fashion-mnist-classification.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Import dependencies"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "pycharm": {
     "is_executing": false
    }
   },
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.svm import LinearSVC\n",
    "from sklearn.svm import SVC\n",
    "from sklearn.metrics import confusion_matrix\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "from sklearn.neural_network import MLPClassifier\n",
    "\n",
    "plt.style.use('ggplot')\n",
    "random_state = 42"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Read data "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABG0AAACCCAYAAAD8MzWDAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjAsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+17YcXAAAgAElEQVR4nO2dd5hV1dnF10lTE0WxIRZAQQR7QcTekRasGBNBTD5LosYWY2yJihGiiRV9NELEWD7FSuyF8KGChaIiQkCRIoKoIMZETWKS8/0xzHbtxZzDnXFm7r1n1u95fPIe9p5z9z273pN3vW+SpimMMcYYY4wxxhhjTGXxtXI3wBhjjDHGGGOMMcasjF/aGGOMMcYYY4wxxlQgfmljjDHGGGOMMcYYU4H4pY0xxhhjjDHGGGNMBeKXNsYYY4wxxhhjjDEViF/aGGOMMcYYY4wxxlQghX5pkyTJN5IkOS9JkreSJPlnkiTvJklyTbnbZeqH+7H6SZLksCRJXl/Rf/OSJDm73G0y9SdJkq2TJPlzkiSfJUmyOEmSIUmSfL3c7TKlkSTJUUmSvJAkybIkSf6RJMnsJEkuSpLkW+Vum2kYSZJskiTJ35MkSZMkWbPc7TGl432x+kmSZPyKuVfXf7uXu32mNDwXi0HR+/Eb5W5AE3MbgAMAXApgFoDNAGxdzgaZBnEb3I9VS5IkewJ4EMCtAM4BsBuAK5Ik+W+apteWtXGmZJIkaQ1gLICZAA4F0BHAVah5+X9RGZtmSmc9AOMA/BbAxwC6A7gEwEYATitfs8xX4LcA/g7gO+VuiCkd74uF4RQAreTfhgDYCcDk5m+OqS+ei8WgJfRjkqZpudvQJCRJ0gvAIwB2SNN0ZrnbYxqG+7H6SZLkKQDfTtN0b/q3qwD8EMBGaZr+q2yNMyWTJMn5AM4F0D5N009W/Nu5WPGjv/bfTHWRJMnlAE4F0Dot6oGgoCRJsg+AMQCGoublzVppmv69vK0ypeB9sZis8FpcAmB0mqY/KXd7zKrxXCwGLaEfiyyP+hGAcf6hX/W4H6ufHQE8I//2NIDWAOw+XD30BvCUvJy5B8AaAPYtT5NMI7AMgOVRVcYKWeJw1Py/+kvL3BxTf7wvFpNeqOnDu8vdEFMynovFoPD9WOSXNrsBeDNJkhuSJPlkRQyGB5Mk2bjcDTP1wv1Y/awOQN9w1153bea2mIbTBTXyxECapu8A+GxFmakSkiT5epIk306SZC8ApwO4yV42VcePAawG4MZyN8Q0CO+LxeQYAO8CeL7cDTEl47lYDArfj0V+abMRgONR8+btGNS4R+0C4KEkSZIytsvUD/dj9TMHwK7yb91X/O+6zdwW03BaoyYOirJ8RZmpHj5d8d/zAJ4F8PPyNsfUhyRJ1gNwGYCz0zT9otztMQ3C+2LBSJLk2wD6A7jXL8GrCs/FYlD4fizyS5tkxX+Hpmn6eJqmowEMQk0HHlDWlpn64H6sfm4GcFiSJCcmSdI6SZJDANRGdP9vGdtlTEtlDwB7A/gZaoJK31De5ph6cjmAl9I0fbzcDTENxvti8fguagKCWxpVXXguFoPC92ORX9osBzA9TdNl9G8TUOMq5cxD1YP7sfq5FcBNK/77CDXR3S9bUbakXI0y9WY5gLXr+PfWK8pMlZCm6Stpmk5I0/Rq1MijfpIkScdyt8usmiRJtkFNrLchSZKskyTJOgC+vaJ47SRJ1ihf60w98L5YPI4BMCdN0ynlboipF56LxaDw/VjklzZ/QY2HhpKgIG/cWgjuxyonTdP/pGl6GoANAGwPoA2Al1YUv5T5h6bSmAWJXZMkyWao+cE4q86/MNXAKyv+d/OytsKUypYAvgngRdS8LF2OL+PavIua4MSmwvG+WCySJFkbNcH67WVTZXguFoOW0I/fKHcDmpBHAVyaJMn6aZrWZlbYBzWHnWnla5apJ+7HgpCmae0PDCRJcgqAF9I09Y/96uEJAD9PkmStNE3/tuLfvgfgc9TERTHVyZ4r/ndeWVthSmUCgP3l33oB+AWAPgDmNnuLTIPxvlgYDkdNYHC/tKlSPBeLQZH7MSlqrKwkSVoBeAPAIgBDAawF4AoAs9I0PbicbTOl436sfpIk6QFgLwCvAWgF4PsADgGwV5qmr5ezbaZ0kiRpDWAmaubjFQC2AHA1gGvTNL2onG0zpZEkyZMAxgKYAeA/qHlh8zMAj6Zpekw522YaTpIkxwMYBWCtNE3/XubmmBLwvlgsVqytG6VpumO522Lqh+diMWgJ/VhYT5s0TT9JkuQAANcDuAc1MVD+BOCssjbM1Av3YyH4AjUeGZegRtL2PIA90zSdXs5GmfqRpunyJEkORE3Q2kdQk0nqGtT0q6kOJqMmG18HAP9GjVfG+agJ4GeMaT68LxaEJEnWB3AggF+Wuy2mQXguFoPC92NhPW2MMcYYY4wxxhhjqpkiByI2xhhjjDHGGGOMqVr80sYYY4wxxhhjjDGmAvFLG2OMMcYYY4wxxpgKxC9tjDHGGGOMMcYYYyoQv7QxxhhjjDHGGGOMqUDqlfI7SZKypZrafPPNg73aaqsFe/78+VG9f/zjH/W+9ze+ET+Gjh07BvvTTz8N9rvvvlvvezcWaZomjXGfcvZhFuutt150vfbaawf7iy++CPZ//vOfqN4///nPYC9btqyJWteoLE3TdIPGuFEl9mNLochzsQXhuQhg9dVXj655/1xrrbWisg02+PJx/f3vfw/2Bx980EStWzUtaS7yPpkk2V/7a1/78v+LK2ff1APPxQLQkuZigfFcLACei4WgzrlYr5c25eTyyy8P9hZbbBHsE044Iar3xhtv1Pve66yzTnQ9YsSIYE+dOjXYZ511Vr3v3ZLRg2VWevn+/ftH17169Qo2Hzo/+eSTqN6cOXOCPWrUqJLakdemrPY1Igua+gOMqYtS52Ie/HL73//+91duk8Jt9FxsHngvBYDZs2cHe7fddovKTjzxxGC/+OKLwb722mubqHWG6du3b7C/9a1vBVvn9hprrBHs66+/PvN+zTzf8vBcNKYy8Fw0pjKocy6W9aWNHgpPOeWUYB9wwAFRGXtS8IHl0Ucfjeq1b98+2J9//nmw2WMDiA827LEBAB9++GGwv/Od7wSbD7QA8PDDDwdbD0cLFy5ES6fUg+B+++0XXe++++7B/u9//xts9aKaPHlysEePHh2VffbZZ3W2o8yHU2O+Mg35saX12MOC17iuXbtG9ebNmxfsRYsWlfRZ3/zmN6PrLbfcMtjLly8P9nvvvZfbxizyvAw8v+umW7duwb7uuuuCvdVWW0X1/vWvf9VpA8D7778fbN6f+WUOAGyzzTZfrbEFhL1fdIxmjdlf/epX0fUll1wSbPZ04j0SiD1VDzzwwKjs0EMPXeXnAhX1QscYY4wxcEwbY4wxxhhjjDHGmIrEL22MMcYYY4wxxhhjKhC/tDHGGGOMMcYYY4ypQJL66JUbI5I0xyHZeOONozLO1KTxS1hfz21W3T1rx19//fVgswYciHX4b731VmZ7ORaOZtrg7Br8uQDw2muvBfvwww/PvH+pVGM08LZt20bXHAeBgw3fd999Ub0BAwYEe+DAgcF+4IEHonrXXHNNsIcMGRKVcUDqO+64I9gLFsSxnTh+UTMwNU3Tbquutmoc1b18VOpc1Fgya665ZrA5zgUQr71//etfg83zEgA6d+4c7HXXXbfOv9HP1mx8H330UbDHjBkTbF2TOWbKkiVLorKGZAVcBYWYi6XGHuH9iGMYLV26NKrH8VG43wCgXbt2wV5//fWDzbHjAKBTp06ranajUUlzsTHiwHTo0CHY48ePz7w/2xrThs8p2g5OsqAxc0qFzzr62Q2kEHOxpVNJc9E0GM/FAuC5COy8887RNZ+HNY4tn4M4SzHHzwXidxS6t379618PNse05XMTEGehXsX+WedctKeNMcYYY4wxxhhjTAXilzbGGGOMMcYYY4wxFUizpPw++eSTg92mTZtgawpZdkvSFN3sDrzJJpsEW12U2G2b5VEPPfRQVO9HP/pRnZ8LZLsX//vf/47qsbRGpQnsmnXUUUcF+/7770eR6NGjR3R9wgknBFtT+vIz+uSTT4KtaWfZ3X7YsGHB5rS1QCzX+P3vfx+VHX300cHmNKcq3WCXNpVYqXzDmKZC01jnyStatWoV7E033TTY6sq52mqrBfvb3/52VMZrF3/2XXfdFdVjWRJ/bp7Lp8pl2AWUZVq8F2gbu3TpEpWxDJbXa5W2NpJco2rIGienn356dM19t2zZsmDresjPWcu473gtz5MXtyRKlUTttttu0XXPnj2D/f3vfz/YvDcB8fljnXXWyaz32WefBVv3MD6LnXrqqcEePnx4VI/TiystbY4ZY4ypDDQUSdZ+pHvYxx9/HGwObQLE59fTTjst2FtuuWVU78c//nGwDzrooMz7c5v03cDdd98d7FGjRtXZ9jzsaWOMMcYYY4wxxhhTgfiljTHGGGOMMcYYY0wF4pc2xhhjjDHGGGOMMRVIs8S04fgiGj+GYU246rT5mvVnCqfkYn34SSedFNVjHb7GgihVm8YxGPR7cUrc4447LthFiGnDMWj0ub788svB5rTbQJxmmDX5GgODU7Oxjn/q1KlRvV133TXYL774YlQ2adKkYP/lL38JtsYd6Nq1a7BvueWWqIz1/3/7299gTFORF9OGY2wBwI477hhsjl+h82j58uXBXrx4cVTGax7fY4MNNojqsX6Y4+DoGpy3FnJ8Mr6fxrzi2F9z5syJyjS+Si3bb799dM1zXdM6FgF9Drw+8rrMumwgXr94bOl6qOOQ4X5dY401gq3rsonjxQDAgAEDgr3ZZptFZRw3ivtXtfo853hsb7TRRlE9jmOjsax4Tfjggw+C/cMf/jCq16dPn2CPHTs2KrvgggtgjDFFRX9/8t633nrrRWWXXXZZsE855ZSmbZjJpXfv3sHWczPvd7feemtUNn78+GC///77wf7d734X1fvOd74T7PPOOy8qu/jii4M9Y8aMYJ9xxhlRvV/84heZ7S8Fe9oYY4wxxhhjjDHGVCB+aWOMMcYYY4wxxhhTgTSLPOrAAw8M9sKFC4Ot7kvs1qtu2pyKlP9OXYhZ2jR58uRga5owdkNWl3NOlcv318/KSocLxCmt+/fvjyJx6aWXBnv+/PlRGadI4/TrQOxaxq7806dPj+qxC/fWW28dbE0Dv2DBgmCzZAQAunfvHuy2bdsGe+nSpVE9TmPLfQbE7m8XXnghTNPQt2/fYPM8B4BnnnmmuZtTJ7UpAll+2ZjkpdFt3759dM0yTZ4TvB4BsczwiSeeiMo++uijYLNEhucDEEuYHn744WCfffbZUb3/+Z//CfbIkSOjMp6n48aNC7bKqDjN9+zZs6OyLHmPrrv8rN58800Ujby00meddVaweQ8DsvtbJb95EmXen7fYYotgP/7446tqdotgxIgRwf7ud78blfGexn0BxJJEfuYq71t33XWDzVLImTNnRvU23HDDYOscY1nbVVddFWx162fJ1aBBg6IydkHfaaedYIwxRSJvn9Xz2M477xzsG2+8MdhjxoyJ6vF5hM9Epn7k9Q3vsw899FBUxuEuOGQLEKfvZsm9pgbn/fNPf/pTVMbyqLlz5wZbf7c++OCDme0vBXvaGGOMMcYYY4wxxlQgfmljjDHGGGOMMcYYU4E0izyK4Qwh6iaUl9WCM5xwmcqe2H2JXee1Ht9DpQlZ2U5URsX3YLkPsLJcp9rZdtttg83u3fpM+NktWrQoKmNZG7ttv/3225n3YHdxllfp/TW7E0tDWOLxzjvvRPVYfqVlLLFid/ElS5bArBqWOPJc0Ywm8+bNCzZnWQFiWRqPO81gMmvWrHq375hjjomuf/CDHwSbs6ABX/a5ukQ2Fa1btw62ZodhiV/nzp2D/eSTT0b1WMKkMkbObtezZ89gcxR9AJgwYUKwWRLTrl27qB67/m666aZRGWc4GjVqVLD1Gd95553Bfumll6KyHXbYIdgsG2F3WCDOzMPZAoBYOlut5GVf5Hmk8L7Ie6lmTmQ4wx8QS3J4DGo/tiQ46xLLwDl7BBDvaZxFEYj3NbZ1bLdp0ybYLOXVLGwdO3YMto55XpN5vnE2TSCWzalsmLNCciaMK664AqaGvCxsee79DD9PlZBz3zV1m1gu8Prrr9f7c41pavLGNlPq3Mtj2LBh0TXvp5yNVmWl3EaV2fM9+CzL4SYAYPDgwcF+4YUXorLa30p5Z4QikJdplTMKn3vuuVE9XkP19zmfsfk3yN133x3VY8nbvffem9lGzkDFfQbEYQYagj1tjDHGGGOMMcYYYyoQv7QxxhhjjDHGGGOMqUD80sYYY4wxxhhjjDGmAmn2mDasf9e4BZw+UuOhcGwZ1uypfo+12BxvhfX4QBxnQePdcIwbvp/qITmVtKb4ZN13EeCYEq+++mqwNZ0ZPxONM8JxFdhW3T33Facv1bg1HONE4xKxzn/ZsmWZ9+A0qhpbZ9KkScH+/ve/H+xrrrkGZtXwfGEdKut3gTiV8N577x2V8XjiOCoaD4tT82mcDh4bHI9I01vzmNSUjK+88gqAlVOSNxWcVlLnGKf25nVMU1Hyd+BxDgDrr79+sGu/GxBrcQHg9ttvDzav3RrT5rTTTgu2xsBg/TavyRozZd999w32DTfcEJXx9+R+U30z33+99daLyooQ0yaPvfbaK9j6XHhN5TJdN3luatpwrsv1mmtOVCK873McNU1Z/+yzzwb7nnvuico4vhTPdd77gDjGDfeFnm1ee+21zHtwPB2eD9rXHEtOY5Dxvn7kkUcGu6XHtMmK4dbQen369Am2xmcYOHBgsHv16lVS+0qN56FjYdq0acHm1LgA8Oc//7mkexpTLkod93x24PMGEMfj5Nh+QPy7gT9Lf9/yvNIzalbsTj2HcowbjWlT9Fg2teiZheF+09+fF110UbA1rswDDzwQbD7nzpw5M6qXF8eG40Hyvnv++edH9Tj24MSJEzPvl4U9bYwxxhhjjDHGGGMqEL+0McYYY4wxxhhjjKlAmkUexfIjdm266aabonojRoyo82+A2H2MJUssFQBi92K21R2NXVT53kDsYsV/p27C7I72xBNPoEhoKu899tgj2OPGjQs2p1UH4lRq+lzZDZylcJrKm1P68v01VSq7o6uLPqcoZzdwdVvv379/sN99992ojNMH9+jRI9iWR9WfPBdVTrXOMicg7leWZLCUCYjnuspz2LWV/077myVDWtbU81vnEc8/lo8B8fq3cOHCYG+11VZRPXbPvfbaa6Oym2++OdhTpkwJdrdu3aJ6J598crB5vl188cVRPZbm3HfffVHZBhtsEOznn3++zvbpPbbZZpuobJNNNgk2pxRX91LeD3QsFR1eH3Xt5T7gPZjXZCCep3myJ55jLVkexXOO96399tsvqnf88ccHW+fzbrvtFuyePXsGe+jQoVG9rDWB1y0g3vtOP/30qOzqq68ONsut9B48XlgmDADHHXdc5t+1ZLL2uLwUtXmovJNhSdwxxxwTbB4/QDxOWNYOAIsXLw42y8k1rffo0aODrTJbUzqnnnpqdM1r9PXXX9/czSkUpc6pvLmYJy/q27dvsPW3Ec8xlm7zuVNRSSv/3uV9RPdWlS62FLLeISj82/3ggw+Oyh555JFg6/mS4XMo/zYB4vANXbp0icoOP/zwYPNYYtk/EKeC5z0YWFmOVRf2tDHGGGOMMcYYY4ypQPzSxhhjjDHGGGOMMaYC8UsbY4wxxhhjjDHGmAqkWWLaZOkNJ0yYEF2zzk/TwbI+MCslN5CdzlR1iKx903twLAS+h8YJ0JRrReLQQw+Nrlnjzhp81Tizlo/TNQNxnBDWbaq+k3WcrOHUNHysJdV0vpwynuM2dOjQIarHmnyOlQHEMVRYD6668aeffhomH547qh3mMaSppDkFNY8F1SZz/2s8LJ7fvBbpfOZ4N5ousKnReZQX04Zjy7Amtnfv3lE9Thepz5XvyWscx7cBstddbgMA3HXXXcHW9fTNN98MNj9/7Sduo6Yq5lhZTz75ZLB1TeB7qqa8Y8eOweY0ndUMP2uOi8RrFxDHZuP5p2vesmXLgs1rNBCv0zomWyp33313sI899thg69jjGEC6V2233XbB7tevX7A11huvhXw/1fgPGzYssx08//j+Op+fe+65YOs6yXumav5bMlmpvHWd4/nH65zGTeNzEKdxVy677LJga4whHjMcRxCI105eG/PWjpEjR2a2o2jkpYBmOPZQp06dojI+K/7tb3+LyjhVO6cc5rPrqsgacxrDk39P6dqt/V1k9LdoqbFSOA7JkiVLojLeg/mMqvfjvtL4qByTkdcHjeOpsQ9bCqXGLOJ6v/nNb6IyfsbXXXddVHbSSScFm+f6LbfcEtXj37S6tw4YMCDYfEb9+c9/HtUbM2ZMsHW9cEwbY4wxxhhjjDHGmCrFL22MMcYYY4wxxhhjKpBmkUdlufBpum52B81Lr83ufeqKz7DrG7tGaZnCroWffvppnW0H8lO6VTt//vOfo2t22dxnn32CranT+NlNnz49Kvvoo4+CvXTp0mB/8MEHUT0eF9xP/Pf6d+q+ym7BfD8dV9wOdRNl+Rt/9tSpU2FWDc/7vHSK5513XrB1PrMMgOUzer+8uZg1n9VNmN2h89rbFMyfPz+6ZgmFSkXZHZvbqdKm9u3bB1vloSz/+tWvfhVsTUvKz4jnoj4f7mt1zWbpxa677hps/c7sTty1a9eojOcf92GbNm2ievzZnNIWWNmtuQgMHjw42Pycda3kvYvHlqZF52erY4brttTUo0qWO7yebXhd07PIiBEj6vy7GTNmRPV4LeR6OhfPPvvsYLNEEojTOesYYXiud+vWLSrjvVXlxi2ZLElU3l5yzz33BFvHDEtfVT7Lc5jPLW+88UZUj887evZh2Ruv8ypX5++lMroikyeJuvLKK4N9zjnnBHvSpElRPZY99ejRIyrj58xyjWuuuSaqN3HixMx2ZMlGNFW0nrFbKnkpv/PqsYT1pZdeisr4fMZ/p/dgdA/gdvDfqaRuhx12yLxnkSlVHnXUUUcFW2XIvAerPIpDfrDU/8ILL4zq8dzk38FAvO9yv6kEnVOPNyS0hj1tjDHGGGOMMcYYYyoQv7QxxhhjjDHGGGOMqUCaXR6VB7uUqmsiu3Kym5O6cmZF8lbYPU2jfPNnsVuWtomzzRQNdS1jl0C2NYMWu4UddNBBURk/S37++lzZ3Zf7V/uax5VKm9glme/Prt1A7No6duzYqEzdlU39yHJp3GyzzaLrn/70p8FW92LOcMLZbFTeWKo7eqkZ5dg1vTnQ78PzT+ciZ/jhaPb77rtvVI8zjqgUgp/rfvvtF2zO8Kawe72u6dx+zSiz0UYbBXv06NHBVgkUZ0pQuRtnPmGJh2aB4mfVEuYvZ/nLy37BZdx36qbNe5rKyViKtnDhwvo3toD85Cc/CTZLTHR+rL322sHWvUplgrWoxIHnC88xnW+crU1lEXwm4oyQelbi9WH27NlRGX83lmCq/KaastJkSfgbSt49WFLO2Z40axC71b/33ntRGWed4v1O102+1mxkvN/xOqD7IkuiGuPZlEJtf+jn5f2WyJKn5a2LeRx22GHBVjnFJptsUudnnXXWWVG9gQMHBlv31q233jrY8+bNC/btt98e1Zs2bVqwVU7Bf8d7oe59nH2RZZZAnCmuWsmTPeVJlrLGxtChQ6NrPlfobwg+3/A80t+H/FmamZGvef7pb57dd9892JzxDShORsy6KHV95mdy/vnnR2X8W++pp56KynivvfPOO4N95plnRvU4y9Qf//jHqIwzTXHZbbfdFtXbeeedg215lDHGGGOMMcYYY0xB8EsbY4wxxhhjjDHGmArEL22MMcYYY4wxxhhjKpBmiWnTEB2sxjlhnR+ndc2LS8H6Qk1tyho2jYHBZXkpxYuc/lBjbHAfsj1hwoSoHmsPr7jiiqiMYyKwFp619UAcT4Q12awd1b/TuB+s/+f7qdb3hhtuQCk0tua9mih1LJTK1VdfHV2zLlvnG38294G2ied63pzluZ2njy81DldjodrqvM/nuFFs//KXv4zq9evXL9ga64DjJ7DGnbX6ADB37txgc/wOjqkAxDExhg8fntn2V199NdgaT4Xnt+qzf//732fesyXTuXPnYOvzzIL3T017zymcNY4Gr50aH6OlwmcHjkugae95fmvsl6xU7Zrm+Yknngg2n3O6dOkS1eNxsN5660VlWfES8tK7P/roo1EZ7+Pbb799sDlGBwBMmTIFlUpjrO+6B3EsxLy5yDHEOF6QPr+sOFRa1qpVq8zP4r/Tc25W7EaNo8GfxalxAeBPf/pT5md/FUpJxVxqDK88eE8DgO9973vB5jTPOo922WWXYA8aNCjYmq6b9zE9D3MsK153+TwExGflX/ziF1EZf0+OIcVx74B4jGicqyLEtMk7h5Z6XuUzzLHHHhuVPf7448HWec9rPc8xjUnGZTo++XzGsXC0Hl/36dMnKss7d1U7ef3G++QhhxwSbD1D9u7dO9gaT3HzzTcP9n333Rfse+65J6o3derUYGtMoeeffz7Yp512WrA5Dg4AHHnkkcHmuFkAMGbMGKwKe9oYY4wxxhhjjDHGVCB+aWOMMcYYY4wxxhhTgTSLPKpU2PVRU1Cyuym7darchV3VuJ6mWGPXNXUNZnkF1+P7FZ0891J+Xipj43TOmr6ZU5GyrEPdcTnVLEugVOI2c+bMYOelaGa3UXVNz6MlS6KYvLFQampNdqvu379/VMYuh5omkWUY/FmaTpbXC3UlZ7dUntvaXu5jdpcEVpYBNjV54y3rmatLPq9XeeOeP0vXOHYL5xTi+oz579idG4glH+xSquvup59+GmxOd5uHuio3NL1rtcJu77ymqhQia2/VtN78d/pseQ9liUxLQmWGLBNk2cE666wT1eOU3x9++GFUxpJBXidVqsgu+yyLGzt2bFRPpcIMSxBZksHpUAHgoosuCvZxxx0Xlc2ZMyfYnG5cx0sl09D9nNc5nWNZa4+OBXad5/3or3/9a1SPn227du2iMl7rWQqj6w+IxTQAAB2TSURBVHeeNISvee3VdZnn/RFHHBGVXXDBBWgKslJ+l7q+9+jRI9inn356VMbjV5/X9OnTg81nW54PAPDrX/862Px8NJVwr169gq2S4jPOOCPYN910U7D17LHtttsGm8+8QHze5r2a/x2Iv6emlt9jjz2C/cILL6AlwdJUlvqxbBGIzzD8vIB4LvLc0TMSl+X95uR1VCVWDKeOrlYa4zfWvffeG2wev/o7o2/fvsFWKSqHATj55JODffjhh0f1WL6kUlE+9/JcvOqqq6J6/BtnxowZqC/Vs8saY4wxxhhjjDHGtCD80sYYY4wxxhhjjDGmAimrPEpdxFjapK7+XJddJDWLRanZo/KkU1lR9dWVstRsHUVDJVEMu4hrBHt2wWVXfn3+nNGkdevWwVb3YZZEqcSKP5vL8twNTWnkyeOYTp06BZuz/7BMDojdzDXrSlY9zRDFY0glVgy7Y2r2N75Hz549o7I77rgj856NgbrS5rmKluoiztIXXWvZBVfLGJZacBvz+knXZJawsqxG3cW5f9kdOY+WJofS5879yM9Zn0vW3NFxxuuj9iOvy3lyxCJLSfOyZvFcGTduXFTGEg1dM1nqxP2mWWRYRsr9rvsit1E/a+HChcFmScY222wT1Xv33XeRBX82y3by1oRyoOe1PCkvj+G8PU0lUQyvZ7xfHHDAAVG9xx57LNg8j1RWypK6xYsXR2Usi+Szjsr3+HvqGYnPryyfVVkeryvaxtprzTT3VclaQ1jSt++++0ZlO+20U7B5n1m6dGlUj+fL+uuvH5Xxnsl/d+GFF0b1Tj311GC3bds22Jy9Bogzvum5hJ/rwQcfHGyVfjOceQyI+5vP1/r8eEwfeOCBURlLQ6pVHpWXDY6fhda75ZZbgs3SUe43IF6jVWrOzz0vxEeefJT3XW5vnlSxCBkcG+Os8NZbbwWbM0bp82FZqmZaHThwYLBZmq/n0J/97GfB1j1kn332CTZnedN2jB8/Ptg33nhjVDZy5MhgZ+1D9rQxxhhjjDHGGGOMqUD80sYYY4wxxhhjjDGmAvFLG2OMMcYYY4wxxpgKpKwxbVR/m6dLZP0YawVZZ6+whlDvrfpehrW+/HeqYeN01KYGfl4aR4g173nPlccFl33yySdRPdYK5mnUWReq6TeLQF76ZYU1pPzMVG+bp93P0lpeeuml0fWAAQPqvB+n3ARinbz2I8911g6rXjgvjg1r01k7rM+NteScMrQclJoKMa8ex5fSOAW8xvHz1/hSrPnnOBcca0rRvhgxYkSwTzrppGDnpbHNW59bMjruuc95XupYyEp1nzfvtR91/a1Fx0Kp8YiqEU2ly3C8GI6VAMR692eeeSYqO/PMM4PNKYfXXXfdqB7HTOF5qvFEOFYXx80A4r7ffffdg619PXr06GD/8Ic/jMo4lhzH6qm0Oat7WN6eViocS0bTO3OaV+6TP/7xj1G9bt26BZvTek+ZMiWzns5nPktxTBVdv3lN1ZhDvBdy/+elBtd71O6TTz/9NBqLb37zmyG+yCmnnBKV6W8Ghr8Px83TOIb8vDTuIj8v7s/LL788qtenT59g8/PRM2/v3r2DvWDBgqjs2WefDTb/jtFUzhxrRecz93dezBRuo95D04g3Bbpvcbv1vJoVpy4vdprGfsmKN/roo49G13y+GTt2bLC33HLLqB6nYed4nEAc44b3YF1v+Hyp34WfDz8Pjd3I37novz+zzra8LgLAiSeeGGyOW8N7EwC8/vrrwd5zzz2jMk7L3aVLl2BzLCMA+N///d9g33bbbVHZ9ttvH2wef1deeWVUj/uX1xEgjtmlv5NqsaeNMcYYY4wxxhhjTAXilzbGGGOMMcYYY4wxFUhZ5VEdO3aMrtn1Ut082U2MXY/U7S5LdqMun/x36lrHf8euhOrupm3MolSpQ7XA7l36TDhNpaYRZffNbbfdNtgqcWM5C7ubanpGdg9Ul3x2cZs/f36w2eW1KOSldyyV+riOd+/ePdjsVsjuh0CcJnPIkCHBVhdrlrktWrQoKuOxxn+nLs+cblddY3l8ZUn0gHjeq5tz7Xd+4403UEnkrS38jHSd5DIeL3nuuPzs1EWY63344YdR2fHHHx9sdhHXdIecarHSpBaVgkqR+LnzuNd+5HHCc13nIqfDZZlN3j00pWWR5VFdu3aNrlmCuMsuuwR72rRpUT2Wy+RJ11gCpXOAzyI8f1Xyy+eZPMkEjxftQ/4sXQt5TeZ6uiaUGx3bgwcPDrbKbPgsx89W03UfddRRwWaZGAA8+OCDweZ9TO/BMgmeb3o24T1N5z2v5/w99SzLa7ues3jc6XrBsIRIx+QWW2yxUnu+Kq1atcJBBx0EYOXfCCzRzEupnCcL4+evc5Hlb3m/R2bPnh1slsewhFjby30NxHKKHXbYAU0Jn4F1LD3wwANN+tlA/m+lUs+eeb+dsuRQQPw7hM8YQCyX4v5gmQoQz4G8cyPb+r24TNOG8z15Lupn8XxW+Wy1ozK5rHFx7rnnRtcs5eWQDJdccklUj+VvvAYAsYT/lVdeCfbUqVOjepwKXucN74WHH354sDt16hTVu/rqq4Ota1NeaIta7GljjDHGGGOMMcYYU4H4pY0xxhhjjDHGGGNMBeKXNsYYY4wxxhhjjDEVSLPEtMlK5V2rh62F9Vx5cWY0ngXDOlf+XP0b1kCqdo41haxT0zaxDlzTr3Fch6LFtMlKyQfEuuK5c+dGZTNmzAg2P2NNJcupUznVno4jTo+q+nJOD80268SLiqZ3ZjheEOtjdWxzatsTTjghKuvbt2+w77rrrmDff//9Ub1+/frV2QZOYQmsHMeG4VgQrP/MSyGpcMwHHp95MWD0edSmgMxKw1eJsG46L10mj4M8vTbH79D4UrzeaYwNnn88rlSHzs9f0xgzRVtP64OmnM7S0+fBz0xTPXPMCo7vkEep9YqAxn7h+BA8j/S8wTElNt5446js4YcfDjbHj+GU3EC89vD+qfERsuY2EO8NvBZqDB7+LN0zOfYKpzb/3ve+F9XjtMvlYNiwYdH1gQceGOxZs2ZFZXxW4eene9OPf/zjYGv8gfbt2wd76623DrbOMY5twumdNa4M70G6zvFc5/U1L7aHxk/ge/A4yUurremia8drKbEYSmXZsmUYNWoUgPi7AUCbNm2Crc+E9yTeg3R+8PfjPgOy43ZpbB2O38RjZ/LkyVE9jp2hZwfuX97vdN5wjDD9znz/vLh+5UZTLHMqd42F9eqrrwb75ZdfDrY+F34WPC4AYKeddgo2n1f1nMhrMY/h999/P6rH41DPSDxveWzpvsj313ZknW3z4lNqvLLaMV9pscVKJS+20WGHHRbsKVOmRGXXXXddsC+66KJgazptjku0//77R2UcY3Xo0KHB1t80l112WbB79+4dlfG4GD58eLCvv/76qB7PdR1LpfSdPW2MMcYYY4wxxhhjKhC/tDHGGGOMMcYYY4ypQMqa8ltTYbG7m6bkYjfkLLkVkO3qr3KHPNdTdtNidzdNmciSKHX/GzNmTEntLTLahyyD4mepaSRZ6sQuhpqClt1B1W2X3cy4nrrg5UktqkWG8eSTT0bX7I6dl0Kbn6fKXbh/1JXz9ttvD/bixYuDPXDgwKgeu36/9dZbwVb5Fvex9g+7inLbVR6VJ5/kfuRxoWsCj08du7XrQF4K3cak1PHG7VF5DKeE1Hbzs+Qx0rZt26jezJkzg62pa5lu3boF+7HHHovK+DlzynRtL9dT12IeZ5U8F5saTfPJz5DnaV7Kb94jdVzw/XTf4muup/KPIjNy5Mjo+pBDDgl2bZpiYOXnyjJSdtcHgN/+9rfB3m677YK96667RvW4T3n+6vPnucju4gBw0kknBZslQjy/gLivVe7IUq8lS5YEW+U35YbTpwOxLFe/U+fOnYPN6yHLiYF4n9Q9iKU2LHNT+Te3g8+Xum/xHqTjKetMqfM+T7bE6wCvqXrO5c/WM9jzzz8PoOkkGffee290zc9YwyuwdDEvFX2ePJSlVPx89B48Rljupn3dGPDz1/7k78JjTmWcLNfbb7/9ojJ+xq+99tpXamsWgwcPjq45DTevIUDcr2eccUaw9XcCP/c8+SDvmXqu69q16yrbDsTrhT5b7h/+LvpZfL7Rcy6fd7iP8+RRml6+GkM/5J1fea3hPVKlfywZ32effYKtct2111472Oecc05UxufcSZMmBfu+++6L6rGsT5/3oEGDgs3rsP626t69e7BZCgisvA/XhT1tjDHGGGOMMcYYYyoQv7QxxhhjjDHGGGOMqUCaRR6VFRWaI3wDsduTShc4eje7lOq92cWK76fu9nmZNti1kt0ntU3sjn700UdHZSyPyouKXTTYvVSzBLFbIcswNFI7R3Rnt8S8DEccGRyI3Z3Z3Zdd5IDYjS0vK1mlsfrqq4eMBh06dIjKOPuWSptYSsXupfpsecxqZjSWNbJkRqU17LLKfa9zT126GXY3zpvPPGdVQsL3Z3dTvQfPdXWBrXX9byqpY30kP1lSF4W/t7r5Z8nJ1PWUszKwK6uOF56z2p/cNyx91Kweed/F1MCZe4B4nvL41XHKz5bdr1XSwuMkT/bE4zUv20zRuPnmm6NrlkXzWqvyG37mLA0AYrlplqwXiOU4LIWYN29eVI/7tF27dlEZ99umm24abJ2znE1DpSG8b/Deo/t9OWjTpk2Q6WofcCYaltvrNct6dWznSS30uhZd17gP+FyrewCPGd0zeZ3mMpWJ89qu7eB75EnB+fyk+2JTSVVr26P3X7BgQZ02EO9PfGbhcQ5kS1GAeF9jWyXdnNmN56Wup6VKuvPgPtXfEjzmuExlWnzGHjJkSOb9G5NWrVqhR48eAFbOhMdjSvcZnhOzZ88Otp7Peb3V8wg/d16vNMtsbVZQIN5bVY7I7dWzFI9R7mP9Xnyt4zorrIfON+5vzUJYqeStLTwHNFzK3nvvHWzO2nfEEUdE9Xg/5f342GOPjeqxXKp2XNbyzjvvBJszFv/f//1fVO/tt98O9nHHHReVcXbkO+64A1lMnDgx2CpxKwV72hhjjDHGGGOMMcZUIH5pY4wxxhhjjDHGGFOB+KWNMcYYY4wxxhhjTAVS1pTfnN4SyE83ytpi1itqDIYsfWleykTVtXKsBdaJql6VtZe18UVaOtxPGn+B0zWyrX3NmmPWbWpslTlz5gRbNa2sUWcdqOq681JiVjKbbbYZrrnmmjrL+NmyZheIU+JxHCDV37J+WLXXWbF/tB95zmWlq9V6GmtA9ei1qA6bU3Bqe7M0wnlpSlX73LNnTwDA008/nfk3zUWpcQR4/un3YbL6CcjW3au2ntNx5mn1WV+u8SK4HRpvyNSg84P7gdc27ce8tLFM3p7J11rWUlAtPD//vNh1PJ5ZMw/Eenpe73R94nXso48+CrbGhuI0ojoXec3n9mocMI5Pdumll0ZltWuhUgkxbb72ta+FNV7PZLxGafwZTmM+d+7cYOs84nOGrl9Z8c50rnAsDr6HxjzhearPlj+L11T+HlpP28tznceC1uOzAZ95gS/HSV5cuobQkFg5PNY5hovGTDTNw+eff4433ngDQHw+A4ANN9ww2Hqu537ksa0xvrhM+5jjWfG80s/KitumaznH9tM4M5zSmf+Of58A8Zqt45t/x/Jc0nnFa5j+9q1d7xoSJ6UUdH3jtuk+w2eRvLncq1evYPfv3z8qGz9+fLDHjh0b7ClTpkT1pk+fHuxu3boFW3+nXHLJJcHm2Df6WY8++miwdRzw3qD9WyqcXlxjrJZCyzx5GWOMMcYYY4wxxlQ4fmljjDHGGGOMMcYYU4GUVR6laenY3UpdRdkVjF001dWbr9mVlV3pgNjVS13Q2G2UXZS1HruAsYscEKch41RgRYfd5NQl8s033ww2u/uqmxm7Li9evDjYKvXh+33wwQdRGbuM8z3UHXLNNdcMNrseAk2X3rkxWLRoES644AIAwE9/+tOojF3s81Kc83evz3dl99A810e+J9v8uUDsPq59MGLEiGDfeeedwWaZFxCn9+P+BrJll9p2dgvXVOm1KXbLIafTvuHrPClSVvpSvQevrerKz+nj8z6Ln5fW48/mNVMlbvz8S72HPpumSkFbKXCqZyD+vjw2da/i/uG1Uecby3jU/ZqfNe+RpaauLQJ6ZuG1kF3v9dkNGzYs2EOHDo3KsqQuOhez3NHVhZvvoXJg/jsuU4nVVVddFezvfve7URnLjtgVXyVW5eC9994LKY2HDx8elR122GHB7tevX1S21157BZvngEqFOPVvXtpePqNq6nHuA05prKnbp02bFmyWvAGxO/+PfvSjYF933XVRPT5LaR9z+3kM6njiPmbpNfDl92xJa4ApjS+++CKcxVgGA8Qp1FUWs//++web5ah6buzQoUOwdfzx2slnBz0v8D7G80HPS1xP1xVOC81r78iRI6N6fE89R/J349++2t6s0B3Al7+plixZgsaktg3alqwwCQr34YcffhiVnXjiicHmPRKI1zxe11Q6z+vfrFmzgq3hTNq1axdsTv8NxJKoV155Jdi69+WRNeb0OXG/6e+Mkj6n3n9hjDHGGGOMMcYYY5ocv7QxxhhjjDHGGGOMqUD80sYYY4wxxhhjjDGmAmn2mDase1ZtGutvVT/Hmj/WhKmWkTX6rAHXWCZ8D9Wf8z25jao/43toDIG+ffsGuyXFtOF+U/0i6z2XLl0abI1twXEDWMOpaf34HpzKVK/5fnmxMqqJzz77LKS+Gzx4cFTG2sojjjgiKmMtP2s8NQ1tXupfHuvc31qPdbscC0D1+XfccUew77nnHpRC165do+suXboEm9PVAtk6YNUV8/zW9eLee+8FsPJa0VQ0JJ6S/g3H49I5xnDcBk0zzM85L17MggULgq1xP3jN5zTkGm+DU3qqHpnXf67X0mLaaOw0foY8nvNSiubFROL1UOez9lcteamui4aeWXgt5NTdGs8tD74HP3PtG+5fLtN+4vms44Dvwf2maWx5zmosiaz4Ug1JX9qULF++PLoeNWpUnXYeO+64Y3TNz0LTB3OcDu7/vPPNxIkTS2pHHtdff32wdS7ymfovf/lLVMb91bp162Br2nCOycP1gC/3RWPqA8+Pm266KSrT61q22mqr6JrjGnbv3j0q43HK65Wuqbw+ckrum2++Oao3derUOtuUx5lnnhldc+xJjSXH8ax4vdDzmP7OYWrTq+ed9RpC7TPSvWTjjTcONseVUfh3xnbbbReVcVvPO++8qOwHP/hBsAcNGlTn3wDA1ltvHWyObapxL2+77bZga/w1jtt18cUXr/wlSoDHVt45lM9ib731Vr0/x542xhhjjDHGGGOMMRWIX9oYY4wxxhhjjDHGVCDNrg3ZcMMNg60u8FkpCIHYJSovNSy7h3KqLZXBsEuxyiSyXM41/SPLKdR1vE+fPsG+8sorUSTy0iazy7Cm6GZ5VLdu3YLNaS+B+LmyS6S6cLOruqZ53mmnnYLNbo/a3ieeeALVSu0Y1jnA1/fff39Upte1aDrQzp07B1vd47PmB/cvEM8/Ts9eH1g6wG6Ff/jDH6J6LE3g/layJB5axilbgZXT0peTLNdLXU95HqmEgt1D+buq+zu77U6aNCnzs3iMbLPNNpntYBf9vHGlaXL182rRZ5EnKyuCdEplezwn8lJJ8v7Hz0Hvx+h85lSkWf1RdNQ1m9dalqK8+OKLUT1+dgrPsbznynOY5yxLDgFg7ty5wW7Tpk1Uxmsyy4113j/11FPBPvjgg6Mynqfsjl5qCthq4rXXXit3E+rFjTfeWO4mGLMS+hurIZJa/Z3A1yNGjGhYw5qQ5557Lve62tDU7IcddliwtX8nTJgQbN7TJk+eHNU78sgjg61SMN5rOWX8BRdcENXr2LFjsPl35TPPPBPV49+j+nvn6KOPRl00VH5faniDhswDe9oYY4wxxhhjjDHGVCB+aWOMMcYYY4wxxhhTgTS7PIrdhDXjkrp0Z/0du3qrSxW7Yq211lrBVgkJu/iqmz67+eZl2uA2qdu0uosViawsFkDsrp8nl8nLOsTkZRTj7FScnQiI5TLc1zpe8sZcpaPP/qvAzxmoHLfwrGxNKnOqZplbXeRJRRtyD3UHZakhR+l/8skno3q6Rteic5bnn7p8nnvuucF+/fXXg81R/4E4q4qun3nyEqYxnlslkycb5u+etzZwPZW0cH/nSc9YWtNcGdUqAc2IyNm8WBbI4xyI95lqG5fTp0+PrtlVnWVU48ePb64mGWOqiJaUYbBIbLHFFrjiiisArBzugs+AGjqEf3+NGzcu2GeffXZU7+WXXw720KFDozLOlMX2LrvsEtV75JFHgs1tVMncs88+G2zNuJtF3hlIz8A8xvN+07IkPet8nYc9bYwxxhhjjDHGGGMqEL+0McYYY4wxxhhjjKlA/NLGGGOMMcYYY4wxpgJp9pg23bt3//LDv5H98VrGsUhYO6YpS1lzxmkxVeOfp/9nnT/H+tC4CvzZGtOGU1+3a9cu2Kz1q1by4iUsWrQo2BrThtODvv3228HWVH4c24K1gRp76OOPPw62pvzmMdK2bdtga+ybpUuX1vEtamjMmDHGNJRS01prHBhenzgGCRDHuOEU6X369InqbbbZZsFu1apVsDUWCscY4nkJxHpkjmOjel6Os6NrR9Zc1GeRF4OsCGy++ebRNe9JvN+pxpzXUe4r3Wc///zzYOt44mfL99eU00Xm8ssvj64vvPDCYHN67VtvvbXZ2tQY5KXk5e8IAHvuuWewBwwYEOzly5dH9TTlqjHGmOph3rx5GDRoEACgffv2Udnee+8d7P333z8q4zh3v/71r4O9xx57RPX4zDdv3ryojPfTM844I9gzZ86M6r3wwgvB3m677YLN6cQB4Omnn0YWfLbJS+vNZRqnqdRYdfwuoyGxb+1pY4wxxhhjjDHGGFOB+KWNMcYYY4wxxhhjTAXS7PIodgdSuQu76bMrPgBstNFGdd5D3eizXL/VdYld87UdLCvIcisHYimQuqOz23q1pfhcFXnp++bPnx9sTucGxFKLddZZJ9j67FjWwDIzTucOxBIPvh8AdOzYMdjcF1dffXVUr9QUwaW6zxnTEHTs5Y2xrLGoMhVe4zQFIV936dIl2Jr6neWELJ3R9vIc07nIbq6MpormdVfvwZKM4cOHB1slUNyu+jzTaoHTVgKxyzKvo7w3AbF8tEOHDsFWKdtee+2VWcb7JLsyT5w4sZSmF4Ldd989uu7Xr1+w2e35vffey7yHStJ4Py31rFCqC3ep5P0NpysHgNatWwebx4SexYwxxlQvaZqG0B8axoKvR44cmXmPIUOGZJZ16tQp2PwbH4jPh3z2VGk+n3tmzJhR59+visY4G/I99F0B87vf/S7YHDKkVOxpY4wxxhhjjDHGGFOB+KWNMcYYY4wxxhhjTAXilzbGGGOMMcYYY4wxFUizx7S5+eabg33AAQdEZZz+a86cOVEZp4HmNM0c8wSI0ztzXBxNrTVr1qxgT58+PSrj1JWsP9f4CXxPjbfC6VcXLFiAloimWFt//fWDzbEzVE/IesBp06YFe5NNNonqcT9p/IXRo0cHm/s6T2tYahwRYxqb+mhqs+pyam0AeO655zL/hq/z4jpxGcfIKbUeEGuLOX4Hx8EBELTTwMrraRZFjFuTR54+nJ+nxiEplY033jjYml6c90nej1sSDz30UHTNcdp4/GpMIUZjwmXNxUqBz1sA8Jvf/CbYfAZ67LHHmq1Nxhhjqhv+na+/+auZvN+ZeWeDUrCnjTHGGGOMMcYYY0wF4pc2xhhjjDHGGGOMMRVIUh933CRJPgTQMrU+5aV9mqYbrLraqnEflhX3Y/XjPiwG7sfqx31YDNyP1Y/7sBi4H6sf92ExqLMf6/XSxhhjjDHGGGOMMcY0D5ZHGWOMMcYYY4wxxlQgfmljjDHGGGOMMcYYU4H4pY0xxhhjjDHGGGNMBeKXNsYYY4wxxhhjjDEViF/aGGOMMcYYY4wxxlQgfmljjDHGGGOMMcYYU4H4pY0xxhhjjDHGGGNMBeKXNsYYY4wxxhhjjDEViF/aGGOMMcYYY4wxxlQg/w9+e9RDtrcl6gAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 1440x288 with 10 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "df = pd.read_csv('fashion-mnist.csv')\n",
    "\n",
    "y = df.loc[:, 'y'].copy().values\n",
    "X = df.drop(['y'], axis=1).copy().values\n",
    "\n",
    "n = 10\n",
    "plt.figure(figsize=(20, 4))\n",
    "for i in range(n):\n",
    "    ax = plt.subplot(2, n, i + 1)\n",
    "    plt.imshow(X[i].reshape(28, 28))\n",
    "    plt.gray()\n",
    "    plt.title('{}'.format((y[i])), fontsize=15)\n",
    "    ax.get_xaxis().set_visible(False)\n",
    "    ax.get_yaxis().set_visible(False)\n",
    "\n",
    "plt.show()\n",
    "plt.close()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Split data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_valid, y_train, y_valid = train_test_split(\n",
    "    X, y, test_size=0.5, random_state=random_state, shuffle=True, stratify=y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Counter({7: 519, 2: 515, 3: 508, 8: 507, 6: 506, 1: 500, 9: 500, 4: 494, 0: 492, 5: 459})\n",
      "Counter({7: 520, 2: 515, 3: 508, 8: 506, 6: 506, 9: 501, 1: 500, 4: 494, 0: 492, 5: 458})\n"
     ]
    }
   ],
   "source": [
    "print(collections.Counter(y_valid))\n",
    "print(collections.Counter(y_train))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Applied algorithms "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def print_accuracy_confusion(clf, X, y):\n",
    "    print('Accuracy on validation set: {}'.format(\n",
    "        (y_valid == clf.predict(X_valid)).sum()/len(y_valid)))\n",
    "    print(confusion_matrix(y_valid, clf.predict(X_valid)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## SVM with Linear Kernel "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "lin_clf = LinearSVC()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/alireza/anaconda3/envs/ml-project/lib/python3.7/site-packages/sklearn/svm/base.py:929: ConvergenceWarning: Liblinear failed to converge, increase the number of iterations.\n",
      "  \"the number of iterations.\", ConvergenceWarning)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "LinearSVC(C=1.0, class_weight=None, dual=True, fit_intercept=True,\n",
       "          intercept_scaling=1, loss='squared_hinge', max_iter=1000,\n",
       "          multi_class='ovr', penalty='l2', random_state=None, tol=0.0001,\n",
       "          verbose=0)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lin_clf.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on validation set: 0.7628\n",
      "[[317   9  20  27   7   1 100   2   9   0]\n",
      " [  2 477   5   7   3   0   5   0   1   0]\n",
      " [ 11   6 344   6  44   2  93   1   7   1]\n",
      " [ 36  20  16 384   7   1  36   0   4   4]\n",
      " [  5   1  65  20 258   1 136   0   8   0]\n",
      " [  5   0   7   3   4 364   4  44  12  16]\n",
      " [ 60   6  61  17  34   1 318   0   9   0]\n",
      " [  1   0   0   0   0  23   3 455   3  34]\n",
      " [  7   1  12   6   3   4  19   4 450   1]\n",
      " [  3   0   1   1   0  16   4  28   0 447]]\n"
     ]
    }
   ],
   "source": [
    "print_accuracy_confusion(lin_clf, X_valid, y_valid)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## SVM with RBF Kernel "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Hyperparameter testing"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Choose the best gamma based on performance on validation set: "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# https://scikit-learn.org/stable/auto_examples/svm/plot_rbf_parameters.html\n",
    "gamma_range = np.logspace(-8, 3, 12)\n",
    "results = []\n",
