homework_cv_graditent.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Digit Recognizer\n",
    "https://www.kaggle.com/c/digit-recognizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import cv2\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Загружаем исходные данные"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train = np.loadtxt('./data/digit/train.csv', delimiter=',', skiprows=1)\n",
    "test = np.loadtxt('./data/digit/test.csv', delimiter=',', skiprows=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# сохраняем разметку в отдельную переменную\n",
    "train_label = train[:, 0]\n",
    "# приводим размерность к удобному для обаботки виду\n",
    "train_img = np.resize(train[:, 1:], (train.shape[0], 28, 28))\n",
    "test_img = np.resize(test, (test.shape[0], 28, 28))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(42000, 28, 28)"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "train_img.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Визуализируем исходные данные"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAzsAAAC3CAYAAAA1tUARAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nSUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1365190b8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize=(20, 10))\n",
    "for i, img in enumerate(train_img[0:5], 1):\n",
    "    subplot = fig.add_subplot(1, 7, i)\n",
    "    plt.imshow(img, cmap='gray');\n",
    "    subplot.set_title('%s' % train_label[i - 1]);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Вычисляем X и Y составляющие градиента с помощью оператора Собеля"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_sobel_x = np.zeros_like(train_img)\n",
    "train_sobel_y = np.zeros_like(train_img)\n",
    "for i in range(len(train_img)):\n",
    "    train_sobel_x[i] = cv2.Sobel(train_img[i], cv2.CV_64F, dx=1, dy=0, ksize=3)\n",
    "    train_sobel_y[i] = cv2.Sobel(train_img[i], cv2.CV_64F, dx=0, dy=1, ksize=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "test_sobel_x = np.zeros_like(test_img)\n",
    "test_sobel_y = np.zeros_like(test_img)\n",
    "for i in range(len(test_img)):\n",
    "    test_sobel_x[i] = cv2.Sobel(test_img[i], cv2.CV_64F, dx=1, dy=0, ksize=3)\n",
    "    test_sobel_y[i] = cv2.Sobel(test_img[i], cv2.CV_64F, dx=0, dy=1, ksize=3)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Вычисляем угол и длину вектора градиента"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_g, train_theta = cv2.cartToPolar(train_sobel_x, train_sobel_y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "test_g, test_theta = cv2.cartToPolar(test_sobel_x, test_sobel_y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA04AAAF3CAYAAACSShvNAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzs3X98VNW1///3EiwqglWEEEGMVqoIYpSItuVjVaT68Aeo\n1F9XK15yoVXbWvHWhv741taLN7bXH5WqFaUl0qp4bRHaAhZQ2osVlepYlVZBjEKMAQVErSLg+v4x\nZzDJPjAhZDJnJq/n45FH5qw5Z2ZPWJyZPWfvtc3dBQAAAADYvt3y3QAAAAAASDo6TgAAAACQBR0n\nAAAAAMiCjhMAAAAAZEHHCQAAAACyoOMEAAAAAFnQcQIAAMAuM7MDzewxM1tmZi+a2VVRfD8zm29m\ny6Pf+zY6ZqKZrTCzl8zs1EbxIWb2fHTfbWZmUbyLmc2I4k+aWVl7v050XHScAAAA0Ba2SLrG3Y+Q\ndLykK83sCElVkha6e39JC6NtRfddKGmgpNMk3WFmnaLHulPSOEn9o5/TonilpPXufqikWyTd2B4v\nDJDoOAEAAKANuHu9uz8T3X5X0j8k9ZE0SlJNtFuNpLOj26MkPeDum9z9VUkrJA01s1JJ3d19ibu7\npHubHZN5rIckDc9cjQJyjY4TAAAA2lQ0hO5oSU9KKnH3+uiuNyWVRLf7SFrV6LDVUaxPdLt5vMkx\n7r5F0juSerT5CwBidM53A1pr//3397Kysnw3Awn1t7/97S1379kez0UuYkfaMxcl8hE7xrkR7WHr\n1q3aa6+91Lt3b+27777vpFIpVVRUuCQNGTJEqVRKZrZW0oxctcHMxksaL0ldu3Ydcvjhh+fqqVDg\ndua8WLAdp7KyMi1dujTfzUBCmdlr7fVc5CJ2pD1zUSIfsWOcG5Frmzdv1plnnqmvfOUrmjBhgiTp\nsMMO0+9//3uVlpaqvr5eJ554ol5++eXXJNVJOrDR4X2jWF10u3lcjY5ZbWadJe0j6e3m7XD3KZKm\nSFJFRYWTi9ienTkvMlQPAAAAu8zdVVlZqQEDBmzrNEnSyJEjVVOTnpZUU1OjUaNGZe6aLenCqFLe\nwUoXgXgqGta30cyOj+YvXSppVqNjxkS3vyzp0WgeFJBzdJwAAACwyx5//HFNnz5djz76qMrLy1Ve\nXq45c+aoqqpK8+fPV//+/bVgwQJVVVVJktz9RUkPSlomaZ6kK919a/RwV0i6R+mCEa9ImhvFp0rq\nYWYrJE1QVKEPaA8FO1QPAAAAyTFs2DBt7+LPwoULY+PuPknSpJj4UkmDYuIfSjpvlxoKtBJXnAAA\nAAAgCzpOAAAAAJAFHScAAAAAyCLrHCcz20PSXyR1ifZ/yN1/aGb7KV1/v0xSraTz3X19dMxESZWS\ntkr6prs/EsWHSJomaU9JcyRd5e5uZl2UXhV6iNIlJS9w99o2e5WNlFX9cYf311afkYunBVAAivn8\nUMyvDUD+cG7JD/7u+dGSK06bJJ3s7kdJKpd0mpkdr3QVk4Xu3l/SwmhbZnaEpAslDZR0mqQ7zKxT\n9Fh3ShqndLnJ/tH9UrqTtd7dD5V0i6Qb2+C1AQAAAECbyNpx8rT3os3dox+XNEpSTRSvkXR2dHuU\npAfcfZO7v6p0GcmhZlYqqbu7L4nq7d/b7JjMYz0kaXhUtx8AAAAA8q5F5cijK0Z/k3SopNvd/Ukz\nK4kWKJOkNyWVRLf7SFrS6PDVUWxzdLt5PHPMKkly9y1m9o6kHpLeataO8ZLGS1K/fv1a0nQAQAe2\no+EsDGUBAOyMFhWHcPet7l4uqa/SV48GNbvflb4KlVPuPsXdK9y9omfPnrl+OgAAEu/DDz/U0KFD\nddRRR2ngwIH64Q9/KElat26dRowYof79+2vEiBFav379tmPMbKKZrTCzl8zs1EbxIWb2fHTfbZnR\nH2bWxcxmRPEnzaysfV8lAOTfTlXVc/cNkh5Tem5SQzT8TtHvNdFudZIObHRY3yhWF91uHm9yjJl1\nlrSP0kUiACCRVq1apZNOOklHHHGEBg4cqJ/97GeSpOuuu059+vRReXm5ysvLpfT5TBIfVpEbXbp0\n0aOPPqrnnntOqVRK8+bN05IlS1RdXa3hw4dr+fLlGj58uKqrqyUxFxkAWitrx8nMeprZp6Pbe0oa\nIemfkmZLGhPtNkbSrOj2bEkXRm/4Byt94n0qGta30cyOjz4UXNrsmMxjfVnSo769pacBIAE6d+6s\nm266ScuWLdOSJUt0++23a9myZZKkq6++WqlUSqlUSpLekfiwitwxM+29996SpM2bN2vz5s0yM82a\nNUtjxqTfWseMGaOHH344cwhzkQGgFVpyxalU0mNm9ndJT0ua7+5/kFQtaYSZLZd0SrQtd39R0oOS\nlkmaJ+lKd98aPdYVku5R+iT9iqS5UXyqpB5mtkLSBEUV+gAgqUpLS3XMMcdIkrp166YBAwaorq5u\nR4fwYRU5s3XrVpWXl6tXr14aMWKEjjvuODU0NKi0tFSS1Lt3bzU0NGR23zavOJKZc9xHLZyLrPQX\nAj1y9XoAIImyFodw979LOjom/rak4ds5ZpKkSTHxpZIGxcQ/lHReC9oLAIlTW1urZ599Vscdd5we\nf/xxTZ48Wffee68qKiokKXNVKWeFc4BOnToplUppw4YNOuecc/TCCy80ud/M1B59boo4AShmOzXH\nCQDQ1HvvvafRo0fr1ltvVffu3XX55Zdr5cqVSqVSmW/7D8z2GLvKzMab2VIzW7p27dpcPx0S7NOf\n/rROOukkzZs3TyUlJaqvTxe/ra+vV69evTK75WwuMkWcABQzOk4A0EqbN2/W6NGjdfHFF+vcc8+V\nJJWUlKhTp07abbfdNG7cOEnqGu3Oh1XkxNq1a7VhwwZJ0gcffKD58+fr8MMP18iRI1VTkx7pWVNT\no1GjRmUOYS4yALQCHScAaAV3V2VlpQYMGKAJEyZsi2e+4ZekmTNnStIH0SYfVpET9fX1OumkkzR4\n8GAde+yxGjFihM4880xVVVVp/vz56t+/vxYsWKCqqvT0YeYiA0DrtGgBXABAU48//rimT5+uI488\nMlN2XDfccIPuv/9+pVIpmZnKysqkT+YovWhmmQ+rWxR+WJ0maU+lP6g2/rA6Pfqwuk7pqnxAE4MH\nD9azzz4bxHv06KGFCxfGHsNcZADYeXScAKAVhg0bpriLP6effnqTbTPbnLnNh1UAAAoXQ/UAAAAA\nIAs6TgAAAACQBR0nAAAAAMiCjhMAAAAAZEHHCQVl7Nix6tWrlwYN+mQe/XXXXac+ffqovLxc5eXl\nmjNnzrb7zGyima0ws5fM7NRG8SFm9nx0321RGWhFpaJnRPEnzays/V4dAAAAkoqOEwrKZZddpnnz\n5gXxq6++WqlUSqlUaltVMzM7QunyzQMlnSbpDjPrFB1yp6RxSq+l0z+6X5IqJa1390Ml3SLpxhy+\nHAAAABQIypGjoJxwwgmqra1t6e6jJD3g7pskvRqthTPUzGoldXf3JZJkZvdKOlvptXNGSbouOv4h\nST83M2PR0abKqv643ftqq89ox5YAAAC0D644oShMnjxZgwcP1tixY7V+/fpMuI+ixUcjq6NYn+h2\n83iTY9x9i6R3JPXIYdMBAABQAOg4oeBdfvnlWrlypVKplEpLS3XNNdfk/DnNbLyZLTWzpWvXrs35\n8wEAACC/6Dih4JWUlKhTp07abbfdNG7cOD311FOZu+okHdho175RrC663Tze5Bgz6yxpH0lvN39O\nd5/i7hXuXtGzZ882fT0AAABIHjpOKHj19fXbbs+cObNxxb3Zki6MKuUdrHQRiKfcvV7SRjM7Pqqm\nd6mkWY2OGRPd/rKkR5nfBAAAgKzFIczsQEn3SiqR5JKmuPvPzOw6pauSZcYpfdfd50THTFS6OtlW\nSd9090ei+BBJ0yTtKWmOpKvc3c2sS/QcQ5T+dv8Cd69to9eIInLRRRdp0aJFeuutt9S3b1/96Ec/\n0qJFi5RKpWRmKisr01133aUZM2bI3V80swclLZO0RdKV7r41eqgr9Ekuzo1+JGmqpOlRIYl1Slfl\nAwAAQJ7tqDiVlPsCVS2pqrdF0jXu/oyZdZP0NzObH913i7v/T+Odm5WAPkDSAjP7bPSBNVMC+kml\nO06nKf2BdVsJaDO7UOkS0Bfs+stDsbn//vuDWGVl5Xb3d/dJkibFxJdKGhQT/1DSebvUSBQtqgkC\nQHHhvI6dkbXjFA1rqo9uv2tm/9AnFcjiUAIaAAAAKED5vqqTZDu1jpOZlUk6WukrRl+Q9A0zu1TS\nUqWvSq1XulO1pNFhmVLPm9XCEtBmlikB/Vaz5x8vabwk9evXb2eaDgBAUVq1apUuvfRSNTQ0yMw0\nfvx4XXXVVbruuut09913K1PA5oYbbth2DEPqkStvzblVH7zytDrttY8OqLxDkgo2F+lAoLkWF4cw\ns70l/VbSt9x9o9LD7g6RVK70FambctLCRqhkBgBAU507d9ZNN92kZcuWacmSJbr99tu1bNkySdLV\nV1+tVCqlVCql008/XVIwpP40SXeYWafo4TJD6vtHP6dF8W1D6iXdovSQeiCw95GnqNd5Pwri5CKK\nQYs6Tma2u9Kdpt+4++8kyd0b3H2ru38s6W5JQ6Pdc1YCGgAANFVaWqpjjjlGktStWzcNGDBAdXV1\nOzpk25B6d39VUmZIfamiIfXRUPnMkPrMMTXR7YckDY+qkgJN7HHgIHXas1tLdycXUVBaUlXPlK40\n9g93v7lRvDSa/yRJ50h6Ibo9W9J9Znaz0sUhMiWgt5rZRjM7XumhfpdKmtzomDGSnhAloHOCy80A\nUPxqa2v17LPP6rjjjtPjjz+uyZMn695771VFRYVuumnbwJCcDakHtodcRDFoyRWnL0j6iqSTzSwV\n/Zwu6Sdm9ryZ/V3SSZKuliR3f1FSpgT0PIUloO9R+huFV9S0BHSPqJDEBElVbfLqAADoIN577z2N\nHj1at956q7p3767LL79cK1euVCqVUmlpqa655pqct8HMxpvZUjNbunbt2uwHoEMgF1EsWlJVb7Gk\nuEugc3ZwDCWgAQBoJ5s3b9bo0aN18cUX69xzz5UklZSUbLt/3LhxOvPMMzObuzKkfvWOhtS7+xRJ\nUySpoqKCkSOQRC6ieLS4OAQA4BNbNq7Vm/dP1Bv3XK437rlCP/vZzyRJ69at04gRI9S/f3+NGDFC\nkjITnWVmE81shZm9ZGanNooPia7grzCz2zLj9c2si5nNiOJPRpVNgSbcXZWVlRowYIAmTJiwLV5f\nX7/t9syZMzVo0LbvLWdLujDKr4P1yZD6ekkbzez4KAcvlTSr0TFjotsMqcdOIRdRLHaqHDkAILJb\nJ+17UqW69D5UH2/6l26//XsaMWKEpk2bpuHDh6uqqkrV1dVasGBBb4nFwZE7jz/+uKZPn64jjzxS\n5eXlktLlnu+//36lUimZmcrKynTXXXdpxowZcvcXzSwzpH6LwiH105QuAT1XTYfUT4+G1K9TOpeB\nwNrZP9Gm15/X1g82avXtYzS1/41atGgRuYiiQMcJAFqh8977qfPe+0mSduuy17ZKZrNmzdKiRYsk\nSWPGjNHEiRP3jQ5hcXDkxLBhwxSXEpmSz3EYUo9c6Tny2ibblZVnqLKycrv7k4soJHScgBzaUTVD\nKhkWjy3vNGyrZNbQ0KDS0lJJUu/evaVPzrMsDg4AQAFjjhMA7IKPP/pAa2fesK2SWWPttbQIi4MD\nAJB7dJwAoJV86xatnXmDuh5xYpNKZpmJ0NHvLdHuLA4OAEABY6ge0AGxIPKuc3e9Pfdn2r3Hgeo+\n9Jxt8ZEjR6qmpkZVVVWqqamRpA3RXSwODhQxzqtA8aPjBCBRCuXDx6a6ZXr/xce0e88yvfGrb6h8\n3vd0ww03qKqqSueff76mTp2qgw46SJLqJVE9CgCAAkfHCQBaYY++A3XQd/6wbTvVqEO3cOHCbbfN\nLNM5onoUAAAFjDlOAAAAAJAFHScAAAAAyIKOEwAAAABkQccJAAAAALKg4wQAAAAAWVBVD0BB2VG5\n8qSUKgcAAMUna8fJzA6UdK+kEkkuaYq7/8zM9pM0Q1KZpFpJ57v7+uiYiZIqJW2V9E13fySKD9En\na5XMkXSVu7uZdYmeY4iktyVd4O61bfYqgQ4o23pIAACg4+HzQeu1ZKjeFknXuPsRko6XdKWZHSGp\nStJCd+8vaWG0rei+CyUNlHSapDvMrFP0WHdKGiepf/RzWhSvlLTe3Q+VdIukG9vgtQEAAABAm8h6\nxcnd6/XJyvfvmtk/JPWRNErSidFuNZIWSfpOFH/A3TdJejVa8X6omdVK6u7uSyTJzO6VdLakudEx\n10WP9ZCkn5uZubvv+ksEihPfGAEAALSfnZrjZGZlko6W9KSkkqhTJUlvKj2UT0p3qpY0Omx1FNsc\n3W4ezxyzSpLcfYuZvSOph6S3dqZ9APIvW4cul/OQ8vncQL6sWrVKl156qRoaGmRmGj9+vK666iqt\nW7dOF1xwgWpra1VWVqYHH3xw2zEMqQeQD4X+Pt3iqnpmtrek30r6lrtvbHxfdGUo51eHzGy8mS01\ns6Vr167N9dMBAJB4nTt31k033aRly5ZpyZIluv3227Vs2TJVV1dr+PDhWr58uYYPH67q6mpJDKkH\ngNZqUcfJzHZXutP0G3f/XRRuMLPS6P5SSWuieJ2kAxsd3jeK1UW3m8ebHGNmnSXto/Q3Wk24+xR3\nr3D3ip49e7ak6QAAFLXS0lIdc8wxkqRu3bppwIABqqur06xZszRmzBhJ0pgxY/Twww9nDtk2pN7d\nX5WUGVJfqmhIffSFaGZIfeaYmuj2Q