MNIST.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch                        # The core package of Torch\n",
    "import torch.nn as nn               # The nn package is used for building neural networks\n",
    "import torch.nn.functional as F     # Contains all the functions in the torch.nn library\n",
    "\n",
    "from sklearn.model_selection import train_test_split\n",
    "import scikitplot as skplt\n",
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train data shape: (42000, 785)\n",
      "Test data shape: (28000, 784)\n"
     ]
    }
   ],
   "source": [
    "# read data\n",
    "train_data = pd.read_csv('./dataset/train.csv')\n",
    "test = pd.read_csv('./dataset/test.csv')\n",
    "\n",
    "# check that we loaded data correctly\n",
    "print(\"Train data shape:\", train_data.shape)\n",
    "print(\"Test data shape:\", test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.LineCollection at 0x221bbe960c8>"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "label_counts  = train_data[\"label\"].value_counts().sort_index()\n",
    "label_counts.plot.bar()\n",
    "plt.hlines(xmin = -1, xmax = 10, y = 4000, linestyles='dotted')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_all = train_data.iloc[:,1:]     # only features (pixels)\n",
    "train_all_label = train_data[\"label\"] # labels\n",
    "\n",
    "# convert to numpy array\n",
    "train_all_numpy = train_all.to_numpy()\n",
    "train_all_label_numpy = train_all_label.to_numpy()\n",
    "test_numpy = test.to_numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAVwAAAD7CAYAAAAmeCzOAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjMsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+AADFEAAAgAElEQVR4nO2dd5hU1fnHP+8uVXqRIl0FBTUGRcFONEZjQ40aibFEEn92sCTYEjV2jcQukqigMSJiN3Yl9gJixQ1FBESRIiAIIuzu+/vj3Lkz7O603ZlbhvfzPDxz7jnn3vsy371nzj3lfUVVMQzDMIpPWdgGGIZhbCpYg2sYhhEQ1uAahmEEhDW4hmEYAWENrmEYRkBYg2sYhhEQoTS4InKgiMwUkTkickEYNhiFx3QtXUzbwiBBr8MVkXJgFrA/sBCYCgxX1c8CNcQoKKZr6WLaFo5GIdxzV2COqs4FEJGJwDAgrXhNpKk2o0VA5kWP1axYpqqbh21HFkzXPImJrpCntqZrel3DaHC7AV+mHC8EBmc6oRktGCz7FdWoKPOSTp4ftg05YLrmSUx0hTy1NV3T6xpGgyt15NUa1xCRU4BTAJqxWbFtMhqO6Vq6ZNXWdM2NMCbNFgI9Uo67A1/XrKSq41R1kKoOakzTwIwz6o3pWrpk1dZ0zY0wGtypQF8R6SMiTYBjgSdDsMMoLKZr6WLaFojAhxRUtVJEzgSeB8qBe1R1RtB2GIXFdC1dTNvCEcYYLqr6DPBMGPc2iofpWrqYtoXBdpoZhmEEhDW4hmEYARHKkMKmQrs32wMwsc8rAOx43el+WZeb3wrFpk2FRl27AKBtWwFQMbJdrTpDf1oBwMf3bu/nNVnlVju1nviuy7CIKJFGGiWbsJl3DnSJavexzRkf+GVaWRmkWWmxHq5hGEZAWA+3wHR+u7WfvqOHm2PYoI0BEOssFYXyDu399OKjtgHg1T//HYDm0iTr+eNHzfbTh7b8HIDdDzkTgH5XrwWgasbMwhhrFBRp3txPzznoro3KDjl3Lz9tPVzDMIxNDGtwDcMwAsKGFArE3Ot3A2Bi9xv9vKbitjgOmT4cgC3Gf+qXVQVoW6lS3rkTAFX/Tg4bvLft7V4q+1BCgpNap+5Sda+oM4feDcCbu7k+yWWn/t6v0ewD55ukaunSfE02NnGsh2sYhhEQ1sNtIMt/53q2bw//GwAty5r5ZTd8OwCAzictA6Bq1aqArSttVuy7JQBvbHtH0e6xR1O3xujFe8f5eT+5zU2odb/GerhRZsHIHf1096ujsQzTeriGYRgBYT3celK+zdYADDtnCgBtvJ7tx+uTo7NP/G1fANp++3bA1uWOiLQHVFVXhG1Lrqw7dFcA+p5dvwgvO4x1PdTNFrl1enueNtUvu7HLe1nPf/a06wE44ts/AtBxXHT13ZTpe8DnfvqHq0M0JIV693BF5JNCGmIEh4j0FJGJIrIUeBeYKiJLvLze4VpnFAt7ZsMnYw9XRI5MVwR0Kbw5RkA8BNwEHKeqVeAHCjwamAgMCdE2owHYMxttsg0pPAQ8QB2hUoBmdeT5iMg9wCHAElXd3str712zNzAPOCZOr7IbfjHIT+9746sAnNv+fxvV+cP1I/305vdF9lWzo6o+lJrhNbwTReSKTCdGQdfKM9wk5L09/5u2zkVLdgLg4Y93qlW29Stu95i8+SEAsya38csO7XwMAP3/PReA67tMq3V+t3IXQqbJEUtcxrhaVcLEntkIk21I4WPgb6r6u5r/gJVZzh0PHFgj7wLgZVXtC7zsHRvB876I3CEig0VkC+/fYBG5A/ggy7njMV2jjD2zESZbD3cUkG4t0xGZTlTV1+oYDxwGDPXSE4D/AqOz2BA6i8/eHYD3R9/m51V7HYhZG9YDMOKz4wHo+thcv040dm/XyQnACOByXERWwUVlfQq4O9OJoekqyTiG5RmcUgy6yk2ItVjiJi/7Tn4366WrVn6XPPDSj7/mRlWuPsad34jyWuf9uuf7ADx4/C/9vLb3h/5WY89shMnY4Krq6xnKar9rZaezqi7yzl8kIp3SVbQooMVDVdcDd3r/CoHpGhHCemZN19zIeVmYiByiqk+nOy40qjoOb3SstbQPxc9Wo949ATjulOfT1jl62h8A6HGU27Yb4V5tncRF1+o9f+qnp2yfvhPe9WU3rlo1c059TQRg63PeAWCPGWcD8O7lt9eqc1Zb9zZz+y9/8PPa3t+g2xaUILUN7XndsMFPHv35AQA8vFX65zVs8lkWtkuW41xYLCJdAbzPJfW4hlFYTNfSxbSNGDn3cFX10kzHOfIkcCJwrff5RD2uUVQSDlEA9n7KRQQY1W6Wl5McR/yich0ALZ5pFZhtxSAuuq7cOv0E++eVyR6mrN+Qtl596PzKInePPyfvsVWj5umqR4q4aNsQqtet89NfTPRWpFwc3R5ufdfhAqCqj2Y490HcYHtHEVkIXIoTbZKIjAAW4NZ9GgFjupYupm20ydbDPTRDmQJpxVPV4WmK9stmlFF0TNfSxbSNMNlWKfwuKEMiQ+uWfrLmpoZURu3k/q7bR9hPQjriqGuzldVpyy5aMMxPVy8urAevyrnzADj2o5P9vKk7P7hRnRt2meynx7Vzw6RVK8LZGxBHbRuCNE76Pf5u1x9DtCQ3cpo0E5HOInK3iDzrHQ/wXjGMGGO6li6mbTTJddJsPHAvcLF3PAu33S/jIvk40ah7NwB2nZzs1ZalTJIBnLNosJ/WH9ZRAown4rqWd+wAwLU3pl8y/NCWL/jpQ3u4rbk0cFlYTZpMSgmzvvPGZYdultxn8I+muUeaKDLjibi2hUCaNfXTs/f/R4iW5Eauy8I6quokvIjvqlqJRYkpBUzX0sW0jSC59nDXiEgHPIcYIjIE+C7zKfFiydgWAFzUMenBLjFqOPLrPQD4Yp/k71P12rWB2VZEIq+rNHYh5oc0zVKxyLT6MvrjgzWIvLabIrk2uOfi1uNtJSJvApsDRxXNKiMoTNfSxbSNIDk1uKo6XUT2AbbBrf6fqaqFXWFuBI7pWrqYttEkpwZXRJoBpwN74l5RXheRsaoa+5mjxGTZ/t1qLwH7vtq9Rr5/y0AA2q6N3xKwTMRB10pvmdfAqcf5eR/s8kBY5sSGOGi7KZLrkMJ9wGrgVu94OHA/tusk7piupYtpG0FybXC3UdUdU46niMhHxTAoKBr16gFAq3+vAeDyTs7v9rKq5J75X/7tTwB0vj8aIZaLQPR1rXYT6zIlZVlWBhcsiUgNFT939Ru6ASHhW2Pf295IW6fflOTy1q0Xf9ig+xWQ6Gu7CZLrsrAPvFlOAERkMPBmcUwyAsR0LV1M2wiSzXnNJ7jxn8bACSKywDvuBdQvRnVEmD/c9XA/6H3rRvmjvzrIT3e+pTR7tnHUtdu/Z/vpK0/eHoBLOn5aq14iBtlFrzjPUW9e6TartHgke+SHVBr16A7A/JtdvLPz2z9Xq86SKrc0cJur1/h5VRqK62afOGq7KZFtSOGQQKwwgsZ0LV1M2wiTzXnN/NRjL7xGxsifUWbJ6bv76UdPu8FLuf/OmV/tCcC3x7VPOSNdaKh4E0ddq5YmndK8crHTqs11roeZiLyQytWdpgNw6p/chpZ5ywbWqtNohRuvr27W2M+rbu4eib29Mdvz289Ma9ORM04EoPVns9LWCZo4atsQ5v6zT8rRq6HZkSu5Oq85TERmA1/g/lfzgGeznNNDRKaISIWIzBCRkV5+exF5UURme5/tMl3HKB6ma+mSr7amazDkOml2BTAEmKWqfXD+MbMNwFcC56lqf+/cM0RkABZ2OUqYrqVLvtqargGQ67KwDar6rYiUiUiZqk4RkesyneBF+kxE+1wtIhW4kNyBh10u33xzAM4f+ZCf16fRxm9Z0+90QQrbzy2tzQ1ZiKWuzZ5+D4D7u7nw5EdefINf1q1844ixY7t7QWz/XTuY7dQf3QTXFo2SSwFrnp+J9Y8nwjF9nvM5AZKXtlHQtT5s13WRny6XfEI0hkOuDe5KEWkJvAY8ICJLyCNArbhY9wOBd7Gwy1HCdC1d6q2t6Vo8cm1whwHrgHOA44A2wF9zOdET/RFglKquEpFspwCFDbv81W/6AnBMy9pLexKsb52bXSVGrHXteJd7G/lFtz/6eTNG1A5nno5dmiZszt5AzNqQ3BH722vOA6DzQ26VVUR9HtZL2yjoWl+qNH1UkKiQq/OaNSmHE3K9uIg0xon3QErwusUi0tX7tbSwyyFiupYu9dHWdC0+2TY+rMbzp1mzCFBVbZ3hXMF5l69Q1TEpRYGHXS7zfCRt0GRfpLGUA/Cj50Bp9VaurEuxjYkApaJrgi1vTi7dGrb3wQA80fc/Bbn2V97mhhGjz/fzOj7ketZR7NnWV9so6lqKZFuH26oB194DOB74REQSG8wvwsIuh47pWro0QFvTNQByHcPNG1V9A0g3AGRhl2OK6VqamK7BULQGN0p0usP5RLj3zK38vBZlztft38c6J/h9bypNvwmbAlXfLvfTerDbWbb7kWcAsHS/9cDGAQYTy4cSkyypy4m2fMF5/up/sVtupOvdkFOrpe8UxXajYSwbk7LTrMZ86fIxvfx0cxYHZFFmor9wzTAMo0TYJHq4CZ4c0KFWXhesZ1tKVK9xk/Nt73/b+3T5B7FTTuf35X0gj8XIRqg0f/w9P33Q4xtr3Jz3alYPHevhGoZhBIQ1uIZhGAFhDa5hGEZAWINrGIYRENbgGoZhBIQ1uIZhGAFhDa5hGEZAWINrGIYREKIhh3XOBRFZCqwBloVtSz3oSMPt7qWqmxfCmChhupquEaSousaiwQUQkWmqOihsO/IlrnYHRVy/n7jaHRRx/X6KbbcNKRiGYQSENbiGYRgBEacGd1zYBtSTuNodFHH9fuJqd1DE9fspqt2xGcM1DMOIO3Hq4RqGYcQaa3ANwzACIvINrogcKCIzRWSOiFwQtj3pEJEeIjJFRCpEZIaIjPTy24vIiyIy2/tsF7atUSEO2pqu+WO6ZrhvlMdwRaQcmAXsDywEpgLDVfWzUA2rAxHpCnRV1eki0gp4HzgcOAlYrqrXen987VR1dIimRoK4aGu65ofpmpmo93B3Beao6lxVXQ9MBIaFbFOdqOoiVZ3upVcDFUA3nL0TvGoTcKIaMdHWdM0b0zUDoTS4ebxydAO+TDle6OVFGhHpDQwE3gU6q+oicCIDncKzrLjk+SoZO203VV2htJ/ZIHUNvMH1XjluB34JDACGi8iAdNXryIvuGAggIi2BR4BRqroqbHuCIk9dIWbabqq6Qmk/s0HrGvgYrojsBlymqgd4xxcCqOo16eo2pskvmtEiUDujxGpWLIu6k5N8dE3Ub0yTt0zXaOsK+T+zpmt6XcMIk17XK8fgmpVE5BTgFGCHchoxWPYLyLzo8ZJOnh+2DTmQr66YrrHQFXLQ1nRNkknXMMZwc3rlUNVxnteeIxrTtPhWGQ0lL11VdZDpGhuyamu65kYYDe5CoEfKcXfg63SVVfWZoltkFIK8dDVihWlbIMJocKcCfUWkj4g0AY4FngzBDqOwmK6li2lbIAIfw1XVShE5E3geKAfuUdUZQdthFBbTtXQxbQtHGJNmiWECGyooMUzX0sW0LQxR32lmGIZRMliDaxiGERChDCkYRtEQt4KpUZ9eAFSM6gxA484/+FVm7nVf2tP7v3k8AH3+sg6AqplzXUF1VcFNNRqONEo2YTPvHAjAMYOmAnB1p+l+2Slf7g3Al+du5c5766OgTNwI6+EahmEEREn3cE+Y6TbH3LdwNwDKDk6Gm69et64g9yhr1QqA5UdsD0Db+94uyHWN3ClrkdxG+uVZOwLw0Vm3pa1flWE3+6e7e46iXnIf2919BgC9LnsvWcl6u6EjTd3misWT+vh5cwbdBcBB/zsMgIHz+/tlrw+6B4DVE58D4OjR5wPQ+sF3im9sCtbDNQzDCAhrcA3DMAKipIcUHjh0qPt82U2SnNj2SL+s+pvCDClIF+cUaOg5bijhw/TzMUaBKW/bBoB9Xk/uMj2/ffqhhPowY8TtAOzw45l+Xo+r3iroPYz8mfl3N3Q0Z9BYP6/flBEAbP3bDwDomlL/tDcPAuD+3i8D8NB1fwPg1I9O9utUfTaraPYmsB6uYRhGQJR0D7dq1ucArK52sySzb+rsl/U5dnFB75VYgvKzw0/185o//l666kYh2MLpeX77KUW/1ajfPu6nb93goq50G+P01crKot/fcHw7wk2ATz/0RgD+tnxHv6zvCBc2ra450RlLuwDwyRYbANihyWYAfD68g1+n958Lbm4trIdrGIYREHn3cEVkp0TwtbhwyLT/A+CEAcke55vN2gKFWx6WQMvqch0aXUSkPaCquiJsW3KlUe+eAOw3aVrWut/rjwDsOXWEn3fZdk9vVGef5ov8dLuy5nVeZ0Trhcn0SDdOfPArJ7iMaZ/mYHXwxFHbdCSW/v3mnOcBaF3WDIBnRw/16zT9cWra87scXgHAhdufCMDoJyYBcNqRz/p1nr/K9YIL3SakkrHBFZGdamYBT4jIobjwPLFqeA2HiPQErgf2A1a6LGkNvAJcoKrzQjTPaACmbbTJ1sOdBrwD/JiS1wEYgxsq2bdIdhnF5SHgJuA4Va0CP1Dg0biw1kNCtM1oGKZthMnW4B4DnAXckIi8ICJfqOrPsl1YRO4BDgGWqOr2Xl573B9Eb2AecEwQrzvr5rvdYBcO+czPO2xztxul+suFdZ6TK/KD+y2ataF4ryFFoKOqPpSa4T2cE0XkikwnRkHXzy5wr35Ptns8bZ3X17k/7dGXu51iW6TsABzHlhvVvfHXv/HTp14+GYDjWi3JaseR978CwJ23H+7ndbot9CVjsdY2HZ9f8hMAnm73OgAD3jgJgD4vfODXySUcbtm3Kzc6PqvtXD/9/Nbeb9Gn/6u/odnun6lQVScDBwP7i8jD3utKrmF+xwMH1si7AHhZVfsCL3vHRvC8LyJ3iMhgEdnC+zdYRO4APshy7nhM1yhj2kaYrJNmqvo9cI6IDAQmAC1zubCqviYivWtkDwOGeukJwH+B0bmZWn86fuhNZP268NeuXPgVADctiVWU0hOAEcDluIisgovK+hRwd6YTw9I1sXceYPTe/8laf9LyXYHcfFu0eii5n/66LY9x5508HoCDN/s+7Xn+RNoZyZ724y/uDkDVzDlZ71skYqdtOspbt/bTpx/27EZlW17tlndV57kkr7qDmyzfq1k4S/lyXqWgqh+IyL5Aqwbcr7OqLvKut0hEOqWrmBp2uRmbNeCWRk1UdT1wp/evEJiuESEsbU3X3Mi5wRWRQ1T1aWBVjeOioKrjgHEAraV9rsMYdVL+Y4NOz4uFByU9SfV7NLDb1puaOkZV1y/+nFww84c26Xutq6rdWPr0m38KQBvy8wbV/Ro3Bjv24QMAeH3SPACu7fx+2nNSl4yVP/4GAA//xC1d0w3r87p/IQlS20I+rwk+/9N2fvqstv8FoP9rvwNgyxnxDKmWz8aHXbIc58JiEekK4H1mn5kwio3pWrqYthEjnyGFSzMd58iTwInAtd7nE/W4Rt40/c71On/U4o/b3Dn0fj/9d/pnqBkN4qLrzJOTb8iZ/Nle/I1bqdjmXw3zc1o15wsAZhzj/K1eMClZlqm3e1Jr50jn4bJeDbp/IYiLtumoalo7T+a44Yr6bqdecHn5Rsepq4vKvl8LQHW9rpwb2TY+HJmpXFXTvjSLyIO4wfaOIrIQuBQn2iQRGQEswK0NNALGdC1dTNtok62He2iGMgXSiqeqw9MUxWo6v0QxXUsX0zbCZGxwVfV3QRlSTJo85/ZYP712cz9v1nUdAdjqd0sB0B9/rH1iHkx5xU3SnDf8JT+vvEN7AKq+Xd6gaxeaUtG1Ll5/yE2ubUFhNiAkhhY+O6q3n/fOS25IYUgdr7wJpL8LVqgffpa+UhEoJW2P+fmbtfK2nOw2LtT3tf9nPTdernf258m1omXzFtTzqrmT06SZiHQWkbtF5FnveID3imHEGNO1dDFto0muk2bjgXuBi73jWbjtfhkXUkeNWy461k9/dNOtABz5E+9vcOonDbp280Vuc0W/xsmAht/t1w+AlpOCDVSXB+MpAV2DoHLuPD+9tCqxIH9V2vozf+fq9B1ZRKMyM56YapvwBnd8u4l+3sPfdwNA5n9d5zm5Uiaub1wurq857/3uftmWfNmga+d0/xzrdVTVSXg9eVWtBCx0afwxXUsX0zaC5NrDXSMiHfD8KIjIEOC7ollVJFpMftdPf3qD65E2+5tbVvjDPg27dvfJ8wBYdF76raARpCR0DZrznv4tAIcec0fIlmQk9tpWk/QtfffCPQEoW1m/Xmhim/Cglm5MvUpdT3ezRcH6r861wT0Xtx5vKxF5E9gcOKpoVhlBYbqWLqZtBMmpwVXV6SKyD7ANzhnGTFXdUFTLjKJjupYupm00yanBFZFmwOnAnrhXlNdFZKyqxsoJbF18/b171WhHw4JKVi12QxPXLR3q57U7fT4A1c+5e1StSj/JEgalrGsxqW6RfSi0zaxwwwXGWVtt7tbbdU9pnQZ3mAfAVMrrOCM70tEt0dyuaWLSzV28zdxgvYblOqRwH7AauNU7Hg7cj+06iTuma+li2kaQXBvcbVR1x5TjKSLyUTEMCorfvvN7AIYPcIEI3/WWc9Xl3al8a7effsUuyTDrS5y7VX491C2wb1m+GoDRHSqSJ7rABPS98jT3eXZy0i4ilJyuxWLFibv56Q8PGuOl0u986PxPtzkiOD91tYivthtcr/O76sItqvj6oC0A+GkT1+QlvMq1mJ+c5C6mD4UEub73fODNcgIgIoOB2ttAjLhhupYupm0Eyea85hPcj3Rj4AQRWeAd9wKC3bNYYLr+2/VO/jLWbXjod8PpADT+LvkbtP2+swC4tZfzANamrIlf9vv5zl/qKzc6D//Nl7lf438MS64vm3PYWAA6vxOt0OmlrOuoEc5VwOQXhgJQ/XHD4lOVbb8tACsO+MHPaynpe7Y73HEmAD3WZ480UQxKQVtt5r7fbuUNc2S+5qjBfnryH6/3Uu6aOz9xDgB9Pwz2rTPbkMIhgVhhBI3pWrqYthEmm/Oa+anHXniNZkW1KCBavOOckty9ym3te+Cw22vVOXn6iQD8/Jk/AdDlvaSDm0YvuzG6mhEFtlma9FLPYYWzt5DEUdcrl23rpy/skL6jlvBHe+VZLhJUvz/kd5/y/n0BqDjHxb6avL/7u0iM/dVF/9dP8tN9rvF6TBrO6G0ctc2Frk2c05qyzXoAUL12bdq6VT9zDozuv/FGP69nI9ezPXXhXgD0v/EbAIKObJar85rDRGQ28AXwKi5c8rNZzukhIlNEpEJEZojISC+/vYi8KCKzvc92Dfw/GPXEdC1d8tXWdA2GXCfNrgCGALNUtQ/OP2a2AfhK4DxV7e+de4aIDMDCLkcJ07V0yVdb0zUAcl0WtkFVvxWRMhEpU9UpInJdphO8SJ+JaJ+rRaQCF7Y5tLDLqVQtdX5wH+nvgpA+Qu1gpD34NO/rln/9bcMMC5bY6PrW/j399JUvuH7CJR3T6zPjl24o4NZP3BDPfQ/un7buQUclJ7hOan8fANs2TkyMpX9Exqxwww99R6/w8yoLuJSpgeSlbZSeV53thvtO+XJvP29cj9ecAUe7Iep2E2pPSjbq5pZ+zTnMTW4nhhEATv9qDwAW/p8LfVT9RTjzh7k2uCtFpCXwGvCAiCwhj+EPcbHuBwLvYmGXo4TpWrrUW1vTtXjk2uAOA9YB5wDHAW2Av+Zyoif6I8AoVV0lktsSqWKEXTZqERtdE1unAZ66xS29u+Sv6Xu4TcX9aZ/ffqb7PGNmjnfKEMbBI9Gz/e/BAwConF98P6r1oF7aRuF5TURfeffxnZOZZ7ke7vkX/xuAMZUuGtDifZO/IZP3dd7bEhOcL/+Q1PL9cS4iS4cPw1mulyBX5zVrUg4n5HpxEWmME++BlOB1i0Wkq/draWGXQ8R0LV3qo63pWnyybXxYTd27EwVQVW1dR1niXMF5l69Q1TEpRaGFXQ6CquXJ8bwrl20PwKrebswx7ZcVMHHXtcM9bineoOZuk8G0C28r1q0Y+50b8/vn7cnYjF3u/RCA6rXR69nWV9so6FqTnk8s9dMLTnPLwH7lBVT51fV1+SJ2zVml52f9out+75d0+Ge4PdsE2dbhtmrAtfcAjgc+EZEPvbyLsLDLoWO6li4N0NZ0DYBcx3DzRlXfANINAFnY5ZhiupYmpmswFK3B3VRJDbf+ySq3TEV3ipYf3Njj7eLqdLt7TTzskYP8oooLegPwf/u+DCQnzTLxk3eO99M/LHQdxIQ/287jnDe5ThuSYdeD8CplQFXFbD995j6/AWDVwK4ALPuNG2I4aMsZfp2py9zwj9y2OQAdnorGMEIq4XpJNgzD2ISwHm6BKWuW3La+S1u3rX3mU/3CMqe08Xq6lYu+8bP6jnTpV2jhfe6U9TLdmZG2zNYjRoPKeQsA2Mz77PmYy09dGNicL7zUF0QV6+EahmEEhPVwC0z1umTIqFd2cL2sLXgrXXXDMDYhrIdrGIYRENbgGoZhBIQ1uIZhGAFhDa5hGEZAWINrGIYRENbgGoZhBIRoSMHu8kFElgJrgGVh21IPOtJwu3up6uaFMCZKmK6mawQpqq6xaHABRGSaqg4K2458iavdQRHX7yeudgdFXL+fYtttQwqGYRgBYQ2uYRhGQMSpwR0XtgH1JK52B0Vcv5+42h0Ucf1+imp3bMZwDcMw4k6ceriGYRixxhpcwzCMgIh8gysiB4rITBGZIyIXhG1POkSkh4hMEZEKEZkhIiO9/PYi8qKIzPY+24Vta1SIg7ama/6YrhnuG+UxXBEpB2YB+wMLganAcFX9LFTD6kBEugJdVXW6iLQC3gcOB04Clqvqtd4fXztVHR2iqZEgLtqarvlhumYm6j3cXYE5qjpXVdcDE4FhIdtUJ6q6SFWne+nVQAXQDWfvBK/aBJyoRky0NV3zxnTNQCgNbh6vHN2AL1OOF3p5kUZEegMDgXeBzqq6CJzIQKfwLCsueb5Kxk7bTVVXKDa5A3EAABMISURBVO1nNkhdA29wvVeO24FfAgOA4SIyIF31OvKiOwYCiEhL4BFglKpuMvHR89QVYqbtpqorlPYzG7SugY/hishuwGWqeoB3fCGAql6Trm5jmvyimReFdVNkNSuWRd3JST66Juo3pslbpmu0dYX8n1nTNb2uYQSRrOuVY3DNSiJyCnAKsEM5jRgs+wVkXvR4SSfPD9uGHMhXV0zXWOgKOWhruibJpGsYY7g5vXKo6jjPa88RjWlafKuMhpKXrqo6yHSNDVm1NV1zI4wGdyHQI+W4O/B1usqq+kzRLTIKQV66GrHCtC0QYTS4U4G+ItJHRJoAxwJPhmCHUVhM19LFtC0QgY/hqmqliJwJPA+UA/eo6oyg7TAKi+laupi2hSOMSbPEMIENFZQYpmvpYtoWhqjvNDMMwygZrME1DMMICGtwDcMwAsIaXMMwjIAIZdLMMAyj4Oy6AwBzznLNWqMmVX5R747LAXhm241Xs12weGc//cTzQwDo9Z8fACh748OCm2g9XMMwjICwBtcwDCMgbEjBKCka9egOQMX57nP3Xf4HwIRer9Sq+4OuB+CQz4718+bPce5PW8/e+NHo9u85