Week10 - Boosting.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Boosting techniques"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We will implement,\n",
    "* AdaBoost\n",
    "* GradientBoosting\n",
    "* XGBoost\n",
    "\n",
    "for MNIST digit recognition classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from keras.datasets import mnist\n",
    "\n",
    "from sklearn.ensemble import AdaBoostClassifier\n",
    "from sklearn.ensemble import GradientBoostingClassifier\n",
    "\n",
    "from sklearn.model_selection import train_test_split, cross_validate, cross_val_score, ShuffleSplit\n",
    "from sklearn.metrics import confusion_matrix, ConfusionMatrixDisplay, classification_report\n",
    "\n",
    "from sklearn.pipeline import Pipeline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Loading MNIST data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "(X_train, y_train), (X_test, y_test) = mnist.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of training data (60000, 28, 28)\n",
      "Shape of test data (10000, 28, 28)\n",
      "Shape of training labels (60000,)\n",
      "Shape of test labels (10000,)\n"
     ]
    }
   ],
   "source": [
    "print(\"Shape of training data\", X_train.shape)\n",
    "print(\"Shape of test data\", X_test.shape)\n",
    "\n",
    "print(\"Shape of training labels\", y_train.shape)\n",
    "print(\"Shape of test labels\", y_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Flattening \n",
    "X_train = X_train.reshape(X_train.shape[0], 28*28)\n",
    "X_test = X_test.reshape(X_test.shape[0], 28*28)\n",
    "\n",
    " #Normalising\n",
    "X_train = X_train/255\n",
    "X_test = X_test/255"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of training data (60000, 784)\n",
      "Shape of test data (10000, 784)\n",
      "Shape of training labels (60000,)\n",
      "Shape of test labels (10000,)\n"
     ]
    }
   ],
   "source": [
    "print(\"Shape of training data\", X_train.shape)\n",
    "print(\"Shape of test data\", X_test.shape)\n",
    "\n",
    "print(\"Shape of training labels\", y_train.shape)\n",
    "print(\"Shape of test labels\", y_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "cv = ShuffleSplit(n_splits=10, test_size=0.2, random_state=42)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Some helper functions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "def train_classifiers(estimator, X_train, y_train, cv, name):\n",
    "    estimator.fit(X_train, y_train)\n",
    "    cv_train_score = cross_val_score(estimator, X_train, y_train, cv = cv, scoring='f1_macro')\n",
    "\n",
    "    print(f\"On an average, {name} model has f1 score of \" \n",
    "            f\"{cv_train_score.mean():.3f} +/- {cv_train_score.std():.3f} on the training set.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "def eval(estimator, X_test, y_test):\n",
    "    y_pred = estimator.predict(X_test)\n",
    "\n",
    "    print(\"# Classification report\")\n",
    "    print(classification_report(y_test, y_pred))\n",
    "\n",
    "    print(\"# Confusion matrix\")\n",
    "    disp = ConfusionMatrixDisplay(confusion_matrix=confusion_matrix(y_test, y_pred))\n",
    "    disp.plot()\n",
    "    plt.title('Confusion matrix')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Adaboost for MNIST multiclass classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "On an average, AdaBoostClassifier model has f1 score of 0.712 +/- 0.016 on the training set.\n"
     ]
    }
   ],
   "source": [
    "abc_pipeline = Pipeline([(\"classifier\", AdaBoostClassifier())])\n",
    "train_classifiers(abc_pipeline, X_train, y_train.ravel(), cv, \"AdaBoostClassifier\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# Classification report\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.84      0.90      0.87       980\n",
      "           1       0.85      0.94      0.90      1135\n",
      "           2       0.79      0.58      0.67      1032\n",
      "           3       0.68      0.67      0.68      1010\n",
      "           4       0.72      0.72      0.72       982\n",
      "           5       0.69      0.59      0.64       892\n",
      "           6       0.72      0.86      0.78       958\n",
      "           7       0.68      0.78      0.73      1028\n",
      "           8       0.66      0.68      0.67       974\n",
      "           9       0.62      0.55      0.58      1009\n",
      "\n",
      "    accuracy                           0.73     10000\n",
      "   macro avg       0.73      0.73      0.72     10000\n",
      "weighted avg       0.73      0.73      0.73     10000\n",
      "\n",
      "# Confusion matrix\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAT4AAAEWCAYAAAD/x/trAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/YYfK9AAAACXBIWXMAAAsTAAALEwEAmpwYAABre0lEQVR4nO2dd3gU1feH37MlvRcghC5F6SgdRVQEVGxfFSvYewVRsXd/FgQpNuwVFRuISlFRioCgIL2EAAkQ0ntI2d37+2M2BUjZ3exCYu77PPMkOztz5u7d2TO3nHs+opRCo9FomhKm410AjUajOdZox6fRaJoc2vFpNJomh3Z8Go2myaEdn0ajaXJox6fRaJoc2vH9xxCRQBH5QURyRWROPexcLSKLvFm244WInCYi2493OTQNB9FxfMcHEbkKmACcCOQD64HnlVLL62l3LHA3MFgpZatvORs6IqKATkqphONdFk3jQbf4jgMiMgF4DXgBaA60Ad4ALvSC+bbAjqbg9FxBRCzHuwyaBohSSm/HcAPCgQLgslqO8cdwjAec22uAv/O9YcA+4H4gDUgBrne+9zRQCpQ5r3Ej8BTwaRXb7QAFWJyvrwMSMVqdu4Grq+xfXuW8wcAaINf5d3CV934HngVWOO0sAmJq+Gzl5X+wSvkvAs4FdgBZwCNVju8PrARynMfOBPyc7y11fpZC5+e9vIr9h4CDwCfl+5znnOC8xsnO1y2BdGDY8b439HbstuNegKa2AaMAW7njqeGYZ4BVQDMgFvgTeNb53jDn+c8AVqfDKAIine8f6ehqdHxAMJAHdHG+Fwd0c/5f4fiAKCAbGOs870rn62jn+78Du4DOQKDz9Ys1fLby8j/hLP/NTsfzORAKdAMOAe2dx58CDHRetx2wFbivij0FdKzG/ksYD5DAqo7PeczNwBYgCFgITD7e94Xeju2mu7rHnmggQ9XeFb0aeEYplaaUSsdoyY2t8n6Z8/0ypdRPGK2dLh6WxwF0F5FApVSKUmpzNcecB+xUSn2ilLIppWYD24DzqxzzgVJqh1LqEPAV0LuWa5ZhjGeWAV8AMcA0pVS+8/pbgF4ASqm/lVKrnNfdA7wNnO7CZ3pSKVXiLM9hKKXeARKA1RjO/tE67Gn+Y2jHd+zJBGLqGHtqCeyt8nqvc1+FjSMcZxEQ4m5BlFKFGN3D24AUEflRRE50oTzlZYqv8vqgG+XJVErZnf+XO6bUKu8fKj9fRDqLyHwROSgieRjjojG12AZIV0oV13HMO0B3YIZSqqSOYzX/MbTjO/asBEowxrVq4gDGJEU5bZz7PKEQo0tXTouqbyqlFiqlzsZo+WzDcAh1lae8TPs9LJM7vIlRrk5KqTDgEUDqOKfWUAURCcEYN30PeEpEorxQTk0jQju+Y4xSKhdjfOt1EblIRIJExCoi54jIy87DZgOPiUisiMQ4j//Uw0uuB4aKSBsRCQceLn9DRJqLyIUiEozhjAswuolH8hPQWUSuEhGLiFwOdAXme1gmdwjFGIcscLZGbz/i/VSgg5s2pwFrlVI3AT8Cb9W7lJpGhXZ8xwGl1KsYMXyPYQzsJwN3Ad87D3kOWAtsADYC/zj3eXKtxcCXTlt/c7izMjnLcQBjpvN0jnYsKKUygdEYM8mZGDOyo5VSGZ6UyU0mAldhzBa/g/FZqvIU8JGI5IjImLqMiciFGBNM5Z9zAnCyiFzttRJrGjw6gFmj0TQ5dItPo9E0ObTj02g0TQ7t+DQaTZNDOz6NRtPkaFALuIMj/VRUfKDX7eZuNnvdJgBSVziZBzSyySbxRR1QRyBevQw3rvr1NsUUUqpK6vWljTwjWGVm2es+EPh7Q8lCpdSo+lzPFzQoxxcVH8i9Xw30ut2F3cO8bhNA/P29blOVNK5FBKaAAJ/Y9VW0QaOqXx88VFY7fqm3jYwsO6sXtnLpWGvcrrpW2RwXGpTj02g0jQGFXVUX59540I5Po9G4hQIcvhuMOCZox6fRaNzGUe3KxsaDdnwajcYtFIoy3dXVaDRNCQXYdVfXN+z52I9931gRgZBODro/d4icdWa2vxoADjAHKbo/f4jgNorkL60kfeGHmIz93Z4qJuQE955IfYflcduzBzCbFD/PjuKrmc3dLnNMXAkPvJpIREwZKOGn2bHM/bAF19y7j1FXpJObZQXgw1daseb3CLftA1j9Hbz6bQJWP4XZolj2YwSfTG5R94kuYjIpZizYQWaKlSeudTfpiUFMXAkTJ+8iMqYMpYSfv2jG3A9b0OGkQu5+bjdWf4XdLrz+eDt2bHAtjWBNdTtuwj4GnZ2NwyHkZFp4dWIHstL8PCq3r+rWl9/ZxTencc6VWSgFu7cF8OqENpSV+D48V4/x1YKIjMJIAWQG3lVKvejKecWpQtJnfgyZW4A5ANbfH8jBn60kvuNHn+mHCDnBQdIXVhLf9qfH88XEnVdG68vLAEhbYmHbywH0fbvI5XKaTIo7X9jPw1d0ICPFyoyfdrJqYThJO90L1XDYhHeeb0PC5mACg+3M+GET65aHA/Dd+y345p04t+xVR1mJ8OBlJ1BcZMZsUUz5PoE1v4Wy7Z/getsGuOimDJJ3BhAU4lqcVnXYbcI7L7Rll7Meps/bxLrlYdw4KYnPprdi7R8R9BuWw42Tknjoqq4u2aypbr+eFcfHU4zQiguvO8jV9+xnxmPtPSq3r+rWV3ajW5Ry0Q0Z3HzGiZQWm3j0rT0MuzCbxV9F18tuXSjA3sjjIX32aBARM/A6cA5G7rYrRcS1uxxQNrCXgMMGjkPgH+sAAVuhEdtkyxf8Y43Kt1RpNNgPuR/+1KVPEQf2+HEwyR9bmYnf50YwaGSue0aArHQ/EjYbN/OhQjPJCYFEtyh1207tCMVFRkC2xaowW5XXYnJj4krpf1YeP39ev7yc2el+7DqsHgKIbmG0/sodalCojUw3WmY11W1RQWVwekCgA6XqE/vmq7r13Xdmtij8AxyYzAr/QAeZB63eMVwHDhe3hoovW3z9gQSlVCKAiHyBIZ+4pa4TA5or2l1XytLhoZgCFDGD7cQMsdP96WL+uT0QUwBYghUDPy+sOCdptpU9H/mjyqDv+6639gCiW5SRfqDyR5iRYuXEk92zcSTN40s4oWsR29eH0O2UfC4Yl8rw/2WwY0Mw7zzfhoI8z6veZFLMXLiDlu1K+eHDaLav805r77anD/Duc3EEhXjvlm0WX8IJ3YrYvj6Yt59ty3MfbeOmh5MQk+L+S7t5ZLNq3QJcOzGZ4RdnUphv5qGrqsuc7zq+qltf2M086MfXbzXjk7+2UFIs/PNHGP8s9U2wflUUqtGP8flyMCAeI8FmOfs4XKMBABG5RUTWisjagiyjdVSWa3RZhy4sYNhvBdgPwYEfrOz52I+T3zzEsF8LiL+ojG0vV3ZF21xZxtAFBXSaUEzi256N8XiLgCA7j725k7efbUNRgZn5nzXn+tN7cce53clKt3Lzo0n1su9wCHec3YWrT+lKl95FtO1ylJ6O2wwYnkdOhoWEjUF1H+wiAUF2HntjB28/25aiAgvnXZ3KrOfaMu7UPsx6ri33vZTomc0qdQvw0eTWjB3SmyVzozl/XGodFmrHF3XrK7sh4TYGjczl2oFduerk7gQE2Tnzf1leKG3tKAVlLm51ISLvi0iaiGyqsi9KRBaLyE7n30jnfhGR6SKSICIbROTkKudc6zx+p4hcW9d1j3uSAqXULKVUX6VU35Aow2FlrrIQGO/AL0phskKzs2zkrDOTv91MRE+jq9TinDJy1h+9BjfuHBtpv7nX3M88aCW2ZWWXNCaujIwUz7oMZouDx9/cyZK50axYaHQZczKsOByCUsKC2c3o0quwDiuuUZhn5t8/Q+h3Rn69bXXtV8jAEXl8tHoLD7+5l16nFvDgjCP1hVzHbHHw2Bs7WTIvhj+d9TD8kgxWLIgEYNlPUXTpWeC2zSPrtiq/zY3m1FHZHpe5Kt6sW1/Z7XNaAQeT/MjNsmC3CSt+jqBrX+/cW7Uj2F3cXOBDjIzYVZkE/KqU6gT86nwNxrBZJ+d2C4YeC07NlCeBARg9zSfLnWVN+NLx7QdaV3ndChfFaQLiFDkbzNgPGU+XrNVmgk+wYyuAwj1GkTP/tBDcweiSFe6t/BjpSy0EtXGvq7Z9fRDx7Utp3roEi9XBsAtzWLUo3C0bBorxL+0mKSGQb9+rnMiIiq10qoNHZrNnh+eJGMKjbASHGc7fL8DByUMLSE6o/3rZD/4vjmv6duXaAV35v9vb8u/yEF6++0h9IVdR3PfibpJ3BfJdlXrITLXSY4Dxg+89OI/9e9wpd/V127JdpZjaoLOzSU70vC58Vbe+spu238pJJxfhH+AAFL1PzXd7Qs4TFOBQrm112lJqKYbsQVUuBD5y/v8RlcJcFwIfK4NVQISIxAEjgcVKqSylVDawmKOd6WH4coxvDdBJRNpjOLwrMLQT6iSip50WZ9tYOSYYMUPoiXZaX1ZGQHPF+vGBIGANU3R/1uguJH1uJXOVBZMFLGGKHi+4141w2IXXH43nhc8TMZlh0RdR7N3h/g3UrW8Bw/+Xye5tgbz+o9Fy//CVVgy7IJMOJxljhqn7/Jn+SDu3bZcT1byMidOSMJnAZIKlP4Sz+hffj+u4g1EPGezeFsjM+RsBozs6/ZEO3Pr4HswWKC0Rpj/qerhMTXU7ckw6rToUoxSk7vdnxqPtPC63r+rWV3a3rwtm2Y/hvL5wO3abkLA5kJ8/8+2MbjkutuY8pblSKsX5/0GgPLaspuEzl4bVquJTzQ0RORdDxs8MvK+Uer6241t3D1c6O0sjyh6Czs7iU3yUnSVPZdXLcLeefuqLH5u5dGzPNvv3AlVFqWYppWZVPUZE2gHzlVLdna9zlFIRVd7PVkpFish84EWl1HLn/l+Bh4BhQIBS6jnn/seBQ0qpyTWVy6dxfEqpnzCkCTUazX8EBZQpl0fJMpRSfd28RKqIxCmlUpxd2TTn/pqGz/ZjOL+q+3+v7QLHfXJDo9E0LhSCHZNLm4fMA8pnZq8F5lbZP845uzsQyHV2iRcCI0Qk0jmpMcK5r0Ya7JI1jUbTcHHUK1C8EhGZjdFaixGRfRizsy8CX4nIjcBeoFwv+SfgXCABKAKuB1BKZYnIsxjzCgDPKKVqjevRjk+j0biFkaTAO45PKXVlDW+dVc2xCrizBjvvA++7el3t+DQajZsIdtfH+Bok2vFpNBq3MDIwa8fnNXI3m30SerLwwHqv2wQY2bK39436SLXMV+pijuLiug9qQPgiBMlXNNTQG6WEUuUj5cJjRINyfBqNpnHg8G0As8/Rjk+j0biFMbmhu7oajaZJoSc3NBpNE0NPbmg0miaJ3UsBzMcL7fg0Go1bKIQy1bhdR6MrvSdqaGN6dCMixsasJdsByMs288Jt7Ujd50fzVqU8+vYeQiPszHkjlt++NRJc2u2QvDOALzduIizSzpolobz1eDx2h3DOlZlcfndabZcEfKeu5QtlLV+VdcKUJAYMzycnw8KtZ3apt72qeEsRzhfKeDXZPe3cLK65dz+tOx7i3ou6snOjaypzNeENdUB30ZMbtSAi7wOjgbTydDP1xVM1tOc/S+SVe9tUvP5qZjP6nJrP5Xen8eWMZnw5sxk3PZbCZXekc9kd6QCsWhTGt+/EEhZpx26H1x9pxf99sYuYuDLuPrczA10QI/KFupavlLV8pQS26Mso5n0QwwPTkus+2E28oQjnK2W8muzu2R7Is7d35J7n93hc5nK8pQ7oLgpp9F1dX7rtD6kjC6q7eKqGFhp5+A9j5cJwho8x1jAPH5PFygVHZ1te8n0kwy4y0phvXxdEy3YlxLUtxeqnGHZhNisXupKh2TfqWr5R1vJNWTetDiE/2/vPV28pwvlKGa8mu8m7AtmX6HkG7qp4Sx3QExyYXNoaKj4rWQ0ppetFdWpoMXFlbtvJzrAS3dwGQFQzG9kZhzuO4iJh7e+hnHqucRMZmhyV13FHk8NkUryxeDtfbtjMuqUh9VbXqqqsNXvdJgrzzF5T1vJ2WX1JuSKccniv5XGketsF41J58+eNjH8pkZAwm9fsegtv/R7cRSmwK5NLW0PluJesqspaGcd+iY4IiBzetFm1OJxufQsJi/S8C1WOt9W1fKms5SuFMW/jM0U4HyjjVacK19gxJjfMLm0NlePu+KqqrFmpfR2lt9TQImPKyEw1ul+ZqRYiog9/mv8xN6KimwvlT9bK63jyZPWWutaxUNbylcKYt/CFIpwvlPHqUoWrL95UB3QXHyci9TkNt2TV4C01tIEj8vjlK+NG/OWrqMPGRQrzTGxYFcLgUXkV+7r0LmL/bn8OJvlRVir8PjeSgSPyjrJ7JL5Q1/KVspavlMB8gbcV4XyjjFe9XW/iPXVA91AIDuXa1lBpVOEsnqqhjT+/E7lZFq4+pStj7z/I5Xel8vxt7VjwRTTN4o1wlnJW/BzBKUPzCQiqlKg0W+DO5/fxyFUdcNiFEVdk0a5L3VlJfKGu5StlLV8pgU16Yy89BxUQHmXj07Vb+OTV5iycfWyUwFzBV8p4Ndm1+jm4/am9hEfZeOb9HSRuCeLRa0/0qOzeUgf0hIbcmnMFn6msVU0pDaQCTyql3qvtnDCJUgPkqMSr9UanpcJnaakaG009LdVq9Wu9VdZadw9TE+a4poY4oevivz0QG/I5Pmvx1ZJSWqPRNGrE17q6PqdRdXU1Gs3xx5CXbLgztq6gHZ9Go3ELpQRHA47RcwXt+DQajds05OBkV9COT6PRuIWRj0+P8Wk0miaFzsDsVcRkwhTovWVI5ZzTaYjXbQLsePckr9s88Z4tXrcJoMo8X2taq12bb9aG2k/v4xO7lpWbvW7THBnhdZsAttS6U5+5jReimhQ06OBkV2hQjk+j0TR8ytfqNma049NoNG7TkFNOuYJ2fBqNxi2MtFS6q6vRaJoYeoxPo9E0KYzsLLqrq9FomhDGkjXt+HyK1c/BK7M3Vah/LV8QzafTWtO8VTGTXttJWGQZOzeFMHliR2xlrn8ZVj8Hr3y+Caufo9Lu9DZMeGknPfrlUejMljvloU4kbq09BXv7hzbgCDCjTIBJSHq8K37JRTT/ZC+mEgdl0X4cvLkDjkDDZsV7xXYQIemxk1DWmsteUx08+OpOOvUowGYTdvwbwvTHO2C3uV4H41/ZzYAzc8jJtHLbCEMPqkPXIu5+fg9+/g7sdmHmY23Z8W/9UqZ7qgoXG1XAQ7cvIzL8EArhx986892CboQGl/DYPb/TPDaf1PRQnp0+jIJCf4IDS5l051KaRRdgNivm/NidhX90qvUavlJZi29byKQXN1S8bhFfxKdvdSQ4tIyRF+8nL9tIGf/RzI6sXRHrst0j+WjVZg4VmHE4wG4T7j7Xu0p21aNbfDUiIq2Bj4HmGA+JWUqpae7aKSsVJo3t5lT/cjD5i82s/SOCi29I4fsP4vjjxxjueiaRkZel8ePnrsshlpUKk8ZVtbuJtUsjAXjv5bYsXxDjVjmTJ3bGEVqZ/bbFR3tIv6w1h7qEErY8g8iFB8m8KB7sirh3d5NyU3tKWwdhKrChzLWPl9RUB0vmxfDy/R0BeGjqTkaNca8OFs+J4YePmjFxyu6KfTc+nMxn01qy9vcI+p2Rw00P7+PBKzzLFwf1U4WzO0y89Vk/EvbEEBhQxpvPz+PvjfGMHLqTdZvi+OKHkVxx/gauOH8D737RjwtGbGXvvnAenzyc8NBiPnj1G35d3gGbvebQC1+prO3fG8zdVw4CDC2Tjxf8wZ9LmnH2BfuZ+1lbvv2knUd2q+PByzqS5wNBp9rw1soNERkP3IThIzYC1wNxwBdANPA3MFYpVSoi/hg+5RQgE7hcKbXHk+v60m3bgPuVUl2BgcCdItLVfTNV1L8sCotT/avXwFyWLTB+PL98F8ugs93VnTjCrsU7qmLlWFNLONTZaCkVdQ0j5G8jlX3w5lxKWgVS2toI1HaEWMBU101UfR2s+SMSEEDYviGEGDfVwTb9FUp+zhE/GEWFXGNwqJ3MtPqnMvdUFS4rJ4iEPcYD6FCxlaT94cREFjL4lCQWLTMc/qJlHRnS16mHoYSgQBugCAwoI7/AH7uj9lvcVyprVenVP5OUfUGkp3hHXe14Uz6r68pWGyISD9wD9HVK0JqBK4CXgKlKqY5ANnCj85QbgWzn/qnO4zzClyprKUqpf5z/5wNbgXhPbJlMipnz/mX26rWsWx5OSlIAhflmHHajYjMO+hHd3P2b1bC7ntmr1rBuRTjb/w0F4NrxSbzxw3pueWQ3Vj9HHVYAgVZTd9LmmS2E/2Ho8pa2DCB4fQ4AIWuzsGYZ5bOmloBA/NQdtHlmC5E/p7hR1so6KC8rGNoOZ12UwdqlEa5/+Bp465k23PTIPj5ZuZ6bHk3mg5da1cuet1Thmsfk07FdFtt2xRIZXkxWjvHgyMoJJDLcyIb9/aKTaNMyhy9f/5J3XvqeNz4egHJj9tFXKmtDRx7kj4WVLfHRlycx88s/uffJTYSE1nPlixJemL2LmT9v55yrM+pnyw0cyuTS5gIWIFBELEAQkAKcCXztfP8j4CLn/xc6X+N8/ywRzzL3HpOOuoi0A/oAq6t5r0JlrVRVn87d4RDuuqAXY089hc69CmjdwTvqX4bd3ow9rS+dexbQtlMhH0xuy80j+3DvJT0JDbdx2S3767ST/NCJJD3Rlf33dSJiSRqBO/I5eF07Ipak0+aZLZiKHSiL8f2IQxGYUEDKTe1JfqgLIetyCNxat37HkXXQtlNRxXt3Pr2bTX+Fsnlt/VPFj74mjbefbc3YQb15+5k2jH95T73seUMVLsC/jCfHL+GNT/pTdMjviHelYhVW35772bU3isvvvJxbH76Qu65bRVCgaw9EX6msWSwOBgxNZ/ni5gD8NKc1N11wGndfMYjsDH9unLDdI7vlTLi4I3eN6sKj13Tggusy6D6goF72XMFNzY2Y8t+3c7ulwo5S+4HJQBKGw8vF6NrmKKXKnzT7qGwwxQPJznNtzuM90jHwueMTkRDgG+A+pdRRv/CqKmt+UrteQGG+hQ2rwjixTz7BoXZMZuOWj2lRSmbqkT8I1ynMt7BhdTh9h+aQne4HCGWlJhZ904zOPeu+kWyRxrXtYVYK+kQQsLuQsrhA9k/oTNITXcnvH0VZrH/FsUWdQnGEWlH+Zgp7hBOwt6g280eXdVUYfYfmAHDV3cmER5Ux64V27n7sahl+SSYrfjbGOpf9GEnnXvX7IdVXFc5sdvDU+N/4dUUHlq9pB0B2bgBREUadRUUUkZNr3DejTt/JsjVtAeFAahgH00No3bJugW1fqawB9B2Swa5tYeRkGd9/TpZ/pd1vW9G5W/0EwDMPGvdebqaVFT+Hc2Jv1+8lT1GATZlc2oCM8t+3c5tVbkdEIjFace2BlkAwMMrnHwAfOz4RsWI4vc+UUt96YiM8qozgUMP5+/nb6TMkl+RdQWxYHcZpozIBGH5xOit/cU++7yi7g3NITgwkskJdSzH47Cz27qg9aYKU2JFie8X/QVvyKIkPxJzn7MI4FNE/ppAzrBkAhd3C8N9/CCmxg10RuCOf0pa1O/xq6yAxkJFjUjnltBxeuq+TW1262shMs9JzoCEr2XtIPgf2HE9VOMXEW5azd38E3/zUvWLvyn/aMOK0BABGnJbAn3+3Ma6VGczJ3Y2hg4iwQ7SOyyMlLfRos0dcwzcqawZDRx3ezY2MqdTRGHxmGnt31VW+mvEPtBMYbK/4/5TT89mz/diIDXmpqzsc2K2USldKlQHfAkOACGfXF6AVUN7t2g+0BnC+H44xyeE2vpzVFeA9YKtSaoqndiJjS5n4SgImE4hJseynaP5aEklSQiCTXtvBuAlJ7NoSzKI5zdy3+3ICJpMy7P4cw19Lovi/jzcRHmVDRJG4NZgZT5xQqx1Lno2Wrxs/QhyK/P5RFHUPJ+KXVCKWGNk1CvpEkjfEaJE7gi1kn92cNs9vBaCwRziFPSM8qoP521aSdsCfKXMMFa8/F0Xx+czWLtfBpOm76Dkon7BIG5+sWs+nU+OZ9lA7bnsqCbNZUVpiYtqkdi7bq476qMJ175LG2aftIjEpkrdemAvA+1+dzBfzevDYPb8z6owdpGWE8Oy0MwD49NvePHDbMt558TsQeGd2X/Lya3cEvlJZA/APsNFnQCYzn6/M4nPDvTvo0DkfBaQdCGTG8x7M9zmJjLXx5HvGjLzZDEu+j2Dt7/Uf7qgT70lHJgEDRSQIOAScBawFlgCXYszsXgvMdR4/z/l6pfP935SHamm+VFk7FViGMUVdPkPwiFLqp5rOCTfHqIGB5/miMN63CWybptNS6bRUjSst1WrHL/VWWYs8sZk68/1LXTr22yFv1qqyJiJPA5djRIGswwhticdwelHOfdcopUpEJAD4BGO+IAu4QimV6Mln8KXK2nJo5GlaNRpNtXhrra5S6kngySN2JwL9qzm2GLjMG9dt8Cs3NBpNw0InItVoNE0OhWCrIzC8oaMdn0ajcRstNqTRaJoWSnd1vYtS4HBhiZi7mHzTLO9y+4a6D3KTfV919LpNgJYX+2a22NLW9fAZdzD9tc0ndjE3Hq0I8fM8KL9GSurvsPQYn0ajaZJox6fRaJoUCqkz601DRzs+jUbjNnpyQ6PRNCmUntzQaDRNEW8lxTheaMen0WjcxGtJCo4bDd7xxcSVMHHyLiJjylBK+PmLZsz9sAUdTirk7ud2Y/VX2O3C64+3Y8cG10VxjqWAT/uTirjnhT0EBDlI3efHy/eeQFFB3WEVUmgn4vUDWJKMVEY5d7UkZH4Wlv0lzvcdqGAT6VNPAJsi4vUDWBOLwa44dEYEBZe4rhsS27KUB6YlERFrAwU/fRrN9+95JoJzwZhERl6QjIhi4bw2zP2yAzfctYX+p6ZiKzORsj+I157rTWGB62ntfSo6VY3d+/4vgU7dCxGB/XsCePXBjhXp/12hJrGhuZ+3BeDia/Zw04QdXHnmMPJyXA9bqUkc6aaHkxhwVg62MuHA3gCmPNCewnzf/bwbe4vPl9lZAoClgD+Gg/3auSC5RsJN0WpgwLmH7YuMLSWqWRm7nGIw0+dt4tlbO3Hr43v57v041v4RQb9hOVx6ywEeuqqGFD/VxvEpAoIchwn4vP1cO0LDbaz5IwIwBHw2rQmrUcCnuown3fvnU1xkYuKU3RWOb/q8zbzzfGs2rg5jxJh0WrQu4eNXq0/pvr9KHF/EtP2Udg2i6OxIKFNIqQMVXPnjC/vgII4gMwWXxxK4NJeANflk398KKXEQe/cuMp9ri72Z8aOqK44vqlkZUc3LSNgYRGCwnZkLdvD0De3rzJ13ZBxf2w55PPjMOibceCplNuHZqX8x8+UetGhZxL9/R+Owm7j+DiMl1wdv1JzdxpF+ZJq16r+vi29I4c+FURWiU7u3BbkluFST3aSEQIoKDMdx8yN7yMm0Muft6pUTTGG159QrFxsaf+0A0lMCiWlezD1PbKZ1u0LuvXpgjY7Pnp1z1L4o5++hqjjSM7d0JiaulPV/huGwCzc8lAzA+y8dHWO5quRn8hyZ9fJawZ3iVNfp17t07Npz/6/W7CzHC1/OSZcAZyqlegG9gVEiMtBdI9npfuw6TAwmgOgWRuuvXBQnKNRGZpq7wZ7HTsAnvn0JG1cbP45/loUx5JzsuktXaMdvSxFFwyOMHVY5zOmhFIEr8jh0Wlj5x0GKHWBXUOIAi1TIWbpCVpqVhI1G0tXyeo6Jcz/lVOt2BezYEkFJiRmH3cTGdVEMPj2FdX/F4rAbt9u2zRFEN3NXPuAYiU457ZY7PVD4+zugHu2DI8WGbr5/Gx+81tkjcauaxJH+WRZeoUGzbV2w2/etuzgQl7aGii/TUimgPG+51bnVq3nZLL6EE7oVsX19MG8/25bnPtrGTQ8nISbF/Zd2c9ueyaSY/v0GWrYtZv6nLaoV8Hnr2Xb1KTIAe3cGMGhEDisXRTL0vGxi4+q+Kc1pZTjCzETMOIB1TwmlJwSQd2MLVIDhPPy2FGGPsGBvaaQ0PzQojIC/8ml+ww6kxEHeDS1QoZ6tUmjeqpQTuh9i2z+1Z5+ujr27Qhl363ZCw0opLTHTd1AaCdsiDjvm7NHJLPulpdu2j/y+vCk6Vd19MP7FBPoNyyEpIZB3/q+t23bLqSo2NPD0NDLTAti90/PMy+UcKY5UzogxGSyd715GcndQNP6urq9Tz5tFZD2QBixWStUuNkTJUTbKCQiy89gbO3j72bYUFVg47+pUZj3XlnGn9mHWc2257yX38xEeKwGfKQ+0Z/TYNGbM30xgsB1bWd03jdgV1sRiCkdFkj6lA8rfRMi3lSpagcvyOHRaeMVrv52HUCZIfa8zaW91ImRuJuaD7juBgCA7j7+7h7eeaOnSOOSRJO8N5etPT+C5aat5ZupqEneGY3dUft7Lr92J3S4sWei+4J5vRaeOvg+mTurINYNPIXlXIEPP8yjD+WFiQ/4BdsbckMinb9We1dsVjhRHKueKOw9gtwm/fe+RBo+LuCU21CDxqeNTStmVUr0x8ub3F5Hu1RxTKTaEf7V2zBYHj72xkyXzYvjTKQYz/JIMVixwiuL8FEUXF0SBasLXAj77dgXy6Ngu3D26G7/PiyZlb926CPZoK/ZoK2WdjVZX8eBQY+ICwK4IWJXHoSGVTjlwaS4lfUKMLm6EhdITg7Ducs8xmC2Kx9/dw2/fRrLi5wi3zq3Koh/acO/1p/HQHYMpyLdyIMnomg0/N5l+Q1KZ/OTJ1CdHrU9Fp6rcB2A4xT/mxzBkpGeOr6rYUItWRTSPP8TML1by/vylxDQrYdpnq4iMrvmBXx3ViSMBnH1JOgPOzObl+zrg6xzASrm2NVSOyboTpVQORh59DxSUFPe9uJvkXYF8V0UMJjPVSo8BTlGcwXnsd1MU51gK+IRHG2NlIoor7z7Aj5/VPVvqiLRgj7Fgds7g+m8oxNbKeDD4/1uILd4fR0zlrKg91or/RkMJTIodWHcUYYuv/kFSPYoJryaTvDOAb2d5NptbTnikUebY5ocYPCyF3xfFc8rANC65ZhfPPNiPkhL3W5LHTHRqSC77dgcQ17b8oaEYeFYW+xLrLza0NyGUq4efwQ2jh3LD6KFkpPlz79UDyc5073uqThzplKE5XHprCk/d3JmSYt8nYlBKXNoaKr4UG4oFypRSOSISCJyNB8rnhhhMBru3BTJz/kYAPprcmumPdODWx/dgtkBpiTD90Q5u2T2WAj4BQXbOH2foJ6xYEMmir1wLM8m9OY7IqfsRm8LW3I+cu41xscDluZWTGk4Kz4kiYsZ+Yu/ZBUpRdGYEtnauPwy69S9k+GXZJG4J4I3FhtbrB/8Xx5rf3O/qP/LC34SFl2KzCW9O7kFhgZXb7t+E1erg+WnGaMe2zRG8/nJPl236VHSqGruvfLGZoBAbIrB7azAzn2zvll2oXmyovtQkjnT7k3ux+ile+MT47ratC2bGY+6X2RWUotGv1fVlOEtPDNVzM0bL8iul1DO1nVNdOItX8FFaKl8I+OzXaamA6sJZGi51hbN4SnXhLPXFG+EsgR1bqvaTb6n7QGDrxU83yHAWX87qbsBQQ9JoNP8xGnI31hUa/MoNjUbTsFA07PE7V9COT6PRuE0DnrB1Ce34NBqNeyhQDt3i02g0TQzd1dVoNE2Ohhyc7Ao1Oj4RmUEtXXml1D3eLoxSCkdxsbfN+gyxel8FK/6yHV63CZD1Yyef2I26cLdP7CqHb35ZlrjmXrfpyMn1uk0Acwv3YhJdQQ7Uv63zX1irW1strD1mpdBoNI0HBfxXHZ9S6qOqr0UkSClVVNPxGo2m6dDYu7p1LmkQkUEisgXY5nzdS0Te8HnJNBpNA0VQDte2hoora7leA0YCmQBKqX+BoT4sk0ajaegoF7cGiksjnUqpZJHDvLfdN8XRaDQNHtX4JzdcafEli8hgQImIVUQmAlt9XC6NRtOQ8VKLT0QiRORrEdkmIludQ2tRIrJYRHY6/0Y6jxURmS4iCSKyQURO9rT4rrT4bgOmAfHAAWAhcKenF/QGJpNixoIdZKZYeeJa99JRVYc3FcaqU1l7eGYCrToYYTohYXYK8szcee5ROVlrsbmHAWflkpNp4bazD0+x/7+bU7nl8X2M6dWLvOy6v07TvlKCX0ypeG0+aOPQNVGUnhVG8IspmNJsOJpZKJwUZ6SuL7QTPDkVU3oZYofi/0VQenZ4LVeoubwXXJfG+ePScDiEv34L570XqhdcqokJk/cyYHguORkWbh1uCEuFRth45I3dNG9dSmqyH8/f3p6CXPdCNi66cjcjLtqHUkbOvKnP9CAqpoSHnl9PaHgZCdvCePWJXtjcUNuz+jl45fNNWP0clept09sw4aWd9OiXR6Eza/KUhzqRuDXYZbvVKdgBnH/pbs67dA8Ou7Dmz2Z88HoNwltew2stvmnAAqXUpSLiBwQBjwC/KqVeFJFJwCTgIeAcoJNzGwC86fzrNnXeIUqpDOBqT4yDkX4eIzRmv1JqtKd2qnLRTRkk7wyoEBuqL3abMOuZlocpjP2zNLROhbHqWDwnhh8+asbEKZXxbf93V2WqqZsfS6Iwz71EkYvnRBs2px4eMxcTV8opQ/NI3ed6PKGjlR/5M536EXZF+LjdlA0OIWBONmW9gigZE4X/V1kEzMnm0A0xBMzPxdHaj8InWyK5NsJu2UvpsDCw1nzjV1fenoPyGTQihztGdaWs1FSRnNUdFs2JYt6HsTzw2p6KfWPuPMi6FaF89XoLxtx5kMvvTOW9F1xPax8dW8z5l+/l9stPo7TEzKQX1nH6iBT6Dk7n+8/bsXRxS+6ctIkRFybz0zeu626UlQqTxnWrot62ibVLjYzh773cluULXJf+LKdthzxGXpB8mILdXyuaE9vsEAOHpnLX2KHYyswViWB9iqP+JkQkHGO+4DoApVQpUCoiFwLDnId9BPyO4fguBD526vmscrYW45RSKbiJK7O6HUTkBxFJF5E0EZkrIu40s+7Fi13jmLhS+p+Vx8+fe09MxVsKY1C9yloliqHnZfH7PPf0EAybRzvLW59M5t0X4j0eRLb8W4QjzoqjmRXrqgJKhxtJR0uHh2Fd5UzlL8AhByiFHFJGK7AOv11deUePTeerN1pQVmrccrmZrmvqVthdfbTdQSNy+WWOU2VtTjSDRua4bddsUfj52zGZHfgH2MnK8Kdnv0yW/2ZkTv71x3gGnp7mptUj1Nssqt4hIDUp2J37v73M+eQEbGXG9XKz3cno7AHlcXyubBBTrqnj3Kom8msPpAMfiMg6EXlXRIKB5lWc2UGgPOo8Hkiucv4+5z63caXt/jnwFRAHtATmALNdMS4irYDzgHc9KVx13Pb0Ad59Ls5nU+X1URiri+79C8jOsHLAzTT51THw7BwyD/qxe6vn5fRbWkDp6UYSTcmxo6IMh60izUiO0ZouHh2BObmU8LG7CbtzL0W3xILJ/bqPb19Mt/4FvDZ3Ky9/tZ3OPQs9LndVImNsZKUZTjQrzUJkjHvJYTPTA/j20/Z8+MPvfPrzbxQWWkjYGk5hvrVCDjMjLYDoZu6vKDKZFDPnrWf2qjWsWxFeod527fgk3vhhPbc8shurn+tNp727QunWK4vQsFL8/e30HZRGbPNi4lsX0q1XFlPeXc6Lb/xJp5Ny3C6ru7ihuZFRrqnj3GZVMWMBTgbeVEr1AQoxurVVrqN8Mj/siuMLUkp9opSyObdPAVd/ua8BD1JLw7iqylpZLSprAAOG55GTYalonXmb+iqM1cWwCzLdbu1Vh3+AgyvuSuHjV92XaKygTGFdXUDpqSFHv1dlBt/6TyH2Dv7kftKevBltCHorDYrcH2IwWxSh4Tbuu/BE3n2+FY+8kYj372dxu1UVElrGwKGp3HDh6Yw950wCAuycMjjdK6Ux1Nt6M/a0vnTuWUDbToV8MLktN4/sw72X9CQ03MZlt+x32V5NCnYmsyI0rIwJNw3h/ZknMem5v/F5LIl3Jjf2AfuqqC9+jeEIU0UkDsD5t7y5vR+omvK7lXOf29To+JwzK1HAzyIySUTaiUhbEXkQ+KkuwyIyGkhTSv1d23FVVdasNaisldO1XyEDR+Tx0eotPPzmXnqdWsCDM/bWVRSX8JbCWE2YzIoho7JZ+kP9u+hxbUto0bqUNxds4aMVG4mJK2XmT1uIjHW9e25dW4j9hABUpLOVF2FGsozWkmTZUBGG4/dfnEfZ4BAQwdHSD0dzK+Zk94cBMlL8nKp4wo5/g3EoCI+qf+r+7AwLUc2M8kQ1KyMn072Jjd79M0g9EERejj92u4k/l7TgpJ7ZBIeWYTIbz+uYZsVkpnneSi/Mt7BhdTh9h+aQne4HCGWlJhZ904zObqoDVqdgl5kewJ+/twCEHVsiUQ4hLMK3guJudHVrNqHUQYyokS7OXWcBW4B5wLXOfdcCc53/zwPGOWd3BwK5nozvQe2TG+WPjfLS31q1zMDDddgeAlwgIuditBDDRORTpdQ1nhQUDPGbD/7PUJbqOaiAS29L4+W7PRd6rsR7CmM10efUPJJ3BZJxsP6JDfZsD+SKk3tVvP5oxUbuHn2SS7O65fgtzaf09MrWXtmAYPx+yaNkTBR+v+RRNtB4z9HMiuXfImzdA5FsG+b9pThauD8+9+eiCHoNymfDylDi2xdjtSpys+q/YH7V4nCGX5bJV6+3YPhlmaxcVPuM85GkHwykS48c/P3tlJSY6NUvk4StYWxcG82pZx5k6eKWnHXeflYvdS9hQHhUGbYyoTDfYqi3Dc5hzjvxRMaWOp2fYvDZWezd4V7vJTyyhNxs/woFu/tvOhWHEnqeksmGf2Jo2boAi9VBXo73E2hURbzXoLwb+Mw5o5sIXI9To0dEbgT2AmOcx/4EnAskAEXOYz2itrW69ZJoUko9jNM5isgwYGJ9nJ4v8abCWHUqawu/jGXY+Zn8Ps+z1t6kGYmVNldv4NMpLVn4pfuzghUUO7CsK6Lwrsofc/FlUQS/mIL/4jwcsRYKHzYeMIeuiCJ4aiphdxgt60PXxaDCax8GqK68i76MZsIre3lr8WZspcLkCe1wNyRi0szd9ByUT3iUjU/XbOSTV+P4cmYLHn1rN6OuyCRtnxHO4g7bN0ew4tcWTPt0BXa7kLg9jJ+/a82aFc148Pn1jL19J4nbw1g4173Qm8jYUia+nIDJpAz1tp9j+GtJFP/38SbCo2yIKBK3BjPjCffExatTsFv8Q2vue/RfXv/0D2w2YcqzvfGprq4S8NIYu1JqPVCdGNFZ1Ryr8FIonUsqa04h8K5UGdtTSn3s8kUqHV+t4SxhEqUGyFGft8Hii7RUKC/ECVRD1lzfSA3qtFS+S0tlivFe5EI5fx74jNyS1Hp5Lf+2rVXcw/e6dOze2x9onCprIvIkRkxNV4ym5jnAcsBlx6eU+h0jFkej0fwXaMDrcF3BlVndSzGanQeVUtcDvQD3BlI0Gs1/iyaQpOCQUsohIjYRCcOYWvaNirRGo2n4/JcTkVZhrYhEAO9gzPQWACt9WSiNRtOw8eKs7nHBlbW6dzj/fUtEFgBhSqkNvi2WRqNp0PxXHV9tKV9E5GSl1D++KZJGo2no/JdbfK/W8p4CzvRyWRARTAH1X8d6FCbX0wm5g+PQIa/bNHd2L67LVaIvSa77IA/Y+aJvIhU6PrrOJ3btaRlet2mKivC6TQBHVo73jdq8lEP4vzrGp5Q641gWRKPRNBIa+IytK2hBcY1G4z7a8Wk0mqaG+GaB0TFDOz6NRuM+jbzF50oGZhGRa0TkCefrNiLS3/dF02g0DRFRrm8NFVemO98ABgFXOl/nA6/7rEQajabh44V8fMcTV7q6A5RSJ4vIOgClVLYzd9YxISauhImTdxEZU4ZSws9fNGPuhy1of2Ihdz+3h4BgO2n7/Hl5/AkUFbjec7f6OXhl9iasfqpSBWtaax58dSedehRgswk7/g1h+uMdsLuhrlWV4DAb4ycn065LMUrBlPvbsPVv1xS17ntwLf0HHiQnx587bjgbgBtu3cCAwSnYykykHAhh6kunUFjoR59TUrnulk1YLQ7KbCbef6sH/66rO39cTXU7afrOKqpwNgryLNw1ukettn6/4DMKbX7YlWB3CBcvvISHeq/kzPgkyhwmkgrCeGjVMPLLjGSzXSIyea7/UkIsZTgQLl5wMaWOmr+/mLgSHng1kYiYMlDCT7NjmfthC256OIkBZ+VgKxMO7A1gygPtKcx3/T6oThWv/UlF3PPCHgKCHKTu8+Ple09wOyN3fNtCJr1UGecfF3+IT948gcy0AK6+bRet2xcyfmx/dm5xb9m71c/BK59tMNTbzLB8YTSfzmhL74E53PjgbsQExUVmXp3UiZSkQLdsu0UDbs25Qp1pqURkNTAYWON0gLHAImeO/LrO3YPRQrQDtrrS04SbotXAgHMP2xcZW0pUszJ2bQ4mMNjO9HmbePbWTtw/OZF3X2jDxr/CGHFZGs1blfDJ1BqWEFcbx6cICHJUUcHazNvPtSM03MaaPyIAeGjqTjatCePHz1tUa7auOL6Jr+1l0+oQFsyOxmJ14B/ooDCv9h9leRxf957pHDpk4f6H11Y4vj59U/n3n1gcDhPX37IRgA9m9aBDxxxysv3Jygykbbtcnn15OePGnHf4p92776hr1VS3SQmVyTFvemQvRflmPp9RfT66nc/1BgzHd/HC/5FdUvljO7VFMitT47ErEw/0XgXAK+sHYhYHc0d9w8SVZ7ItJ5oIv2LyyvxwqMrv6cg4vihnWROcZZ3xwyaeuaUzMXGlrP8zDIdduOEhI1bx/ZdqWUp+RLqr7v3zKS4yMXHK7grHN33eZt55vjUbV4cxYkw6LVqX8PGrNefjqyuOz2RSfLxwKRPG9cc/wI7DIdz92Fbem9qpVseniqq7v464bz/fwNvPd+D+l3bwzB1dSU4M4ryrUujSI58pD3c+6uxVBfPItWfUqykW0Kq1an3XBJeOTXh4QoNMS+VKU2Y68B3QTESex0hJ9YIb1zhDKdXb0w+fne7Hrs1GK6lcAS26RRnx7YvZ+Jch3vLP8nBOHZXlpuUjVLCshgrWmj+M9OggbN8QQkwLz1J4B4Xa6TGgkAWzjZxqtjJTnU6vKps2xJKfd3jDet3a5jgcxle2bUsUMbHGDyMxIYKsTMPh7N0Thr+/HYu17kDVmuq2EsXQc7P4/QfPkp4uP9gau9OZrc9oTosgQ2Do1Lh9bM+JYluOoT+SUxpwmNOrjqx0PxIOK2sg0S1K+WdZOA678Tveti7Y7e+rOlW8+PYlbFztvLeWhTHknGy3bB5Jr/5ZHNwXSFpKIMm7Q9i/13Ud3aOpTr3NuF/L5VaDQ2xkpvmwU6aMWV1XtoaKK2t1PxORvzFSUwlwkVLKa3KR7tAsvoQTuhWxfX0we3cEMujsbFYujuK0c7OIiXPfQZlMiunfb6Bl22Lmf9qiQgULwGxxcNZFGbz1bDuPytqiTQm5mRbun5pEh67F7NwQyJtPxFNyyDsiRiPO2cPSJUe3bIYM3U/CzogKqUFXqVq35XTvl092pmuqcArhwzN+QgGzd57El7sOF7S+7IRt/LjXaM22D81BIXxwxo9E+Rczf+8JvLO1t8tlbR5fwgldi9i+/nChpBFjMlg6v/7JO/fuDGDQiBxWLopk6HnZxHpwb1Xl9JEH+X1B9b0GTzCZFNO/XU/LNoeY/3kc2zeE8tqjHXlm1mZKS