Week8 - SVM.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Support Vector Machines"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Here we implement Multiclass MNIST digit recognition classifier with SVMs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Import libraries\n",
    "\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "import numpy as np\n",
    "from pandas.plotting import scatter_matrix\n",
    "from sklearn.preprocessing import OrdinalEncoder\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "\n",
    "from keras.datasets import mnist\n",
    "from sklearn.svm import SVC \n",
    "from sklearn.metrics import confusion_matrix, ConfusionMatrixDisplay, classification_report\n",
    "from sklearn.metrics import plot_roc_curve, roc_auc_score\n",
    "from sklearn.preprocessing import StandardScaler, MinMaxScaler\n",
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.metrics import plot_confusion_matrix\n",
    "from sklearn.model_selection import cross_val_score, GridSearchCV, StratifiedShuffleSplit"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Loading MNIST dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "(X_train, y_train), (X_test, y_test) = mnist.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of training features:  (60000, 28, 28)\n",
      "Shape of training labels:  (60000,)\n",
      "Shape of test features:  (10000, 28, 28)\n",
      "Shape of test labels:  (60000,)\n"
     ]
    }
   ],
   "source": [
    "print('Shape of training features: ', X_train.shape)\n",
    "print('Shape of training labels: ', y_train.shape)\n",
    "print('Shape of test features: ', X_test.shape)\n",
    "print('Shape of test labels: ', y_train.shape)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 720x504 with 15 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize=(10,7))\n",
    "for i in range(15):\n",
    "    ax = fig.add_subplot(3, 5, i+1)\n",
    "    ax.imshow(X_train[i], cmap=plt.get_cmap('gray'))\n",
    "    ax.set_title('Label (y): {y}'.format(y=y_train[i]))\n",
    "    plt.axis('off')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = X_train.reshape(X_train.shape[0], 28*28)\n",
    "X_test = X_test.reshape(X_test.shape[0], 28*28)\n",
    "\n",
    "#Normalising\n",
    "X_train = X_train/255\n",
    "X_test = X_test/255"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's consider the forst 10,000 images in the training dataset and first 2000 images in the testing dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = X_train[0:10000, :]\n",
    "y_train = y_train[0:10000]\n",
    "\n",
    "X_test = X_test[0:2000, :]\n",
    "y_test = y_test[0:2000]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Linear SVM for MNIST multiclass classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training accuracy: 91.07 %\n"
     ]
    }
   ],
   "source": [
    "pipe_1 = Pipeline([('scaler', MinMaxScaler()),\n",
    "                    ('classifier', SVC(kernel='linear', C=1))])\n",
    "\n",
    "pipe_1.fit(X_train, y_train)\n",
    "\n",
    "acc = cross_val_score(pipe_1, X_train, y_train, cv=2)\n",
    "print('Training accuracy: {:.2f} %'.format(acc.mean()*100))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAATgAAAEWCAYAAADy2YssAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/YYfK9AAAACXBIWXMAAAsTAAALEwEAmpwYAABCUUlEQVR4nO2dd5xU5fX/32dmtheWpXdBWhAFCQJiQ+xd80uMiYnGxC9qMJpoYkmMLYlfY+z9i7HXYCxRRCASG7FSFRGkSF/KAssuW2dnzu+PexcWhN07e++zOzM+79frvnbmlnPPPnPnzFPPR1QVi8ViSUdCbe2AxWKxmMIGOIvFkrbYAGexWNIWG+AsFkvaYgOcxWJJW2yAs1gsaYsNcGmGiOSIyOsisl1EXvRh51wRmRGkb22BiLwpIue3tR+WtsEGuDZCRH4sIrNFZIeIlLhfxMMDMP19oAvQQVV/0FIjqvqsqh4fgD+7ISLjRERF5JU99g9z97/j0c6NIvJMc+ep6kmq+mQL3bWkODbAtQEicgVwN3ALTjDqDTwInBGA+T7AV6paH4AtU2wGDhWRDo32nQ98FdQNxME+3992VNVurbgB7YAdwA+aOCcLJwCud7e7gSz32DhgLXAlsAkoAS5wj90E1AFR9x6/AG4Enmlkez9AgYj7/mfACqAC+Bo4t9H+WY2uGwt8Cmx3/45tdOwd4E/Af107M4CO+/jfGvx/GJjo7gsD64DrgXcanXsPsAYoB+YAR7j7T9zj/1zQyI+/uH5UA/3dfRe6xx8CXmpk/6/ATEDa+rmwm5nN/sK1PocC2cArTZzzB2AMMBwYBowCrmt0vCtOoOyBE8QeEJH2qnoDTq3wH6qar6qPNuWIiOQB9wInqWoBThCbv5fzioE33HM7AHcCb+xRA/sxcAHQGcgEftvUvYGngPPc1ycAC3GCeWM+xSmDYuA54EURyVbVaXv8n8MaXfNTYAJQAKzaw96VwIEi8jMROQKn7M5XN9pZ0g8b4FqfDkCpNt2EPBe4WVU3qepmnJrZTxsdj7rHo6o6FacWM6iF/sSBoSKSo6olqvrFXs45BViqqk+rar2qPg8sBk5rdM7jqvqVqlYDk3EC0z5R1Q+AYhEZhBPontrLOc+o6hb3nnfg1Gyb+z+fUNUv3Guie9irwinHO4FngF+p6tpm7FlSGBvgWp8tQEcRiTRxTnd2r32scvfttLFHgKwC8hN1RFUrgR8CFwMlIvKGiAz24E+DTz0avd/QAn+eBi4FjmYvNVoR+a2IfOmOCJfh1Fo7NmNzTVMHVfVjnCa54ARiSxpjA1zr8yFQC5zZxDnrcQYLGujNN5tvXqkEchu979r4oKpOV9XjgG44tbJHPPjT4NO6FvrUwNPAL4Gpbu1qJ24T8irgbKC9qhbh9P9Jg+v7sNlkc1NEJuLUBNe79i1pjA1wrYyqbsfpTH9ARM4UkVwRyRCRk0TkNve054HrRKSTiHR0z292SsQ+mA8cKSK9RaQdcG3DARHpIiJnuH1xtThN3fhebEwFBrpTWyIi8kNgCDClhT4BoKpfA0fh9DnuSQFQjzPiGhGR64HCRsc3AvslMlIqIgOBPwM/wWmqXiUiw1vmvSUVsAGuDXD7k67AGTjYjNOsuhR41T3lz8Bs4DPgc2Cuu68l9/o38A/X1hx2D0oh14/1wFacYHPJXmxsAU7F6aTfglPzOVVVS1vi0x62Z6nq3mqn04FpOFNHVgE17N78bJjEvEVE5jZ3H7dL4Bngr6q6QFWXAr8HnhaRLD//gyV5ETuAZLFY0hVbg7NYLGmLDXAWiyVtsQHOYrGkLTbAWSyWtKWpyaatTlZRjuZ0LQjcbnxJMq87t+wVaf6UFvEtH1OroZI6rfVVuiccnadbtsY8nTvns9rpqnqin/v5IakCXE7XAsb9/fuB2608cnPgNgEIhYO3Gff24CQNJsoAkJCZCKf13+4fu491pm8bpVtjfDy9p6dzM7otb27liVGSKsBZLJZUQInp3uaDJx82wFksloRQIJ4ibX0b4CwWS8LE97qiL/mwAc5isSSEokRtE9VisaQjCsRsE7Xl1N5aQf0HtUj7ELlPFu/cH32pmugr1RCCyKGZZF6Sj9YrdX+tIPZVPcQgcmI2mT/JbcL63hk5rpyL/7SecEh58/liJt/fxdf/0KlbHb+7ZyVFHetBYepzHXn10c6+bJry1aRdU+WQkRXn9heXkJGphCPK+1Pb88yd3Zu/sBlSqWxN2m0O2wcHiMiJOHn1w8DfVfVWT06dmEXkrGxqb6nYuS82t476WbXkPNYeyRR0m1NFjr1di0Yh98litEapPm8rkWOyCHXzPn0hFFIm3rKOa8/pR2lJBvdNXcpH09uxeml2Iv/ubsRiwqSbe7JsYS45eTHuf3Mxc98rYPXSnBbbNOWrSbumyiFaK1x9zkBqqsKEI8odLy1m9tuFLJ6XcN7PnaRa2Zqy2xwKxFIkSYexlQwiEgYeAE7CyR32IxEZ4uXa8PBMpHB316L/qiHz3Fwk05kfJe3d4wLUKFqvUKsQESQvsTlUgw6uYv3KTDaszqI+GuKdfxVx6AnbE7KxJ1s3ZbBsoVOTrK4Ms2ZpNh27Rpu5qm18NWnXVDmAUFPl/IhFIkokoqj6mzuXamVryq4X4h63tsbkUq1RwDJVXaGqdcAL+JDF0zX1xD6LUn3RNqp/VUbsS+dLEh6XBdlC1VlbqPrBFjLOyflGcGyODl2jbF6fufN9aUkGHbsF8SV06NKzlv2HVrF4Xp5vW6Z8NV0GEGw5gFODeeDNRbwwbwFzZxWyZL4/u6lWtq3xme0NRYl53NoakwGuB7snKFzL7jn8ARCRCa4A8uy6sup9GtMYaHmc7IeLyLwkj9obylFV4l/WQwhyX+lA7j86EP1HNfH1ybMaIDs3xh8nreDhG3tStcPMrP9UwEQ5xOPCxJOG8JPRBzJoWCV9Bu77+bEEhypEPW5tTZsvtlfVSao6UlVHZhbtu18m1ClE5MgsRITwkAzH8+1K/b9rCI/ORCKCtA8RPjCD+OLEfsW2bMigU/e6ne87dotSWpLR0n9pJ+GI8sdJK/jPK8X89832vu2BOV9N2QUz5dCYyvIICz4sYOQ4f82zVCtbk59Z0wgxj1tbYzLArQN6NXrfEx8iJeEjsojNcwJXfE29I5zXTpAuYeJznQ9Zq5XYF1FCfRIbO1kyP5cefevo0quWSEaccWeU8dGMdi111UW54vZVrFmWzcuPBDeyZcZXc3ZNlUO74ih5hc660sysOCOOqGDNcn+d66lWtuY+s6ZRIK7etrbG5Cjqp8AAEemLE9jOwREHbpaam8qJz4ui2+NU/b8tZFyQS+TkbGpvraDq/K1IRMj6fQEiQsZZOdTeWk7VeVtBIXJyNqH9E/u34jHhgT/04JbnVhAKw4wXiln1lb8vywGHVHLs97ey4stsHpz+JQCP/7U7n/7H3wNowleTdk2VQ3HnKFfeuZJwGCSkvDelPZ/MLPJlM9XK1pRdLyRD7cwLRjUZRORk4G6caSKPqepfmjq/aHBntdlEkqf/0BM2m0hK8bHOpFy3+ircAw7K1Bfe8DaX8aDe6+ao6kg/9/OD0Xlwrur6VJP3sFgsrYsCUW3z7ntPJOVKBovFkrwoQqztxyc9YQOcxWJJmLjPSdWthQ1wFoslIZzF9jbAWSyWtESI2T44i8WSjjgZfW2AS5j4knojUzqmr58fuE2AE7oPN2I3pTA0rUX59i5rS3ZUhTpNjc8nqQKcxWJJDeK2D85isaQjziCDbaJaLJa0xA4yWCyWNCWVBhlSw0uLxZJUxFQ8bU0hIr1E5G0RWSQiX4jI5e7+YhH5t4gsdf+2d/eLiNwrIstE5DMRGdGcnzbAWSyWhFCEqEY8bc1QD1ypqkOAMcBEV9bgGmCmqg4AZrrvwZE/GOBuE4CHmrtByjVRW6oitGldBn+7vDdlmzNAlJN/soWzLizlydu68uH0dohAUccov717NR261vOfl9sz+YHOqEJOXpxf3bqG/Q+oaRVfrd2msYpl5u02RVCDDKpaApS4rytE5EucrN9nAOPc054E3gGudvc/pU4KpI9EpEhEurl29opJ0ZnHRGSTiCwMymaDitB15/blf8YN4ugzyug9wFvQCUeUCdev55F3F3PPlKW8/kRHVn2Vxfcv2cTDM5fw0FtLGH1sOc/c1RWALr1q+dtLy/i//yzh3N9s4J6rejVzh+B8tXabpkGpa8L4IVx++iBOO38zvQf4T1eeSmVg0m5zKN6ap24TtWODJIG7TdibTRHZDzgY+Bjo0ihobQAaorYnGYTGmGyiPgGcGKRBPypCHbrUM+Ag50uQmx+nV/9aSksyyCvYpf1TUx1C3G6DAw6poqDImcQ6