{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "anaconda-cloud": {},
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "file_extension": ".py",
      "mimetype": "text/x-python",
      "name": "python",
      "nbconvert_exporter": "python",
      "pygments_lexer": "ipython3",
      "version": "3.7.1"
    },
    "colab": {
      "name": "deep_learning.ipynb",
      "provenance": []
    },
    "accelerator": "TPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-ta7ICN6mdv8",
        "colab_type": "text"
      },
      "source": [
        "# Introduction to deep learning with `keras` (and `tensorflow` backend)\n",
        "\n",
        "For this notebook you need to install:\n",
        "\n",
        "- `tensorflow`\n",
        "- `keras`\n",
        "- `keras-vis`\n",
        "\n",
        "The easiest way is to first install the `anaconda` distribution for `python` (preferably `Python 3`) and then use in a terminal\n",
        "\n",
        "    conda install pip\n",
        " \n",
        "`pip` is the python package installation tool. Then, you can simply use\n",
        "\n",
        "    pip install tensorflow\n",
        "    pip install keras\n",
        "    pip install keras-vis\n",
        "\n",
        "which should install everything easily."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RdL139QbmiCa",
        "colab_type": "code",
        "outputId": "f4fcb6e8-ce1c-4598-9c00-b3880b7bb62a",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 510
        }
      },
      "source": [
        "# A executer sous Google Colab uniquement\n",
        "!pip install git+https://github.com/raghakot/keras-vis.git"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Collecting git+https://github.com/raghakot/keras-vis.git\n",
            "  Cloning https://github.com/raghakot/keras-vis.git to /tmp/pip-req-build-mw8dipsy\n",
            "  Running command git clone -q https://github.com/raghakot/keras-vis.git /tmp/pip-req-build-mw8dipsy\n",
            "Requirement already satisfied (use --upgrade to upgrade): keras-vis==0.5.0 from git+https://github.com/raghakot/keras-vis.git in /usr/local/lib/python3.6/dist-packages\n",
            "Requirement already satisfied: keras>=2.0 in /usr/local/lib/python3.6/dist-packages (from keras-vis==0.5.0) (2.2.5)\n",
            "Requirement already satisfied: six in /usr/local/lib/python3.6/dist-packages (from keras-vis==0.5.0) (1.12.0)\n",
            "Requirement already satisfied: scikit-image in /usr/local/lib/python3.6/dist-packages (from keras-vis==0.5.0) (0.16.2)\n",
            "Requirement already satisfied: matplotlib in /usr/local/lib/python3.6/dist-packages (from keras-vis==0.5.0) (3.1.3)\n",
            "Requirement already satisfied: h5py in /usr/local/lib/python3.6/dist-packages (from keras-vis==0.5.0) (2.8.0)\n",
            "Requirement already satisfied: scipy>=0.14 in /usr/local/lib/python3.6/dist-packages (from keras>=2.0->keras-vis==0.5.0) (1.4.1)\n",
            "Requirement already satisfied: keras-preprocessing>=1.1.0 in /usr/local/lib/python3.6/dist-packages (from keras>=2.0->keras-vis==0.5.0) (1.1.0)\n",
            "Requirement already satisfied: pyyaml in /usr/local/lib/python3.6/dist-packages (from keras>=2.0->keras-vis==0.5.0) (3.13)\n",
            "Requirement already satisfied: keras-applications>=1.0.8 in /usr/local/lib/python3.6/dist-packages (from keras>=2.0->keras-vis==0.5.0) (1.0.8)\n",
            "Requirement already satisfied: numpy>=1.9.1 in /usr/local/lib/python3.6/dist-packages (from keras>=2.0->keras-vis==0.5.0) (1.17.5)\n",
            "Requirement already satisfied: imageio>=2.3.0 in /usr/local/lib/python3.6/dist-packages (from scikit-image->keras-vis==0.5.0) (2.4.1)\n",
            "Requirement already satisfied: PyWavelets>=0.4.0 in /usr/local/lib/python3.6/dist-packages (from scikit-image->keras-vis==0.5.0) (1.1.1)\n",
            "Requirement already satisfied: pillow>=4.3.0 in /usr/local/lib/python3.6/dist-packages (from scikit-image->keras-vis==0.5.0) (6.2.2)\n",
            "Requirement already satisfied: networkx>=2.0 in /usr/local/lib/python3.6/dist-packages (from scikit-image->keras-vis==0.5.0) (2.4)\n",
            "Requirement already satisfied: python-dateutil>=2.1 in /usr/local/lib/python3.6/dist-packages (from matplotlib->keras-vis==0.5.0) (2.6.1)\n",
            "Requirement already satisfied: kiwisolver>=1.0.1 in /usr/local/lib/python3.6/dist-packages (from matplotlib->keras-vis==0.5.0) (1.1.0)\n",
            "Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /usr/local/lib/python3.6/dist-packages (from matplotlib->keras-vis==0.5.0) (2.4.6)\n",
            "Requirement already satisfied: cycler>=0.10 in /usr/local/lib/python3.6/dist-packages (from matplotlib->keras-vis==0.5.0) (0.10.0)\n",
            "Requirement already satisfied: decorator>=4.3.0 in /usr/local/lib/python3.6/dist-packages (from networkx>=2.0->scikit-image->keras-vis==0.5.0) (4.4.1)\n",
            "Requirement already satisfied: setuptools in /usr/local/lib/python3.6/dist-packages (from kiwisolver>=1.0.1->matplotlib->keras-vis==0.5.0) (45.2.0)\n",
            "Building wheels for collected packages: keras-vis\n",
            "  Building wheel for keras-vis (setup.py) ... \u001b[?25l\u001b[?25hdone\n",
            "  Created wheel for keras-vis: filename=keras_vis-0.5.0-py2.py3-none-any.whl size=38886 sha256=f4d2bcdddfb00a1c269b502232b21994814525130a54dc5aee78de6470d2ac21\n",
            "  Stored in directory: /tmp/pip-ephem-wheel-cache-cutqj_0e/wheels/c5/ae/e7/b34d1cb48b1898f606a5cce08ebc9521fa0588f37f1e590d9f\n",
            "Successfully built keras-vis\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "676K1nqoml2M",
        "colab_type": "text"
      },
      "source": [
        "# A executer sous Google Colab uniquement\n",
        "!pip install git+https://github.com/raghakot/keras-vis.git"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HwBL-iLKmdv-",
        "colab_type": "code",
        "outputId": "273852ad-f8ae-453a-aadb-d724857f6e7e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 114
        }
      },
      "source": [
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "%matplotlib inline\n",
        "\n",
        "import pickle as pkl\n",
        "import os\n",
        "\n",
        "import tensorflow\n",
        "import keras\n",
        "from keras import backend as K\n",
        "from keras.datasets import mnist\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Dense, Dropout, Flatten\n",
        "from keras.layers import Conv2D, MaxPooling2D\n",
        "from keras import activations\n",
        "\n",
        "from vis.visualization import visualize_activation\n",
        "from vis.utils import utils\n",
        "\n",
        "print(\"tensorflow version %s (should be at least 0.12.1)\" % tensorflow.__version__)\n",
        "print(\"keras version %s (should be at least 2.0.7)\" % keras.__version__)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<p style=\"color: red;\">\n",
              "The default version of TensorFlow in Colab will soon switch to TensorFlow 2.x.<br>\n",
