Exercise_3_Convolution_Neural_Networks_for_Image_Analysis.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Exercise 3 - Convolution Neural Networks for Image Analysis",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DFUsJZSU-wJF"
      },
      "source": [
        "# Exercise 3 - Convolution Neural Networks for Image Analysis"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XusAABHD-58D"
      },
      "source": [
        "In this workshop we will usea CNN to classify images in the  [Fashion MNIST](https://github.com/zalandoresearch/fashion-mnist) dataset. \n",
        "\n",
        "We will classify the fashion products in Fashion MNIST  into 10 classes. The dataset contains 70,000 grayscale images, with resolution of 28x28 pixels.   \n",
        "We will use 60,000 images to train the classification CNN and 10,000 images to evaluate the accuracy of the network. "
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "NRZEWyM6_fb1"
      },
      "source": [
        "## Load the data"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3A9P3pEB_j13"
      },
      "source": [
        "Generally we have to download the dataset from a repository and upload to use in our application. However, Keras python library provides us access to directly load a limited number of benchmark datasets.   \n",
        "Thus, we can access and load the Fashion-MNIST dataset directly from Keras."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "n7iIFLbF-Acy"
      },
      "source": [
        "# Load Tensorflow library\n",
        "from keras.datasets import fashion_mnist"
      ],
      "execution_count": 1,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-HWedSm3DPXh"
      },
      "source": [
        "The dataset is already setup as training and testing sets. (60K for training and 10K for testing)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6migTJqA_yhW",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "494bd36e-2a19-44a5-b559-d74280a0bd46"
      },
      "source": [
        "(X_train, y_train), (X_test, y_test) = fashion_mnist.load_data()"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-labels-idx1-ubyte.gz\n",
            "32768/29515 [=================================] - 0s 0us/step\n",
            "40960/29515 [=========================================] - 0s 0us/step\n",
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/train-images-idx3-ubyte.gz\n",
            "26427392/26421880 [==============================] - 0s 0us/step\n",
            "26435584/26421880 [==============================] - 0s 0us/step\n",
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-labels-idx1-ubyte.gz\n",
            "16384/5148 [===============================================================================================] - 0s 0us/step\n",
            "Downloading data from https://storage.googleapis.com/tensorflow/tf-keras-datasets/t10k-images-idx3-ubyte.gz\n",
            "4423680/4422102 [==============================] - 0s 0us/step\n",
            "4431872/4422102 [==============================] - 0s 0us/step\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rlpHxpHbAL9f",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "f26aeea4-7a7a-44d4-f34e-45f80a5de1ea"
      },
      "source": [
        "# View the dataset shape\n",
        "print('Train data shape:', X_train.shape)\n",
        "print('Test data shape:', X_test.shape)"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Train data shape: (60000, 28, 28)\n",
            "Test data shape: (10000, 28, 28)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kIPjfH5RAXne"
      },
      "source": [
        "X_train contains 60,000 train images with a resolution of 28x28.  \n",
        "The *labels* are an array of integers, ranging from 0 to 9. These correspond to the *class* of clothing the image represents:\n",
        "\n",
        "<table>\n",
        "  <tr>\n",
        "    <th>Label</th>\n",
        "    <th>Class</th> \n",
        "  </tr>\n",
        "  <tr>\n",
        "    <td>0</td>\n",
        "    <td>T-shirt/top</td> \n",
        "  </tr>\n",
        "  <tr>\n",
        "    <td>1</td>\n",
        "    <td>Trouser</td> \n",
        "  </tr>\n",
        "    <tr>\n",
        "    <td>2</td>\n",
        "    <td>Pullover</td> \n",
        "  </tr>\n",
        "    <tr>\n",
        "    <td>3</td>\n",
        "    <td>Dress</td> \n",
        "  </tr>\n",
        "    <tr>\n",
        "    <td>4</td>\n",
        "    <td>Coat</td> \n",
        "  </tr>\n",
        "    <tr>\n",
        "    <td>5</td>\n",
        "    <td>Sandal</td> \n",
        "  </tr>\n",
        "    <tr>\n",
        "    <td>6</td>\n",
        "    <td>Shirt</td> \n",
        "  </tr>\n",
        "    <tr>\n",
        "    <td>7</td>\n",
        "    <td>Sneaker</td> \n",
        "  </tr>\n",
        "    <tr>\n",
        "    <td>8</td>\n",
        "    <td>Bag</td> \n",
        "  </tr>\n",
        "    <tr>\n",
        "    <td>9</td>\n",
        "    <td>Ankle boot</td> \n",
        "  </tr>\n",
        "</table>\n",
        "\n",
        "Each image is mapped to a single label. Since the *class names* are not included with the dataset, store them here to use later when plotting the images:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oGMrPcgJAUnf"
      },
      "source": [
        "class_names = ['T-shirt/top', 'Trouser', 'Pullover', 'Dress', 'Coat', \n",
        "               'Sandal', 'Shirt', 'Sneaker', 'Bag', 'Ankle boot']"
      ],
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3TjH6DZ7Bj_b"
      },
      "source": [
        "## Explore and pre-process the data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uVSDAnwnAlkD"
      },
      "source": [
        "# Load visualization library\n",
        "import matplotlib.pyplot as plt"
      ],
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "cguV38FYDgSQ"
      },
      "source": [
        "We will plot an image from the dataset to get a feel for the data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_jnlp88TBtTS",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 265
        },