    "rbf_classifiers = []\n",
    "for gamma in gamma_range:\n",
    "    rbf_svc = SVC(kernel='rbf', gamma=gamma)\n",
    "    rbf_svc.fit(X_train, y_train)\n",
    "    rbf_classifiers.append(\n",
    "        (rbf_svc, gamma, (y_valid == rbf_svc.predict(X_valid)).sum()/len(y_valid)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "best = max(rbf_classifiers, key=lambda res: res[2])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best gamma is 1e-07 with 0.8422 accuracy\n"
     ]
    }
   ],
   "source": [
    "print('Best gamma is {} with {} accuracy'.format(best[1], best[2]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on validation set: 0.8422\n",
      "[[414   2   4  31   1   0  31   0   9   0]\n",
      " [  2 474   8  14   1   0   1   0   0   0]\n",
      " [ 11   0 383   8  60   0  45   0   8   0]\n",
      " [ 26   3   1 462   7   0   8   0   1   0]\n",
      " [  2   0  33  16 384   0  57   0   2   0]\n",
      " [  0   0   0   0   0 401   0  41   5  12]\n",
      " [ 89   1  68  16  41   0 286   0   5   0]\n",
      " [  0   0   0   0   0  15   0 464   1  39]\n",
      " [  3   0   1   2   1   2   9   0 489   0]\n",
      " [  0   0   0   0   0  10   0  36   0 454]]\n"
     ]
    }
   ],
   "source": [
    "print_accuracy_confusion(best[0], X_valid, y_valid)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## K-Nearest Neighbors "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Choose the best k based on performance on validation set: "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "k_opts = [5, 10, 15, 20]\n",
    "neigh_classifiers = []\n",
    "for k in k_opts:\n",
    "    neigh = KNeighborsClassifier(n_neighbors=k)\n",
    "    neigh.fit(X_train, y_train)\n",
    "    neigh_classifiers.append(\n",
    "        (neigh, k, (y_valid == neigh.predict(X_valid)).sum()/len(y_valid)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "best = max(rbf_classifiers, key=lambda res: res[2])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best k is 1e-07 with 0.8422 accuracy\n"
     ]
    }
   ],
   "source": [
    "print('Best k is {} with {} accuracy'.format(best[1], best[2]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on validation set: 0.8422\n",
      "[[414   2   4  31   1   0  31   0   9   0]\n",
      " [  2 474   8  14   1   0   1   0   0   0]\n",
      " [ 11   0 383   8  60   0  45   0   8   0]\n",
      " [ 26   3   1 462   7   0   8   0   1   0]\n",
      " [  2   0  33  16 384   0  57   0   2   0]\n",
      " [  0   0   0   0   0 401   0  41   5  12]\n",
      " [ 89   1  68  16  41   0 286   0   5   0]\n",
      " [  0   0   0   0   0  15   0 464   1  39]\n",
      " [  3   0   1   2   1   2   9   0 489   0]\n",
      " [  0   0   0   0   0  10   0  36   0 454]]\n"
     ]
    }
   ],
   "source": [
    "print_accuracy_confusion(best[0], X_valid, y_valid)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Decision Tree"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DecisionTreeClassifier(class_weight=None, criterion='gini', max_depth=None,\n",
       "                       max_features=None, max_leaf_nodes=None,\n",
       "                       min_impurity_decrease=0.0, min_impurity_split=None,\n",
       "                       min_samples_leaf=1, min_samples_split=2,\n",
       "                       min_weight_fraction_leaf=0.0, presort=False,\n",
       "                       random_state=42, splitter='best')"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dt = DecisionTreeClassifier(random_state=random_state)\n",
    "dt.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on validation set: 0.73\n",
      "[[321   8   8  37   8   3  91   0  15   1]\n",
      " [  4 469   0  19   1   2   4   0   1   0]\n",
      " [ 15   5 310  14  79   3  71   0  16   2]\n",
      " [ 32  24   9 388  19   2  30   0   4   0]\n",
      " [  1   7  85  34 296   0  66   0   5   0]\n",
      " [  2   2   1   2   0 353   0  49  34  16]\n",
      " [ 83   3  74  30  63   1 239   1  11   1]\n",
      " [  0   1   0   0   1  34   1 434   6  42]\n",
      " [  4   8   7  11  12   9  13  10 430   3]\n",
      " [  1   0   0   1   3  24   2  48  11 410]]\n"
     ]
    }
   ],
   "source": [
