hello world in ML.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.datasets import fetch_openml"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "mnist = fetch_openml('mnist_784', cache = False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "def sort_by_target(mnist):\n",
    "    reorder_train = np.array(sorted([(target, i) for i, target in enumerate(mnist.target[:60000])]))[:, 1]\n",
    "    reorder_test = np.array(sorted([(target, i) for i, target in enumerate(mnist.target[60000:])]))[:, 1]\n",
    "    mnist.data[:60000] = mnist.data[reorder_train]\n",
    "    mnist.target[:60000] = mnist.target[reorder_train]\n",
    "    mnist.data[60000:] = mnist.data[reorder_test + 60000]\n",
    "    mnist.target[60000:] = mnist.target[reorder_test + 60000]\n",
    "mnist.target = mnist.target.astype(np.int8) # fetch_openml() returns targets as strings\n",
    "sort_by_target(mnist)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array([[0., 0., 0., ..., 0., 0., 0.],\n",
       "        [0., 0., 0., ..., 0., 0., 0.],\n",
       "        [0., 0., 0., ..., 0., 0., 0.],\n",
       "        ...,\n",
       "        [0., 0., 0., ..., 0., 0., 0.],\n",
       "        [0., 0., 0., ..., 0., 0., 0.],\n",
       "        [0., 0., 0., ..., 0., 0., 0.]]),\n",
       " array([0, 0, 0, ..., 9, 9, 9], dtype=int8))"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mnist[\"data\"], mnist[\"target\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(70000, 784)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X, y = mnist[\"data\"], mnist[\"target\"]\n",
    "X.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(70000,)"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "import matplotlib as mpl\n",
    "import matplotlib.pyplot as plt\n",
    "some_digit = X[36000]\n",
    "ome_digit_image = some_digit.reshape(28, 28)\n",
    "plt.imshow(ome_digit_image, cmap = mpl.cm.binary,\n",
    "           interpolation=\"nearest\")\n",
    "plt.axis('off')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "5"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y[36000]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_digit(data):\n",
    "    reshape_data = data.reshape(28,28)\n",
    "    plt.imshow(reshape_data, mpl.cm.binary, interpolation = 'nearest')\n",
    "    plt.axis('off')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train, X_test, y_train, y_test = X[:60000], X[60000:], y[:60000], y[60000:]\n",
    "# shuffle the train dataset\n",
    "shuffle_index = np.random.permutation(60000)\n",
    "X_train, y_train = X_train[shuffle_index], y_train[shuffle_index]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Binary Classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train_5 = (y_train == 5)\n",
    "y_test_5 = (y_test == 5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "SGDClassifier(alpha=0.0001, average=False, class_weight=None,\n",
       "       early_stopping=False, epsilon=0.1, eta0=0.0, fit_intercept=True,\n",
       "       l1_ratio=0.15, learning_rate='optimal', loss='hinge', max_iter=5,\n",
       "       n_iter=None, n_iter_no_change=5, n_jobs=None, penalty='l2',\n",
       "       power_t=0.5, random_state=123, shuffle=True, tol=-inf,\n",
       "       validation_fraction=0.1, verbose=0, warm_start=False)"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.linear_model import SGDClassifier\n",
    "\n",
    "sgd_clf = SGDClassifier(max_iter = 5, tol=-np.infty, random_state=123)\n",
    "sgd_clf.fit(X_train, y_train_5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ True])"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sgd_clf.predict([some_digit])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.95725, 0.9634 , 0.9629 ])"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.model_selection import cross_val_score\n",
    "cross_val_score(sgd_clf, X_train, y_train_5, cv = 3, scoring = 'accuracy')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.base import BaseEstimator\n",
    "\n",
    "class Never5Classifer(BaseEstimator):\n",
    "    def fit(self, X, y = None):\n",
    "        pass\n",
    "    def predict(self, X):\n",
    "        return np.zeros((len(X),1), dtype = bool)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.9083 , 0.90995, 0.9107 ])"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nvr_5_clf = Never5Classifer()\n",
    "cross_val_score(nvr_5_clf, X_train, y_train_5, cv = 3, scoring = 'accuracy') "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Confusion Matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import cross_val_predict\n",
    "y_train_pred = cross_val_predict(sgd_clf, X_train, y_train_5, cv = 3)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[54136,   443],\n",
       "       [ 1886,  3535]], dtype=int64)"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.metrics import confusion_matrix\n",
    "confusion_matrix(y_train_5, y_train_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[54579,     0],\n",
       "       [    0,  5421]], dtype=int64)"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "confusion_matrix(y_train_5, y_train_5) ## misclassify would be 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.8886375062845651\n",
      "0.6520937096476664\n",
      "0.8296370967741935\n",
      "0.7590850396605793\n"
     ]
    }
   ],
   "source": [
    "from sklearn.metrics import precision_score, recall_score\n",
    "print(precision_score(y_train_5, y_train_pred)) \n",
    "print(recall_score(y_train_5, y_train_pred))\n",
    "print(4115/(845+4115))\n",
    "print(4115/(1306+4115))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.7522076816682626"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## f1 score, harmonic mean of precision and recall\n",
    "# whereas the regular mean treats all values equally, the harmonic mean gives much more weight to low values\n",
    "# so only get a high F1 score if both recall and precision are high\n",
    "from sklearn.metrics import f1_score\n",
    "f1_score(y_train_5, y_train_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([  0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,  86., 131., 225., 225., 225.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,  13.,  73., 197.,\n",
       "       253., 252., 252., 252., 252.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   4.,\n",
       "        29.,  29., 154., 187., 252., 252., 253., 252., 252., 233., 145.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,  29., 252., 253., 252., 252., 252.,\n",
       "       252., 253., 204., 112.,  37.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "       169., 253., 255., 253., 228., 126.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,  98., 243., 252., 253., 252., 246.,\n",