0wUFZgZP6ZX3YbqQyPv6royq9sGI7/9DxhiH4XOfa6ggEUi8vcReppV7VeorJXWorIWEGTnsTd28PazbSkqsDD1oQ6MviaV6XM3Ehhsx1bm/jicoYLVi7GnnkLnXgW07VRU8d6dT+9m01+hbF7rfvp5ALMZOvYoYv7HMdw5sgvFRSYuv8tdbdbqufzqbdjtJpb8crjja9Mujxtu2cSMKTUus66WI+u2nGEXZPKHi6pwVyy+kAsXXMINS87lms6b6Rd7oOK927v9g81hYu6eTgCYRXFK7EEm/Hkmly++gBGtdzOo+dFd8RrL+uZO3n62zWHjblfceQC7Tfjt+/qr2E15oD2jx6YxY/5m573lec/QYnEw4PR0li/2XuZnh0O466I+jD29f4V628XXHeCJW7ox9vT+LPq2OTc/7JvM2BU0gXx8P1IpOhSAIQK8HejmwrmnKqX2i0gzYLGIbFNKLa16gFNncxYYY3zVGTFbHDz2xk6WzIvhz4XGE31fYiCPXnsSAPHtD9H/jBwXilM9hfkWNqwKo+/QHPbuDOKqu5MJjyrjuce61H1yDWSkWElPsbJ9ndGCWv5jBGO84PiGj9xD/0EpPHL/aVTVVYiOKeLxZ1by6ot9OXigGsnIGqiubsFQhRs8Mot7Lujukp3UQ8bnzCoJZPG+9vSMTmdNekv+1347Z8bvZeyvoyvKe/BQMGvS4irGA38/0IZuURmsTK1d18JscfD4mztZMjeaFVXKevYl6Qw4M5tJV58IXtCa2LcrkEfHGt99fPti+p/peWr5vqdmsGtbKDlZ3hf5rlRvy6bDiYVs32D0WJb+FMNz7272+vWq0pBDVVyhzmaSUqqHUqqn828noD8u5uNTSu13/k3DGCf0IP5Pcd+Lu0neFch378VV7A2PNsaiRBRX3HmAnz53XwUrONSQN/Tzt9NnSC7JiYGMHJPKKafl8NJ9nZxjJ56RnW4l44AfrU4wZkd7n5pP0o763fyn9DvIpVfs4OlHB1NSUvnMCg4u5ekX/+SDd7qzZZM743HV1y1AnyG57NsVSMbBusscaC4j2FJa8f+pLfaxMzeSoXFJ3NJ1Pbf+MYpie6Uy27IDrekSkUWAuQyzOOjfLIWE3Mg6yzr+pd0kJQTybZWynjI0h0tvTeGpmztTUuydYYSq99aVdx/gx888V947fdRB/vBiNzc88oj7dnAOybuCCAq1Ed/OGPPtMySHpF2+0Z7+r+D2yg2l1D8iMqCu40QkGDAppfKd/48AnnH3et36FjD8fxns3hbIzPnGTOZHk1vTsl0xo8emAvDnwkgWzXHv5oyMLWXiKwmYTBgqWD9F89eSSOZvW0naAX+mzNlk2F4UxeczPUs4/frj8Tw0Yy8Wq+Jgkh+vTnB1hAAefGw1PXtnEBZewsdf/cSnH57EmKu2Y7U6eH7yMgC2b4li5tSTOf/iXbRsWcCV47Zy5Thj+PWxB04lN6f2sbma6nbN7xGcPjqT339wrdsYE3CIN4YuBMAiinl7O7I0pQ2/nj8bP5OdD8/8EYD1Gc14Ys1Q8sr8eX9bD74b+R0Ko8X3+4HaZUKNsmaye1sgr/9ofDcfvtKK25/ci9VP8cInhjretnXBzHjMdWGl6lTxAoLsnD/OaJ2vWBDJoq88m9zxD7DTZ0AWM547qWLfoDPSuP2hbYRHlvLU9PUkbg/l8TtdH5qIbFbKxBd3YDIrRGDZghj++j2K6Y915NHpW1EKCnItTH3k6Bldr9LIW3yuhLNUnbc2YSidRyulRtZxXgeMVh4YDvZzpdTztZ1TXTiLV9BpqaoNZ/EG5eEs3sZXaamODGfxBr5KS1V9OEv98Eo4S8vWqt0troWzbH+6YYazuNLiC63yvw1jzO+buk5SSiViCBNpNJr/Go28xVer4xMRMxCqlJp4jMqj0WgaOELjn9yoLfW8RSllE5Ehx7JAGo2mEfBfdXzAXxjjeetFZB4wBygsf1Mp9a2Py6bRaBoiDTzziiu4MsYXAGRiaGyUx/MpQDs+jaap0oCXo7lCbY6vmXNGdxOVDq+cRu7vNRpNffgvt/jMQAjVh8L75GMrQNm9/yhRpWV1H+QBpiDvB4nad+zyuk0Ak7/3Vw4AdHrCNxLL7Zf7Jpfbrn41L4v0FHtautdtAlBHqJlnJr30+/oPO74UpZTbAccajeY/TgNfh+sKtUX2Ntz0qRqN5rjizdTzImIWkXUiMt/5ur2IrBaRBBH5sjzxsYj4O18nON9v52n5a3N8Z3lqVKPR/MfxbnaWe4Gqqe5eAqYqpToC2cCNzv03AtnO/VOdx3lEjY5PKeVuZk+NRtNE8FYiUhFpBZwHvOt8LRgRJF87D/kIuMj5/4XO1zjfP8t5vNv4ZhGrRqP57+Jqa8+1Ft9rwINUBshEAzlKKZvz9T4g3vl/PJAM4Hw/13m822jHp9Fo3ELc2ICY8kTDzq0iIbGIjAbSlFJ/H9MPQCMQFI+JK+GBqbud6lrw0+exzP2gBSHhNh55fRfNW5WQus+fF+44gQI3NC0mTN7LgOG55GRYuHW4kSY9NMLGI2/spnnrUlKT/Xj+9vYU5Lqp3Pb5JkMBq1y5bXobJry0kx798ih0Zgye8lAnErd6rrtQH/W2qtSksgZwwbiDjB6bisMu/LUkgvdfcj2lljcV7HI+t5H3vR0E/DsKsU9YSX+2jJKtCiwQ0M1E7CMWxCLk/2wn52ObsbIgCGInWfHv7N6zfcKUJAYMzzfuizM9T0R7JN76zo6k77A8bnv2AGaT4ufZUXw103uZnmvF9fG7jFqyswwBLhCRczEWSoQB04CI8iWzQCtgv/P4/UBrYJ+IWIBwjMUVblNnWqr6ICIRGH337hhVdYNSqsYkpmGmaDXQOuqwfVHNnOpam5zqWvM388wtnTj70gzycyx89WYcY25PISTcxvsvVp83T9mPFt7pPiCf4kIzD7y2p8Lx3fjoPsPm6y0Yc+dBQsPtvPdC/FHnlmMKPDLf3ZHKbZt4+7n2nHvlQf5aEsnyBXXndXMUFdV5jCfqbdXF8dWkshYRU8YVdx7gyRu7UFZqIjy6jNxMazVW8ZmCXfs/HNjSFPtvLqX1l36YAoSDD5cSNNiEOUoIGmxcN+2xMgL6mAi/1ELxvw6s7QVzmFC4wk72OzZafXj4597Vr7jWeuo+oIDiIhMPTEt23fG5MMzkyXdWVxyfyaR4b/k2Hr6iAxkpVmb8tJP/u6MtSTtrzsO4Wv1KnsqqV8RGUPPWqtOVrqWl2jDNtbRUIjIMmKiUGi0ic4BvlFJfiMhbwAal1BsicifQQyl1m4hcAfxPKTXGk8/g667uNGCBUupEjBRVbosUZaX5kbDpCHWt5qUMOjuHX74xuve/fBPN4BE5btndtDqU/JzDM/YOGpHLL3OcNudEM2ikezarV8By00Qd1Fe9rSo1qaydd3UaX73VkrJS4/ao0enViPcU7JRNoUqcf4vBEisEDzEjIogI/t1M2NKMSg7oZcIcZvymA3pU7neHTatDyM/2bkfIm99ZVbr0KeLAHj8OJvljKzPx+9wIBo30PE2+W/hWc+MhYIKIJGCM4b3n3P8eEO3cPwGY5OkFfOb4RCQcGIqz0EqpUqVUTn1sNm9VrgQWQkRMGVlOCb2sNKvRFa4nkTE2stKsTpsWImNsdZxxNCaTYua89cxetYZ1K8IrlNuuHZ/EGz+s55ZHdmP18zx6vqp62+sLt3PfK0n4B9YtJVkXVVXW4tsX071fPlO/3cTLs7fQuWeB2/aMeviX2avXsm55eLUKdmuXRtRqw9JMiLjGwt7zS9hzTgmmYAgaWPmwUjZF/k92ggYdnXI+f66doMHeSUVfX3z1nUW3KCP9QKWMZEaKlZg436xSOgwfyEsqpX5XSo12/p+olOqvlOqolLpMKVXi3F/sfN3R+X6ipx/Bly2+9kA68IEzOPFdZwr6w6iqslamau6GBATZeeytBN5+pnU1qvbig0By8ai1Zii39WbsaX0rFLA+mNyWm0f24d5LehIabuOyW/bXbagGfKHedqTKmtmsCA23Mf5/3Xj3/9rw8IwE3H18e0PBzp6nKFzqoO1cf9r97I+jGPJ/qnQY6S/aCOxjIrDP4bfxobV28ubZib6rYQxh+1Jx77jRyFXWfOn4LBhprd5USvXBSGl1VNNUKTVLKdVXKdXXKtWPTZgtDh5/K4El30ezYoHRXcjJsBLVzOgqRTUrJTfD3e7Y0WRnWIhqVua0WUZOpuc/nEoFrByy0/0AoazUxKJvmnnUgiqnOvW2jj08T1FencpaxkE/Viw0uqU7NoSgHBAe5X7rFw5XsAMqFOxmvdCuznMP/eXA2lIwRwpiEULOMFO8wWhGZL1jw56jiB5/+HdUstNB2nM2Wky2Yo5oGIuPvP2dlZN50Epsy8rhgpi4MjJS6v87cAVvrtw4HvjS8e0D9imlVjtff43hCN1EMf7lPYa61ruVA+Grfolg+CXGhM7wSzJZuTiivuVl1eJwhl/mtHlZJisXhbt1/lHKbYNzSE4MJDK2/OZUDD47i707PE9u4F31tupV1lYujqTXwHzAkO60WBW5Wa4/BLylYGdpIRRvdOAoViilKFpjx6+9kPe9jaKVdpo/Z0VMlXbKDioOPlhG86et+LVtOJFavlDcA9i+Poj49qU0b12Cxepg2IU5rHLznvWYRt7i81lfQCl1UESSRaSLUmo7xhK4Le7a6da3gOGXZLJ7ayCv/1SprvXlG3E88kYCIy9PJ22/P8/f4Z5Iz6SZu+k5KJ/wKBufrtnIJ6/G8eXMFjz61m5GXZFJ2j4jnMUdImNLmfhyAiaTMpTbfo7hryVR/N/HmwiPsiGiSNwazIwn6icoVB/1tqrUpLK2aE4s419K5M2fN2ArE159oAPuLN32loJdQHcTIWeZ2HdNKZjBv4sQdrGZxKElWFoI+28wHijBZ5iJutlC9rs2HLmK9Jec8pAWaPWxew5m0ht76TmowLgv1m7hk1ebs3B2/UXKvfWdVcVhF15/NJ4XPk/EZIZFX0Sxd0ftynreoiG35lzB1+EsvTHCWfyAROB6pVR2TcdXF87iDaoLZ/EGR4ez1B9Xwlk8wVdpqXylYNf+D99kuqwrnMUjPFs1VTc++G16JZwltrU68X+uhbOsm9V4VdY8Rim1HmhwH1qj0XjOf1psSKPRaGpEOz6NRtPUEB8OkR0LtOPTaDTu0cBnbF1BOz6NRuM2eoxPo9E0OdxZjtYQaViOTymUzftrDcXsmzWbyubZaobjga/K6qtQoV39fNOkaLkqtO6D3CTlNO8rtwGYOrb1uk1JXO4dQ7rFp9FomhQNfDmaK2jHp9Fo3Ec7Po1G05TQAcwajaZJIo7G7fm049NoNO6h4/iOLa1OKOaRN/dUvG7RppRPJrfgu3ebuW1r/Ct7GHBWLjmZFm47u1vF/guuS+P8cWk4HMJfv4Xz3gutXLYZE1fCA68mOoWRhJ9mxzL3wxaMm7CPQWdn43AIOZkWXp3YoSJ7tCdcfHMa51yZhVKwe1sAr05oQ1mJ+8kCaqoDgP/dnMotj+9jTK9e5NUjFbs3v7Oq1FcUqGB2KUXzykDAeoKJiMcCsGcqsh87hCNPYe1iJvKpAMRauZ7/0G9lZD9STMwHQfidVHukgDdFsu6buJb+A1PIyfHnjptGAHDDLRsYMCgFm81EyoFgpr7cl8JCP8xmB/dO/JuOHbMxmRW/LW7LV7NPdLt+6qKxh7P4MvV8FxFZX2XLE5H76mNz364A7hhxIneMOJG7RnWh5JCJFT9HeGRr8ZxoHhvX6bB9PQflM2hEDneM6sqtw7vx9dvuKVY5bMI7z7fh1hE9ue9/XTl/XCptOh7i61lx3H5OD+48rzt//RbB1fd4noE5ukUpF92QwV3ndubWs07EbIZhF9aY8KZWqqsDgJi4Uk4ZmkfqPs+dczne/M6qsujLKB692r20YeXY0xwUflVK7AdBNPs8GOWAQ4tt5L1eQsiVfjT/OgRTmBiO0YmjUFH4VRnWbq79ZBx24Z3nWnPr8B7cd1FXzh+XRptOh7j8jhTWrwjjxmE9Wb8ijDF3pNRp65eFbXn84VMP27fu72bcfuPZ3Hnz2ezfF8KYq7YBcNrp+7Ba7dxx8wjuvf0szhmdSLPmhW7Ujos08nx8PnN8SqntSqneSqnewClAEfCdt+z3PjWflL3+pO337Me56a+jxYZGj03nqzdaeCyyk5XuR8LmI4SRWpQelio/INDhUhLO2jBbFP4BDkxmhX+gg8yDnmXdra4OAG59Mpl3X4j3+o1b3++sKvUVBVJ2qogYKUwxQulaOwFnGDaDzrVSvLQy9jF/VgkhY/0QP9e+O2+KZG3aGEt+3uF1tu7vFjgcxn26bUs0MTFGRmelICDAjsnkwM/fjs1moqjI+1mZG3sG5mPV1T0L2KWU2ustg8MuzOH37yO8ZQ6A+PbFdOtfwLUP7Ke0xMS7z7VixwbPtE+bx5dwQldDGAng2onJDL84k8J8Mw9d5XnXI/OgH1+/1YxP/tpCSbHwzx9h/LO0du0Kdxh4dg6ZB/3YvdXzLNE14YvvzBPMzUyEXO1H6kUFiL/g39+M9UQTEgpiEecxgj3d+OWWbrNjT1UEDLFQ8GndynBH4muRrBHn7GHp78aQzPKlrRg45ACfzZmPv7+dWW/2oiC//g+aw1D4JFfgseRY5ee+Aphd3RuHiQ3hWgS8xepg4Ihcls6P8GIRjZZUaLiN+y48kXefb8UjbyTiSbMnIMjOY2/u5O1n21S09j6a3JqxQ3qzZG40549L9biMIeE2Bo3M5dqBXbnq5O4EBNk5839ZHturin+AgyvuSuHjV1t6xV5VfPWdeYIjT1G81Eazb4NpPj8YVQwlK6tfgaIcirxpJYTd41kiV1+LZF1+1VbsdmHJL0ZG5y4nZuGwC9eMGc3115zD/y7bQYs4zzVeasLbKmvHGp87PhHxAy4A5lT3/mFiQ7h2c/U7I5+EjUHkeEFgqCoZKX6sWOAU2fk3GIdyX2THbHHw+Js7WTI3mhVO8Z6q/DY3mlNHeTYmB9DntAIOJvmRm2XBbhNW/BxB177eGcOJa1tCi9alvLlgCx+t2EhMXCkzf9pCZGz9WyW++s48oWSNDUtLE+ZIE2IRAoZZKN1gR+UbXV8Ae5rCHCuoIrAlOsi8o4jUiwoo3Wwn64FDlG6te6mer0Wyho/cQ/9BKbzyQn/KpQGGnZXM32taYLebyM0JYMumGDp19vx+q47yOL7G3NU9Fi2+c4B/lFKeN3OOYNhF2T7pMv25KIJeg8pFdoqxuimyA4rxL+02hJGqiPe0bFeZ7nzQ2dkkJ3qesj5tv5WTTi7CP8ABKEO4Zqd3UuDv2R7IFSf34tohPbh2SA8yUvy469yuZKfX31n56jvzBHNzE6Wb7BUiRiVrbVjam/A7xUzxEuNBV/RTGQGnWTCFCC0WhtD8e2Pz62Ym6pXAOmd1fS2SdUq/g1x6+XaefmwIJSWV92haWiC9+hjSlf4BNk7smklyspfXJyvl+tZAORZjfFdSQzfXE/wD7Zw8NJ9pD9UsUuMKk2Yk0nNQPmGRNj5ZvYFPp7Rk0ZfRTHhlL28t3oytVJg8oR3uiOwY4j2Z7N4WyOs/VgojjRyTTqsOxSgFqfv9mfFoO4/LvX1dMMt+DOf1hdux24SEzYH8/JlnYjjV1cHCL2M8LltNeOs7q0p9RIH8upsJONNCxrVFYAZrZxPBF1kJGGIh+/FD5L1dgrWzmaALPHf43hTJevDR1fTslU5YeAkff/Ejn37UlTFXbsNqdfD8y0sB2L41mpmvncz87zsy/sE1vPneIkQUixe0Y09ihMefoyYacmvOFXwtNhQMJAEdlFK5dR0fJlFqgGm498vho+ws+MCuKnV/8NwVfJahxkfZWXzVWmjq2VlWJn5A7qGUeoUVhEa0Un2G3uvSsct+eLBJig0VAvXX5tNoNA2Kxt7ia1QrNzQaTQNAAfbG7fm049NoNG6jW3wajabp0YBnbF1BOz6NRuM2usWn0WiaFg08AYErNCjHJyYTppAQ79s1+yZO257n/aVAllbxXrcJYE856BO7piDvr+kF34kj+SL0ZO/n7qfFcoV21yV632hp/VfhCCBemNwQkdbAx0BzDFc6Syk1TUSigC+BdsAeYIxSKltEBJgGnIuR9OQ6pdQ/nlz7WK3V1Wg0/yFEKZe2OrAB9yulugIDgTtFpCswCfhVKdUJ+NX5GoxVYJ2c2y3Am56WXzs+jUbjHq7m4qvD7ymlUspbbEqpfGArEA9cCHzkPOwj4CLn/xcCHyuDVUCEiMThAQ2qq6vRaBoDbq3DjRGRtVVez1JKzTryIBFpB/QBVgPNlVLlGVoPYnSFwXCKyVVO2+fcV3c21yPQjk+j0biNG7O6GXUtWROREOAb4D6lVJ4xlGeglFIi3p9D1l1djUbjPl7KziIiVgyn95lS6lvn7tTyLqzzb5pz/36gaqaLVs59bqMdn0ajcQ9lzOq6stWGc5b2PWCrUmpKlbfmAdc6/78WmFtl/zgxGAjkVukSu0Wj6eqaTIrp36wnI9WPp27rRvNWxUyaso2wCBs7N4cw+cHO2Mrc8+Mmk2LanH/ITPXnqTu602tgNjdO3I2YFMWFZqY82oWUpEC3bE6YvJcBw3MN9a/hXQE47bxsxk5IoXWnYu4Z3YWdHqSzv+Dy3Yy8MAkRWDi3DXO/aM+pZ6Zw1c07aN2ugPHXDyFhW4RbNqtTWbtm/AFGXZlBbqZxa3z4cjxrloS7Zdfq5+CVzzdh9XNgtiiWL4jm0+ltKt6/7fFERlySxv96D3TZZk0Kdjc9nMSAs3KwlQkH9gYw5YH2FOa7fluPf2U3A87MISfTym0jugPQoWsRdz+/Bz9/B3a7MPOxtuz417UwKym0E/3mPqzJJSCQeXs8pV2CCfk5g9AFWSgTFJ8cSs7YOEz5NmJeTcIv4RCFwyLIvqnuUCZf1K1HeKfzOQQYC2wUkfXOfY8ALwJficiNwF5gjPO9nzBCWRIwwlmu9/TCPnV8IjIeuAmjmjYC1yulims/q3ouHHeApF1BBIUY8V03TNzD9x/G88dPsdz1dAIjL03lx9nuTfBcOHY/ybuCCAoxUivd9UQCz9zVjeTEIM674gBX3JrE1Efdi9FaNCeKeR/G8sBreyr27dkewDM3d+Cel5LcslVO2w75jLwwiQnXn0qZTXj2tb/4a3kz9iaG8PxDp3DXpI0e2V08J5ofPmrGxKm7D9v/3bvN+GZWixrOqpuyUmHSuG4UF5kxWxxM/mITa5dGsm19KJ26FxAS5n4qq3IFu4TNwQQG25nxwybWLQ/nn+XhvP9yaxx24YaHkrn8jhTef8n1vH+L58QYdTClsg5ufDiZz6a1ZO3vEfQ7I4ebHt7Hg1e4ppMS+cEBDvUJJWNiWyhzIKUK/00FBK3JI2VyR7CaMOUa97Cymsi9vDnW5GKsSa79LHxRt57gQqhKnSilllNzwsuzqjleAXfW+8L4Vl4yHrgH6KuU6g6YMbQ33CameQn9h2Wx8OvyyR1Fr4E5LFtoJM385btmDDor0y2b0c1L6Hd6Fgu/qfyBK0WFYw0OtZGV7r5Iy6bVRyuXJScEsq8eWZdbtytgx+YISkrMOOwmNq6LZvCwgyTvCWV/kucB3zWprNUfobjIsGuxKCwWhVJGC/vGh/bw3svu55mrScHun2XhOOzGb2fbumBiWriXz9CogyOe/4qKh2FwqJ3MNNcSkkqhnYAthRSeGWnssJpQwWZCFmWRe1EzsBo/N0e4cT0VYKLkpGCU1Z2foffr1iN0BuY67QeKSBkQBBzwxMitjyTy3ivtCQw2nFJYpI3CPEvFDZ9x0J/o5u7d8LdO2sX7k9sTGFz5hJz2RGeefmsTpcUmigotjL+ityfF9Tp7E0MYd/t2QsNKKS0x03dwGglb3et+usMF16Yz/JIsdmwI4p3nWlGQ6/5tYjIppn//Ly3bFDP/sxZs/zeUC689wKpfo8j24IFSlSMV7MoZMSaDpfOP1jlxl7eeacPzH+/g5keTERNM+N9JLp1nSSvFHmYh6vV9+O0tprRDINnXt8R6oISArYVEzD6IsprIGdeC0o6er3jxZd26hAIasJCQK/hSV3c/MBkjA3MKxkDkoiOPq6qyVlpNL7j/sCxysqwkbPbeUrb+p2caNrccno33onH7ePK27ow7cyCLv2vOLQ/5YMmQByTvCeXrjzvw3IzVPDPtLxJ3hGF31E+btybmfxLL9ad1545RJ5GVZuXmx/Z5ZMfhEO66oDdjT+tL554FdO+Xy2mjMpn3iUfxphVUp2AHcMWdB7DbhN++r3/e29HXpPH2s60ZO6g3bz/ThvEv73HpPHEo/HYfomBkNAdf6YTyNxH2fRo4FKYCO6kvnEDO2BbETEmqV2vIV3XrKoJrqza80R32FT5r8YlIJEakdXsgB5gjItcopT6tepwzmHEWQLg55qia6npyHgPPzKLf0DVY/R0Ehdi57dFEgsNsmMwKh12IaVFCZqrrT7quJ+cx8IxM+g3NMmwG23nqzU20bl/E9g2GRu3Sn2N5dtYmTz++11n0QxsW/WAMYo+7fRuZad4RGDqSqipoC2bH8PQHCfWyV5hvYcPqcHoOyCOubTHv/2IsrfQPdPDeL/9w4/CTXbZVk4Ld2ZekM+DMbCZdfSLuaKTUxPBLMnnzKaOul/0YyX0v7a7jDANblBV7tJXSTkZrrmhQOGHfpWOPslI0IAxEKO0UhDIJpjx7RZfXU7xZt27jaNxNPl+GswwHdiul0pVSZcC3wGB3jXw4pR1jT+/PdWf148UJXfh3VTgvT+zChtXhnDYyw7jQxWms/M31J/2HU9sz7syBXH/2AF66/yQ2rI7gmbu6ERRqI75tEQB9BmWTvMs3C/A9ITzSWFwf2/wQg4cd5PeFvklmENWschH74JE57Nnu3qw2QHhUGcGhxrCEn7+dPoNzSNgczNWD+3HdGadw3RmnUHLI5OYPs3oFu1OG5nDprSk8dXNnSoq9M16ZmWal50BDba/3kHwO7HHtIeOItGKLtmLZb3xXARsLKGvlz6H+YQRsMiRALQdKEJvCEeZZWX1Tt25S3tV1ZWug+HKMLwkYKCJBwCGMWZq1tZ/iOu+/0p5JU7cx7r697NoazKI5zes+qRYcdmH6E515dNoWHA6hIM/Ca491dtvOpJm76Tko31D/WrORT16NIz/Hwh3PJhMeZePZj3axa3Mgj17TyS27j7z4N2HhZdhswpuvdKewwMqg0w9y28TNhEeU8tTUNSTuCOOJewe4XtZqVNZ6DsqnQ9ciUELqPj+mP+z+YHlkbCkTX07AZFKISbHs5xj+WlK/sbeaFOxuf3IvVj/FC59sB4wJjhmPtXfZ7qTpuyrrYNV6Pp0az7SH2nHbU0mYzYrSEhPTJrVz2V72DS2Jnp6M2BS25n5k3tEK5S9Ev7mfFhN2gEXIvLMVOFcntLxjG1LkQGyKwDV5pD3WHlvrmh2tL+rWExpyN9YVfK2y9jRwOUYWhnXATUqpGvMChZtj1MCQC7xfjsaUlireN+M0vkpLJf6uicC7i6/SUuHw/v3emNJSrSqaT649o17jAeFBLdWgTje6dOzCDc81SZW1J4EnfXkNjUZzrGnYoSqu0GhWbmg0mgaCVlnTaDRNkcY+xqcdn0ajcR/t+DQaTZNC4ZNJomOJdnwajcZN9OSGV1EOB47CouNdDJcRq/erz5GZ5XWbAPTyTciF2rDTN3bL3Ft77SrSr4fXbfpEDQ3Y/n/dvW6z+OVfvWNIOz6NRtOkUIC9AS/LcAHt+DQajZsoUNrxaTSapobu6mo0miaFntXVaDRNEt3i02g0TQ7t+I4tF92YxjlXZiACP38ew3fvNfPIji/U0GpSAhs3YR+Dzs7G4RByMi28OrEDWWmuJ061+jl4ZfYmrH6qUllrWmsefHUnnXoUYLMJO/4NYfrjHbDbas9EM/6eVQzot5+c3ABuu+u8iv0XjN7O+eftxOEQ/lrTkvc+7ENoaAmPTVpG505ZLP61PW+83c+l8npbuaw6JkxJYsDwfOP7O9O9UJ3xd69kQN99Rh3cc37F/gvO28b55+4w6mBtPO99dDIWi5177lhNpxOyUAreercvGzbVLcTkTTW0dk+tw+FvBpOgTELyA92J+mkf4SvTsIcYiWMzRremqFsEpsIy4t7bSUBSIXkDYkm/rJ1bdeMSSoH92Iga+Qpfq6zdC9yMkRb3HaXUa/Wx17bLIc65MoN7Rp9IWZnwwqcJrP41zOVEkVXxhRpaTUpgX8+K4+MprQC48LqDXH3PfrdyxpWVCpPGVlXW2szaPyJYMi+Gl+/vCMBDU3cyakwaP35e+49y8a8d+OHHzkwcv7JiX88eqQwasI877j6HMpuZ8HBDAqC01MzHn/WkbZtc2rXNcbm83lYuq45FX0Yx74MYHpiW7Pa5FXVw358V+3r2OGjUwb3nHVYH54wwMlDffu9owsOLee6J37hn4jkoVXtmJ2+roe27+yQcIYeLHmUPiyPnrMPTmCmLiczzWuOfUoRfyiG3ruEWjbzF50uVte4YTq8/0AsYLSId62OzTcditq0PpqTYhMMubFgVwpBzcjyy5Qs1tJqUwKpqQwQEOur80RzNEcpaVkNZa80fkRjPFGH7hhCXFMY2bW5Gfv7hrc3R5+7kq6+7UWYzrpGba9RBSYmFzVuaUVbmXrZgbyqX1XiN1SHkZ3v23N60pTn5BYfnERw9agdffXN0HbRpncu/G1pU7Cso9KNTR1cU/Y6PGpryN1N8QigOt5TbPLmQVlmriZOA1UqpIgAR+QP4H/Cypwb3bA/guocOEBpho7TYRL8z89i5oeGkh6/KkUpg105MZvjFmRTmm3noKvdbOoay1gZati1m/qeGslY5ZouDsy7K4K1n23lU1viWeXTrlsa1Y/+ltMzMu+/3YcfO+ov2VMVT5bJjRXzLfLp1TePaa9ZTWmrm3Q9OZkdCDIm7IxnYfx9LlrYjNqaITidkEhtTxA4XFqx4Tw1NiH9jGwC5Q5qTN8QY3olYdpCwNekUtw4h4+I2OIKO1ciVavSzur58LGwCThORaGf6+XOBo5Seq6qslVFjcmbAaJF99UZz/u/znTz/aQKJmwNxNMChhuqUwD6a3JqxQ3qzZG40549LddumoazVi7GnnkLnXgW07VS5tO/Op3ez6a9QNq8N86i8ZrMiNKSU+yaO4N33e/PIQ8sxYha8h6fKZccKs9lh1MEDo3j3w5N55MFlgGLhLyeQnhnEjFd/5rab1rJlWywOFxXuvKWGlnxfV5If7MGB208kYlkqAQl55J7anD1P9CbpwR7Yw63EfOfZ8IxHKFDK4dLWUPGlvORW4CVgEbAAWA8c5aaUUrOUUn2VUn2t1J3GfOEXMdx17klMvLQzBbnmenVNfUFNSmDl/DY3mlNHZXtsvzDfwoZVYfQdmgPAVXcnEx5VxqwX2nlsMyMjiBUrWwPCjp0xOBxCeFjtDyF3GX5JJit+NoS2l/0YSede3k/bXx8yMoNYseroOnA4TMx6ry93jj+Pp18YRkhIKfsPhNZpryrVqaF9uOTvCjW0urBHGK1De6iVgp6RBOwtxB5mBZOAScgd1IyApGNcn3aHa1sDxacDAUqp95RSpyilhgLZwI762gyPNlTAYluWMuScHJZ8H1lfk16keiWwlu0q9YIHnZ1NspvO+ihlrSG5JCcGMnJMKqeclsNL93XyYNywkj9XtaJXT6MVGt8yD6vFQW6ed7U0PFUuO1b8ubo1vXpUqQOrUQf+fjb8/Y2679MrBbvdRFJyRJ32vKWGJiV2pNhe8X/QtlxK4wIx51aO54ZsyKI0zn01PI9RypCXdGVroPh6VreZUipNRNpgjO/VPXdfB0/MSiQ00o7dJsx8tDWFeZ59BF+oodWkBDZyTDqtOhSjFKTu92fGo+3cKmtkbCkTX0nAZMJQ1vopmr+WRDJ/20rSDvgzZY5xrT8XRfH5zKNGEw7/3BNX0LNHKmFhJXzywXd8+nlPFv3SgQn3rOatmT9is5mY/NpAyvVpP3p3LkFBZVgsDgYN3MejT5xJUnJ47dfwsnJZtdd4Yy89BxUY39/aLXzyanMWznZtXHLS/cvo2d1ZB+99y6eze7LolxOYcPdK3pr+g7MOBgNCREQxzz/1Kw6HkJkVxCtTXVNI9ZYamjm/jJbvOgcUHYr8U6Ip6hpB848T8N9fBAJlUf6kXV4ZJdDuqXWYiu2ITRG8IYsDd5xIaZyXx8Ib8MSFK/haZW0ZEA2UAROUUrXmxAmTKDXAPMJn5fE2vkhLJWbvaMMeiTrJ9fAZt9BpqZAtjSctVcrL0yhJSq6fypo5Rg0MPK/uA4FFhR83SZW103xpX6PRHA8adqiKKzS6lRsajeY4o5MUaDSapoYCVCNfsubj8G6NRvOfQzkTkbqy1YGIjBKR7SKSICKTjkHpAd3i02g0HqC80NUVETPwOnA2sA9YIyLzlFJb6m28DnSLT6PRuI93Wnz9gQSlVKJSqhT4ArjQ52XHx+Es7iIi6cBeFw6NATJ8UARtt3GVtbHZbQhlbauUiq3PxURkgfOarhAAFFd5PUspNctp51JglFLqJufrscAApdRd9SmfKzSorq6rX4iIrPVFbJC227jK2tjsNqay1oZSatSxupav0F1djUZzvNjP4YlLWjn3+Rzt+DQazfFiDdBJRNqLiB9wBTDvWFy4QXV13WCWtuszu42prI3NbmMqq89RStlE5C5gIWAG3ldKbT4W125QkxsajUZzLNBdXY1G0+TQjk+j0TQ5Gp3j88USFxF5X0TSRGSTN+w5bbYWkSUiskVENjsV57xhN0BE/hKRf512n/aG3Sr2zSKyTkTme9HmHhHZKCLrRWStl2xGiMjXIrJNRLaKyCAv2OziLGP5lici93mhuIjIeOf3tUlEZouIVzKxisi9TpubvVXWJoFSqtFsGAOgu4AOgB/wL9DVC3aHAicDm7xY1jjgZOf/oRjZp71RVgFCnP9bgdXAQC+WewLwOTDfizb3ADFevhc+Am5y/u8HRPjgXjuIEfBbX1vxwG4g0Pn6K+A6L9jtjqFtE4QxUfkL0NGb9fBf3Rpbi88nS1yUUkuBrPraOcJmilLqH+f/+cBWjB9Afe0qpVS5wILVuXllhkpEWgHnAe96w56vEJFwjIfVewBKqVKlVI6XL3MWsEsp5cpKIlewAIEiYsFwVAe8YLNCyVApZQPKlQw1ddDYHF88UFVBeh9ecCa+RkTaAX0wWmfesGcWkfVAGrBYKeUVu8BrwIOAt8USFLBIRP4WkVu8YK89kA584OyWvysiwV6wW5UrgNneMKSU2g9MBpKAFCBXKbXIC6ZdUjLUHE1jc3yNDhEJAb4B7lNK5XnDplLKrpTqjRHp3t8p3l4vRGQ0kKaU+ru+tqrhVKXUycA5wJ0iMrSe9iwYQxNvKqX6AIWA11IaOYNpLwDmeMleJEbPpD3QEggWkWvqa1e5qGSoOZrG5viO2xIXTxARK4bT+0wp9a237Tu7d0sAb6ydHAJcICJ7MIYQzhSRT71gt7zFg1IqDfgOY8iiPuwD9lVp6X6N4Qi9xTnAP0op9wWQq2c4sFspla6UKgO+BVxTLaoD5QMlw6ZAY3N8x22Ji7uIiGCMQW1VSk3xot1YEYlw/h+IkctsW33tKqUeVkq1Ukq1w6jX35RS9W6ViEiwiISW/w+MwOii1aesB4FkEeni3HUW4M0cblfipW6ukyRgoIgEOe+LszDGfOuNiDRz/i1XMvzcG3b/6zSqJWvKR0tcRGQ2MAyIEZF9wJNKqffqaXYIMBbY6ByPA3hEKfVTPe3GAR85kziagK+UUl4LPfEBzYHvjN87FuBzpdQCL9i9G/jM+QBMBK73gs1y53w2cKs37AEopVaLyNfAP4ANWIf3lpl9IyLlSoZ3+mCS5z+JXrKm0WiaHI2tq6vRaDT1Rjs+jUbT5NCOT6PRNDm049NoNE0O7fg0Gk2TQzu+RoSI2J1ZQzaJyBznMiVPbX3oVLnCueSray3HDhMRtwNunVlZjlLjqmn/EccU1PZ+Ncc/JSIT3S2jpmmiHV/j4pBSqrdSqjtQCtxW9U3nAni3UUrdpGoXcR6Gl1YaaDQNAe34Gi/LgI7O1tgyEZkHbHEmMHhFRNaIyAYRuRWMlSQiMtOZy/AXoFm5IRH5XUT6Ov8fJSL/OPP9/epMsHAbMN7Z2jzNuXrkG+c11ojIEOe50SKyyJkb7l2MFFq1IiLfO5MXbD4ygYGITHXu/1VEYp37ThCRBc5zlonIiV6pTU2TolGt3NAYOFt252AsTAdjnWp3pdRup/PIVUr1ExF/YIWILMLIDtMF6IqxmmIL8P4RdmOBd4ChTltRSqksEXkLKFBKTXYe9zkwVSm13LlUaiFGiqQngeVKqWdE5DzgRhc+zg3OawQCa0TkG6VUJhAMrFVKjReRJ5y278JY8XCbUmqniAwA3gDO9KAaNU0Y7fgaF4FVlr8tw1gLPBj4Sym127l/BNCzfPwOCAc6YeSvm62UsgMHROS3auwPBJaW21JK1ZSjcDjQ1bkMDSDMmYVmKM58cEqpH0Uk24XPdI+IXOz8v7WzrJkYqbG+dO7/FPjWeY3BwJwq1/Z34RoazWFox9e4OORMR1WB0wEUVt0F3K2UWnjEced6sRwmjKzPxdWUxWVEZBiGEx2klCoSkd+BmlKyK+d1c46sA43GXfQY33+PhcDtzpRYiEhn58L7pcDlzjHAOOCMas5dBQwVkfbOc6Oc+/Mx0ueXswgjSQDO43o7/10KXOXcdw4QWUdZw4Fsp9M7EaPFWY4JKG+1XoXRhc4DdovIZc5riIj0quMaGs1RaMf33+NdjPG7f8QQT3obo2X/HbDT+d7HwMojT1RKpQO3YHQr/6Wyq/kDcHH55AZwD9DXOXmyhcrZ5acxHOdmjC5vUh1lXQBYRGQr8CKG4y2nECPJ6iaMMbxnnPuvBm50lm8zXpAe0DQ9dHYWjUbT5NAtPo1G0+TQjk+j0TQ5tOPTaDRNDu34NBpNk0M7Po1G0+TQjk+j0TQ5tOPTaDRNjv8HjrNjL+2zOTYAAAAASUVORK5CYII=",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "eval(abc_pipeline, X_test, y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Gradient