eERVwqmgU01JKZXsWsUys3a9ECfkaQNKGyQJ3G3SnrZEJB94Cfi1qpY3PubW1lqctNJYgFPV94CtQdoMSkVow5pMli/MYfCIKgAev7Ur5353CP95uT3n/e6btd1pzxdzyNEV39jfGr5au03zbVYsazNVLYWYhjxtzSEiGTjB7VlVfdndvVFEurnHuwGb3P0JyyC0+SBDY1WtKLXG71ddGeJPF+7HxTev21l7u+CaDTw7ZxHjv7eN1x7rtNv58/+bz/TnO/CLP6w37pslMaxiWdvgDDKEPW1NISICPAp8qap3Njr0GnC++/p84F+N9p/njqaOAbY31f8GSRDgGqtqZZDV5Ll+VYTqo/CnC/dj/Pe2cfjJ36zKjz9rG7Om7tIKWLEom7t/24sbH/+awuLE1lymmpJSqtm1imVtqarlDDJ42ZrhMOCnwHgRme9uJwO3AseJyFLgWPc9ONnBVwDLgEeAXzZ3gzYPcIngR0VIFe68sje9BtTy/y7ataB/3YpdVfwPp7ejV3+nFrlpbQY3X9iX3927ip77J16zTDUlpdSyaxXLTNptDkWIq7etSTuqs1RVVPUgVR3ublNVdYuqHqOqA1T1WFXd6p6vqjpRVfdX1QNVdXZzvqbUNBE/KkJffJLHzH8W0/c71Vxy7CAALrh2PdOe78Da5VmEQtC5Rx2X/XUtAM/e1ZWKbWHuv9Zp8ocjyv3TvmoVX63dprGKZWbteiFV1qIaU9USkedx5rJ0BDYCN6jqo01dUyjFOlqOCdwXmy4pBTGk1pVyqmUBE4SqVq+hhXrFi2M8nXvFkH+np6qWqv7IlG2LxdKWJIdqvRdSqolqsVjaHkc2MDVGrW2As1gsCaEqxG26JIvFkq7YfHAWiyUtcfLB2T44i8WSltiMvi3HwPSAE3ocHLhNgNMXlQZu87WhwaT92RMJmfnF1fp6I3YlbKgT20A5SFbTK3BaSrzaQGaQAGbJKFbZ3mKxpCkNa1FTARvgLBZLwqSKJoMNcBaLJSGcdEm2iWqxWNIU2wdnsVjSEiebiG2iWiyWNMRZqmUDXOCYUlO64o7VjD62nLLSCBcdMzjh6+f9IY+N72aSVRzn6NecRJrbF4f57KY86quE3B5xRty2g4z8XZlbqtaHePu0IgZNrKL/z71PBzBVBhlZcW5/cQkZmUo4orw/tT3P3Nndt10wp/wUCin3TlnElg0Z3PDzgb7tmSyDM3+2nhPP3ogqrPwqjzuv7k+0zl+QMKku1jS2BoeI9AKewlHEUWCSqt7jx2aDmtKyhbnk5MW4/83FzH2vgNVLc3z5OmNyMa893pHf3bO6Rdf3PquWvufWMO+a/J37Flyfz5DfVdLxkHpWv5TF8seyGXzZLuWnL27LpfMRdXsz1ySmyiBaK1x9zkBqqsKEI8odLy1m9tuFLJ6X3/zFTdCg/HTtOf0oLcngvqlL+Wh6O1Yv9Z+37Myfb2TNsmxy84NJgWSqDDp0qeWM80q46KTh1NWGufaeJRx1ailvvewvGJl6FryQKisZTIbhfYm6thhTakoLP86noqzl83o6jKwns93uefV2rAzRYaQzCbbT2CjrZ+zKHFzyVga5PeIU9E/8i2mqDECoqXLKIBJRIhFFA+hINqX81LFrHYeML2PaC52aP9kzZsoAnISpmdlxQmElKyfO1k2ZzV/UDOaehaZpGEX1q2zfGphU1SpR1bnu6wqgQdQ1EIJUUzJBQf8YG2Y6+fHXT8+keoPzxamvhGWP5jDol1W+7xF0GYRCygNvLuKFeQuYO6uQJfOTV6nqohtW8+gtvdB48+cmgoky2LIxi5ce7c5T787huQ8+paoizNxZRf6dbURrfx/iGvK0tTWt4sEeoq57HktYVSsV1JSG/3kHK1/I5t3vt6O+UghlODW8JQ/k0u+8GiI+n0MTZRCPCxNPGsJPRh/IoGGV9BnoX0zZBKPGl1G2JcKyhcF/mU2UQX5hPWOO2coF47/LuYeNJCsnztGnb27+Qo+09vchKE2G1sD4IENToq7gqGoBk8BJWd6cPRNqSiYo6Bfn0L87Wqo7VobY+J5Ti9n2WYT1MzJZdEcu0QpBBMJZ0Pdc7wMNpsugsjzCgg8LGDluO6u+8tefY0L56YCRFYw5toxR4xaQkRUntyDOVXcv57Zf7+/LbmOCLIPhY8vYuDab7Vud//uDGcUMGVHO26/5b163xfdBgfokqJ15wWiA24eoqw/MqCmZoHaLkNVB0Th89XAu+53tBLDDn9kV4xffn0MkVxMKbqbKoF1xlPp6obI8QmZWnBFHVDD5If/2Gys/bdmQwbgzyrh1Yh9fNh+/rReP3+aIAR00ppz/N2FDIMHNVBlsLsli8PAKsrJj1NaEGH7odpYu9Ddw4dB234dkaH56weQo6r5EXVuMKTWlax5YyUGH7qBdcT3PzP6Cp2/vyvQXOni+fs5v8yn9JIO6MmHG0UUMurSaWJXw9XPOSGG34+ro9b1gRK1NlUFx5yhX3rmScBgkpLw3pT2fzCzy7W9bKj8liqkyWLKggFnTOnDfq58Ri8HyRfm8+Q//AcnUs9AsSdL89IJJVa3DgfeBz4GGruDfq+rUfV1TKMU6Onx88M4E3RPtcvoXNl2SsXRJGf5HGfeKgWchldIlfRyb4VtVq/3gzjr+se97Ovflwx5KW1WtWZAik2UsFktCpEoNLqVWMlgslrbHJry0WCxpiyLUx7/lgwwWiyV9SZWlWjbAWSyWxFDbRG05hkY8TTBlRGArz3Yy9NPEF+B7YeF3g1mQviehbDNTPuK1wUyr2ZNQTvAL0Y2Iw0DSfhdsH5zFYklrbICzWCxpiSLE7CCDxWJJV+wgg8ViSUvUDjJYLJZ0JqhEoKaxAc5isSRI6iy2T6kA51ccprXt5hXU8+tbV9BnYDWqcNfV/Vg8r8DTtWtvilPxPkSKYcBkp0O3eomy/hZF64AwdL9GyB0qlE1VNj+poBDKg+7XCjkDE38ATYnD+CmHfWHiM+vRt5pr7/lq5/tuvWt5+u5evPpEN192k1UsyQ+pUoMzNhQiItki8omILBCRL0TkJr82Z0wu5g/n9gvCvVaxe/H1q5j9bhETjhvGxFMOZM0y73Ow2p8m7Hff7g/RhnuUzhOE/s+H6HKxsOFeJxNMZg/o94gwYHKIzhcK6/+ceIaYBnGY687ty/+MG8TRZ5TRe0Aw87v8lMO+MPGZrfs6h0tPH8alpw/jsjMPoqY6xAczin3bbRCHmTB+CJefPojTzt9M7wH+MwWbem6bQxVicfG0tTUmx3prgfGqOgwYDpwoImP8GPQrDtOadnML6hk6qoLpk52srfXREJUV3ivMeSOE8B5pvUQgXum8ju2AjI7uvYYJ4ULnYco9EKKbEvfXlDiM33LYF6aehQaGj91OyepsNq33nwopWcWS/BBHPG3NISKPicgmEVnYaN+NIrJOROa728mNjl0rIstEZImInNCcfZPpkhTY4b7NcDczyeeSkK49a9m+NcIVt62g33eqWLowj4dv7kNtdcsfyK6/FVZNVEruVohDv8e/+QBtexUKxiZue2/iMINH+BfGMVEOrcFRp5Ty7hTvSU+9kuxiSV5QAm2iPgHcjyMx2pi7VPX2xjtcVb5zgAOA7sBbIjJQVfe5TMfobD0RCYvIfGAT8G9VDUR0JhUIR5T+B1TyxrNduPS0A6mpCnH2xet92dz6otL1SmHw1BDdrhDW3byHVOGnyrZ/KV0ua/umQQMmysE0kYw4o4/ZxvtTgw1wqSCW5I3gRGdU9T1gq8cbnwG8oKq1qvo1sAwY1dQFRgOcqsZUdTjQExglIkP3cs4kVR2pqiMzMJMZtS0oLcmkdEMmSxY4ufdnTSum/1B/NaKyKVA43nldeBxUf7HrWM1SZd2flN53CpGixAOcCXEYMFMOphl5VBnLF+VRtiW4rMKpIpbkFVVvG9CxoQLjbhM83uJSEfnMbcI2FFgPYE2jc9bSjBRpq6y3UNUy4G3gxNa4XzKwrTSTzSVZ9OjrdCYPH1vuW3E8oxNUznFeV34KmY7uCnUlyurfKr3+JGT1aVntrbE4TCQjzrgzyvhohv/c/ibKwTTjTi3lndc7BmgxdcSSvKIqnjagtKEC426TPJh/CNgfp+++BLijpX6aFJ3pBERVtUxEcoDjgL/6selXHKa17T50Yx+uuns5GRlxSlZnc9dV3ke81vw+TuVsqC+DxSfF6XyR0P06oeR2hZgimdDjOieYbX5Eqd8O629VQCEM/Z9J7LfLpDiMn3LYF6Y+s6ycGAcftp17rwtudDJZxZJaijOKaq5upKobG16LyCPAFPftOqBXo1N7uvv2iUnRmYOAJ4EwTk1xsqre3NQ1hVKso0PHGvHHBCEDYiND/msqXZKZ1Ds2XRLEawz1HRtIl/Rx/C3fojM5/btr39u9tTS/POumZkVnXGH4Kao61H3fTVVL3Ne/AUar6jkicgDwHE6/W3dgJjCgqUEGk6Oon+Go2VssljQjqFFUEXkeGIfTV7cWuAEYJyLDcQZsVwIXOffUL0RkMrAIqAcmNhXcIMVWMlgslrZHkcACnKr+aC+7H23i/L8Af/Fq3wY4i8WSMKkyodUGOIvFkhgKmgTLsLxgA5zFYkmYVFlsbwOcxWJJGEOTLwJnnwFORO6jiaa2ql5mxKNUKTlA6+sDt2lqOsfEpV81f1ILeGDAQCN2JWLmtzdeldyrKBoTys0N3KZU+5+/FvBaVKM09RTNbjUvLBZL6qBAqgc4VX2y8XsRyVXV1Pn5s1gsxkiVhlaz9VUROVREFgGL3ffDRORB455ZLJYkRdC4t62t8dIgvxs4AdgCoKoLgCMN+mSxWJId9bi1MZ56clV1jchu0bjJ5REWiyWN0fQYZGhgjYiMBVREMoDLgS/NumWxWJKaJKidecFLgLsYuAcnsdx6YDow0aRTTWFK+cmE3YysOLe/uISMTCUcUd6f2p5n7uzepr5WlESY+buuVJWGQeCAH25n2M/KqCkLMf3yblSsy6CgR5QT7i0hu92uKSsbP8vipbN7c/xdJfQ/aUcTdwjW332RjGXb2nZNqYB5I01qcKpaCpzb0huISBhnysk6VT21pXZgl/LTtef0o7Qkg/umLuWj6e1YvdRfyh5TdqO1wtXnDKSmKkw4otzx0mJmv13I4nn5beZrKKwcdu1mOh1QS90OYfJZfeh1WBWLXy6k59gqvnvRNub8X3vm/l8xY68qBSAegw//1olehyc+iP5tKtvWttugAtZwj6f/OycQFTBPmJmuGTheRlH7icjrIrLZVb/5l4gkkg0wsCatKeUnU3ZBqKlycu9HIkokor77Lvz6mtc5RqcDnPxlmflK+/3rqNwY4euZ+Qw+qxyAwWeV8/VbuwLF508Vsf8JFeQWJz6x+dtUtq1ttzFBqoA1S8M8OC9bG+NlFPU5YDLQDSfJ3IvA816Mi0hP4BTg7y11sDF7U37q2M2//Jopu+D8sj7w5iJemLeAubMKWTLfn5pSkL6Wr41QuiiLLsNqqCoNk9fZGTvK7RRzmrDAjg0RVvw7n6E/btkX8ttatq1htzGmVMD2RQKaDG2KlwCXq6pPq2q9uz0DeK1b3w1cRRMV2nRV1WogHhcmnjSEn4w+kEHDKukz0L/gbxDUVQrTLu3O4X/YTGbB7h+PiLMBzPpLJw79XSnSKuodiZGsZdvamFIBa5JUnyYiIg2N+TdF5BrgBRyXfwhMbc6wiJwKbFLVOSIybl/nuSIUk8BJWd6UTVPKT6bsNqayPMKCDwsYOW47q75qedrsIHyNRWHapd0ZeHo5+5/gDBjkdoxRucmpxVVuCpPTwanNbVqYzYzfOJ3W1dvCrHo3j1BE6XdcZav52xzJVLatabcBEypgzZIEzU8vNPW7PAdncOBsnJTBbwPvAJfgBLnmOAw4XURW4gTH8SLyjB9nTSk/mbLbrjhKXqHTb5WZFWfEERWsWe6vY9mvr6rw9u+70n7/Oob/vGzn/v3G72DxK4UALH6lkL7HOIHvvLe/5rx3nK3/CRUceeMmz8EtCH/3RTKWbWvbbSB4FbDmEfW2tTVNrUXt68ewql4LXAvg1uB+q6o/8WPTlPKTKbvFnaNceedKwmGQkPLelPZ8MrOoTX0tmZPNklcL6TColhdO6w3AmCu38N2LtjLt8u58+WI7Z5rIPSW+/AzK332RjGXb2nbBjApYs6hAEizD8oInVS1XsHkIjfreVPUpzzfZFeCanCZSKMU6Wo7xarbNMZHSx0QKJrDpkhowVb4mMJEu6aPqN9geK/UVnbL69NJu117u6dxVl/yuWVUtkzT7FInIDTiqN0Nw+t5OAmYBngOcqr6D07y1WCzpQBI0P73gZWzs+8AxwAZVvQAYBgTXgWCxWFKPVB9FbUS1qsZFpF5ECoFN7K4ubbFYvk2kQ8LLRswWkSLgEZyR1R3AhyadslgsyU0yjJB6wcta1F+6Lx8WkWlAoatab7FYvq2keoATkRFNHVPVuWZcslgsyU461ODuaOKYAuMD9gVEkIzgZ2NrtK75k1piNxZ83s9w+/aB2wRz0zm2TjFjt/j05UbsIsH3HYWyzCxwj1cHv/RM4wGlAUn1PjhVPbo1HbFYLClCkoyQesEKP1sslsSxAc5isaQrkiIJL22As1gsiZMiNTgvGX1FRH4iIte773uLyCjzrlkslmTEayaRZBhp9bJU60HgUOBH7vsK4AFjHlksluQnRVKWe2mijlbVESIyD0BVt4lIK2bW+yahkHLvlEVs2ZDBDT8PZpqCCdWjK+5YzehjyykrjXDRMYNbbOfXf1rMqKO2ULY1g1+e6VSeDz9+E+dOXEmvflX85pwRLP2i0Le/fsog7+4NZHxaSbxdmPIH9wMg59lSsqZvJ97Oecyqz+tA9JB8wkuqybt/k3OhKtU/7kB0bIHne3XqVsfv7llJUcd6UJj6XEdefbSz5+ubIqjPbE/yCur59a0r6DOwGlW46+p+LJ7n/X/eG6Z89UQS1M684KUGF3WVsRRARDrhUVNHRFaKyOciMl9EZvvwczfO/PlG1iwLJp8W7FI9uu7cvvzPuEEcfUYZvQfU+LY7Y3IxfzjXf56ut17tyh8vOmi3fauW5fHny4eycHYweQ/8lkHtsYVU3NTjG/trzmxP+X19KL+vD9FDHCGbWJ8syu/uTfl9fai4uSd5D2yCmPdvTCwmTLq5JxPGD+Hy0wdx2vmb6T0gmDljQX1me3Lx9auY/W4RE44bxsRTDmTNspZnHm7AlK9eSKcm6r3AK0BnEfkLTqqkWxK4x9GqOjyonFAdu9ZxyPgypr3QKQhzgDnVo4Uf51NRFvZvZ04RFdt3r2yvWZHHupXB5QvzWwb1Q3PRAo//a3YIwk7zReo0YYnNrZsyWLbQ+d+rK8OsWZpNx67BiLgE9Zk1JregnqGjKpg+2Xlm66MhKiv8j++Z8NUT6oyietnaGi9rUZ8VkTk4KZMEOFNV20zZ/qIbVvPoLb3IzQ9uFcHeVI8Gj0hcAzSVMVUGWVPKyPxPObH+2VRd2AnNd76Q4SXV5N2zkfCmKDuu6Loz4CVKl5617D+0isXz/ClqmaRrz1q2b41wxW0r6PedKpYuzOPhm/tQW90GwSkokqB25gUvo6i9gSrgdeA1oNLd5wUFZojIHBGZsA/7u1S1tOkm0ajxZZRtibBsYfI+zJZd1JxcxPZH+lJ+bx/ixRFy/75557HYoBzKH9yP7Xf1JufFrVCX+M99dm6MP05awcM39qRqR/IGi3BE6X9AJW8824VLTzuQmqoQZ1+8vq3d8keK5IPz0kR9A5ji/p0JrADe9Gj/cFUdgZMFeKKIHLnnCao6SVVHqurIDGm6X+2AkRWMObaMJ2ct4Jr7ljNsbAVX3e1/zWJrKD8lOybKQNtHnJpZSKg9oR3hr775AxbvlYXmhAivSmy9cDii/HHSCv7zSjH/fdPM+t2gKC3JpHRDJksWOH2Qs6YV039oarcQguqDE5HHXEH5hY32FYvIv0Vkqfu3vbtfROReEVkmIp81lRCkgWYDnKoeqKoHuX8HAKPwmA9OVde5fzfh9OP5mj/3+G29+OmY4Zx/+DBu/dX+LPiggNt+vb8fk4B51aNUwEQZyNZd+gcZH+4g1sdZlB7aEN05qBDaFCW8to5450SCqXLF7atYsyyblx/xP9ptmm2lmWwuyaJHX2cgZPjYclYv9T/IkCY8AZy4x75rgJluvJnpvgenojTA3SYADzVnPOGeTlWdKyKjmztPRPKAkKpWuK+PB25O9H6tgSnVo2seWMlBh+6gXXE9z8z+gqdv78r0FxIX573qb4s46JAyCouiPDXzA555oC8V2yNc8vultCuOcuODn7NiST5/nDCsxb76LYO820rI+LwKKY9RdP4Kqs7tQMbnVYRX1IJAvHMGlZc6wSiyqJrsf251a3dQeUkXtJ33JuYBh1Ry7Pe3suLLbB6c7nQHP/7X7nz6H/8/SkF9Znvy0I19uOru5WRkxClZnc1dV/kf/TTlqycCan6q6nsist8eu8/A0YEBeBJHz+Vqd/9T6ihlfSQiRSLSTVX3KQHXrKqWiFzR6G0IGAF0UNUTmrmuH06tDZxA+pyq/qWpawpDHXRMxp7B3D+m0iWZSL0TLioK3CZAbNs2I3ZTLl2SBj+0ZyxdUm1t4DY/jr9FuW719eBmd++l+024ovkTgSU3XbEKKG20a5Ir9r4TN8BNUdWh7vsyVS1yXwuwTVWLRGQKcKuqznKPzQSuVtV9TkHzUoNrPBuxHqcv7qXmLlLVFTgCNRaLJd3wXoMr9TNFTFVVpOUz6poMcO4E3wJV/W1Lb2CxWNILwfgk3o0NTU8R6YYjdAWwjt0Fr3q6+/bJPgcZRCSiqjHgML/eWiyWNMPsNJHXgPPd1+cD/2q0/zx3NHUMsL2p/jdougb3CU5/23wReQ14EahsOKiqL7fQeYvFksoEuAxLRJ7HGVDoKCJrgRuAW4HJIvILYBVwtnv6VOBkYBnO3NwLmrPvpQ8uG9iCo8GgODVUBWyAs1i+rQQl7aD6o30cOmYv5yowMRH7TQW4zu4I6kJ2Bbad90rkJhaLJb1IhoX0XmgqwIWBfPa+FNrcvxcykEPKwHQOx66XhSCJYWo6h6ky6HDmCiN2I/8JJv3RnkTHNdll0yJMTOcAoJkpXG1KErvWmKYCXImqJuXEXIvF0oYkyTpTLzQV4No+HafFYklK0qGJ+o1OPovFYgFSvwanqltb0xGLxZI6JEMySy9Y2UCLxZIYadIHZ7FYLN9ASJ0O+pQLcE++P5+qHWHicSFWD5edMdS3TRPqRCaVn5JZAWxPMrLi3P7iEjIylXBEeX9qe565s7una+v/WoZ+WAtFITKe2KXBEXu5kvgrlRAWQmOyCF9ciJbUU3/+ZujlPNKhIZmEr0w8fVIqlS2Y8dcTtgYHIlIE/B0YilMkP1dVT8kym+LqHw+mfFtwGXdnTC7mtcc78rt7Vgdms0H5adnCXHLyYtz/5mLmvlfgO9Fhg/rVtef0o7Qkg/umLuWj6e1YvdRf/joTZQAQrRWuPmcgNVVhwhHljpcWM/vtQhbPy2/22tCJOchZedTfUrZzX3xeLTqrhsijnZBMQbc10uboHiHj0ZaLEaVa2Zry1wupMooa/EzV3bkHmKaqg3FSJ7WZWE1TmFAnMqX8lOwKYN9EqKly7EYiSiSiqEdB4NCwLCjY/dz4v6oI/TgfyXRVudoH53Oqla0pfz2RIpoMxmpwItIOOBL4GYCq1gG+M0+qwi1PLUEVpj7fmTefNzPjPUiCVH5KRQWwUEi5740v6b5fLa8/1Ykl81teDrqmHv28jvpHKyBTCF1SQGiwWx4bYkQv3IzkhQj9ooDQQYnpk6da2baZv2pHUQH6ApuBx0VkGDAHuFxVKxuf5KptTQDIpnmdzyt/MIQtGzNp1yHK/z69mDXLs1n4iX9Vd1OkivKTSeJxYeJJQ8grrOf6ScvpM7CaVV+1sKkeA8rjhB/sgC6OEruxDHm+E3QIE/lHZ6RdCF0Spf66rcgTnZA8042UbylJUDvzgslPP4KTbukhVT0YJ9XSNXuelIiqFsCWjc4v1vYtGXwwvT2DhlU2c0XbYUL5KZUVwCrLIyz4sICR41rejJJOYeTIbESE0HcynSd4exzJFKSd8zjLoAykewRdU9+0sT1ItbJtS3/TSdm+pawF1qrqx+77f+IEvBaTlRMjJy+28/WII8pZuSRZ1YnMKD+lmgJYu+IoeYVOoMnMijPiiArWLG95J7gcnoXOc77UuqYeogrtQmhZDHWVunR9PbquHumeWAMl1cq2Tf39tvfBqeoGEVkjIoNUdQnO0q9Ffmy27xjl+v9bCkA4DG+/1oE57xX59tWEOpEp5adkVwDbk+LOUa68cyXhMEhIeW9Kez6ZWeTp2vqbt6Hz62B7nOj3NxK+oIDQybnE/lpG9GebIQPC1xYhIsQX1BF7fIeTAycE4SvaIYWJ/X6nWtma8tcLyVA780Kzqlq+jIsMx5kmkokjGH2Bqu4zH1BhqIOOyTopcD+0zpSqloEKcDzW/DktwVC6JAmb6VeMvNXy6R5NYSJdkrF0XAa+mx/rTN+qWrmdeung73lT1Zo36Yo5fkRn/GJ0Hpyqzgfa7J+zWCzB0wqiM4GRcisZLBZLEmADnMViSVckmbMNN8IGOIvFkhhJMkLqBRvgLBZLwtg+OIvFkrbYpVotQRWNJjb73BMmpnMAoeyswG1qzMw0ESPlCmi9GbtGpnMA+38a/Dyx5aPMqGqF8vyvXd4TqQrou2BrcBaLJS1JkmVYXrABzmKxJI4NcBaLJR2xE30tFktaI/HUiHA2wFkslsSw8+DMYErIxZTdHn2rufaer3a+79a7lqfv7sWrT3TzZdeE8E6qieT4sbnp5iiVs2KE2wu9/+GMhG+4to7oKudbG9+hhPKFXs9lofXKpj9HqVusaAwKTg7T/oLEvjYmRWfO/Nl6Tjx7I6qw8qs87ry6P9E680k+v/XTRERkEPCPRrv6Ader6t0ttWlKyMWU3XVf53Dp6cMAJ2330/+dwwczin3ZbCBo4Z1UEsnxa7Pg1DDtzg6z8YZdGhld/7dR6u+7ooTynYQbO96KQx30eiGLeI2y5uxa8k8IkdHdexAxJTrToUstZ5xXwkUnDaeuNsy19yzhqFNLeevlVkjjnyI1OGOhXlWXqOpwVR0OfBeoAl7xY9OUkIspu40ZPnY7Jauz2bQ++LlzQZBKIjl+beaMCBHaR5Z7VWXHWzHyT3C/GgLxatB6RWtAMoRQXmLZhswJ+jhZozOz44TCSlZOnK2bEtOhaCmpktG3tZqoxwDLVXVVUAaDFHJpDbtHnVLKu1P8JzkE88I7yS6SY1JspWaeEukgZPZ2Alz+MSGq3o2x8qRatAY6/iZCuF1yyB5v2ZjFS49256l351BXG2LurCLmzioyf2PFSK46E7SWIsc5wPN7OyAiE0RktojMjuJtRrgpIRdTdiMZcUYfs433pwYT4K78wRAuPW0o110wiNN+upGho8oDsQtWJGfHjBj5x+/6v2u+UAjBfm9m0ftfWZQ9GyO6Njk6oPIL6xlzzFYuGP9dzj1sJFk5cY4+fXOr3Fvi3ra2xniAE5FM4HTgxb0d3010huabbyaEXEzaBRh5VBnLF+VRtiWY5oMp4Z1UEckxJbai9Url2zHyj9sV4HZMi5E7NoxEhEixkD0sRO2XyVF7GT62jI1rs9m+NYNYfYgPZhQzZERwP3b7omEeXCo0UVujBncSMFdVN/o3ZUbIxZxdh3GnlvLO6x0DsWVOeCd1RHJMia1UfxIno48Q6bKrCRrpKlR/6lRF4tVKzcI4GfslRxN1c0kWg4dXkJUdA5Thh25nzfLmpTd9o+p9a2Naow/uR+yjeZoopoRcTNkFJwgdfNh27r2un29bYE54J5VEcvza3PiHOqrnxImVwcpTaiieEKHwjIjTPD1h92Z5ux+E2XRzlNVnO90nBaeFyRqQWL3AlOjMkgUFzJrWgfte/YxYDJYvyufNfwT/A703kqF25gXTojN5wGqgn6o2O8xVKMU6Ony8MX+CxmYTwZxIjiFSKptIbvC1sY+qprA9VuqrClpQ1FMPPvJyT+e+//pVaS06UwkE07NusViShqBqcCKyEqgAYkC9qo4UkWKcObT7ASuBs5tS42uK1hpFtVgs6YICMfW2eeNod85sQ03vGmCmqg4AZrrvW4QNcBaLJWEMj6KeATzpvn4SOLOlhmyAs1gsieN9FLVjwzxXd5uwpyVghojMaXSsi6o2pHTeALR45CSlFttbLJbkIIHaWWkzgwyHq+o6EekM/FtEFjc+qKoq0vK6oK3BWSyWxNAEtuZMqa5z/27CWas+CtgoIt0A3L+bWupq8tXgNPj1HRI2s+TIxNQLyQwuQ0hrIGLGX1NiNstHB5tEAaB2ep/AbQJknRDY0u2daNz/90sA8T6AsG87zjSykKpWuK+PB24GXgPOB251//6rpfdIvgBnsViSnoCU7bsAr4gIOLHoOVWdJiKfApNF5BfAKuDslt7ABjiLxZIYAWX0VdUVwLC97N+Ck4HINzbAWSyWBEmOdaZesAHOYrEkTKqsRbUBzmKxJI6twVkslrREgxlFbQ1SKsCZUifKyIpz+4tLyMhUwhHl/anteebO7oHYDoWUe6csYsuGDG74+cBAbJpSUjKh1gWQV1DPr29dQZ+B1ajCXVf3Y/G8Al82TXxmfpXFIneUEvqoCi0KE32kBwDhp7YRfnMH2s75fGI/b098VC5Elcg9W5CvaiEE9ZcUo8MSy+tnUq2rWVIjvpkNcCLyG+BCnOL4HLhAVWtaas+UOlG0Vrj6nIHUVIUJR5Q7XlrM7LcLWTwv37ftM3++kTXLssnNDyatkGklpaDVugAuvn4Vs98t4i8TBxLJiJOV7X8ulonPzK+yWOy4fGKnFxC5rXT3/d8rJPaD3fPqhd6scP6PST1gW4yMP2wken82hLxnMjL1ffBCQNNEjGNsJYOI9AAuA0aq6lAgjKPN0GLMqRMJNVWO3UhEiUQUVf9ZWzt2reOQ8WVMe6GTb1uNaSslpZaQW1DP0FEVTJ/slEF9NERlRRC/q8F/Zn6VxfSgbLTA21cqtCpKfLibm659GPJDyFd1TV+0BybVuprFZvTdaT9HRKJALrDe8P1aTCik3PfGl3Tfr5bXn+rEkvn+FaUuumE1j97SK7DaG5hVUjKh1tW1Zy3bt0a44rYV9PtOFUsX5vHwzX2orfb/xTTxmTUQpLJY+LVyQm/tQAdmUT+hPRSEiffLJPRhFfGj82BTPbK0Ftlcjw5OTlnJ3VAgCQRlvGBSF3UdcDtORt8SYLuqztjzvJaoapkgHhcmnjSEn4w+kEHDKukzsNqXvVHjyyjbEmHZwmDlB00qKZlQ6wpHlP4HVPLGs1249LQDqakKcfbFwfzOBf2ZNRCksljstELqnuhJ9KHuaHGYyCQnb2P8xHy0Y4SMieuJPLyV+JDslFkZLiii3ra2xmQTtT1OXqe+QHcgT0R+sud5iapqmaayPMKCDwsYOc6fOPEBIysYc2wZT85awDX3LWfY2Aquunu5b/9MKimZUOsqLcmkdEMmSxY4fWOzphXTf2gwGqYNBPWZgQFlsfZhCAuEhNhJ+chi90c8LMQuKSb6cA/qb+qCVMbRnim0Djke97a1MSZ/M44FvlbVzaoaBV4Gxhq8X4tpVxwlr9BZ3J2ZFWfEERWsWe4vd//jt/Xip2OGc/7hw7j1V/uz4IMCbvv1/r59NaWkZEqta1tpJptLsujR16ldDR9b7rnTvilMfGZGlMW27EoaEP5vFbqfG8Rq4lDtBACZUw0h0D7J25e6Gw1NVC9bG2OyD241MEZEcoFqnLVls/0YNKVOVNw5ypV3riQcBgkp701pzyczi3zbNYEpJSVTal0AD93Yh6vuXk5GRpyS1dncdZV/hTETn5lfZbHILZsJfVYD22Nk/ngN9T8tIvRZDbK8DgS0S4T6y93ntSxGxu83ggh0DBO9OvGBKFPfBy8kQ/PTC6ZVtW4CfgjUA/OAC1V1nx1thVKso0PHBu+HoXRJSPAVYFPpkkylH3IzQQSOKX81HvzzXjutV+A2wUy6pI/jb1GuW319aO1yu+uhA37h6dzpn/05rVW1bgBuMHkPi8XS2iTHFBAvpNRKBovFkgQ0qGqlADbAWSyWhEmVPjgb4CwWS+LYAGexWNISBQwM1pjABjiLxZIgdpAhqdBYcGtBdzdsQFUrw8xHEsr1Pxl4b8S2B7OK4hvEzXxmobxgl86BmekcALzVI3ibFwc0DckGOIvFkpYoEEuCZQoesAHOYrEkiBrRLzaBDXAWiyVxbBPVYrGkJXYU1WKxpDW2BmexWNIWG+CCx5SKkCm7I8eVc/Gf1hMOKW8+X8zk+/2nNerRt5pr7/lq5/tuvWt5+u5evPpEt4Rt/fpPixl11BbKtmbwyzNHAXD48Zs4d+JKevWr4jfnjGDpF4W+/PWrVLUvTJQtmFEs8/18/W0rfFQDRSF4tOuu/a/sgH/tcLI6js6Gi4pgcR3c6WQNRoHzC+Fw//n3dkMVTE29ChijSZJF5HIRWSgiX4jIr/3amzG5mD+c6z+XWGvYDYWUibes47pz+/I/4wZx9Bll9B7QYkGxnaz7OodLTx/GpacP47IzD6Km2snq2xLeerUrf7zooN32rVqWx58vH8rC2d5yoDVHg1LVhPFDuPz0QZx2/mZ6D/CXWtxU2TYoll121kFccsrBhELKUaeWNn9hM/h+vk7Ig//tuPu+eTXwQTVM6gKPdYWzXRnG/SLwUGdn/60d4a5tZhbGp4jojMmU5UOB/wFGAcOAU0Wkvx+bplSETNgddHAV61dmsmF1FvXREO/8q4hDT/CfUrsxw8dup2R1NpvWtyzV+8I5RVRs370Sv2ZFHutWBjcp2K9S1d4wWbYmFMt8P18HZUHhHl/V1yvhnALIdFO7tXftZ4ecFOkAdQYDTIoEOJNN1O8AH6tqFYCIvAt8D7jN4D2Thg5do2xev+vLUVqSweARwWoRHHVKKe9OaZ0MrkEQlFKVqbI1qVgWOGvr4fNaeGy7E+QuKoLBbpl8WQt/2wYbY3Bt8a6AFxiaMqOoJpuoC4EjRKSDm7