              "We recommend you <a href=\"https://www.tensorflow.org/guide/migrate\" target=\"_blank\">upgrade</a> now \n",
              "or ensure your notebook will continue to use TensorFlow 1.x via the <code>%tensorflow_version 1.x</code> magic:\n",
              "<a href=\"https://colab.research.google.com/notebooks/tensorflow_version.ipynb\" target=\"_blank\">more info</a>.</p>\n"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Using TensorFlow backend.\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "tensorflow version 1.15.0 (should be at least 0.12.1)\n",
            "keras version 2.2.5 (should be at least 2.0.7)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7gXzbNdzmdwB",
        "colab_type": "text"
      },
      "source": [
        "# 0. Introduction: about `tensorflow` and `keras`\n",
        "\n",
        "The `numpy` library does some expensive operations outside Python using efficient code (Fortran, C/C++). However, switching back to python after each operation cause a big overhead because of unnecessary copies of the data. \n",
        "\n",
        "The library `tensorflow` does all the computations outside of Python: the python API is used to define a graph of operations, that will run entirely using C++ binaries. This architecture allows to get rid of the overhead. Besides, knowing the computational graph beforehand allows to parallelize and/or distribute the computation more easily. As a result, `tensoflow` can run the computations on multiple CPUs or GPUs, and on multiple servers.\n",
        "\n",
        "However, for quick an easy model prototying, the library `keras` is simpler to use than `tensorflow`. \n",
        "Deep learning models can be constructed thanks to `keras` in few lines of python. So in this notebook, we won't see direct calls to `tensorflow`, but only to `keras`, even if the computations are done by `tensorflow`.\n",
        "\n",
        "# 1. Handwritten digit recognition with MNIST\n",
        "\n",
        "For the first part of this tutorial, we will use the [MNIST](http://yann.lecun.com/exdb/mnist) dataset.\n",
        "This dataset contains images representing handwritten digits. \n",
        "Each image is 28 x 28 pixels, and each pixel is represented by a number (gray level). \n",
        "These arrays can be flattened into vectors of 28 x 28 = 784 numbers.\n",
        "You can then see each image as a point in a 784-dimensional vector space. \n",
        "You can find interesting visualisations of this vector space [http://colah.github.io/posts/2014-10-Visualizing-MNIST/](http://colah.github.io/posts/2014-10-Visualizing-MNIST/).\n",
        "\n",
        "## 1.1. Introduction\n",
        "\n",
        "The labels in $\\{0, 1, 2, \\ldots, 9\\}$ giving the digit on the image are be represented using one-hot encoding: labels in $\\{0, 1, 2, \\ldots, 9\\}$ are replaced by labels in $\\{ 0, 1\\}^{10}$, namely $0$ is replaced by $(1, 0, \\ldots 0)$, $1$ is replaced by $(0, 1, 0, \\ldots 0)$, $2$ is replaced by $(0, 0, 1, 0, \\ldots, 0)$, etc.\n",
        "\n",
        "Also, MNIST data is grayscale pixels in $\\{0, \\ldots, 255\\}$. The pixels should be normalized to belong to $[0, 1]$.\n",
        "Indeed, working with big floats can lead to important numerical errors, in particular in deep learning models.\n",
        "To convince yourself of the damages that can result from numerical instability, see the following cell:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QLA5mDFwmdwB",
        "colab_type": "code",
        "outputId": "79d33e8f-d71e-4ec1-d2f1-39dc1d9e1091",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        }
      },
      "source": [
        "foo = 1e9\n",
        "for i in range(int(1e6)):\n",
        "    foo += 1e-6\n",
        "foo -= 1e9\n",
        "print(\"Numerically instable sum: %f \" % foo)\n",
        "\n",
        "bar = 1\n",
        "for i in range(int(1e6)):\n",
        "    bar += 1e-6\n",
        "bar -= 1\n",
        "print(\"Numerically stable sum: %f \" % bar)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Numerically instable sum: 0.953674 \n",
            "Numerically stable sum: 1.000000 \n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "K2Pm7Zy4mdwE",
        "colab_type": "text"
      },
      "source": [
        "## 1.2. Load the data\n",
        "\n",
        "MNIST is a very old and standard benchmark dataset for image classification, so it's built-in (ready to be downloaded) in all machine learning libraries (including `keras` and `tensorflow`)."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "scrolled": true,
        "id": "4sD8o7AzmdwF",
        "colab_type": "code",
        "outputId": "b607564f-d9c0-43f6-e723-dc3a8778ca24",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 119
        }
      },
      "source": [
        "# Number of classes\n",
        "num_classes = 10\n",
        "# input image dimensions\n",
        "img_rows, img_cols = 28, 28\n",
        "\n",
        "# the data, shuffled and split between train and test sets\n",
        "(x_train, y_train), (x_test, y_test) = mnist.load_data()\n",
        "\n",
        "if K.image_data_format() == 'channels_first':\n",
        "    x_train = x_train.reshape(x_train.shape[0], 1, img_rows, img_cols)\n",
        "    x_test = x_test.reshape(x_test.shape[0], 1, img_rows, img_cols)\n",
        "    input_shape = (1, img_rows, img_cols)\n",
        "else:\n",
        "    x_train = x_train.reshape(x_train.shape[0], img_rows, img_cols, 1)\n",
        "    x_test = x_test.reshape(x_test.shape[0], img_rows, img_cols, 1)\n",
        "    input_shape = (img_rows, img_cols, 1)\n",
        "\n",
        "x_train = x_train.astype('float32')\n",
        "x_test = x_test.astype('float32')\n",
        "\n",
        "print('x_train shape:', x_train.shape)\n",
        "print('x_test shape:', x_test.shape)\n",
        "print('y_train shape:', y_train.shape)\n",
        "print('y_test shape:', y_test.shape)\n",
        "\n",
        "print(x_train.shape[0], 'train samples')\n",
        "print(x_test.shape[0], 'test samples')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "x_train shape: (60000, 28, 28, 1)\n",
            "x_test shape: (10000, 28, 28, 1)\n",
            "y_train shape: (60000,)\n",
            "y_test shape: (10000,)\n",
            "60000 train samples\n",
            "10000 test samples\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PZLN3KS-mdwH",
        "colab_type": "text"
      },
      "source": [
        "## 1.3. A first look at the data\n",
        "\n",
        "In the next cell we illustrate the first for elements of the training data: \n",
        "pixels grayscale of the digit and their corresponding label."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "scrolled": true,
        "id": "CS9ayifUmdwI",
        "colab_type": "code",
        "outputId": "989cf10d-836e-4727-e3fb-ae1d7e3bbf9b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 159
        }
      },
      "source": [
        "plt.figure(figsize=(8, 2))\n",