frp6+UoAdMP6Qplt5El5h/YA/G9Mbz/vxaG3ANCzUfNa9cs8dxDVNdx7XN15WjJ9gksvO84NKfx86v/5Zd1/VZh9HtbDNQzDCIiS7uF+O2I3AFbu536xOj3ZzC9rvmTDRnWbzV4MQGXXZMy4ypZNst7jq32cZ6Q9DvgYgFfnbu2X9btiDQBVFbPztt1IT/nWfQBY8KuuAGx7yCy/7KEtN54UWVW9DoBH1nStdZ1m4v4GXt7u0WTmdmluen4yeeWy7QGY+NhQAHpdlewlWa+3uJQP6AfAKU+4TW8Hb/ZiSqnr2R468zAA1m5IPr9l4nq21VqX47ONGdnnZQBe3XWcn7fLvWcDsO1prqdbvW5dfcy3Hq5hGEZQZOzhishy4FHgQeAVjXKI3zr4zv0YUrHP3QBU71Ptl5V5vzXVuLyn13QAYJdmSa9zXcubb1SnLOX3qWZe4viDTq/6df5KcmwwSojItsDfgWrgbODPuGB5s4ATVbUiRPOysvujzrynOjxSq+wXFS7m38J3XRitPo+tBkCnfVqr7voDdwFgr3/c7OftdZvrynZ954eN6i7aLTkuuMvhnwDw6R9uA2DrTqf6Zf1Ofy+f/0rBEZGTVfUeL90dFwhxZ+Az4CRVnZXp/KhSttlmAGw5wfn2PnQzFw2nOqXOrtOOA6DTEe6Nsnl1FfXh7nZuqdilp/T383Y86HMA1rdp7TKK1MNdCnwI/BVYKCI3i8iQet3JiBLjgDuAfwGvAM8B7YArgNtCtMtoOGempMcAk4D2wA3AnaFYZPhka3DXqOptqroHsBvwFXCHiMwVkauLb55RJFqp6lOq+iCwQVUnquMpXMNrlAb9VPUuVa1W1cdwDa8RItkmzfwRZlVdAFwPXC8i20Dm92URuQc4BFiiqtt7ee2Bh4DewDzgGFVdUV/js+JZP/CWswDoNmW1X/TVz1o17NJD3NKgD3f9l5fjfruGv5JcStKvYlrN06JCeUp6TI2yjDOFUdD1kXH7AvBAB/fZ+4mVflmjDz9zeSwAMoeLXdfOfQ0/uz45I9bt1rfqrNstOVLE4rvcb9JT77nXy9P3ftkve7lpR3ffH3/M+v8oEt1F5BbcX//mItJYVRMzxI0znRgFbdMx+59ufPDJLe5OWAvATu8d79fpebrbTVZZz6GEBFUr3H+v23XJv4UfrmvQJX2y9XCn1JWpqjNV9fIs544HDqyRdwHwsqr2BV72jo3gud0LD42q3pHIFJGtgZeynDse0zXK/BF4H5gGXAS0BBCRLmSP0jAe07aoZOzhquq59b2wqr4mIr1rZA8DhnrpCcB/gdH1vUc2jt7/TQCenLiny3jvE7+sWwPnNpq+2gVILqS+feVWAAy4bJFfp7JhtygaqnpXmvw5wKgs54aua6fbNu6FVqepl43WD77jPjPUqfrZTgB8fmzypeD5A28CYCtvgf0+Fxznl7X48d16WlMYVHVCmvxvcA1wpnND1zYdD+3mlmiVeU3W9m+dCECf07/x61QuXRq0WXmT87IwETkk03GOdFbVRQDeZ6cM9ztFRKaJyLQNhPZ6VvKYrqVLkNqarrmRz8aHXYCnMxwXFFUdh5tNp7W0b9BytDV9C7MYvbxtGz+9f0e3NCmxZXDCnQcB0Glh3WOAESa2uhaKxJIjgDl/2RGAy46YBMCxLacDsKhqrV/n8kUHAPDVIe68FkvD7dVmIDBti6Hrd8clF0Rt09i9jSTeKBM926oMvdryzsnfBmncOGEoAJVfhRN0OOcerqpemuk4RxaLSFcA73NJPa5hFBDTtXQxbaNHto0PR2YqV9VHM5XXwZPAicC13ucTeZ6fHc8nJsCpHdyywydn71mQS/+4U3Lb7ilt3dzS3p/8GoCu97mF9Q2bHw2GWOpaB2UtWgAwf9SOfp6mmYdv/k2y0/VDF/dWsq6bm7y/dK+kucNaOF2PmukW4dxxl9sS3PbthX6dyoVfeak1DbC+OJSCtok3jiHnJlf5NJWNha2rZ9toy94AVIx08ysPH3aLX/bTJq6pW1HtNrTs8tLGW3Wh/tt18yHbkMKhGcoUtwutTkTkQdxge0cRWQhcihNtkoiMABYAR+dlrVEoTNfSxbSNMNlWKfyuvhdW1eFpivar7zWNwmC6li6mbbTJadJMRDoDVwNbqOovRWQAsJuq3p3l1FCZum6Lgl7vyNte8NMJHwrfTXGvLy1XzS3ovYIgrromWP6rnwDwlxMe9POObvltzuf/Z21LAC685yQ/b+KLbo9+I8/3QktvA0VUl/ilI87alrV3G0tu6FJ72fCBFUcAsOSPzt/xH0dM8suGNHfLQPs0SngFLKcmbcpc2axfuGVm/a88wy/b6vx3Gmh5dnKdNBsPPA8kWrBZZFmvacSC8Ziupcp4TNvIkeuysI6qOklELgRQ1UoRieb8UMrmhnv23QuAXjgPQw3tpZzSZp6frq73cvtIER9d66DtfW8DcN/jyYnS+3rU9ntbk/+d4npQf9r/KQD2POIDv2zuG9sCJeG3NLbaVi1dBsCpX+7j543t4fZWP9f/MQDK+tcVwcH1Xi/4xnmBe2r29rWu/dgQt+enX2O3g/2uI/7hl4257WAAKuctaPD/IR25/l2tEZEOeFvTPY9h3xXNKiMoTNfSxbSNILn2cM/FLQ/ZSkTeBDYHjiqaVQUiuXynYfwwbFcAypiekut+q1Id4sSQWOpak6pVq5IHM1alr+jRd6T7fKJZDwDmjR7ol901wbmWuGS2Gyvc7DC3QD5EZzT1JbbaJr7rOVcll/utvdNFdmgpLsLKvEq3EeUXr5/l19nmqu+BZISVPnxc69qvf+aWdm7b5ksAhjZLRn65YkBnAJoWsYebU4OrqtNFZB9gG5ybnpkpHoiMmGK6li6mbTTJdZVCM+B0YE/cK8rrIjJWVYu/UtgoGqZr6WLaRpNchxTuA1YDt3rHw4H72UQWQS/f1n1NqQP0e398DACtUybpYkhsdK3eK/na32SO88hWueibdNVzu6a3s6jn5Un/F1c/4/yrXvLgRADefM/5YX3vhJ8kz/so0hGIEsRG23Q0eyrp0u/YL0cAoI3cUF7ZGjfssHVFcsIzlxnBKj8klnuWP1ifnPzebN53OV+nvuTa4G6jqjumHE8RkY+KYZARKKZr6WLaRpBcG9wPRGSIqr4DICKDgTeLZ1a0KNvNeYAvSwbAYN1/3AB7az4PxaYCEXldG/VyE1uX3pcMWf3n4a63QwN7uHWhU90byzVnngTAsBvdZM2Zjzzm17n1179ydd+fQYSJvLb5UO1F8kiQdy/U87FyYItEWDfny/jKBcmd0FWfFT++ZjbnNZ/gxn8aAyeIyALvuBcuCqgRQ0zX0sW0jTbZerj1cVhcchzSy/VkUsdwO8yI9dxDbHRdeKTr4Z4w9WQ/r9c7tZf7FJomz00F4IX5bhF9s0eS47xtb3FjyN8d6OLiVa+O1NLA2GgbJD8Z67Zq92zUfKP8uc9s6ae7Ufg3pppkc14zP/VYRDqR2M5hxBbTtXQxbaNNrsvCDgNuxO3LXoJ7PakAtstwTg/cTGkXXNipcap6c1SigOZC5b47A3B5p0Q8JclUPXbESdcfVzUt1KXyIrGI/h83H+bnvfeX2wHY+8DTAGj5cPQiPuSrbSk8rzWZ/9fd/PQznZ1miTUJ273m3pi2vOl9v04Q4Udy3dp7BTAEmKWqfXDu2rINwFcC56lqf+/cMzyPRRYFNDqYrqVLvtqargGQa4O7QVW/BcpEpExVpwA/zXSCqi5S1eleejXu17UbLgpoIrLoBODwelluFALTtXTJS1vTNRhyXRa2UkRaAq8BD4jIEvJwvuWFXh4IvEuNKKDeGFOkSXoGKwEfUhsTeV1bLHLf/YEnJMOtfLL55kDmAIKFptOE5AL7sSN7AbD4cBectOXDgZmRD/XWNu7Pa8L3yYwRt/t55eKe3XkbnL+Frf/qfDFUBewjI9cWZBjwA3AO8BzwOZlDefh4oj8CjFLV7J5FkudZ2OXiY7qWLvXS1nQtLrk6r0mNljchbcUaiEhjnHgPpASvWywiXb1fy7RRQKMUTjsR3aHUJs3ioGvbZ9zS0YOv/tDPe+lIN1nV8R/LXUZ18d28pgYYfPe7PgCctIPzx/t6BBcB1EfbOD6v5a1b++k5F7j5wAeH3wxAdUrEh++94JHD7vwTAN0qksv8giRjD1dEVovIqjr+rRaRjL9+IiLA3UCFqo5JKUpEAYUAI7waSUzX0qW+2pquwZBtHW6rBlx7D+B44BMRSXRPLiKGUUATY7hjVybDpDd5fw4Qj7DoNYmTrglft2ffdLqfN+WSvwEwqP+5APS7KLkRonrt2kLcthbzrkouMRrT1bVHR40/D4CehNNbqosGaBu553X9AYP89PfdXJj09ve4t4qvRu8OwMnHP+fXeaLdFC9VO5bZT590TpD7XRuuVrlOmuWNqr4Bad/BLQpoTDFdSxPTNRhKbtrdMAwjqhSth1sKzP+9GzBITJpNGHOQX9Zh1duh2LSp0vmW5KvgUDkfgGnnu6GFcfslvRA+fo3rjLV7diYAVSvquSlqiPN/O/s094hU/PwWv2hb7/V02xudZ7GSCCcaQdZ0beynH7nsBgC++bPbcbhzExfuqrqO/WH3rnL+N258dJif1+/iaDyv1sM1DMMICOvhZuBfQ+4GkpNmHe6Oxq/kpk6Xm11v95gPzwDg67PX+2VnXeJCn297hfPoddr04wBo83gLv07jta5XtKqnm1xZOzg50XbtILca6oDN3gFgwqq+AOx020i/Tr9r3P2tZ1tc2nyRXIqX6OsObLJxH/G6b5OuIca/OBSAfmPdyrXes6P3vFoP1zAMIyCsh1uDRj26++khzdzqmA2hbrsw0lH2qttu2/3VZN6Tbd3SvRsuc5uqBu/ixnJHXfmCX+frynYAHN7CbfP8w5d7+GV/fHG4O/9tN2HfbrL7G+i+LjpLvzYVEvoCnNRzz6z1t8K9lUR5qab1cA3DMALCerg10WR3doO638rbV24VljVGnlStdJFXtx7lejvfevl/Zpdade/0U8ldsH3Z2LetjdMahcR6uIZhGAFhDa5hGEZA2JBCDSoXfuWnD+m2c4iWGIZRalgP1zAMIyBENfprnkRkKW5mY1nYttSDjjTc7l6qunkhjIkSpqvpGkGKqmssGlwAEZmmqoOy14wWcbU7KOL6/cTV7qCI6/dTbLttSMEwDCMgrME1DMMIiDg1uOPCNqCexNXuoIjr9xNXu4Mirt9PUe2OzRiuYRhG3IlTD9cwDCPWRL7BFZEDRWSmiMwRkQvCticdItJDRKaISIWIzBCRkV5+exF5UURme5/twrY1KsRBW9M1f0zXDPeN8pCCiJQDs4D9gYXAVGC4qn4WqmF1ICJdga6qOl1EWgHvA4cDJwHLVfVa74+vnaqODtHUSBAXbU3X/DBdMxP1Hu6uwBxVnauq64GJwLAs54SCqi5S1eleejVQAXTD2TvBqzYBJ6oRE21N17wxXTMQ9Qa3G/BlyvFCLy/SiEhvYCDwLtBZVReBExnoFJ5lkSJ22pquOWG6ZiDqDa7UkRfdMRBARFoCjwCjVHVV2PZEmFhpa7rmjOmagag3uAuBHinH3YGvQ7IlKyLSGCfeA6r6qJe92BsvSowbLQnLvogRG21N17wwXTMQ9QZ3KtBXRPqISBPgWODJkG2qExER4G6gQlXHpBQ9CZzopU8EngjatogSC21N17wxXTPdN8qrFABE5CDgJqAcuEdVrwrZpDoRkT2B14FPSEZmuQg3LjQJ6AksAI5W1eWhGBkx4qCt6Zo/pmuG+0a9wTUMwygVoj6kYBiGUTJYg2sYhhEQ1uAahmEEhDW4hmEYAWENrmEYRkBYg2sYhhEQ1uAahmEEhDW4hmEYAfH/j7dj1RN1a14AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 9 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def plot_img(data, label):\n",
    "    fig, axs = plt.subplots(3, 3) # 9 images\n",
    "    k = 0\n",
    "    for i in range(3):\n",
    "        for j in range(3):        \n",
    "            axs[i, j].imshow(data[k].astype('uint8').reshape(28, 28))   # plot image            \n",
    "            axs[i, j].set_ylabel(\"label:\" + str(label[k].item()))       # print label\n",
    "            k +=1\n",
    "plot_img(train_all_numpy, train_all_label_numpy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(33600, 784)\n",
      "(33600,)\n",
      "(8400, 784)\n",
      "(8400,)\n"
     ]
    }
   ],
   "source": [
    "# split train on train and validation\n",
    "train, validation, train_label, validation_label = train_test_split(train_all_numpy, train_all_label_numpy, test_size=0.2)\n",
    "\n",
    "print(train.shape)\n",
    "print(train_label.shape)\n",
    "print(validation.shape)\n",
    "print(validation_label.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<BarContainer object of 10 artists>"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "unique, counts_train = np.unique(train_label, return_counts=True)\n",
    "plt.subplot(1, 2, 1)\n",
    "plt.bar(unique, counts_train/len(train_label))\n",
    "unique, counts_val = np.unique(validation_label, return_counts=True)\n",
    "plt.subplot(1, 2, 2)\n",
    "plt.bar(unique, counts_val/len(validation_label))\n",
    "#print (np.asarray((unique, counts_train/len(train_label), counts_val/len(validation_label))).T)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_all_tensor = torch.as_tensor(train_all_numpy).type(torch.FloatTensor)\n",
    "train_all_label_tensor = torch.as_tensor(train_all_label_numpy)\n",
    "test_tensor = torch.as_tensor(test_numpy).type(torch.FloatTensor)\n",
    "\n",
    "train_tensor = torch.as_tensor(train).type(torch.FloatTensor)\n",
    "train_label = torch.as_tensor(train_label)\n",
    "\n",
    "validation_tensor = torch.as_tensor(validation).type(torch.FloatTensor)\n",
    "validation_label = torch.as_tensor(validation_label)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "                   # nn.Module - Base class for all neural network modules.\n",
    "class FNet(nn.Module):             \n",
    "    def __init__(self):\n",
    "        super(FNet, self).__init__()\n",
    "                   # 784 inputs, connects to hidden layer with 600 nodes\n",
    "        self.fc1 = nn.Linear(in_features=784, out_features=600)\n",
    "                   # 600 nodes connects to hidden layer with 500 nodes\n",
    "        self.fc2 = nn.Linear(in_features=600, out_features=500) \n",
    "                   # 500 nodes connects to hidden layer with 250 nodes\n",
    "        self.fc3 = nn.Linear(in_features=500, out_features=250) \n",
