handwritten_digit_recognition_CPU.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "MHvuWIUHdIpt"
   },
   "source": [
    "# Handwritten Digit Recognition\n",
    "- adapted from https://github.com/amitrajitbose/handwritten-digit-recognition\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "oGjRmijsaXJ3"
   },
   "source": [
    "### Necessary Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "TOyGrPT5ASDc"
   },
   "outputs": [],
   "source": [
    "# Import necessary packages\n",
    "%matplotlib inline\n",
    "%config InlineBackend.figure_format = 'retina'\n",
    "\n",
    "import numpy as np\n",
    "import torch\n",
    "import torchvision\n",
    "import matplotlib.pyplot as plt\n",
    "from time import time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Collecting google-colab\n",
      "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/70/9f/d3ec1275a089ec017f9c91af22ecd1e2fe738254b944e7a1f9528fcfacd0/google-colab-1.0.0.tar.gz (72kB)\n",
      "\u001b[K     |████████████████████████████████| 81kB 1.7MB/s eta 0:00:01\n",
      "\u001b[?25hCollecting google-auth~=1.4.0 (from google-colab)\n",
      "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/56/80/369a47c28ce7d9be6a6973338133d073864d8efbb62747e414c34a3a5f4f/google_auth-1.4.2-py2.py3-none-any.whl (64kB)\n",
      "\u001b[K     |████████████████████████████████| 71kB 3.0MB/s eta 0:00:011\n",
      "\u001b[?25hCollecting ipykernel~=4.6.0 (from google-colab)\n",
      "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/18/c3/76775a650cae2e3d9c033b26153583e61282692d9a3af12a3022d8f0cefa/ipykernel-4.6.1-py3-none-any.whl (104kB)\n",
      "\u001b[K     |████████████████████████████████| 112kB 2.2MB/s eta 0:00:01\n",
      "\u001b[?25hCollecting ipython~=5.5.0 (from google-colab)\n",
      "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/07/63/c987612bcf82c56eaacaf6bf01e31e53a244a0a3a0fb036ec5adc377e0fe/ipython-5.5.0-py3-none-any.whl (758kB)\n",
      "\u001b[K     |████████████████████████████████| 768kB 1.7MB/s eta 0:00:01\n",
      "\u001b[?25hCollecting notebook~=5.2.0 (from google-colab)\n",
      "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/96/a2/d951ebb7855743f989287486ded73cd3f66915c3ecd9c5c5a0c7ca12377a/notebook-5.2.2-py2.py3-none-any.whl (8.0MB)\n",
      "\u001b[K     |████████████████████████████████| 8.0MB 390kB/s eta 0:00:011\n",
      "\u001b[?25hRequirement already satisfied: six~=1.12.0 in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from google-colab) (1.12.0)\n",
      "Collecting pandas~=0.24.0 (from google-colab)\n",
      "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/fc/43/fd867e3347559845c8f993059d410c50a1e18709f1c4d4b3b47323a06a37/pandas-0.24.2-cp37-cp37m-macosx_10_6_intel.macosx_10_9_intel.macosx_10_9_x86_64.macosx_10_10_intel.macosx_10_10_x86_64.whl (15.9MB)\n",
      "\u001b[K     |████████████████████████████████| 15.9MB 391kB/s eta 0:00:01\n",
      "\u001b[?25hCollecting portpicker~=1.2.0 (from google-colab)\n",
      "  Downloading https://files.pythonhosted.org/packages/49/2c/a75ef568273036aa61319a554164e6031e31708106ea6ca10e17265e1703/portpicker-1.2.0.tar.gz\n",
      "Collecting requests~=2.21.0 (from google-colab)\n",
      "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/7d/e3/20f3d364d6c8e5d2353c72a67778eb189176f08e873c9900e10c0287b84b/requests-2.21.0-py2.py3-none-any.whl (57kB)\n",
      "\u001b[K     |████████████████████████████████| 61kB 3.1MB/s eta 0:00:01\n",
      "\u001b[?25hCollecting tornado~=4.5.0 (from google-colab)\n",
      "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/e3/7b/e29ab3d51c8df66922fea216e2bddfcb6430fb29620e5165b16a216e0d3c/tornado-4.5.3.tar.gz (484kB)\n",
      "\u001b[K     |████████████████████████████████| 491kB 2.0MB/s eta 0:00:01\n",
      "\u001b[?25hRequirement already satisfied: rsa>=3.1.4 in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from google-auth~=1.4.0->google-colab) (4.0)\n",
      "Requirement already satisfied: cachetools>=2.0.0 in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from google-auth~=1.4.0->google-colab) (4.0.0)\n",
      "Requirement already satisfied: pyasn1-modules>=0.2.1 in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from google-auth~=1.4.0->google-colab) (0.2.8)\n",
      "Requirement already satisfied: traitlets>=4.1.0 in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from ipykernel~=4.6.0->google-colab) (4.3.2)\n",
      "Requirement already satisfied: jupyter-client in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from ipykernel~=4.6.0->google-colab) (5.3.1)\n",
      "Requirement already satisfied: setuptools>=18.5 in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from ipython~=5.5.0->google-colab) (41.4.0)\n",
      "Requirement already satisfied: pygments in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from ipython~=5.5.0->google-colab) (2.4.2)\n",
      "Requirement already satisfied: pexpect; sys_platform != \"win32\" in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from ipython~=5.5.0->google-colab) (4.7.0)\n",
      "Requirement already satisfied: pickleshare in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from ipython~=5.5.0->google-colab) (0.7.5)\n",
      "Requirement already satisfied: decorator in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from ipython~=5.5.0->google-colab) (4.4.0)\n",