        "for i in range(4):\n",
        "    plt.subplot(1, 4, i+1)\n",
        "    plt.imshow(x_train[i].reshape(28, 28), \n",
        "               interpolation=\"none\", cmap=\"gray_r\")\n",
        "    plt.title('Label=%d' % y_train[i], fontsize=14)\n",
        "    plt.axis(\"off\")\n",
        "plt.tight_layout()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAiUAAACOCAYAAADuKXjvAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 576x144 with 4 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "scrolled": true,
        "id": "LkvErln1mdwL",
        "colab_type": "code",
        "outputId": "5b9d7cf7-e376-4af7-b4ad-da731ad081b0",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 297
        }
      },
      "source": [
        "n_rows = 4\n",
        "n_cols = 8\n",
        "plt.figure(figsize=(8, 4))\n",
        "for i in range(n_rows * n_cols):\n",
        "    plt.subplot(n_rows, n_cols, i+1)\n",
        "    plt.imshow(x_train[i].reshape(28, 28),\n",
        "               interpolation=\"none\", cmap=\"gray_r\")\n",
        "    plt.axis(\"off\")\n",
        "plt.tight_layout()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAjcAAAEYCAYAAACgOtfQAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nmaWKT27MzMwsVf4PQi9PgS/eOu0AAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 576x288 with 32 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "H5hN1dK8mdwO",
        "colab_type": "text"
      },
      "source": [
        "The first character is a 5 digit, encoded in grayscale matrix as follows"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "uCOchBb3mdwO",
        "colab_type": "code",
        "outputId": "70792c0a-5d36-4b83-8cce-7a3949caac63",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 493
        }
      },
      "source": [
        "print(np.array2string(x_train[0].astype(np.int).reshape(28, 28), \n",
        "                      max_line_width=150))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[[  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   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   0   0   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   0   0   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   0   0   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   0   0   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   3  18  18  18 126 136 175  26 166 255 247 127   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0  30  36  94 154 170 253 253 253 253 253 225 172 253 242 195  64   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0  49 238 253 253 253 253 253 253 253 253 251  93  82  82  56  39   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0  18 219 253 253 253 253 253 198 182 247 241   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0  80 156 107 253 253 205  11   0  43 154   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0  14   1 154 253  90   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0 139 253 190   2   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  11 190 253  70   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  35 241 225 160 108   1   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0   0   0  81 240 253 253 119  25   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0  45 186 253 253 150  27   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0  16  93 252 253 187   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0 249 253 249  64   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0  46 130 183 253 253 207   2   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0   0   0  39 148 229 253 253 253 250 182   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0   0   0  24 114 221 253 253 253 253 201  78   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0   0   0  23  66 213 253 253 253 253 198  81   2   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0   0   0  18 171 219 253 253 253 253 195  80   9   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0  55 172 226 253 253 253 253 244 133  11   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
            " [  0   0   0   0 136 253 253 253 212 135 132  16   0   0   0   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   0   0   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   0   0   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   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0]]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XwJAMaEemdwR",
        "colab_type": "text"
      },
      "source": [
        "## 1.4. Normalization and preprocessing of the data\n",
        "\n",
        "We need to normalize the images and one-hot encode the labels.\n",
        "\n",
        "**Warning:** call this cell only once (otherwise you'll devide several times by 255), which might be problematic later on."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_56iIXO5mdwR",
        "colab_type": "code",
        "outputId": "bb2761dd-35eb-4f32-d9dc-03c1f9840020",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        }
      },
      "source": [
        "x_train /= 255\n",
        "x_test /= 255\n",
        "print(x_train.min(axis=None), x_train.max(axis=None))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "0.0 1.0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "EIBLhezimdwT",
        "colab_type": "code",
        "outputId": "8a74276d-8232-4036-a600-a5d116a17088",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 187
        }
      },
      "source": [
        "# convert class vectors to binary class matrices\n",
        "y_train = keras.utils.to_categorical(y_train, num_classes)\n",
        "y_test = keras.utils.to_categorical(y_test, num_classes)\n",
        "y_train[:10]"
      ],
      "execution_count": 0,
      "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": 9
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "33MnZMzxmdwW",
        "colab_type": "text"
      },
      "source": [
        "<a id='softmax'></a>\n",
        "# 2. Softmax regression for MNIST\n",
        "\n",
        "## 2.1. Description of the model\n",
        "\n",
        "We want to classify these pictures to predict the digit $k$ they represent for $k \\in \\{0, \\ldots, 9\\}$.\n",
        "A simple model allowing to do that is softmax regression.\n",
        "\n",
        "The idea behind this model is to produce a score for each input image $x$ using a simple linear model. \n",
        "To do so, we assume that belonging to a class $k$ (corresponding to digit $k$) can be expressed by a weigthed sum of the pixel intensities, with weights $W_{k, 1}, \\ldots, W_{k, 784}$ and to a bias $b_k$ capturing variability independent of the input:\n",
        "$$\n",
        "\\text{score}_k(x_i) = \\sum_{j=1}^{784} W_{k, j} x_j + b_k,\n",
        "$$\n",
        "These scores are sometimes called the \"logits\" in the deep learning community.\n",
        "We then use the softmax function to convert the scores into predicted probabilities $p_k$:\n",
        "$$\n",
