
            "Epoch 6/10\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0183 - accuracy: 0.9970 - val_loss: 0.2431 - val_accuracy: 0.9852\n",
            "Epoch 7/10\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0242 - accuracy: 0.9968 - val_loss: 0.2892 - val_accuracy: 0.9870\n",
            "Epoch 8/10\n",
            "1688/1688 [==============================] - 7s 4ms/step - loss: 0.0224 - accuracy: 0.9974 - val_loss: 0.3042 - val_accuracy: 0.9847\n",
            "Epoch 9/10\n",
            "1688/1688 [==============================] - 7s 4ms/step - loss: 0.0230 - accuracy: 0.9972 - val_loss: 0.3372 - val_accuracy: 0.9867\n",
            "Epoch 10/10\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0269 - accuracy: 0.9971 - val_loss: 0.3315 - val_accuracy: 0.9862\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5lJ2NHm6SCvP"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "koW9-zDDdM5w",
        "outputId": "658a90a9-4077-47da-e5a8-0e15af4f642a"
      },
      "source": [
        "train_acc = history_1.history['accuracy']\n",
        "val_acc = history_1.history['val_accuracy']\n",
        "epochs = range(4, 11)\n",
        "plt.plot(epochs, train_acc, 'g', label='Training accuracy')\n",
        "plt.plot(epochs, val_acc, 'b', label='validation accuracy')\n",
        "plt.title('Training and Validation accuracy')\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pI7ubJNSEB1C"
      },
      "source": [
        "#### Prediction:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GUoFPzqoCNaA"
      },
      "source": [
        "prediction = best_model_1.predict(X_test)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "VtOwY5lwEYgP",
        "outputId": "d76f5a45-b773-4d20-d005-16dc46ba10c4"
      },
      "source": [
        "prediction"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[0.000000e+00, 0.000000e+00, 0.000000e+00, ..., 1.000000e+00,\n",
              "        0.000000e+00, 0.000000e+00],\n",
              "       [0.000000e+00, 0.000000e+00, 1.000000e+00, ..., 0.000000e+00,\n",
              "        0.000000e+00, 0.000000e+00],\n",
              "       [0.000000e+00, 1.000000e+00, 0.000000e+00, ..., 4.572769e-37,\n",
              "        0.000000e+00, 0.000000e+00],\n",
              "       ...,\n",
              "       [0.000000e+00, 0.000000e+00, 0.000000e+00, ..., 0.000000e+00,\n",
              "        0.000000e+00, 0.000000e+00],\n",
              "       [0.000000e+00, 0.000000e+00, 0.000000e+00, ..., 0.000000e+00,\n",
              "        0.000000e+00, 0.000000e+00],\n",
              "       [0.000000e+00, 0.000000e+00, 0.000000e+00, ..., 0.000000e+00,\n",
              "        0.000000e+00, 0.000000e+00]], dtype=float32)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 64
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2Cr4lUCdG3qv",
        "outputId": "41e9d707-5631-4af2-c4a2-74c36283d652"
      },
      "source": [
        "print(np.argmax(prediction[0]))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "7\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "u60o2WK9HfrB"
      },
      "source": [
        "our model predicts that, the 1st number in dataset is 7\n",
        "\n",
        "lets see the label of this image in y_test"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "9rDqxO8aHBuH",
        "outputId": "4fcca406-47ec-4c36-d37f-48f757a87ea7"
      },
      "source": [
        "y_test[0]"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([0., 0., 0., 0., 0., 0., 0., 1., 0., 0.], dtype=float32)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 66
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TW0NpOJgHzTY"
      },
      "source": [
        "So prediction of our model is correct."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7b4flIlFRNwA"
      },
      "source": [
        "### 2. Using Leaky relu activation function"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "59pnR9muRZNQ"
      },
      "source": [
        "lrelu = lambda x: tf.keras.activations.relu(x, alpha=0.1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vdQXJSLFI9V-"
      },
      "source": [
        "import keras\n",
        "\n",
        "def build_model(hp):\n",
        "  model = keras.Sequential([\n",
        "                            keras.layers.Conv2D(\n",
        "                                filters = hp.Int('filter_1', min_value = 32, max_value = 100, step = 10),\n",