b8ZOAbqU+TRVUr1YhkxBl9zDben5oaAS5IpSpTmFQsC5yYQkUc7u/sBLc/bdlVY/pOltNsfbAzPFcefE1OQTXuaWtrTMoGfgn8FZgBTAPmA9/omUw2Va2g2LIhg07ddwn5duwWpbQkuHTkI48qY/miPMq2JL9QSdBKVabK1qRiWeB0CjuDByJOzU2A7XsElD4ZkBOCr/11CeyVWNzb1sYYHWRQ1UdV9buqeiSwDfiquWvShSXzc+nRt44uvWqJZMQZd0YZH80IpuMeYNyppbzzesfmT2xzgleqMlW2phTLjHBYDsx3Wzxroo7qSbsQlNTvGlTYWO8c6xpwjVk1ZWQDjU4TEZHOqrpJRHrj9L+N8WPPlIqQCbvxmPDAH3pwy3MrCIVhxgvFrPrKr6ydQ1ZOjIMP28691/kb+b3qb4s46JAyCouiPDXzA555oC8V2yNc8vultCuOcuODn7NiST5/nDCsxffwq1S1N0yVrSnFMt/P15+3wIJap4b2wxJn6seJeU4/2y82QETg6vZObW5hLTxf4ewT4LL20M5Al0ASDCB4wbSq1vtAByAKXKGqM5s635SqljEMlJ2ptEaSZab5b9MlQbwq2MGjnRhIl/Txxc9RvmSjP1WtcEcdk3OKp3NnVD6V1qpaR5i0b7FY2oLkmALihZRayWCxWJIAu9jeYrGkK4rBLNkBY3QU1WKxpCHqJrz0sjWDiJwoIktEZJmIXBO0q7YGZ7FYEkYDaKKKSBh4ADgOWAt8KiKvqeoi38ZdbA3OYrEkTjA1uFHAMlVdoap1wAvAGUG6aXSaSKKIyGbAi0RRR8B/mgdrt7VsWrvmbCZqt4+qdvJzMxGZ5t7TC9lA41Qvk1R1kmvn+8CJqnqh+/6nwGhVvdSPf41Jqiaq14IXkdkm5tZYu6nla6rZTSVfm0JVT2yte/nFNlEtFktbsY7dE3D0dPcFhg1wFoulrfgUGCAifUUkEzgHeC3IGyRVEzUBJlm7xuymkq+pZjeVfDWOqtaLyKXAdCAMPKaqXwR5j6QaZLBYLJYgsU1Ui8WSttgAZ7FY0paUC3AmlnaIyGMisklEFgZhz7XZS0TeFpFFrqrY5QHZzRaRT0RkgWv3piDsNrIfFpF5IjIlQJsrReRzEZkvIrMDslkkIv8UkcUi8qWIHBqAzUGujw1beRBqcK7t37if10IReV5EAkkOGLRyXdqhqimz4XRELgf6AZnAAmBIAHaPBEYACwP0tRswwn1dgJPNOAhfBch3X2cAHwNjAvT7CuA5YEqANlcCHQN+Fp4ELnRfZwJFBp61DTgTY/3a6gF8DeS47ycDPwvA7lAc7ZNcnAHDt4D+QZZDqm+pVoMzsrRDVd8Dtvq1s4fNElWd676uAL7EedD92lVV3eG+zXC3QEaKRKQncArw9yDsmUJE2uH8KD0KoKp1qloW8G2OAZarqpeVNV6IADkiEsEJSOsDsLlTuU5V64EG5TqLS6oFuB7Amkbv1xJA0DCNiOwHHIxT2wrCXlhE5gObgH+raiB2gbuBq4Cgk+krMENE5ojIhADs9QU2A4+7zem/i0jQqXrPAZ4PwpCqrgNuB1YDJcB2VZ0RgGlPynXfZlItwKUcIpIPvAT8WlUDye+tqjFVHY4z83uUK7LtCxE5FdikqnP82toLh6vqCOAkYKKIHOnTXgSnS+EhVT0YqAQCS7XjTjo9HXgxIHvtcVoafYHuQJ6I/MSvXfWoXPdtJtUCnPGlHUEiIhk4we1ZVX05aPtus+xtIIi1gYcBp4vISpym/3gReSYAuw01GFR1E/AKTleDH9YCaxvVXP+JE/CC4iRgrqpuDMjescDXqrpZVaPAy8DYIAzrt1i5zgupFuCML+0IChERnD6iL1X1zgDtdhKRIvd1Dk4urcV+7arqtaraU1X3wynX/6iq71qGiOSJSEHDa+B4nKaVH183AGtEZJC76xggsBxiwI8IqHnqshoYIyK57nNxDE6frG9EpLP7t0G57rkg7KYLKbVUSw0t7RCR54FxQEcRWQvcoKqP+jR7GPBT4HO3vwzg96o61afdbsCTbrLAEDBZVQOb0mGALsArzveaCPCcqk4LwO6vgGfdH7oVwAUB2GwIwscBFwVhD0BVPxaRfwJzcRRM5xHc8qqXRKRBuW6igcGWlMYu1bJYLGlLqjVRLRaLxTM2wFkslrTFBjiLxZK22ABnsVjSFhvgLBZL2mIDXAohIjE3y8VCEXnRXZ7TUltPuKpGuEudhjRx7jgRSXhiqptF5BvqS/vav8c5O5o6vpfzbxSR3ybqoyW9sQEutahW1eGqOhSoAy5ufNBdyJ0wqnqhNi22O46AZt5bLK2JDXCpy/tAf7d29b6IvAYschfi/01EPhWRz0TkInBWVojI/W4uvbeAzg2GROQdERnpvj5RROa6+eZmuokCLgZ+49Yej3BXU7zk3uNTETnMvbaDiMxwc5P9HSe1U5OIyKvuIvwv9lyILyJ3uftnikgnd9/+IjLNveZ9ERkcSGla0pKUWslgcXBraifhLLAGZx3mUFX92g0S21X1EBHJAv4rIjNwspkMAobgrC5YBDy2h91OwCPAka6tYlXdKiIPAztU9Xb3vOeAu1R1lrtEaDpO6p4bgFmqerOInAL8wsO/83P3HjnApyLykqpuAfKA2ar6GxG53rV9Kc4KgItVdamIjAYeBMa3oBgt3wJsgEstchot+3ofZ63rWOATVf3a3X88cFBD/xrQDhiAkz/teVWNAetF5D97sT8GeK/BlqruK0fescAQd/kVQKGbNeVI3HxkqvqGiGzz8D9dJiJnua97ub5uwUnZ9A93/zPAy+49xgIvNrp3lod7WL6l2ACXWlS7aZJ24n7RKxvvAn6lqtP3OO/kAP0I4WQRrtmLL54RkXE4wfJQVa0SkXeAfaXyVve+ZXuWgcWyL2wfXPoxHbjETdWEiAx0F5C/B/zQ7aPrBhy9l2s/Ao4Ukb7utcXu/gqctOsNzMBZ7I573nD35XvAj919JwHtm/G1HbDNDW6DcWqQDYSAhlroj3GavuXA1yLyA/ceIiLDmrmH5VuMDXDpx99x+tfmiiOi8384NfVXgKXusaeAD/e8UFU3AxNwmoML2NVEfB04q2GQAbgMGOkOYixi12juTTgB8gucpurqZnydBkRE5EvgVpwA20AlTjLPhTh9bDe7+88FfuH69wUBpKy3pC82m4jFYklbbA3OYrGkLTbAWSyWtMUGOIvFkrbYAGexWNIWG+AsFkvaYgOcxWJJW2yAs1gsacv/B7ZgS5xaGEHdAAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y_pred = pipe_1.predict(X_test)\n",
    "cm = confusion_matrix(y_test, y_pred)\n",
    "disp = ConfusionMatrixDisplay(confusion_matrix=cm)\n",
    "disp.plot()\n",
    "plt.title('Confusion Matrix')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.91      0.96      0.94       175\n",
      "           1       0.96      0.99      0.98       234\n",
      "           2       0.89      0.89      0.89       219\n",