      "Collecting prompt-toolkit<2.0.0,>=1.0.4 (from ipython~=5.5.0->google-colab)\n",
      "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/64/27/5fd61a451d086ad4aa806dc72fe1383d2bc0e74323668672287f616d5d51/prompt_toolkit-1.0.18-py3-none-any.whl (245kB)\n",
      "\u001b[K     |████████████████████████████████| 245kB 4.6MB/s eta 0:00:01\n",
      "\u001b[?25hCollecting simplegeneric>0.8 (from ipython~=5.5.0->google-colab)\n",
      "  Downloading https://files.pythonhosted.org/packages/3d/57/4d9c9e3ae9a255cd4e1106bb57e24056d3d0709fc01b2e3e345898e49d5b/simplegeneric-0.8.1.zip\n",
      "Requirement already satisfied: appnope; sys_platform == \"darwin\" in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from ipython~=5.5.0->google-colab) (0.1.0)\n",
      "Requirement already satisfied: terminado>=0.3.3; sys_platform != \"win32\" in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from notebook~=5.2.0->google-colab) (0.8.2)\n",
      "Requirement already satisfied: nbformat in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from notebook~=5.2.0->google-colab) (4.4.0)\n",
      "Requirement already satisfied: ipython-genutils in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from notebook~=5.2.0->google-colab) (0.2.0)\n",
      "Requirement already satisfied: jinja2 in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from notebook~=5.2.0->google-colab) (2.10.1)\n",
      "Requirement already satisfied: nbconvert in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from notebook~=5.2.0->google-colab) (5.6.0)\n",
      "Requirement already satisfied: jupyter-core in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from notebook~=5.2.0->google-colab) (4.5.0)\n",
      "Requirement already satisfied: python-dateutil>=2.5.0 in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from pandas~=0.24.0->google-colab) (2.8.0)\n",
      "Requirement already satisfied: numpy>=1.12.0 in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from pandas~=0.24.0->google-colab) (1.16.4)\n",
      "Requirement already satisfied: pytz>=2011k in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from pandas~=0.24.0->google-colab) (2019.2)\n",
      "Requirement already satisfied: idna<2.9,>=2.5 in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from requests~=2.21.0->google-colab) (2.8)\n",
      "Requirement already satisfied: certifi>=2017.4.17 in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from requests~=2.21.0->google-colab) (2019.11.28)\n",
      "Collecting urllib3<1.25,>=1.21.1 (from requests~=2.21.0->google-colab)\n",
      "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/01/11/525b02e4acc0c747de8b6ccdab376331597c569c42ea66ab0a1dbd36eca2/urllib3-1.24.3-py2.py3-none-any.whl (118kB)\n",
      "\u001b[K     |████████████████████████████████| 122kB 1.3MB/s eta 0:00:01\n",
      "\u001b[?25hRequirement already satisfied: chardet<3.1.0,>=3.0.2 in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from requests~=2.21.0->google-colab) (3.0.4)\n",
      "Requirement already satisfied: pyasn1>=0.1.3 in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from rsa>=3.1.4->google-auth~=1.4.0->google-colab) (0.4.8)\n",
      "Requirement already satisfied: pyzmq>=13 in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from jupyter-client->ipykernel~=4.6.0->google-colab) (18.1.0)\n",
      "Requirement already satisfied: ptyprocess>=0.5 in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from pexpect; sys_platform != \"win32\"->ipython~=5.5.0->google-colab) (0.6.0)\n",
      "Requirement already satisfied: wcwidth in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from prompt-toolkit<2.0.0,>=1.0.4->ipython~=5.5.0->google-colab) (0.1.7)\n",
      "Requirement already satisfied: jsonschema!=2.5.0,>=2.4 in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from nbformat->notebook~=5.2.0->google-colab) (3.0.2)\n",
      "Requirement already satisfied: MarkupSafe>=0.23 in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from jinja2->notebook~=5.2.0->google-colab) (1.1.1)\n",
      "Requirement already satisfied: testpath in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from nbconvert->notebook~=5.2.0->google-colab) (0.4.2)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: entrypoints>=0.2.2 in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from nbconvert->notebook~=5.2.0->google-colab) (0.3)\n",
      "Requirement already satisfied: mistune<2,>=0.8.1 in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from nbconvert->notebook~=5.2.0->google-colab) (0.8.4)\n",
      "Requirement already satisfied: defusedxml in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from nbconvert->notebook~=5.2.0->google-colab) (0.6.0)\n",
      "Requirement already satisfied: pandocfilters>=1.4.1 in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from nbconvert->notebook~=5.2.0->google-colab) (1.4.2)\n",
      "Requirement already satisfied: bleach in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from nbconvert->notebook~=5.2.0->google-colab) (3.1.0)\n",
      "Requirement already satisfied: attrs>=17.4.0 in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from jsonschema!=2.5.0,>=2.4->nbformat->notebook~=5.2.0->google-colab) (19.1.0)\n",
      "Requirement already satisfied: pyrsistent>=0.14.0 in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from jsonschema!=2.5.0,>=2.4->nbformat->notebook~=5.2.0->google-colab) (0.15.4)\n",