Boosting for MNIST multiclass classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[1;32md:\\Data Science\\IITM\\MLP\\Week10 - Boosting.ipynb Cell 17\u001b[0m in \u001b[0;36m<cell line: 2>\u001b[1;34m()\u001b[0m\n\u001b[0;32m      <a href='vscode-notebook-cell:/d%3A/Data%20Science/IITM/MLP/Week10%20-%20Boosting.ipynb#ch0000020?line=0'>1</a>\u001b[0m gbc_pipeline \u001b[39m=\u001b[39m Pipeline([(\u001b[39m\"\u001b[39m\u001b[39mclassifier\u001b[39m\u001b[39m\"\u001b[39m, GradientBoostingClassifier(n_estimators\u001b[39m=\u001b[39m\u001b[39m10\u001b[39m))])\n\u001b[1;32m----> <a href='vscode-notebook-cell:/d%3A/Data%20Science/IITM/MLP/Week10%20-%20Boosting.ipynb#ch0000020?line=1'>2</a>\u001b[0m train_classifiers(gbc_pipeline, X_train, y_train\u001b[39m.\u001b[39;49mravel(), cv, \u001b[39m\"\u001b[39;49m\u001b[39mGradientBoostingClassifier\u001b[39;49m\u001b[39m\"\u001b[39;49m)\n",
      "\u001b[1;32md:\\Data Science\\IITM\\MLP\\Week10 - Boosting.ipynb Cell 17\u001b[0m in \u001b[0;36mtrain_classifiers\u001b[1;34m(estimator, X_train, y_train, cv, name)\u001b[0m\n\u001b[0;32m      <a href='vscode-notebook-cell:/d%3A/Data%20Science/IITM/MLP/Week10%20-%20Boosting.ipynb#ch0000020?line=0'>1</a>\u001b[0m \u001b[39mdef\u001b[39;00m \u001b[39mtrain_classifiers\u001b[39m(estimator, X_train, y_train, cv, name):\n\u001b[0;32m      <a href='vscode-notebook-cell:/d%3A/Data%20Science/IITM/MLP/Week10%20-%20Boosting.ipynb#ch0000020?line=1'>2</a>\u001b[0m     estimator\u001b[39m.\u001b[39mfit(X_train, y_train)\n\u001b[1;32m----> <a href='vscode-notebook-cell:/d%3A/Data%20Science/IITM/MLP/Week10%20-%20Boosting.ipynb#ch0000020?line=2'>3</a>\u001b[0m     cv_train_score \u001b[39m=\u001b[39m cross_val_score(estimator, X_train, y_train, cv \u001b[39m=\u001b[39;49m cv, scoring\u001b[39m=\u001b[39;49m\u001b[39m'\u001b[39;49m\u001b[39mf1_macro\u001b[39;49m\u001b[39m'\u001b[39;49m)\n\u001b[0;32m      <a href='vscode-notebook-cell:/d%3A/Data%20Science/IITM/MLP/Week10%20-%20Boosting.ipynb#ch0000020?line=4'>5</a>\u001b[0m     \u001b[39mprint\u001b[39m(\u001b[39mf\u001b[39m\u001b[39m\"\u001b[39m\u001b[39mOn an average, \u001b[39m\u001b[39m{\u001b[39;00mname\u001b[39m}\u001b[39;00m\u001b[39m model has f1 score of \u001b[39m\u001b[39m\"\u001b[39m \n\u001b[0;32m      <a href='vscode-notebook-cell:/d%3A/Data%20Science/IITM/MLP/Week10%20-%20Boosting.ipynb#ch0000020?line=5'>6</a>\u001b[0m             \u001b[39mf\u001b[39m\u001b[39m\"\u001b[39m\u001b[39m{\u001b[39;00mcv_train_score\u001b[39m.\u001b[39mmean()\u001b[39m:\u001b[39;00m\u001b[39m.3f\u001b[39m\u001b[39m}\u001b[39;00m\u001b[39m +/- \u001b[39m\u001b[39m{\u001b[39;00mcv_train_score\u001b[39m.\u001b[39mstd()\u001b[39m:\u001b[39;00m\u001b[39m.3f\u001b[39m\u001b[39m}\u001b[39;00m\u001b[39m on the training set.\u001b[39m\u001b[39m\"\u001b[39m)\n",
      "File \u001b[1;32mc:\\Python310\\lib\\site-packages\\sklearn\\model_selection\\_validation.py:515\u001b[0m, in \u001b[0;36mcross_val_score\u001b[1;34m(estimator, X, y, groups, scoring, cv, n_jobs, verbose, fit_params, pre_dispatch, error_score)\u001b[0m\n\u001b[0;32m    512\u001b[0m \u001b[39m# To ensure multimetric format is not supported\u001b[39;00m\n\u001b[0;32m    513\u001b[0m scorer \u001b[39m=\u001b[39m check_scoring(estimator, scoring\u001b[39m=\u001b[39mscoring)\n\u001b[1;32m--> 515\u001b[0m cv_results \u001b[39m=\u001b[39m cross_validate(\n\u001b[0;32m    516\u001b[0m     estimator\u001b[39m=\u001b[39;49mestimator,\n\u001b[0;32m    517\u001b[0m     X\u001b[39m=\u001b[39;49mX,\n\u001b[0;32m    518\u001b[0m     y\u001b[39m=\u001b[39;49my,\n\u001b[0;32m    519\u001b[0m     groups\u001b[39m=\u001b[39;49mgroups,\n\u001b[0;32m    520\u001b[0m     scoring\u001b[39m=\u001b[39;49m{\u001b[39m\"\u001b[39;49m\u001b[39mscore\u001b[39;49m\u001b[39m\"\u001b[39;49m: scorer},\n\u001b[0;32m    521\u001b[0m     cv\u001b[39m=\u001b[39;49mcv,\n\u001b[0;32m    522\u001b[0m     n_jobs\u001b[39m=\u001b[39;49mn_jobs,\n\u001b[0;32m    523\u001b[0m     verbose\u001b[39m=\u001b[39;49mverbose,\n\u001b[0;32m    524\u001b[0m     fit_params\u001b[39m=\u001b[39;49mfit_params,\n\u001b[0;32m    525\u001b[0m     pre_dispatch\u001b[39m=\u001b[39;49mpre_dispatch,\n\u001b[0;32m    526\u001b[0m     error_score\u001b[39m=\u001b[39;49merror_score,\n\u001b[0;32m    527\u001b[0m )\n\u001b[0;32m    528\u001b[0m \u001b[39mreturn\u001b[39;00m cv_results[\u001b[39m\"\u001b[39m\u001b[39mtest_score\u001b[39m\u001b[39m\"\u001b[39m]\n",
      "File \u001b[1;32mc:\\Python310\\lib\\site-packages\\sklearn\\model_selection\\_validation.py:266\u001b[0m, in \u001b[0;36mcross_validate\u001b[1;34m(estimator, X, y, groups, scoring, cv, n_jobs, verbose, fit_params, pre_dispatch, return_train_score, return_estimator, error_score)\u001b[0m\n\u001b[0;32m    263\u001b[0m \u001b[39m# We clone the estimator to make sure that all the folds are\u001b[39;00m\n\u001b[0;32m    264\u001b[0m \u001b[39m# independent, and that it is pickle-able.\u001b[39;00m\n\u001b[0;32m    265\u001b[0m parallel \u001b[39m=\u001b[39m Parallel(n_jobs\u001b[39m=\u001b[39mn_jobs, verbose\u001b[39m=\u001b[39mverbose, pre_dispatch\u001b[39m=\u001b[39mpre_dispatch)\n\u001b[1;32m--> 266\u001b[0m results \u001b[39m=\u001b[39m parallel(\n\u001b[0;32m    267\u001b[0m     delayed(_fit_and_score)(\n\u001b[0;32m    268\u001b[0m         clone(estimator),\n\u001b[0;32m    269\u001b[0m         X,\n\u001b[0;32m    270\u001b[0m         y,\n\u001b[0;32m    271\u001b[0m         scorers,\n\u001b[0;32m    272\u001b[0m         train,\n\u001b[0;32m    273\u001b[0m         test,\n\u001b[0;32m    274\u001b[0m         verbose,\n\u001b[0;32m    275\u001b[0m         \u001b[39mNone\u001b[39;49;00m,\n\u001b[0;32m    276\u001b[0m         fit_params,\n\u001b[0;32m    277\u001b[0m         return_train_score\u001b[39m=\u001b[39;49mreturn_train_score,\n\u001b[0;32m    278\u001b[0m         return_times\u001b[39m=\u001b[39;49m\u001b[39mTrue\u001b[39;49;00m,\n\u001b[0;32m    279\u001b[0m         return_estimator\u001b[39m=\u001b[39;49mreturn_estimator,\n\u001b[0;32m    280\u001b[0m         error_score\u001b[39m=\u001b[39;49merror_score,\n\u001b[0;32m    281\u001b[0m     )\n\u001b[0;32m    282\u001b[0m     \u001b[39mfor\u001b[39;49;00m train, test \u001b[39min\u001b[39;49;00m cv\u001b[39m.\u001b[39;49msplit(X, y, groups)\n\u001b[0;32m    283\u001b[0m )\n\u001b[0;32m    285\u001b[0m _warn_or_raise_about_fit_failures(results, error_score)\n\u001b[0;32m    287\u001b[0m \u001b[39m# For callabe scoring, the return type is only know after calling. If the\u001b[39;00m\n\u001b[0;32m    288\u001b[0m \u001b[39m# return type is a dictionary, the error scores can now be inserted with\u001b[39;00m\n\u001b[0;32m    289\u001b[0m \u001b[39m# the correct key.\u001b[39;00m\n",
