dl_intro.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "dl_intro.ipynb",
      "provenance": [],
      "collapsed_sections": [
        "OlylxH9QYLFd"
      ],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/nbustos/CompMet/blob/master/dl_intro.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Jt8IL6_Cdmpw"
      },
      "source": [
        "#**Introduction to Deep Learning**\n",
        "### *Michaela Buchanan - Mark III Systems*"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CpfsmL9Sd3G4"
      },
      "source": [
        "---\n",
        "Imports for all the code below. Please run first before anything else!\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "YBtll6NYXKEb",
        "outputId": "a5792444-893e-4277-d2d4-2ab86c630121"
      },
      "source": [
        "import tensorflow as tf\n",
        "print(tf.__version__)\n",
        "\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "\n",
        "from sklearn.preprocessing import OneHotEncoder\n",
        "from skimage import color\n",
        "from skimage import io\n",
        "\n",
        "from keras.layers import Conv2D, Dense, Flatten, MaxPooling2D, Dropout, BatchNormalization\n",
        "from keras.models import Sequential, load_model\n",
        "from tensorflow.keras.optimizers import Adam"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "2.8.0\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "OlylxH9QYLFd"
      },
      "source": [
        "\n",
        "\n",
        "---\n",
        "### Introduction\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "6P5E6nafyte5"
      },
      "source": [
        "In this workshop we will be working with the CIFAR-10 dataset which consists of 10 classes of images. Our goal is to create a model that can classify images into these 10 categories. To do this, we will be experimenting with building a deep neural network using Tensorflow Keras. While this neural network will be very basic in comparison to many common algorithms used today, it should serve as a good starting point for learning about how to implement deep learning algorithms in Python. "
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GMR1lB8fpHme"
      },
      "source": [
        "---\n",
        "### Data Analysis and Preprocessing"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qTGUkV4fzKHF"
      },
      "source": [
        "The 10 classes in CIFAR are denoted by numbers in our Y variables. The list of classes and their associated numbers can be found below:\n",
        "\n",
        "```\n",
        "0.   airplane\n",
        "1.   automobile\n",
        "2.   bird\n",
        "3.   cat\n",
        "4.   deer\n",
        "5.   dog\n",
        "6.   frog\n",
        "7.   horse\n",
        "8.   ship\n",
        "9.   truck\n",
        "```\n",
        "Knowing the dimensions of our images and the number of entries in our dataset are both important for correctly setting up our neural network. All of our images need to be the same size so we have the same number of neurons in the input layer of our model. This shouldn't be a problem with this dataset as all the images are uniform but it is a caveat you will have to watch out for with your own datasets. Let's start by taking a look at our CIFAR-10 dataset.\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 351
        },
        "id": "PV6UP8cTXWL-",
        "outputId": "c9a3ad55-6f1d-40bc-cab6-ed8a3948d1b0"
      },
      "source": [
        "# Load in our dataset from Keras\n",
        "cifar10 = tf.keras.datasets.cifar10\n",
        "(x_train, y_train), (x_test, y_test) = cifar10.load_data()\n",
        "\n",
        "# Use matplotlib to show sample image\n",
        "print(y_train[5])\n",
        "plt.imshow(x_train[5])"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading data from https://www.cs.toronto.edu/~kriz/cifar-10-python.tar.gz\n",
            "170500096/170498071 [==============================] - 4s 0us/step\n",
            "170508288/170498071 [==============================] - 4s 0us/step\n",
            "[1]\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7f81305f5110>"
            ]
          },
          "metadata": {},
          "execution_count": 2