      "Requirement already satisfied: webencodings in /Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages (from bleach->nbconvert->notebook~=5.2.0->google-colab) (0.5.1)\n",
      "Building wheels for collected packages: google-colab, portpicker, tornado, simplegeneric\n",
      "  Building wheel for google-colab (setup.py) ... \u001b[?25ldone\n",
      "\u001b[?25h  Created wheel for google-colab: filename=google_colab-1.0.0-py2.py3-none-any.whl size=102291 sha256=c0dfed5400d50f5f7e32ed41998d8e347dc0f7ae8de06798d445b5f07fc16c25\n",
      "  Stored in directory: /Users/elena/Library/Caches/pip/wheels/38/0d/59/701e300a337b2a2e07b27fe74dbfff0bc56ac58f711566ee67\n",
      "  Building wheel for portpicker (setup.py) ... \u001b[?25ldone\n",
      "\u001b[?25h  Created wheel for portpicker: filename=portpicker-1.2.0-cp37-none-any.whl size=13368 sha256=d7591476f4809a5c274f4d2413a1e152044c9661971158bb8e38fdbd2f6256f7\n",
      "  Stored in directory: /Users/elena/Library/Caches/pip/wheels/4a/45/47/1e126be9d4605e71f00d6e6fb151611f2f4cb9770b050c7d2d\n",
      "  Building wheel for tornado (setup.py) ... \u001b[?25ldone\n",
      "\u001b[?25h  Created wheel for tornado: filename=tornado-4.5.3-cp37-cp37m-macosx_10_9_x86_64.whl size=422223 sha256=79385651975516493fb351029fa3940bda781ee73040800e743606a241d8520b\n",
      "  Stored in directory: /Users/elena/Library/Caches/pip/wheels/72/bf/f4/b68fa69596986881b397b18ff2b9af5f8181233aadcc9f76fd\n",
      "  Building wheel for simplegeneric (setup.py) ... \u001b[?25ldone\n",
      "\u001b[?25h  Created wheel for simplegeneric: filename=simplegeneric-0.8.1-cp37-none-any.whl size=5075 sha256=dbf9c0f367f2d1aacaa69c65234937c787b64d7aa39a00a5c9ab82c9a48b6928\n",
      "  Stored in directory: /Users/elena/Library/Caches/pip/wheels/a9/28/53/f24776b4c5bcbe91aaf1f1e247bd6fadd17191aa12fac63902\n",
      "Successfully built google-colab portpicker tornado simplegeneric\n",
      "\u001b[31mERROR: tensorboard 2.1.0 has requirement google-auth<2,>=1.6.3, but you'll have google-auth 1.4.2 which is incompatible.\u001b[0m\n",
      "\u001b[31mERROR: jupyter-console 6.0.0 has requirement prompt-toolkit<2.1.0,>=2.0.0, but you'll have prompt-toolkit 1.0.18 which is incompatible.\u001b[0m\n",
      "Installing collected packages: google-auth, tornado, prompt-toolkit, simplegeneric, ipython, ipykernel, notebook, pandas, portpicker, urllib3, requests, google-colab\n",
      "  Found existing installation: google-auth 1.10.1\n",
      "    Uninstalling google-auth-1.10.1:\n",
      "      Successfully uninstalled google-auth-1.10.1\n",
      "  Found existing installation: tornado 6.0.3\n",
      "    Uninstalling tornado-6.0.3:\n",
      "      Successfully uninstalled tornado-6.0.3\n",
      "  Found existing installation: prompt-toolkit 2.0.9\n",
      "    Uninstalling prompt-toolkit-2.0.9:\n",
      "      Successfully uninstalled prompt-toolkit-2.0.9\n",
      "  Found existing installation: ipython 7.8.0\n",
      "    Uninstalling ipython-7.8.0:\n",
      "      Successfully uninstalled ipython-7.8.0\n",
      "  Found existing installation: ipykernel 5.1.2\n",
      "    Uninstalling ipykernel-5.1.2:\n",
      "      Successfully uninstalled ipykernel-5.1.2\n",
      "  Found existing installation: notebook 6.0.1\n",
      "    Uninstalling notebook-6.0.1:\n",
      "      Successfully uninstalled notebook-6.0.1\n",
      "  Found existing installation: pandas 0.25.1\n",
      "    Uninstalling pandas-0.25.1:\n",
      "      Successfully uninstalled pandas-0.25.1\n",
      "  Found existing installation: urllib3 1.25.4\n",
      "    Uninstalling urllib3-1.25.4:\n",
      "      Successfully uninstalled urllib3-1.25.4\n",
      "  Found existing installation: requests 2.22.0\n",
      "    Uninstalling requests-2.22.0:\n",
      "      Successfully uninstalled requests-2.22.0\n",
      "Successfully installed google-auth-1.4.2 google-colab-1.0.0 ipykernel-4.6.1 ipython-5.5.0 notebook-5.2.2 pandas-0.24.2 portpicker-1.2.0 prompt-toolkit-1.0.18 requests-2.21.0 simplegeneric-0.8.1 tornado-4.5.3 urllib3-1.24.3\n"
     ]
    }
   ],
   "source": [
    "!pip install google-colab"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "XPuuTDEfAfDy"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/elena/anaconda2/envs/torch/lib/python3.7/site-packages/IPython/utils/traitlets.py:5: UserWarning: IPython.utils.traitlets has moved to a top-level traitlets package.\n",
      "  warn(\"IPython.utils.traitlets has moved to a top-level traitlets package.\")\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "from google.colab import drive"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "uLdtrS4zaeEs"
   },
   "source": [
    "### Download The Dataset & Define The Transforms"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 119
    },
    "colab_type": "code",
    "id": "sZD2NGz2Ak6w",
    "outputId": "74eec0da-d867-406b-be2c-4013a2162bf1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz to drive/My Drive/mnist/MNIST_data/MNIST/raw/train-images-idx3-ubyte.gz\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c533469c62e740fc9ecc59086f50bb47",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Extracting drive/My Drive/mnist/MNIST_data/MNIST/raw/train-images-idx3-ubyte.gz to drive/My Drive/mnist/MNIST_data/MNIST/raw\n",