        "outputId": "5edc006d-0b1d-4b57-d2b1-48a1553691ee"
      },
      "source": [
        "image_index = 1\n",
        "plt.figure()\n",
        "plt.imshow(X_train[image_index])\n",
        "plt.colorbar()\n",
        "plt.grid(False)\n",
        "plt.show()"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "rYGKlpKUCRcm"
      },
      "source": [
        "Display the first 25 images from the training set and display the class name below each image.   \n",
        "It is important to verify that the data is in the correct format before building the model."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "brodZC5qBvNc",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 589
        },
        "outputId": "dcb1031b-0459-4cc6-e095-41f49006ef31"
      },
      "source": [
        "plt.figure(figsize=(10,10))\n",
        "for i in range(25):\n",
        "    plt.subplot(5,5,i+1)\n",
        "    plt.xticks([])\n",
        "    plt.yticks([])\n",
        "    plt.grid(False)\n",
        "    plt.imshow(X_train[i], cmap=plt.cm.binary)\n",
        "    plt.xlabel(class_names[y_train[i]])\n",
        "plt.show()"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 720x720 with 25 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Yd-lPPNaCfDQ"
      },
      "source": [
        "A main step of image processing is to scale the data. With this step, we will be able observe  the range of data."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LPn5SFTjCeI7"
      },
      "source": [
        "# Import numpy library to process matrices\n",
        "import numpy as np"
      ],
      "execution_count": 8,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eZk_duGmB3XY",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "2565d4f0-1936-48f5-e859-fb157fa45db6"
      },
      "source": [
        "print('min:', np.min(X_train[image_index]))\n",
        "print('max:', np.max(X_train[image_index]))"
      ],
      "execution_count": 9,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "min: 0\n",
            "max: 255\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MURb0-JMDNjr"
      },
      "source": [
        "It is evident that the data ranges from 0-255.   \n",
        "We scale these values to a range of 0 to 1 before feeding to the neural network model.  \n",
        "For this, we divide the values by 255.   \n",
        "It's important that the training set and the testing set are preprocessed in the same way:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gEnN_wVdDLaf"
      },
      "source": [
        "X_train = X_train / 255.0\n",
        "X_test = X_test / 255.0"
      ],
      "execution_count": 10,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "p2U0u1n-L1j3"
      },
      "source": [
        "Prepare for CNN processing by reshaping the image into (# training/testing samples, height, width, 1)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xXdZsPpcMCDD"
      },
      "source": [
        "height, width = 28, 28"
      ],
      "execution_count": 11,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rrAKyhukL9Bi"
      },
      "source": [
        "X_train = X_train.reshape(X_train.shape[0], height, width, 1)\n",
        "X_test = X_test.reshape(X_test.shape[0], height, width, 1)"
      ],
      "execution_count": 12,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ClwHrfPUCwQt"
      },
      "source": [
        "## Enable GPU\n",
        "\n",
        "CPUs are designed for more general computing workloads.  GPUs in contrast are less flexible, however GPUs are designed for parallel  computations. Deep Neural Networks (DNN) are structured in a very uniform manner such that at each layer of the network thousands of identical artificial neurons perform the same computation.   Therefore the structure of a DNN fits quite well with the kinds of computation that a GPU can efficiently perform.\n",
        "\n",
        "To enable GPU in Google Colab, \n",
        "\n",
        "\n",
        "1. Click on Runtime menu on the toolbar\n",
        "2. Select change runtime type\n",
        "3. Select GPU for Hardware Acceleration option list.\n",
        "4. Select SAVE.\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "IyLSG4IyDtbX"
      },
      "source": [
        "## Model Building"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "HdCD6-1NDwsq"
      },
      "source": [
        "In this workshop, we will develop 3 deep learning models with increasing complexity to evaluate the classification accuracy.  \n",
        "\n",
        "\n",
        "1.   3 Layer DNN\n",
        "2.   1 Layer CNN\n",
        "3.   3 Layer CNN"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "hGno2vDBEvnp"
      },
      "source": [
        "### 3-DNN Model"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-7TX5AmwFIc8"
      },
      "source": [
        "In this DNN model, we flatten the 28x28 image into a 784 input feature vector.  \n",
        "It should be noted that with flattening we will face the inherent problems discussed in the class (i.e., scale, rotation, location, high-demensionality)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ud-2924qDfqP"
      },
      "source": [
        "# Import Keras libraries\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Dense, Flatten"
      ],
      "execution_count": 13,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZoTnr8D9E7oQ"
      },
      "source": [
        "dnn3_model = Sequential()"
      ],
      "execution_count": 14,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "v0_5zNC3FAwx"
      },
      "source": [
        "dnn3_model.add(Flatten(input_shape=(width, height, 1)))  # Add Keras Flatten layer to conver (28x28) image -> 784 feature vector"
      ],
      "execution_count": 15,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kKztFDk1FvBs"
      },
      "source": [
        "dnn3_model.add(Dense(128, activation='relu')) \n",
        "dnn3_model.add(Dense(64, activation='relu')) \n",
        "dnn3_model.add(Dense(10, activation='softmax'))  # In the final layer we use softmax activation to classify the input into 10 classes."