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAPsAAAD5CAYAAADhukOtAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAfTElEQVR4nO2dbWyc15Xf/2feOBySEilRL5RMvVi27Miu36I6Tm24qYMN3CCAk+0iSD4EBhqsF8UGbYDtByMFmhToh2zRJEiBIoWyMdZbZPPSTdJ4F8Z6vd5N0mxS25JjS45lx7Isy6IkUiLFt+G8z+mHGady9v4vaYkcevf+f4Cg4T28z3PmznPm4dz/nHPM3SGE+MdPZr0dEEL0BgW7EImgYBciERTsQiSCgl2IRFCwC5EIuauZbGb3A/gqgCyAP3L3L8Z+f6hY9NGhoaCt3Y5IgEaGC3k6pZnh72OlLDkggPrSErXNlivB8dYV+L6MCRbxP5vjL1uWTCtG1mposERtMWm22WpTm2WywfFKrU7nLCyUqS26jhFblhgzkTntmBwdU6pjl0HEyTaZ2OTLCyPnWqrVUG80gie74mA3syyA/w7gtwCcAfCsmT3m7i+xOaNDQ/j8b38saKuU+UWQzYWvYBsfo3NmS/3UdsvGArWdPvoLavvznz8fPletQedkWfQhfgHk+4rUtmnLKLVt6A+f7/pdW+icD9x9J7U1G/y5XZxbpLb80Ehw/PiJN+icp370c2oDuQYAoC/PbRvz4Te5Qq5F59Qjz7kZjqMOzqOzL9tHbUsevvYvVfm7R4a4+H9eeJHPoZbluRPACXc/6e51AN8G8MBVHE8IsYZcTbDvBPDmZT+f6Y4JId6FrPkGnZk9ZGaHzezwQrW61qcTQhCuJtgnAIxf9vM13bG34e6H3P2gux8cKvLPoUKIteVqgv1ZANeb2V4zKwD4BIDHVsctIcRqc8W78e7eNLPPAHgCHentEXf/ZWxOs1HDpYnXw45EZJx8LrwrOeE1OufVCt9RveU911Jbu86PuW00vAveHzlXTI+J7cYv1bgfczOXqG3RwrvMtWpYNgSAW+94H7U1lvhHr4vT3I9txbAa0q7P0zn9fXyt2uDXx9ahQWq7+drrguMXpv7eH6G/plJZoLbFRa5AIMPlzb5ck9p2bN8YHG8UttI5J146FXYhoilelc7u7o8DePxqjiGE6A36Bp0QiaBgFyIRFOxCJIKCXYhEULALkQhXtRv/Tqm3M3i9Gk4IWKrM0XkFI/JPKyxZAEDGeLLLxTcmqe3I2TPU9vJUWGryGpdVYvJaMfIlo0aTJ2ogkhFX7A+v72yFS1fPHHuV2sY28zWuNWN5e2EZrS9yxeXzsVQ0brph3z5q27Nrd3B8eIhn+p0/d4q70eBS5OAIT8xq5XliVqkvLOftGOWS4pvZsP9m/NrQnV2IRFCwC5EICnYhEkHBLkQiKNiFSISe7sa3DaiQ+m8zGb77bK1wUsjmSC22wQ3hskgAUC3znf/ZBZ6AMl8NJ7x4xPdWi9uy5HgAkIu9Dzd4wkiZJPIMRuqqPfPCUWrbf104kQQAbty3i9pyhfBu8Z49fOe83OaJJJPnLlDb/AJP8kFxIDh88N5b6JTnn/0xtVWaXHlZaPAd/ukyvx43VcI7/DuzPCGnuhiOo0hlLN3ZhUgFBbsQiaBgFyIRFOxCJIKCXYhEULALkQg9ld4MTfTZTNA2VuKSxjDCksymEZ5c8Lpz2WKgP9K5g/XVAVCy8HI1Bni3j0aTy2vVSJ25VuR9uL/EJZ5CX3ittke65+y4ZpzaLi7yxI/z81zyet/7wl1mZibP0zm//a/uprbH/+IJavv5z/4vte26+Y7g+H23vJfOeW3iJLW9/nfPUttcPdzaDAAWI72c3vNPwz5WGrzG3+hoOIkql+MJYLqzC5EICnYhEkHBLkQiKNiFSAQFuxCJoGAXIhGuSnozs1MAFgC0ADTd/WD09zOGwkD4lNcO8VY3ez08Z2Mh0ihyjteSKw1zqaxcWKK2dj6cwXbwtrB0AgDbtvLndfLECWp78zRvT5TJ8uwwb4alsmIkM+/97+P+X+DLgWd+/CNqe+WVcEZcqxI54ADPDJstc5lyscHvWSfOTQfHy+0snVNu8uNNzXI/akVeM+763bzl2PC2HcHxC9Nh3wHgvvtuCo4/ceSv6ZzV0Nn/hbtfXIXjCCHWEP0ZL0QiXG2wO4C/MrMjZvbQajgkhFgbrvbP+HvcfcLMtgJ40sxedvefXP4L3TeBhwBgiNQ0F0KsPVd1Z3f3ie7/UwB+AODvfSHa3Q+5+0F3P9hPvrcthFh7rjjYzWzAzIbeegzgQwBeXC3HhBCry9X8Gb8NwA+67Y1yAP7U3f8yNqHthsV6+O6+MRsuDAgAjYvh7J83Z7k8dc+tN1JbpV6mtp2Rgn3FUjgj7q5h7vuBLaPUttTmGXYX+/hHnqU5ng3VqofHc3WeBbj79OvU1j/LsxE3bRmmtsaLvwiOx2TDn790nNpeOXuW2qpNLodNnA5LsFPTvIDlnbffRW27h3mG4H/70/9NbfUKz/Y78mxYzJqcfI3OueOD4es72+ZrccXB7u4nAdx6pfOFEL1F0psQiaBgFyIRFOxCJIKCXYhEULALkQg9LTiZQwZbsuFMtZ3gWUgbNoQL+T1/iWe2Xarxfm67t/Pii78ztZfa8vNhyW7zq9yPvtfOUVurzYtR7gm38ur40eLGTC68vi3jklftmeeobWNE1mqPcsmxxQoszvPsuw1ZnjVWK3O5dBO/dFDycFHM+fNv0Dk737Of2oYGeKblnft2UtvUHNFEAZxfDGcCLi2Fi7MCwMlXXw2O1yJFTHVnFyIRFOxCJIKCXYhEULALkQgKdiESoae78cVsBjcOhVsXDUzzylbZTHhnd/8119A5C5M80QHOd7N3xto/FcLzspFdU4sku/D9WaCWibwPF3iSTN7D58tF2g/lM1wVaAzxrW5f4ju/zVrYjxb42m/L8BW5r5/v/NeNtzxq7dgWHC+eOkXnLPHDAUQZAoCbbryO2saW+HMba4STjfbvC9emA4DrRsPKRfGJn9I5urMLkQgKdiESQcEuRCIo2IVIBAW7EImgYBciEXoqvbUaNcycPRm01Zpckqlkw7LR0kaeONG/xOWk6nFe26uV5YkaTdK6KpPlskpfRPIy8KSKZkQebLX5MT0fTnjhAmDcltvK2xYNzfJ7RZU8tfpu3uJppLlIbQNVvsbNSJ28xalwQtTS2b+jc84dfoHaNtzEk2Smz3O5t17aRG3NcK4OlqZ5rcH5fHg9Wi2+FrqzC5EICnYhEkHBLkQiKNiFSAQFuxCJoGAXIhGWld7M7BEAHwEw5e43d8c2AfgOgD0ATgH4uLtznaBLs9XC9OJs0PZmucrntcNyQsG20zmlEd52abrCWyFtz/KMsv5q+L2xNc9lvlqd2zDKfRzYzzOoqhGJavHifHC8r82lvGykblntAl8r9HEZzYbDsmguklXYnufXQP9NXAJEgUuwpamwrlWe4K3DZl8+QW3t05PUNrSJZ8TNDHO5dPp8+PU8N8VrG+4thOsotpr8elvJnf2PAdz/G2MPA3jK3a8H8FT3ZyHEu5hlg73bb/03E7YfAPBo9/GjAD66yn4JIVaZK/3Mvs3d36qRfB6djq5CiHcxV71B5+6OyDcuzewhMztsZoeXmvyrqEKIteVKg33SzMYAoPv/FPtFdz/k7gfd/WApF6nmL4RYU6402B8D8GD38YMAfrg67ggh1oqVSG/fAvABAKNmdgbA5wF8EcB3zezTAN4A8PGVnKzpbVyqhuWV80tcTmqQtkuj27bQOT6+ldr6RrhE0jfPs4ZyZ8NZTXXSvgcAFsEll9ZgP7Xld+/ifhj/ODQwHPal8avTdE4jIg9WI8Uoh+49QG1Ls6SA6Csv0zloRu4953hB0lo7LOcCQH57uGjj9n9+F53T18//Ap35Fc+YHF7i8zbu5pLu6fNhOa8/y2XKfD5cFdOMS6zLBru7f5KYPrjcXCHEuwd9g06IRFCwC5EICnYhEkHBLkQiKNiFSISeFpwsFAoYHw/3Z8u8zrOQ+klBvladSxN9Fi68CACXyuHMMAD42Zs802hHNZwBdiOIg4hnvVUimVf1517i8yIlIm3nzuB4dT/PEFxqhvvvAcAt+7i8Vs7wbLPK2VPB8cJcJLtxA2+yVj8dkQ4nw9IsAOS3hr/vtbSNS7P5TRupbeSDd1Db7JvnqG14lMtydwzuDo4/+VOeSNo3HJadM1ke0rqzC5EICnYhEkHBLkQiKNiFSAQFuxCJoGAXIhF6Kr3l8zls3xEuarMwwbOaSiMkk8d4JlE+w7N/zl2cprY/euGX1HbD5rDU9G+LA3ROKfJ26mWe6TdzjEtvM1u4NHSyFpah6hG5bsf+cGYYAOwa4eeqn+PFFweJDGVt3rMNC/w168vwDMH5Cs86bJ0M9xb0s+fpnEtD/LoauCEsHQPAjr37qK1KMtsAYEspfP3cfjMvOjq+N+xHvo/Ll7qzC5EICnYhEkHBLkQiKNiFSAQFuxCJ0NPd+Ja3MNcKf7k/53N0Xj4XdrMeqdE12+TJKTMVPq/pfEnm8+Ed4Yk8TyQZdl7Trp7hNnfekmmuzXefz0yFd+M3ZIp0ziW+0Y3HJh6jthtI0g0A7NsUPt/mPp6QUz7FE4NaFZ7s4i2+jpcuhesGeotfA/Ui341vzHHVqH70VWorRdSQWjGctLX7wE3cj7NvBMe9wdUO3dmFSAQFuxCJoGAXIhEU7EIkgoJdiERQsAuRCCtp//QIgI8AmHL3m7tjXwDwuwDe0jU+5+6PL3ssOAoeboeUa/NabaOZsDRRz0ZaNUUkiKUqb8m0cwtvKXXN3vHg+MQil/ngXHIpEMkFAKzJX5p6m8tyY5tHg+M5vlSYv8CTQnyGy3xnp7kcNlcKJ2TsqvHXOXORS2+o8CeQibSNqjTDPi61+PXhEZmyVIkkWE3w+oWlSFumcjP83IZr/DmP3rI/bGhE1pda/j9/DOD+wPhX3P227r9lA10Isb4sG+zu/hMAMz3wRQixhlzNZ/bPmNlRM3vEzEZWzSMhxJpwpcH+NQD7ANwG4ByAL7FfNLOHzOywmR1erEY+OAoh1pQrCnZ3n3T3lru3AXwdwJ2R3z3k7gfd/eBgsadfxRdCXMYVBbuZjV3248cAvLg67ggh1oqVSG/fAvABAKNmdgbA5wF8wMxuA+AATgH4vZWcLNPOoL8SzhA72+S1zrZmwi2DRiqzdE5uirfiaS7wtjrvObCX2nbdcH1wfOaFV+icMeNtf5Dnslze+ftw/yKXvHIku6pU4qltv3rtFLWNlrkf1+7ZRG1nCmEJaPIEf136F/g+sDUjLa9afI2rRJ6tZ/jzqpf5x82ZVrgFGACUShuobaHO5dJyLfzcZiZ43brcrnD2YKvV4nOopYu7fzIw/I3l5gkh3l3oG3RCJIKCXYhEULALkQgKdiESQcEuRCL0tuBk2zFXDksyP5rjckdzc3j87kgrof4pnslVbPBMrtvfex+17RgPt+P582eO0TlztbBsCACtHM9QakQku37nGVTVM+Hnnd3EZbJrR8KZcgBQbfFCoLkB3mrolnvC37Oa4QoUZo5MUVutzaW3do4XiKyQtRoYIBcVAPTzdl6VAn9d2pv5t8ar4PPOXwhLjnOzvLjlpZfDxS3LVX696c4uRCIo2IVIBAW7EImgYBciERTsQiSCgl2IROip9OatBurzZ4O2E9M8w6fSCEs8w9dwyejWPJe1hiLVF/eOh4tKAsCGwbB8VYsUL6wtcVshzzOUqh6Zl+GSV6Eefm6VGZ5RliG99ACgHemnNznN5c1Lx18KjpeKXIJaKA5yWz/vp1cbHKK2cjmcIVga5VLkTJ3LVwtN/pplGrzw6Lnzi3xeMSz1zUeKpg7MhyXRZiTrTXd2IRJBwS5EIijYhUgEBbsQiaBgFyIRerobv6Evgw/tDu88XpjhO7HPvh5OXHnyFE/S6L+WJzOUBnnixFCW7/o2FsK7tC3jO6DlSCJMMcuXv5WNvA8bt7VJbbWZMt8N9kiJ70KZ+9+YjbRQeu10cLwUub/UIzXcjjV5Bs2pizyBpkg6fRXafOc8H6mCbI1IEtIsVzzKzhWD3GC4DVgrz8+1e2Q4OF7I8hZUurMLkQgKdiESQcEuRCIo2IVIBAW7EImgYBciEVbS/mkcwJ8A2IZOu6dD7v5VM9sE4DsA9qDTAurj7s77KgEo5g37d4RP+a9Lu+i88b6J4PjfvMLlpKdO8USY23bvoLbF116ntlny3phtE30HwGyd17vbUuJyTMt5wkijzZ/bBQ/7crHEpc1qJDFoyPglMrCR+98mCTmYnqdz+vq4XHqmyqWy6RZP1tmeD8tapQG+HkMD3A+vcCnyYp37mMvy6yA7E7bd7DzhaXAhfA1kIrX6VnJnbwL4A3c/AOAuAL9vZgcAPAzgKXe/HsBT3Z+FEO9Slg12dz/n7s91Hy8AOA5gJ4AHADza/bVHAXx0rZwUQlw97+gzu5ntAXA7gKcBbHP3t1pynkfnz3whxLuUFQe7mQ0C+B6Az7r72z54ubsD4V7BZvaQmR02s8MXlvhnQyHE2rKiYDezPDqB/k13/353eNLMxrr2MQDBLyi7+yF3P+juB7eUevpVfCHEZSwb7GZm6PRjP+7uX77M9BiAB7uPHwTww9V3TwixWqzkVns3gE8BOGZmz3fHPgfgiwC+a2afBvAGgI8vd6C2t1EjUtSmIs/wef/+cK25i2UueR2Z4Blxxye5Qnh9ROKpF8LL5W3+nrlQ5dlaXuPSSizzyiPyCoitv69Ipyw4l5Pmd/GtmM033UhtWfLSHHvix3TOeGStrhnZQm2o8ey7Yi7syFykXlx5mstk2yMS5o5R3lKqkOGvZ34mfK3uXuDS8vgwy3rjcbRssLv7TwGwI3xwuflCiHcH+gadEImgYBciERTsQiSCgl2IRFCwC5EIPf2Wi8FgpMiiRQoKjg2HZaN/tncjnTMfaeFzapZLK0sR6WIraQ2VLfAildUml8mqCwvUlmvwIpaFfD+1sRVpTl6gcza0+Dcba/N8rWYaXPocHhkJj0eKZear/Fw7I5lohcg9ywbCxUUtz4+XWeRS3rYcf60j6jEyNf56LpHrYGMkU27frnBM9B3ha6E7uxCJoGAXIhEU7EIkgoJdiERQsAuRCAp2IRKhp9KbA3AP6xPejkhN7bAsd2ATd//CGM9OKte4zNeMFBQc3RzOvCoOcglwNpKh1qjzwpHNiK2W5T5mLFyockPkbZ3nwwH1eZ49iCr3w8+H+69dQ3OqgHw2Uviywv3YmuVS5CUis/YNhaVBAGg3+GI1l2apbb7GpbKI8oZ2rRwcHzuwlc7Zuyt8LfaRzExAd3YhkkHBLkQiKNiFSAQFuxCJoGAXIhF6XO7V0CaJEC3wdkdohnemN+b4zu7t4+G6dQAwvTBDbfXJc9TWKId3TQsDfDe4Gkn8aHgkaSHS4qkVSZKxVnhNmhE/6vlIBgf4Drk1uR+tLKmvl+HnajX5uTyy819shVs8AYA3wkkt54t8V73Rx2sDtsN5NQCA/AD3Y2mJJ9cUSMuuLbu20znFXNjHjPH11Z1diERQsAuRCAp2IRJBwS5EIijYhUgEBbsQibCs9GZm4wD+BJ2WzA7gkLt/1cy+AOB3AbxV3Oxz7v549FiZDAr94dpf2SKv7VWfDbfBiUlQO4b58f7JHJdxjs9OUtv5s6eD4/OV+eA4ACy2eZ22aiZSjy2SQNN0/rwzHn5JyxFJZokkJwFALnI/aNf4c2vXwmtsEemNta4CgGqOP+d2RLIrk2NW+3gyFDL8XMU8197aLS6vDZBkLgC4bttQcHykwNdjaTosHbYjcuhKdPYmgD9w9+fMbAjAETN7smv7irv/1xUcQwixzqyk19s5AOe6jxfM7DiAnWvtmBBidXlHn9nNbA+A2wE83R36jJkdNbNHzIwnCAsh1p0VB7uZDQL4HoDPuvs8gK8B2AfgNnTu/F8i8x4ys8NmdvjiEv8KqBBibVlRsJtZHp1A/6a7fx8A3H3S3Vvu3gbwdQB3hua6+yF3P+juB0dL/LvDQoi1ZdlgNzMD8A0Ax939y5eNj132ax8D8OLquyeEWC1Wsht/N4BPAThmZs93xz4H4JNmdhs6ctwpAL+3ojNmwtltnT8eiJMkqaya4R8L8hHZYtcYl+VeP8PlkzqpFdZq8zmzTW67aHz5h7I8C9CcPzcjEtscV8lwvh6R8iLZctmIZEePF7HlI5mPk5EswDlw/xfJ894ZkQCHI5Judoa37NqW49X83jvOM9j2jYcv8FIlLDkDQI3IfO3WVUhv7v5TIFglMKqpCyHeXegbdEIkgoJdiERQsAuRCAp2IRJBwS5EIvS84CTa4feXWoW3zmESTyyDyiPtkwYHwpl3ADC6gUtlMxfCLY0WSKsjAJjL8vfTn0XkpBGurmFDRKYcINJbI8MPON+MZJtFZK2Y8JYlGX2FiKRYih+RWnLGdcUSed7tBs+Uq5OinQDQH1mPjYP8mGhEMiMvhf2f38BfZyNFWFuRzEHd2YVIBAW7EImgYBciERTsQiSCgl2IRFCwC5EIPZbeuDTgEcnAiHxVIP2uAMArkUIZEVlr6wA/5nPHwlm802cvBMcBoBnJbLsQkZrmI9lypVZEaiKH7ItIgF7gzzkTKYrJMuwAIJcLy0Yt0tcMAOZb/DVrRgopeuSYBeZ+RHprR9Yqk+MXTxvc/9lF3lsu62Ff+jLhQpQAYO3wddWKFDjVnV2IRFCwC5EICnYhEkHBLkQiKNiFSAQFuxCJ0FvpzQyZfFiSyUfkMCM2y0bcjxTea5V5Ib+xIV6McnM+fMx8tULnbGhzeaoaKeYYK/TYzHF5pUykl0pkfRGRvLKRjDiLSIcZIh16pFimR7LXYvlweeMZcXlyjfRH1ncwcgscMH5dkcujCzfWKuFCppHLFKVM+DqNSdi6swuRCAp2IRJBwS5EIijYhUgEBbsQibDsbryZFQH8BEBf9/f/zN0/b2Z7AXwbwGYARwB8yt159kaXTC58yqxH3ndYokN0Nz7STipSu27Q+FO496YdwfG5JT7nF6cvUtvFGk/GqEZ2VWuRvek2WZN25H09WreMSSEAInkwyERq3jGykR3ySP4J+jP8OihlwtfBUI47P5ThqsDmyCVXiixIHvy1LpC18lbk+iAKUDuSFLSSO3sNwH3ufis67ZnvN7O7APwhgK+4+3UALgH49AqOJYRYJ5YNdu/wluKX7/5zAPcB+LPu+KMAPromHgohVoWV9mfPdju4TgF4EsBrAGbdf52IewbAzrVxUQixGqwo2N295e63AbgGwJ0AblzpCczsITM7bGaHL5aX/UgvhFgj3tFuvLvPAvhbAO8HMGz26zIs1wCYIHMOuftBdz84GqkCI4RYW5YNdjPbYmbD3cf9AH4LwHF0gv53ur/2IIAfrpWTQoirZyWJMGMAHjWzLDpvDt91978ws5cAfNvM/jOAXwD4xrJHymSAQpEYucxgLHmCyHgA0CTtcQCgHXnaMbljjOTIfORWvl2xLc+lkBOTvCXQZJn7f6kZSa5ph5NCahHpqmn8OXssWSfSyilLbNGElogEGMn9wUBEgu0j/vdFkm42ZHnSykhEshuI1K4r5rmPObKMjQa/BpZIQk47UoNu2WB396MAbg+Mn0Tn87sQ4h8A+gadEImgYBciERTsQiSCgl2IRFCwC5EIFqsJtuonM7sA4I3uj6MAeEpY75Afb0d+vJ1/aH7sdvctIUNPg/1tJzY77O4H1+Xk8kN+JOiH