        "p_k(x_i) = \\text{softmax}(\\text{score}_k(x_i)) = \\frac{\\exp(\\text{score}_k(x_i))}{\\sum_{k'=1}^{10}\\exp(\\text{score}_{k'}(x_i))}\n",
        "$$\n",
        "for $k=1, \\ldots, 10$."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DNdUwo3rmdwW",
        "colab_type": "text"
      },
      "source": [
        "## 2.2. The computational graph for training of softmax regression\n",
        "\n",
        "To train the model weights $W_{k, j}$ and $b_k$ for $k=1, \\ldots, 10$ and $j=1, \\ldots, 784$, we consider the goodness-of-fit given by the negative log-likelihood.\n",
        "For the considered model, the negative log-likelihood of a sample with input $x_i \\in \\mathbb R^{784}$ and label $y_i \\in \\{0, 1\\}^{10}$, it is given by the cross-entropy between the scores $p_k(x_i)$ and the label $y_i$:\n",
        "$$\n",
        "- \\sum_{k=1}^{10} y_{i, k} \\log(p_k(x_i))\n",
        "$$\n",
        "For this first model, we will simply use stochastic gradient descent over small batches of data. It can be done easily with TensorFlow, as it will automatically and efficiently compute the gradient from your graph, then apply an optimization algorithm of your choice to perform the parameters update."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "scrolled": true,
        "id": "qhB7r_QamdwX",
        "colab_type": "code",
        "outputId": "47a1178e-170e-484b-fdb8-cc1cb6c21244",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 391
        }
      },
      "source": [
        "# We use a sequential model: we stack layers\n",
        "model = Sequential()\n",
        "# First we need to flatten the data: replace 28 * 28 matrices by flat vectors\n",
        "# This is always necessary before feeding data to a fully-connected layer (Dense object)\n",
        "model.add(Flatten(input_shape=input_shape, name='flatten'))\n",
        "# We add one dense (fully connected layer) with softmax activation function\n",
        "# Since it's the first layer, we need to give the size of input data\n",
        "model.add(Dense(num_classes, activation='softmax', name='dense_softmax'))\n",
        "\n",
        "# We \"compile\" this model, \n",
        "model.compile(\n",
        "    # specifying the loss as the cross-entropy\n",
        "    loss=keras.losses.categorical_crossentropy,\n",
        "    # We choose the Adagrad solver, but you can choose others\n",
        "    optimizer=keras.optimizers.Adagrad(),\n",
        "    # We will monitor the accuracy on a testing set along optimization\n",
        "    metrics=['accuracy']\n",
        ")\n",
        "model.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:66: The name tf.get_default_graph is deprecated. Please use tf.compat.v1.get_default_graph instead.\n",
            "\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:541: The name tf.placeholder is deprecated. Please use tf.compat.v1.placeholder instead.\n",
            "\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:4432: The name tf.random_uniform is deprecated. Please use tf.random.uniform instead.\n",
            "\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/optimizers.py:793: The name tf.train.Optimizer is deprecated. Please use tf.compat.v1.train.Optimizer instead.\n",
            "\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:3576: The name tf.log is deprecated. Please use tf.math.log instead.\n",
            "\n",
            "Model: \"sequential_1\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "flatten (Flatten)            (None, 784)               0         \n",
            "_________________________________________________________________\n",
            "dense_softmax (Dense)        (None, 10)                7850      \n",
            "=================================================================\n",
            "Total params: 7,850\n",
            "Trainable params: 7,850\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TFZ5o8yLmdwZ",
        "colab_type": "text"
      },
      "source": [
        "## 2.3. Run the training of the model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "scrolled": false,
        "id": "KoO3zv30mdwa",
        "colab_type": "code",
        "outputId": "f6508c6e-b4da-483e-ee35-ba6a93eaf17d",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 731
        }
      },
      "source": [
        "batch_size = 64\n",
        "epochs = 10\n",
        "\n",
        "# Run the train\n",
        "history = model.fit(x_train, y_train,\n",
        "                    batch_size=batch_size,\n",
        "                    epochs=epochs,\n",
        "                    verbose=1,\n",
        "                    validation_data=(x_test, y_test))\n",
        "score = model.evaluate(x_test, y_test, verbose=0)\n",
        "print('Test loss:', score[0])\n",
        "print('Test accuracy:', score[1])"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/tensorflow_core/python/ops/math_grad.py:1424: where (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.\n",
            "Instructions for updating:\n",
            "Use tf.where in 2.0, which has the same broadcast rule as np.where\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:1033: The name tf.assign_add is deprecated. Please use tf.compat.v1.assign_add instead.\n",
            "\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:1020: The name tf.assign is deprecated. Please use tf.compat.v1.assign instead.\n",
            "\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:3005: The name tf.Session is deprecated. Please use tf.compat.v1.Session instead.\n",
            "\n",
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/10\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:190: The name tf.get_default_session is deprecated. Please use tf.compat.v1.get_default_session instead.\n",
            "\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:197: The name tf.ConfigProto is deprecated. Please use tf.compat.v1.ConfigProto instead.\n",
            "\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:207: The name tf.global_variables is deprecated. Please use tf.compat.v1.global_variables instead.\n",
            "\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:216: The name tf.is_variable_initialized is deprecated. Please use tf.compat.v1.is_variable_initialized instead.\n",
            "\n",
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:223: The name tf.variables_initializer is deprecated. Please use tf.compat.v1.variables_initializer instead.\n",
            "\n",
            "60000/60000 [==============================] - 2s 26us/step - loss: 0.4935 - acc: 0.8752 - val_loss: 0.3675 - val_acc: 0.9034\n",
            "Epoch 2/10\n",
            "60000/60000 [==============================] - 1s 24us/step - loss: 0.3622 - acc: 0.9024 - val_loss: 0.3329 - val_acc: 0.9110\n",
            "Epoch 3/10\n",
            "60000/60000 [==============================] - 1s 24us/step - loss: 0.3379 - acc: 0.9077 - val_loss: 0.3187 - val_acc: 0.9131\n",
            "Epoch 4/10\n",