      ],
      "execution_count": 16,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vYJr_sT2GRsk"
      },
      "source": [
        "Softmax is an activation function that turns logits into probabilities that sum to one. E.g.,\n",
        "\n",
        "---\n",
        "\n",
        "\n",
        " \n",
        "![alt text](https://cdn-images-1.medium.com/max/906/1*670CdxchunD-yAuUWdI7Bw.png)\n",
        "\n",
        "For a complete understanding of available activation functions, please refer the following resources:  \n",
        "\n",
        "\n",
        "1.   [ML Cheatsheat](https://ml-cheatsheet.readthedocs.io/en/latest/activation_functions.html)\n",
        "2.   [Towards data science](https://towardsdatascience.com/activation-functions-neural-networks-1cbd9f8d91d6)\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "l0gwey-RIE-r"
      },
      "source": [
        "**Compiling the model**\n",
        "\n",
        "Before the model is ready for training, it needs a few more settings. These are added during the model's compile step:  \n",
        "\n",
        "* Loss function —This measures how accurate the model is during training. We want to minimize this function to \"steer\" the model in the right direction. [Further details](https://keras.io/losses/) \n",
        "* Optimizer —This is how the model is updated based on the data it sees and its loss function. [Further details.](https://keras.io/optimizers/)\n",
        "* Metrics —Used to monitor the training and testing steps. The following example uses accuracy, the fraction of the images that are correctly classified. [Further details.](https://keras.io/metrics/) "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hAzocclQGQWI"
      },
      "source": [
        "dnn3_model.compile(optimizer='adam', \n",
        "              loss='sparse_categorical_crossentropy',\n",
        "              metrics=['accuracy'])"
      ],
      "execution_count": 17,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pfXpA1jNEjvg"
      },
      "source": [
        "Initially, we will train the model with 10 learning epochs. Based on the learning curve, visualized below, you could update this hyper-parameter to best fit the training data."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NYNDvwNrIWRO",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "8489dcf8-9473-4476-c876-6bba59593217"
      },
      "source": [
        "# Train the model\n",
        "epochs = 10\n",
        "batch_size = 32\n",
        "validation_split = 0.1\n",
        "history = dnn3_model.fit(X_train, y_train, epochs=epochs, batch_size=batch_size, validation_split = validation_split)"
      ],
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/10\n",
            "1688/1688 [==============================] - 8s 3ms/step - loss: 0.5029 - accuracy: 0.8215 - val_loss: 0.3863 - val_accuracy: 0.8565\n",
            "Epoch 2/10\n",
            "1688/1688 [==============================] - 5s 3ms/step - loss: 0.3704 - accuracy: 0.8635 - val_loss: 0.3695 - val_accuracy: 0.8648\n",
            "Epoch 3/10\n",
            "1688/1688 [==============================] - 5s 3ms/step - loss: 0.3336 - accuracy: 0.8768 - val_loss: 0.3664 - val_accuracy: 0.8663\n",
            "Epoch 4/10\n",
            "1688/1688 [==============================] - 5s 3ms/step - loss: 0.3115 - accuracy: 0.8838 - val_loss: 0.3295 - val_accuracy: 0.8793\n",
            "Epoch 5/10\n",
            "1688/1688 [==============================] - 5s 3ms/step - loss: 0.2949 - accuracy: 0.8899 - val_loss: 0.3855 - val_accuracy: 0.8638\n",
            "Epoch 6/10\n",
            "1688/1688 [==============================] - 5s 3ms/step - loss: 0.2765 - accuracy: 0.8986 - val_loss: 0.3403 - val_accuracy: 0.8842\n",
            "Epoch 7/10\n",
            "1688/1688 [==============================] - 5s 3ms/step - loss: 0.2658 - accuracy: 0.8996 - val_loss: 0.3153 - val_accuracy: 0.8853\n",
            "Epoch 8/10\n",
            "1688/1688 [==============================] - 5s 3ms/step - loss: 0.2535 - accuracy: 0.9055 - val_loss: 0.3284 - val_accuracy: 0.8810\n",
            "Epoch 9/10\n",
            "1688/1688 [==============================] - 5s 3ms/step - loss: 0.2461 - accuracy: 0.9069 - val_loss: 0.3108 - val_accuracy: 0.8903\n",
            "Epoch 10/10\n",
            "1688/1688 [==============================] - 5s 3ms/step - loss: 0.2348 - accuracy: 0.9116 - val_loss: 0.3182 - val_accuracy: 0.8868\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "I2huim_yJjO3",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "dcfcad03-b52f-4603-85bf-c00d5ee9f97f"