/owXIhEU7EIkwnoG+6F1PPflyI+3Iz/ezj8aP9btM7sQorfoz3ghEmFdgt3M7jezV8zshJk9vB4+dP04ZWbHzOx5Mzvcw/M+YmZTZvbiZWObzOxJM3u1+//IOvnxBTOb6K7J82b24R74MW5mf2tmL5nZL83s33XHe7omET96uiZmVjSzZ8zsha4f/6k7vtfMnu7GzXfM7J0ViHD3nv4DkEWnrNW1AAoAXgBwoNd+dH05BWB0Hc57L4A7ALx42dh/AfBw9/HDAP5wnfz4AoB/3+P1GANwR/fxEIBfATjQ6zWJ+NHTNUEnE3iw+zgP4GkAdwH4LoBPdMf/B4B/806Oux539jsBnHD3k94pPf1tAA+sgx/rhrv/BMDMbww/gE7hTqBHBTyJHz3H3c+5+3PdxwvoFEfZiR6vScSPnuIdVr3I63oE+04Ab17283oWq3QAf2VmR8zsoXXy4S22ufu57uPzALatoy+fMbOj3T/z1/zjxOWY2R506ic8jXVck9/wA+jxmqxFkdfUN+jucfc7APxLAL9vZveut0NA550dnTei9eBrAPah0yPgHIAv9erEZjYI4HsAPuvubyvj08s1CfjR8zXxqyjyyliPYJ8AMH7Zz7RY5Vrj7hPd/6cA/ADrW3ln0szGAKD7/9R6OOHuk90LrQ3g6+jRmphZHp0A+6a7f7873PM1CfmxXmvSPfc7LvLKWI9gfxbA9d2dxQKATwB4rNdOmNmAmQ299RjAhwC8GJ+1pjyGTuFOYB0LeL4VXF0+hh6siZkZOjUMj7v7ly8z9XRNmB+9XpM1K/Laqx3G39ht/DA6O52vAfgP6+TDtegoAS8A+GUv/QDwLXT+HGyg89nr0+j0zHsKwKsA/hrApnXy438COAbgKDrBNtYDP+5B50/0owCe7/77cK/XJOJHT9cEwC3oFHE9is4by3+87Jp9BsAJAP8LQN87Oa6+QSdEIqS+QSdEMijYhUgEBbsQiaBgFyIRFOxCJIKCXYhEULALkQgKdiES4f8Bjj+JdOtlST4AAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AT0uv2RwTk1g"
      },
      "source": [
        "If you look above, we got the number 1 when we printed the y associated with our image we displayed. This makes sense as we are looking at an image of a car, and according to the list above the image should be in the automobile class. The low resolution is nothing to be alarmed about. CIFAR-10 images are purposely low resolution in order to keep the size of the input layer reasonable as this dataset is primarily for people looking to learn how to use DL, not real world model training. Feel free to look at different images in the dataset by replacing the number 5 in these two lines with another number.\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jb1x9Ncp0DtC"
      },
      "source": [
        "Now we need to figure out some important information about our dataset. Let's start by checking the dimensions of the images we are using."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "70kU0jXQX3Yn",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "9fc4fdd5-e2ce-45ab-ec39-44109c98f818"
      },
      "source": [
        "print(\"single image shape: \",x_train[5].shape)\n",
        "print(\"training dataset shape: \", x_train.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "single image shape:  (32, 32, 3)\n",
            "training dataset shape:  (50000, 32, 32, 3)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ww0kbBNEWlJr"
      },
      "source": [
        "Our model isn't going to know what to do with integer output for our Y like we have now so we need to convert it to one-hot encoding. One-hot encoding creates an array that has the lenght of the number of options we have. Then we set 1 to the corresponding category of the entry we are dealing with and the rest to 0. This will probably make more sense when you take a look at the example below."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "4cxAIgNiZIUY",
        "outputId": "3f0f1bfb-bf6a-4468-a482-7244235ad84c"
      },
      "source": [
        "# creates and formate one-hot encoder\n",
        "one_hot_encoder = OneHotEncoder(sparse=False)\n",
        "one_hot_encoder.fit(y_train)\n",
        "\n",
        "# actually one-hot encodes our data\n",
        "y_train = one_hot_encoder.transform(y_train)\n",
        "y_test = one_hot_encoder.transform(y_test)\n",
        "\n",
        "print(y_train[5])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "[0. 1. 0. 0. 0. 0. 0. 0. 0. 0.]\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8xiHYGteZubb"
      },
      "source": [
        "Here you can see for our 5th training sample, the 1 spot (remember we start counting at 0) has a 1 in it while the rest of the spots in the array are 0. This gives us the same information the Y did previously but in a format that will be much easier for the model to associate with the 10 output neurons we will be creating."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Qon-hkTDi8as"
      },
      "source": [
        "Our x entries are going to consist of three numbers ranging from 0-255 representing the RGB of that pixel. Since we know this, we can easily normalize our data so that it's easier to train with."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Bsa0Rt_ljIuu"
      },
      "source": [
        "x_train = x_train.astype('float32')\n",
        "x_test = x_test.astype('float32')\n",
        "x_train = x_train / 255\n",
        "x_test = x_test / 255"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "tgVhb_M7fO4O"
      },
      "source": [
        "---\n",