       "       130.,  38.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,  98.,\n",
       "       240., 252., 252., 253., 252., 252., 252., 221.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0., 225., 252., 252., 236., 225., 223.,\n",
       "       230., 252., 252.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "       146., 252., 157.,  50.,   0.,   0.,  25., 205., 252.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,  26., 207., 253.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,  29.,  19.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,  73., 205., 252.,  79.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0., 120., 215., 209., 175.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,  19., 209., 252., 220.,  79.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0., 174., 252., 252., 239.,\n",
       "       140.,   0.,   0.,   0.,   0.,   0.,  29., 104., 252., 249., 177.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0., 174., 252., 252., 223.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0., 174., 252., 252., 223.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0., 141., 241., 253.,\n",
       "       146.,   0.,   0.,   0.,   0., 169., 253., 255., 253., 253.,  84.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0., 178., 252., 154.,  85.,  85., 210., 225.,\n",
       "       243., 252., 215., 121.,  27.,   9.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,  66.,\n",
       "       208., 220., 252., 253., 252., 252., 214., 195.,  31.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,  19.,  37.,  84., 146., 223.,\n",
       "       114.,  28.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,   0.,\n",
       "         0.,   0.,   0.])"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "## increasing precision reduces recall, and vice versa. This is called the precision/recall tradeoff\n",
    "some_digit"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([226129.12172593])"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# threshold\n",
    "y_scores = sgd_clf.decision_function([some_digit])\n",
    "y_scores"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ True])"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "threshold = 0  # the SGDCLASSIFIER uses a threshold = 0\n",
    "y_some_digit_pred = (y_scores > threshold)\n",
    "y_some_digit_pred"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ True])"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# what if we rasie the threshold\n",
    "threshold  = 200000\n",
    "y_some_digit_pred = (y_scores > threshold)\n",
    "y_some_digit_pred\n",
    "### raising the threshold decreases recall\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_scores = cross_val_predict(sgd_clf, X_train, y_train_5, cv = 3, method = 'decision_function')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import precision_recall_curve\n",
    "precisions, recalls, thresholds = precision_recall_curve(y_train_5, y_scores)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAecAAAEPCAYAAAB4Ggy7AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJzs3Xd4FNX6wPHv2U3vkIQeOgiIFAlFQEGkowiC9B8iKE3wquC1AApYwAsXGyqiFGkKoiBNQAVBuLRQlQ6hJCAQSEgCaZtkfn8cUsAEEthkNsn7eZ59dnZ2dvbdlH3nnDnzHmUYBkIIIYRwHBazAxBCCCHEzSQ5CyGEEA5GkrMQQgjhYCQ5CyGEEA5GkrMQQgjhYCQ5CyGEEA7mjslZKTVbKXVJKfVXNs8rpdQnSqkTSqkDSqkH7R+mEEIIUXTkpOU8F2h/m+c7ANVu3AYDX9x7WEIIIUTRdcfkbBjGZiDyNps8CcwztO2An1KqtL0CFEIIIYoaJzvsoywQlulx+I11f9+6oVJqMLp1jaenZ4MaNWrY4e2FEEIIx3D0KFy7ppcrV4ZixfTyhQtw7tzuy4ZhBOZkP/ZIziqLdVnWBDUMYyYwE8A5yNk41/scFmXBqqz63mJNf+zj6oOPqw8BHgF4OnviZHFKv3k4e+Dh7IGL1QVXJ1eCfILwdfPFqqxYLdab7l2dXPF28cbVyRVPZ0+8XLywWqx2+NhCCCFEhoQE8PMDpWDWLGjVCipU0M/99Rc88IA6k9N92SM5hwNBmR6XA87f6UXJKclcuHbBDm+fewqFi9UFH1cfvFy88HH1wdPFUyd7qyu+br54OHvg7eKNp7MnzlZnXKwueLt442J1uelAwcvFK/3m6+aLm5Mbfm5+eDh74Gp1Ramsjl2EEEIUNtu3Q2Ii1K0Lzz5783O1a+duX/ZIziuAEUqp74DGQLRhGP/o0r5VnVJ1WPvKWlKMFFKNVFJSb9wbKSSnJhOdEM3luMtcib9CSqpel5yajC3VRmxiLEkpSSSlJHEt6RpnY84SZ4sjJTWFFCPlpvvrtuvE2+JJTEnkasJVEpMTMTBITEkkIi6CiLgIO/wIsmZRFoq7F8fX1RcfV5/01ru3q07yHs4eFHMrhp+bHyU9S1LauzTlfctT0rMkrk6uuFhdcHNyw8lij1+TEEKIvPT77/q+Zct739cdv/WVUt8CLYEApVQ48DbgDGAYxgxgDdAROAHEAc9mvaebOVucKe1tzrixlNQUElMSiU6I5rrtOlcTrpKQnEBSShJxtjhiE2OJT44nKj4q/SAgITmB2KTYjAMFIxlbio2YxBjibHFcTbiafiAQlRCVvr/LcZe5HHf5rmNVKHxcfSjlVYqSXiXxd/fHx9UHf3d/irkXo7RXaXxcffBz88PLxYuagTXxdfWVFrsQRcjx41C1KkRFQb9+EBsLW7aAs7NusVWsCJUqwZNPwiOP6NdMmgQlS8J990HNmlC8ePb7P3kSoqNhxAi4fh3694dRo/Rzf/wB3bvr7tyaNW++1aunYygq8jU5G4bR+w7PG8AL9x5K/rFarHhY9HnrvJSYnEh0YjRX4q4QZ4tLPyBISE4gMSWRmMQYYhJjuBJ3hQvXL3Dh2gXOXD1DZHwkSSlJJKYkEmeLIzoxmujEaI5eOZqzz6esBPkGUdqrNCW9SuLj6kNF34rUCqxF7RK18XXzpaRnSZytRei/Roh7sGIFTJsGwcHQvLlOckrB3r3QuzfUrw9lyujzi5UqQbVqejCQi4t+/YEDOoEGBoJhgM2mk1aVKlC2rN4XQHg4XL6sX+vjo7cLC4Ny5fS+UlJ016mfn97Xv/4F332n9zF5Mjz6KHTsCC+9pPdns+kY9+7VjytU0Mk5KQnGjoXU1IzPGBAAjRtDo0bQqxdUrw6ffgrr1sHq1Tf/PCIydTh+9hlcuqRvx47BTz/p9U5OsH69jgn0e1nsUPbqwgU4eFAnfn//nL3GMODll/XBS/360Lq1Piix94HDU0+Bpyc8/PC970v6S/OQq5MrJZxKUMKzxF3vw5Zi42rCVcJiwrgSd0Un6oRoIuMjORd7jpjEGCLjI7mWdI2IuAhORJ4gzhbH6aunOX319G33ndb6LuVVKn3wXZBPEDUDalLGuwwPln4QXzffu45diIIkNVV/efv7w9WrGV+0q1ZlbLNpE0RG6uQMsGyZHp17NIvj5s6dMxJVx45w7lzW73vmDJQvr+8rVsxYHxCgEzXAmDHw7rs6ES9aBGvW3LyPc+cgNFS3mrt318k8LAyaNNGf69QpfUtr0dls8NZbOpkePQpHjuj3Wr1a3+rV0zHt339zYnZxgeHDYdCgjHUtW8LixTBxok7ohw/r22OPZSRmm0237P39oWFDaNMGGjTQBwvZJWzDyDhoyax7d9i6VS/36qUPmkrfoRP26FH4+GO9PG+evn/tNX1AA3D+PJw9C7Vq6YOiu/Xii/pmD5KcHZyz1ZlAz0ACPXM0+h6AeFs8Z6PPcun6JSLiIoiMj+RE5An2/L2Hs9FniUqISj+ffyX+CiciT2S7ryCfIGoG1qSkZ0kCPQIJ8AigvG95Aj0D8XLxIsAjAH93f/w9cngIK4QD2bsXVq6EHTt0wps8Wbc6IyL0SNtx427efvBg6NQpI2nUqaO3q1cPSpSA06d1kjx+XHfrpmnRQifVNKVK6Utsrl3TiRR0F3MaN7eMxAwZyadPH52AQcdgGHrw0Xvv6bjS9t2ly81xN25882NPT3j77YzHhqEPDrZsgZ074cEHdQxvvql7BKKjdSu9cuV//gyfe04nST+/fz6XZscOnfzOntU/85kz9XoPD/1zW7ECHnggY/uQEP1zrVJFHww8/3zGc6GhGcvr1sHFi/rnYxj6Mz3/PARlHqKMjj9Nt26wZ49OxGmWLdNd9qAPInr21J83MOdfu3andK90/gsODjZCQkJuu010dDSXL18mKSkpn6IqOgzDSB+Al2qkpt+SUpJISU3BlmrDlmLL8f6cLE64WF1wsbrgbHXG2eKMRVmwKMtdnf92cXEhICAAX19puQt9DhXA21u3Eg8fhr//1t2SFSrAQw/p59etgy++0NtXq5bRfbx7t/7ynz1bb/fWW/DOO1m/14wZMGQIbN6sW8p//627QgcMyF03aHYtv+wkJOjLbYKDdWv3yBHdcn3sMZ3A0iQn6y7j69f1/j3y9uyc3URG6t/b6tX6AOCvv3RitVr1zzgtEU6apHscduzQj5s31+e1QXfrO91oUm7dqg8y6tbVj69e1Qc8pUvD8uW6hZ728//lF2jbVif83377Z2xz58JHH+n4MqebatXgq6/0wdWdzJmjD16aNs3+70QptdswjOA7782Bk3NCQgJnz56lXLlyuLu7ywAnExiGwXXb9YyR8ik2klOTSUpJIjk1OX1EfFJqEtn9HSmlcLO6ZVyyduPStNv9Pg3DID4+nvDwcMqXL4+bm1tefUThgOLi9Bf3nj0wdKheN348TJiQ9fbVq2d0K3/1lW7dZic5WSeDyMibz1d26aJbdPv23b4FKOwrLEz/7lq3zlgXFKTPvQN4eenu6IED9eMlS3SrFiA+Xrfu08TE6K7yEzc6AuvXhx499MGbzQajR+tu9sWLs48nORmWLtXvuWuXPhgYN0532d9O2vXNSUm6xyO7wXW5Sc4O260dERFBYGAgHgXlsLAQUkrh5eJ1x+0MwyDOFkd8cjxxtjg94C05Mf2yuPjkeOKT42+6bM3F6oKns6e+HtzJFXcnd9yc3FBKoZTCw8ODgIAAIiIiCLq1j0oUSCkpeiDP0aO6pZvWnfvCC/D11xnnHhMS9H1AgO4ydXLSXZtLl+rXg25J1qunW05ly2a8x8CBusVZs6b+Qo6O1u9Xq5buQrbeqD/k5KTPe772mh68kzZwS+SvoKCbu6ANQ3cnjx+vewZGjcpIzKB/14MH69/9rcfsPj66h+Stt/QphMwD4Vq10gPW7sTJSXfR9+qlE22/fvDGGzfHl1W7YscOfX1znTq3H/WeGw7bcj5+/DgVK1bEuSiNwy+EklOTibfFE50YzbWka8Tb4kkxUrLc1qIsuDu5p1d+87R6EnE+gurVqudz1MKezp3Tg3i2b89YlzYICnQiPXIk69fGxurWE+iu3tOndYJNq7okCqfERH1K4eGHwd0996+/fh1++EGPkt+4UQ/SeuaZe4spLEyfYnj/ff33nNmECfqA4sUXMwaeZaVQtJyTk5NxcnLY8EQOOVmc8Hb1xtvVO31dqpFKQnIC15KukZicqFvWtnhsqTau265z3XZdb2jAlStXmHxoMp3v60zn+zpjUTIFeUFx6+jjzDJ3iE2apL/4PD31NblVquhzh7eO4rVYsh6QJAofV1d9jvhueXrqa7HT2KMNOnOmHuj39NN6dHfmEeL2vL45jcO2nA8fPkzNzMMdRaFnS7GRkJyALdXGtaRrRMVH8fepv+mwvgMA1f2r81SNp+hftz81A+Vvw9EYhh5sU726bhVHRNw8kKlHD11v2OvOZ0qEcDgpKbrH5tw53RV//Lg+iEhI0AeTCQn6fPPtrr3OTctZmiHCYThbnfF29aa4e3HK+5anTsk6lPYuzcSWEynuXpxjV44xeetkan1ei8ApgTz6zaN8vP1jIuNvN6OpyCuxsboARffuumvaYtHncdeu1c8HBOjzyu+9p7ukFy+WxCwKLqtVjzIH3dNTqZK+/nzpUp2YH3gg50VRckKSs3BYSukJSsa1GEf4y+H82ONHet7fEx9XHy7HXeb307/z0rqXKDetHCPXjOR/Yf8jJTXr89ni3mXuZFuzRg/AGTFCn9vLfM44JkbfK6W/xN58M3eXFAnhqMqU0Qeb7u768q/339fFXcqVs2+XNkhyzldz585NH42slMLb25u6desyffp0kpOT8y2O8ePH5/rStJYtW9LS3n99ueDu7E7Xml35rvt3RL0WxZmXzjC782weqfAI8cnxTN81nWazm1H/y/rMCJnBtaRrpsVa2ERG6pKPI0dmrDt1St+7uelBO+PH60RsGPqSFSEKqzffhCtXdA/RmDG6DOrZszBlin3fR84556O5c+fy7LPP8v3331OuXDliYmL4/vvv+frrrxk3bhwT73QxnZ2Eh4cTHh5OkyZNcvyaQ4cOAVArc1mdfJCTv4Md4TuYt38e8w/MJzZJV6uwKAvtq7bn/VbvU7dU3fwItVAKD9dFIM7cmIX2xAk9YOvkSX2pSSH7FxUiTxWKIiSFOTkfP36cqlWrpq9/9NFH2b17NzFp/YGZGIaBzWbDpYheiJmbv4NrSdeYvXc28w/MJ+S8/tuyKAvP1H2GcY+Mo1KxSnkZaqEyZ47uul66NGPdhg0ZtZKFELknA8IKmIYNGxIbG8ulS5eoWLEi/fr1Y/bs2dSoUQMXFxdW36g8HxcXx2uvvUalSpVwcXGhUqVKvPfee6RmnloGXcBl+PDhBAUF4erqSlBQEP/3f/9HYmIikHW39scff0zNmjVxd3enWLFiBAcHs2zZsvTns+rWPnr0KF27dsXPzw93d3eaNGnC2rTRQDekvdfx48fp1KkTXl5eVKhQgYkTJ/4j7nvl5eLFi41fZNfzuzj70ll63N+DVCOVOfvmUOWTKrRf0P62dcSFHpE6caIu/JCWmFu21CNUJTELkX8KXHJWKvtbWjF10Mu32zazBg2y3y5zKcDdu/PmM506dQqr1YrXjaGsGzduZNq0abz99tusXbuWOnXqkJycTLt27fj666/517/+xc8//8xzzz3HO++8w6uvvpq+r6ioKJo2bcrixYt55ZVXWLNmDf/5z3+w2WzZ1ihfuHAho0aNonfv3qxZs4aFCxfSvXt3IiOzHwV9/vx5mjdvzv79+5k+fTpLlizBz8+PTp068fPPP/9j+65du9KqVSuWL19Oly5dePvtt/nmm2/u8SeXvSDfIBZ3X0zI8yF0rdEVA4N1J9cRPDOYL3Z9QXJq/p3jd3TXr+uyl4ahR6SmdeAEB8Off+oiDmXKmBujEEWOYRim3Bo0aGDczqFDh7Jcr79Csr59+WXGdl9+efttM3vwwey3e/75jO1CQm4b8h3NmTPHAIwjR44YNpvNiIyMNGbMmGFYLBbjySefNAzDMCpUqGC4u7sbf//9902vnTdvngEYmzZtumn9u+++azg7OxsXL140DMMwxo0bZ1gsFmPPnj3ZxvH2228bZPohvPDCC0b9+vVvG3uLFi2MFi1apD8eNWqUYbVajePHj6evS05ONqpXr37TvtLea/bs2Tftr3bt2kabNm1u+56Gkf3fQW6FRYcZj8x5xGA8BuMx/Cb7Ga3ntTZm7Jph2FJsdnmPgmjfvoy/9dRUve78ecOIjjY3LiEKIyDEyGGOLHAt59ul3Myt3MGDb79tZrt3Z79d5tZ4gwb2+Qw1atTA2dmZ4sWLM3z4cPr27cvstOlygCZNmlCqVKmbXrN27VoqVKhA06ZNSU5OTr+1bdsWm83G9hu1EdevX0/Dhg2pX79+juNp2LAh+/btY+TIkfz666/ExcXd8TWbN2+mSZMmN507t1qt9O7dm3379v3j/HmntPnsbqhduzZnz57NcYz3qpxPOX5/5ncWdF1AteLVuJpwlV9Df2Xo6qE8POdh9l3Yl2+xOIKdOzNqFKe5eFHfly59b3PaCiHundTHNMGyZcsoV64c3t7eVKhQ4R+zLpXOYubwS5cucebMmWxrjV+5ciX9vm7d3I1O7t+/PwkJCcyaNYvPP/8cZ2dnOnbsyLRp06iYTf3FyMjILA8ASpUqhWEYREVF4ZPpG774LdXgXV1dSUib4SCfKKXoW6cvfR7oQ2hUKOtOrmPshrFsD99O/S/r06JCC4YFD+Pp+58utGVCo6L0rD6//HLz+rVr9TzAQgjHIMnZBLVr176pxXmrrK5B9vf3p1KlSixZsiTL16Ql0YCAAM6dO5ereJRSDBkyhCFDhhAVFcX69esZNWoUPXv2ZEfapKq3KF68OBcuXPjH+gsXLqCU+kcydiRKKaoUr8Lw4sN58r4nGf3LaBb/tZhNZzax6cwmJm2ZxFst3qJLjS6FIkknJGTM4JOUlJGYPTx0AZG2bf9Zx1oIYS75lywg2rdvT1hYGF5eXgQHB//jFhAQAEDbtm3ZuXMn+/fvv6v3KVasGD179qRHjx789ddf2W7XokULtm/fzunTp9PXpaSksHjxYurXr4+3t3e2r3UkZX3K8m23b7kw+gJvt3ib4u7F2X9xP92WdOPRbx5lycEl2c5V7eiSkvSUe15eGdPlBQbqKRr37NEDwdq3l8QshCOSf8sCom/fvjRt2pTHHnuMadOm8dtvv/Hzzz8zffp02rZtm36e+OWXX6Zy5cq0bt2ajz/+mA0bNrBkyRL69u1LbGxslvsePHgwo0aNYunSpWzevJmvv/6a+fPn0/Y208K8/PLL+Pn50aZNGxYtWsSqVat44oknOHbsGO+9916e/AzyUgnPEoxvOZ7T/zrNm83fxNPZk81nNtNzaU9az29NeEy42SHmmGHo6faaN4dp0/TlUdOm6ecsFpg+XU9EL4RwXNKtXUA4Ozuzbt06Jk+ezMyZMzl16hSenp5UqVKFTp06pRcp8fPzY+vWrYwdO5bJkydz5coVSpYsSatWrbItZNKsWTPmzJnD/PnziY6OpkyZMvTr148JEyZkG0+ZMmXYsmULr732GsOGDSMxMZF69eqxevVq2rdvnyc/g/zg7erNe4+9xysPvcLsvbOZtGUSG05toMb0Gvyr8b945aFX8PewY3V7O9u+HZ57Dg4e1I+tVj35/Nix5sYlhMgdqRAmHJrZfwdnrp6h1w+92B6uR8M7WZx4sdGLTGk7xeHOR0+apOv+Avj56SQ9ZoxeFkKYTyqECWEnFfwq8L+B/2Nt37W0qNCC5NRkpm2fRqOvGrEjPOvBcmYZPjxj+fBhXYhfErMQBZMkZyHuQClFu6rt+H3A7yzruYxibsXY/fdumsxqwvDVwzkfe9602I4fh2PH9LK7u74kKjVVLosSoqCT5CxELnSp0YUzL51h9EOjsSgLX4R8QYWPKjDh9wmkGvatFX4nW7dC9eq65vWJE+DiAu3aydzJQhQGkpyFyCVvV2+mtJ3Clme30K5KO5JTkxm/aTwj1ozAlmLL8/dPTYVPPtGjsUHXvb5Rll0IUUhIchbiLj0U9BBr+61lSfclWJWVL0K+oMPCDiQk513ls+Rk6NoV/vUv/bhRI/jtN+nGFqKwkeQsxD16+v6nWdVnFZ7Onvx26jfaL2hPxPUIu79PSgo4O8OKFfrx1Kn60impgy1E4SPJWQg7aF+1Pb/1/w1/d382ndlE1U+r8svJX+78wlxITIQqVfS55fXrdfUvOb8sROEkyVkIO2lcrjE7ntvB/YH3E5MYQ9sFbRmxZgTxtvh72u/Fi7o728MDPvhAz73cpo2dghZCOCRJzkLYUZXiVdg/dD8vN3kZheKzXZ9R+4vazNs/j6SUpFzvb+NGqFED0kqld+sG/fvbOWghhMOR5CyEnVktVqa1m8a2Qduo5FeJ0KhQnln+DI/Ne4xzMTmfMSwuDp54Aq5ehUWL8jBgIYTDkeScj+bOnYtSKv3m4uJClSpVePPNN/N9buNbVaxYkQEDBqQ/Tos186xTIncal2vMoRcOMavzLEp6lmTL2S00nd2Uy3GX7/jakyehWjU9c1RAAEycmA8BCyEchiRnE3z//fds27aN1atX065dOyZNmsSrr75qdlgiD7g5uTGw/kD2DNlDwzINORt9lqcWP8XfsX9n+5offoCqVeH8jcJjS5eCp2c+BSyEcAg5Ss5KqfZKqaNKqRNKqdezeL68UmqjUmqvUuqAUqqj/UMtPOrVq0eTJk1o06YNn3/+Oa1bt2bWrFmkpuZvhSmRf8p4l+G77t9R3L04f5z9gxZzW7Dvwr5/bBcfD9276+Vy5eDyZWjRIp+DFUKY7o7JWSllBT4DOgC1gN5KqVq3bDYWWGIYRn2gF/C5vQMtzB588EHi4+O5fDmju/PUqVP07duXwMBAXF1dqVevHsuWLfvHa/fv30/Xrl3x9/fH3d2d++67j0mTJqU/v379ejp27Ejp0qXx8PCgdu3a/Pe//yUlJSVfPpvIULlYZbYP2k51/+ocjzxOo68aserYqpu22bZNj8oG+PNP8Hfc2SmFEHkoJ/M5NwJOGIYRCqCU+g54EjiUaRsDSCuF4AvkyUwAaoJjXNRpvG3faTZPnz6Nr68v/je+icPCwmjcuDElSpTgww8/JDAwkMWLF9OtWzeWL19O586dAdi5cyctW7akatWqfPjhh5QrV47jx49z4MCB9H2Hhoby2GOPMXLkSNzc3AgJCWH8+PFEREQwefJku34OcWfV/KuxZ/Aeev/Qm5XHVvLU4qf46omveKbeMwC0aqVHaEdEyIxSQhRlOUnOZYGwTI/Dgca3bDMeWK+UGgl4Aq2z2pFSajAwGKB8+fK5jbXQSElJITk5mdjYWJYtW8YPP/zARx99hNVqBWD8+PEYhsGmTZvSE3a7du0ICwvjrbfeSk/Oo0ePxt/fn+3bt+Nxo7nVqlWrm95r6NCh6cuGYfDwww+TlJTE1KlTef/997FYZNhBfvN08WRZz2UMWz2Mr/Z8xYCfBnDk8lHGNZ+Ih5sTjRqZHaEQwmw5Sc5ZNVdvbTr2BuYahvFfpdRDwHylVG3DuHmaHsMwZgIzAYKDg3Pd/LR3i9UsNWrUuOnx8OHDGTFiRPrjtWvX0rFjR3x9fUlOTk5f365dO1599VViYmJwcnJi69atvPrqq+mJOSt///0348ePZ+3atZw/f/6m/V26dIlSUpTZFFaLlS8f/5LKxSozZsMYJm+dxPKdIewZvRJ3F1ezwxNCmCwnzaZwICjT43L8s9t6ELAEwDCMbYAbEGCPAAujZcuWsWvXLtasWUPr1q35/PPPmTdvXvrzly5dYt68eTg7O990SxvRfeXKFaKiokhNTaVcuXLZvk9qaiqdO3dm1apVjB07lg0bNrBr1y7GjBkDYPrlW0WdUorXm79Ol2trIN6PI7Zf6Dy3HympMh5AiKIuJy3nXUA1pVQl4Bx6wFefW7Y5CzwGzFVK1UQnZ/tX/i8kateuTdWqVQHdDV2nTh1effVVunXrhqenJ/7+/jz88MO89tprWb6+TJkypKSkYLFYOHcu+6IWJ0+eJCQkhPnz59OvX7/09StXrrTvBxJ3bfx4+HFKOyj5O54jWvDr30sZvno40ztOx9nqbHZ4QgiT3LHlbBhGMjACWAccRo/KPqiUmqiU6nxjs1HA80qp/cC3wADDMApHH3Qec3V1ZcqUKVy6dInPP9eD3Nu3b8+BAwe4//77CQ4O/sfN1dUVDw8PmjdvzoIFC4iPz7p2c1xcHADOzhlf8jabjYULF+b9BxN3NGkSTJigl6e9VpeVfZfhZHFi5p6ZPDznYS5eu2hugEII0+Sk5YxhGGuANbeseyvT8iGgmX1DKzo6d+5Mw4YNmTp1KiNGjGDixIk0atSIRx55hBEjRlCxYkWioqL466+/CA0NZfbs2QBMnTqVFi1a8NBDDzFq1CjKlStHaGgo+/bt49NPP6VmzZpUqFCBMWPGYLVacXZ25sMPPzT50wqAX36BN9/Uy089BS+/DPAovz/zO09//zQ7zu3g4TkPs2XgFkp4ljAzVCGECWSoroN49913uXTpEjNmzKB8+fKEhIRQt25d3nzzTdq0acOwYcPYtGnTTaOxGzZsyNatWwkKCmLkyJF07NiRKVOmpJ+HdnFxYfny5ZQqVYr+/fvzwgsv8Mgjj/D66/+oIyPy2Wef6fvx43VFsDTNyjdj+3PbqVKsCscjj9N0VlPOx+bJlYlCCAemzOp9Dg4ONkJCQrJ9/vDhw9SsWTMfIxKOqLD+HaSmwrJl0LkzOGdxajksOoxW81pxIvIElYtVZk2fNdwXcF/+ByqEsBul1G7DMIJzsq20nIXIJ3FxEB6uly0WPf1jVokZIMg3iG2DtnGf/32ERoXy+LePczXhav4FK4QwlSRnIfJBcjL07g2NG8Nff+XsNQEeAWwduJU6JetwIvIEPZf2xJZiy9tAhRAOQZKzEHksJUWT8QdMAAAgAElEQVS3kFesgIQEuFEILkf8PfxZ1nMZgR6BrD+5nmd/epZUQyZIEaKwk+QsRB4bNEjfOznBypWQ21PolYtVZlWfVXg4e7Dwz4UMWD4AuVJRiMLNoZOzfAEVbYXh979oEXzzjV7++GNo2vTu9tOobCNW91mNi9WF+QfmM3TVUKkkJkQh5rDJ2dnZOdviGqJoiI+Pv6mASkFz4AA895xe/uwzGD783vbXsmJLlj69FDcnN2bumUmfH/twPen6vQcqhHA4DpucS5Qowblz54iLiysULSiRc4ZhEBcXx7lz5yhRouAW4Ni9W59vHjAAhg2zzz6fuO8J1vZdi7eLN0sOLqHN/DbSghaiEHLY65wBYmJiuHTpEjabjFAtapydnSlRogQ+Pj533tiBHT0KZcuCl5d997v7/G6az2lOQnICU9tMZVTTUfZ9AyGE3eXmOmeHTs5CFESJiXpEtlOOiuPevWWHl/HUkqdwtjizbdA2GpRpkLdvKIS4J1KERAiTXLoEbm668ldSUt6+V9eaXRnRcAS2VBt9fuxDnC0ub99QCJFvJDkLYSeJifDEE3r5l19ydz3z3ZrSdgq1Amtx7MoxRq2Trm0hCgtJzkLYycSJsHOnXj5xIn+Ss5uTGwu6LsDJ4sSM3TOY+r+pef+mQog8J8lZCDs4fRqmTdPLv/8OFSrk33vXL12fGZ1mAPDqL6/y/h/v59+bCyHyhCRnIe6RYcCQIbo0Z8+e0KJF/scw6MFBfNz+YwDGbBjDoj8X5X8QQgi7keQsxD1atAjWr4dixXQVMLO82PhF3m+lW81DVw3l9NXT5gUjhLgnkpyFuEdt20LfvvDhh1CypLmxvN78dTpU7UBsUixPf/80SSl5PGRcCJEnJDkLcY8CA2HBAnjmGbMjAaUUC59aSHnf8oScD+H1X183OyQhxF2Q5CzEXTp8WJ9ndjTF3IvxXbfvsCorH27/kG/2fWN2SEKIXJLkLMRd+PNPqFULKlfWJTodzUNBD/Fph08BePanZ/nl5C8mRySEyA1JzkLkUmJiRhd2TAxUqWJuPNkZGjyU4cHDMTDo/n13Ll67aHZIQogckuQsRC717w979+rlsLC8r6F9t5RSfNLhE5qUa0JMYgzPrXxOZngTooCQ5CxELsyeDUuW6OU//tCXTzkyq8XKd92+w9fVl1XHVvHB1g/MDkkIkQOSnIXIoRMnYNAgvfzcc9C8ubnx5FQFvwrMeXIOCsWYDWNY/Ndis0MSQtyBJGchcig0VN/XqQMzZ5obS251rdmVcY+MI9VIZcBPAzh46aDZIQkhbkOSsxA51LYtbNsGmzeDUmZHk3sTHp3AgHoDSEhO4Onvn+ZqwlWzQxJCZEOSsxC50KQJ+PqaHcXdm95hOjUDanL48mHaLWgnc0AL4aAkOQtxG4YBXbromtk2m9nR3DtPF09W9VlFed/y7Dy3k3YL2nEt6ZrZYQkhbiHJWYjb+PBD+OknePdduH7d7Gjso3Kxyqzus5oSniXYcnYLA38aKJdYCeFgJDkLkY2YGBg9Wi9PmgR+fubGY0+1S9Rm04BNeLt48/2h7/n3L/8m1Ug1OywhxA2SnIXIxtixulsbYOBAc2PJCzUCajCv6zwUiqnbpvKfrf8xOyQhxA2SnIXIwpYt8Omn4OwMISFgKaT/KV1qdOHbbt8C8NbGtzgRecLkiIQQIMlZiH+IjtYlOgFefhkaNDA3nrzWs3ZP+jzQB1uqjW5LuskAMSEcQI6Ss1KqvVLqqFLqhFIqywlilVI9lFKHlFIHlVKL7BumEPknJgaqVoV69WD8eLOjyR+fdviUqsWrcuDiAYasGiIDxIQw2R2Ts1LKCnwGdABqAb2VUrVu2aYa8AbQzDCM+4GX8iBWIfJFUBCsWwe//w7u7mZHkz+KuxdnRa8VeDp7sujPRUzbNs3skIQo0nLScm4EnDAMI9QwjCTgO+DJW7Z5HvjMMIwoAMMwLtk3TCHyXkoKJCToZaUKdrGRu1EzsCazOs8C4M0Nb7Lx1EaTIxKi6MpJci4LhGV6HH5jXWbVgepKqa1Kqe1KqfZZ7UgpNVgpFaKUComIiLi7iIXII1Om6PPLadNBFkU9a/dkZKORJKUk0f377pyMPGl2SEIUSTlJzllVEb71hJQTUA1oCfQGvlZK/eOqUMMwZhqGEWwYRnBgYGBuYxUiz+zfD2+9BYcOwaUi3u/zUfuP6FStE5HxkTz6zaOcvnra7JCEKHJykpzDgaBMj8sB57PY5ifDMGyGYZwCjqKTtRAOLzFRj8622WDYMGjXzuyIzGVRFhZ1W0STck0IiwljwPIBUqBEiHyWk+S8C6imlKqklHIBegErbtlmOfAogFIqAN3NHWrPQIXIK+PHw4EDUKWK7toW4OPqw4peK/Bz82PTmU28u/lds0MSoki5Y3I2DCMZGAGsAw4DSwzDOKiUmqiU6nxjs3XAFaXUIWAj8KphGFfyKmgh7GXTJpg8WS/PmweenubG40gCPQNZ9NQiFIqJmyayLWyb2SEJUWQos65nDA4ONkJCQkx5byFAj8xOu1Rq2DD4/HNz43FUo9aNYtr2aVQuVpnjI49jUVK7SIi7oZTabRhGcE62lf8yUWS5usL8+VCjhp4SUmRtUutJlPMpR2hUKB9s+cDscIQoEiQ5iyJLKejXDw4f1jW0RdZcrC580ekLAMZtHMefF/80OSIhCj9JzqLISU2FUBmumCuPV3+cFxq+QIqRwnMrn5PR20LkMUnOoshJ68r+j8yQmCvvtnqXQI9Adp7bSc+lPc0OR4hCTZKzKFKuXoV//1tf01y6tNnRFCx+bn7p3dtLDy3l5+M/mxyREIWXJGdRpDz5pK4A1ry5Pt8scqdbrW688+g7AAxZNYSL1y6aHJEQhZMkZ1Fk/PorbN6sl2fM0APCRO692vTV9OphDb9qSGiUnMAXwt4kOYsi4dgxeOopvdyzJ9x/v7nxFGSuTq4s67mMWoG1CIsJ4/FFjxOdEG12WEIUKpKcRZHwzjsQGwtdu8LChWZHU/CV8irFtkHbqF2iNocvH6bfsn6YVdBIiMJIkrMoEv77X32O+bPPwGo1O5rCIa3+tq+rL6uOrWLshrFmhyREoSHJWRQJJUroS6hkhLZ9VSpWidlPzgbg/S3v82XIlyZHJEThIMlZFFo2m57MIj7e7EgKt6dqPsW0ttMAGLp6qFxiJYQdSHIWhdbkyfDMM9CiBcjp0Lz18kMvM6HlBACGrxlObGKsyREJUbBJchaF0vbtMEHnCiZNksum8sPrzV+nVmAtTl89zbiN48wOR4gCTZKzKHQSEqBHD0hJgZdfhsceMzuiosHF6sKcJ+dgURY+3fkpm89sNjskIQosSc6i0Hn/fQgLgwce0F3bIv80KtuIUQ+NItVI5aW1L5Gcmmx2SEIUSJKcRaGycye8957uxv7sM3BxMTuiomdCywmU8irF3gt7eW/ze2aHI0SBJMlZFCpLl+opIV95BR5+2OxoiiZ3Z3dmd9aXV03cPJGQ8yEmRyREwSPJWRQqH3wAy5frimDCPB2qdeDFRi+SaqTScWFHdp3bZXZIQhQokpxFoaKUnnnK3d3sSMQHbT6gXZV2RMRF0PKblhyKOGR2SEIUGJKcRYFns8Gzz8Lu3WZHIjJzc3JjZe+VdKvZjThbHC+seUHqbwuRQ5KcRYH30Ucwdy48/TQky+Bgh+JsdWbmEzPxd/fn99O/MyNkhtkhCVEgSHIWBdqePTBmjF7+9FNwcjI3HvFPxd2LM+NxnZRH/zKaE5EnTI5ICMcnyVkUWLGx0KuX7tYePhw6dTI7IpGd7rW60+eBPsTZ4ui/rD8pqSlmhySEQ5PkLAqs0aPh+HFdbGTqVLOjEXcyvcN0yniXYVv4NiZummh2OEI4NEnOokBauBBmztRFRr75RkZnFwTF3Isx58k5gJ5ecv+F/SZHJITjkuQsCiRnZ/Dzg08+gfr1zY5G5FTbKm3pXbs3yanJdFnchctxl80OSQiHJMlZFEg9esCxYzB4sNmRiNya1XkWdUvW5fTV0wxZNcTscIRwSJKcRYESHp6xHBgoU0EWRO7O7izvtRwvFy9+PPwjU7ZOMTskIRyOJGdRYKxcCZUr6+uaRcFW0a8iMx+fCcC/f/03X4Z8aXJEQjgWSc6iQAgJgT599GVTUmikcOj9QG8+6/gZAMNWD2Pv33tNjkgIxyHJWTi8qCho2BCuXYPevfWMU6JwGN5wOMOCh2Fg8MZvb0h5TyFukOQsHJrNBq1a6WVfX5g1CyzyV1uoTGg5AV9XX9adXMfnuz43OxwhHIJ8zQmHZRi68te+ffrx1q1yPXNhFOgZmF7e8+V1L7Ph1AaTIxLCfJKchcMKDYXvvwc3N9i+He6/3+yIRF7pVbsXrzR5BVuqjWeWP0NYdJjZIQlhqhwlZ6VUe6XUUaXUCaXU67fZrrtSylBKBdsvRFFUVakCJ0/CunXQuLHZ0Yi89kGbD2hctjHhMeH0XNpT6m+LIu2OyVkpZQU+AzoAtYDeSqlaWWznDbwI7LB3kKJoyTwa298fHnnEvFhE/nGyOLGqz6r0+tv/3fZfs0MSwjQ5aTk3Ak4YhhFqGEYS8B3wZBbbvQP8B0iwY3yiiElNhZEj4f33IUx6NoucAI8AZnWeBcC4jeP448wfJkckhDlykpzLApm/JsNvrEunlKoPBBmGsep2O1JKDVZKhSilQiIiInIdrCj8PvgAZsyA//wHUqRXs0hqX7U9w4KHkZSSRJv5bVh5dKXZIQmR73KSnLMqkJh+MaJSygJ8CIy6044Mw5hpGEawYRjBgYGBOY9SFAm//gpjxujlhQuhYkVTwxEm+rj9xzz/4PMkpiTyzPJnOB973uyQhMhXOUnO4UBQpsflgMz/Kd5AbeB3pdRpoAmwQgaFidw4e1YXGDEMGDsWOnUyOyJhJmerM18+/iWtK7cmKiGKZ396VgqUiCIlJ8l5F1BNKVVJKeUC9AJWpD1pGEa0YRgBhmFUNAyjIrAd6GwYRkieRCwKnYQE6NoVLl+Gtm1h/HizIxKOQCnFgq4L8HX1Zf3J9Uz931SzQxIi39wxORuGkQyMANYBh4ElhmEcVEpNVEp1zusAReE3dSrs2aMntfj2W7BazY5IOIqSXiXTB4i99utrLD+y3OSIhMgfyqyuouDgYCMkRBrXAiIjoXt3mDwZGjUyOxrhiMb/Pp4Jmybg6ezJ3iF7qeZfzeyQhMg1pdRuwzBydMpXKoQJ0xUvDr/9JolZZO/tFm/TrWY3rtuu8+9f/y3nn0WhJ8lZmGLbNl03O+1yKZXVNQFC3KCUYmrbqXi5eLH8yHLe/+N9s0MSIk9Jchb5LiwMnnoKvvgCvv7a7GhEQVHRryIzH58JwNiNY1lxdMUdXiFEwSXJWeSrqCho3x4uXIBHH4VBg8yOSBQkvR/ozevNdHn/gT8N5FzMOZMjEiJvSHIW+SYuDtq1g0OH9AxTP/wATk5mRyUKmvcee492VdpxJf4KfX/sK+efRaEkyVnki9RU6N8fdu2C8uVhzRooVszsqERBZFEW5nedT4BHAJvObOLL3V+aHZIQdifJWeSL6dN1S9nHB1av1glaiLsV6BnIlDZTABi2ehhz9801NyAh7EySs8gXzz6rq4B9/z3Urm12NKIweKbuM7zY6EUAnv3pWT7b+ZnJEQlhP5KcRb7w9oYff9TlOYWwB6UUH3f4mPdb6cuqXlz7Ij8f/9nkqISwD0nOIs/8+it06QLx8WZHIgqzNx5+g2HBw0g1Uum3rB+R8ZFmhyTEPZPkLPLEhg3QuTP89BN89ZXZ0YjC7pMOn1C/VH0i4yPp8l0XLsddNjskIe6JJGdhd+vX6ykf4