      },
      "source": [
        "# Visualize the training curve\n",
        "plt.plot(history.history['loss'])\n",
        "plt.plot(history.history['val_loss'])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'validation'], loc='upper right')\n",
        "plt.show()"
      ],
      "execution_count": 19,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NeftI0WOJDcm",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "9d4a6da2-6f06-49c0-c2f5-121decab6e92"
      },
      "source": [
        "# Visualize accuracy (mean squared error)\n",
        "plt.plot(history.history['accuracy'])\n",
        "plt.plot(history.history['val_accuracy'])\n",
        "plt.title('model accuracy')\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'validation'], loc='upper right')\n",
        "plt.show()"
      ],
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8NhftIe4J1Ya",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "5ee60d53-87f5-46ec-9cad-696af80ad809"
      },
      "source": [
        "test_loss, test_acc = dnn3_model.evaluate(X_test, y_test)\n",
        "print('Test accuracy:', test_acc)"
      ],
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "313/313 [==============================] - 1s 2ms/step - loss: 0.3317 - accuracy: 0.8851\n",
            "Test accuracy: 0.8851000070571899\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "NAOA4U3yKvw8"
      },
      "source": [
        "### 1-CNN"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7ikjZvn-LLv4"
      },
      "source": [
        "As discussed in the lecture, CNN is specifically designed to process two-dimensional data spaces (focusing on image data).  \n",
        "First, we will design a single layer CNN. This will include one convolutional layer, one max pooling layer and two fully connected layers."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pxLaLrbeKxWX"
      },
      "source": [
        "# Import required libraries\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Dense, Flatten, Conv2D, MaxPooling2D"
      ],
      "execution_count": 22,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Mim4gnhxMmzF"
      },
      "source": [
        "# Set input shape\n",
        "input_shape = (width, height, 1)"
      ],
      "execution_count": 23,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VpdMjo2nZGfU"
      },
      "source": [
        "Construct the CNN model."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "9KIAwoYwLdmD"
      },
      "source": [
        "cnn1_model = Sequential()"
      ],
      "execution_count": 24,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "62awqnJ5ZKHE"
      },
      "source": [
        "Here we will use 3x3 filters (or kernals) to learn."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lxUzzQMtLj5A"
      },
      "source": [
        "cnn1_model.add(Conv2D(32, kernel_size=(3, 3), activation='relu', input_shape=input_shape))   # We will use 32 of 3x3 kernals to learn\n",
        "cnn1_model.add(MaxPooling2D(pool_size=(2, 2)))   # We use 2x2 max pooling"
      ],
      "execution_count": 25,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XtQM_8QDZRRM"
      },
      "source": [
        "Flattent the previous output before sending to fully connected layers."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HKyeW1uZMz7w"
      },
      "source": [
        "cnn1_model.add(Flatten())"
      ],
      "execution_count": 26,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7wgKQuLgZhq7"
      },
      "source": [
        "We will use one fully connected (FC) layer with 128 nodes."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Tgert5CQM-BE"
      },
      "source": [
        "cnn1_model.add(Dense(64, activation='relu'))\n",
        "cnn1_model.add(Dense(10, activation='softmax'))"
      ],
      "execution_count": 27,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Tj7fxj6VNA9R"
      },
      "source": [
        "# Compile the model\n",
        "cnn1_model.compile(optimizer='adam', \n",
        "              loss='sparse_categorical_crossentropy',\n",
        "              metrics=['accuracy'])"
      ],
      "execution_count": 28,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TMy-8xdfZo2U"
      },
      "source": [
        "As before, initially we will train the model for 10 epochs to evaluate the performance.  \n",
        "Then, based on the accuracy/learning curve we will refine the hyper-parameters."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4RkPsb3rNIxG",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "985431eb-c3b1-4788-fd97-65987030fb8b"