5KGm5m1x+sDgKTI2nGKToxTJf3D3W9udNdsSWOi22MkzWoU\nv9DMupjZwUp/Y/VUNKxvo5kdHz3mpc2OyTzWlyU9yvwmAAB2Tm1trZ599lkdd9xxamhoUGlpqSSp\nd+/eamhoyOy2bXh8JDN0vo9aOKReUmZIPQB0GC2Z4/QFSV+R9LyZpaLYdyVVS3rQzColvSbpfEly\n9xfN7EFJy5SuyHelu2+NjrtCn4ydnhv9SOmO2fSokMQ6pYcQAACAFnrvvfc0evRo3XrrrerevXuT\n+8xM7XGByMzGSxovSf369cv58wH5VOjzdbDzWlJVb7Gk7Z1th2/nmEmSJsXEl0oaFBP/UNJ52doC\noH1QsQ8oLJs3b9bo0aN18cUX69xzz5UklZSUqL6+XqWlpaqvr1evXr20YcMGadeG1K/ONqRe0hRJ\nqqioYOQIOjQWbC8+O1VVD0Db4ZsqAG3B3VVZWakBAwZowoQJ2+IjR45UTU2NqqqqVFNTo1GjRumn\nP/2plB4ef5+Z3SzpAH0ypH6rmW00s+OVrp57qaTJ0cNlhtQ/IYbUA+igWlxVDwAAJM/jjz+u6dOn\n69FHH1V5ebnKy8s1Z84cVVVVaf78+erfv78WLFigqqoqSekh9ZIyQ+rnKRxSf4/SBSNeUdMh9T2i\nIfUTFC16DwAdCVecAAAoYMOGDdP2Lv4sXLgwNs6QegBJlPSpAlxxAgAAAIAs6DihoIwdO1a9evXS\noEGffCG6bt06jRgxQv3799eIESO0fv36bfeZ2UQzW2FmL5nZqY3iQ8zs+ei+2zLrkURl9GdE8SfN\nrKz9Xh0KyVtzbtWqyRfrjalXbItdd9116tOnT5PhUhnkIgAAhY2OEwrKZZddpnnz5jWJVVdXa/jw\n4Vq+fLmGDx+u6upqSZKZHaF0afuBkk6TdIeZdYoOu1PSOKUnRfeP7pekSknr3f1QSbdIujHHLwkF\nau8jT1Gv834UxK+++mqlUimlUimdfvrpkshFAACKAXOcUFBOOOEE1dbWNonNmjVLixYtkiSNGTNG\nJ554YuauUZIecPdNkl6NJjUPNbNaSd3dfYkkmdm9ks5WehL0KEnXRcc/JOnnZmZUj0Jzexw4SFve\naci+Yxq5CAAoCkmfh5RLXHFCwWtoaFBpaakkqXfv3mpo2PZhdttK95HVUaxPdLt5vMkx7r5F0juS\nejR/TjMbb2ZLzWzp2rVr2+7FoOBNnjxZgwcP1tixYxsPG81ZLgIAgPZBxwlFxcwUTRHJKXef4u4V\n7l7Rs2fPnD8fCsPll1+ulStXKpVKqbS0VNdcc027PC8deQAAco+OEwpeSUmJ6uvrJUn19fXq1atX\n5q7MSvcZfaNYXXS7ebzJMWbWWdI+kt7OVdtRXEpKStSpUyfttttuGjdunJ566qnMXTnNRTryAADk\nHh0nFLyRI0eqpqZGklRTU6NRo0Zl7pot6cKoOtnBSk+8f8rd6yVtNLPjowpml0qa1eiYMdHtL0t6\nlDklaKlMB16SZs6c2bj6I7kIoEOIqzhK9VsUCzpOKCgXXXSRPve5z+mll15S3759NXXqVFVVVWn+\n/Pnq37+/FixYoKqq9IL27v6ipAclLZM0T9KV7r41eqgrJN0jaYWkV5SejC9JUyX1iCbvT5BU1X6v\nDoVk7eyf6M3p/6nN6+q0+vYxmjp1qq699lodeeSRGjx4sB577DHdcsstkshFAB1HXMVRqt+iWFBV\nDwXl/vvvj40vXLgwNu7ukyRNiokvlTQoJv6hpPN2qZHoEHqOvLbJdmXlGaqsrNzu/uQigI4gruIo\n1W9RLLjiBAAAgJzJR/VbIBe44gQkVEdeJwEAUJzaq/qtmY2XNF6S+vXrl/Pna2t8BkgmrjgBAAAg\nZ/JR/ZZqo8iFrB0nM/ulma0xsxcaxa4zszozS0U/pze6j+ooAAAAkET1WxSPllxxmqZPKpk0dou7\nl0c/cySqowAAAHRkcRVHqX6LYpF1jpO7/2UnrgJRHQUAgHY0duxY/eEPf1CvXr30wgvpwSHXXXed\n7r77bmWGKN1www06/fT04BAzm6j0l5ZbJX3T3R+J4kOU/rJ0T0lzJF3l7m5mXSTdK2mI0kOiLnD3\n2vZ7hSgkcRVHJarfojjsyhynb5jZ36OhfPtGMaqjAADQji677DLNmzcviF999dVKpVJKpVLbOk2S\n9hAjQwCgVVpbVe9OSddL8uj3TZLGtlWjtqfQK6Qk2Y6qt9RWn9GOLQEA7IwTTjhBtbW1Ld3905Ju\nZ2QIAOy8Vl1xcvcGd9/q7h9LulvS0OiunFVHiZ6XCikAALTA5MmTNXjwYI0dO1br16/PhD8lRoYA\nQKu0quNkZqWNNs+RlKm4R3UUAADy7PLLL9fKlSuVSqVUWlqqa665pl2e18zGm9lSM1u6du3adnlO\nAGgvWYfqmdn9kk6UtL+ZrZb0Q0knmlm50kP1aiV9VUpXRzGzTHWULQqro0xTetLpXDWtjjI9Gi6w\nTumx1wAAoJVKSkq23R43bpzOPPPMzOZHav3IkNUtGRkiaYokVVRU8CUogKLSkqp6F8WEp+5gf6qj\nAACQR/X19SotTQ8OmTlzpgYN2vb2u0HpkSE3SzpAn4wM2WpmG83seElPKj0yZHJ0TGZkyBNiZAjQ\nJnY0t7wQHr+jam1xCAAAkAAXXXSRFi1apLfeekt9+/bVj370Iy1atEipVEpmprKyMt11112Z3T+U\nNFOMDAGAnUbHCUC74lswoG3df//9QayysnK7+zMyBABah44TgA6DThsAAGitXVkAFwAAAAA6BDpO\nAAAAAJAFHScAAAAAyIKOEwAAAABkQccJAAAAALKg4wQArfDWnFu1avLFemPqFdti69at04gRI9S/\nf3+NGDFC69ev33afmU00sxVm9pKZndooPsTMno/uu83MLIp3MbMZUfxJMytrv1cHAACaoxw5ALTC\n3keeom7HnKm3/3jztlh1dbWGDx+uqqoqVVdXq7q6WpJkZkcovWjoQEkHSFpgZp+NFh69U9I4SU9K\nmiPpNKUXHq2UtN7dDzWzCyXdKOmC9nuFAAAUlh0tO1JbfcYuPz5XnACgFfY4cJA67dmtSWzWrFka\nM2aMJGnMmDF6+OGHM3eNkvSAu29y91clrZA01MxKJXV39yXu7pLulXR2o2NqotsPSRqeuRoFAADa\nHx0nAGgjDQ0NKi0tlST17t1bDQ0Nmbv6SFrVaNfVUaxPdLt5vMkx7r5F0juSeuSq7QAAYMfoOAFA\nDpiZ2usCkZmNN7OlZrZ07dq17fKcAAB0NHScAKCNlJSUqL6+XpJUX1+vXr16Ze6qk3Rgo137RrG6\n6HbzeJNjzKyzpH0kvR33vO4+xd0r3L2iZ8+ebfNiAABAE3ScAKCNjBw5UjU16WlJNTU1GjVqVOau\n2ZIujCrlHSypv6Sn3L1e0kYzOz6av3SppFmNjhkT3f6ypEejeVAAACAPqKoHAK2wdvZPtOn157X1\ng41affsYTe1/o6qqqnT++edr6tSpOuigg/Tggw/qpz/9qdz9RTN7UNIySVskXRlV1JOkKyRNk7Sn\n0tX05kbxqZKmm9kKSeuUrsoHoEDtqNqX1DYVvwDkVtaOk5n9UtKZkta4+6Aotp+kGZLKJNVKOt/d\n10f3TVS6jO5WSd9090ei+BB98uFgjqSr3N3NrIvSlaSGKD0M5QJ3r22zVwgAOdBz5LVNtisr0x96\nFi5cGLu/u0+SNCkmvlTSoJj4h5LO2/WWAgCAttCSoXrTlF5XpLEqSQvdvb+khdF287VKTpN0h5l1\nio7JrFXSP/rJPOa2tUok3aL0WiUAAKAFxo4dq169emnQoE/63yzGDABtL+sVJ3f/S8xJcpSkE6Pb\nNZIWSfqOGq1VIunVaIjJUDOrVbRWiSSZWWatkrnRMddFj/WQpJ+bmTGWHwCQTbbhT8Wq8ev+cNNh\n6nT6UXr5jzdvi5//8Z+DxZhvvPFGSdpDLMYMAK3S2uIQJdGkZkl6U1JJdDuna5VQchcAgKZ2cjHm\nT4vFmAGgVXa5ql50gm2Xq0OU3AUAILsdLMb8KbEYMwC0Sms7Tg3Rt1OKfq+J4jldqwQAAOwcFmMG\ngLbR2o5T4/VFxqjpuiOsVYK8KCsr05FHHqny8nJJGiClK0Ca2XwzWx793jez/85OkAaAQrGDxZg/\nEosxA0CrZO04mdn9kp6QdJiZrTazSknVkkaY2XJJp0TbcvcXJWXWKpmncK2Se5QeT/2Kmq5V0iMq\nJDFBUYX0BdS5AAAgAElEQVQ+oDUee+wxpVIpSfpHFGrLCpAAUBB2sBjzBvEFJwC0Skuq6l20nbuG\nb2f/gl6rZEcVmlicriC1ZQVIAEic5osx7zPsYlX9+v8LFmOOfChppliMGQB2WtaOE1AozEynnHKK\nOnXqJEn7R+EdVYBc0ujwzETozdr+BGkASJzmizFLUo8ePViMGYlTVlambt26Zd6ntw2plzRDUpmk\nWknnu/v66L6JSpfD3yrpm+7+SBQfok86+XMkXcVVULSHXa6qByTF4sWLlUqlNHfuXEnqZWYnNL6/\nLStAMgEaAICdx5B6FDI6TigaffqkLwxFk6A3SBqqtq0AuQ0ToAEAaBON1wmrUdP1w3Z2zTEgpxiq\nh6Lw/vvv6+OPP1a3bt30/vvvS1J3SS/ok0nN1QorQN5nZjdLOkCfTJDeamYbzex4SU8qPUF6cvu+\nGgBAEu1oHjSyY0g9Ch0dJxSFhoYGnXPOOZKkLVu2SNIGd59nZk9LejCqBvmapPOldAVIM8tUgGzp\nBGkAQIJl69hQ5Cm/Fi9erD59+mjNmjUqKSmJHVJvZm02pF7SeEnq169fWzwkQMcJxeGQQw7Rc889\nt23bzN6UJHd/W21UARIAALTejobUu3t9Ww+plzRFkioqKigcgTbBHCcAAADk1Pvvv6933313222F\nQ+qlcEj9zq45BuQUV5wAAAAKWCEMUWRIPYoBHScAAADkFEPqUQwYqgcAAAAAWdBxAgAAAIAs6DgB\nAAAAQBZ0nACgjZWVlenII49UeXm5JA2QJDPbz8zmm9ny6Pe+mf3NbKKZrTCzl8zs1EbxIWb2fHTf\nbVEFKQAAkAd0nAAgBx577DGlUilJ+kcUqpK00N37S1oYbcvMjpB0oaSBkk6TdIeZdYqOuVPSOKXL\n8PaP7gcAAHlAxwkA2scoSTXR7RpJZzeKP+Dum9z9VUkrJA2NFoLs7u5L3N0l3dvoGAAA0M4oRw4A\nbczMdMopp6hTp06StH8ULokWbpSkNyWVRLf7SFrS6PDVUWxzdLt5PO75xksaL0n9+vVri5cA5Ewh\nrDkEAHF26YqTmdVG4+9TZrY0ijGOH0CHtnjxYqVSKc2dO1eSepnZCY3vj64geVs9n7tPcfcKd6/o\n2bNnWz0sikDj+XYVFRWSeJ8GgNZqi6F6J7l7ubtXRNuM4wfQofXpk74w1KtXL0naIGmopIZo+J2i\n32ui3eskHdjo8L5RrC663TwO7JTMfLulS5dmQrxPA0Ar5GKOE+P4AXRY77//vt59991ttyV1l/SC\npNmSxkS7jZE0K7o9W9KFZtbFzA5W+kPpU9Gwvo1mdnz07f6ljY4BdgXv0wDQCrs6x8klLTCzrZLu\ncvcpYhw/gA6soaFB55xzjiRpy5YtkrTB3eeZ2dOSHjSzSkmvSTpfktz9RTN7UNIySVskXenuW6OH\nu0LSNEl7Spob/QAt1ni+3Ve/+tVMOGfv0wBQzHa14zTM3evMrJek+Wb2z8Z3urubWZuO45c0RZIq\nKira7HEBoK0ccsgheu6557Ztm9mbkuTub0saHneMu0+SNCkmvlTSoNy0FB3B4sWL1adPH61Zs0Yj\nRoyQpL0b39/W79N8wQmgmO3SUD13r4t+r5E0U4zjBwAgMRrPt4uuhHZVDt+nKVQCoJi1uuNkZl3N\nrFvmtqQviXH8AAAkQvP5dn/6058k6QPxPg0ArbIrQ/VKJM2MKpJ2lnQf4/gBAEiG5vPt/u3f/k1P\nPPHERknV4n0aAHZaqztO7r5S0lExccbxAwCQZ83n20nS97//fd6nAaCVclGOHAAAAACKyq5W1UNC\nlFX9Md9NAAAARYbPF8AnuOIEAAAAAFnQcQIAAACALBiqBwAAUMSyDberrT6jnVoCFLai6zgxFre4\ncLIHgI6F8z6ApCq6jhMAAABaji+dgZZhjhMAAAAAZEHHCQAAAACyYKgeAACAGLIGYMfoOCErJuoC\n6Ihyee7jAzp2FjkD5B8dJwDADuXzy5Ncfljkg2jHw785gF1Bx6kdceUmHm9kQMfGOQDg/wFQCOg4\nAQCQA3wQBoDiQsdpJ+T6itGOHr+jXo0C0D74kL/z+JsBQMeSmI6TmZ0m6WeSOkm6x92r89ykndZR\nx+InuW2tVQz5iOJALiIpkpKLxfieg52XlHxEx5KIjpOZdZJ0u6QRklZLetrMZrv7svy2DB0R+Yik\nKJRc5INs8SuUXETHQD4iX5KyAO5QSSvcfaW7fyTpAUmj8twmdFzkI5KCXERSkItIEvIReZGUjlMf\nSasaba+OYkA+kI9ICnIRSUEuIknIR+RFIobqtZSZjZc0Ptp8z8xeitltf0lvtV+rWoU2to397cbt\ntvGgXD5xC3NRSvbfMcltk5LdvqBtduN2981pLkpFcW5MctukZLcvtm35ysciyEUp2e0ruLYlPBel\nAvybJkhBta8tcjEpHac6SQc22u4bxZpw9ymSpuzogcxsqbtXtG3z2hZtbBs5bGPWfGxJLkrJ/jsm\nuW1SstvXjm3rEOfGJLdNSnb7yMW2l+T20bZteJ9OgI7YvqQM1XtaUn8zO9jMPiXpQkmz89wmdFzk\nI5KCXERSkItIEvIReZGIK07uvsXMvi7pEaXLSv7S3V/Mc7PQQZGPSApyEUlBLiJJyEfkSyI6TpLk\n7nMkzWmDh8p6WTYBaGPbyFkbO0g+JrltUrLb125tIxcTIcntIxfbXpLbR9siHSQfk9w2qQO2z9y9\nrR8TAAAAAIpKUuY4AQAAAEBiFVXHycxOM7OXzGyFmVXluz3NmdmBZvaYmS0zsxfN7Kp8tymOmXUy\ns2fN7A/5bkscM/u0mT1kZv80s3+Y2efy3abmkpyLhZCHSc7BQsi/xsjFXUMutq2k5iO5uGvIxbZF\nPu6aXOZj0QzVM7NOkl6WNELphdCelnSRuy/La8MaMbNSSaXu/oyZdZP0N0lnJ6mNkmRmEyRVSOru\n7mfmuz3NmVmNpP9z93uiajp7ufuGfLcrI+m5WAh5mOQcTHr+NUYu7jpyse0kOR/JxV1DLrYt8nHX\n5DIfi+mK01BJK9x9pbt/JOkBSaPy3KYm3L3e3Z+Jbr8r6R9K2ErXZtZX0hmS7sl3W+KY2T6STpA0\nVZLc/aMEnpwTnYtJz8Mk52CB5F9j5OIuIBfbXGLzkVxsPXKx7ZGPrZfrfCymjlMfSasaba9WgpKs\nOTMrk3S0pCfz25LArZKulfRxvhuyHQdLWivpV9El4nvMrGu+G9VMweRiQvMwyTlYCPnXGLm4a8jF\ntlUQ+Ugu7jRyMYfIx52W03wspo5TwTCzvSX9VtK33H1jvtuTYWZnSlrj7n/Ld1t2oLOkYyTd6e5H\nS3pfUqLGJheKJOZhAeQg+ZcD5GKrkIs5QC62CrmYI+Rjq+Q0H4up41Qn6cBG232jWKKY2e5K/yf4\njbv/Lt/taeYLkkaaWa3Sl61PNrNf57dJgdWSVrt75puXh5T+D5Ikic/FBOdh0nOwEPKvMXKx9cjF\ntpfofCQXW41czAHysdVymo/F1HF6WlJ/Mzs4mgh2oaTZeW5TE2ZmSo+5/Ie735zv9jTn7hPdva+7\nlyn993vU3S/Jc7OacPc3Ja0ys8Oi0HBJiZksGUl0LiY5D5OegwWSf42Ri61ELuZEYvORXGw9crHt\nkY+tl+t87NxWD5Rv7r7FzL4u6RFJnST90t1fzHOzmvuCpK9Iet7MUlHsu9Hq12i5b0j6TXSyWynp\n3/PcniYKIBfJw12T6PxrjFwsegWTi1Li85Fc3DXkYtsiH3dNzvKxaMqRAwAAAECuFNNQPQAAAADI\nCTpOAAAAAJAFHScAAAAAyIKOEwAAAABkQccJAAAAALKg4wQAAAAAWdBxAgAAAIAs6DgBAAAAQBZ0\nnAAAAAAgCzpOAAAAAJAFHScAAAAAyIKOEwAAAABkQccJAAAAALKg41QgzOzrZrbUzDaZ2bQs+15t\nZm+a2UYz+6WZdWmnZqKDMLP9zGymmb1vZq+Z2b/tYF/yETnDuRFJY2b9zexDM/v1DvYhF5FTvE/n\nBh2nwvGGpP+S9Msd7WRmp0qqkjRc0kGSDpH0o5y3Dh3N7ZI+klQi6WJJd5rZwOY7kY9oB5wbkTS3\nS3p6e3eSi2gnvE/ngLl7vtuAnWBm/yWpr7tftp3775NU6+7fjbZPlnSfu/duv1aimJlZV0nrJQ1y\n95ej2L2S3nD3qmb7ko9oF5wbkQRmdqGkcyUtk3Sou18Ssw+5iJzifTp3uOJUfAZKeq7R9nOSSsys\nR57ag+LzWUlbMifjyHNK515z5COSglxETplZd0k/ljQhy67kInKN9+kcoeNUfPaW9E6j7Y3R7255\naAuK0976JK8yNio+x8hHJAW5iFy7XtJUd1+dZT9yEbnG+3SOdM53A9Dm3pPUvdH2PtHvd/PQFhSn\n5jkmpfMsLsfIRyQFuYicMbNySadIOroFu5OLyDXep3OEK07F50VJRzXaPkpSg7u/naf2oPi8LKmz\nmfVvFDtK6dxrjnxEUpCLyKUTJZVJet3M3pT0n5JGm9kzMfuSi8g13qdzhI5TgTCzzma2h6ROkjqZ\n2R5mFnfF8F5JlWZ2hJntK+kHkqa1Y1NR5Nz9fUm/k/RjM+tqZsMkjZQ0PWZ38hE5xbkRCTFF0mck\nlUc/v5D0R0mnxuxLLiKneJ/OHTpOheP7kj5QumTkJdHt75tZPzN7z8z6SZK7z5P0E0mPSXpN0quS\nfpifJqOIXSFpT0lrJN0n6XJ3f5F8RB5wbkTeufu/3P3NzI/Sw58+dPe15CLyhPfpHKAcOQAAAABk\nwRUnAAAAAMiCjhMAAAAAZEHHCQAAAACyoOMEAAAAAFkkpuNkZqeZ2UtmtsLMqvLdHnRs5COSglxE\nUpCLSBLyEfmQiKp6ZtZJ6cW6RkhaLelpSRe5+7IdHJP/hiPR3N1ac9zO5iO5iGzaKxejY8hH7FBr\n8pFcRC7wPo2kaGkuJuWK01BJK9x9pbt/JOkBSaPy3CZ0XOQjkoJcRFKQi0gS8hF5kZSOUx9Jqxpt\nr45iTZjZeDNbamZL261l6Iiy5iO5iHbCuRFJQS4iSXifRl50zncDdoa7T5E0ReKyK/KLXESSkI9I\nCnIRSUEuIheScsWpTtKBjbb7RjEgH8hHJAW5iKQgF5Ek5CPyIikdp6cl9Tezg83sU5IulDQ7z21C\nx0U+IinIRSQFuYgkIR+RF4kYqufuW8zs65IekdRJ0i/d/cU8NwsdFPmIpCAXkRTkIpKEfES+JKIc\neWswXhXZtLbM6c4iF5FNe+WiRD4iO86NSApyEUlRaOXIAQAAACCx6DgBAAAAQBZ0nAAAAAAgCzpO\nAAAAAJAFHScAAAAAyIKOEwAAAABkkYh1nFAYhg0bFsT+/d//PYj16dMniD366KNB7IEHHmiy/frr\nr+9C69BSu+++exDr2rVrEDv88MOD2FFHHRXEPvOZzzTZNgsreq5YsSKIPffcc0Hs1VdfDWINDQ1B\nDB1PXF7F5ePnPve5rI/117/+NYj9/e9/D2KFulwHgMK2xx57BLGBAwcGsaFDhwaxuHPg0Ucf3WS7\npKQk2CfuvXbRokVBbMaMGUHsmWeeCWL/+te/glgx4IoTAAAAAGRBxwkAAAAAsqDjBAAAAABZ0HEC\nAAAAgCysUCe/mllhNrxAnHDCCUHs1ltvDWLNJxxK0gsvvBDENm3aFMSqqqqabC9YsGBnmpiVu4ez\nyXMgKbnYuXNY66Vv375B7Lzzzgti55xzThA77LDDWvS8W7ZsabL94YcfBvts3rw5iK1fvz6IpVKp\nIFZTUxPECm0ianvlopScfGxrcfn9k5/8JIiNHDmyyXZcUYmHH344iF177bVBbOvWrTvTxIJRLOfG\ns846K4h9+ctfDmIfffRRELvvvvuC2J///Ocg9vHHH7eyda3XqVOnIBZXdGmvvfYKYu+++24Qq6ur\na5uG5UCx5GJL7bZbeL3ikEMOCWIXX3xxELvkkkuCWM+ePVv0vM3PZc3ft6X4c+y+++4bxFavXh3E\n4t6np0+fHsRefvnlHbYzn1qai1xxAgAAAIAs6DgBAAAAQBZ0nAAAAAAgi0QtgGtmtZLelbRV0hZ3\nr8hvizqGuIVOv/e97wWxAQMGBLG4BXD/3//7f0Hs5JNPDmJxY32TIum5GLeIbdy8tK997WtB7Mwz\nzwxi77zzThBbunRpi2LNxyzH/bvGzZc65phjgtj5558fxOIW+LvjjjuC2MyZM4PYunXrgljc2O4k\nS3outpe4eR+nn356EGuef3H5eMYZZwSxiRMnBrFinePUWknLxbj3rriFQ7t16xbExo0bF8SeeOKJ\nIBY3ZzPX4uYujRo1KojFzVGPm881ZcqUtmlYwiQtH5uLO2d96UtfCmITJkwIYieddFIQW758eRC7\n8847g1hcHjdfcP61114L9unXr18Qaz5nVJL+4z/+I4j953/+ZxCLW2T3xhtvDGKvvPJKEEuyRHWc\nIie5+1v5bgQgchHJQS4iKchFJAn5iHaV3K/8AQAAACAhktZxckkLzOxvZja++Z1mNt7MlppZOF4I\naFvkIpJih7kokY9oN+QikoT3abS7pA3VG+budWbWS9J8M/unu/8lc6e7T5E0RUpOTX4ULXIRSbHD\nXJTIR7QbchFJwvs02l2iOk7uXhf9XmNmMyUNlfSXHR+FnXHAAQcEsSuvvDKIffGLXwxiN910UxCL\nm/wYN8F/7ty5QSzhC6ElOhfjJl3GTdiMm2D6wAMPBLFf/OIXQeypp54KYi1ZMLtr165BrEuXLkFs\n//33D2JxE1GvuOKKIDZp0qQgdtBBBwWxadOmBbEk512cpOdie4lbyLa0tDSINZ/Mv2bNmmCfz3zm\nM0GsrKwsiBVaruRa0nJx1apVQezHP/5xEDv44IOD2MqVK4NYXNGZxYsXt7J1rbfnnnsGsXPPPTeI\n/e53vwticQWb4hZ8jvt/UWiSlo/Nxb0Xjh07Nogdd9xxQSyu6MPUqVODWNyi8a31+uuvB7Gf//zn\nQWzRokVB7Prrrw9iZ599dhCLW/g+7rXW1tZup5X5l5ihembW1cy6ZW5L+pKkF/LbKnRE5CKSglxE\nUpCLSBLyEfmSpCtOJZJmRt8qdpZ0n7vPy2+T0EGRi0gKchFJQS4iSchH5EViOk7uvlLSUfluB0Au\nIinIRSQFuYgkIR+RL4kZqgcAAAAASZWYK05oe3ETTOMm31900UVB7E9/+lMQu/vuu4PYo48+GsTi\nJt3GTTCMm9iLUI8ePYLYWWedFcROOOGEILZw4cIgNnHixCD25ptvtqgtgwcPDmInn3xyk+1u3boF\n+zzzzDNB7K9//WsQmzx5chB74YVw2Pp3vvOdIPbVr341iH3qU58KYrfccksQq6urC2IoDnGT+wcN\nGhTEmuexRHGIpJs1a1YQiyt6MH58WDn93XffDWJxBRjizlMff/xxS5vYKrvvvnsQO+aYY4LYt771\nrSBWXV0dxPr27RvEiqE4RJLEFUA67bTTgtipp54axP785z8Hse9973tBLC5n44oixRVe2rRpU5Pt\ntWvXBvu8//77Qexf//pXEIt7T44rHta5c9jFiCte0rygjyT9z//8TxCLe/35wBUnAAAAAMiCjhMA\nAAAAZEHHCQAAAACyoOMEAAAAAFlQHKJI7LvvvkFs1KhRQeyaa64JYnGTBL/yla8EsdmzZwexPfbY\nI4hNmjQpiMVNsN26dWsQQ+jzn/98ELvsssuCWFyBhwceeCCIxU3EHDp0aBArKysLYieeeGIQGzZs\nWJPtuNXSr7rqqiD29NNPB7Ff//rXQWzOnDlB7Nvf/nYQi5sUPXr06BY974wZM4IYikPcKveXXHJJ\nEDv22GOD2C9+8YtcNAlt5IMPPghif/vb34JYXNGHuMI5w4cPD2Jxk/7jnrctxRWfiCv21HzCvxRf\nWKJ79+5t0zBs15FHHhnE4op3xOVO3Pte3L9tXAGbK664IojFFb9p/hlxw4YNwT5PPvlkELvvvvuC\n2F/+8pcglkqlglhcgYef/OQnQezyyy8PYnHte+ihh5ps56vAGFecAAAAACALOk4AAAAAkAUdJwAA\nAADIgo4TAAAAAGRBcYgC1a9fvybb48aNC/aJmzT49ttvB7G4ifvXX399ECsvLw9it912WxCLm8y/\nZcuWIIaW+dznPhfEDjjggCB28803B7Fnn302iH3nO98JYnEFKPr06RPEPvroo6zP8cQTTwT7xE26\njitIceWVVwax1157LYjFFRuZNm1aELvhhhuCWNxq7s1Xbo8rtIHCtGbNmiAWN/l+v/32a4/mIMee\neeaZIDZixIggdtdddwWxs846K4jFFUDKdXGIuMIA69evD2L7779/EFu3bl0QM7O2aRi2af43Pf30\n04N94gpG3HrrrUFs7ty5QSzu3zauKFRcwaa4/wNvvPFGk+1evXoF+5xxxhlBLO5zQNznucWLFwex\n5u+rUnxxiIkTJwaxb3zjG0GseUGxu+++O9gn1/83Ja44AQAAAEBWdJwAAAAAIAs6TgAAAACQRbt2\nnMzsl2a2xsxeaBTbz8zmm9ny6He4kiuQA+QjkoJcRFKQi0gS8hFJ097FIaZJ+rmkexvFqiQtdPdq\nM6uKtsPZ6x1Y3Mrlo0aNarIdt/LyihUrgtj48eOzPpYkXXLJJUFs3rx5QexXv/pVEKuvrw9iCTVN\nCczHffbZp8n2UUcdFezj7kEsbtXvr33ta0Es7t87rhjCwoULg9g///nPINZ8ImpcQYqampogFpez\ncSutX3vttUFs0qRJWdshxa9mHleU4thjj22y/fvf/z7YJ8emKYG5mCRxOR+3unzXrl2bbL/77rs5\na1ORmqYCzsVXXnkliO29995BbPny5UGsee5IUufO7V9D6/333w9izSf3S1KPHj2CWNy5vGfPnm3T\nsPyYpgTm42c/+9km2yNHjgz2ifu3mDVrVhCLO0fFFVSKK3ISV4AhrshS889lccUh4gqKjR07NojF\nFXi4/fbbg9hvf/vbIPbII48EsbiiU//1X/8VxCZMmNBke9WqVcE+Dz/8cBCLe+/YFe16xcnd/yKp\necmXUZIyn6pqJJ3dnm1Cx0U+IinIRSQFuYgkIR+RNEmY41Ti7pmu8JuSSvLZGHR45COSglxEUpCL\nSBLyEXmTqHWc3N3NbLvX1MxsvKRwrBmQAzvKR3IR7YlzI5KCXESS8D6N9paEK04NZlYqSdHvcLXC\niLtPcfcKd69ot9aho2lRPpKLaAecG5EU5CKShPdp5E0SrjjNljRGUnX0O5w514HsvvvuQSxuRerm\nk+jXrAnPG3GT6o855pggFjf5b+nSpUFs8uTJQSxucl6By3s+9u3bt8n2YYcdFuxTV1cXxOImO3/+\n858PYnGTnb///e8HsZdffjmIrV27Noi11uzZs4PYQQcdFMTiCpXETQD/7ne/G8TiJs4effTRQWzw\n4MFNtvNQHCJO3nMxST7++OMgFlcQpPn5kuIQbaJgcjHuHBWXO3H7tfUk8tbatGlTEHvvvfeC2Je+\n9KUg1qlTpyD2xS9+MYg9+OCDrWxdIuQ9Hy+44IIm20OGDAn2ufnmm4NYXIGluCIfF154YRDbY489\nglhcAYaWFOiK+8w4bdq0IPavf/0riMUVdvrxj38cxOLa+5vf/CaIPfbYY0Hs1ltvDWJ33nlnk+24\nzwZxnyu2bt0axHZFe5cjv1/SE5IOM7PVZlapdOKPMLPlkk6JtoGcIx+RFOQikoJcRJKQj0iadr3i\n5O4XbeeusO4ikGPkI5KCXERSkItIEvIRSZOEOU4AAAAAkGh0nAAAAAAgiyQUh0Aje+21VxCLmwDX\nfNXnb3/728E+n/rUp4LYxIkTg9iHH34YxO66664gtnjx4iCGttd8omjcqvdxk+LjCovErQ4e9+8d\nN2H17bff3mE7d1Xcc8ZNpj3xxBODWNyq6iUl4VIecQUu4v5OhxxySJPtuAnWbT3BFDsn7u//1FNP\nBbHm+bLPPvvkqklIoLfeeiuIbd68OYi99NJLQSwuxwYOHBjE4orO5LqwxLp1zdeAlYYNGxbE4grn\nnHDCCUEsrkjQli1bWtm6jmfkyJFNtjdu3Bjs86tf/SqIxRX5iPt3jCvo8de//jWILVy4cIft3Bm1\ntbVB7LbbbgticYWo4opDxBVs+uCDD4LYr3/96yD2v//7v0HsBz/4QZPtioqwWGLcZ+i2LhDEFScA\nAAAAyIKOEwAAAABkQccJAAAAALKg4wQAAAAAWVAcIo/iijeccsopQSxudfD/+7//a7I9ZcqUYJ/5\n8+cHsQMPPDCIVVVVBbEnnngiiKHtxU3QHTx4cJPtuNW3V65cGcTi9osTN4k1KZOC6+rqgljchNjz\nzjsviMVNsH322WeDWNxE8cMPP7zJdpcuXYJ94lZQR/vZbbfwe75DDz00iHXt2rXJdlwxFBSvVatW\nBbEXXnghiF1//fVB7MknnwxilZWVQSyVSgWxDRs2tLSJrfLiiy8GsZNPPjmI3X333UEsrphOaWlp\nEIv72yFev379mmzHnWfi8i7O0UcfHcTiihjFFYJ48803W/QcrRX3uubOnduiY+OKjJ1//vlBbObM\nmUEsrhhK88Inxx57bLDPEUccEcTi/l/vCq44AQAAAEAWdJwAAAAAIAs6TgAAAACQBXOc8uj4448P\nYt/73veCWNwY0+aLkl1zzTXBPnELqM2YMSOI3XnnnUHso48+CmJoe3FzaZovrBe3sOKCBQuCWNwi\nb3ELOvbs2bNF7ciHuFyPe62jR48OYl/4wheC2G9/+9sgFrcAX9++fZtsx809Q37FLUocN6+t+Vy0\nuLlRKF5xC3dPnTo1iP33f/93EItbUDNunkrceeXhhx9uaRNbJW6+Ztxin3HtiFtYvE+fPkGMOU75\nsf/++wexuHnHuZ7P1FItnff097//PYiVl5cHseYL0Evxc5yeeeaZJtvHHHNMsM+gQYOCGHOcAAAA\nAKYqZZUAABQLSURBVKCd0XECAAAAgCzoOAEAAABAFu3acTKzX5rZGjN7oVHsOjOrM7NU9HN6e7YJ\nHRf5iKQgF5EU5CKShHxE0rT3DOhpkn4u6d5m8Vvc/X/auS3tKm6x27POOiuIfeYznwli1dXVQWzT\npk1Ntr/+9a8H+7z00ktBbNq0aUEsKYuf5sE05TkfzSyINS/esH79+mCfl19+OYitXbs2iMVN4uzR\no0cQS3IxhOXLlwexuOIlBx10UBCLKwTx8ccfB7HmxTHi/l1ybJo66LmxpeL+TXr37h3Emuf8vvvu\nm7M2FalpKrJc/N3vfhfE4t5XL7744iAWd76IW1A218Uhnn/++SAWtyj3YYcdFsTee++9INZ8oXVJ\nevrpp4NYXIGhdjZNRZaPxSjus0ZczsYVhxgwYEAQW7p0aRDbuHFjk+24z67NF0DPhXa94uTuf5EU\nlsoA8oB8RFKQi0gKchFJQj4iaZIyx+kbZvb36JIsXw8i38hHJAW5iKQgF5Ek5CPyIgkdpzslHSKp\nXFK9pJu2t6OZjTezpWYWXsMD2kaL8pFcRDvg3IikIBeRJLxPI2/y3nFy9wZ33+ruH0u6W9LQHew7\nxd0r3D1cpQ5oAy3NR3IRuca5EUlBLiJJeJ9GPuV9RriZlbp7fbR5jqRwme4C03yiuSSdcsopQWzk\nyJFBLG6V8nvuuSeINV+5fI899gj2+cEPfhDEFi5cGMTiJr92VO2dj3FFDp577rkm26NGjQr2Of74\n44PY66+/HsRWrlwZxD772c8GsWOPPTaI/eEPfwhiuZ4ovPvuuwexuImje+65ZxCrq6sLYmVlZUFs\nr732CmJr1qxpsp2ACdFFeW5EYSr0XGxoaAhib7zxRhD7/Oc/H8ReffXVIBZXxCmueEnz9+UDDjgg\n2GfgwIEtakdcMYcNGzYEsdGjRwexuOIQX/3qV4NY3IT8Z555JojlW6HnYzEqLS0NYvvvv38QiyvY\nFPc5KM7hhx/eZDvus3bc/9e21q4dJzO7X9KJkvY3s9WSfijpRDMrl+SSaiWF/5uBHCAfkRTkIpKC\nXESSkI9ImnbtOLn7RTHhqe3ZBiCDfPz/27vXGKmqNY3jzyuCGJSjMIAEUUD5IE7ERjPeUEmM0BKj\naLxgFBghwajBBk0Y4ph4w0uig0K84CE0KFFU5OgBIgKHcJGoKCCgDcwoZhDlLiIIihjWfOjyDF3v\nanYJTdXe1f9fQuh6KLrXrnqqqldXr7WRFnQRaUEXkSb0EWlT8jVOAAAAAJB2TJwAAAAAIEHJN4co\nR507d3bZPffc47LYwrnHHnvMZQ8++KDL8hfzjx071l1n7ty5LmMjiHSJLYqcM2dOnct9+/Z116ms\nrEz8f1L8bPZXXnmly0aPHu2yk08+uaCvsX37dpcVolmzZi6LLYoePny4y2KLnWObWfTr189lp512\nmsvyb6f9+/e76wAoHx999JHLLr/8cpdVV1e7bMCAAS578sknXZb/Ol1RUVHQ2JYsWeKyadOmueyt\nt95yWdu2bV02adIkl8WeV7t27eqyNG4OUW5im5fEdOnSxWUtWrRw2d69e496TIcT22Ap9ti57LLL\nXFZTU+Oyzz//3GXdu3d3Wa9evepc3rhxo7vO6tWrXdbQeMcJAAAAABIwcQIAAACABEycAAAAACAB\nEycAAAAASMDmEEepY8eOLhs6dKjLevbs6bJx48a5LHb28SFDhrhs3rx5dS6/9NJL7jpbtmxxGdLv\n008/rXN51apV7jpXX321y6699lqXvfnmmy5r06aNy4YNG+aykSNHuqx58+Yuy1/IfODAAXedWK8v\nvvhil1VVVbnszDPPdNlzzz3nsoULF7osdgyxx0X+IuvYMaC0YouFY4uU161bV4zhIONmzpzpsptu\nusllixcvdtntt9/usuuuu85ly5cvr3M5tonT2rVrXRb7HmLgwIEui210E3t+v/fee13WsmVLl8We\npxGXvwFDbJOG2G28e/duly1dutRl27Ztc9kNN9zgsvzvBSXpk08+cVkhTjzxRJedd955Los9TmKb\nWMU2qYh9T/Lrr7+67KGHHnJZft/Hjx/vrvPtt9+6rKHxjhMAAAAAJGDiBAAAAAAJmDgBAAAAQAIm\nTgAAAACQgM0h/oTjj/c3V79+/VwWW8QZW/z38ssvu2zBggUu27Bhg8tefPHFOpeLsSAOxZF/f0+e\nPNld5+GHH3ZZbIOHH374wWUTJ0502a5du1w2YsQIlz3++OMu+/333+tc3r9/v7vOccf5n9G0atXK\nZXv27HHZmDFjXPbGG2+4rE+fPi4755xzXBZbFB47cznSpbKy0mWx5+TZs2cXYzjIuM8++8xlq1ev\ndlns+fLjjz922fbt2112/vnn17lcXV3trhPbcCd/gyDJb2AjxZ/fY8+Nse8/Qggu++mnn1yGuPwO\n3HLLLe46se8Fp0yZ4rL8TUSk+P14//33u+z55593Wew1btOmTXUuxzYWueiii1zWo0cPl7Vu3dpl\nNTU1Lps6darLZsyY4bK7777bZbHvradPn17ncmyjiVivGxrvOAEAAABAAiZOAAAAAJCAiRMAAAAA\nJCjqxMnMOprZAjNbY2Y1ZlaVy1uZ2Twz+yr396nFHBcaH7qINKGPSAu6iLSgi0ijYm8O8bukB0II\nK8zsZEnLzWyepH+XND+E8LSZjZI0StJ/FHlsiTp16uSy2267zWX79u1z2TPPPOOyu+66y2Wnnuof\n/7EF+YsWLapvmChMaruYv9nCBx984K4TW9j5wAMPuKyqqsplW7ZscVlsMWnsbN69e/d2WefOnetc\nji3Yj4l1OLZwNHb8sQWrQ4cOdVlssXNs84DYY7bIUtvHtOjWrZvLYpuJ5C+Ej21MYmYNN7Dy0yi6\nuG3bNpfFNqK58847XRbbiGb37t0uW7lyZZ3Ljz76qLvOihUrXLZu3TqX7dy502Xt2rVzWZMmTVwW\n24gnI1Lbxddee63O5QsuuMBd57777nPZ+vXrXRZ7TZo2bZrLWrZs6bL+/fu7LLaJRP73FbHX6YMH\nD7os1sXYxmax1+7YsQ4aNMhlsdsp9nXHjRuX+PmLoajvOIUQNocQVuQ+3iNpraQOkq6X9Gruaq9K\n8ttpAA2ILiJN6CPSgi4iLegi0qhk25GbWSdJFZKWSmoXQtic+6ctkvyPUWr/z1BJ/sfKwFGgi0gT\n+oi0oItIC7qItCjJ5hBmdpKk6ZKGhxDqvL8dajdhj27EHkL4awjhwhDChUUYJhoBuog0oY9IC7qI\ntKCLSJOiT5zMrKlqHwCvhxD+lou3mln73L+3l+R/+RhoYHQRaUIfkRZ0EWlBF5E2Rf1VPatdkTtR\n0toQwqGrMGdIGiTp6dzffy/muAqVvwheim8Y8eGHH7psyZIlLrvmmmtcFlu0vGvXLpfFFu6jcFnq\nYuzs8LEzZp9++ukuGzx4sMuGDBnistGjRxf0NWJZs2bNDntZii863b9/v8tievXq5bJHHnnEZeee\ne67LJkyY4LL8M76nQZb6WCqxXsXOEl/Ic2Mxzi6fVY2liwcOHHDZu+++W1CWFrGu792712WtW7d2\n2c8//+yy5s2bN8zAGkiauzh//vw6l5999ll3nSeeeMJlI0eOdNn27dtdtmzZMpc9/fTTLpszZ47L\nKioqXNa2bds6l3fs2OGus2rVKpd98cUXLot9T9qzZ0+XjRgxwmU333yzy7Zu3eqyp556ymXLly93\nWSkUe43TZZIGSPrCzP7YbuZB1Zb/bTMbImmDpFuKPC40PnQRaUIfkRZ0EWlBF5E6RZ04hRCWSKpv\nH9irijkWNG50EWlCH5EWdBFpQReRRiXZHAIAAAAAsoSJEwAAAAAkKNl5nLIo/8zL9Ylt8HDppZe6\n7KyzznLZli1bXMZGEMj3/fffu+yVV15x2dlnn+2yO+64w2UdOnRw2cyZM122aNEil33zzTd1Lv/2\n22/uOl26dHHZjTfe6LLYhindu3d32XHH+Z/5jB071mWTJ0922XfffecypN+mTZtcdsIJJ7gsf8Oe\nWFdi/49eIGtiG+xs3LjRZbHXgVjfY5tdIS5/c5H333/fXadr164uGzZsmMtimy69/vrrLps9e7bL\nVqxY4bKFCxe6LN8ZZ5zhsthr7cCBA13Wo0cPl8U2cfrLX/7istjmaS+88ILL5s6d67JCN5Q61njH\nCQAAAAASMHECAAAAgARMnAAAAAAgARMnAAAAAEjA5hB/wtdff+2yL7/80mW9e/cuKIstWn777bdd\nVlNTU+gQ0YjF+jlmzBiX7du3z2Wxs35XVla67JdffnFZ/iLZmKZNm7qsVatWLtu5c6fLVq9e7bJ3\n3nnHZdOmTXPZtm3bEseGbIhtTBI7C/348eMTP1fszPeLFy8+soEBJRLbiOfHH390Wbt27Qq6Xuw5\nGYWJbdgU2/Qg9ho3ePBgl40aNcplVVVVLot1IIRQ7zj/cPzx/tv/Jk2auCz2+n7w4EGXrVu3zmVT\np051WWyDi9iGJrGvkRa84wQAAAAACZg4AQAAAEACJk4AAAAAkMAK+V3INDKzVAw8tnapf//+Ljvl\nlFNcNn/+fJfNmjXLZRs2bDjC0TVuIQR/JuJjIC1djImto2vfvr3LKioqXNanTx+XXXHFFS6L/f58\nvq1bt7ps3rx5Lout8VuzZo3LYieFLvQE1aVQrC5K6e7j0WjTpo3Lbr31VpcNGDAg8XNNmTLFZbGT\nUO7YsaPA0WULz43lIbYmZcKECS4z83d37KS41dXVLps0adIRjq4wja2LzZo1c1m3bt1cFnv9veqq\nq1wWux9btGiROI7Y+t+VK1e6LLbGOLaeadWqVS6LrV1K85yj0C7yjhMAAAAAJGDiBAAAAAAJmDgB\nAAAAQIKiTpzMrKOZLTCzNWZWY2ZVufwRM/vezFbm/vQt5rjQ+NBFpAl9RFrQRaQFXUQaFXVzCDNr\nL6l9CGGFmZ0sabmkfpJukfRzCOHZP/G50rvCDKlwuIV+dBHFlLTolD6imHhuLA+xzX/69vVziEsu\nucRlsROxvvfeey5bv379EY6uMHQRaVHo5hD+1MHHUAhhs6TNuY/3mNlaSR2KOQZAootIF/qItKCL\nSAu6iDQq2RonM+skqULS0lw0zMxWm1m1mZ1aqnGh8aGLSBP6iLSgi0gLuoi0KMnEycxOkjRd0vAQ\nwm5JL0vqIul81f504b/q+X9DzWyZmS0r2mBR1ugi0oQ+Ii3oItKCLiJNin4CXDNrKmmWpDkhhDGR\nf+8kaVYI4V8TPg+/r4rDKmBdCV1EURTyu9P0EcXCc2N5KPc1ThJdRPGkco2T1Z6+eqKktYc+AMys\nfe53WSXpBklfFnNcaHzoItKEPiIt6GJ2HDx40GWzZs0qKMsCuog0Kvauej0lfSjpC0l/POIflHSb\nat9yDZL+V9Jdhzwo6vtc/PQAh5WwWw9dRNEU8FNV+oii4bkRaUEXkRaFvuNU9F/Vayg8CJCk0AfB\n0aKLSFKsLkr0Ecl4bkRa0EWkRaFdLNmuegAAAACQFUycAAAAACABEycAAAAASMDECQAAAAASMHEC\nAAAAgARMnAAAAAAgARMnAAAAAEhwfKkHcBR2SNqQ+/hfcpezKuvjl9J3DGcW8WvRxXRJ2zEUs4vS\n//cxbbfDkcj6MaRx/KV4bkzj7fBncQwNj9fpI5P18UvpO4aCu5jZE+AeysyWhRAuLPU4jlTWxy+V\nxzE0hKzfDlkfv1Qex9AQyuF2yPoxZH38DaUcbgeOoXxk/XbI+vilbB8Dv6oHAAAAAAmYOAEAAABA\ngnKZOP211AM4Slkfv1Qex9AQsn47ZH38UnkcQ0Moh9sh68eQ9fE3lHK4HTiG8pH12yHr45cyfAxl\nscYJAAAAAI6lcnnHCQAAAACOmUxPnMys0sz+28y+NrNRpR5PIcys2sy2mdmXh2StzGyemX2V+/vU\nUo7xcMyso5ktMLM1ZlZjZlW5PDPHcCzQxeKji/XLWh+z3kWJPtYna12Ust9HuhhHF4uvHLuY2YmT\nmTWR9KKkayR1k3SbmXUr7agKMllSZV42StL8EEJXSfNzl9Pqd0kPhBC6SbpY0r252z1Lx9Cg6GLJ\n0MWIjPZxsrLdRYk+OhntopT9PtLFPHSxZMqui5mdOEn6N0lfhxC+CSH8JulNSdeXeEyJQgiLJe3M\ni6+X9Gru41cl9SvqoP6EEMLmEMKK3Md7JK2V1EEZOoZjgC6WAF2sV+b6mPUuSvSxHpnropT9PtLF\nKLpYAuXYxSxPnDpI2njI5e9yWRa1CyFszn28RVK7Ug6mUGbWSVKFpKXK6DE0ELpYYnSxjnLpY2bv\nR/r4T+XSRSmj9yNd/Ce6WGLl0sUsT5zKUqjd5jD1Wx2a2UmSpksaHkLYfei/ZeUYcHhZuR/pYvnL\n0v1IH8tfVu5Hulj+snI/llMXszxx+l5Sx0Mun57LsmirmbWXpNzf20o8nsMys6aqfQC8HkL4Wy7O\n1DE0MLpYInQxqlz6mLn7kT465dJFKWP3I1106GKJlFsXszxx+kxSVzPrbGbNJPWXNKPEYzpSMyQN\nyn08SNLfSziWwzIzkzRR0toQwphD/ikzx3AM0MUSoIv1Kpc+Zup+pI9R5dJFKUP3I12MooslUJZd\nDCFk9o+kvpL+R9J6Sf9Z6vEUOOapkjZLOqDa37EdIqm1ancV+UrSPyS1KvU4DzP+nqp9S3W1pJW5\nP32zdAzH6Hahi8UfP12s/7bJVB+z3sXcMdDH+O2SqS7mxpzpPtLFem8Xulj88ZddFy13YAAAAACA\nemT5V/UAAAAAoCiYOAEAAABAAiZOAAAAAJCAiRMAAAAAJGDiBAAAAAAJmDgBAAAAQAImTgAAAACQ\ngIkTAAAAACT4P9SDCLYKOZ2gAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1372c4a20>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize=(20, 10))\n",
    "for i, img in enumerate(train_g[:5], 1):\n",
    "    subplot = fig.add_subplot(1, 7, i)\n",
    "    plt.imshow(img, cmap='gray');\n",
    "    subplot.set_title('%s' % train_label[i - 1]);\n",
    "    subplot = fig.add_subplot(3, 7, i)\n",
    "    plt.hist(train_theta[i - 1].flatten(),\n",
    "             bins=16, weights=train_g[i - 1].flatten())"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Вычисляем гистограммы градиентов"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Гистограммы вычисляются с учетом длины вектора градиента\n",
    "train_hist = np.zeros((len(train_img), 26))\n",
    "for i in range(len(train_img)):\n",
    "    hist, borders = np.histogram(train_theta[i],\n",
    "                                 bins=26,\n",
    "                                 range=(0., 2. * np.pi),\n",
    "                                 weights=train_g[i])\n",
    "    train_hist[i] = hist"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "test_hist = np.zeros((len(test_img), 26))\n",
    "for i in range(len(test_img)):\n",
    "    hist, borders = np.histogram(test_theta[i],\n",
    "                                 bins=26,\n",
    "                                 range=(0., 2. * np.pi),\n",
    "                                 weights=test_g[i])\n",