    "                   # connection between the last hidden layer \n",
    "                   # and the output layer (with 10 nodes)\n",
    "        self.fc4 = nn.Linear(in_features=250, out_features=10)  \n",
    "                                                                \n",
    "    def forward(self, x):\n",
    "        x = x.view(-1,784)          # Put all the entries of the image in the vector\n",
    "        x = F.relu(self.fc1(x))     # Input x into first layer and apply a ReLU\n",
    "                                    # to the nodes in this layer\n",
    "        x = F.relu(self.fc2(x))        \n",
    "        x = F.relu(self.fc3(x))\n",
    "        x = self.fc4(x)\n",
    "        return x     "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_accuracy(predictions, true_labels):\n",
    "    _, predicted = torch.max(predictions, 1)\n",
    "    corrects = (predicted == true_labels).sum()\n",
    "    accuracy = 100.0 * corrects/len(true_labels)\n",
    "    return accuracy.item()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "def training (dataloader, epochs, model, criterion, optimizer):\n",
    "\n",
    "    train_accuracies, train_losses = [], []\n",
    "    \n",
    "    # set the train mode\n",
    "    model.train()\n",
    "    \n",
    "    # loop over training dataset multiple times\n",
    "    for epoch in range(epochs):        \n",
    "\n",
    "        train_loss = 0 \n",
    "        train_accuracy = 0\n",
    "        num_batch = 0\n",
    "        \n",
    "        # iterate over all batches\n",
    "        for data, labels in dataloader:\n",
    "            \n",
    "            # zero the parameters gradient to not accumulate gradients from previous iteration\n",
    "            optimizer.zero_grad()\n",
    "            \n",
    "            # put data into the model\n",
    "            predictions = net(data)\n",
    "            \n",
    "            # calculate loss\n",
    "            loss = criterion(predictions, labels)\n",
    "            \n",
    "            # calculate accuracy\n",
    "            accurasy = get_accuracy(predictions, labels)\n",
    "            \n",
    "            # compute gradients\n",
    "            loss.backward()\n",
    "            \n",
    "            # change the weights\n",
    "            optimizer.step()\n",
    "            \n",
    "            num_batch += 1\n",
    "            train_loss += loss.item()\n",
    "            train_accuracy += accurasy\n",
    "    \n",
    "        epoch_accuracy = train_accuracy/num_batch\n",
    "        epoch_loss = train_loss/num_batch        \n",
    "        train_accuracies.append(epoch_accuracy)\n",
    "        train_losses.append(epoch_loss)\n",
    "        \n",
    "        print(\"Epoch: {}/{} \".format(epoch + 1, epochs),\n",
    "              \"Training Loss: {:.4f} \".format(epoch_loss),\n",
    "              \"Training accuracy: {:.4f}\".format(epoch_accuracy))\n",
    "    \n",
    "    return train_accuracies, train_losses"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 1/40  Training Loss: 0.3055  Training accuracy: 91.4597\n",
      "Epoch: 2/40  Training Loss: 0.0873  Training accuracy: 97.3206\n",
      "Epoch: 3/40  Training Loss: 0.0533  Training accuracy: 98.3068\n",
      "Epoch: 4/40  Training Loss: 0.0343  Training accuracy: 98.8682\n",
      "Epoch: 5/40  Training Loss: 0.0255  Training accuracy: 99.1178\n",
      "Epoch: 6/40  Training Loss: 0.0252  Training accuracy: 99.2158\n",
      "Epoch: 7/40  Training Loss: 0.0277  Training accuracy: 99.1059\n",
      "Epoch: 8/40  Training Loss: 0.0258  Training accuracy: 99.1712\n",
      "Epoch: 9/40  Training Loss: 0.0279  Training accuracy: 99.0702\n",
      "Epoch: 10/40  Training Loss: 0.0171  Training accuracy: 99.4445\n",
      "Epoch: 11/40  Training Loss: 0.0241  Training accuracy: 99.1920\n",
      "Epoch: 12/40  Training Loss: 0.0184  Training accuracy: 99.4237\n",
      "Epoch: 13/40  Training Loss: 0.0145  Training accuracy: 99.5485\n",
      "Epoch: 14/40  Training Loss: 0.0102  Training accuracy: 99.7208\n",
      "Epoch: 15/40  Training Loss: 0.0207  Training accuracy: 99.3910\n",
      "Epoch: 16/40  Training Loss: 0.0196  Training accuracy: 99.3940\n",
      "Epoch: 17/40  Training Loss: 0.0106  Training accuracy: 99.6792\n",
      "Epoch: 18/40  Training Loss: 0.0133  Training accuracy: 99.6346\n",