      },
      "source": [
        "# Train the model\n",
        "epochs = 10\n",
        "batch_size = 32\n",
        "validation_split = 0.1\n",
        "history_1 = cnn1_model.fit(X_train, y_train, epochs=epochs, batch_size=batch_size, validation_split = validation_split)"
      ],
      "execution_count": 29,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/10\n",
            "1688/1688 [==============================] - 15s 4ms/step - loss: 0.4234 - accuracy: 0.8514 - val_loss: 0.3309 - val_accuracy: 0.8810\n",
            "Epoch 2/10\n",
            "1688/1688 [==============================] - 9s 6ms/step - loss: 0.2859 - accuracy: 0.8967 - val_loss: 0.2731 - val_accuracy: 0.9003\n",
            "Epoch 3/10\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.2415 - accuracy: 0.9109 - val_loss: 0.2678 - val_accuracy: 0.9022\n",
            "Epoch 4/10\n",
            "1688/1688 [==============================] - 6s 3ms/step - loss: 0.2097 - accuracy: 0.9229 - val_loss: 0.2655 - val_accuracy: 0.9018\n",
            "Epoch 5/10\n",
            "1688/1688 [==============================] - 6s 3ms/step - loss: 0.1847 - accuracy: 0.9321 - val_loss: 0.2448 - val_accuracy: 0.9135\n",
            "Epoch 6/10\n",
            "1688/1688 [==============================] - 6s 3ms/step - loss: 0.1654 - accuracy: 0.9383 - val_loss: 0.2690 - val_accuracy: 0.9070\n",
            "Epoch 7/10\n",
            "1688/1688 [==============================] - 6s 3ms/step - loss: 0.1461 - accuracy: 0.9456 - val_loss: 0.2552 - val_accuracy: 0.9137\n",
            "Epoch 8/10\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.1288 - accuracy: 0.9530 - val_loss: 0.2772 - val_accuracy: 0.9100\n",
            "Epoch 9/10\n",
            "1688/1688 [==============================] - 6s 4ms/step - loss: 0.1142 - accuracy: 0.9581 - val_loss: 0.2750 - val_accuracy: 0.9145\n",
            "Epoch 10/10\n",
            "1688/1688 [==============================] - 6s 3ms/step - loss: 0.1035 - accuracy: 0.9614 - val_loss: 0.3061 - val_accuracy: 0.9087\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XKxLPmfENX5w",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "9151c10e-2f44-4335-c6af-b67d5d8e63e9"
      },
      "source": [
        "cnn1_score = cnn1_model.evaluate(X_test, y_test)\n",
        "print('Test loss:', cnn1_score[0])\n",
        "print('Test accuracy:', cnn1_score[1])"
      ],
      "execution_count": 30,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "313/313 [==============================] - 1s 3ms/step - loss: 0.3271 - accuracy: 0.9023\n",
            "Test loss: 0.3270799219608307\n",
            "Test accuracy: 0.9023000001907349\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yj0trFTsNljB",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "d86b66da-3a3f-49c5-a80f-5e45524f0960"
      },
      "source": [
        "# Visualize the training curve\n",
        "plt.plot(history_1.history['loss'])\n",
        "plt.plot(history_1.history['val_loss'])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'validation'], loc='upper right')\n",
        "plt.show()"
      ],
      "execution_count": 31,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yR-gHsmLnzZH"
      },
      "source": [
        "By analyzing the learning curve, you should be able to identify whether training the model longer would provide better results."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "KfgXh5cgj7NH"
      },
      "source": [
        "### 3-CNN"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "HXKbUov0j7NM"
      },
      "source": [
        "In this step, we will design a complex three layer CNN intending an improvement in the accuracy."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "tubeQl30j7NQ"
      },
      "source": [
        "# Import required libraries\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Dense, Flatten, Conv2D, MaxPooling2D"
      ],
      "execution_count": 32,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YKRi5t1Lj7Nb"
      },
      "source": [
        "# Set input shape\n",
        "input_shape = (width, height, 1)"
      ],
      "execution_count": 33,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Uk433vWAj7Ng"
      },
      "source": [
        "cnn3_model = Sequential()"
      ],
      "execution_count": 34,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "msnxlKZoj7Nk"
      },
      "source": [
        "# CNN Layer 1\n",
        "cnn3_model.add(Conv2D(32, kernel_size=(3, 3), activation='relu', input_shape=input_shape))\n",