      "           3       0.85      0.88      0.86       207\n",
      "           4       0.90      0.94      0.92       217\n",
      "           5       0.89      0.85      0.87       179\n",
      "           6       0.90      0.93      0.91       178\n",
      "           7       0.89      0.87      0.88       205\n",
      "           8       0.89      0.83      0.86       192\n",
      "           9       0.89      0.84      0.86       194\n",
      "\n",
      "    accuracy                           0.90      2000\n",
      "   macro avg       0.90      0.90      0.90      2000\n",
      "weighted avg       0.90      0.90      0.90      2000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Nonlinear SVM for MNIST multiclass classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training accuracy: 82.87 %\n"
     ]
    }
   ],
   "source": [
    "pipe_2 = Pipeline([('scaler', MinMaxScaler()),\n",
    "                    ('classifier', SVC(kernel='rbf', gamma =0.1, C=1))])\n",
    "\n",
    "pipe_2.fit(X_train, y_train)\n",
    "\n",
    "acc = cross_val_score(pipe_2, X_train, y_train, cv=2)\n",
    "print('Training accuracy: {:.2f} %'.format(acc.mean()*100))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y_pred = pipe_2.predict(X_test)\n",
    "cm = confusion_matrix(y_test, y_pred)\n",
    "disp = ConfusionMatrixDisplay(confusion_matrix=cm)\n",
    "disp.plot()\n",
    "plt.title('Confusion Matrix')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.99      0.89      0.94       175\n",
      "           1       0.99      0.98      0.98       234\n",
      "           2       0.47      0.99      0.63       219\n",
      "           3       0.93      0.89      0.91       207\n",
      "           4       0.95      0.82      0.88       217\n",
      "           5       0.98      0.78      0.87       179\n",
      "           6       0.99      0.73      0.84       178\n",
      "           7       0.96      0.79      0.87       205\n",
      "           8       0.95      0.74      0.84       192\n",
      "           9       0.95      0.84      0.89       194\n",
      "\n",
      "    accuracy                           0.85      2000\n",
      "   macro avg       0.92      0.84      0.86      2000\n",
      "weighted avg       0.91      0.85      0.86      2000\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(classification_report(y_test, y_pred))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Using GridSearchCV for HPT "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "scaler = StandardScaler()\n",
    "\n",
    "X_train =  scaler.fit_transform(X_train)\n",
    "X_test = scaler.fit_transform(X_test)\n",
    "\n",
    "C_range = np.logspace(-2, 10, 13)\n",
    "gamma_range = np.logspace(-9, 3, 13)\n",
    "param_grid = dict(gamma = gamma_range, C = C_range)\n",
    "cv = StratifiedShuffleSplit(n_splits=3, test_size=0.2, random_state=42)\n",
    "grid = GridSearchCV(SVC(kernel = 'rbf'), param_grid=param_grid, cv=cv)\n",
    "grid.fit(X_train, y_train)\n",
    "\n",
    "print(\"The best parameters are %s with a score of %0.2f\" %(grid.best_params_, grid.best_score_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = grid.best_estimator_.predict(X_test)\n",
    "cm = confusion_matrix(y_test, y_pred)\n",
    "disp = ConfusionMatrixDisplay(confusion_matrix=cm)\n",
    "disp.plot()\n",
    "plt.title('Confusion Matrix')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(classification_report(y_test, y_pred))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.10.2 64-bit",
   "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": "3.10.2"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "369f2c481f4da34e4445cda3fffd2e751bd1c4d706f27375911949ba6bb62e1c"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}