      "Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz to drive/My Drive/mnist/MNIST_data/MNIST/raw/train-labels-idx1-ubyte.gz\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "32df956a1e514251ba6fa33b290238e9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Extracting drive/My Drive/mnist/MNIST_data/MNIST/raw/train-labels-idx1-ubyte.gz to drive/My Drive/mnist/MNIST_data/MNIST/raw\n",
      "Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz to drive/My Drive/mnist/MNIST_data/MNIST/raw/t10k-images-idx3-ubyte.gz\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6f08d85773d34112a5572f80c791cffa",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Extracting drive/My Drive/mnist/MNIST_data/MNIST/raw/t10k-images-idx3-ubyte.gz to drive/My Drive/mnist/MNIST_data/MNIST/raw\n",
      "Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz to drive/My Drive/mnist/MNIST_data/MNIST/raw/t10k-labels-idx1-ubyte.gz\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "761c7d2b933142d7a5528c8f285059d1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=1, bar_style='info', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Extracting drive/My Drive/mnist/MNIST_data/MNIST/raw/t10k-labels-idx1-ubyte.gz to drive/My Drive/mnist/MNIST_data/MNIST/raw\n",
      "Processing...\n",
      "Done!\n"
     ]
    }
   ],
   "source": [
    "### Run this cell\n",
    "\n",
    "from torchvision import datasets, transforms\n",
    "\n",
    "# Define a transform to normalize the data\n",
    "transform = transforms.Compose([transforms.ToTensor(),\n",
    "                              transforms.Normalize((0.5,), (0.5,)),\n",
    "                              ])\n",
    "\n",
    "# Download and load the training data\n",
    "trainset = datasets.MNIST('drive/My Drive/mnist/MNIST_data/', download=True, train=True, transform=transform)\n",
    "valset = datasets.MNIST('drive/My Drive/mnist/MNIST_data/', download=True, train=False, transform=transform)\n",
    "trainloader = torch.utils.data.DataLoader(trainset, batch_size=64, shuffle=True)\n",
    "valloader = torch.utils.data.DataLoader(valset, batch_size=64, shuffle=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "GcAfrn2falkK"
   },
   "source": [
    "### Exploring The Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 68
    },
    "colab_type": "code",
    "id": "xOjlOyjcCezX",
    "outputId": "fde25724-ea30-46f6-cf0f-4a7807c2ee0e"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<class 'torch.Tensor'>\n",
      "torch.Size([64, 1, 28, 28])\n",
      "torch.Size([64])\n"
     ]
    }
   ],
   "source": [
    "dataiter = iter(trainloader)\n",
    "images, labels = dataiter.next()\n",
    "print(type(images))\n",
    "print(images.shape)\n",
    "print(labels.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 265
    },
    "colab_type": "code",
    "id": "EuBvOWmGDHOq",
    "outputId": "127e0264-be67-4f12-b03f-dac28b7280af"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfcAAAHwCAYAAAC7cCafAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAAWJQAAFiUBSVIk8AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8QZhcZAAAcR0lEQVR4nO3dfaxtdXkn8O8jGLilvCippa3TgFSE1gJzr20pFOQlIrZRr+VlbCOSRmvbqWOhQJxWqdfWSUzagC84aqqWBMlgg6mN06sykSugYI2XKmNE0QIypioCAiIX5eU3f+x12tvTc+7L3vuefc5vfz7Jzjp7rfXs33MXi/M9a++116rWWgCAfjxl1g0AANMl3AGgM8IdADoj3AGgM8IdADoj3AGgM8IdADoj3AGgM8IdADoj3AGgM8IdADoj3AGgM3vPuoE9oaruTHJAkrtm3AoAjOvQJA+11g7b3cIuwz3JAevWrXv6UUcd9fRZNwIA47jtttuybdu2sWp7Dfe7jjrqqKdv3bp11n0AwFg2bNiQW2655a5xamf6mXtVPbOqPlBV/1JVP6yqu6rqbVX1tFn2BQBr2cyO3Kvq8CQ3JXlGkr9P8pUkv5zkj5KcUVUntNbum1V/ALBWzfLI/X9mFOyva61tbK3999baqUkuS/KcJP9jhr0BwJo1k3AfjtpPz+hs9nctWvymJD9Icm5V7bfCrQHAmjert+VPGabXttae3H5Ba+37VfWZjML/uCSfXO5Fqmq5M+aOnEqXALAGzept+ecM09uXWf61YXrECvQCAF2Z1ZH7gcP0wWWWL8w/aEcv0lrbsNT84Yh+/XitAcDa5vKzANCZWYX7wpH5gcssX5j/wAr0AgBdmVW4f3WYLveZ+rOH6XKfyQMAy5hVuG8ZpqdX1b/roar2T3JCkkeSfHalGwOAtW4m4d5a++ck12Z0x5s/XLT4zUn2S3Jla+0HK9waAKx5s7xxzH/N6PKz76iq05LcluRXMvoO/O1J3jDD3gBgzZrZ2fLD0fvzklyRUahfmOTwJG9PcpzrygPAeGZ6y9fW2v9L8juz7AEAeuN77gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ2ZWbhX1V1V1ZZ5fHtWfQHAWrf3jMd/MMnblpj/8Eo3AgC9mHW4P9Ba2zTjHgCgKz5zB4DOzPrIfZ+qekWSn03ygyS3JrmhtfbEbNsCgLVr1uF+SJIrF827s6p+p7V2/c6Kq2rrMouOnLgzAFijZvm2/N8kOS2jgN8vyS8meW+SQ5N8rKqOmV1rALB2zezIvbX25kWzvpTk96vq4SQXJtmU5GU7eY0NS80fjujXT6FNAFhzVuMJde8ZpifNtAsAWKNWY7h/d5juN9MuAGCNWo3hftwwvWOmXQDAGjWTcK+qo6rqPxyZV9WhSS4fnn5wJXsCgF7M6oS6/5Lkwqq6Ick3knw/yeFJfiPJvkk2J/mrGfUGAGvarMJ9S5LnJPnPSU7I6PP1B5J8OqPvvV/ZWmsz6g0A1rSZhPtwgZqdXqQGANh9q/GEOgBgAsIdADoj3AGgM8IdADoj3AGgM8IdADoj3AGgM8IdADoj3AGgM8IdADoj3AGgM8IdADoj3AGgM8IdADoj3AGgMzO5nzvApO65556xa5/xjGdMsZPd89hjj01Uv2XLlil1wq46/fTTZ93CbnPkDgCdEe4A0BnhDgCdEe4A0BnhDgCdEe4A0BnhDgCdEe4A0BnhDgCdEe4A0BnhDgCdEe4A0BnhDgCdEe4A0Bm3fIU59uSTT45d+9GPfnSisV/zmtdMVP+9731v7Nq99tprorFn6dFHHx279qlPfepEY++3334T1U/ikEMOGbv23HPPnWhst3wFAGZOuANAZ4Q7AHRGuANAZ4Q7AHRGuANAZ4Q7AHRGuANAZ4Q7AHRGuANAZ4Q7AHRGuANAZ4Q7AHRGuANAZ4Q7AHTG/dxhQj/60Y8mqv/85z8/du0Xv/jFmY39gQ98YKKxJ3XQQQeNXXvEEUdMNPY555wzdu13vvOdicae5J7qGzdunGjsY445ZqJ6Vo4jdwDojHAHgM4IdwDojHAHgM4IdwDojHAHgM4IdwDojHAHgM4IdwDojHAHgM4IdwDojHAHgM4IdwDojHAHgM645Ssk2bZt29i1r33taycae5Jbp77hDW+YaOzXv/71Y9ceeOCBE439xBNPTFT/J3/yJ2PXHnLIIRONDaudI3cA6MxUwr2qzqqqd1bVjVX1UFW1qvrgTmqOr6rNVXV/VW2rqlur6vyq2msaPQHAvJrW2/JvTHJMkoeTfDPJkTtauapemuTDSR5N8qEk9yd5cZLLkpyQ5Owp9QUAc2dab8tfkOSIJAck+YMdrVhVByT56yRPJDm5tfaq1trFSY5NcnOSs6rq5VPqCwDmzlTCvbW2pbX2tdZa24XVz0ryE0mubq19frvXeDSjdwCSnfyBAAAsbxYn1J06TD++xLIbkjyS5Piq2mflWgKAfsziq3DPGaa3L17QWnu8qu5M8gtJnpXkth29UFVtXWbRDj/zB4CezeLIfeHLsQ8us3xh/kEr0AsAdGdNX8SmtbZhqfnDEf36FW4HAFaFWRy5LxyZL3d5q4X5D6xALwDQnVmE+1eH6RGLF1TV3kkOS/J4kjtWsikA6MUswv26YXrGEstOSvJjSW5qrf1w5VoCgH7MItyvSXJvkpdX1fMWZlbVvkneMjx99wz6AoAuTOWEuqramGTj8HThdku/WlVXDD/f21q7KElaaw9V1e9mFPKfqqqrM7r87Esy+prcNRldkhYAGMO0zpY/Nsl5i+Y9a3gkyTeSXLSwoLX2kap6fpI3JDkzyb5Jvp7kj5O8YxevdAcALGEq4d5a25Rk027WfCbJr09jfJjUa17zmrFrN2/ePNHY119//di1J5544kRjV9XYtZdeeulEYwN7jvu5A0BnhDsAdEa4A0BnhDsAdEa4A0BnhDsAdEa4A0BnhDsAdEa4A0BnhDsAdEa4A0BnhDsAdEa4A0BnhDsAdGZa93OHmbr77rsnqv/Qhz40du1VV1010dgnnXTSRPUAizlyB4DOCHcA6IxwB4DOCHcA6IxwB4DOCHcA6IxwB4DOCHcA6IxwB4DOCHcA6IxwB4DOCHcA6IxwB4DOCHcA6IxwB4DOuJ87Xbjvvvsmqn/iiSfGrt2yZctEY//ar/3a2LU/9VM/NdHYQJ8cuQNAZ4Q7AHRGuANAZ4Q7AHRGuANAZ4Q7AHRGuANAZ4Q7AHRGuANAZ4Q7AHRGuANAZ4Q7AHRGuANAZ4Q7AHTGLV/pws///M9PVH/99dePXXviiSdONPZVV101du3hhx8+0dgXX3zx2LW/9Vu/NdHY8+qxxx6bqP7yyy8fu/bUU0+daOxjjjlmonpWjiN3AOiMcAeAzgh3AOiMcAeAzgh3AOiMcAeAzgh3AOiMcAeAzgh3AOiMcAeAzgh3AOiMcAeAzgh3AOiMcAeAzgh3AOhMtdZm3cPUVdXW9evXr9+6deusW2EOPPzwwxPVX3rppWPXvuUtb5lo7EnuLX7SSSdNNPa11147Uf0+++wzUT2sdhs2bMgtt9xyS2ttw+7WOnIHgM5MJdyr6qyqemdV3VhVD1VVq6oPLrPuocPy5R5XT6MnAJhXe0/pdd6Y5JgkDyf5ZpIjd6Hmi0k+ssT8L02pJwCYS9MK9wsyCvWvJ3l+ki27UPOF1tqmKY0PAAymEu6ttX8N86qaxksCAGOa1pH7OH66qn4vycFJ7ktyc2vt1t15gapa7nT4XflYAAC6NMtwf8Hw+FdV9akk57XW7p5JRwDQgVmE+yNJ/iKjk+nuGOYdnWRTklOSfLKqjm2t/WBnL7Tcd/+GI/r1U+kWANaYFf+ee2vtntban7XWbmmtPTA8bkhyepJ/TPJzSV690n0BQC9WzUVsWmuPJ3nf8HSyS18BwBxbNeE++O4w3W+mXQDAGrbawv24YXrHDtcCAJa14uFeVeur6j+MW1WnZXQxnCRZ8tK1AMDOTeVs+aramGTj8PSQYfqrVXXF8PO9rbWLhp8vTfLsqropo6vaJaOz5U8dfr6ktXbTNPoCgHk0ra/CHZvkvEXznjU8kuQbSRbC/cokL0vyS0lelOSpSb6T5G+TXN5au3FKPQHAXJrW5Wc3ZfQ99V1Z9/1J3j+NcWE1+PEf//GJ6i+44IKdr7SMo48+eqKxL7zwwrFr/+mf/mmisd/xjndMVH/xxRdPVA89W20n1AEAExLuANAZ4Q4AnRHuANAZ4Q4AnRHuANAZ4Q4AnRHuANAZ4Q4AnRHuANAZ4Q4AnRHuANAZ4Q4AnRHuANCZad3PHRjT/vvvP3btxo0bJxr72GOPHbv2zDPPnGjsN73pTRPV/8zP/MzYtb/927890diw2jlyB4DOCHcA6IxwB4DOCHcA6IxwB4DOCHcA6IxwB4DOCHcA6IxwB4DOCHcA6IxwB4DOCHcA6IxwB4DOCHcA6IxwB4DOuJ87U7Nt27aJ6q+99tqxa1/84hdPNPZTnjKff+ceeuihY9fefPPNE419+OGHT1T/l3/5l2PXnnPOORONvffefnWyus3nbzQA6JhwB4DOCHcA6IxwB4DOCHcA6IxwB4DOCHcA6IxwB4DOCHcA6IxwB4DOCHcA6IxwB4DOCHcA6IxwB4DOuG8hU/Oxj31sovqrrrpq7NoXvvCFE4297777TlQ/jz73uc9NVP/MZz5zovp169aNXTvp7Yn333//iephT3PkDgCdEe4A0BnhDgCdEe4A0BnhDgCdEe4A0BnhDgCdEe4A0BnhDgCdEe4A0BnhDgCdEe4A0BnhDgCdEe4A0BnhDgCdcT93puazn/3sRPWbN28eu/aiiy6aaOwDDjhg7NozzjhjorFn6brrrhu79l3vetdEYx988MET1b/yla8cu9b92OndxEfuVXVwVb26qv6uqr5eVduq6sGq+nRVvaqqlhyjqo6vqs1Vdf9Qc2tVnV9Ve03aEwDMs2kcuZ+d5N1JvpVkS5K7k/xkkt9M8r4kL6qqs1trbaGgql6a5MNJHk3yoST3J3lxksuSnDC8JgAwhmmE++1JXpLkH1prTy7MrKo/TfK5JGdmFPQfHuYfkOSvkzyR5OTW2ueH+ZckuS7JWVX18tba1VPoDQDmzsRvy7fWrmutfXT7YB/mfzvJe4anJ2+36KwkP5Hk6oVgH9Z/NMkbh6d/MGlfADCv9vTZ8o8N08e3m3fqMP34EuvfkOSRJMdX1T57sjEA6NUeO1u+qvZOsnA66/ZB/pxhevvimtba41V1Z5JfSPKsJLftZIytyyw6cve6BYB+7Mkj97cmeW6Sza21T2w3/8Bh+uAydQvzD9pTjQFAz/bIkXtVvS7JhUm+kuTcPTFGkrTWNiwz/tYk6/fUuACwmk39yL2qXpvk7Um+nOSU1tr9i1ZZODI/MEtbmP/AtHsDgHkw1XCvqvOTvDPJlzIK9m8vsdpXh+kRS9TvneSwjE7Au2OavQHAvJhauFfV6zO6CM0XMgr2e5ZZdeF6l0tds/OkJD+W5KbW2g+n1RsAzJOphPtwAZq3Jtma5LTW2r07WP2aJPcmeXlVPW+719g3yVuGp++eRl8AMI8mPqGuqs5L8ucZXXHuxiSvq6rFq93VWrsiSVprD1XV72YU8p+qqqszuvzsSzL6mtw1GV2SFgAYwzTOlj9smO6V5Pxl1rk+yRULT1prH6mq5yd5Q0aXp903ydeT/HGSd2x/HXoAYPdMHO6ttU1JNo1R95kkvz7p+Kwer3jFKyaqf/DB5S59sHNf/vKXJxp727ZtY9dedtllE409S+vWrRu79gUveMFEY19yySUT1T/3uc+dqB56tqcvPwsArDDhDgCdEe4A0BnhDgCdEe4A0BnhDgCdEe4A0BnhDgCdEe4A0BnhDgCdEe4A0BnhDgCdEe4A0BnhDgCdEe4A0JmJ7+cOC44++uiJ6t/73vdOqROA+ebIHQA6I9wBoDPCHQA6I9wBoDPCHQA6I9wBoDPCHQA6I9wBoDPCHQA6I9wBoDPCHQA6I9wBoDPCHQA6I9wBoDPCHQA6I9wBoDPCHQA6I9wBoDPCHQA6I9wBoDPCHQA6I9wBoDPCHQA6I9wBoDPCHQA6I9wBoDPCHQA6I9wBoDPCHQA6I9wBoDPCHQA6I9wBoDPCHQA6I9wBoDPCHQA6I9wBoDPCHQA6I9wBoDPCHQA6I9wBoDPCHQA6I9wBoDPCHQA6I9wBoDPCHQA6I9wBoDPCHQA6M3G4V9XBVfXqqvq7qvp6VW2rqger6tNV9aqqesqi9Q+tqraDx9WT9gQA82zvKbzG2UneneRbSbYkuTvJTyb5zSTvS/Kiqjq7tdYW1X0xyUeWeL0vTaEnAJhb0wj325O8JMk/tNaeXJhZVX+a5HNJzswo6D+8qO4LrbVNUxgfANjOxG/Lt9aua619dPtgH+Z/O8l7hqcnTzoOALBrpnHkviOPDdPHl1j201X1e0kOTnJfkptba7fu4X4AoHt7LNyrau8krxyefnyJVV4wPLav+VSS81prd+/iGFuXWXTkLrYJAN3Zk1+Fe2uS5ybZ3Fr7xHbzH0nyF0k2JHna8Hh+RifjnZzkk1W13x7sCwC6tkeO3KvqdUkuTPKVJOduv6y1dk+SP1tUckNVnZ7k00l+Jcmrk7x9Z+O01jYsM/7WJOt3v3MAWPumfuReVa/NKJi/nOSU1tr9u1LXWns8o6/OJclJ0+4LAObFVMO9qs5P8s6Mvqt+ynDG/O747jD1tjwAjGlq4V5Vr09yWZIvZBTs94zxMscN0zum1RcAzJuphHtVXZLRCXRbk5zWWrt3B+uuX3xJ2mH+aUkuGJ5+cBp9AcA8mviEuqo6L8mfJ3kiyY1JXldVi1e7q7V2xfDzpUmeXVU3JfnmMO/oJKcOP1/SWrtp0r4AYF5N42z5w4bpXknOX2ad65NcMfx8ZZKXJfmlJC9K8tQk30nyt0kub63dOIWeAGBuTRzuw/XhN+3G+u9P8v5JxwUAluZ+7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQGeEOAJ0R7gDQmWqtzbqHqauq+9atW/f0o446atatAMBYbrvttmzbtu3+1trBu1vba7jfmeSAJHcts8qRw/QrK9JQH2yz8dhu47Hddp9tNp7VvN0OTfJQa+2w3S3sMtx3pqq2JklrbcOse1krbLPx2G7jsd12n202nl63m8/cAaAzwh0AOiPcAaAzwh0AOiPcAaAzc3m2PAD0zJE7AHRGuANAZ4Q7AHRGuANAZ4Q7AHRGuANAZ4Q7AHRmrsK9qp5ZVR+oqn+pqh9W1V1V9baqetqse1uthm3Ulnl8e9b9zUpVnVVV76yqG6vqoWF7fHAnNcdX1eaqur+qtlXVrVV1flXttVJ9z9rubLeqOnQH+16rqqtXuv9ZqKqDq+rVVfV3VfX1Yd95sKo+XVWvqqolf4/P+/62u9utt/1t71k3sFKq6vAkNyV5RpK/z+jevb+c5I+SnFFVJ7TW7pthi6vZg0netsT8h1e6kVXkjUmOyWgbfDP/dk/oJVXVS5N8OMmjST6U5P4kL05yWZITkpy9J5tdRXZruw2+mOQjS8z/0hT7Ws3OTvLuJN9KsiXJ3Ul+MslvJnlfkhdV1dltuyuS2d+SjLHdBn3sb621uXgk+USSluS/LZp/6TD/PbPucTU+ktyV5K5Z97HaHklOSfLsJJXk5GEf+uAy6x6Q5J4kP0zyvO3m75vRH5wtyctn/W9ahdvt0GH5FbPue8bb7NSMgvkpi+YfklFgtSRnbjff/jbedutqf5uLt+WHo/bTMwqqdy1a/KYkP0hyblXtt8KtsUa11ra01r7Wht8KO3FWkp9IcnVr7fPbvcajGR3JJskf7IE2V53d3G4kaa1d11r7aGvtyUXzv53kPcPTk7dbZH/LWNutK/Pytvwpw/TaJf5Df7+qPpNR+B+X5JMr3dwasE9VvSLJz2b0h9CtSW5orT0x27bWjFOH6ceXWHZDkkeSHF9V+7TWfrhyba0ZP11Vv5fk4CT3Jbm5tXbrjHtaLR4bpo9vN8/+tnNLbbcFXexv8xLuzxmmty+z/GsZhfsREe5LOSTJlYvm3VlVv9Nau34WDa0xy+5/rbXHq+rOJL+Q5FlJblvJxtaIFwyPf1VVn0pyXmvt7pl0tApU1d5JXjk83T7I7W87sIPttqCL/W0u3pZPcuAwfXCZ5QvzD1qBXtaav0lyWkYBv1+SX0zy3ow+n/pYVR0zu9bWDPvfeB5J8hdJNiR52vB4fkYnR52c5JNz/lHaW5M8N8nm1tontptvf9ux5bZbV/vbvIQ7Y2qtvXn47Oo7rbVHWmtfaq39fkYnIq5Lsmm2HdKr1to9rbU/a63d0lp7YHjckNG7bP+Y5OeSvHq2Xc5GVb0uyYUZfevn3Bm3s2bsaLv1tr/NS7gv/KV64DLLF+Y/sAK99GLhhJSTZtrF2mD/m6LW2uMZfZUpmcP9r6pem+TtSb6c5JTW2v2LVrG/LWEXttuS1ur+Ni/h/tVhesQyy589TJf7TJ7/6LvDdM28TTVDy+5/w+d/h2V0Ys8dK9nUGjeX+19VnZ/knRl95/qU4czvxexvi+zidtuRNbe/zUu4bxmmpy9xVaL9M7qowyNJPrvSja1hxw3TufkFMYHrhukZSyw7KcmPJblpjs9cHsfc7X9V9fqMLkLzhYwC6p5lVrW/bWc3ttuOrLn9bS7CvbX2z0muzegksD9ctPjNGf01dmVr7Qcr3NqqVlVHLXUCSVUdmuTy4ekOL7lKkuSaJPcmeXlVPW9hZlXtm+Qtw9N3z6Kx1ayq1i91adWqOi3JBcPTudj/quqSjE4E25rktNbavTtY3f422J3t1tv+VvNyLYklLj97W5Jfyeg78LcnOb65/Oy/U1WbMjr55IYk30jy/SSHJ/mNjK52tTnJy1prP5pVj7NSVRuTbByeHpLkhRn9VX/jMO/e1tpFi9a/JqPLgV6d0eVAX5LR15auSXLOPFzYZXe22/D1o2dn9P/tN4flR+ffvsd9SWttIay6VVXnJbkiyRMZvbW81Fnwd7XWrtiuZu73t93dbt3tb7O+RN5KPpL8p4y+2vWtJD/KKLDeluRps+5tNT4y+hrI/8rozNIHMrrww3eT/J+Mvidas+5xhttmU0aXqlzucdcSNSdk9AfR95JsS/J/Mzoi2GvW/57VuN2SvCrJ/87oypIPZ3Q51bszulb6ibP+t6yibdaSfMr+Ntl2621/m5sjdwCYF3PxmTsAzBPhDgCdEe4A0BnhDgCdEe4A0BnhDgCdEe4A0BnhDgCdEe4A0BnhDgCdEe4A0BnhDgCdEe4A0BnhDgCdEe4A0BnhDgCdEe4A0Jn/D8sPiawfqBBMAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "image/png": {
       "height": 248,
       "width": 251
      },
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(images[0].numpy().squeeze(), cmap='gray_r');"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 265
    },
    "colab_type": "code",
    "id": "F9CppCcqDLtB",
    "outputId": "0f59838b-90b6-4370-e23a-0e4a0ba90c2a"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 60 Axes>"
      ]
     },
     "metadata": {
      "image/png": {
       "height": 229,
       "width": 349
      },
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "figure = plt.figure()\n",
    "num_of_images = 60\n",
    "for index in range(1, num_of_images + 1):\n",
    "    plt.subplot(6, 10, index)\n",
    "    plt.axis('off')\n",
    "    plt.imshow(images[index].numpy().squeeze(), cmap='gray_r')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "lGyau0mOaP2m"
   },
   "source": [
    "### Defining The Neural Network"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "n-NR96UtFSkB"
   },
   "source": [
    "![](https://raw.githubusercontent.com/dmlc/web-data/master/mxnet/image/mlp_mnist.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 153
    },
    "colab_type": "code",
    "id": "3WJXInzQGcAy",
    "outputId": "be31aea8-542d-46ab-9816-85bc418664f9"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sequential(\n",
      "  (0): Linear(in_features=784, out_features=128, bias=True)\n",
      "  (1): ReLU()\n",
      "  (2): Linear(in_features=128, out_features=64, bias=True)\n",
      "  (3): ReLU()\n",
      "  (4): Linear(in_features=64, out_features=10, bias=True)\n",
      "  (5): LogSoftmax()\n",
      ")\n"
     ]
    }
   ],
   "source": [
    "from torch import nn\n",
    "\n",
    "# Layer details for the neural network\n",
    "input_size = 784\n",
    "hidden_sizes = [128, 64]\n",
    "output_size = 10\n",
    "\n",
    "# Build a feed-forward network\n",
    "model = nn.Sequential(nn.Linear(input_size, hidden_sizes[0]),\n",
    "                      nn.ReLU(),\n",
    "                      nn.Linear(hidden_sizes[0], hidden_sizes[1]),\n",
    "                      nn.ReLU(),\n",
    "                      nn.Linear(hidden_sizes[1], output_size),\n",
    "                      nn.LogSoftmax(dim=1))\n",
    "print(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "oxSLypv2LOD-"
   },
   "outputs": [],
   "source": [
    "criterion = nn.NLLLoss()\n",
    "images, labels = next(iter(trainloader))\n",
    "images = images.view(images.shape[0], -1)\n",
    "\n",
    "logps = model(images)\n",
    "loss = criterion(logps, labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 187
    },
    "colab_type": "code",
    "id": "Pj4I2lLgLVWw",
    "outputId": "f9032f51-5acb-431a-cc42-4de0abe331a5"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Before backward pass: \n",
      " None\n",
      "After backward pass: \n",
      " tensor([[ 2.3738e-03,  2.3738e-03,  2.3738e-03,  ...,  2.3738e-03,\n",
      "          2.3738e-03,  2.3738e-03],\n",
      "        [ 1.8307e-05,  1.8307e-05,  1.8307e-05,  ...,  1.8307e-05,\n",
      "          1.8307e-05,  1.8307e-05],\n",
      "        [ 2.1192e-04,  2.1192e-04,  2.1192e-04,  ...,  2.1192e-04,\n",
      "          2.1192e-04,  2.1192e-04],\n",
      "        ...,\n",
      "        [ 1.4289e-03,  1.4289e-03,  1.4289e-03,  ...,  1.4289e-03,\n",
      "          1.4289e-03,  1.4289e-03],\n",
      "        [-5.3521e-04, -5.3521e-04, -5.3521e-04,  ..., -5.3521e-04,\n",
      "         -5.3521e-04, -5.3521e-04],\n",
      "        [-1.9371e-03, -1.9371e-03, -1.9371e-03,  ..., -1.9371e-03,\n",
      "         -1.9371e-03, -1.9371e-03]])\n"
     ]
    }
   ],
   "source": [
    "print('Before backward pass: \\n', model[0].weight.grad)\n",
    "\n",
    "loss.backward()\n",
    "\n",
    "print('After backward pass: \\n', model[0].weight.grad)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "F0ZVHVbvI_yt"