      "Epoch: 19/40  Training Loss: 0.0156  Training accuracy: 99.5722\n",
      "Epoch: 20/40  Training Loss: 0.0105  Training accuracy: 99.7119\n",
      "Epoch: 21/40  Training Loss: 0.0143  Training accuracy: 99.5663\n",
      "Epoch: 22/40  Training Loss: 0.0160  Training accuracy: 99.5366\n",
      "Epoch: 23/40  Training Loss: 0.0192  Training accuracy: 99.4415\n",
      "Epoch: 24/40  Training Loss: 0.0098  Training accuracy: 99.7059\n",
      "Epoch: 25/40  Training Loss: 0.0141  Training accuracy: 99.6227\n",
      "Epoch: 26/40  Training Loss: 0.0095  Training accuracy: 99.7237\n",
      "Epoch: 27/40  Training Loss: 0.0119  Training accuracy: 99.6792\n",
      "Epoch: 28/40  Training Loss: 0.0054  Training accuracy: 99.8515\n",
      "Epoch: 29/40  Training Loss: 0.0071  Training accuracy: 99.8277\n",
      "Epoch: 30/40  Training Loss: 0.0141  Training accuracy: 99.6851\n",
      "Epoch: 31/40  Training Loss: 0.0187  Training accuracy: 99.4742\n",
      "Epoch: 32/40  Training Loss: 0.0114  Training accuracy: 99.7089\n",
      "Epoch: 33/40  Training Loss: 0.0020  Training accuracy: 99.9317\n",
      "Epoch: 34/40  Training Loss: 0.0091  Training accuracy: 99.7475\n",
      "Epoch: 35/40  Training Loss: 0.0052  Training accuracy: 99.8574\n",
      "Epoch: 36/40  Training Loss: 0.0059  Training accuracy: 99.8634\n",
      "Epoch: 37/40  Training Loss: 0.0116  Training accuracy: 99.7000\n",
      "Epoch: 38/40  Training Loss: 0.0142  Training accuracy: 99.6762\n",
      "Epoch: 39/40  Training Loss: 0.0096  Training accuracy: 99.7237\n",
      "Epoch: 40/40  Training Loss: 0.0119  Training accuracy: 99.6822\n"
     ]
    }
   ],
   "source": [
    "torch.manual_seed(0)                                      # set seed to make results reproducible\n",
    "\n",
    "batch_size = 128                                          # Set the batch_size\n",
    "train_dataset = torch.utils.data.TensorDataset(train_tensor, train_label)\n",
    "trainloader = torch.utils.data.DataLoader(train_dataset, batch_size = batch_size, shuffle = True)\n",
    "\n",
    "\n",
    "epochs = 40                                               # set number of epohs\n",
    "\n",
    "net = FNet()                                              # initialize network\n",
    "criterion = nn.CrossEntropyLoss()                         # set criterion\n",
    "optimizer = torch.optim.Adam(net.parameters(), lr = 3e-4) # set optimizer\n",
    "\n",
    "                                                          # start training process\n",
    "train_accuracies, train_losses = training(trainloader, epochs, net, criterion, optimizer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "def train_curves(epochs, train_losses, train_accuracies):\n",
    "    iters = range(1, epochs+1)\n",
    "    fig, (ax1, ax2) = plt.subplots(ncols=2, figsize=(15, 5))\n",
    "    fig.suptitle('Training Curve')\n",
    "    ax1.plot(iters, train_losses)\n",
    "    ax1.set_xlabel(\"Iterations\")\n",
    "    ax1.set_ylabel(\"Loss\")\n",
    "    ax2.plot(iters, train_accuracies, color = 'g')\n",
    "    ax2.set_xlabel(\"Iterations\")\n",
    "    ax2.set_ylabel(\"Training Accuracy\")\n",
    "    plt.show()\n",
    "    \n",
    "train_curves(epochs, train_losses, train_accuracies)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss:  0.17723943293094635 Accuracy:  97.17857360839844\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x221961aec48>"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# set net in test (evaluation) mode\n",
    "net.eval()    \n",
    "# get predictions\n",
    "val_predictions = net(validation_tensor)\n",
    "val_loss = criterion(val_predictions, validation_label)\n",
    "val_accurasy = get_accuracy(val_predictions, validation_label)\n",
    " \n",
    "print(\"Loss: \", str(val_loss.item()), \"Accuracy: \", str(val_accurasy))\n",
    "\n",
    "# to get class with the maximum score as prediction\n",
    "_, val_predicted = torch.max(val_predictions.data,1)            \n",
    "\n",
    "# confusion matrix\n",
    "skplt.metrics.plot_confusion_matrix(validation_label, val_predicted, figsize=(5,5))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 9 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_img(validation, val_predicted)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 9 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# wrong predicted\n",
    "\n",