            "60000/60000 [==============================] - 1s 24us/step - loss: 0.3249 - acc: 0.9110 - val_loss: 0.3090 - val_acc: 0.9145\n",
            "Epoch 5/10\n",
            "60000/60000 [==============================] - 1s 24us/step - loss: 0.3164 - acc: 0.9131 - val_loss: 0.3028 - val_acc: 0.9156\n",
            "Epoch 6/10\n",
            "60000/60000 [==============================] - 1s 24us/step - loss: 0.3101 - acc: 0.9148 - val_loss: 0.2990 - val_acc: 0.9168\n",
            "Epoch 7/10\n",
            "60000/60000 [==============================] - 1s 24us/step - loss: 0.3054 - acc: 0.9157 - val_loss: 0.2957 - val_acc: 0.9162\n",
            "Epoch 8/10\n",
            "60000/60000 [==============================] - 1s 24us/step - loss: 0.3015 - acc: 0.9166 - val_loss: 0.2921 - val_acc: 0.9175\n",
            "Epoch 9/10\n",
            "60000/60000 [==============================] - 1s 24us/step - loss: 0.2983 - acc: 0.9175 - val_loss: 0.2899 - val_acc: 0.9180\n",
            "Epoch 10/10\n",
            "60000/60000 [==============================] - 1s 24us/step - loss: 0.2956 - acc: 0.9182 - val_loss: 0.2885 - val_acc: 0.9180\n",
            "Test loss: 0.2884747538626194\n",
            "Test accuracy: 0.918\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "scrolled": false,
        "id": "D-b4i24nmdwc",
        "colab_type": "code",
        "outputId": "9b2f9b71-8ff8-4ddc-b7f8-08adcf6d24ac",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 369
        }
      },
      "source": [
        "plt.figure(figsize=(7, 5))\n",
        "plt.plot(history.epoch, history.history['acc'], lw=3, label='Training')\n",
        "plt.plot(history.epoch, history.history['val_acc'], lw=3, label='Testing')\n",
        "plt.legend(fontsize=14)\n",
        "plt.title('Accuracy of softmax regression', fontsize=16)\n",
        "plt.xlabel('Epoch', fontsize=14)\n",
        "plt.ylabel('Accuracy', fontsize=14)\n",
        "plt.tight_layout()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfAAAAFgCAYAAABEyiulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 504x360 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pxu_8V-0mdwe",
        "colab_type": "text"
      },
      "source": [
        "**QUESTION**\n",
        "\n",
        "- Run 70 epochs and look at the training and testing accuracy curves"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "M7mm0veSmdwf",
        "colab_type": "text"
      },
      "source": [
        "## 2.4. Visualisation of the model weights\n",
        "\n",
        "Weight matrices plots show that the learned weights are consistant with the digits they should predict (see below).\n",
        "You should be able to see rough shapes corresponding to the digits 0, 1, 2, 3, etc."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ug98LYD7mdwg",
        "colab_type": "code",
        "outputId": "8a9b09ea-fac6-454a-b04d-71ac5ccf6960",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 346
        }
      },
      "source": [
        "weights, biases = model.get_layer('dense_softmax').get_weights()\n",
        "imgs = weights.reshape(28, 28, 10)\n",
        "\n",
        "fig = plt.figure(figsize=(10, 5))\n",
        "vmin, vmax = imgs.min(), imgs.max()\n",
        "for i in range(10):\n",
        "    ax = plt.subplot(2, 5, i + 1)\n",
        "    im = imgs[:, :, i]\n",
        "    mappable = ax.imshow(im, interpolation=\"nearest\", \n",
        "                         vmin=vmin, vmax=vmax, cmap='bwr')\n",
        "    ax.axis('off')\n",
        "    ax.set_title(\"%i\" % i)\n",
        "plt.tight_layout()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAsgAAAFJCAYAAACGr9uWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nG2OMMcYYU+AfwuyHOXIY9sYAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 720x360 with 10 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "E9UPWJ2nmdwi",
        "colab_type": "text"
      },
      "source": [
        "## 2.5. Conclusion with MNIST\n",
        "\n",
        "You should have reached an accuracy better than 0.9 with this simple model. \n",
        "**Too easy !** You almost solved the problem using a simple softmax regression. \n",
        "Weight matrices plots show that the learned weights are consistant with the digits they should predict (see below).\n",
        "You should be able to see rough shapes corresponding to the digits 0, 1, 2, 3, etc.\n",
        "\n",
        "# 3. Towards 0.99 accuracy on MNIST\n",
        "\n",
        "Now, let's build a better model for MNIST using more layers. \n",
        "Let's start with a feed-forward neural net (FFNN) with one hidden layer and relu activation.\n",
        "\n",
        "## 3.1. Feed-Forward Neural Network (FFNN)\n",
        "\n",
        "The softmax regression you just trained is a linear model, with 7850 parameters. \n",
        "It is easy to fit, numerically stable, but might be too simple for some tasks. \n",
        "The idea behind neural networks is to have a nonlinear model, while keeping the nice features of linear ones. \n",
        "The idea is to keep parameters into linear functions, and link these small linear model using non linear operations.\n",
        "\n",
        "A simple nonlinearity which is often used to do this is the rectified linear unit\n",
        "$\\text{ReLU}(x) = \\max(0, x)$\n",
        "\n",
        "The derivative of this function is very easy to compute, and it is parameter-free. If we stack models such as softmax regression and ReLUs, it is still very easy to compute the gradient using the chain rule, as the model is a combination of simple functions.\n",
        "\n",
        "The backpropagation algorithm allows efficient computation of complex derivatives as long as the function is made of simple blocks with simple derivatives. \n",
        "This algorithm efficiency is based on data reuse: when working with parallel architectures such as GPUs, you want to minimize communication (data transfer) as it is very time consuming in comparison to the computing time.\n",
        "\n",
        "**QUESTION**\n",
        "\n",
        "- Create the graph for a fully connected feed-forward neural network with one hidden layer with 128 units and a relu activation function. Use what you did for softmax regression (you need to add a **single line** to the code creating the softmax regression)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wIW0XCovmdwj",
        "colab_type": "code",
        "outputId": "91a56423-dd4b-4c46-86ad-cba0a468140b",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 221
        }
      },
      "source": [
        "model_ffnn = Sequential()\n",
        "\n",
        "model_ffnn.add(Flatten(input_shape=input_shape))\n",
        "\n",
        "## TODO : add a line here that adds the extra layer\n",
        "\n",
        "model_ffnn.add(Dense(num_classes, activation='softmax'))\n",
        "model_ffnn.compile(\n",
        "    loss=keras.losses.categorical_crossentropy,\n",
        "    optimizer=keras.optimizers.Adagrad(),\n",
        "    metrics=['accuracy']\n",
        ")\n",
        "\n",