    "    test_hist[i] = hist"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Нормируем вектор гистограммы"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# По умолчанию используется L2 норма\n",
    "train_hist = train_hist / np.linalg.norm(train_hist, axis=1)[:, None]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "test_hist = test_hist / np.linalg.norm(test_hist, axis=1)[:, None]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Разбиваем выборку на обучение и валидацию"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "y_train, y_val, x_train, x_val = train_test_split(\n",
    "    train_label, train_hist, test_size=0.2, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.ensemble import GradientBoostingClassifier\n",
    "\n",
    "model = GradientBoostingClassifier().fit(x_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "pred_val = np.zeros_like(y_val)\n",
    "for i in range(len(y_val)):\n",
    "    \n",
    "    pred_val[i] = model.predict(x_val[i])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(8400,)"
      ]
     },
     "execution_count": 64,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred_val.shape\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Оцениваем качество решение на валидационной выборке"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy: 0.64880952381\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import accuracy_score\n",
    "print('Accuracy: %s' % accuracy_score(y_val, pred_val))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "             precision    recall  f1-score   support\n",
      "\n",
      "        0.0       0.45      0.45      0.45       816\n",
      "        1.0       0.98      0.96      0.97       909\n",
      "        2.0       0.49      0.44      0.46       846\n",
      "        3.0       0.64      0.68      0.66       937\n",
      "        4.0       0.73      0.69      0.71       839\n",
      "        5.0       0.62      0.64      0.63       702\n",
      "        6.0       0.50      0.49      0.50       785\n",
      "        7.0       0.85      0.81      0.83       893\n",
      "        8.0       0.52      0.57      0.55       835\n",
      "        9.0       0.67      0.70      0.69       838\n",
      "\n",
      "avg / total       0.65      0.65      0.65      8400\n",
      "\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import classification_report\n",
    "print(classification_report(y_val, pred_val))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[364   4  64  33  32  26 107   8 146  32]\n",
      " [  0 874   3   0  20   0   1   6   1   4]\n",
      " [ 62   0 370 128   7  99  47  30  83  20]\n",
      " [ 19   0  98 640   1  85   8   4  71  11]\n",
      " [ 16   9  11   2 580   2  90  36   6  87]\n",
      " [  6   0  84 107   0 448   8   8  38   3]\n",
      " [126   1  50  15  58  10 385   7  69  64]\n",
      " [  8   3  23   7  40  12  27 720   6  47]\n",
      " [145   0  37  58  11  39  40   5 479  21]\n",
      " [ 56   1  19  17  46   5  56  28  20 590]]\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import confusion_matrix\n",
    "print(confusion_matrix(y_val, pred_val))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Предсказания на тестовыйх данных"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [],
   "source": [
    "pred_test = np.zeros(len(test_img), np.uint8)\n",
    "for i in range(len(test_img)):\n",
    "    \n",
    "    pred_test[i] = model.predict(test_hist[i])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [],
   "source": [
    "#pred_test = np.zeros(len(test_img), np.uint8)\n",
    "#for i in range(len(test_img)):\n",
    "#    distances = np.linalg.norm(centroids - test_hist[i], axis=1)\n",
    "#    pred_test[i] = np.argmin(distances)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Визуализируем предсказания"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAzsAAAC3CAYAAAA1tUARAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAAgEL6f2jTO9kfMmbCAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1de9d0d30>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize=(20, 10))\n",
    "for i, img in enumerate(test_img[0:5], 1):\n",
    "    subplot = fig.add_subplot(1, 7, i)\n",
    "    plt.imshow(img, cmap='gray');\n",
    "    subplot.set_title('%s' % pred_test[i - 1]);"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Готовим файл для отправки"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "with open('submit_cv.txt', 'w') as dst:\n",
    "    dst.write('ImageId,Label\\n')\n",
    "    for i, p in enumerate(pred_test, 1):\n",
    "        dst.write('%s,%s\\n' % (i, p))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Your submission scored 0.57"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.6.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}