   },
   "outputs": [],
   "source": [
    "from torch import optim\n",
    "\n",
    "# Optimizers require the parameters to optimize and a learning rate\n",
    "optimizer = optim.SGD(model.parameters(), lr=0.01, momentum=0.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 289
    },
    "colab_type": "code",
    "id": "dS9JqXhhLdkr",
    "outputId": "4ec0a101-bc31-45ee-ea9b-39804b6cab16"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Initial weights -  Parameter containing:\n",
      "tensor([[ 0.0301, -0.0045,  0.0166,  ..., -0.0314,  0.0094,  0.0293],\n",
      "        [ 0.0233, -0.0306, -0.0328,  ..., -0.0197, -0.0138, -0.0075],\n",
      "        [-0.0308, -0.0263, -0.0259,  ..., -0.0062,  0.0017,  0.0052],\n",
      "        ...,\n",
      "        [ 0.0297,  0.0309, -0.0332,  ...,  0.0275, -0.0031, -0.0256],\n",
      "        [ 0.0137,  0.0205,  0.0031,  ..., -0.0323,  0.0347, -0.0357],\n",
      "        [ 0.0184,  0.0231, -0.0339,  ..., -0.0339,  0.0036,  0.0322]],\n",
      "       requires_grad=True)\n",
      "Gradient - tensor([[-0.0020, -0.0020, -0.0020,  ..., -0.0020, -0.0020, -0.0020],\n",
      "        [-0.0001, -0.0001, -0.0001,  ..., -0.0001, -0.0001, -0.0001],\n",
      "        [ 0.0010,  0.0010,  0.0010,  ...,  0.0010,  0.0010,  0.0010],\n",
      "        ...,\n",
      "        [ 0.0015,  0.0015,  0.0015,  ...,  0.0015,  0.0015,  0.0015],\n",
      "        [ 0.0022,  0.0022,  0.0022,  ...,  0.0022,  0.0022,  0.0022],\n",
      "        [ 0.0011,  0.0011,  0.0011,  ...,  0.0011,  0.0011,  0.0011]])\n"
     ]
    }
   ],
   "source": [
    "print('Initial weights - ', model[0].weight)\n",
    "\n",
    "images, labels = next(iter(trainloader))\n",
    "images.resize_(64, 784)\n",
    "\n",
    "# Clear the gradients, do this because gradients are accumulated\n",
    "optimizer.zero_grad()\n",
    "\n",
    "# Forward pass, then backward pass, then update weights\n",
    "output = model(images)\n",
    "loss = criterion(output, labels)\n",
    "loss.backward()\n",
    "print('Gradient -', model[0].weight.grad)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 170
    },
    "colab_type": "code",
    "id": "wy0KOQ95LgYN",
    "outputId": "d742c1e9-50e2-47f0-ef0b-bdeea6064390"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Updated weights -  Parameter containing:\n",
      "tensor([[ 0.0301, -0.0045,  0.0167,  ..., -0.0313,  0.0094,  0.0293],\n",
      "        [ 0.0233, -0.0306, -0.0328,  ..., -0.0197, -0.0138, -0.0075],\n",
      "        [-0.0308, -0.0263, -0.0259,  ..., -0.0062,  0.0017,  0.0052],\n",
      "        ...,\n",
      "        [ 0.0297,  0.0309, -0.0332,  ...,  0.0275, -0.0031, -0.0256],\n",
      "        [ 0.0137,  0.0204,  0.0030,  ..., -0.0323,  0.0347, -0.0357],\n",
      "        [ 0.0184,  0.0230, -0.0339,  ..., -0.0339,  0.0036,  0.0322]],\n",
      "       requires_grad=True)\n"
     ]
    }
   ],
   "source": [
    "# Take an update step and few the new weights\n",
    "optimizer.step()\n",
    "print('Updated weights - ', model[0].weight)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "wstRGu4FaJBe"
   },
   "source": [
    "### Core Training Of Neural Network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "import tqdm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 306
    },
    "colab_type": "code",
    "id": "XCsoAdjdLjPb",
    "outputId": "6e7a5f80-f945-4e5c-c538-8ef445b6ad3e"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 0 - Training loss: 0.6382603662561125\n",
      "Epoch 1 - Training loss: 0.2859742457448229\n",
      "Epoch 2 - Training loss: 0.22385910253868554\n",
      "Epoch 3 - Training loss: 0.1810971234601412\n",
      "Epoch 4 - Training loss: 0.15141980831962087\n",
      "Epoch 5 - Training loss: 0.12902044818035638\n",
      "Epoch 6 - Training loss: 0.11346033318980989\n",
      "Epoch 7 - Training loss: 0.10111659060794312\n",
      "Epoch 8 - Training loss: 0.09049932678132805\n",
      "Epoch 9 - Training loss: 0.08147087670973878\n",
      "Epoch 10 - Training loss: 0.07475941698761907\n",
      "Epoch 11 - Training loss: 0.06735960688582981\n",
      "Epoch 12 - Training loss: 0.061582124334321156\n",
      "Epoch 13 - Training loss: 0.05887969804387579\n",
      "Epoch 14 - Training loss: 0.0542472667231568\n",
      "\n",
      "Training Time (in minutes) = 2.432295314470927\n"
     ]
    }
   ],
   "source": [
    "optimizer = optim.SGD(model.parameters(), lr=0.003, momentum=0.9)\n",
    "time0 = time()\n",
    "epochs = 15\n",
    "for e in range(epochs):\n",
    "    running_loss = 0\n",
    "    for images, labels in trainloader:\n",
    "        # Flatten MNIST images into a 784 long vector\n",
    "        images = images.view(images.shape[0], -1)\n",
    "    \n",
    "        # Training pass\n",
    "        optimizer.zero_grad()\n",
    "        \n",
    "        output = model(images)\n",
    "        loss = criterion(output, labels)\n",
    "        \n",
    "        #This is where the model learns by backpropagating\n",
    "        loss.backward()\n",
    "        \n",
    "        #And optimizes its weights here\n",
    "        optimizer.step()\n",
    "        \n",
    "        running_loss += loss.item()\n",
    "    else:\n",
    "        print(\"Epoch {} - Training loss: {}\".format(e, running_loss/len(trainloader)))\n",