        "                                kernel_size = hp.Choice('kernel_1', values = [3, 5]),\n",
        "                                activation = lrelu,\n",
        "                                input_shape = (28, 28, 1)\n",
        "                            ),\n",
        "                            keras.layers.Conv2D(\n",
        "                                filters = hp.Int('filter_2', min_value = 32, max_value = 64, step = 10),\n",
        "                                kernel_size = hp.Choice('kernel_2', values = [3, 5]),\n",
        "                                activation = lrelu\n",
        "                            ),\n",
        "                            keras.layers.Flatten(),\n",
        "                            keras.layers.Dense(\n",
        "                                units = hp.Int('dense_1', min_value = 32, max_value = 100, step = 10),\n",
        "                                activation = lrelu\n",
        "                            ),\n",
        "                            keras.layers.Dense(10, activation='softmax')\n",
        "  ])\n",
        "\n",
        "  model.compile(optimizer = keras.optimizers.Adam(hp.Choice('lerning_rate', values = [1e-2, 1e-3])),\n",
        "                loss = 'categorical_crossentropy',\n",
        "                metrics = ['accuracy'])\n",
        "  \n",
        "  return model"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Uuw9mudARyNy",
        "outputId": "f1f6016d-5421-40f5-a1be-5f003460174e"
      },
      "source": [
        "tuner_1 = RandomSearch(build_model, objective='val_accuracy', max_trials = 5)\n",
        "\n",
        "tuner_1.search(X_train, y_train, epochs = 3, validation_split = 0.1)\n",
        "\n",
        "best_model_2 = tuner.get_best_models(num_models=1)[0]\n",
        "\n",
        "best_model_2.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Reloading Oracle from existing project ./untitled_project/oracle.json\n",
            "INFO:tensorflow:Reloading Tuner from ./untitled_project/tuner0.json\n",
            "INFO:tensorflow:Oracle triggered exit\n",
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d (Conv2D)              (None, 24, 24, 82)        2132      \n",
            "_________________________________________________________________\n",
            "conv2d_1 (Conv2D)            (None, 22, 22, 52)        38428     \n",
            "_________________________________________________________________\n",
            "flatten (Flatten)            (None, 25168)             0         \n",
            "_________________________________________________________________\n",
            "dense (Dense)                (None, 82)                2063858   \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 10)                830       \n",
            "=================================================================\n",
            "Total params: 2,105,248\n",
            "Trainable params: 2,105,248\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "yiQeHeHsSEme",
        "outputId": "dabcbbf0-feda-4fcb-9268-0c24343582f3"
      },
      "source": [
        "history_2 = best_model_2.fit(X_train, y_train, epochs = 10, validation_split=0.1, initial_epoch=3)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 4/10\n",
            "1688/1688 [==============================] - 7s 4ms/step - loss: 0.0493 - accuracy: 0.9847 - val_loss: 0.0945 - val_accuracy: 0.9755\n",
            "Epoch 5/10\n",
            "1688/1688 [==============================] - 7s 4ms/step - loss: 0.0395 - accuracy: 0.9881 - val_loss: 0.0787 - val_accuracy: 0.9813\n",
            "Epoch 6/10\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0312 - accuracy: 0.9905 - val_loss: 0.0763 - val_accuracy: 0.9825\n",
            "Epoch 7/10\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0282 - accuracy: 0.9922 - val_loss: 0.0992 - val_accuracy: 0.9820\n",
            "Epoch 8/10\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0230 - accuracy: 0.9938 - val_loss: 0.0884 - val_accuracy: 0.9835\n",
            "Epoch 9/10\n",
            "1688/1688 [==============================] - 7s 4ms/step - loss: 0.0244 - accuracy: 0.9936 - val_loss: 0.1204 - val_accuracy: 0.9832\n",
            "Epoch 10/10\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0253 - accuracy: 0.9943 - val_loss: 0.1370 - val_accuracy: 0.9833\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "id": "MLsVzRDBb20T",