        "cnn3_model.add(MaxPooling2D((2, 2)))\n",
        "\n",
        "# CNN Layer 2\n",
        "cnn3_model.add(Conv2D(64, kernel_size=(3, 3), activation='relu'))\n",
        "cnn3_model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "\n",
        "# CNN Layer 3\n",
        "cnn3_model.add(Conv2D(128, kernel_size=(3, 3), activation='relu'))"
      ],
      "execution_count": 35,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vTvTcnfOj7No"
      },
      "source": [
        "cnn3_model.add(Flatten())  # Flatten the previous output before sending to fully connected layers"
      ],
      "execution_count": 36,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iLdtzybmj7Nr"
      },
      "source": [
        "cnn3_model.add(Dense(128, activation='relu'))\n",
        "cnn3_model.add(Dense(10, activation='softmax'))"
      ],
      "execution_count": 37,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zcuHfQn5j7Nv"
      },
      "source": [
        "# Compile the model\n",
        "cnn3_model.compile(optimizer='adam', \n",
        "              loss='sparse_categorical_crossentropy',\n",
        "              metrics=['accuracy'])"
      ],
      "execution_count": 38,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ROuuFRIvj7N0",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "53e6c880-38b5-4e3d-cf2a-08e319fabda4"
      },
      "source": [
        "# Train the model\n",
        "epochs = 10\n",
        "batch_size = 32\n",
        "validation_split = 0.1\n",
        "history_3 = cnn3_model.fit(X_train, y_train, epochs=epochs, batch_size=batch_size, validation_split = validation_split)"
      ],
      "execution_count": 39,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/10\n",
            "1688/1688 [==============================] - 8s 4ms/step - loss: 0.4818 - accuracy: 0.8236 - val_loss: 0.3524 - val_accuracy: 0.8688\n",
            "Epoch 2/10\n",
            "1688/1688 [==============================] - 7s 4ms/step - loss: 0.3096 - accuracy: 0.8865 - val_loss: 0.2896 - val_accuracy: 0.8958\n",
            "Epoch 3/10\n",
            "1688/1688 [==============================] - 8s 5ms/step - loss: 0.2621 - accuracy: 0.9035 - val_loss: 0.2682 - val_accuracy: 0.9012\n",
            "Epoch 4/10\n",
            "1688/1688 [==============================] - 7s 4ms/step - loss: 0.2307 - accuracy: 0.9137 - val_loss: 0.2586 - val_accuracy: 0.9030\n",
            "Epoch 5/10\n",
            "1688/1688 [==============================] - 7s 4ms/step - loss: 0.2030 - accuracy: 0.9251 - val_loss: 0.2437 - val_accuracy: 0.9112\n",
            "Epoch 6/10\n",
            "1688/1688 [==============================] - 7s 4ms/step - loss: 0.1805 - accuracy: 0.9325 - val_loss: 0.2469 - val_accuracy: 0.9107\n",
            "Epoch 7/10\n",
            "1688/1688 [==============================] - 7s 4ms/step - loss: 0.1583 - accuracy: 0.9404 - val_loss: 0.2562 - val_accuracy: 0.9072\n",
            "Epoch 8/10\n",
            "1688/1688 [==============================] - 7s 4ms/step - loss: 0.1433 - accuracy: 0.9449 - val_loss: 0.2533 - val_accuracy: 0.9093\n",
            "Epoch 9/10\n",
            "1688/1688 [==============================] - 7s 4ms/step - loss: 0.1252 - accuracy: 0.9522 - val_loss: 0.2658 - val_accuracy: 0.9118\n",
            "Epoch 10/10\n",
            "1688/1688 [==============================] - 7s 4ms/step - loss: 0.1113 - accuracy: 0.9576 - val_loss: 0.2849 - val_accuracy: 0.9133\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_C_rWaxUj7N5",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "d5ce1345-9321-4648-e471-78209fb6572f"
      },
      "source": [
        "cnn3_score = cnn3_model.evaluate(X_test, y_test)\n",
        "print('Test loss:', cnn3_score[0])\n",
        "print('Test accuracy:', cnn3_score[1])"
      ],
      "execution_count": 40,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "313/313 [==============================] - 1s 3ms/step - loss: 0.2879 - accuracy: 0.9108\n",
            "Test loss: 0.28788453340530396\n",
            "Test accuracy: 0.9107999801635742\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RUeiwKS7oEez",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "ad69a3cf-80c5-4342-b9d2-60ff0ae4238a"
      },
      "source": [
        "# Visualize the training curve\n",
        "plt.plot(history_3.history['loss'])\n",
        "plt.plot(history_3.history['val_loss'])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'validation'], loc='upper right')\n",
        "plt.show()"
      ],