        "model_ffnn.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_2\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "flatten_1 (Flatten)          (None, 784)               0         \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 10)                7850      \n",
            "=================================================================\n",
            "Total params: 7,850\n",
            "Trainable params: 7,850\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "scrolled": false,
        "id": "k370b0X8mdwl",
        "colab_type": "code",
        "outputId": "362ac784-2d3d-4719-ea55-d78c3c6a45fc",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 408
        }
      },
      "source": [
        "batch_size = 32\n",
        "epochs = 10\n",
        "\n",
        "# Run the train\n",
        "history = model_ffnn.fit(x_train, y_train,\n",
        "                         batch_size=batch_size,\n",
        "                         epochs=epochs,\n",
        "                         verbose=1,\n",
        "                         validation_data=(x_test, y_test))\n",
        "score = model_ffnn.evaluate(x_test, y_test, verbose=0)\n",
        "print('Test loss:', score[0])\n",
        "print('Test accuracy:', score[1])"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 60000 samples, validate on 10000 samples\n",
            "Epoch 1/10\n",
            "60000/60000 [==============================] - 3s 45us/step - loss: 0.4656 - acc: 0.8805 - val_loss: 0.3540 - val_acc: 0.9082\n",
            "Epoch 2/10\n",
            "60000/60000 [==============================] - 3s 43us/step - loss: 0.3531 - acc: 0.9037 - val_loss: 0.3263 - val_acc: 0.9122\n",
            "Epoch 3/10\n",
            "60000/60000 [==============================] - 2s 42us/step - loss: 0.3317 - acc: 0.9089 - val_loss: 0.3128 - val_acc: 0.9153\n",
            "Epoch 4/10\n",
            "60000/60000 [==============================] - 3s 42us/step - loss: 0.3202 - acc: 0.9116 - val_loss: 0.3051 - val_acc: 0.9169\n",
            "Epoch 5/10\n",
            "60000/60000 [==============================] - 3s 42us/step - loss: 0.3124 - acc: 0.9136 - val_loss: 0.2996 - val_acc: 0.9175\n",
            "Epoch 6/10\n",
            "60000/60000 [==============================] - 3s 42us/step - loss: 0.3068 - acc: 0.9151 - val_loss: 0.2958 - val_acc: 0.9196\n",
            "Epoch 7/10\n",
            "60000/60000 [==============================] - 3s 42us/step - loss: 0.3024 - acc: 0.9161 - val_loss: 0.2929 - val_acc: 0.9191\n",
            "Epoch 8/10\n",
            "60000/60000 [==============================] - 3s 42us/step - loss: 0.2989 - acc: 0.9174 - val_loss: 0.2901 - val_acc: 0.9196\n",
            "Epoch 9/10\n",
            "60000/60000 [==============================] - 3s 42us/step - loss: 0.2958 - acc: 0.9177 - val_loss: 0.2883 - val_acc: 0.9208\n",
            "Epoch 10/10\n",
            "60000/60000 [==============================] - 3s 42us/step - loss: 0.2933 - acc: 0.9186 - val_loss: 0.2869 - val_acc: 0.9211\n",
            "Test loss: 0.286909355956316\n",
            "Test accuracy: 0.9211\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "scrolled": false,
        "id": "-wsHJxIomdwn",
        "colab_type": "code",
        "outputId": "ef86d012-2e48-4e6d-d771-99de30910f15",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 369
        }
      },
      "source": [
        "plt.figure(figsize=(7, 5))\n",
        "plt.plot(history.epoch, history.history['acc'], lw=3, label='Training')\n",
        "plt.plot(history.epoch, history.history['val_acc'], lw=3, label='Testing')\n",
        "plt.legend(fontsize=14)\n",
        "plt.title('Accuracy of softmax regression', fontsize=16)\n",
        "plt.xlabel('Epoch', fontsize=14)\n",
        "plt.ylabel('Accuracy', fontsize=14)\n",
        "plt.tight_layout()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfAAAAFgCAYAAABEyiulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 504x360 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "eXCn8menmdwp",
        "colab_type": "text"
      },
      "source": [
        "## 3.2. Saving the model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bWzriBfmmdwq",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "path = '/'\n",
        "model_ffnn.save(os.path.join(path, 'mnist_ffnn.h5'))\n",
        "with open(os.path.join(path, 'mnist_ffnn_history.pkl'), 'wb') as f:\n",
        "    pkl.dump(history.history, f)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "QLE7N1b6mdws",
        "colab_type": "text"
      },
      "source": [
        "## 3.3. Visualize the layer\n",
        "\n",
        "Using the function defined below, we can visualize the units of the layers.\n",
        "It uses the `visualize_activation` function from the `keras-vis` library that we imported at the beginning of this notebook.\n",
        "It is based on the activation maximization principle: it finds the input that maximizes the output activation of a layer, see the documentation of the `keras-vis` library for more details."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4_6QT_gImdwt",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def visualize_layer(model, layer_name):\n",
        "    layer_idx = utils.find_layer_idx(model, layer_name)\n",
        "    model.layers[layer_idx].activation = activations.linear\n",
        "    model = utils.apply_modifications(model)\n",
        "    imgs = [visualize_activation(model, layer_idx, filter_indices=k,\n",
        "                                 input_range=(0., 1.), tv_weight=10,\n",
        "                                 lp_norm_weight=0) for k in range(10)]\n",
        "    vmin = min(img.min(axis=None) for img in imgs)\n",
        "    vmax = max(img.max(axis=None) for img in imgs)\n",
        "    plt.figure(figsize=(10, 8))\n",
        "    for k in range(10):\n",
        "        img = imgs[k]\n",
        "        ax = plt.subplot(3, 4, k + 1)\n",
        "        ax.imshow(img.reshape(28, 28), cmap='bwr', vmin=vmin, vmax=vmax)\n",
        "        plt.title('Percep. of %d' % k, fontsize=16)\n",
        "        plt.axis('off')\n",
        "    plt.tight_layout()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "y11B8xaCmdwv",
        "colab_type": "code",
        "outputId": "ac499331-a1f7-454c-e8fb-3fbc55e032c8",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 221
        }
      },
      "source": [
        "model_ffnn.summary()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_2\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "flatten_1 (Flatten)          (None, 784)               0         \n",
            "_________________________________________________________________\n",
            "dense_1 (Dense)              (None, 10)                7850      \n",
            "=================================================================\n",
            "Total params: 7,850\n",
            "Trainable params: 7,850\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "scrolled": false,