        "outputId": "d6561a36-462a-4bbe-d9a3-d20761131a1c"
      },
      "source": [
        "train_acc = history_2.history['accuracy']\n",
        "val_acc = history_2.history['val_accuracy']\n",
        "epochs = range(4, 11)\n",
        "plt.plot(epochs, train_acc, 'g', label='Training accuracy')\n",
        "plt.plot(epochs, val_acc, 'b', label='validation accuracy')\n",
        "plt.title('Training and Validation Accuracy')\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "X97YfNIkTx1C"
      },
      "source": [
        "#### Prediction"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "4mdRJ_IkSOBi",
        "outputId": "a8d88a15-2125-44b4-ab4b-439d56808882"
      },
      "source": [
        "prediction_lrelu = best_model_2.predict(X_test)\n",
        "print(np.argmax(prediction_lrelu[0]))\n",
        "print(y_test[0])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "7\n",
            "[0. 0. 0. 0. 0. 0. 0. 1. 0. 0.]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gcKLZXsTT_CE"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7KAyXcu-UlrJ"
      },
      "source": [
        "## 3. Using Prelu activation function:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9vTASZmWUve3"
      },
      "source": [
        "from keras.activations import elu\n",
        "\n",
        "ELU = lambda x: elu(x, alpha=0.4)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RMcms97XUyQA"
      },
      "source": [
        "import keras\n",
        "\n",
        "def build_model(hp):\n",
        "  model = keras.Sequential([\n",
        "                            keras.layers.Conv2D(\n",
        "                                filters = hp.Int('filter_1', min_value = 32, max_value = 100, step = 10),\n",
        "                                kernel_size = hp.Choice('kernel_1', values = [3, 5]),\n",
        "                                activation = ELU,\n",
        "                                input_shape = (28, 28, 1)\n",
        "                            ),\n",
        "                            keras.layers.Conv2D(\n",
        "                                filters = hp.Int('filter_2', min_value = 32, max_value = 64, step = 10),\n",
        "                                kernel_size = hp.Choice('kernel_2', values = [3, 5]),\n",
        "                                activation = ELU\n",
        "                            ),\n",
        "                            keras.layers.Flatten(),\n",
        "                            keras.layers.Dense(\n",
        "                                units = hp.Int('dense_1', min_value = 32, max_value = 100, step = 10),\n",
        "                                activation = ELU\n",
        "                            ),\n",
        "                            keras.layers.Dense(10, activation='softmax')\n",
        "  ])\n",
        "\n",
        "  model.compile(optimizer = keras.optimizers.Adam(hp.Choice('lerning_rate', values = [1e-2, 1e-3])),\n",
        "                loss = 'categorical_crossentropy',\n",
        "                metrics = ['accuracy'])\n",
        "  \n",
        "  return model"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "-6asLlnJU1H5",
        "outputId": "259754a7-814f-450d-93a1-1bf7ee3c1e36"
      },
      "source": [
        "tuner_3 = RandomSearch(build_model, objective='val_accuracy', max_trials = 5)\n",
        "\n",
        "tuner_3.search(X_train, y_train, epochs = 3, validation_split = 0.1)\n",
        "\n",
        "best_model_3 = tuner.get_best_models(num_models=1)[0]\n",
        "\n",
        "best_model_3.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Reloading Oracle from existing project ./untitled_project/oracle.json\n",
            "INFO:tensorflow:Reloading Tuner from ./untitled_project/tuner0.json\n",
            "INFO:tensorflow:Oracle triggered exit\n",
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "conv2d (Conv2D)              (None, 24, 24, 82)        2132      \n",
            "_________________________________________________________________\n",
            "conv2d_1 (Conv2D)            (None, 22, 22, 52)        38428     \n",
            "_________________________________________________________________\n",
            "flatten (Flatten)            (None, 25168)             0         \n",
            "_________________________________________________________________\n",