        "### Creating our Model\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CG0wzh353Hgs"
      },
      "source": [
        "Now it's time to start creating our neural network. Luckily, Keras makes this pretty easy, especially when you are using what's called a Sequencial model which is what we will be doing today. With a Sequential Keras model, all you need to do is initialize the model and then add layers to it using premade functions in Keras as you will see in a minute. Sequential models are great for situations where you have only one input and one output vector and you are not planning on doing any layer sharing and are sticking to a linear model topology. To start with, let's define some of the variables we need. First we have our batch size which designates how many training samples we go through before we backpropogate changes through the model. Then we have epochs which is how many times we go through the whole dataset. Feel free to change these variables to see what happens. You can do as many or as few epochs but batch_size must be 1 < batch_size < 50,000 since we have 50,000 entries in x_train. num_classes just defines how many categories we have that we are classifying into in our output layer."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gXHgt4VZh8NM"
      },
      "source": [
        "# feel free to experiment with these\n",
        "batch = 32\n",
        "num_epochs = 20\n",
        "\n",
        "num_classes = 10"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fAjYAEm5iBoe"
      },
      "source": [
        "Let's create our model. "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "bS9KKKsnrDcr",
        "outputId": "70cc7065-1273-40c8-8011-a21a5fa9f992"
      },
      "source": [
        "# initialize model\n",
        "model = Sequential()\n",
        "\n",
        "model.add(Conv2D(input_shape=x_train[0,:,:,:].shape, filters=96, kernel_size=(3,3), activation='relu'))\n",
        "model.add(Conv2D(filters=96, kernel_size=(3,3), strides=2, activation='relu'))\n",
        "model.add(Conv2D(filters=192, kernel_size=(3,3), activation='relu'))\n",
        "model.add(Conv2D(filters=192, kernel_size=(3,3), strides=2, activation='relu'))\n",
        "\n",
        "model.add(Flatten())\n",
        "model.add(BatchNormalization())\n",
        "model.add(Dense(256, activation='relu'))\n",
        "model.add(Dense(num_classes, activation=\"softmax\"))\n",
        "# model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
        "\n",
        "model.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " conv2d (Conv2D)             (None, 30, 30, 96)        2688      \n",
            "                                                                 \n",
            " conv2d_1 (Conv2D)           (None, 14, 14, 96)        83040     \n",
            "                                                                 \n",
            " conv2d_2 (Conv2D)           (None, 12, 12, 192)       166080    \n",
            "                                                                 \n",
            " conv2d_3 (Conv2D)           (None, 5, 5, 192)         331968    \n",
            "                                                                 \n",
            " flatten (Flatten)           (None, 4800)              0         \n",
            "                                                                 \n",
            " batch_normalization (BatchN  (None, 4800)             19200     \n",
            " ormalization)                                                   \n",
            "                                                                 \n",
            " dense (Dense)               (None, 256)               1229056   \n",
            "                                                                 \n",
            " dense_1 (Dense)             (None, 10)                2570      \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 1,834,602\n",
            "Trainable params: 1,825,002\n",
            "Non-trainable params: 9,600\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gYTkeuwS_Fng"
      },
      "source": [
        "The last line prints a nice summary of the model we have created and the output shapes from each layer. If you are having trouble getting your data to flow through your layers, this model summary can be a great tool for debugging your model. Now that we have defined our model structure, we can use Keras to compile it."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dCW3YDAn_EtQ"
      },
      "source": [
        "model.compile(\n",
        "    loss='categorical_crossentropy', \n",
        "    optimizer=Adam(learning_rate=0.0001), \n",