      "File \u001b[1;32mc:\\Python310\\lib\\site-packages\\joblib\\parallel.py:1046\u001b[0m, in \u001b[0;36mParallel.__call__\u001b[1;34m(self, iterable)\u001b[0m\n\u001b[0;32m   1043\u001b[0m \u001b[39mif\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mdispatch_one_batch(iterator):\n\u001b[0;32m   1044\u001b[0m     \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_iterating \u001b[39m=\u001b[39m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_original_iterator \u001b[39mis\u001b[39;00m \u001b[39mnot\u001b[39;00m \u001b[39mNone\u001b[39;00m\n\u001b[1;32m-> 1046\u001b[0m \u001b[39mwhile\u001b[39;00m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49mdispatch_one_batch(iterator):\n\u001b[0;32m   1047\u001b[0m     \u001b[39mpass\u001b[39;00m\n\u001b[0;32m   1049\u001b[0m \u001b[39mif\u001b[39;00m pre_dispatch \u001b[39m==\u001b[39m \u001b[39m\"\u001b[39m\u001b[39mall\u001b[39m\u001b[39m\"\u001b[39m \u001b[39mor\u001b[39;00m n_jobs \u001b[39m==\u001b[39m \u001b[39m1\u001b[39m:\n\u001b[0;32m   1050\u001b[0m     \u001b[39m# The iterable was consumed all at once by the above for loop.\u001b[39;00m\n\u001b[0;32m   1051\u001b[0m     \u001b[39m# No need to wait for async callbacks to trigger to\u001b[39;00m\n\u001b[0;32m   1052\u001b[0m     \u001b[39m# consumption.\u001b[39;00m\n",
      "File \u001b[1;32mc:\\Python310\\lib\\site-packages\\joblib\\parallel.py:861\u001b[0m, in \u001b[0;36mParallel.dispatch_one_batch\u001b[1;34m(self, iterator)\u001b[0m\n\u001b[0;32m    859\u001b[0m     \u001b[39mreturn\u001b[39;00m \u001b[39mFalse\u001b[39;00m\n\u001b[0;32m    860\u001b[0m \u001b[39melse\u001b[39;00m:\n\u001b[1;32m--> 861\u001b[0m     \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49m_dispatch(tasks)\n\u001b[0;32m    862\u001b[0m     \u001b[39mreturn\u001b[39;00m \u001b[39mTrue\u001b[39;00m\n",
      "File \u001b[1;32mc:\\Python310\\lib\\site-packages\\joblib\\parallel.py:779\u001b[0m, in \u001b[0;36mParallel._dispatch\u001b[1;34m(self, batch)\u001b[0m\n\u001b[0;32m    777\u001b[0m \u001b[39mwith\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_lock:\n\u001b[0;32m    778\u001b[0m     job_idx \u001b[39m=\u001b[39m \u001b[39mlen\u001b[39m(\u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_jobs)\n\u001b[1;32m--> 779\u001b[0m     job \u001b[39m=\u001b[39m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49m_backend\u001b[39m.\u001b[39;49mapply_async(batch, callback\u001b[39m=\u001b[39;49mcb)\n\u001b[0;32m    780\u001b[0m     \u001b[39m# A job can complete so quickly than its callback is\u001b[39;00m\n\u001b[0;32m    781\u001b[0m     \u001b[39m# called before we get here, causing self._jobs to\u001b[39;00m\n\u001b[0;32m    782\u001b[0m     \u001b[39m# grow. To ensure correct results ordering, .insert is\u001b[39;00m\n\u001b[0;32m    783\u001b[0m     \u001b[39m# used (rather than .append) in the following line\u001b[39;00m\n\u001b[0;32m    784\u001b[0m     \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_jobs\u001b[39m.\u001b[39minsert(job_idx, job)\n",
      "File \u001b[1;32mc:\\Python310\\lib\\site-packages\\joblib\\_parallel_backends.py:208\u001b[0m, in \u001b[0;36mSequentialBackend.apply_async\u001b[1;34m(self, func, callback)\u001b[0m\n\u001b[0;32m    206\u001b[0m \u001b[39mdef\u001b[39;00m \u001b[39mapply_async\u001b[39m(\u001b[39mself\u001b[39m, func, callback\u001b[39m=\u001b[39m\u001b[39mNone\u001b[39;00m):\n\u001b[0;32m    207\u001b[0m     \u001b[39m\"\"\"Schedule a func to be run\"\"\"\u001b[39;00m\n\u001b[1;32m--> 208\u001b[0m     result \u001b[39m=\u001b[39m ImmediateResult(func)\n\u001b[0;32m    209\u001b[0m     \u001b[39mif\u001b[39;00m callback:\n\u001b[0;32m    210\u001b[0m         callback(result)\n",
      "File \u001b[1;32mc:\\Python310\\lib\\site-packages\\joblib\\_parallel_backends.py:572\u001b[0m, in \u001b[0;36mImmediateResult.__init__\u001b[1;34m(self, batch)\u001b[0m\n\u001b[0;32m    569\u001b[0m \u001b[39mdef\u001b[39;00m \u001b[39m__init__\u001b[39m(\u001b[39mself\u001b[39m, batch):\n\u001b[0;32m    570\u001b[0m     \u001b[39m# Don't delay the application, to avoid keeping the input\u001b[39;00m\n\u001b[0;32m    571\u001b[0m     \u001b[39m# arguments in memory\u001b[39;00m\n\u001b[1;32m--> 572\u001b[0m     \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mresults \u001b[39m=\u001b[39m batch()\n",
      "File \u001b[1;32mc:\\Python310\\lib\\site-packages\\joblib\\parallel.py:262\u001b[0m, in \u001b[0;36mBatchedCalls.__call__\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m    258\u001b[0m \u001b[39mdef\u001b[39;00m \u001b[39m__call__\u001b[39m(\u001b[39mself\u001b[39m):\n\u001b[0;32m    259\u001b[0m     \u001b[39m# Set the default nested backend to self._backend but do not set the\u001b[39;00m\n\u001b[0;32m    260\u001b[0m     \u001b[39m# change the default number of processes to -1\u001b[39;00m\n\u001b[0;32m    261\u001b[0m     \u001b[39mwith\u001b[39;00m parallel_backend(\u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_backend, n_jobs\u001b[39m=\u001b[39m\u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_n_jobs):\n\u001b[1;32m--> 262\u001b[0m         \u001b[39mreturn\u001b[39;00m [func(\u001b[39m*\u001b[39margs, \u001b[39m*\u001b[39m\u001b[39m*\u001b[39mkwargs)\n\u001b[0;32m    263\u001b[0m                 \u001b[39mfor\u001b[39;00m func, args, kwargs \u001b[39min\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mitems]\n",
      "File \u001b[1;32mc:\\Python310\\lib\\site-packages\\joblib\\parallel.py:262\u001b[0m, in \u001b[0;36m<listcomp>\u001b[1;34m(.0)\u001b[0m\n\u001b[0;32m    258\u001b[0m \u001b[39mdef\u001b[39;00m \u001b[39m__call__\u001b[39m(\u001b[39mself\u001b[39m):\n\u001b[0;32m    259\u001b[0m     \u001b[39m# Set the default nested backend to self._backend but do not set the\u001b[39;00m\n\u001b[0;32m    260\u001b[0m     \u001b[39m# change the default number of processes to -1\u001b[39;00m\n\u001b[0;32m    261\u001b[0m     \u001b[39mwith\u001b[39;00m parallel_backend(\u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_backend, n_jobs\u001b[39m=\u001b[39m\u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_n_jobs):\n\u001b[1;32m--> 262\u001b[0m         \u001b[39mreturn\u001b[39;00m [func(\u001b[39m*\u001b[39margs, \u001b[39m*\u001b[39m\u001b[39m*\u001b[39mkwargs)\n\u001b[0;32m    263\u001b[0m                 \u001b[39mfor\u001b[39;00m func, args, kwargs \u001b[39min\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mitems]\n",
      "File \u001b[1;32mc:\\Python310\\lib\\site-packages\\sklearn\\utils\\fixes.py:117\u001b[0m, in \u001b[0;36m_FuncWrapper.__call__\u001b[1;34m(self, *args, **kwargs)\u001b[0m\n\u001b[0;32m    115\u001b[0m \u001b[39mdef\u001b[39;00m \u001b[39m__call__\u001b[39m(\u001b[39mself\u001b[39m, \u001b[39m*\u001b[39margs, \u001b[39m*\u001b[39m\u001b[39m*\u001b[39mkwargs):\n\u001b[0;32m    116\u001b[0m     \u001b[39mwith\u001b[39;00m config_context(\u001b[39m*\u001b[39m\u001b[39m*\u001b[39m\u001b[39mself\u001b[39m\u001b[39m.\u001b[39mconfig):\n\u001b[1;32m--> 117\u001b[0m         \u001b[39mreturn\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mfunction(\u001b[39m*\u001b[39margs, \u001b[39m*\u001b[39m\u001b[39m*\u001b[39mkwargs)\n",
      "File \u001b[1;32mc:\\Python310\\lib\\site-packages\\sklearn\\model_selection\\_validation.py:686\u001b[0m, in \u001b[0;36m_fit_and_score\u001b[1;34m(estimator, X, y, scorer, train, test, verbose, parameters, fit_params, return_train_score, return_parameters, return_n_test_samples, return_times, return_estimator, split_progress, candidate_progress, error_score)\u001b[0m\n\u001b[0;32m    684\u001b[0m         estimator\u001b[39m.\u001b[39mfit(X_train, \u001b[39m*\u001b[39m\u001b[39m*\u001b[39mfit_params)\n\u001b[0;32m    685\u001b[0m     \u001b[39melse\u001b[39;00m:\n\u001b[1;32m--> 686\u001b[0m         estimator\u001b[39m.\u001b[39mfit(X_train, y_train, \u001b[39m*\u001b[39m\u001b[39m*\u001b[39mfit_params)\n\u001b[0;32m    688\u001b[0m \u001b[39mexcept\u001b[39;00m \u001b[39mException\u001b[39;00m:\n\u001b[0;32m    689\u001b[0m     \u001b[39m# Note fit time as time until error\u001b[39;00m\n\u001b[0;32m    690\u001b[0m     fit_time \u001b[39m=\u001b[39m time\u001b[39m.\u001b[39mtime() \u001b[39m-\u001b[39m start_time\n",