+P1dcwjRpgdkSjsnCxO/NDjB8p6l+WPs3/Q6KtGXLp+yeywhLhrkpyFXf38s76WOSEBnn8eZs4Ei/yViXxQqVgltgzcQr1S9Th19RQDfxpIqpFqdlhC3BX52hR2s3w5dOyolwcMgBkzJDGL/FXRryI/9vgRX1dfVh9fzdsb3zY7JCHuinx1CrtZfmOq3cGDYfZsSczCHJWKVeK77t9hURbe/eNdvt4jBdxFwSNfn8JuvvoKFizQLWaZZUqYqX3V9nzU7iMAXljzAutPrjc5IiFyR5KzuGupqfDhhxATox87O0PfvpKYhWMY0WgEw4KHkZSSRKdFnZi/f77ZIQmRY5KcxV1JSdEjsV95BXr0AJl7QDgapRTTO07n303/TXJqMv2X92fO3jlmhyVEjkhyFrkWFwe9esHcueDhoRO0tJaFI7IoCx+0+YD/tP4PAK/+8irHrhwzOSoh7kySs8iViAho1QqWLtUlOdevl5KcwvGNbjqatlXaciX+Ck1nNWXXuV1mhyTEbUlyFjl24AA89BDs2KFnldq6FZo1MzsqIe5MKcXi7otpU7kNV+Kv8PCch1l3Yp3ZYQmRLUnOIsfmz4eTJ6F+fZ2gH3jA7IiEyDk/Nz9W91nNgHoDSExJpPcPvaXMp3BYkpxFjk2eDJ9+Cn/8AaVKmR2NELnnbHVmVudZPFrxUaISoujyXRfibHFmhyXEP0hyFtk6dkyX4rx4UT+2WnWdbE9Pc+MS4l5YlIVZnWdRxrsMW8O2MnTVUCnzKRyOJGeRpWXLIDhYD/gaM8bsaISwr0rFKvFz359xd3Jn/oH5DF01lJjEGLPDEiKdJGdxk+RkeP11eOopiI2F7t11oREhCps6JevwQ48fsCorX+35iuCZwZyLOWd2WEIAkpxFJqGhusrXBx/oLuz//heWLNGXTAlRGHWo1oHtz22ncrHKHI88zkOzHuLgpYNmhyWEJGehxcRAlSoZjzdskOIiomgILhPM5gGbaVy2MWExYTSf05w/zvxhdliiiJPkLADw8dHlOAHCw+GRR8yNR4j8VNanLBuf2cgT1Z/gasJVWs1rJZXEhKkkORdRcXEwdix8913GupkzdY3ssmXNi0sIs7g7u/Njzx95vPrjJKcm02x2M/4X9j+zwxJFlCTnImjVKrj/fnjvPXj5ZUhI0Otl/mVR1DlZnFj01CLaVmnL5bjLtF/Qnq/3fI0hM7uIfCZfx0XImTPQpQs88QScPg116sAPP4Cbm9mRCeE4vF29Wd1nNU/XeprYpFieX/k8A1cMlGuhRb6S5FwEJCXBpElQqxb89BN4eenLo3bvhqZNzY5OCMfjZHFicffFTGkzBXcnd+bum8u4DePMDksUIZKciwAnJ1iwQJ9n7tEDjhyBl17S64UQWVNKMbrpaH7o8QMKxaQtk5i1Z5bZYYkiIkfJWSnVXil1VCl1Qin1ehbPv6KUOqSUOqCU+k0pVcH+oYqcSkyEzz+HczfqKVgs8NFH8MsvsHixDPgSIjc6VOvAO4++g4HBcyuf4+2Nb8s5aJHn7piclVJW4DOgA1AL6K2UqnXLZnuBYMMw6gBLgf/YO1BxZ/HxMG0alCwJL7wA776b8VybNtC6tXmxCVGQjXlkDDMfn4lFWZi4eSJdF3flzNUzZoclCrGctJwbAScMwwg1DCMJ+A54MvMGhmFsNAwjbWqX7UA5+4YpbufcOXjrLahYEUaNguhoPZ1ju3ZmRyZE4fF8g+f5ttu3eLt489PRn6j/ZX22h283OyxRSOUkOZcFwjI9Dr+xLjuDgJ+zekIpNVgpFaKUComIiMh5lCJbH38MFSrAO+/ApUs6Kc+fD/v365HZQgj76XF/Dw4OP0ibym2ISoiizfw2ci20yBM5Sc5ZFXDM8oSLUqofEAxMyep5wzBmGoYRbBhGcGBgYM6jFOmuXtUDutK0bKnPKT/9NGzcqJNyv35SdlOIvBLkG8TK3itpU7kN15Ku0WFhB2buninnoYVd5SQ5hwNBmR6XA87fupFSqjUwBuhsGEaifcIToKt2/fEH9O8PpUvD889nPFe3Lvz9t56gomVLScpC5AdXJ1dW9VlFz/t7EpMYw5BVQ2i3oB3nY//x1SjEXclJct4FVFNKVVJKuQC9gBWZN1BK1Qe+RCfmS/YPs2iKiNAzQ9WqpWtdz5+vq3m5uenBX2n8/c2LUYiiysXqwrfdvmV+1/kUdy/OL6G/UG9GPXaE7zA7NFEI3DE5G4aRDIwA1gGHgSWGYRxUSk1USnW+sdkUwAv4Xim1Tym1IpvdiRxav15f8jR6tO7GLlUK3ngDTpzQl0S5u5sdoRBCKUW/Ov34c9ifPFDiASLiInhk7iMsP7Lc7NBEAafMOk8SHBxshISEmPLejiYiAtas0dM2jhyp18XEQLlyusX8/PPQqZMUDRHCkcUmxjJk1RC+/etbLMrCm83fZHzL8VgtVrNDEw5CKbXbMIzgnGwrX/cmMAz46y9YtgzWroUdOyA1VXdPDx8OVquewjE8XN8LIRyft6s3C59aSEnPkny842Pe/eNdohOj+aj9R1iUFGMUuSPJOZ+tWAFDhsCFCxnrnJ11kZDOncFm08kZJDELUdAopfiw/Ye0r9qex799nE93fsqhiEPM6jyLCn5SOFHknBzO5ZHYWD014yuvwOzZGevLltWJuWRJGDRIT0Rx5YpuQQ8fLjNECVEYtKvajtV9VuPv7s9vp36jzow67D6/2+ywRAEi55ztJCYGtm2D33/Xt5AQSE7WzzVrBlu26OXUVDh0SM+nLJc9CVG4Xbp+iYE/DWT18dUUdy/Or//3K/VL1zc7LGGS3JxzluR8F1JS4PBhCAoCX1+9btCgm1vIFgs0agSPPaa7rFu0MCdWIYS5EpMTefr7p1l5bCX+7v6s67eOBmUamB2WMIEMCLOjuDg9eGvPHti3T9/+/FOvX7JEV+YCePhhvV3LlvrWrJmcMxZC6IIlS3sspevirqw5voZms5vxRvM3GPvIWBnJLbIlLecbUlLgzBk4fx6aN9frbDbw9tZTMN6qfHldz7p///yNUwhRMCWlJDF45WC+2f8NANX9q/NDjx+oXaK2yZGJ/CLd2ndw/Dhs2AAnT2bcjh3TVbf8/CAyMuN8cIMGOjk/+CDUq6dvdepAQIApoQshCrglB5cwZNUQriZcxc3JjWltpzE0eChKBqEUekU6OcfHQ2jozYn35EkYPBi6dtXbfPWVfnyr0qX1QK0ff9QtZtADuCwypl0IYUdXE67yyrpXmLNvDgBda3Rl9pOz8XPzMzkykZcK9Tlnw9At25Mn4fJl6Ngx47nq1XWrOCsNGmQk5wcfhAEDoEqVjFu1alCs2D9fJ4lZCGFvfm5+zH5yNq0qtWLY6mEsO7KMvy79xdIeS6lTso7Z4QkH4FAt55QUPSexh0fGKOg//oAFC/S54PBwOHUKoqP1c56e+nritN6gOnX0KOqKFW9OvFWq6O7oClIDQAjhYP669BcdF3YkLCYMgOfqP8fERydS2ru0yZEJeysQ3dqlSwcbTzwRwsWLOvGePw8XL+oEPWOGrqIFMGcODBx482u9vDKS7vz5OpmDrlFdrJjUoBZCFCxR8VEMWTWEHw//SIqRgp+bH5+0/4R+dfrJuehCpEAkZyenYCMl5Z/nnAMCYPx4eOEF/fjYMfj1VyhTRlfXqlABAgOlgIcQovA5FHGI0etH8/OJnwHoWK0jC7ouoJh7FufcRIFTIJJz+fLBxptvhlCihE68ZcroaRFdXEwJRwghHIJhGMzdN5eRP4/kuu06VYtX5f1W79O9VndpRRdwBSI5O9p1zkII4UhORp7k8W8f58jlIwDUKVmHSY9NomO1jnd4pXBUuUnOMhZZCCEcUJXiVdg/dD8TWk6gpGdJDlw8QKdFnRi6aihh0WFmhyfymCRnIYRwUC5WF95q8RZnXjrDuEfGYVVWvtz9JeU/Ks/jix7n+JVsrh0VBZ4kZyGEcHCuTq5MfHQi+4buo0uNLrhYXVh9fDXVp1enyidVmL9/PimpKWaHKexIkrMQQhQQtUvUZlnPZRx54Qj/V+f/8HLxIjQqlP7L+1Pt02p8GfIlCckJZocp7EAGhAkhRAFlS7Gx8M+FvLP5HUKjQgFdfax/nf688fAblPIqZXKEIjMZrS2EEEVISmoK3+z/huk7p7P3wl4A3JzcGFhvIK889ApVilcxOUIBkpyFEKLI2ndhHxM2TWD5keUAWJWVLjW6MLjBYNpUbiPXSptIkrMQQhRxBy8d5J3N77D44OL0ddX9q9O6UmvGtxxPoGegidEVTZKchRBCAHDsyjGWHFzC1P9NJTpRzxrk5uRGr9q9GNpgKI3KNpLWdD6R5CyEEOIm8bZ4toZtZfKWyfx26rf09TUDajKo/iD+r+7/UcKzhIkRFn6SnIUQQmTrROQJpu+czrd/fcul65cAcLI40aZyG4Y3HE6nap2kNZ0HJDkLIYS4I1uKjTXH1/D13q9ZdWxV+vqKfhXpUasHnap34qFyD+FsdTYxysJDkrMQQohcORF5gll7ZvHN/m/4+9rf6es9nT1pGtSUmgE1CS4TzBP3PYGfm5+JkRZckpyFEELclZTUFDaf2cyyI8tYf3I9R68cvel5V6srPe7vQcuKLWkW1Izq/tWlCzyHJDkLIYSwiwvXLrD17Fb+vPQn606uY3v49pueD/AIoGlQU5oFNaNZUDMalGmAm5ObSdE6NknOQggh8sSRy0dYfWw1W8O2sjVsa/qAsjQuVheCywTTPKg5XWp0oXG5xliUTOMAkpyFEELkA8MwCI0K1Yn6rE7WByMO3rRNaa/SdL6vM49Xf5waATUo71seF6uLSRGbS5KzEEIIU0TFR/G/sP+x7Mgy1hxfc9PgMgB3J3fuL3E/Zb3LUsa7DD6uPrSs2JJ6pepR0rNkoT5/LclZCCGE6QzDYO+Fvfx05Cc2nt7ImegznI0+m+32Pq4+VPKrRKVilajsV5lq/tWoVrwaxdyL4efmR0nPkni6eObjJ7AvSc5CCCEc0uW4yxy/cpzzsefZHr6d2KRYDkUcYv/F/cQkxtzx9T6uPpT3LU8pr/2jOZoAAAvVSURBVFI8UOIBqhWvRnH34hRzL0Zx9+KU9y1PoEegQ7bA7Z6clVLtgY8BK/C1YRiTb3neFZgHNACuAD0Nwzh9u31KchZCCJHGMAyuxF/hVNQpQqNCOXX1FIciDhEWE8bVhKtExUdxPvY8tlTbHfflanWlhGcJfFx98Hb1xtvFO+P+xrKXixeezp54unji4exx22UvFy+sFus9f8bcJGenHOzMCnwGtAHCgV1KqRWGYRzKtNkgIMowjKpKqV7AB0DP3IcuhBCiKFJKEeARQIBHAA3LNsxyG8MwuHj9IhevXeRM9BkOXDxAWHQYUQlRRMZHciX+CqFRocQkxhAWE2a/2FD4uPrg5+ZHMfdieLl4YVEWrMqKRVn0siXTcjbrc+OOyRloBJwwDCMUQCn1HfAkkDk5PwmMv7G8FJiulFKGWX3mQgghCh2lFKW8SlHKqxR1S9Wl832ds9zuetJ1Ll2/RGxSLLGJsVneX0u6RpwtjutJ17luu66X0+4zr0u6TmxSLNGJ0UQnRnMm+ky+fNacJOeyQOZDkHCgcXbbGIaRrJSKBvyBy5k3UkoNBgbfeHhNKXVz6Zn8FcAt8RUx8vmL7ucvyp8d5PPL5zfv81fI6YY5Sc5ZnVW/tUWck20wDGMmMDMH75nnlFIhOe37L4zk8xfdz1+UPzvI55fPXzA+f046wcOBoEyPywHns9tGKeUE+AKR9ghQCCGEKGpykpx3AdWUUpWUUi5AL2DFLdusAJ65sdwd2CDnm4UQQoi7c8du7RvnkEcA69CXUs02DOOgUmoiEGIYxgpgFjBfKXUC3WLulZdB24lDdK+bSD5/0VWUPzvI55fPXwCYVoRECCGEEFmTqUKEEEIIByPJWQghhHAwBT45K6VGKqWOKqUOKvX/7Z1/jJTFGcc/33KCCUK4E4Wr0Mg1xURsqtQ2vdriSW1FINgfpkLTBGv6h6BNiW0MSGKpSWPF1KK1ERIbW1soIIXS8KMUhFJs4GirojZKORRbKmqpUrQYKjr9Y57l9l7f3b292+y7e/t8kjc788wz78y8M3PP7rxzz2hxnnyBpC5LuypPPsVkXZLm58nHSeqUdEDSKtv8hqQhFu+y9PNLlVFtJH1bUpA00uKSdJ/V7SlJE/N0Z1sbD0ianSf/qKSnLc99Mse0klokbTX9rZKaS5VRxXbfLek5K3+dpBF5aQ3T/+VQqP31gKSxknZIetbm+zdNXvYYrdQ8yAJJgyQ9IWmDxSs2dsudH9VG0ghJa2zePyupfcD2fwihbi/gCmAbMMTi59rnhcA+YAgwDjhI3Mw2yMJtwGDTudDyrAZmWngpMMfCc4GlFp4JrCpWRgbPYCxxs96LwEiTTQU2E////BNAp8lbgOfts9nCzZa2F2i3PJuBq02+GJhv4fnAXcXKqHLbPwc0WfiuvLo1TP+X+bwKtr8eLqAVmGjhYcDfrB/KGqOVnAcZPYdbgBXAhkqO3b7Mjwza/jPg6xYeDIwYqP2f+YTrZ0etBq5MkS8AFuTFt9gDbwe2JPWsI47S/Yf+tF4ur4WbTE+FysjgGawBPgIcots4LwNm5ensJ/5hmwUsy5MvM1kr8Fye/LReLq+FW4H9xcrIcCx8AVjeaP1f5jNKbX/W9epHe9YTff6XNUYrOQ8yaPMY4FFgMrChkmO3L/Ojym0fDryAbWRO9utA6/96X9YeD3zallt2Ssp5S09zOXpeEfnZwLEQwqmEvMe9LD3nmrTQvaqGpBnAP0MI+xJJ5bb/PAsn5QCjQghHAOzz3BJlZMUNxG+60CD93wfqsc6p2BLtJUAn5Y/RSs6DarMEuBV41+KVHLt9mR/VpA34F/CQLes/KGkoA7T/e+O+M1MkbQNGpyQtJNa/mbhk8TFgtaQ2CrsTTfsyEoroUyStVy5L+0uJ9t9GXNp9T7YUWbE696Utmbc/hLDedBYCp4DlJepWd/1fYeqxzu9B0lnAr4B5IYTjKnxubzXmQdWQNB14NYTwF0kdOXGKal/Hbl/mRzVpAiYC3wghdEq6l7jEXIi67v+aN84hhCsLpUmaA6wNca1hr6R3iU7Ni7kcTZMfBUZIarJvh/n6uXsdVk/XpL1xa9pvCrVf0oeJ74v22R+nMcDjkj5epG6HgY6E/PcmH5OiD/CKpNYQwhFJrcCrJs+0/TlsM8d04DM2DkrVra76v8LUY517IOkMomFeHkJYa+Jyx2gl50E1uQyYIWkqcCZxmXcJlR275c6PanIYOBxC6LT4GqJxHpj9X+33BhV+B3EjcIeFxxOXKgRMoOeGh+eJmx2aLDyO7g0PEyz/I/Tc8DDXwjfRc1PFagunlpHhszhE9zvnafTcCLHX5C3EdzbNdr0AtFjan0w3txFiqsnvpudGiMXFyqhym6cQjy49JyFvuP7v5fMq2P56uGysPQwsScjLGqOVnAcZPosOujeEVWTs9mV+ZNDuXcAFFl5k/TIg+z/zCdfPjhoM/AJ4BngcmJyXtpC483A/tuPO5FOJuzwPEpdGc/I24k69LhuIuR3gZ1q8y9LbSpWR0bM4RLdxFvBjq9vTwKV5ejdYW7qAr+XJL7XneBC4n27vcWcTN6AcsM+WUmVUsc1dxC9kT9q1tFH7v4xnltr+eriATxGXGZ/K6/OpfRmjlZoHGT6LDrqNc8XGbrnzI4N2Xwz82cbAr4nGdUD2v7vvdBzHcZwao953azuO4zjOgMONs+M4juPUGG6cHcdxHKfGcOPsOI7jODWGG2fHcRzHqTHcODtOhVE8IazUdch0fyrpcIlbVgVJi6xuFXFOlLtfL/Q6rNyOSpTrOAOBmvcQ5jh1SHsivo7o0GFRnuxk1WrjOE7d4cbZcSpMCGFPflzSSeBoUt5fJA0JIbiRd5wBiC9rO04NIOkSSbsknbAD3W9MpF9vS7+TJD0i6RjxRKZc+uWSHpX0hqT/Stoi6aLEPa6S9EdJ/5H0pqT9km5Pqc44SRtN50VJt0t6X+JeF0haJ+mYpLck7ZE0pRftPEfSCknHLe/DxDN5HcfJw42z42TPcGAF0RXtNUT/vg9IuiJFdznRF/C12Ik8kqYRXQq+CXwV+AowDNglaazptAG/Ibp5vQ6YAdwDDE0pYx2wHfg80UXid4HZuURJ7wceI54jfjPwZeAYsFHS1SXaupZ4UMltVo9TwI9K5HGchsOXtR0ne4YRDxLYASDpD8SjQGcBOxK6a0IItyZk9wI7QwjX5ASSdhAPMfgWMI941N5gYE4I4bipbS9Qnx+EEB6y8DZJk60uOdktRJ/G7SGELitvE/EQku/Rfa52DyR9lugfe1YIYaWJt0jaTM/TgByn4fFfzo6TPSdyhhnA3iMfAD6QorsuPyLpQ8AHgeWSmnIXcALYDUwy1SeBt4GVkq6VVOyw+I2J+DOJukwC9uQMs9X5HeCXwMWShhe4bzvwDvHIx3xWpug6TkPjxtlxsuf1FNlJ4qlCSY4k4jkj+xOi8c2/phNP08EM6VXEOf9z4GVJnZIuTynjtRJ1aUmpB8DLxJOAmlPSAFqB10MIbyfkrxTQd5yGxZe1Hae+SP7f8L/tcwGwLUX/f6czxl/nOyQNAS4D7iC+Jz4/hHC0jDq8BoxOkY+2+iWNe44jQLOkMxIGelQZZTtOQ+DG2XHqm/3ETV4TQgjf700GWzbfLuksYD0wDijHOO8E5plRPwQgaRBxg9cTIYQ3CuTbDQwCvkTPpeyZZZTtOA2BG2fHqWNCCEHSTcB6SYOB1URDOwr4JPD3EMI99q9Zk4BNwD+AkcRf2y8R3ymXww+B64Gtkr4DHAfmAuOBaUXqulXSY8AySSOJ79WvAy4qlMdxGhV/5+w4dU4IYRPR8A4FHgS2AIuJy8y7TW2fpd8J/A64n/gvWZNDCG+VWd5LxF3XfwUeANYQ30NPCyH8tkT2LxK/INwJrCL+QLi5nPIdpxFQCCVd3zqO4ziOU0X8l7PjOI7j1BhunB3HcRynxnDj7DiO4zg1hhtnx3Ecx6kx3Dg7juM4To3hxtlxHMdxagw3zo7jOI5TY7hxdhzHcZwa4//r2e1H1BZrfQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 576x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def plot_precision_recall_vs_threshold(precisions, recalls, thresholds):\n",
    "    plt.plot(thresholds, precisions[:-1], \"b--\", label=\"Precision\", linewidth=2)\n",
    "    plt.plot(thresholds, recalls[:-1], \"g-\", label=\"Recall\", linewidth=2)\n",
    "    plt.xlabel(\"Threshold\", fontsize=16)\n",
    "    plt.legend(loc=\"upper left\", fontsize=16)\n",
    "    plt.ylim([0, 1])\n",
    "\n",
    "plt.figure(figsize=(8, 4))\n",
    "plot_precision_recall_vs_threshold(precisions, recalls, thresholds)\n",
    "plt.xlim([-700000, 700000])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x432 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def plot_precision_vs_recall(precisions, recalls):\n",
    "    plt.plot(recalls, precisions, \"b-\", linewidth=2)\n",
    "    plt.xlabel(\"Recall\", fontsize=16)\n",
    "    plt.ylabel(\"Precision\", fontsize=16)\n",
    "    plt.axis([0, 1, 0, 1])\n",
    "\n",
    "plt.figure(figsize=(8, 6))\n",
    "plot_precision_vs_recall(precisions, recalls)\n",
    "#save_fig(\"precision_vs_recall_plot\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.9322813938198553\n",
      "0.52315071020107\n"
     ]
    }
   ],
   "source": [
    "y_train_pred_90 = (y_scores > 70000)\n",
    "print(precision_score(y_train_5, y_train_pred_90))\n",
    "print(recall_score(y_train_5, y_train_pred_90))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# The ROC curve is another common tool used with binary classifiers. It plots the true position rate (another name for recall) against the false positive rate.  The FPR is the ratio of negative instances that are incorrectly classified as positive(1- true negative rate). The TNR is also called specificity, hence the ROC curve plots sentitivity(recall) versus 1 - specificity\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import roc_curve\n",
    "fpr, tpr, threshold = roc_curve(y_train_5, y_scores)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "def plot_roc_curve(fpr, tpr, label = None):\n",
    "    plt.plot(fpr, tpr, linewidth = 2, label = label)\n",
    "    plt.plot([0,1],[0,1],'k--')\n",
    "    plt.axis([0,1,0,1])\n",
    "    plt.xlabel('False Positive Rate')\n",
    "    plt.ylabel('True Positive Rate')\n",
    "\n",
    "plot_roc_curve(fpr, tpr)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9642913019917457"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.metrics import roc_auc_score\n",
    "roc_auc_score(y_train_5, y_scores)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Since the ROC curve is so similar to the precision/recall curve, as a rule of thumb, you should prefer the PR curve whenever the positive class is rare or when you care more about the false positives than the false engatives, and the ROC curve otherwise."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "D:\\anaconda\\lib\\site-packages\\sklearn\\ensemble\\forest.py:248: FutureWarning: The default value of n_estimators will change from 10 in version 0.20 to 100 in 0.22.\n",
      "  \"10 in version 0.20 to 100 in 0.22.\", FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\ensemble\\forest.py:248: FutureWarning: The default value of n_estimators will change from 10 in version 0.20 to 100 in 0.22.\n",
      "  \"10 in version 0.20 to 100 in 0.22.\", FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\ensemble\\forest.py:248: FutureWarning: The default value of n_estimators will change from 10 in version 0.20 to 100 in 0.22.\n",
      "  \"10 in version 0.20 to 100 in 0.22.\", FutureWarning)\n"
     ]
    }
   ],
   "source": [
    "from sklearn.ensemble import RandomForestClassifier\n",
    "forest_clf = RandomForestClassifier(random_state = 123)\n",
    "y_probas_forest = cross_val_predict(forest_clf, X_train, y_train_5, cv = 3, method = 'predict_proba')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "y_scores_forest = y_probas_forest[:,1]\n",
    "fpr_forest, tpr_forest, threshold_forest = roc_curve(y_train_5,y_scores_forest)\n",
    "plt.plot(fpr,tpr,'b:',label = 'SGD')\n",
    "plot_roc_curve(fpr_forest, tpr_forest, 'Random Forest')\n",