        "    metrics=['acc'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MzchUyZcCRqd"
      },
      "source": [
        "And finally we can start training our model. To do this, all we have to do is run the fit function on our model as shown below.\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "WQThjEfUCm6-",
        "outputId": "e337c2c7-4d42-42b4-89c8-18acfda1eaed"
      },
      "source": [
        "history = model.fit(\n",
        "    x_train, \n",
        "    y_train, \n",
        "    epochs=num_epochs, \n",
        "    batch_size=batch,\n",
        "    validation_data=(x_test, y_test))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "1563/1563 [==============================] - 42s 18ms/step - loss: 1.4371 - acc: 0.4834 - val_loss: 1.1866 - val_acc: 0.5762\n",
            "Epoch 2/20\n",
            "1563/1563 [==============================] - 29s 18ms/step - loss: 1.0341 - acc: 0.6375 - val_loss: 1.1564 - val_acc: 0.6152\n",
            "Epoch 3/20\n",
            "1563/1563 [==============================] - 28s 18ms/step - loss: 0.8192 - acc: 0.7154 - val_loss: 0.8991 - val_acc: 0.6887\n",
            "Epoch 4/20\n",
            "1563/1563 [==============================] - 29s 19ms/step - loss: 0.6595 - acc: 0.7713 - val_loss: 0.8920 - val_acc: 0.6991\n",
            "Epoch 5/20\n",
            "1563/1563 [==============================] - 29s 18ms/step - loss: 0.5353 - acc: 0.8169 - val_loss: 0.8711 - val_acc: 0.7148\n",
            "Epoch 6/20\n",
            "1563/1563 [==============================] - 28s 18ms/step - loss: 0.4209 - acc: 0.8596 - val_loss: 0.8811 - val_acc: 0.7186\n",
            "Epoch 7/20\n",
            "1563/1563 [==============================] - 28s 18ms/step - loss: 0.3204 - acc: 0.8952 - val_loss: 0.8552 - val_acc: 0.7287\n",
            "Epoch 8/20\n",
            "1563/1563 [==============================] - 29s 18ms/step - loss: 0.2420 - acc: 0.9246 - val_loss: 0.9249 - val_acc: 0.7250\n",
            "Epoch 9/20\n",
            "1563/1563 [==============================] - 27s 17ms/step - loss: 0.1783 - acc: 0.9484 - val_loss: 1.0702 - val_acc: 0.7073\n",
            "Epoch 10/20\n",
            "1563/1563 [==============================] - 28s 18ms/step - loss: 0.1340 - acc: 0.9616 - val_loss: 1.0414 - val_acc: 0.7155\n",
            "Epoch 11/20\n",
            "1563/1563 [==============================] - 28s 18ms/step - loss: 0.1061 - acc: 0.9708 - val_loss: 1.0602 - val_acc: 0.7301\n",
            "Epoch 12/20\n",
            "1563/1563 [==============================] - 28s 18ms/step - loss: 0.0833 - acc: 0.9771 - val_loss: 1.1204 - val_acc: 0.7271\n",
            "Epoch 13/20\n",
            "1563/1563 [==============================] - 27s 18ms/step - loss: 0.0670 - acc: 0.9819 - val_loss: 1.1667 - val_acc: 0.7248\n",
            "Epoch 14/20\n",
            "1563/1563 [==============================] - 27s 18ms/step - loss: 0.0604 - acc: 0.9840 - val_loss: 1.1437 - val_acc: 0.7330\n",
            "Epoch 15/20\n",
            "1563/1563 [==============================] - 27s 17ms/step - loss: 0.0516 - acc: 0.9863 - val_loss: 1.2830 - val_acc: 0.7181\n",
            "Epoch 16/20\n",
            "1563/1563 [==============================] - 27s 18ms/step - loss: 0.0472 - acc: 0.9866 - val_loss: 1.2266 - val_acc: 0.7280\n",
            "Epoch 17/20\n",
            "1563/1563 [==============================] - 27s 18ms/step - loss: 0.0449 - acc: 0.9867 - val_loss: 1.3407 - val_acc: 0.7272\n",
            "Epoch 18/20\n",
            "1563/1563 [==============================] - 27s 18ms/step - loss: 0.0377 - acc: 0.9898 - val_loss: 1.3656 - val_acc: 0.7143\n",
            "Epoch 19/20\n",
            "1563/1563 [==============================] - 27s 18ms/step - loss: 0.0377 - acc: 0.9890 - val_loss: 1.4042 - val_acc: 0.7232\n",
            "Epoch 20/20\n",
            "1563/1563 [==============================] - 27s 18ms/step - loss: 0.0374 - acc: 0.9887 - val_loss: 1.3404 - val_acc: 0.7308\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "oFnqStkTYSAp",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 573
        },
        "outputId": "cd13dc0a-efe4-4f6d-98a7-d1bfa23f104c"
      },
      "source": [
        "#  Ploting the accuracy graph\n",
        "plt.plot(history.history['acc'])\n",
        "plt.plot(history.history['val_acc'])\n",
        "plt.title('model accuracy')\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'val'], loc='best')\n",
        "plt.show()\n",
        "# Ploting the loss graph\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', 'val'], loc='best')\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Jxi5ux2xt_Xy"
      },
      "source": [
        "As you can see, our accuracy and loss look great for our training data. However, the same cannot be said for our validation data. This is usually a red flag that our model is overfitting to our training data. In the next section, we are going to make some changes to this model to see if we can get the ovefitting under control."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "V87QIOEct5sB"
      },
      "source": [
        "\n",
        "\n",
        "---\n",
        "\n",
        "###Improving our Model"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5CQHgQuRvgjF"
      },
      "source": [
        "Below is a very similar model to the one we created earlier. The only difference is the addition of 2 dropout layers. Dropout layers will ignore a random selection of neurons during training which can help with overfitting."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0c1v7IaRgE9-",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "0918bc54-8a2f-4abd-809c-9fcaee45413a"
      },
      "source": [
        "model2 = Sequential()\n",
        "\n",
        "model2.add(Conv2D(input_shape=x_train[0,:,:,:].shape, filters=96, kernel_size=(3,3), activation='relu'))\n",
        "model2.add(Conv2D(filters=96, kernel_size=(3,3), strides=2, activation='relu'))\n",
        "\n",
        "# added dropout layer\n",
        "model2.add(Dropout(0.2))\n",
        "model2.add(Conv2D(filters=192, kernel_size=(3,3), activation='relu'))\n",
        "model2.add(Conv2D(filters=192, kernel_size=(3,3), strides=2, activation='relu'))\n",
        "\n",
        "# added dropout layer\n",
        "model2.add(Dropout(0.5))\n",
        "model2.add(Flatten())\n",
        "model2.add(BatchNormalization())\n",
        "model2.add(Dense(256, activation='relu'))\n",
        "model2.add(Dense(num_classes, activation=\"softmax\"))\n",
        "model2.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
        "model2.summary()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"sequential_1\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " conv2d_4 (Conv2D)           (None, 30, 30, 96)        2688      \n",
            "                                                                 \n",
            " conv2d_5 (Conv2D)           (None, 14, 14, 96)        83040     \n",
            "                                                                 \n",
            " dropout (Dropout)           (None, 14, 14, 96)        0         \n",
            "                                                                 \n",
            " conv2d_6 (Conv2D)           (None, 12, 12, 192)       166080    \n",
            "                                                                 \n",
            " conv2d_7 (Conv2D)           (None, 5, 5, 192)         331968    \n",
            "                                                                 \n",
            " dropout_1 (Dropout)         (None, 5, 5, 192)         0         \n",
            "                                                                 \n",
            " flatten_1 (Flatten)         (None, 4800)              0         \n",
            "                                                                 \n",
            " batch_normalization_1 (Batc  (None, 4800)             19200     \n",
            " hNormalization)                                                 \n",
            "                                                                 \n",
            " dense_2 (Dense)             (None, 256)               1229056   \n",
            "                                                                 \n",
            " dense_3 (Dense)             (None, 10)                2570      \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 1,834,602\n",
            "Trainable params: 1,825,002\n",
            "Non-trainable params: 9,600\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "YEQTRM-7zMYC"
      },
      "source": [
        "Let's try training this model with our dropout layers to see if it helps with our overfitting issue."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_3bF-VNMzHT1",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "0798d8c7-8b5b-4910-febd-8ae24f35c832"
      },
      "source": [
        "history2 = model2.fit(\n",
        "    x_train, \n",
        "    y_train, \n",
        "    epochs=num_epochs, \n",
        "    batch_size=batch,\n",
        "    validation_data=(x_test, y_test))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/20\n",
            "1563/1563 [==============================] - 30s 19ms/step - loss: 1.6593 - accuracy: 0.3956 - val_loss: 1.5160 - val_accuracy: 0.4528\n",