    "print_accuracy_confusion(dt, X_valid, y_valid)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Neural Network "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Choose the best activation function based on performance on validation set: "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/alireza/anaconda3/envs/ml-project/lib/python3.7/site-packages/sklearn/neural_network/multilayer_perceptron.py:566: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (100) reached and the optimization hasn't converged yet.\n",
      "  % self.max_iter, ConvergenceWarning)\n"
     ]
    }
   ],
   "source": [
    "activation_opts = ['logistic', 'tanh', 'relu']\n",
    "nn_classifiers = []\n",
    "for activation in activation_opts:\n",
    "    nn = MLPClassifier(hidden_layer_sizes=(100, 100),\n",
    "                       solver='sgd', activation=activation, max_iter=100)\n",
    "    nn.fit(X_train, y_train)\n",
    "    nn_classifiers.append(\n",
    "        (nn, activation, (y_valid == nn.predict(X_valid)).sum()/len(y_valid)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on validation set: 0.724\n",
      "[[407   7  10  33   2   0  22   0  11   0]\n",
      " [  8 469   5  13   1   0   3   0   1   0]\n",
      " [ 12   0 385   7  66   0  31   0  14   0]\n",
      " [ 50   2   1 432   4   0  19   0   0   0]\n",
      " [  8   3 129  21 306   0  23   0   4   0]\n",
      " [  0   0   0   0   0 171   0 155  16 117]\n",
      " [113   1 251  11  55   0  63   0  12   0]\n",
      " [  0   0   0   0   0   6   0 459   1  53]\n",
      " [  3   0  10   8   4   2   7   4 469   0]\n",
      " [  0   0   0   0   0   3   0  37   1 459]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on validation set: 0.775\n",
      "[[384   8   5  32   4   1  35   0  23   0]\n",
      " [  2 469   8  14   2   0   5   0   0   0]\n",
      " [ 11   2 240   5 176   0  66   0  15   0]\n",
      " [ 29   3   0 429  25   0  20   0   2   0]\n",
      " [  1   2  22  14 408   0  38   0   9   0]\n",
      " [  0   0   0   0   0 376   0  44   6  33]\n",
      " [113   2  39  15 132   0 186   0  19   0]\n",
      " [  0   0   0   0   0  19   0 449   1  50]\n",
      " [  2   1   2   2   8   5   4   2 481   0]\n",
      " [  0   0   0   0   0  12   0  34   1 453]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on validation set: 0.1016\n",
      "[[  0   0   0 492   0   0   0   0   0   0]\n",
      " [  0   0   0 500   0   0   0   0   0   0]\n",
      " [  0   0   0 515   0   0   0   0   0   0]\n",
      " [  0   0   0 508   0   0   0   0   0   0]\n",
      " [  0   0   0 494   0   0   0   0   0   0]\n",
      " [  0   0   0 458   0   0   0   0   1   0]\n",
      " [  0   0   0 506   0   0   0   0   0   0]\n",
      " [  0   0   0 519   0   0   0   0   0   0]\n",
      " [  0   0   0 507   0   0   0   0   0   0]\n",
      " [  0   0   0 500   0   0   0   0   0   0]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for nn, activation, accuracy in nn_classifiers:\n",
    "    print_accuracy_confusion(nn, X_valid, y_valid)\n",
    "    plt.scatter(list(range(1, len(nn.loss_curve_)+1)), nn.loss_curve_)\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "best = max(nn_classifiers, key=lambda res: res[2])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best activation is tanh with 0.775 accuracy\n"
     ]
    }
   ],
   "source": [
    "print('Best activation is {} with {} accuracy'.format(best[1], best[2]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy on validation set: 0.775\n",
      "[[384   8   5  32   4   1  35   0  23   0]\n",
      " [  2 469   8  14   2   0   5   0   0   0]\n",
      " [ 11   2 240   5 176   0  66   0  15   0]\n",
      " [ 29   3   0 429  25   0  20   0   2   0]\n",
      " [  1   2  22  14 408   0  38   0   9   0]\n",
      " [  0   0   0   0   0 376   0  44   6  33]\n",
      " [113   2  39  15 132   0 186   0  19   0]\n",
      " [  0   0   0   0   0  19   0 449   1  50]\n",
      " [  2   1   2   2   8   5   4   2 481   0]\n",
      " [  0   0   0   0   0  12   0  34   1 453]]\n"
     ]
    }
   ],
   "source": [
    "print_accuracy_confusion(best[0], X_valid, y_valid)"
   ]
  }
 ],
 "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": "3.7.3"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}