    "plt.legend(loc = 'lower right')\n",
    "plt.show() # the random forest roc is closer to the top left corner"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9923078555535422"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "roc_auc_score(y_train_5, y_scores_forest)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([5], dtype=int8)"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sgd_clf.fit(X_train,y_train)\n",
    "sgd_clf.predict([some_digit])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[-120360.21365249, -458422.21222651, -380424.43428467,\n",
       "         -73653.53488748, -501161.61837561,  226129.12172593,\n",
       "        -849225.61893952, -395667.46908569, -678340.42021237,\n",
       "        -752388.70367284]])"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "some_digit_scores = sgd_clf.decision_function([some_digit])\n",
    "some_digit_scores"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9], dtype=int8)"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sgd_clf.classes_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "5"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sgd_clf.classes_[5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n",
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "D:\\anaconda\\lib\\site-packages\\sklearn\\linear_model\\stochastic_gradient.py:144: FutureWarning: max_iter and tol parameters have been added in SGDClassifier in 0.19. If both are left unset, they default to max_iter=5 and tol=None. If tol is not None, max_iter defaults to max_iter=1000. From 0.21, default max_iter will be 1000, and default tol will be 1e-3.\n",
      "  FutureWarning)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([5], dtype=int8)"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# if you want to force sklearn to use OvO or OvA\n",
    "from sklearn.multiclass import OneVsOneClassifier\n",
    "ovo_clf = OneVsOneClassifier(SGDClassifier(random_state = 123))\n",
    "ovo_clf.fit(X_train, y_train)\n",
    "ovo_clf.predict([some_digit])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "D:\\anaconda\\lib\\site-packages\\sklearn\\ensemble\\forest.py:248: FutureWarning: The default value of n_estimators will change from 10 in version 0.20 to 100 in 0.22.\n",
      "  \"10 in version 0.20 to 100 in 0.22.\", FutureWarning)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "array([[0. , 0. , 0. , 0. , 0.1, 0.7, 0. , 0.1, 0.1, 0. ]])"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "forest_clf.fit(X_train, y_train)\n",
    "forest_clf.predict([some_digit])\n",
    "# this time sklearn did not have to run OvO or OvA because Random Forest clf can directly classify instances into multiple classes\n",
    "# you can call predict_proba() to get the list of probabilityies that the classifier assigned to each instance for each class\n",
    "forest_clf.predict_proba([some_digit])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.8619776 , 0.8760438 , 0.84587688])"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# evaluate\n",
    "cross_val_score(sgd_clf, X_train, y_train, cv = 3, scoring = 'accuracy')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.91066787, 0.90864543, 0.91058659])"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# any improvement?\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "scaler = StandardScaler()\n",
    "X_train_scaled = scaler.fit_transform(X_train.astype(np.float64))\n",
    "cross_val_score(sgd_clf, X_train_scaled, y_train, cv = 3, scoring = 'accuracy')  # boosted the accuracy to 90%"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[5741,    3,   18,   12,   11,   46,   40,    9,   39,    4],\n",
       "       [   1, 6481,   46,   28,    6,   41,    8,    9,  112,   10],\n",
       "       [  51,   32, 5331,  105,   79,   25,   86,   60,  174,   15],\n",
       "       [  45,   41,  134, 5377,    4,  219,   33,   53,  127,   98],\n",
       "       [  23,   30,   38,    8, 5357,   10,   51,   38,   81,  206],\n",
       "       [  71,   43,   28,  198,   78, 4602,  115,   26,  172,   88],\n",
       "       [  38,   23,   42,    2,   39,   81, 5638,    8,   47,    0],\n",
       "       [  24,   24,   72,   31,   62,   12,    4, 5795,   16,  225],\n",
       "       [  52,  158,   68,  177,   13,  157,   58,   26, 5010,  132],\n",
       "       [  43,   35,   25,   88,  158,   38,    2,  210,   84, 5266]],\n",
       "      dtype=int64)"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train_pred = cross_val_predict(sgd_clf, X_train_scaled, y_train, cv = 3)\n",
    "conf_mx = confusion_matrix(y_train, y_train_pred)\n",
    "conf_mx"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 288x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.matshow(conf_mx, cmap = plt.cm.gray)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP4AAAECCAYAAADesWqHAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAADFhJREFUeJzt3W9oXfUdx/HPp0n/pKnDwvxDW1kdGWopzow4dRUR/+Hc2B440IHi9sAKbupUGFMRnwnCHBOcw1JXxOn2oKs4xtgf2BTmg7K0ipvNprVbNZt/ulatNlGb5LsHScA513OunN89ufu+XyC08fTLl/S+c25uzz1xRAhALovaXgBA9xE+kBDhAwkRPpAQ4QMJET6QUGvh277Y9l9t77b9nbb2qMv2CbZ/b3vM9rO2b2h7pzps99l+yvYv2t6lDttH295q+y9zn+uz2t6piu0b5x4Tf7b9E9vL2t6pSivh2+6T9ANJn5e0TtJXba9rY5cOTEm6OSJOkXSmpG/0wM6SdIOksbaX6MA9kn4VESdL+rQW+O62V0u6XtJIRKyX1Cfp8na3qtbWGf+zknZHxJ6IeE/STyV9uaVdaomIlyNi59yv39LsA3J1u1sdme01kr4gaXPbu9Rh+2OSzpH0gCRFxHsR8Ua7W9XSL2nAdr+k5ZL+2fI+ldoKf7Wkl973+3Et8Ijez/ZaScOStre7SaXvS/q2pJm2F6npk5L2Sdoy9+3JZtuDbS91JBHxD0nflfSipJclvRkRv2l3q2pthe8P+VhPXDtse4Wkn0n6VkQcbHuf/8X2FyW9FhE72t6lA/2SPiPphxExLOmQpAX9+o/tlZp9tnqipFWSBm1f0e5W1doKf1zSCe/7/Rr1wNMj24s1G/3DEbGt7X0qbJD0Jdt/1+y3UufZ/nG7K1UalzQeEfPPpLZq9gvBQnaBpL9FxL6IOCxpm6TPtbxTpbbC/6OkT9k+0fYSzb4Y8vOWdqnFtjX7vedYRHyv7X2qRMQtEbEmItZq9vP7u4hY0GeiiHhF0ku2T5r70PmSdrW4Uh0vSjrT9vK5x8j5WuAvSEqzT626LiKmbH9T0q81+yrojyLi2TZ26cAGSVdK+pPtp+c+dmtE/LLFnf4fXSfp4bkTwh5JX295nyOKiO22t0raqdl/+XlK0qZ2t6pm3pYL5MOVe0BChA8kRPhAQoQPJET4QEKth297Y9s7dKLX9pXYuRt6bd/Ww5fUU58w9d6+Ejt3Q0/tuxDCB9BlRS7gsd1zVwX199e7iHFmZkaLFtX/ejkz0/4b4yJCs1eT1rNsWZn7SHTyWJuamqr9dyJJ77777kdZqdLSpUtrHdfpvpI0OTn5UVaqFBGVf9mtXLL7UfX19RWbvXLlyiJzSz0gpc5C6sQpp5xSZG7Jz8Xzzz9fZO7Q0FCRuZK0a1fzb0OYmpqqdRxP9YGECB9IiPCBhAgfSIjwgYRqhd9r98AHcGSV4ffoPfABHEGdM37P3QMfwJHVCb+n74EP4L/VuXKv1j3w596d1FNvVACyqhN+rXvgR8Qmzd1dtBev1QcyqfNUv+fugQ/gyCrP+D16D3wAR1Dr3XlzPzSCHxwB/J/gyj0gIcIHEiJ8ICHCBxIifCChYvfc6+TmjnVNT083PnPeqlWrisw9cOBAkbmS9OqrrxaZ+/bbbxeZu2fPniJzpTKPN0m66KKLisyVpN27dzc+s+7NXTnjAwkRPpAQ4QMJET6QEOEDCRE+kBDhAwkRPpAQ4QMJET6QEOEDCRE+kBDhAwkRPpAQ4QMJET6QEOEDCRE+kBDhAwkRPpAQ4QMJET6QkCOa/1H2K1asiNNOO63xuaVuJy2VudWxJN12221F5krS3r17i8x94oknisy9+OKLi8yVyn0u7rzzziJzJenCCy9sfObBgwc1NTVVea9xzvhAQoQPJET4QEKEDyRE+EBChA8kRPhAQpXh2z7B9u9tj9l+1vYN3VgMQDn9NY6ZknRzROy0fZSkHbZ/GxG7Cu8GoJDKM35EvBwRO+d+/ZakMUmrSy8GoJyOvse3vVbSsKTtJZYB0B21w7e9QtLPJH0rIg5+yP/faHvU9ujU1FSTOwJoWK3wbS/WbPQPR8S2DzsmIjZFxEhEjPT313npAEBb6ryqb0kPSBqLiO+VXwlAaXXO+BskXSnpPNtPz/13SeG9ABRU+Zw8Iv4gqfL9vQB6B1fuAQkRPpAQ4QMJET6QEOEDCRW50iYi9M477zQ+d/ny5Y3PnHf//fcXmXvNNdcUmStJpS6UmpiYKDJ33bp1ReZK0vHHH19k7pNPPllkriRdddVVjc985JFHah3HGR9IiPCBhAgfSIjwgYQIH0iI8IGECB9IiPCBhAgfSIjwgYQIH0iI8IGECB9IiPCBhAgfSIjwgYQIH0iI8IGECB9IiPCBhAgfSIjwgYQcEY0PXbJkSRx77LGNz122bFnjM+cNDAwUmfvCCy8UmStJk5OTReYODg4WmTs8PFxkriS98cYbReaefvrpReZK0n333df4zA0bNmjHjh2VP+SWMz6QEOEDCRE+kBDhAwkRPpAQ4QMJET6QUO3wbffZfsr2L0ouBKC8Ts74N0gaK7UIgO6pFb7tNZK+IGlz2XUAdEPdM/73JX1b0kzBXQB0SWX4tr8o6bWI2FFx3Ebbo7ZHZ2b4+gAsZHXO+Bskfcn23yX9VNJ5tn/8wYMiYlNEjETEyKJF/GMBsJBVFhoRt0TEmohYK+lySb+LiCuKbwagGE7NQEL9nRwcEY9LerzIJgC6hjM+kBDhAwkRPpAQ4QMJET6QUEev6td11FFH6dxzz2187u7duxufOW///v1F5k5MTBSZK0lXX311kbmbN5d5S8Zll11WZK4k3X333UXm3n777UXmStKtt97a+Mzx8fFax3HGBxIifCAhwgcSInwgIcIHEiJ8ICHCBxIifCAhwgcSInwgIcIHEiJ8ICHCBxIifCAhwgcSInwgIcIHEiJ8ICHCBxIifCAhwgcSckQ0PnRgYCCGhoYan/v66683PnPeMcccU2TuM888U2SuJK1fv77I3OHh4SJzH3zwwSJzJWlgYKDI3FNPPbXIXEnavn17kbkR4apjOOMDCRE+kBDhAwkRPpAQ4QMJET6QEOEDCdUK3/bRtrfa/ovtMdtnlV4MQDl1f0z2PZJ+FRFfsb1E0vKCOwEorDJ82x+TdI6kr0lSRLwn6b2yawEoqc5T/U9K2idpi+2nbG+2PVh4LwAF1Qm/X9JnJP0wIoYlHZL0nQ8eZHuj7VHbo9PT0w2vCaBJdcIflzQeEfPvKNiq2S8E/yEiNkXESESM9PX1NbkjgIZVhh8Rr0h6yfZJcx86X9KuolsBKKruq/rXSXp47hX9PZK+Xm4lAKXVCj8inpY0UngXAF3ClXtAQoQPJET4QEKEDyRE+EBChA8kVPff8TsyPT2tAwcOlBhdzNq1a4vMPXjwYJG5knTccccVmbtt27Yic/v7izzcJEmTk5NF5j733HNF5krS4cOHG595xhln1DqOMz6QEOEDCRE+kBDhAwkRPpAQ4QMJET6QEOEDCRE+kBDhAwkRPpAQ4QMJET6QEOEDCRE+kBDhAwkRPpAQ4QMJET6QEOEDCRE+kFCR254uXbpUQ0NDjc+94447Gp8579577y0y97HHHisyV5LOPvvsInMfffTRInPvuuuuInOlcnfD3bt3b5G5kvTQQw81PnP//v21juOMDyRE+EBChA8kRPhAQoQPJET4QEKEDyRUK3zbN9p+1vafbf/E9rLSiwEopzJ826slXS9pJCLWS+qTdHnpxQCUU/epfr+kAdv9kpZL+me5lQCUVhl+RPxD0nclvSjpZUlvRsRvSi8GoJw6T/VXSvqypBMlrZI0aPuKDzluo+1R26OHDx9uflMAjanzVP8CSX+LiH0RcVjSNkmf++BBEbEpIkYiYmTx4sVN7wmgQXXCf1HSmbaX27ak8yWNlV0LQEl1vsffLmmrpJ2S/jT3ZzYV3gtAQbXejx8Rd0gq92Z4AF3FlXtAQoQPJET4QEKEDyRE+EBChA8k5IhofOjg4GCcfPLJjc+dmJhofOa8Q4cOFZl7ySWXFJkrSZdeemmRuTfddFORubt27SoyVyr393fttdcWmStJW7ZsKTI3Ilx1DGd8ICHCBxIifCAhwgcSInwgIcIHEiJ8ICHCBxIifCAhwgcSInwgIcIHEiJ8ICHCBxIifCAhwgcSInwgIcIHEiJ8ICHCBxIifCChInfZtb1P0t6ah39c0r8aX6KcXttXYuduWCj7fiIijqk6qEj4nbA9GhEjrS7RgV7bV2Lnbui1fXmqDyRE+EBCCyH8TW0v0KFe21di527oqX1b/x4fQPcthDM+gC4jfCAhwgcSInwgIcIHEvo3AePmFkcIa+IAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 288x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "row_sums = conf_mx.sum(axis = 1, keepdims = True)\n",
    "norm_conf_mx = conf_mx/row_sums\n",
    "np.fill_diagonal(norm_conf_mx, 0)\n",
    "plt.matshow(norm_conf_mx, cmap = plt.cm.gray)\n",
    "plt.show()   # it seems like 8-9, 3-5 are easily getting misclassified"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_digits(instances, images_per_row=10, **options):\n",
    "    size = 28\n",
    "    images_per_row = min(len(instances), images_per_row)\n",
    "    images = [instance.reshape(size,size) for instance in instances]\n",
    "    n_rows = (len(instances) - 1) // images_per_row + 1\n",
    "    row_images = []\n",
    "    n_empty = n_rows * images_per_row - len(instances)\n",
    "    images.append(np.zeros((size, size * n_empty)))\n",
    "    for row in range(n_rows):\n",
    "        rimages = images[row * images_per_row : (row + 1) * images_per_row]\n",
    "        row_images.append(np.concatenate(rimages, axis=1))\n",
    "    image = np.concatenate(row_images, axis=0)\n",
    "    plt.imshow(image, cmap = mpl.cm.binary, **options)\n",
    "    plt.axis(\"off\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x576 with 4 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "cl_a, cl_b = 3, 5\n",
    "X_aa = X_train[(y_train == cl_a) & (y_train_pred == cl_a)]\n",
    "X_ab = X_train[(y_train == cl_a) & (y_train_pred == cl_b)]\n",
    "X_bb = X_train[(y_train == cl_b) & (y_train_pred == cl_b)]\n",
    "X_ba = X_train[(y_train == cl_b) & (y_train_pred == cl_a)]\n",
    "\n",
    "\n",
    "plt.figure(figsize = (8,8))\n",
    "plt.subplot(221)\n",
    "plot_digits(X_aa[:25], images_per_row = 5)\n",
    "plt.subplot(222)\n",
    "plot_digits(X_ab[:25], images_per_row = 5)\n",
    "plt.subplot(223)\n",
    "plot_digits(X_ba[:25], images_per_row = 5)\n",
    "plt.subplot(224)\n",
    "plot_digits(X_bb[:25], images_per_row = 5)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "KNeighborsClassifier(algorithm='auto', leaf_size=30, metric='minkowski',\n",
       "           metric_params=None, n_jobs=None, n_neighbors=5, p=2,\n",
       "           weights='uniform')"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# ultilabel classification\n",
    "\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "\n",
    "y_train_large = (y_train >= 7)\n",
    "y_train_odd = (y_train%2 == 1)\n",
    "y_multilabel = np.c_[y_train_large, y_train_odd]\n",
    "\n",
    "knn_clf = KNeighborsClassifier()\n",
    "knn_clf.fit(X_train, y_multilabel)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[False,  True]])"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "knn_clf.predict([some_digit])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# There are many ways to evaluate a multilabel classifier, and selecting the right metric really depends on your project. For example, one approach is to measure the F1 score for each individual label (or any other binary classifier metric discussed earlier), then simply compute the average score."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train_knn_pred = cross_val_predict(knn_clf, X_train, y_train, cv = 3)\n",
    "f1_score(y_train, y_train_knn_pred, average = 'macro') # average = 'macro' --> assumes all labels are equally important, otherwise average = 'weighted'\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "## multiouput-multiclass classification a.k.a multioutput classification\n",
    "\n",
    "\n",
    "# add some noise to the pixel\n",
    "noise = np.random.randint(0, 100, (len(X_train), 784))\n",
    "X_train_mod = X_train + noise\n",
    "noise = np.random.randint(0, 100, (len(X_test), 784))\n",
    "X_tes_mod = X_test + noise\n",
    "y_train_mod = X_train\n",
    "y_test_mode = X_test\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_digit(X_train_mod[30000])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "knn_clf.fit(X_train_mod, y_train_mod)\n",
    "clean_digit = knn_clf.predict([X_train_mod[30000]])\n",
    "plot_digit(clean_digit)\n"
   ]
  }
 ],
 "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}