            "Epoch 2/20\n",
            "1563/1563 [==============================] - 28s 18ms/step - loss: 1.3688 - accuracy: 0.5037 - val_loss: 1.2007 - val_accuracy: 0.5719\n",
            "Epoch 3/20\n",
            "1563/1563 [==============================] - 29s 18ms/step - loss: 1.2155 - accuracy: 0.5643 - val_loss: 1.1973 - val_accuracy: 0.5787\n",
            "Epoch 4/20\n",
            "1563/1563 [==============================] - 28s 18ms/step - loss: 1.1055 - accuracy: 0.6056 - val_loss: 1.0445 - val_accuracy: 0.6372\n",
            "Epoch 5/20\n",
            "1563/1563 [==============================] - 29s 18ms/step - loss: 1.0372 - accuracy: 0.6343 - val_loss: 0.9785 - val_accuracy: 0.6585\n",
            "Epoch 6/20\n",
            "1563/1563 [==============================] - 28s 18ms/step - loss: 0.9827 - accuracy: 0.6527 - val_loss: 0.9198 - val_accuracy: 0.6772\n",
            "Epoch 7/20\n",
            "1563/1563 [==============================] - 28s 18ms/step - loss: 0.9393 - accuracy: 0.6681 - val_loss: 0.9484 - val_accuracy: 0.6742\n",
            "Epoch 8/20\n",
            "1563/1563 [==============================] - 28s 18ms/step - loss: 0.8915 - accuracy: 0.6884 - val_loss: 0.8624 - val_accuracy: 0.7049\n",
            "Epoch 9/20\n",
            "1563/1563 [==============================] - 29s 18ms/step - loss: 0.8616 - accuracy: 0.6971 - val_loss: 0.9627 - val_accuracy: 0.6774\n",
            "Epoch 10/20\n",
            "1563/1563 [==============================] - 28s 18ms/step - loss: 0.8297 - accuracy: 0.7069 - val_loss: 0.8831 - val_accuracy: 0.7000\n",
            "Epoch 11/20\n",
            "1563/1563 [==============================] - 28s 18ms/step - loss: 0.8040 - accuracy: 0.7180 - val_loss: 0.8509 - val_accuracy: 0.7094\n",
            "Epoch 12/20\n",
            "1563/1563 [==============================] - 28s 18ms/step - loss: 0.7864 - accuracy: 0.7259 - val_loss: 0.8230 - val_accuracy: 0.7207\n",
            "Epoch 13/20\n",
            "1563/1563 [==============================] - 29s 18ms/step - loss: 0.7763 - accuracy: 0.7272 - val_loss: 0.9628 - val_accuracy: 0.6786\n",
            "Epoch 14/20\n",
            "1563/1563 [==============================] - 29s 19ms/step - loss: 0.7480 - accuracy: 0.7375 - val_loss: 0.7699 - val_accuracy: 0.7312\n",
            "Epoch 15/20\n",
            "1563/1563 [==============================] - 28s 18ms/step - loss: 0.7389 - accuracy: 0.7403 - val_loss: 0.8773 - val_accuracy: 0.6975\n",
            "Epoch 16/20\n",
            "1563/1563 [==============================] - 28s 18ms/step - loss: 0.7047 - accuracy: 0.7542 - val_loss: 0.7682 - val_accuracy: 0.7396\n",
            "Epoch 17/20\n",
            "1563/1563 [==============================] - 29s 19ms/step - loss: 0.7137 - accuracy: 0.7496 - val_loss: 0.7387 - val_accuracy: 0.7486\n",
            "Epoch 18/20\n",
            "1563/1563 [==============================] - 28s 18ms/step - loss: 0.6855 - accuracy: 0.7585 - val_loss: 0.8060 - val_accuracy: 0.7330\n",
            "Epoch 19/20\n",
            "1563/1563 [==============================] - 28s 18ms/step - loss: 0.6805 - accuracy: 0.7623 - val_loss: 0.9536 - val_accuracy: 0.6872\n",
            "Epoch 20/20\n",
            "1563/1563 [==============================] - 29s 19ms/step - loss: 0.6721 - accuracy: 0.7639 - val_loss: 0.7550 - val_accuracy: 0.7419\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UTDcTFXEzH2T",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "b0319d5a-7ca8-4854-8abf-b57e0554c95d"
      },
      "source": [
        "#  Ploting the accuracy graph\n",
        "plt.plot(history2.history['accuracy'])\n",
        "plt.plot(history2.history['val_accuracy'])\n",
        "plt.title('model accuracy')\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'val'], loc='best')\n",
        "plt.show()\n",
        "# Ploting the loss graph\n",
        "plt.plot(history2.history['loss'])\n",
        "plt.plot(history2.history['val_loss'])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'val'], loc='best')\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "CZolzkLezSYC"
      },
      "source": [
        "As you can see, the dropout layers were very helpful with curbing the model's overfitting tendencies. The training and validation accuracies were much more in line with each other. We did take a bit of a hit in training accuracy but the accuracy we achieved in our earlier model wasn't very meaningful anyway since it didn't generalize to data outside of the training set."
      ]
    }
  ]
}