        "id": "uhzwj4M3mdwx",
        "colab_type": "code",
        "outputId": "a68097d8-341f-4c5f-9311-36a02695d7e7",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 623
        }
      },
      "source": [
        "visualize_layer(model_ffnn, 'dense_1')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "WARNING:tensorflow:From /usr/local/lib/python3.6/dist-packages/keras/backend/tensorflow_backend.py:148: The name tf.placeholder_with_default is deprecated. Please use tf.compat.v1.placeholder_with_default instead.\n",
            "\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAscAAAI8CAYAAAAdowb0AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nDAAA5R9yBAsoagBYrAAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 720x576 with 10 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "n5Q3ObyBmdwz",
        "colab_type": "code",
        "outputId": "9362f26f-03d6-4767-b031-705f1f382e5f",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 307
        }
      },
      "source": [
        "visualize_layer(model_ffnn, 'dense_2')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "error",
          "ename": "ValueError",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mValueError\u001b[0m                                Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-21-3b7353a6c877>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mvisualize_layer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel_ffnn\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'dense_2'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
            "\u001b[0;32m<ipython-input-18-e3fd01377c2d>\u001b[0m in \u001b[0;36mvisualize_layer\u001b[0;34m(model, layer_name)\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mvisualize_layer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlayer_name\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 2\u001b[0;31m     \u001b[0mlayer_idx\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mutils\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfind_layer_idx\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlayer_name\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      3\u001b[0m     \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlayers\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mlayer_idx\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mactivation\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mactivations\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mlinear\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m     \u001b[0mmodel\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mutils\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mapply_modifications\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m     imgs = [visualize_activation(model, layer_idx, filter_indices=k,\n",
            "\u001b[0;32m/usr/local/lib/python3.6/dist-packages/vis/utils/utils.py\u001b[0m in \u001b[0;36mfind_layer_idx\u001b[0;34m(model, layer_name)\u001b[0m\n\u001b[1;32m    151\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    152\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mlayer_idx\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 153\u001b[0;31m         \u001b[0;32mraise\u001b[0m \u001b[0mValueError\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"No layer with name '{}' within the model\"\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mformat\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlayer_name\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    154\u001b[0m     \u001b[0;32mreturn\u001b[0m \u001b[0mlayer_idx\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    155\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
            "\u001b[0;31mValueError\u001b[0m: No layer with name 'dense_2' within the model"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xid325ocmdw1",
        "colab_type": "text"
      },
      "source": [
        "**QUESTION**\n",
        "\n",
        "- Run 60 epochs and look at the training and testing accuracy curves. Use again the visualize_layer function. Can you interpret the displays ?"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "IspRWoGkmdw1",
        "colab_type": "text"
      },
      "source": [
        "<a id='cnn'></a>\n",
        "## 3.2. Convolutional Neural Network\n",
        "\n",
        "In practice, increasing the size of hidden layers is not very effective. \n",
        "It is often a better idea to add more layers. \n",
        "Intuitively, if the phenomenon you try to learn has a hierarchical structure, adding more layers can be interpreted as a way to learn more levels of abstraction. \n",
        "For example, if you are trying to recognize objects, it is easier to express shapes from edges and objects from shapes, than to express objects from pixels. \n",
        "Thus, a good design should try to exploit this hierarchy.\n",
        "\n",
        "In particular cases, such as grid-like data (time series, images), you might want to detect a pattern which can happen in different locations of the data. \n",
        "For example, you try to detect a cat, but the cat can be in the middle or the left of the picture. \n",
        "Thus you need to build a model which is translation invariant: it is easier to learn how to recognize an object independently of its location. \n",
        "\n",
        "### Description\n",
        "\n",
        "When two inputs might contain the same kind of information, then it is useful to share their weights and train the weights jointly for those inputs to learn statistical invariants (things that don't change much on average across time or space). \n",
        "Using this concept on images leads to convolutional neural networks (CNNs), on text, it results on recurrent neural networks (RNNs). \n",
        "When using CNNs, you set weights to a small kernel that will be used to perform a convolution across the image.\n",
        "\n",
        "The image is represented as a 3-dimensional tensor: (width, height, depth). Width and height charecterize the size of the image (eg. 28 x 28 pixels), and depth the color space (e.g. 1 for grey levels, 3 for RGB pictures since each pixel is represented by a triplet $(R,G,B)$).\n",
        "\n",
        "![Output = Relu(patch * W + b](https://drive.google.com/uc?id=18CYlxBF5qb97Rco-DfcFIcy_ph1kfbzB)\n",
        "\n",
        "The convolution will map patches of this image, combined with the convolution kernel, for example\n",
        "\n",
        "$$\n",
        "\\text{output} = \\text{ReLU}(\\text{patch} \\times W + b)\n",
        "$$\n",
        "\n",
        "![title](https://drive.google.com/uc?id=1Jh-D_Yk1kD4cVyGkHvwHNrbQu_bNCxIT)\n",
        "![title](https://drive.google.com/uc?id=1FuzkP7K_fngqZduTMPN8m1Fg90fiAFyz)\n",
        "\n",
        "Depending on the shape of the $W$ weights tensor, the tensor resulting from the convolution can have a different depth. Note that in the context of a CNN, the \"kernel\" can be also called a \"filter\".\n",
        "\n",