    "print(\"\\nTraining Time (in minutes) =\",(time()-time0)/60)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "75j9X1b6ME5K"
   },
   "outputs": [],
   "source": [
    "def view_classify(img, ps):\n",
    "    ''' Function for viewing an image and it's predicted classes.\n",
    "    '''\n",
    "    ps = ps.data.numpy().squeeze()\n",
    "\n",
    "    fig, (ax1, ax2) = plt.subplots(figsize=(6,9), ncols=2)\n",
    "    ax1.imshow(img.resize_(1, 28, 28).numpy().squeeze())\n",
    "    ax1.axis('off')\n",
    "    ax2.barh(np.arange(10), ps)\n",
    "    ax2.set_aspect(0.1)\n",
    "    ax2.set_yticks(np.arange(10))\n",
    "    ax2.set_yticklabels(np.arange(10))\n",
    "    ax2.set_title('Class Probability')\n",
    "    ax2.set_xlim(0, 1.1)\n",
    "    plt.tight_layout()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 261
    },
    "colab_type": "code",
    "id": "Ie9Fffl_Mqp6",
    "outputId": "2e93c6ea-0534-498d-e072-904f62591dfe"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Predicted Digit = 0\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x648 with 2 Axes>"
      ]
     },
     "metadata": {
      "image/png": {
       "height": 235,
       "width": 424
      },
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "images, labels = next(iter(valloader))\n",
    "\n",
    "img = images[0].view(1, 784)\n",
    "# Turn off gradients to speed up this part\n",
    "with torch.no_grad():\n",
    "    logps = model(img)\n",
    "\n",
    "# Output of the network are log-probabilities, need to take exponential for probabilities\n",
    "ps = torch.exp(logps)\n",
    "probab = list(ps.numpy()[0])\n",
    "print(\"Predicted Digit =\", probab.index(max(probab)))\n",
    "view_classify(img.view(1, 28, 28), ps)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "wAEvDtiaM6RQ"
   },
   "source": [
    "### Model Evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 68
    },
    "colab_type": "code",
    "id": "5sBPmaBONPkT",
    "outputId": "25b032aa-737f-49d8-d1b5-76f6b9e6b7b6"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number Of Images Tested = 10000\n",
      "\n",
      "Model Accuracy = 0.9687\n"
     ]
    }
   ],
   "source": [
    "correct_count, all_count = 0, 0\n",
    "for images,labels in valloader:\n",
    "  for i in range(len(labels)):\n",
    "    img = images[i].view(1, 784)\n",
    "    # Turn off gradients to speed up this part\n",
    "    with torch.no_grad():\n",
    "        logps = model(img)\n",
    "\n",
    "    # Output of the network are log-probabilities, need to take exponential for probabilities\n",
    "    ps = torch.exp(logps)\n",
    "    probab = list(ps.numpy()[0])\n",
    "    pred_label = probab.index(max(probab))\n",
    "    true_label = labels.numpy()[i]\n",
    "    if(true_label == pred_label):\n",
    "      correct_count += 1\n",
    "    all_count += 1\n",
    "\n",
    "print(\"Number Of Images Tested =\", all_count)\n",
    "print(\"\\nModel Accuracy =\", (correct_count/all_count))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Feature Extraction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sequential(\n",
      "  (0): Linear(in_features=784, out_features=128, bias=True)\n",
      "  (1): ReLU()\n",
      "  (2): Linear(in_features=128, out_features=64, bias=True)\n",
      "  (3): ReLU()\n",
      "  (4): Linear(in_features=64, out_features=10, bias=True)\n",
      "  (5): LogSoftmax()\n",
      ")\n"
     ]
    }
   ],
   "source": [
    "print(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sequential(\n",
      "  (0): Linear(in_features=784, out_features=128, bias=True)\n",
      "  (1): ReLU()\n",
      "  (2): Linear(in_features=128, out_features=64, bias=True)\n",
      ")\n"
     ]
    }
   ],
   "source": [
    "feature_extractor = nn.Sequential(*list(model.children())[:-3])\n",
    "print(feature_extractor)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "correct_count, all_count = 0, 0\n",
    "l_imgs = []\n",
    "l_features = []\n",
    "for images,labels in valloader:\n",
    "  for i in range(len(labels)):\n",
    "    img = images[i].view(1, 784)\n",
    "    l_imgs.append(images[i].numpy())\n",
    "    # Turn off gradients to speed up this part\n",
    "    with torch.no_grad():\n",
    "        features = feature_extractor(img)\n",
    "    l_features.append(features.numpy())\n",
    "#     # Output of the network are log-probabilities, need to take exponential for probabilities\n",
    "#     ps = torch.exp(logps)\n",
    "#     probab = list(ps.numpy()[0])\n",
    "#     pred_label = probab.index(max(probab))\n",
    "#     true_label = labels.numpy()[i]\n",
    "#     if(true_label == pred_label):\n",
    "#       correct_count += 1\n",
    "#     all_count += 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(10000, 1, 28, 28)\n",
      "(10000, 1, 64)\n"
     ]
    }
   ],
   "source": [
    "l_imgs_np = np.array(l_imgs)\n",
    "print(l_imgs_np.shape)\n",
    "# print(len(l_imgs))\n",
    "l_features_np = np.array(l_features)\n",
    "print(l_features_np.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import scipy.io as sio\n",
    "sio.savemat('data_mnist_all_features.mat', {'names':all_idx, 'features':all_features})"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "name": "handwritten_digit_recognition_CPU.ipynb",
   "provenance": [],
   "version": "0.3.2"
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}