      "execution_count": 41,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "hvzA59d2o9CT"
      },
      "source": [
        "Based on the leanring curve, it seems after first epoch the model starts to overfit.  \n",
        "In such scenarios, **dropout** is used to mitigate the effect of overfitting. "
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FRUZXaROpRza"
      },
      "source": [
        "### 3-CNN with Dropout"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "uH-eSJlMpVZZ"
      },
      "source": [
        "Dropout aims to solve the significant challenge of overfitting. It is one of the biggest advancements in deep learning proposed in recent years.\n",
        "\n",
        "The idea is very simple though, it is to randomly drop units in a deep neural network.\n",
        "\n",
        "Learning the relationship between the inputs and the outputs of a dataset is a very complicated procedure. If you have a very small dataset, the relationship maybe a result of noise in the input sample.\n",
        "\n",
        "Dropout refers to randomly and temporary removing a unit, either in a hidden or a visible layer, and all of its incoming and outgoing connections.\n",
        "\n",
        "Further reading on dropout: \n",
        "\n",
        "\n",
        "*   [Machine Learning Mastery](https://machinelearningmastery.com/dropout-for-regularizing-deep-neural-networks/)\n",
        "*   [Marko Jerkic](https://markojerkic.com/what-is-dropout-deep-learning/)\n",
        "\n",
        "![alt text](https://i1.wp.com/cdn-images-1.medium.com/max/800/1*iWQzxhVlvadk6VAJjsgXgg.png?resize=800%2C398&ssl=1)\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "901wJ7o3p9qF"
      },
      "source": [
        "Let's now attempt the same 3-CNN model with **dropout**."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SZcyXqU8qfqK"
      },
      "source": [
        "# Load dropout from keras library\n",
        "from keras.layers import Dropout"
      ],
      "execution_count": 42,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cNxRpV75oqtR"
      },
      "source": [
        "# Updated model layers\n",
        "cnn3_dp_model = Sequential()"
      ],
      "execution_count": 43,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3_VjK_zla_39"
      },
      "source": [
        "Now in each layer, we will add a dropout. The dropout rate is a hyper-parameter we could select based on our model requirements."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6IdzgRgBqOJD"
      },
      "source": [
        "# CNN Layer 1\n",
        "cnn3_dp_model.add(Conv2D(32, kernel_size=(3, 3), activation='relu', input_shape=input_shape))\n",
        "cnn3_dp_model.add(MaxPooling2D((2, 2)))\n",
        "cnn3_dp_model.add(Dropout(0.3))\n",
        "\n",
        "# CNN Layer 2\n",
        "cnn3_dp_model.add(Conv2D(64, kernel_size=(3, 3), activation='relu'))\n",
        "cnn3_dp_model.add(MaxPooling2D(pool_size=(2, 2)))\n",
        "cnn3_dp_model.add(Dropout(0.3))\n",
        "\n",
        "# CNN Layer 3\n",
        "cnn3_dp_model.add(Conv2D(128, kernel_size=(3, 3), activation='relu'))\n",
        "cnn3_dp_model.add(Dropout(0.3))"
      ],
      "execution_count": 44,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "H_ce3upTqOJH"
      },
      "source": [
        "cnn3_dp_model.add(Flatten())  # Flatten the previous output before sending to fully connected layers"
      ],
      "execution_count": 45,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "98D6ImkWqOJJ"
      },
      "source": [
        "cnn3_dp_model.add(Dense(64, activation='relu'))\n",
        "cnn3_dp_model.add(Dropout(0.3))\n",
        "cnn3_dp_model.add(Dense(10, activation='softmax'))"
      ],
      "execution_count": 46,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iG4s4WqkqOJM"
      },
      "source": [
        "# Compile the model\n",
        "cnn3_dp_model.compile(optimizer='adam', \n",
        "              loss='sparse_categorical_crossentropy',\n",
        "              metrics=['accuracy'])"
      ],
      "execution_count": 47,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NUOt0HYGqOJQ",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "f1ba227e-a563-4a38-c2c3-f2c93eb893d3"
      },
      "source": [
        "# Train the model\n",
        "epochs = 15\n",
        "batch_size = 32\n",
        "validation_split = 0.1\n",
        "history_3_dp = cnn3_dp_model.fit(X_train, y_train, epochs=epochs, batch_size=batch_size, validation_split = validation_split)"
      ],
      "execution_count": 48,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/15\n",