            "dense (Dense)                (None, 82)                2063858   \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 10)                830       \n",
            "=================================================================\n",
            "Total params: 2,105,248\n",
            "Trainable params: 2,105,248\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "YrGQr0gqVDnH",
        "outputId": "b4e5ba6f-4116-4eb2-807b-b679931868ca"
      },
      "source": [
        "history_3 = best_model_3.fit(X_train, y_train, epochs = 20, validation_split = 0.1, initial_epoch=3)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 4/20\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0191 - accuracy: 0.9952 - val_loss: 0.1353 - val_accuracy: 0.9837\n",
            "Epoch 5/20\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0195 - accuracy: 0.9955 - val_loss: 0.1240 - val_accuracy: 0.9835\n",
            "Epoch 6/20\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0211 - accuracy: 0.9959 - val_loss: 0.2495 - val_accuracy: 0.9840\n",
            "Epoch 7/20\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0216 - accuracy: 0.9963 - val_loss: 0.1552 - val_accuracy: 0.9842\n",
            "Epoch 8/20\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0187 - accuracy: 0.9967 - val_loss: 0.2168 - val_accuracy: 0.9832\n",
            "Epoch 9/20\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0204 - accuracy: 0.9965 - val_loss: 0.1785 - val_accuracy: 0.9845\n",
            "Epoch 10/20\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0143 - accuracy: 0.9976 - val_loss: 0.2243 - val_accuracy: 0.9820\n",
            "Epoch 11/20\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0282 - accuracy: 0.9957 - val_loss: 0.2274 - val_accuracy: 0.9842\n",
            "Epoch 12/20\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0195 - accuracy: 0.9972 - val_loss: 0.2580 - val_accuracy: 0.9825\n",
            "Epoch 13/20\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0165 - accuracy: 0.9973 - val_loss: 0.2562 - val_accuracy: 0.9842\n",
            "Epoch 14/20\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0228 - accuracy: 0.9971 - val_loss: 0.3079 - val_accuracy: 0.9835\n",
            "Epoch 15/20\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0198 - accuracy: 0.9974 - val_loss: 0.2811 - val_accuracy: 0.9872\n",
            "Epoch 16/20\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0209 - accuracy: 0.9975 - val_loss: 0.3137 - val_accuracy: 0.9837\n",
            "Epoch 17/20\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0230 - accuracy: 0.9978 - val_loss: 0.2972 - val_accuracy: 0.9852\n",
            "Epoch 18/20\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0190 - accuracy: 0.9980 - val_loss: 0.3895 - val_accuracy: 0.9842\n",
            "Epoch 19/20\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0257 - accuracy: 0.9978 - val_loss: 0.3486 - val_accuracy: 0.9873\n",
            "Epoch 20/20\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.0195 - accuracy: 0.9981 - val_loss: 0.4841 - val_accuracy: 0.9858\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 312
        },
        "id": "yojXryhmZgRo",
        "outputId": "00cce217-0af9-4e2c-d1ac-d342b2a84a74"
      },
      "source": [
        "train_acc = history_3.history['accuracy']\n",
        "print(len(train_acc))\n",
        "val_acc = history_3.history['val_accuracy']\n",
        "epochs = range(4,21)\n",
        "plt.plot(epochs, train_acc, 'g', label='Training accuracy')\n",
        "plt.plot(epochs, val_acc, 'b', label='validation accuracy')\n",
        "plt.title('Training and Validation Accuracy')\n",
        "plt.xlabel('Epochs')\n",
        "plt.ylabel('Accuracy')\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "17\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Q3t4ZXP7VNHt"
      },
      "source": [
        "#### Prediction:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "eOosuhOuVIzK",
        "outputId": "e6f282c2-b51f-4939-ba5d-bcd9296c2a1d"
      },
      "source": [
        "prediction_Prelu = best_model_3.predict(X_test)\n",
        "print(np.argmax(prediction_Prelu[0]))\n",
        "print(np.argmax(y_test[0]))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "7\n",
            "7\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kzN7btOIWeas"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}