    "val_wrong = validation[validation_label != val_predicted]\n",
    "val_lab_wrong = val_predicted[validation_label != val_predicted]\n",
    "\n",
    "plot_img(val_wrong, val_lab_wrong)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 1/40  Training Loss: 0.3055  Training accuracy: 91.4597\n",
      "Epoch: 2/40  Training Loss: 0.0873  Training accuracy: 97.3206\n",
      "Epoch: 3/40  Training Loss: 0.0533  Training accuracy: 98.3068\n",
      "Epoch: 4/40  Training Loss: 0.0343  Training accuracy: 98.8682\n",
      "Epoch: 5/40  Training Loss: 0.0255  Training accuracy: 99.1178\n",
      "Epoch: 6/40  Training Loss: 0.0252  Training accuracy: 99.2158\n",
      "Epoch: 7/40  Training Loss: 0.0277  Training accuracy: 99.1059\n",
      "Epoch: 8/40  Training Loss: 0.0258  Training accuracy: 99.1712\n",
      "Epoch: 9/40  Training Loss: 0.0279  Training accuracy: 99.0702\n",
      "Epoch: 10/40  Training Loss: 0.0171  Training accuracy: 99.4445\n",
      "Epoch: 11/40  Training Loss: 0.0241  Training accuracy: 99.1920\n",
      "Epoch: 12/40  Training Loss: 0.0184  Training accuracy: 99.4237\n",
      "Epoch: 13/40  Training Loss: 0.0145  Training accuracy: 99.5485\n",
      "Epoch: 14/40  Training Loss: 0.0102  Training accuracy: 99.7208\n",
      "Epoch: 15/40  Training Loss: 0.0207  Training accuracy: 99.3910\n",
      "Epoch: 16/40  Training Loss: 0.0196  Training accuracy: 99.3940\n",
      "Epoch: 17/40  Training Loss: 0.0106  Training accuracy: 99.6792\n",
      "Epoch: 18/40  Training Loss: 0.0133  Training accuracy: 99.6346\n",
      "Epoch: 19/40  Training Loss: 0.0156  Training accuracy: 99.5722\n",
      "Epoch: 20/40  Training Loss: 0.0105  Training accuracy: 99.7119\n",
      "Epoch: 21/40  Training Loss: 0.0143  Training accuracy: 99.5663\n",
      "Epoch: 22/40  Training Loss: 0.0160  Training accuracy: 99.5366\n",
      "Epoch: 23/40  Training Loss: 0.0192  Training accuracy: 99.4415\n",
      "Epoch: 24/40  Training Loss: 0.0098  Training accuracy: 99.7059\n",
      "Epoch: 25/40  Training Loss: 0.0141  Training accuracy: 99.6227\n",
      "Epoch: 26/40  Training Loss: 0.0095  Training accuracy: 99.7237\n",
      "Epoch: 27/40  Training Loss: 0.0119  Training accuracy: 99.6792\n",
      "Epoch: 28/40  Training Loss: 0.0054  Training accuracy: 99.8515\n",
      "Epoch: 29/40  Training Loss: 0.0071  Training accuracy: 99.8277\n",
      "Epoch: 30/40  Training Loss: 0.0141  Training accuracy: 99.6851\n",
      "Epoch: 31/40  Training Loss: 0.0187  Training accuracy: 99.4742\n",
      "Epoch: 32/40  Training Loss: 0.0114  Training accuracy: 99.7089\n",
      "Epoch: 33/40  Training Loss: 0.0020  Training accuracy: 99.9317\n",
      "Epoch: 34/40  Training Loss: 0.0091  Training accuracy: 99.7475\n",
      "Epoch: 35/40  Training Loss: 0.0052  Training accuracy: 99.8574\n",
      "Epoch: 36/40  Training Loss: 0.0059  Training accuracy: 99.8634\n",
      "Epoch: 37/40  Training Loss: 0.0116  Training accuracy: 99.7000\n",
      "Epoch: 38/40  Training Loss: 0.0142  Training accuracy: 99.6762\n",
      "Epoch: 39/40  Training Loss: 0.0096  Training accuracy: 99.7237\n",
      "Epoch: 40/40  Training Loss: 0.0119  Training accuracy: 99.6822\n"
     ]
    }
   ],
   "source": [
    "torch.manual_seed(0)                                      # set seed to make results reproducible\n",
    "\n",
    "batch_size = 128                                          # Set the batch_size\n",
    "train_all_dataset = torch.utils.data.TensorDataset(train_all_tensor, train_all_label_tensor)\n",
    "trainloader = torch.utils.data.DataLoader(train_dataset, batch_size = batch_size, shuffle = True)\n",
    "\n",
    "epochs = 40                                               # set number of epohs\n",
    "\n",
    "net = FNet()                                              # initialize network\n",
    "criterion = nn.CrossEntropyLoss()                         # set criterion\n",
    "optimizer = torch.optim.Adam(net.parameters(), lr = 3e-4) # set optimizer\n",
    "\n",
    "                                                          # start training process\n",
    "train_accuracies, train_losses = training(trainloader, epochs, net, criterion, optimizer)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "train_curves(epochs, train_losses, train_accuracies)"
   ]
  }
 ],
 "metadata": {
  "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}