      "File \u001b[1;32mc:\\Python310\\lib\\site-packages\\sklearn\\pipeline.py:382\u001b[0m, in \u001b[0;36mPipeline.fit\u001b[1;34m(self, X, y, **fit_params)\u001b[0m\n\u001b[0;32m    380\u001b[0m     \u001b[39mif\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_final_estimator \u001b[39m!=\u001b[39m \u001b[39m\"\u001b[39m\u001b[39mpassthrough\u001b[39m\u001b[39m\"\u001b[39m:\n\u001b[0;32m    381\u001b[0m         fit_params_last_step \u001b[39m=\u001b[39m fit_params_steps[\u001b[39mself\u001b[39m\u001b[39m.\u001b[39msteps[\u001b[39m-\u001b[39m\u001b[39m1\u001b[39m][\u001b[39m0\u001b[39m]]\n\u001b[1;32m--> 382\u001b[0m         \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_final_estimator\u001b[39m.\u001b[39mfit(Xt, y, \u001b[39m*\u001b[39m\u001b[39m*\u001b[39mfit_params_last_step)\n\u001b[0;32m    384\u001b[0m \u001b[39mreturn\u001b[39;00m \u001b[39mself\u001b[39m\n",
      "File \u001b[1;32mc:\\Python310\\lib\\site-packages\\sklearn\\ensemble\\_gb.py:668\u001b[0m, in \u001b[0;36mBaseGradientBoosting.fit\u001b[1;34m(self, X, y, sample_weight, monitor)\u001b[0m\n\u001b[0;32m    665\u001b[0m     \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_resize_state()\n\u001b[0;32m    667\u001b[0m \u001b[39m# fit the boosting stages\u001b[39;00m\n\u001b[1;32m--> 668\u001b[0m n_stages \u001b[39m=\u001b[39m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49m_fit_stages(\n\u001b[0;32m    669\u001b[0m     X,\n\u001b[0;32m    670\u001b[0m     y,\n\u001b[0;32m    671\u001b[0m     raw_predictions,\n\u001b[0;32m    672\u001b[0m     sample_weight,\n\u001b[0;32m    673\u001b[0m     \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49m_rng,\n\u001b[0;32m    674\u001b[0m     X_val,\n\u001b[0;32m    675\u001b[0m     y_val,\n\u001b[0;32m    676\u001b[0m     sample_weight_val,\n\u001b[0;32m    677\u001b[0m     begin_at_stage,\n\u001b[0;32m    678\u001b[0m     monitor,\n\u001b[0;32m    679\u001b[0m )\n\u001b[0;32m    681\u001b[0m \u001b[39m# change shape of arrays after fit (early-stopping or additional ests)\u001b[39;00m\n\u001b[0;32m    682\u001b[0m \u001b[39mif\u001b[39;00m n_stages \u001b[39m!=\u001b[39m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mestimators_\u001b[39m.\u001b[39mshape[\u001b[39m0\u001b[39m]:\n",
      "File \u001b[1;32mc:\\Python310\\lib\\site-packages\\sklearn\\ensemble\\_gb.py:745\u001b[0m, in \u001b[0;36mBaseGradientBoosting._fit_stages\u001b[1;34m(self, X, y, raw_predictions, sample_weight, random_state, X_val, y_val, sample_weight_val, begin_at_stage, monitor)\u001b[0m\n\u001b[0;32m    738\u001b[0m     old_oob_score \u001b[39m=\u001b[39m loss_(\n\u001b[0;32m    739\u001b[0m         y[\u001b[39m~\u001b[39msample_mask],\n\u001b[0;32m    740\u001b[0m         raw_predictions[\u001b[39m~\u001b[39msample_mask],\n\u001b[0;32m    741\u001b[0m         sample_weight[\u001b[39m~\u001b[39msample_mask],\n\u001b[0;32m    742\u001b[0m     )\n\u001b[0;32m    744\u001b[0m \u001b[39m# fit next stage of trees\u001b[39;00m\n\u001b[1;32m--> 745\u001b[0m raw_predictions \u001b[39m=\u001b[39m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49m_fit_stage(\n\u001b[0;32m    746\u001b[0m     i,\n\u001b[0;32m    747\u001b[0m     X,\n\u001b[0;32m    748\u001b[0m     y,\n\u001b[0;32m    749\u001b[0m     raw_predictions,\n\u001b[0;32m    750\u001b[0m     sample_weight,\n\u001b[0;32m    751\u001b[0m     sample_mask,\n\u001b[0;32m    752\u001b[0m     random_state,\n\u001b[0;32m    753\u001b[0m     X_csc,\n\u001b[0;32m    754\u001b[0m     X_csr,\n\u001b[0;32m    755\u001b[0m )\n\u001b[0;32m    757\u001b[0m \u001b[39m# track deviance (= loss)\u001b[39;00m\n\u001b[0;32m    758\u001b[0m \u001b[39mif\u001b[39;00m do_oob:\n",
      "File \u001b[1;32mc:\\Python310\\lib\\site-packages\\sklearn\\ensemble\\_gb.py:247\u001b[0m, in \u001b[0;36mBaseGradientBoosting._fit_stage\u001b[1;34m(self, i, X, y, raw_predictions, sample_weight, sample_mask, random_state, X_csc, X_csr)\u001b[0m\n\u001b[0;32m    244\u001b[0m     sample_weight \u001b[39m=\u001b[39m sample_weight \u001b[39m*\u001b[39m sample_mask\u001b[39m.\u001b[39mastype(np\u001b[39m.\u001b[39mfloat64)\n\u001b[0;32m    246\u001b[0m X \u001b[39m=\u001b[39m X_csr \u001b[39mif\u001b[39;00m X_csr \u001b[39mis\u001b[39;00m \u001b[39mnot\u001b[39;00m \u001b[39mNone\u001b[39;00m \u001b[39melse\u001b[39;00m X\n\u001b[1;32m--> 247\u001b[0m tree\u001b[39m.\u001b[39;49mfit(X, residual, sample_weight\u001b[39m=\u001b[39;49msample_weight, check_input\u001b[39m=\u001b[39;49m\u001b[39mFalse\u001b[39;49;00m)\n\u001b[0;32m    249\u001b[0m \u001b[39m# update tree leaves\u001b[39;00m\n\u001b[0;32m    250\u001b[0m loss\u001b[39m.\u001b[39mupdate_terminal_regions(\n\u001b[0;32m    251\u001b[0m     tree\u001b[39m.\u001b[39mtree_,\n\u001b[0;32m    252\u001b[0m     X,\n\u001b[1;32m   (...)\u001b[0m\n\u001b[0;32m    259\u001b[0m     k\u001b[39m=\u001b[39mk,\n\u001b[0;32m    260\u001b[0m )\n",
      "File \u001b[1;32mc:\\Python310\\lib\\site-packages\\sklearn\\tree\\_classes.py:1342\u001b[0m, in \u001b[0;36mDecisionTreeRegressor.fit\u001b[1;34m(self, X, y, sample_weight, check_input)\u001b[0m\n\u001b[0;32m   1313\u001b[0m \u001b[39mdef\u001b[39;00m \u001b[39mfit\u001b[39m(\u001b[39mself\u001b[39m, X, y, sample_weight\u001b[39m=\u001b[39m\u001b[39mNone\u001b[39;00m, check_input\u001b[39m=\u001b[39m\u001b[39mTrue\u001b[39;00m):\n\u001b[0;32m   1314\u001b[0m     \u001b[39m\"\"\"Build a decision tree regressor from the training set (X, y).\u001b[39;00m\n\u001b[0;32m   1315\u001b[0m \n\u001b[0;32m   1316\u001b[0m \u001b[39m    Parameters\u001b[39;00m\n\u001b[1;32m   (...)\u001b[0m\n\u001b[0;32m   1339\u001b[0m \u001b[39m        Fitted estimator.\u001b[39;00m\n\u001b[0;32m   1340\u001b[0m \u001b[39m    \"\"\"\u001b[39;00m\n\u001b[1;32m-> 1342\u001b[0m     \u001b[39msuper\u001b[39;49m()\u001b[39m.\u001b[39;49mfit(\n\u001b[0;32m   1343\u001b[0m         X,\n\u001b[0;32m   1344\u001b[0m         y,\n\u001b[0;32m   1345\u001b[0m         sample_weight\u001b[39m=\u001b[39;49msample_weight,\n\u001b[0;32m   1346\u001b[0m         check_input\u001b[39m=\u001b[39;49mcheck_input,\n\u001b[0;32m   1347\u001b[0m     )\n\u001b[0;32m   1348\u001b[0m     \u001b[39mreturn\u001b[39;00m \u001b[39mself\u001b[39m\n",
      "File \u001b[1;32mc:\\Python310\\lib\\site-packages\\sklearn\\tree\\_classes.py:458\u001b[0m, in \u001b[0;36mBaseDecisionTree.fit\u001b[1;34m(self, X, y, sample_weight, check_input)\u001b[0m\n\u001b[0;32m    447\u001b[0m \u001b[39melse\u001b[39;00m:\n\u001b[0;32m    448\u001b[0m     builder \u001b[39m=\u001b[39m BestFirstTreeBuilder(\n\u001b[0;32m    449\u001b[0m         splitter,\n\u001b[0;32m    450\u001b[0m         min_samples_split,\n\u001b[1;32m   (...)\u001b[0m\n\u001b[0;32m    455\u001b[0m         \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mmin_impurity_decrease,\n\u001b[0;32m    456\u001b[0m     )\n\u001b[1;32m--> 458\u001b[0m builder\u001b[39m.\u001b[39;49mbuild(\u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49mtree_, X, y, sample_weight)\n\u001b[0;32m    460\u001b[0m \u001b[39mif\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mn_outputs_ \u001b[39m==\u001b[39m \u001b[39m1\u001b[39m \u001b[39mand\u001b[39;00m is_classifier(\u001b[39mself\u001b[39m):\n\u001b[0;32m    461\u001b[0m     \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mn_classes_ \u001b[39m=\u001b[39m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39mn_classes_[\u001b[39m0\u001b[39m]\n",
      "\u001b[1;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "gbc_pipeline = Pipeline([(\"classifier\", GradientBoostingClassifier(n_estimators=10))])\n",
    "train_classifiers(gbc_pipeline, X_train, y_train.ravel(), cv, \"GradientBoostingClassifier\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "eval(gbc_pipeline, X_test, y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## XGBoost for MNIST multiclass classification"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from xgboost import XGBClassifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "xbc_pipeline = Pipeline([(\"classifier\", XGBClassifier())])\n",
    "train_classifiers(xbc_pipeline, X_train, y_train.ravel(), cv, \"XGBoostClassifier\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "eval(abc_pipeline, X_test, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.10.2 64-bit",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.2"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "369f2c481f4da34e4445cda3fffd2e751bd1c4d706f27375911949ba6bb62e1c"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}