        "Performing the convolution between the image and the kernel consist to move the kernel across the image, and to produce an output for each patch. \n",
        "The way you move across the image is defined by two parameters:\n",
        "\n",
        "- **Stride:** the stride is the number of pixels you are shifting each time you move your kernel during the convolution.\n",
        "- **Padding:** defines what happens when the kernel reaches a border of the image when doing the convolution. \n",
        "\"Valid\" padding means that you stop at the edge, while \"Same\" padding allows to go off the edge and pad with zeros so that the width and the height of the output and input tensors are the same.\n",
        "\n",
        "For example, a convolution with a stride $> 1$ and valid padding results in a tensor of smaller width and height. \n",
        "You can compute the size of a tensor after convolution using the following formulas:\n",
        "\n",
        "#### Valid padding\n",
        "$$\n",
        "\\text{out}_{\\text{height}} = \\bigg\\lceil \\frac{\\text{in}_{\\text{height}} - \\text{kernel}_{\\text{height}} + 1}{\\text{stride}_{\\text{vertical}}} \\bigg\\rceil \\quad \\text{ and } \\quad\n",
        "\\text{out}_{\\text{width}} = \\bigg\\lceil \\frac{\\text{in}_{\\text{width}} - \\text{kernel}_{\\text{width}} + 1}{\\text{stride}_{\\text{horizontal}}} \\bigg\\rceil\n",
        "$$\n",
        "\n",
        "#### Same padding\n",
        "$$\n",
        "\\text{out}_{\\text{height}} = \\bigg\\lceil \\frac{\\text{in}_{\\text{height}}}{\\text{strides}_{\\text{vertical}}} \n",
        "\\bigg\\rceil \\quad \\text{ and } \\quad \n",
        "\\text{out}_{\\text{width}} = \\bigg\\lceil \\frac{\\text{in}_{\\text{width}}}{\\text{strides}_{\\text{horizontal}}} \\bigg\\rceil\n",
        "$$\n",
        "\n",
        "**Example.**\n",
        "Assume the input tensor is 28x28x3 and the convolution kernel takes in 4x4x3 tensors and outputs 1x1x32 tensors (height x width x depth), i.e the kernel takes in a patch of size 4x4 and depth 3, and output a patch of size 1x1 and depth 32. To do so, the weights tensor $W$ should be 3x3x3x32 (in-height, in-width, in-depth, out-depth). \n",
        "If we are using a stride of 1, the output tensor will be 28x28x32 with same padding, and 25x25x32 with valid padding.\n",
        "Using a stride of 2, the output tensor will be 14x14x32 with same padding, and 13x13x32 with valid padding.\n",
        "\n",
        "Striding is an agressive method to reduce the image size. \n",
        "Instead, it can be a better idea to use a stride of 1 and to combine the convolution's outputs being in some neighborhood. Such an operation combining elements of a tensor is called **pooling**. \n",
        "Neighborhoods are define by the pooling window dimension (width x height) and the strides you use when moving this window across the image.\n",
        "\n",
        "**Example.**\n",
        "Max pooling aggregate several outputs in a neighborhood $N$ using a max operation: \n",
        "\n",
        "$$\n",
        "\\text{output}'_i = \\max_{j \\in N}\\text{output}_j, \\quad i \\in N.\n",
        "$$\n",
        "The formulas to compute the size of the ouput tensor are the same as for convolution padding and striding.\n",
        "\n",
        "Many successful architectures stack convolution layers in a \"pyramidal\" way: each convolution layer result in a tensor with increased depth and decreased height and width. \n",
        "Roughly, increasing the depth increases the complexity of the semantic compexity of your representation, and allows to keep the relevant information in a smaller space (height x width). \n",
        "\n",
        "**QUESTION**\n",
        "\n",
        "Implement a CNN having the following structure:\n",
        "\n",
        "1. Convolutional layer with 32 filters and 3 * 3 kernel sizes and 'relu' activation (use the `Conv2D` object)\n",
        "- Convolutional layer with 64 filters and 3 * 3 kernel sizes and 'relu' activation (use the `Conv2D` object)\n",
        "- Max pooling with pool size 2 * 2 (use the `MaxPooling2D` object)\n",
        "- Dropout with probability 0.25 (use the `Dropout` object)\n",
        "- Dense layer with 128 units with relu activation\n",
        "- Dropout with probability 0.5\n",
        "- Dense output layer with softmax activation"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "scrolled": false,
        "id": "FPTS1W-Pmdw2",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model_cnn = Sequential()\n",
        "\n",
        "# TODO Add the required layers in model_cnn here\n",
        "\n",
        "model_cnn.compile(loss=keras.losses.categorical_crossentropy,\n",
        "                  optimizer=keras.optimizers.Adadelta(),\n",
        "                  metrics=['accuracy'])\n",
        "\n",
        "model_cnn.summary()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "scrolled": true,
        "id": "DR1KYbw2mdw4",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "batch_size = 32\n",
        "epochs = 5\n",
        "\n",
        "# Run the train\n",
        "history = model_cnn.fit(x_train, y_train,\n",
        "                        batch_size=batch_size,\n",
        "                        epochs=epochs,\n",
        "                        verbose=1,\n",
        "                        validation_data=(x_test, y_test))\n",
        "score = model_cnn.evaluate(x_test, y_test, verbose=0)\n",
        "print('Test loss:', score[0])\n",
        "print('Test accuracy:', score[1])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "scrolled": false,
        "id": "dViz5Vfbmdw6",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "plt.figure(figsize=(7, 5))\n",
        "plt.plot(history.epoch, history.history['acc'], lw=3, label='Training')\n",
        "plt.plot(history.epoch, history.history['val_acc'], lw=3, label='Testing')\n",
        "plt.legend(fontsize=14)\n",
        "plt.title('Accuracy of softmax regression', fontsize=16)\n",
        "plt.xlabel('Epoch', fontsize=14)\n",
        "plt.ylabel('Accuracy', fontsize=14)\n",
        "plt.tight_layout()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "axSZEspImdw9",
        "colab_type": "text"
      },
      "source": [
        "**QUESTION**\n",
        "\n",
        "Try to find a way to visualize and interpret the layers of this CNN (this question can be skipped at first)."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "UAOWP1P7mdw-",
        "colab_type": "text"
      },
      "source": [
        "## Playground\n",
        "\n",
        "Please visit [http://playground.tensorflow.org](http://playground.tensorflow.org/#activation=tanh&batchSize=10&dataset=circle&regDataset=reg-plane&learningRate=0.03&regularizationRate=0&noise=0&networkShape=4,2&seed=0.99213&showTestData=false&discretize=false&percTrainData=50&x=true&y=true&xTimesY=false&xSquared=false&ySquared=false&cosX=false&sinX=false&cosY=false&sinY=false&collectStats=false&problem=classification&initZero=false&hideText=false) to play with a Neural Network"
      ]
    }
  ]
}