            "1688/1688 [==============================] - 8s 5ms/step - loss: 0.6514 - accuracy: 0.7583 - val_loss: 0.3920 - val_accuracy: 0.8477\n",
            "Epoch 2/15\n",
            "1688/1688 [==============================] - 7s 4ms/step - loss: 0.4387 - accuracy: 0.8380 - val_loss: 0.3444 - val_accuracy: 0.8692\n",
            "Epoch 3/15\n",
            "1688/1688 [==============================] - 8s 4ms/step - loss: 0.3828 - accuracy: 0.8597 - val_loss: 0.3115 - val_accuracy: 0.8805\n",
            "Epoch 4/15\n",
            "1688/1688 [==============================] - 7s 4ms/step - loss: 0.3530 - accuracy: 0.8714 - val_loss: 0.2802 - val_accuracy: 0.8968\n",
            "Epoch 5/15\n",
            "1688/1688 [==============================] - 7s 4ms/step - loss: 0.3343 - accuracy: 0.8772 - val_loss: 0.2741 - val_accuracy: 0.8975\n",
            "Epoch 6/15\n",
            "1688/1688 [==============================] - 8s 5ms/step - loss: 0.3206 - accuracy: 0.8829 - val_loss: 0.2657 - val_accuracy: 0.9017\n",
            "Epoch 7/15\n",
            "1688/1688 [==============================] - 7s 4ms/step - loss: 0.3099 - accuracy: 0.8856 - val_loss: 0.2605 - val_accuracy: 0.9035\n",
            "Epoch 8/15\n",
            "1688/1688 [==============================] - 7s 4ms/step - loss: 0.3006 - accuracy: 0.8906 - val_loss: 0.2531 - val_accuracy: 0.9037\n",
            "Epoch 9/15\n",
            "1688/1688 [==============================] - 8s 4ms/step - loss: 0.2974 - accuracy: 0.8920 - val_loss: 0.2507 - val_accuracy: 0.9087\n",
            "Epoch 10/15\n",
            "1688/1688 [==============================] - 8s 5ms/step - loss: 0.2912 - accuracy: 0.8926 - val_loss: 0.2548 - val_accuracy: 0.9068\n",
            "Epoch 11/15\n",
            "1688/1688 [==============================] - 7s 4ms/step - loss: 0.2851 - accuracy: 0.8944 - val_loss: 0.2493 - val_accuracy: 0.9060\n",
            "Epoch 12/15\n",
            "1688/1688 [==============================] - 8s 5ms/step - loss: 0.2823 - accuracy: 0.8953 - val_loss: 0.2574 - val_accuracy: 0.9013\n",
            "Epoch 13/15\n",
            "1688/1688 [==============================] - 7s 4ms/step - loss: 0.2788 - accuracy: 0.8966 - val_loss: 0.2440 - val_accuracy: 0.9082\n",
            "Epoch 14/15\n",
            "1688/1688 [==============================] - 8s 4ms/step - loss: 0.2737 - accuracy: 0.8994 - val_loss: 0.2368 - val_accuracy: 0.9130\n",
            "Epoch 15/15\n",
            "1688/1688 [==============================] - 8s 5ms/step - loss: 0.2724 - accuracy: 0.8992 - val_loss: 0.2476 - val_accuracy: 0.9065\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Q7Vln5efqOJU",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "87962dbe-42e8-4653-b085-562b5c98450a"
      },
      "source": [
        "cnn3_dp_score = cnn3_dp_model.evaluate(X_test, y_test)\n",
        "print('Test loss:', cnn3_dp_score[0])\n",
        "print('Test accuracy:', cnn3_dp_score[1])"
      ],
      "execution_count": 49,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "313/313 [==============================] - 1s 3ms/step - loss: 0.2607 - accuracy: 0.9041\n",
            "Test loss: 0.2607325315475464\n",
            "Test accuracy: 0.9041000008583069\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "p9aGtsRuqOJY",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "56ef8e68-7a77-413a-ae26-8559468a96a9"
      },
      "source": [
        "# Visualize the training curve\n",
        "plt.plot(history_3_dp.history['loss'])\n",
        "plt.plot(history_3_dp.history['val_loss'])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'validation'], loc='upper right')\n",
        "plt.show()"
      ],
      "execution_count": 50,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "t4Jyfotz9B_8",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 295
        },
        "outputId": "209eb406-92c7-4441-a699-7ab360cef288"
      },
      "source": [
        "plt.plot(history_3_dp.history['accuracy'])\n",
        "plt.plot(history_3_dp.history['val_accuracy'])\n",
        "plt.title('model accuracy')\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'validation'], loc='upper right')\n",
        "plt.show()"
      ],
      "execution_count": 51,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "dhPwcElqt1Df"
      },
      "source": [
        "We can now see the overfitting problem has been solved by adding dropout.  \n",
        "In order to improve accuracy, we could train the model for longer - more epochs (Hint: Try 20, 30, 40, 50, etc.)"
      ]
    }
  ]
}