Approach 4 e0140017Y.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "deca9ad1",
   "metadata": {},
   "source": [
    "Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 182,
   "id": "26041c2a",
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "\n",
    "from __future__ import print_function\n",
    "from __future__ import division\n",
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "from matplotlib import pyplot as plt\n",
    "import seaborn as sns\n",
    "\n",
    "from sklearn.model_selection import train_test_split\n",
    "import statsmodels.api as sm\n",
    "\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.impute import KNNImputer\n",
    "from sklearn.feature_selection import SelectKBest\n",
    "from sklearn.pipeline import Pipeline\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.decomposition import PCA\n",
    "from sklearn.cluster import KMeans\n",
    "from sklearn.ensemble import GradientBoostingRegressor\n",
    "from sklearn.metrics import mean_absolute_error\n",
    "from sklearn.model_selection import KFold, TimeSeriesSplit\n",
    "from scipy import sparse\n",
    "\n",
    "from ax.plot.contour import plot_contour\n",
    "from ax.plot.trace import optimization_trace_single_method\n",
    "from ax import optimize\n",
    "from ax.utils.notebook.plotting import render, init_notebook_plotting\n",
    "\n",
    "from warnings import filterwarnings\n",
    "filterwarnings('ignore')\n",
    "\n",
    "rng = 0"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "358d417d",
   "metadata": {},
   "source": [
    "Segregation of data and viewing some information about these datas."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 183,
   "id": "d2d7f818",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "San Juan\n",
      "features:  (936, 21)\n",
      "labels  :  (936, 1)\n",
      "test:  (260, 21)\n",
      "\n",
      "Iquitos\n",
      "features:  (520, 21)\n",
      "labels  :  (520, 1)\n",
      "test:  (156, 21)\n",
      "\n",
      " San Juan Unique index values\n",
      "Int64Index([1990, 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000,\n",
      "            2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008],\n",
      "           dtype='int64', name='year')\n",
      "\n",
      " Iquitos Unique index values\n",
      "Int64Index([2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009, 2010], dtype='int64', name='year')\n",
      "\n",
      " San Juan Unique index values\n",
      "       total_cases\n",
      "count   936.000000\n",
      "mean     34.180556\n",
      "std      51.381372\n",
      "min       0.000000\n",
      "25%       9.000000\n",
      "50%      19.000000\n",
      "75%      37.000000\n",
      "max     461.000000\n",
      "\n",
      " Iquitos unique index values\n",
      "       total_cases\n",
      "count   520.000000\n",
      "mean      7.565385\n",
      "std      10.765478\n",
      "min       0.000000\n",
      "25%       1.000000\n",
      "50%       5.000000\n",
      "75%       9.000000\n",
      "max     116.000000\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Text(0.5, 0, 'Time')"
      ]
     },
     "execution_count": 183,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABIQAAAJcCAYAAACIb39OAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8rg+JYAAAACXBIWXMAAAsTAAALEwEAmpwYAAEAAElEQVR4nOzdeZxkd13v//f3nNp6m30yS2ZIJiRAtiFg2AUkgBGUG0QEuRARBUVk8Xp/VxBRuV7RwOWKImICgkIQAYlISBQCEohAEgiQjezJTDIzmczWe3dt55zv74+zVFVPdfep6q7qqu7X8/HIY7qrt5OZruW8z2cx1loBAAAAAABg7XBW+gAAAAAAAADQXQRCAAAAAAAAawyBEAAAAAAAwBpDIAQAAAAAALDGEAgBAAAAAACsMQRCAAAAAAAAawyBEAAA6GnGmMcZY6aNMe5KH0s9Y8w/GmP+bKWPoxuMMe82xvz9Sh8HAABYPgRCAACsAcaYrxpj/rTJ7ZcYYx4zxmQ69HNPN8bYVr6/MWa/MeZF8fvW2kestcPWWr8Dx2eNMWcu9/ftN8aYy6PQbdoYUzHGVOve/w9r7Z9ba9+40scJAACWD4EQAABrw6ckvc4YY+bcfqmkf7LWeitwTFghcwM6a+2bo9BtWNKfS/p8/L619iUrc5QAAKCTCIQAAFgb/k3SZknPjW8wxmyU9AuSPm2McYwx7zLGPGiMOWGM+YIxZlPd5/6qMebh6GN/VF/Fs8jX3hD9OR5VmzzLGPN4Y8w3o889boz5J2PMhuh7XSnpcZK+En3+78+tMjLG7DTGXG2MGTXGPGCMeVPdcb43+vmfNsZMGWN+Yoy5MM1f0GJfa4x5ijHmR9HHPi+pMOfrf8EYc6sxZtwY8z1jzN7o9lcbY/YZY9ZF778kqsraOs9x/LfoZ48bY75ljDk7uv2dxpgvzvncvzbGfDh6e70x5hPGmMPGmEPGmD+L2+yMMb9mjPmuMeZDxpgTkt6b5u9kzt/NZ6K343+PNxhjDhhjxowxbzbGPM0Yc3t03B+Z8/W/boy5O/rcrxljTmvl5wMAgOVHIAQAwBpgrS1K+oKkX627+VWS7rHW3ibpbZJeLun5knZKGpP0t5JkjDlH0kclvVbSDknrJZ1a933m/VpJz4v+3BBVm9woyUj6i+hzz5a0W1FAYa29VNIjkl4Wff4HmvzvfE7SwejrXynpz40xF9V9/L9Fn7NB0tWSPjL3Gyyg6dcaY3IKQ7UrJW2S9C+Sfin+ImPMUyR9UtJvKQzerpB0tTEmb639vKTvSfqwMWazpE9IeqO19tjcH26MeYKkf5b0u5K2Svp3heFYLjqulxpjRqLPdRX+G342+vJ/lORJOlPSUyT9rKT6Nq9nSHpI0jZJ72vh72Q+z5B0lqRXS/orSX8o6UWSzpX0KmPM86PjvETSuyW9Ivp/+q/o/xEAAKwgAiEAANaOT0l6pTEmrmz51eg2SXqzpD+01h601pYVBjSvjKpyXinpK9ba71hrK5L+WJKt+74Lfe1JrLUPWGu/bq0tR6HIXyoMkxZljNkt6TmS3mmtLVlrb5X092oMur5jrf33aObQlZKenOZ7L/K1z5SUlfRX1tqqtfaLkn5Q93W/KekKa+3N1lrfWvspSeXo6yTpdyRdJOlbCv8ur5nn579a0rXR309V0gclDUh6trX2YUk/kvSL0edeJGnWWnuTMWabpJdK+l1r7Yy19qikD0n6lbrv/ai19m+stV4UEC7V/4n+Da6TNCPpn621R621hxSGPk+JPu/Nkv7CWnt31Jr455IuoEoIAICV1ZEBkgAAoPdYa79jjDku6eXGmB9IerrCqg1JOk3Sl4wxQd2X+AqrSXZKOlD3fWajtiOl+NqTROHFXytsXxtReIFqLOX/xk5Jo9baqbrbHpZU3xb2WN3bs5IKxphMyjlJTb82+rmHrLX1QdjDdW+fJun1xpi31d2Wi75O1tpxY8y/SPo91VUWNbGz/vtaawNjzAHVKrI+K+k1kj4t6b+rVh10msLA6rCpjYlyVPfvNuft5XCk7u1ik/eH647tr40x/6/u40bh/1P93yEAAOgiKoQAAFhbPq2wmuZ1kr5mrY1P4g9Ieom1dkPdf4Wo2uOwpF3xNzDGDChsi1KKr60PUGJ/Ht1+vrV2XXQs9cOum31N7FFJm+K2qcjjJB1K8f++FIclnWpMw1Dux9W9fUDS++b8HQxaa/9ZkowxF0j6dYWtUh9e4Oc8qjBAUfR1RmFLXfz/9y+SfsYYs0thpVAcCB1QWJG0pe7nr7PWnlv3vRf6e+2kA5J+a87fzYC19nsrdDwAAEAEQgAArDWfVjjn5U2qtYtJ0uWS3he38RhjtkazXyTpi5JeZox5djTL5r1qDHAW+tpjkgJJZ9R9/oikaUkTxphTJf2vOcd4ZM7nJ6y1BxTO4/kLY0whGtz8G5I+k/L/v103KpzP83ZjTNYY8wqFFVaxj0t6szHmGSY0ZIz5eWPMSNSi9xmFc3TeoDBYess8P+cLkn7eGPNCY0xW0v9UGPR8T5KiFrtvSfoHSfustXdHtx+WdJ2k/2eMWWfCQd+Pj+f4rLDLJf2BMeZcKRl+/csrfEwAAKx5BEIAAKwh1tr9CsOFIYVDk2N/Hb1/nTFmStJNCocGy1r7E4WDoz+nsFJmWtJRhUHFYl87q3CA8Xej7VPPlPS/JT1V0oSkayX965zD/AtJ74k+//9r8r/xGkmnK6ym+ZKkP7HWfqONv47UotlJr5D0a5JGFc76+de6j9+iMGT7iML2tweiz5XC/58D1tq/i2YsvU7Snxljzmryc+6NPv43ko5LepnCAduVuk/7rMJQ77NzvvxXFbap3RUdwxcVDgFfUdbaL0l6v6TPGWMmJd0piVX2AACsMNPYCg8AALAwY8ywpHFJZ1lr963w4QAAAKANVAgBAIBFGWNeZowZNMYMKdx8dYek/St7VAAAAGgXgRAAAEjjEoUtWo9KOkvSr1jKjAEAAPoWLWMAAAAAAABrDBVCAAAAAAAAa0xmpQ9AkrZs2WJPP/30lT4MAAAAAACAVeOHP/zhcWvt1mYf64lA6PTTT9ctt9yy0ocBAAAAAACwahhjHp7vY7SMAQAAAAAArDEEQgAAAAAAAGsMgRAAAAAAAMAa0xMzhAAAAAAAwOpSrVZ18OBBlUqllT6UVa9QKGjXrl3KZrOpv4ZACAAAAAAALLuDBw9qZGREp59+uowxK304q5a1VidOnNDBgwe1Z8+e1F9HyxgAAAAAAFh2pVJJmzdvJgzqMGOMNm/e3HIlFoEQAAAAAADoCMKg7mjn75lACAAAAAAAYI0hEAIAAAAAAFhjCIQAAAAAAMCqMz4+ro9+9KMLfs7+/fv12c9+dtHvtX//fp133nnLdWg9gUAIAAAAAACsOssZCK1GrJ0HAAAAAAAd9b+/8hPd9ejksn7Pc3au05+87Nx5P/6ud71LDz74oC644AK9+MUvliT9x3/8h4wxes973qNXv/rVete73qW7775bF1xwgV7/+tfrF3/xF3XppZdqZmZGkvSRj3xEz372sxc9Ft/39c53vlNf/epX5TiO3vSmN+ltb3ub/vRP/1Rf+cpXVCwW9exnP1tXXHGFjDH68Ic/rMsvv1yZTEbnnHOOPve5z2lmZkZve9vbdOedd6pareq9732vLrnkEv3kJz/RG97wBlUqFQVBoKuuukpnnXXWkv/+CIQAAAAAAMCqc9lll+nOO+/UrbfeqquuukqXX365brvtNh0/flxPe9rT9LznPU+XXXaZPvjBD+qaa66RJM3OzurrX/+6CoWC7r//fr3mNa/RLbfcsujP+tjHPqb9+/fr1ltvVSaT0ejoqCTprW99q/74j/9YknTppZfqmmuu0cte9jJddtll2rdvn/L5vMbHxyVJ73vf+3TRRRfpk5/8pMbHx/X0pz9dL3rRi3T55ZfrHe94h1772teqUqnI9/1l+fshEAIAAAAAAB21UCVPN3znO9/Ra17zGrmuq23btun5z3++fvCDH2jdunUNn1etVvXWt75Vt956q1zX1X333Zfq+3/jG9/Qm9/8ZmUyYcyyadMmSdL111+vD3zgA5qdndXo6KjOPfdcvexlL9PevXv12te+Vi9/+cv18pe/XJJ03XXX6eqrr9YHP/hBSVKpVNIjjzyiZz3rWXrf+96ngwcP6hWveMWyVAdJBEIAAAAAAACSpA996EPatm2bbrvtNgVBoEKh0Pb3KpVKestb3qJbbrlFu3fv1nvf+16VSiVJ0rXXXqsbbrhBX/nKV/S+971Pd9xxh6y1uuqqq/TEJz6x4fucffbZesYznqFrr71WL33pS3XFFVfooosuWtL/p8RQaQAAAAAAsAqNjIxoampKkvTc5z5Xn//85+X7vo4dO6YbbrhBT3/60xs+R5ImJia0Y8cOOY6jK6+8MnV71otf/GJdccUV8jxPkjQ6OpqEP1u2bNH09LS++MUvSpKCINCBAwf0ghe8QO9///s1MTGh6elpXXzxxfqbv/kbWWslST/+8Y8lSQ899JDOOOMMvf3tb9cll1yi22+/fVn+fqgQAgAAAAAAq87mzZv1nOc8R+edd55e8pKXaO/evXryk58sY4w+8IEPaPv27dq8ebNc19WTn/xk/dqv/Zre8pa36Jd+6Zf06U9/Wj/3cz+noaGhVD/rjW98o+677z7t3btX2WxWb3rTm/TWt75Vb3rTm3Teeedp+/btetrTniYpHED9ute9ThMTE7LW6u1vf7s2bNigP/qjP9Lv/u7vau/evQqCQHv27NE111yjL3zhC7ryyiuVzWa1fft2vfvd716Wvx8TJ08r6cILL7RphjQBAAAAAID+cPfdd+vss89e6cNYM5r9fRtjfmitvbDZ59MyBgAAAAAAsMbQMgYAAAAAAJDC1772Nb3zne9suG3Pnj360pe+tEJH1D4CIQAAAAAAgBQuvvhiXXzxxSt9GMuCljEAAAAAAIA1hkAIAFo0Uazq9Z/8vo5Mllb6UAAAAACgLQRCANCi+49M6dv3HdOdhyZW+lAAAAAAoC0EQgDQorIXSJL8wK7wkQAAAABAewiEAKBFlSgQCiyBEAAAANCrxsfH9dGPfnTBz9m/f78++9nPLvq99u/fr/POO2+5Dq0nEAgBQIvKni9J8qgQAgAAAHrWcgZCqxFr5wGgRbSMAQAAAC36j3dJj92xvN9z+/nSSy6b98Pvete79OCDD+qCCy7Qi1/84vAw/uM/ZIzRe97zHr361a/Wu971Lt1999264IIL9PrXv16/+Iu/qEsvvVQzMzOSpI985CN69rOfveih+L6vd77znfrqV78qx3H0pje9SW9729v0p3/6p/rKV76iYrGoZz/72briiitkjNGHP/xhXX755cpkMjrnnHP0uc99TjMzM3rb296mO++8U9VqVe9973t1ySWX6Cc/+Yne8IY3qFKpKAgCXXXVVTrrrLOW/NdHIAQALaJlDAAAAOh9l112me68807deuutuuqqq3T55Zfrtttu0/Hjx/W0pz1Nz3ve83TZZZfpgx/8oK655hpJ0uzsrL7+9a+rUCjo/vvv12te8xrdcssti/6sj33sY9q/f79uvfVWZTIZjY6OSpLe+ta36o//+I8lSZdeeqmuueYavexlL9Nll12mffv2KZ/Pa3x8XJL0vve9TxdddJE++clPanx8XE9/+tP1ohe9SJdffrne8Y536LWvfa0qlYp831+Wvx8CIQBoUa1CaIUPBAAAAOgXC1TydMN3vvMdveY1r5Hrutq2bZue//zn6wc/+IHWrVvX8HnValVvfetbdeutt8p1Xd13332pvv83vvENvfnNb1YmE8YsmzZtkiRdf/31+sAHPqDZ2VmNjo7q3HPP1cte9jLt3btXr33ta/Xyl79cL3/5yyVJ1113na6++mp98IMflCSVSiU98sgjetaznqX3ve99OnjwoF7xilcsS3WQRCAEAC2rJIEQiRAAAACwmnzoQx/Stm3bdNtttykIAhUKhba/V6lU0lve8hbdcsst2r17t9773veqVCpJkq699lrdcMMN+spXvqL3ve99uuOOO2St1VVXXaUnPvGJDd/n7LPP1jOe8Qxde+21eulLX6orrrhCF1100ZL+PyWGSgNAy6gQAgAAAHrfyMiIpqamJEnPfe5z9fnPf16+7+vYsWO64YYb9PSnP73hcyRpYmJCO3bskOM4uvLKK1O3Z734xS/WFVdcIc/zJEmjo6NJ+LNlyxZNT0/ri1/8oiQpCAIdOHBAL3jBC/T+979fExMTmp6e1sUXX6y/+Zu/kY1GU/z4xz+WJD300EM644wz9Pa3v12XXHKJbr/99mX5+6FCCABalFQIMUMIAAAA6FmbN2/Wc57zHJ133nl6yUteor179+rJT36yjDH6wAc+oO3bt2vz5s1yXVdPfvKT9Wu/9mt6y1veol/6pV/Spz/9af3cz/2choaGUv2sN77xjbrvvvu0d+9eZbNZvelNb9Jb3/pWvelNb9J5552n7du362lPe5qkcAD16173Ok1MTMhaq7e//e3asGGD/uiP/ki/+7u/q7179yoIAu3Zs0fXXHONvvCFL+jKK69UNpvV9u3b9e53v3tZ/n6M7YETmgsvvNCmGdIEAL3gA1+9Rx/91oP63//tXL3+2aev9OEAAAAAPenuu+/W2WefvdKHsWY0+/s2xvzQWnths8+nZQwAWhRXCHmsnQcAAADQp2gZA4AWVaLhQQGBEAAAALCmfO1rX9M73/nOhtv27NmjL33pSyt0RO0jEAKAFpWrzBACAAAA0rDWyhiz0oexbC6++GJdfPHFK30YJ2lnHBAtYwDQorhCyKdCCAAAAJhXoVDQiRMn2gorkJ61VidOnFChUGjp66gQAoAWlb1w9SQtYwAAAMD8du3apYMHD+rYsWMrfSirXqFQ0K5du1r6GgIhAGgRQ6UBAACAxWWzWe3Zs2elDwPzoGUMAFpUjgKhgNJXAAAAAH2KQAgAWhQHQswQAgAAANCvCIQAoEVxyxhbxgAAAAD0KwIhAGhRUiHkEwgBAAAA6E8EQgDQokq0ZYwKIQAAAAD9ikAIAFqUDJVmhhAAAACAPkUgBAAtYoYQAAAAgH5HIAQALar4bBkDAAAA0N8IhACgReUqgRAAAACA/kYgBAAtqlUIrfCBAAAAAECbCIQAoAWeHySVQQEzhAAAAAD0KQIhAGhBpa4syKNlDAAAAECfIhACgBbEG8Yk1s4DAAAA6F8EQgDQgnJdIMRQaQAAAAD9ikAIAFpQXyHkM0MIAAAAQJ8iEAKAFpQ9P3mbCiEAAAAA/YpACABaQMsYAAAAgNWAQAgAWtAwVJqWMQAAAAB9ikAIAFpAhRAAAACA1YBACABaEFcIZV0jj0AIAAAAQJ8iEAKAFsQVQoWsq4BACAAAAECfIhACgBbEW8aGchnWzgMAAADoWwRCANCCcjWsEBrMUyEEAAAAoH8RCAFAC0pRhdBgzmWGEAAAAIC+RSAEAC1IKoRyGbaMAQAAAOhbBEIA0IJ4qPRgzlXADCEAAAAAfYpACABaUKrWWsaoEAIAAADQrwiEAKAFZS9QznWUcRwCIQAAAAB9i0AIAFpQ9nzls45cx7B2HgAAAEDfIhACgBaUqoHyGVeOMQqClT4aAAAAAGgPgRAAtKDs+cpnHGUcQ8sYAAAAgL5FIAQALSh7gQpZR45j5BEIAQAAAOhTBEIA0IJy1Vc+48p1xNp5AAAAAH2LQAgAWlD2gnCotKFlDAAAAED/IhACgBaUqr4KGVeOYxQQCAEAAADoUwRCANCCuEIowwwhAAAAAH2MQAgAWlCuBkmFkF83Q2hitqpS1V/BIwMAAACA9AiEAKAFJc9PZgjVt4y97hM36/9+7d4VPDIAAAAASI9ACABaUK4GymccuXMqhI5NlXVsqryCRwYAAAAA6REIAUALyp6vQtaV6xhZq6RKyAsCVf1ghY8OAAAAANIhEAKAFpTiCiFjJCmpEvICSyAEAAAAoG8QCAFAStZalT1f+WiotCT5cYWQb1X12ToGAAAAoD8QCAFASl5gFVipkA1nCElSYGkZAwAAANB/Mit9AADQL+K18vmMq6hjTF5dhZBHhRAAAACAPkGFEACkVPbCCqB81pETJUJBYGWtlRdYVagQAgAAANAnUgdCxhjXGPNjY8w10ft7jDE3G2MeMMZ83hiTi27PR+8/EH389A4dOwB0VRwIFTJu0jLmBzaZI0TLGAAAAIB+0UqF0Dsk3V33/vslfchae6akMUm/Ed3+G5LGots/FH0eAPS9pGUs69SGSkfVQZJoGQMAAADQN1IFQsaYXZJ+XtLfR+8bSRdJ+mL0KZ+S9PLo7Uui9xV9/IXR5wNAXytXo5axjKNMPFQ6qM0RokIIAAAAQL9IWyH0V5J+X1J8trNZ0ri11ovePyjp1OjtUyUdkKTo4xPR5zcwxvymMeYWY8wtx44da+/oAaCLyl5cIeTKjXJuLwjkRUEQM4QAAAAA9ItFAyFjzC9IOmqt/eFy/mBr7cestRdaay/cunXrcn5rAOiIUl2FkNOkQoiWMQAAAAD9Is3a+edI+m/GmJdKKkhaJ+mvJW0wxmSiKqBdkg5Fn39I0m5JB40xGUnrJZ1Y9iMHgC5LKoQyrtwoTvdtbd08LWMAAAAA+sWiFULW2j+w1u6y1p4u6VckfdNa+1pJ10t6ZfRpr5f05ejtq6P3FX38m9ZaLpsD6HvJlrGsI9cJHz79wMoLaBkDAAAA0F9a2TI21zsl/Z4x5gGFM4I+Ed3+CUmbo9t/T9K7lnaIANAbki1jmdoMIT+oVQjRMgYAAACgX6RpGUtYa78l6VvR2w9JenqTzylJ+uVlODYA6ClxhVA+49RaxgLLljEAAAAAfWcpFUIAsKYEUfCTcY2cqEIosLWWMS+wokMWAAAAQD8gEAKAlPwo7HGMUcY9uWVMkqq0jQEAAADoAwRCAJBSXCHkmFqFkFfXMibRNgYAAACgP7Q0QwgA1jI/Cn5cx8h1ai1jIhACAAAA0GcIhAAgpbgbzDWmYcuYX/c5tIwBAAAA6AcEQgCQUjww2nFUqxAKbH2BEBVCAAAAAPoCgRAApOTXzRCKAyEvsGHbWIRACAAAAEA/IBACgJTiLWOuY+REgZBvrXy2jAEAAADoM2wZA4CU6reMxTOEAraMAQAAAOhDBEIAkFKc9dRvGfMDKy+ohUAeFUIAAAAA+gCBEACkFM8KcowaAiG/rkKoQoUQAAAAgD5AIAQAKQXWyhjJ1A2V9q1tmBtEyxgAAACAfkAgBAAp+YFNZgc5pr5CiJYxAAAAAP2FQAgAUvKtTbaLZaI/AyqEAAAAAPQhAiEASCmoqxCKW8Y8nxlCAAAAAPoPgRAApBTYcKC0JGXcKBAKbENVEC1jAAAAAPoBgRAApOQH9S1j4cOn5wcNFUK0jAEAAADoBwRCAJBSYG3SKpZzw4fPim/l0TIGAAAAoM8QCAFASvVbxpKWMT+gZQwAAABA3yEQAoCUAqukZSwbVQh5gaVlDAAAAEDfIRACgJSCwCZDpbNRhVDFC1g7DwAAAKDvEAgBQEq+rbWMGWPkOkZeEMgPgiQoqtIyBgAAAKAPEAgBQEpB3ZYxKawSqvpWVd9qIOtKokIIAAAAQH8gEAKAlPy6LWOSlHUcVaO187mMI2MIhAAAAAD0BwIhAEgpsEpaxiQpm3Hk+VZeECjjOsq6Di1jAAAAAPpCZqUPAAD6RRBY1eVByjgmqRDKOkbZ6H0AAAAA6HUEQgCQkh/MaRmLKoICa+W6RtmMQyAEAAAAoC8QCAFASr61cszcodKBrMJ5QhmHljEAAAAA/YFACABSCuZUCGVcR14QyFrJdYxyLi1jAAAAAPoDgRAApBTM3TLmOqp44VyhjOvQMgYAAACgbxAIAUBKvpXMnJYxLwhkorczjpFHyxgAAACAPkAgBAApBYGVW7dlLOuGa+eNCVvGsq6jChVCAAAAAPoAgRAApDR3y1jGMar4gVxjlHUc+RmrskcgBAAAAKD3OSt9AADQL07eMubI8wN5QSDXMdq9aVAPHZtewSMEAAAAgHQIhAAgJdt07byVF1hlXKO9p67XwbGiRmcqK3iUAAAAALA4AiEASOmkljE33Crm+VZZ19H5u9ZLku44NLFShwgAAAAAqRAIAUBKvpWcukAo5zrygrBCyHWMzjs1CoQOjq/QEQIAAABAOgyVBoCU5m4Zy7hGVb+2dn5dIas9W4aoEAIAAADQ8wiEACClk7eMhWvnXWPlOmHB5Z4tQzo4VlypQwQAAACAVAiEACClwFqZuqHSuUy4dt5xpGwUFGUcIz+wK3WIAAAAAJAKgRAApBRYK9fMrRAK5DsmqRzKuARCAAAAAHofQ6UBIKW5LWNZN2wZqwZWGTd8OHUMgRAAAACA3kcgBAApBXO2jGXdsGXMD6yybq1lzCMQAgAAANDjCIQAIKWwZaz2fsYNw5+qHySVQ67jUCEEAAAAoOcRCAFASn5g5ZjGljE/sKp4gbJRyxhDpQEAAAD0AwIhAEgpCOyclrHwIbTsBRrMuZLCljJaxgAAAAD0OgIhAEjJn7NlLFvXPzacD5c2hhVCQdePDQAAAABaQSAEACn5QeNQ6YxTewgdigIhl5YxAAAAAH2AQAgAUrLWyq171KyvECIQAgAAANBPCIQAICXfnjxUOjYUzRBi7TwAAACAfkAgBAApzd0ylnGbt4wFlkAIAAAAQG8jEAKAlILAynUWHirtUiEEAAAAoA8QCAFASr6dGwg1rxCyNgyPAAAAAKBXEQgBQEqB1fwzhPK1GUKSqBICAAAA0NMIhAAgpSCwqisQUqZpy1j4sMocIQAAAAC9jEAIAFI6qWUsCn8cIw1kwwqhuGiICiEAAAAAvYxACABSsNbKntQyFr49lMvIRLfHFUK+TyAEAAAAoHcRCAFACn5U8VNfIRSvnY8HSku1GUI+LWMAAAAAehiBEACkEHeA1c8QyiWBkJvc5iRDpYOuHRsAAAAAtIpACABSiIdEOw0VQuHbw80qhJghBAAAAKCHEQgBQApJy1izGUJ1gZBLIAQAAACgDxAIAUAK8Uyghi1jC80QIhACAAAA0MMIhAAghSAKeOq3jMVDpYebVAixdh4AAABALyMQAoAUmg2VrrWM1YZK0zIGAAAAoB8QCAFACs3WzmcdWsYAAAAA9CcCIQBIYcEtY7laIBS3lBEIAQAAAOhlBEIAkEKzLWPD+Yx+/Tl79MKztyW3xSERM4QAAAAA9LLM4p8CAGhWIWSM0R+/7JyGz3OjNjIqhAAAAAD0MiqEACCFIAj/rN8y1oxLyxgAAACAPkAgBAAp+DYeKr3w59XWzgedPiRg1SlWfH31zsMrfRgAAABrAoEQAKQQV/wsViEUzxAiDwJad91dj+nNn/mRDo0XV/pQAAAAVj0CIQBIIbAnr51vhgohoH2lqi9JKkd/AgAAoHMIhAAghSQQYoYQ0DHxdj7uPwAAAJ1HIAQAKcQnqGaxQMhh7TzQriC633D/AQAA6DwCIQBIIe4AW6xlrDZDiBNaoFVUCAEAAHQPgRAApJB6y5ihQghol0+FEAAAQNcQCAFACmm3jMUVRFQ4AK3zkwohhrIDAAB0GoEQAKRgbcq18074sEogBLQursTzfO4/AAAAnUYgBAApxAHPomvnXSqEgHb5PjOEAAAAuoVACABS8FNWCDFDCGhfUiHE/QcAAKDjCIQAIIW0W8ZqM4SYgQK0ymfLGAAAQNcQCAFACmm3jGUYKg20zWPLGAAAQNcQCAFACkEUCJlFWsYch5YxoF0BW8YAAAC6hkAIAFKIT1TdRbeMUSEEtIsKIQAAgO4hEAKAFFJvGYsDIcsJLdAqZggBAAB0D4EQAKQQpN0yFgdCPie0QKsIhAAAALqHQAgAUojPTxetEGLtPNA21s4DAAB0D4EQAKQQVywskgfJcYwcU6soApBeXFlHhRAAAEDnEQgBQApJy9hiiZCkjONQ4QC0gQohAACA7iEQAoAU/JRbxiTJcahwANqRzBDyWTsPAADQaQRCAJBC2i1jUlQhxFBpoGU+a+cBAAC6hkAIAFKIRwKlaRlzHcMMIaANbBkDAADoHgIhAEjBt+mGSkthIOQFtLwAraJCCAAAoHsIhAAghVZmCLmOocIBaINHhRAAAEDXEAgBQAqtbRkjEALa4UeVdVQIAQAAdB6BEACk0GqFECe0QOviWew+LZcAAAAdRyAEACkELQ6VpkIIaB0VQgAAAN1DIAQAKQRBa0OlCYSA1iVbxnzuPwAAAJ22aCBkjCkYY75vjLnNGPMTY8z/jm7fY4y52RjzgDHm88aYXHR7Pnr/gejjp3f4/wEAOi7eMuYyQwjoGLaMAQAAdE+aCqGypIustU+WdIGknzPGPFPS+yV9yFp7pqQxSb8Rff5vSBqLbv9Q9HkA0Nc8P2xlSRMIOYYZQkA74kAoHuIOAACAzlk0ELKh6ejdbPSflXSRpC9Gt39K0sujty+J3lf08Rcak2IKKwD0sLIXyBgp5y6eo2dcKoSAdlAhBAAA0D2pZggZY1xjzK2Sjkr6uqQHJY1ba73oUw5KOjV6+1RJByQp+viEpM1NvudvGmNuMcbccuzYsSX9TwBAp5WqvgoZV2nybddxCISANsStmcwQAgAA6LxUgZC11rfWXiBpl6SnS3rSUn+wtfZj1toLrbUXbt26danfDgA6quwFKmTTzeF3jQiEgDZ4PhVCAAAA3dLSljFr7bik6yU9S9IGY0wm+tAuSYeitw9J2i1J0cfXSzqxHAcLACulVPWVz7ipPjfjOPKi9dkA0otnB/ncfwAAADouzZaxrcaYDdHbA5JeLOluhcHQK6NPe72kL0dvXx29r+jj37SW6ZAA+lup2kKFkGPE+SzQOo8ZQgAAAF2TWfxTtEPSp4wxrsIA6QvW2muMMXdJ+pwx5s8k/VjSJ6LP/4SkK40xD0galfQrHThuAOiqsuerkE1ZIeQalT2/w0cErD5xqyUtlwAAAJ23aCBkrb1d0lOa3P6QwnlCc28vSfrlZTk6AOgRpWqgfCZdhZBj2DIGtIMtYwAAAN3T0gwhAFirSlVf+bQVQo7hhBZoAxVCAAAA3UMgBAAplL30FUKuQ4UQ0A4qhAAAALqHQAgAUihV088QIhAC2lOrEGIqOwAAQKcRCAFACmUvIBACOsyPlpJ6PvcfAACATiMQAoAUylU/dctYxjHJiS2A9HyfGUIAAADdQiAEACmUvECFbMotY46hwgFoQ1IhRCAEAADQcQRCAJBCqeqrkEm/ZYwKB6B1HlvGAAAAuoZACABSKHuB8ikrhFzHoWUMaENAIAQAANA1BEIAsIiqH8gPLBVCQIdRIQQAANA9BEIAsIhS1ZeklraMeT5rs4FWBHUhkMfaeQAAgI4jEAKARZS98OQ0fcsYFUJAq+oHSXP/AQAA6DwCIQBYRFIh1ErLGDOEgJb4DRVC3H8AAAA6jUAIABZRqrZWIeRQIQS0rD5E5f4DAADQeQRCALCIshdWCOVbqBCiwgFoje9TIQQAANBNBEIAsIi4QqjQwgwhaxuH5AJYGBVCAAAA3UUgBACLKLe4ZSzjGElijhDQgnizWNZlSx8AAEA3EAgBwCKSLWOZ9DOEJKocgFbEm+bzGZf7DgAAQBcQCAHAIkrtVghxUgukFlcI5TIOM4QAAAC6gEAIABYRVwilDYRcJ3xo5aQWSK9WIeQQpgIAAHQBgRAALCKuEErbMuaGBUKc1AItiCuE8lGFkGUGFwAAQEcRCAHAIlptGXPduEKIwbhAWkEUAOWi4JU8FQAAoLMIhABgEbWWsXQPmfEMIfIgIL24xTIOhAhUAQAAOotACAAWUarGrSwpK4RMGAhxQguk5/lRIBRV2HH3AQAA6CwCIQBYRMnzlXWN3KjyZzEuW8aAlsUtY3HwSqAKAADQWQRCALCIihckVQtpZFwCIaBVc1vGuP8AAAB0FoEQACzCD6wyLQRCVAgBrQuCuEIoniHE/QcAAKCTCIQAYBFeEKRuF5PqZwhxQgukRYUQAABAdxEIAcAi/ECtBUJUCAEtq1UIxTOEuP8AAAB0EoEQACzCD4JklXwazBACWndShZDP/QcAAKCTCIQAYBFeYOWY9IGQQ8sY0DLfzp0hxJYxAACATiIQAoBFhEOlW6gQcpiBArQqrgjKM0MIAACgKwiEAGARfmCZIQR02MkVQtx/AAAAOolACAAW4QeWGUJAh8X3l3zWbXgfAAAAnUEgBACLaH+GEDNQgLSSodJu+NKk6nP/AQAA6CQCIQBYRNDyDKHwcwNLhQOQVrx2vpBlhhAAAEA3EAgBwCK8wMp10j9cxjOEPNZmA6l5c1rGKlQIAQAAdBSBEAAswg+sWigQYqg00Ia4QmggCoSqBKoAAAAdRSAEAIvwgiBZJZ9G3DLGliQgPS9pGYsCIY8KIQAAgE4iEAKARQSB2lo7zwwhIL147XytQohACAAAoJMIhABgEV4QtDRUmhlCQOv8KACKh0ozQwgAAKCzCIQAYBF+i2vnmSEEtC7OTwvMEAIAAOgKAiEAWIRvbTIXKI143pBPyxiQmh/EFUK0jAEAAHQDgRAALMLzbVszhBgqDaQX318GcwRCAAAA3UAgBACL8IP2AiGfE1ogNd9vHCpdYcsYAABARxEIAcAifNtmIESBEJDaSWvnuQMBAAB0FIEQACzCD1qdIRQPlabCAUgrrsTLZcKXJrSMAQAAdBaBEAAswvOtHGYIAR3lRYGQ6xg5hkAIAACg0wiEAGARrVYI1WYIEQgBaflBkNzPsq6jCoEQAABARxEIAcAiwhlC6R8uXRPPECIQAtLy6oa351xHVY/7DwAAQCcRCAHAIlqtEHIcI2PCrwOQTv39LJtxaBkDAADoMAIhAFiE5wctbRmTwsHSzBAC0gsrhMKXJVnXEAgBAAB0GIEQACwisGo5EHIdo4BACEjN9y0zhAAAALqIQAgAFuHVDbtNK+M4VAgBLThphhBD2QEAADqKQAgAFuEHra2dlySHGUJAS/yg1pqZdR1VPH+FjwgAAGB1IxACgEV4LQ6VlqSM68gLaHkB0vIahkobKoQAAAA6jEAIABYQBFa2zRlCjEAB0vPrWsayLlvGAAAAOo1ACAAW4NuwSqHVCiHXGPlUCAGpeXMCoYrH/QcAAKCTCIQAYAHxHKBWZwi5rJ0HWuIHVhm3fqg0gRAAAEAnEQgBwALiQKj1GUKsnQdaEVYIhS9Lsi4zhAAAADqNQAgAFhBX+cQnqmlRIQS0xg+C2lBpKoQAAAA6jkAIABYQVwi5rRUIRTOECISAtDy/boZQxlGFQAgAAKCjCIQAYAHx6njXbb1CiEAISM+vWzvPDCEAAIDOIxACgAXEi8LamSFEIASkV79lLOc6qnrcfwAAADqJQAgAFpBUCLWxdp4ZQkB69RVC2YyhQggAAKDDCIQAYAG1GUKtr52nQghIr3HLGDOEAAAAOo1ACAAWkKydb3GqdMZxCISAFtRvGWOGEAAAQOcRCAHAApIKoRZbxhxHBEJAC/zAynXr185z/wEAAOgkAiEAWIDXZstYxnGS+UMAFtcwQ8gNK+wIVQEAADqHQAgAFtBuhZDrGFHgAKRXv2Usmwn/pG0MAACgcwiEAGAB7c8QMvKpEAJSq68QyrnhyxMCIQAAgM4hEAKABcQtY06LLWOOY+RRIgSkNnfLmCTmCAEAAHQQgRAALCCpEHJae7jMOEaB5WQWSGvuDCGJCiEAAIBOIhACgAUsZYaQx0BcIDXPD2ozhKIWzYpHIAQAANApBEIAsIB2Zwi5jmFDEtCChhlCGSqEAAAAOo1ACAAWEK+Ob3WGkMsMIaAlXmDlunNbxrgPAQAAdAqBEAAsIJ4DlGmxZYwZQkBrmCEEAADQXQRCALCAuMqn9RlCDjOEgJSstXO2jEUzhAiEAAAAOoZACAAW0P5QaTFDCEgpvqskM4TiCiGGSgMAAHQMgRAALMAL2m0ZcwiEgJTiWV3JlrEMM4QAAAA6jUAIABYQzwFqZ+08gRCQjj8neI1nCFV8f8WOCQAAYLUjEAKABcQzhDJOaw+XrmOSqgcAC/PmtGYO5lxJ0myFQAgAAKBTCIQAYAFx5UKLeRAVQkALfL+xQmhdIStJmix6K3ZMAAAAqx2BEAAswLftVQhlCISA1JIKoahVbN1ARpI0Waqu2DEBAACsdgRCALCAua0saTnGKLBSQCgELGruDKGBrKuMYzRZJBACAADoFAIhAFiA7zduP0orPrGNK4wAzG/uljFjjNYNZDVBIAQAANAxBEIAsIB2K4RcNwqEqBACFjW3QkiS1g9kNVlihhAAAECnEAgBwAICe/KJahpJhRCBELCoZsHrukKGljEAAIAOIhACgAUsZYZQ/dcDmF+tQqj2smTdQJah0gAAAB1EIAQAC4jXYbc7Q4ih0sDivCb3s3UFZggBAAB0EoEQACwgHgrtmlZnCIUPr1QIAYtrNkNo3UBGk0VmCAEAAHQKgRAALMAPrBwjOa0OlTbMEALSSraMufWBEC1jAAAAnUQgBAAL8ALbcruYVKt0iE90AcyvaYVQIauKF6hU9VfqsAAAAFY1AiEAWEDQZiDkJjOElvuIgNWn6ZaxgawksWkMAACgQwiEAGABXmAbNh+l5VIhBKTWdMtYISNJtI0BAAB0CIEQACwgniHUqjgQYoYQsLhmFULrowqhCQZLAwAAdASBEAAswAsCZdzWHyrjWSjxljIA8/OjSrpMs5YxKoQAAAA6gkAIABbgB1rSDCHPJxACFhPfT9w5Q6UlZggBAAB0CoEQACzAD4KGqoW0aBkD0ktmCNWtnR+JZghNlWgZAwAA6IRFAyFjzG5jzPXGmLuMMT8xxrwjun2TMebrxpj7oz83RrcbY8yHjTEPGGNuN8Y8tdP/EwDQKV5g5ZglBEK0jAGL8pqsnR/MuZKkYoW18wAAAJ2QpkLIk/Q/rbXnSHqmpN8xxpwj6V2S/tNae5ak/4zel6SXSDor+u83Jf3dsh81AHRJ1bfKuq0HQvG2JCqEgMX5yVDp2suSwVxYITRToUIIAACgExYNhKy1h621P4renpJ0t6RTJV0i6VPRp31K0sujty+R9GkbuknSBmPMjuU+cADohmLF00B0YtqK+LyWGULA4ppVCLmOUS7jUCEEAADQIS3NEDLGnC7pKZJulrTNWns4+tBjkrZFb58q6UDdlx2Mbpv7vX7TGHOLMeaWY8eOtXrcANAVsxVfQ1HrSiuoEALSi7eMzR3gPpRzNUsgBAAA0BGpAyFjzLCkqyT9rrV2sv5j1lorqaWzHmvtx6y1F1prL9y6dWsrXwoAXTNb8TXQRiDEDCEgvWYVQlLYNkYgBAAA0BmpAiFjTFZhGPRP1tp/jW4+EreCRX8ejW4/JGl33Zfvim4DgL5TrPjJcNtWZJItY8FyHxKw6tRmCDUGQgM5V7PMEAIAAOiINFvGjKRPSLrbWvuXdR+6WtLro7dfL+nLdbf/arRt7JmSJupaywCgr8xUPA21MUMoPrFlhhCwuPh+knEaX5bQMgYAANA5ac5yniPpUkl3GGNujW57t6TLJH3BGPMbkh6W9KroY/8u6aWSHpA0K+kNy3nAANBNxSW2jAW0jAGLSiqE3JMrhBgqDQAA0BmLBkLW2u9Imm/n8gubfL6V9DtLPC4A6AmzS2wZ8xgqDSxqoRlCRyZLK3FIAAAAq15LW8YAYC0JAqti1W9z7Xw8Q4hACFjMfFvGqBACAADoHAIhAJhHyQtPRNtbO88MISCtuELINaydBwAA6BYCIQCYx0w5PBFtp2WMtfNAen5g5ZhaZV0sXDvPljEAAIBOIBACgHnErSrttIy5tIwBqXmBPWnDmBSvnadCCAAAoBMIhABgHrPVsDJhKRVCDJUGFuf5wUnzg6SwZcwLrCpesAJHBQAAsLoRCAHAPOLKhPa2jIUPrwGBELCoihcol2lWIRRW5zFYGgAAYPkRCAHAPGaTGULtt4xRIQQsruLbpoFQHMbOMEcIAABg2REIAcA84mG2SxoqHdDqAiym6gfKufMHQswRAgAAWH4EQgAwj2J1KS1jcSC0rIcErEoVL1DWPXmG0CAtYwAAAB1DIAQA86jNEFrKljESIWAxVb/5DKFahRAtYwAAAMuNQAgA5jFTDk9CB9ppGTPMEALSqvqBsk1axgZoGQMAAOgYAiEAmEdxCVvGHMfIGMknEAIWVfaaB0JDUXUegRAAAMDyIxACgHnMVn3lXKfpiWoaGccQCAEpLD5UmpYxAACA5UYgBADzKFb8ttrFYi6BEJBKdZ618/H9Lx7wDgAAgOVDIAQA85gpe221i8VcY5ghBKQw/5ax8P43UyYQAgAAWG4EQgAwj9mqv7RAiAohIJX5hkrHt3k+2/oAAACWG4EQAMyjWPHbWjkfy7gOgRCQQmWetfPxtj7fcj8CAABYbgRCADCPihco3+QkNS2HljEglYrXfKg02/oAAAA6h0AIAObhBYFc5+S5JmllHKOAE1lgUfO1jElhlRCBEAAAwPIjEAKAefiBVabJoNu0XIcKISCN+baMSWGVEC1jAAAAy49ACADm4QVWrtP+w2Q4VJphuMBiwi1j81cIUWkHAACw/AiEAGAefmCVWWLLGBVCwOIqfqBspvl9LQxWu3xAAAAAawCBEADMw/PtkmYIuY5RQKsLsCBrrap+oPw8FUKOEfcjAACADiAQAoB5LLVCyHWMPJ8TWWAhXmBlreZvGXMYKg0AANAJBEIAMI+lbhnjRBZYXDXqB8vOM1TadRxaLwEAADqAQAgA5rEcM4TYjgQsrOqF95HcvBVCYqg0AABrTLHiq+IxRLDTCITQV6y1spxgo0uWZ8sYv6/AQsq+L2mBCiFDsAoAwFrzmo/fpP933b0rfRirHoEQ+soLPvgtfebmR1b6MLBGMEMI6LyqH1cINb+vOQ5r5wEAWGuOTJZ0aLy40oex6hEIoW9Ya7X/xKz2HZtZ6UPBGlH1rdx5TlLToEIIWFw1KgfPzTtDiAohAADWGi+wtIx1AYEQ+kZ8Fbnk+St8JFgr/CBY4gwhhxNZYBGVeKj0fDOEDMEqAABrjR9YlQmEOo5ACH0j3kRTqhAIoTvCGULtB0KOY9iOBCwivvrH2nkAABDzqRDqCgIh9I34AYEKIXTLsmwZC3giAxYSh/0LtowRCAEAsKb4gU2qiNE5BELoG0mFUJUHBnTH8mwZW8YDAlahOOyfb+28Y4wCWi8BAFhTvCBQmUKAjiMQQt+Ie0hLVR4Y0B1UCAGdF8+Ho2UMAADEaBnrDgIh9I24QqhIIIQusNbKZ4YQ0HGLtYw5jpHP3QgAgDWFodLdQSCEvlGhZQxdFFckLLVCKCAQAhZUToZKN7+vuUbcjwAAWEOCwCqwokKoCwiE0DeqXnhCUKZCCF0QV/a485ykpuFSIQQsKqkQomUMAABI8qPZgVQIdR6BEPpGxQ+DIGYIoRuWo0LINZzIAothyxgAAKgXP+9TIdR5BELoG5WoQogZQugGLwmE2n+YzLhUCAGLqSQtYwsEQmwZAwBgzSAQ6h4CIfQN1s6jm5IKoSW2jDH7BFhY/Ng+XyDkUGkHAMCaEl9QrfgBr6U7jEAIfSNOiEueL8vVYnSYF62LX8qWsYzjUCEELKISrRBbqGUs4DEfAIA1o/5CULxYCJ1BIIS+EV9FtpYHBnTecswQorIBWFwc9s87VJr7EQAAa0p8YVZisHSnEQihb9SHQKUKDwzoLC+qWnCXOEOIE1lgYbWWsebhq8NQaQAA1pS6PIg5Qh1GIIS+Uf9gUPIYLI3OWpYtY9GJ7EPHpnXhn31dj44Xl+vwgFWj6gdyjJSZp0IoQyAEAMCaUl8hRGdIZxEIoW80VAixaQwdFs/+WcoMIdcYeUGgfcdndHy6ooNjBELAXBUvmHegtBRVCDFDCACANaP+QlCZ876OIhBC36jWVQixeh6dtlwVQoGtVbfVX+0AEKr4wbzzg6QwWGXDCAAAa4fHUOmuIRBC32isEOKBAZ0VzzVZ2pax8GvjFkfaXoCTVf1g3g1jUtR6SYUQAABrRtBQIcR5XycRCKFvVP3aAwMtY+i0pEJonkG3aThRIFSsxBVCnNQCcy3aMmaMKK4DAGDtoEKoewiE0DfqVw4SCKHTajOElrBlLA6Eot9X3ycQAuYqVQPlswtVCFFdBwDAWuJTIdQ1BELoG1WGSqOLlmOGULw1qVjxJFEhBDQzW/E1mMvM+3FaxgAAWFsaK4Q47+skAiH0jYa18yTF6LB4APRSZgjFc1Gmy8wQAuZTrHoazLnzftxl7TwAAGtK/fN+/Tkglh+BEPoGFULopuWoEMq7cSBUlcSWMaCZmbK/cCBkCIQAAFhLGlrGCIQ6ikAIfaPi1VYTEwih02ozhJZeITQTVQh5zBACTlKsLBwIOQ5r5wEAWEvqL6ISCHUWgRD6RsUPtG4gnDNRpGUMHRYPgM4sYah0rWUsnCFElQNwstmqt/AMIcMMIQAA1hIqhLqHQAh9o+IFGilkJVEhhM5blgohN64QYqg0MJ9ixdcAM4QAAECEGULdQyCEnuMHVpOl6km3V/1A+YyjXMZRySMQQmclM4Tc5WgZiyuEeEID5pop+xrMLtIyRoUQAABrBoFQ9xAIoed8/gcH9PwPXN8wRFqSqr5V1nVUyDgqVQiE0FnLu2WMCiGgmSCwKlZ9DebnbxnLOIb7DgAAa4jX0DLGeV8nEQih5xwan9XYbFXjs1WVPT9JhSteoFzG0UDOZe08Om45towxQwhYWFztueBQaWNkrWSpEgIAYE2gQqh75r8kB3TZX33jPt316KR2bRyUJF1/71G9+1/vkGOMrn7bc1TxA2Vdo3zGJSlGxy3vDCG/4XsCCM1WFg+E4vugH9gltXACAID+wFDp7qFCCD3j3semdOehCRWjgdE3PnhCXmBV8QM9eHQmqhByVcg6PDCg42oVQu0/TOapEAIWVEwCoQW2jMWBEBVCAACsCVQIdQ+BEHpG1Q9UrPrJBrF7H5tKPjY6W1HVD5RLKoR4YEBnLUuFUKbxIXbuXCxgrZuphGHpYi1jksRMdgAA1gZmCHUPgRB6RtW3mq34yRXjB45NJycJYzOVZIZQPuOwdh4d50fhzXLMEEq+JxVCQIO4ZWzhtfPhn1QIAQCa8QOrt372R/rhw2MrfShYJvFm3qxrqBDqMAIh9AwvCFT2guSKccULtGN9QSOFjEZnwgqhrOsoT8sYuiCpEFrK2nm38SGWGUJAo6RlbKG186Y2QwgAgLmOT5d1ze2H9dq/v2mlDwXLJC6qL2RdVaiw7yiGSqNnVL3wxf7oTCW5bctwXl5gNTYbVQi5jvIZVxPF6kodJtaI+OQzu4QZQlQIAQuLK4SGFlk7L3H/AQA0NxPNamQL8eoRVwgN5TIq8+/aUQRC6BnV6I5fHwhtHcmr7AUanamo4ltlM044VJoHBnRYJ2YIeT4ntEC92agidOGWMQIhAMD8pkreSh8Clln8Onwwz+zYTqNlDD0jHrh7Yk6F0KahXFQh5CcVQiWGi6HD4vBmSTOE3LkVQjyhAfXSrJ13ovtgwAwhAEAT8TZXSZos0UWwGsQXgYbzGWbHdhiBEHpGfAJePzhs60heGwdzGpupqurbZKg0FULoND8IZEztZLQdxpiGUIgZQkCj2TRr55khBABYwFRdCHT/kakFPhP9In7OH8xRCNBpBELoGc1Wcm8dzmvTUDZqGYtnCDFUGp3nBXZJ1UGx+rYxTmiBRsU0a+dpGQMALKC+Zey+I9MreCRYLl5DhRDnfZ1EIISeUW0yX2XrSF4bh3IqVn35gVXWdVTIuiqTFKPD/MAuaX5QrD4QokIIaDRT8ZV1jbLu/C9H4gohWsYAAM3UB0L7T8ys4JFgudQqhGgZ6zQCIfQMr65CKD4R3zKc16bBXHJ70jLmBbKcHKCDwgqhpT9E1reMUeEANCpWfA0ssHJekjJu+HxAoAoAaCaeIbSukNE0A6ZXhfg5fyjvUiHUYQRC6BmVugqhs04ZVs51tHNDQRuHaoHQcN5VPuvKWqnSpMUMWC6dqBBq1hYJrFVTpapGZyoLrpyXJCeuECIQAgA0MV32VMg62jiUaxgwjf4VRK/D8xlXZSqEOopACD3Dq9vA9EtP3aXr/9fPaHO0ZSz283t3Kh+dYDNHCJ3kBQEzhIAOeuOnbtHVtz2qyeLCG2GStfN1VaFBYLk/AQAkhRcYRgpZDeUymiEQWhW8KBAqZBkq3WkEQugZXl2F0FA+o1M3DEgKB0tL0ovOPkWbhnLKR+0FbBpDJy1bhRBbxoCmbt43KkkaLqSrEKoPgP7mmw/oFR/9bucODgDQN6ZKnkbyGQ0XMg3zhNC//OjCbCHrqOpzEaiTFn4VBnRRfQvYQK52En36liF95jeeoafv2SRJdRVCpMXoHM9nyxjQSbs3DWjLcF7v/6W9C35eHMzWFZFq/4kZHRwrdvLwAAA97oGjU3rRX96gkXxGZ2wd0nA+oyOTpZU+LCwDL7ByTVghJEmlqr9oiznaQ4UQekb9UOm5Q0Z/+qwtyYl1HAgxYAyd5AdWrsuWMaBTipVAZ+9YpydsG1nw8+Iiu/qWsWLFZyYXAKxxtx6YkCRNlT0NFzIaztMytloE0evwQnLeRyFApxAIoSf4gVX9uXJhga0z+UzUMkaFEDpoubaM5RsqhDiBBWKl6uIbxqTmLWOzVZ+AFQDWuC3D9YtnwpYxhkqvDuHr8LoKIWbHdgyBEHrC3Cu9C50kFLIMlUbndWSGkM8JLCBJ1loVUwZCcTBbHwiVKj73JwBY44K6ytGRQlbDeWYIrRZ+3VBpSWwa6yACIfSEuVd6B3IpKoRoGUMHsWUM6Jx4QORCj/WxuFCvsULIa9hMCQBYeype7XlhOB+2jJW9gJbiVaA2Q4hRIZ1GIISeUPXSVwjlswyVRuctW4UQM4SAkxSjK30LtQfH3KhlLJgzQyiw4YwBAMDaVB/8rItmCElijtAqEM8QyictY5z3dQqBEHpCdc6V3oVnCJEUo/Pi3uWlalw7z+8sINWGQ6ZpGYuD2YaWsejxf+5zBwBg7agPhIbrAiHmCPW/eJZnIVPbMobOIBBCT6j66VvGkl5SkmJ0UEcqhJh5AkgKK3wkaSC3+MsQJw6E6iqEZivhi33uUwCwdtUHQiOFrIYLBEKrRW2GUNQZQiFAxxAIoSd4LQyVjiuEGCqNTvL85dkyFgdCrmOYIQREiq1UCMUtY3X3n/jrCYQAYO2qRM8BW4bzOv/U9bUKIQZL9z0vCKIZQlQIdRqBEHpCXCEUve5PuXaeQAid4wXBslYIDWZdAiEg0tIMoeh+GM/gCgKbtIzRhgkAa1c8g/Qbv/c8nXfqeg3RMrZq+IEatowxQ6hzCITQE+KSz3WFrHIZZ8ET8WSoNEkxOsgLrDJuk99Da6VDP0z9ffLRDKGBnMtQaSBSqrQ+QyiuEKp/Uch9CgDWrviiQDZ6rTVCy9iq4QeBMi5bxrqBQAg9IS7737NlSLs2Diz4uQUqhNAF884Q2neD9PGLpGP3pvo+cYXQQI4KISCWtIylWDvvzpkhFM8fksRqYQBYw+IOgzgQomVs9fACK8cYhkp3QWalDwCQpEr0ov5tF52pZz1+84Kfm3WNjKFCCJ1V9efZMlYcDf8sT6X6PkkglHU1Vq0s1+EBfa2VGUKOadwyNlsXCDFDCADWrooXVwiFzxO0jK0efrTttzZDiAtAnUKFEHpCPFS6kHU1mFs4pzTGKJ9xqBBCR5WrfvP5Jl45/DNIF0jGa+cLzBACEnGVTyszhIKoQqj+KiEzhABg7ar6QXShOHyeYO386hFX6sfLhKgQ6hwCIfSEeA5EXPK5mHzG5YEBHVWs+s2rF7xS+GeQ7sVGLip1HcgyQwiIlVppGUsqhML3i9X6ljHuUwCwVoWBUO3cwXWMBrIuLWOrgB/N8nQco1zGYah0BxEIoSfELWNNh/g2QYUQOq1Y9ZufrMYVQjbdE9NQPvweI4WMfE5esQJ+/4u36Y+/fOdKH0aDllrGolcqQZOWMaruAGDtqvr2pIvJw4WMZioEQv1obKai0991rf7tx4eSGUKSVMg4KtMy1jEEQugJ8RyIXMoKoULWJRBCRxUri1UIpQuELj53uz7+qxfq9C1DqtLeghVwz2NTuvexdDOvuqVYqbUJLyYTJUJxhV1jhRD3KQBYqypzKoSk8EJD/fIB9I+jU+FF1w9ed28yQ0iS8lk6QzqJQAg9odpWhRAPDOiMILAqe8E8M4SiwdApK4QKWVcvPmebMo6hmgErouIFPRecFKu+chmn+Sa/OeIKoXjLWKnC2nkAgFT1gmSgdCyfcRhA3OcOjhWjGULxHE6HQKiDCITQE5JAyEk5QyhL6SA6J+5Tbt4yFlcItfb7l3EMJ69YEVU/6LlZO6X5ZnQ1Ec8QatYy1mtBFwCge+bOEJLiLgLCg35Uqev+qK8QKmRcQr4OIhBCT2i1ZSyfcRkuho6JS40XbBlLWSEUcx1H1tZOaoFuqfq254KTYsVXIZvu8T6uIvKbtIyxdh4A1q5whlBjhVBYTdJbz3lIp1L3WmW67CXP/4Us532dRCCEntBqy1iBCiF00IIDb1tcOx+Lf7epEkK3Vf2g4UVWL5h3i18Tzpy180WGSgMA1LxCKJ+hQqhf1V+8OjRerAuEaBnrJAIh9IRqG2vnGSqNTomfdAoLtYy1XCHUWOUAdEvVDxrKsHtBseqnGigt1a+dZ6g0AKCm6gfKZea2jFEh1K/mvlbJ1FcI8W/aMQRC6AnV6AFgbtnnfBgqjU6KNyAta4WQE1cI8YSG7ir34FDpUtVvPqOrCddprK5raBkjYAWANavZ2vk87UV9a+5rlfj5Pzzv663XMasJgRB6QnySnEldIdSY/o/OVPSNu4505Niw9izYMuZHgZBt7YmJCiGslF4cKl2stDBU2mkcKl1kqDQAQPHa+ZO3jDFWoj/Fz+nx64P4+T+XcXi+7yACIfSE+GQlbYXQ3A0CX7jlgN505S2aLnsdOT6sLUkglGvyENn2DKHwe1HRgG6r+japwuwVrcwQSlrGmswQYqg0AKxdbBlbXeIqoKeetkGSFD39K+c6Pdf6vpoQCKEnxKlvNu3a+Tmlg1OlqqyVxmYqHTk+rC3xCWfTGSfJ2vnWwsekZYwTWHSRH1j5ge3JodJNZ3Q14cypEJqt+snMCCruAGDtqvrBSRuKWVHev+ICgZ963EZJ0iOjs5LCCiECoc5Z9OzbGPNJY8xRY8yddbdtMsZ83Rhzf/Tnxuh2Y4z5sDHmAWPM7caYp3by4LF6eL6V65jkhf9i8lm3oRx0NjqBH5+tduT4sLaU0mwZa3OoNDOE0E1x2N5rpdaliq9CJl0gJIX3n/oKoXWFjCSpyv0JANasqtdshhBzRvtV/Frlp07fJEm6+/CUpCgQ6rHXMatJmnKMf5T0c3Nue5ek/7TWniXpP6P3Jeklks6K/vtNSX+3PIeJ1a7qB0kFRRr5jKOS58tGJwiz5fCBf2yWCiEsXa1lbKEKofaGSlPRgG6KX1wFVvJ66MVU2QtUyKYvUnaNUXz4sxVP6wpZSVTcAcBaVvUDZeaMmyhkXFV9y+utPhRXAZ29fUSS9OzHb5Yk5VyXCqEOyiz2CdbaG4wxp8+5+RJJPxO9/SlJ35L0zuj2T9vwLP0mY8wGY8wOa+3hZTtirErNtgQspJB1ZW34dbmM0Wx0Aj9epEIISxe3jHWmQogXKOie+hdQVd+qhaKcjgoDofQH4zhSEF0AmKn4WjcQBkK9VvkEAOieSrOWsehiQ6nqayi/6Kkuekj8nF7IubrlPS/ScPTvl80YAqEOaneG0La6kOcxSduit0+VdKDu8w5Gt53EGPObxphbjDG3HDt2rM3DwGrhBSdvCVhIPpofEZeEFivhPJdxKoSwDOIKoeYzhOKh0q09MWUcZp6g++q3i/VSuXWp6ieP42lkHCepBipWPK2PAiECVgBYu5oNla6dI/TOcx7SiV+n5FxHW4bzyevwvBu2jMWdIVheSx4qHVUDtfyvY639mLX2QmvthVu3bl3qYaDPhSWf6X8d5z7YM0MIy6lU9WWMmp+wLrVCiBYXdFF9BU2vVNN4fiAvsMq3UK6UcU0yf2umXKsQ6qU2OABAd3m+VTYzp2UsChHieZDoH3EV0NyQL14kUeU1dEe0GwgdMcbskKToz6PR7Yck7a77vF3RbcCCqr49qeRzIfGJRBwIzVSYIYTlU6yEK7GNaVK11uYMobgCjgohdFOlBwOh+HG7lRlCWddJjj+cIRSWkVMhBABrV6VZhVCWCqF+VfUDuY5JLqLG4kColyqdV5N2A6GrJb0+evv1kr5cd/uvRtvGnilpgvlBSKPZULiF5Ov6g6Vay9gEFUJYBsWq33x+kMSWMfSV+hCoV/rv4xfprbSM5VwnuTI4U/E1XMjIMVTcAcBaNt/aeYkKoX40X4FAfFuvvI5ZbRadtGWM+WeFA6S3GGMOSvoTSZdJ+oIx5jckPSzpVdGn/7ukl0p6QNKspDd04JixCnktDpVOKoSqjS1jVAhhORSr/vwDb9veMsYMIXRf1av9vvVOhVB438m3MFQ66xpV/UCeH6jiBRrKZZRxHdbOA8Aa1mwpDS1j/aviNZ8pm4vO+wiEOiPNlrHXzPOhFzb5XCvpd5Z6UFh7Wl47n507VJotY1g+parffOV8EEhB9DvWZoUQ/c/opkpDhVBv/O6Vqq23jGWilrF4o+RgzlXWMVQIAcAa5QfhanmGSq8eFT9I2sPqxSERgVBnLHmoNLAcmm0JWMjcB/uZZMsYgRCWLp4hdBK/XHu71S1jzBDCCmhcO98bL6SSCqEWhkpnXUcVz2q2HAdCYYUQQ6UBYG2Kn9PmDpXOUyHUt6reyS2AEjOEOm3RCiGgG7zAtrR2vr4cNAhscsWZtfNYDvPOEIrbxSQp8Fr6nswQwkroxS1j5TYqhHJRy1gc/g/lXWUcw1BpAFijqnUryusVkjmjvfGch8VNlz197IaHNFvxlW1SIRQXAlAh1BlUCGHF+YHVPY9Nacf6gdRfU18hVIyuABSyjiaKVQWcIGCJitVAhWYtY15dhVCLLWNxSyQVQuimhqHSPRIIxVdtW60Q8oJgToUQLWMAsFbFLfhzR07UNhFTIdQvvnP/MX34P+/XDx8ea9oxQoVQZxEIYcX9YP+ojk2V9ZLzt6f+mvq18/FA6Z3rBxRYabJE2xiWplTxNdCseqGhQqi1Fxo5etqxAlbLlrGs66jq2VqFUM5VxmGoNACsVbWWseYVQmUqhPrGRDQDdrxYmWfLGEOlO4lACCvu2tsPayDr6qInnZL6a5IKoaqv2egE4bTNg5KkQ+PF5T9IrCnzt4y1XyE0nA87dKfLrbWaAUtR8eu3jPVGNU0pqepMXyGUcY0qfpAsEBjIucqmrBB68Ni0Hjw23d7BAgB6UhwOzLdljAqh/jFZDF8bl6pB05axHC1jHUUghBX3/X2jetbjN2swl36kVbxlrFRXIfT0PZslSXcemlj+g8SaMluZZ8tYQ4VQa09KSSBUIhBC91R7cqh06xVCuWjLWG2GUDRUOsX98I/+7U798ZfvbO9gAQA9KZ4hN7eiJH5uYYZQ/6jv7sg1mSmbbBnzCfk6gUAIK26qVNWmoVxLX5Ok/1U/CYSetGNEI4WMbj9IIISlmSpVNVLInvyBJVQIDUWB0AwVQuiiSi8OlU4CodZmCFX9+hlC0VDpFBVCJ6YrGpuhlRgAVpOkZWyeCiG2jPWPyWJdILRghVBvVDqvNmwZw4qbqfgaalaNsYCGodJRIDSUy+j8U9frDiqEsARlz1fZC7Su0OThsT4QanGGUNZ1VMg6tIyhq3pxhlCpbhFAWtmMI8+vnyEUDZVOMaR9slRtqRoJAND7ai1jjRUlWdeR6xhmNvaRybrq+WZDpfMMle4oXiFhxc1WvKR6Iq24PLTs1VoIBnOuzt+1XncfnqRvGG2bip6UlrtCSArbxqYIhNBF9SFQr7yQaqtCyAlnCMUVoYP5aKj0PP9PsxVP1oZh0USxqpkKzwkAsJrMN1RaCgMEKoT6x0RdhVDTLWMMle4oAiGsqLLnq+rblgMhY4zyGUdlz08qhAZzrs7buV5V3+rBozOdOFysctbaukCoWYVQ+1vGpDAQomUM3VQ/SLraIy+k4sA+30qFUNwyVvHkOkY515l3qPSND57QOX/8NV1508PR1/jc7wBglYmf35ptpSpkXZW4ONw30reM9cbrmNWGQAgrKp4H0WrLmBQ+2JerdVeMcxltHAxnEdGWg1ZNzFa15w/+XZ/8zj5J81UI1QVC7VQIFTIMlUZXVRtmCPVG73086LOltfMZo6pvNVP2NZhzZYxRxjl5qPRM2dMbP/UDSdLdh6eSgHe24itI0V4GAOgP880QkqRCxmHtfB9pHCq9UCBEyNcJBEJYUUm7V4sVQpKSCqF47fxAzk1mUtAyhlZ9f/+oJOnKmx6WpBQzhFp/oTGUo2UM3dUwQ6hnWsZ85TKOjDl5k8h8sq6jqhdWCA1FGykzrjkp5Do2Va5rD7MNVx2LtA8AwKrwX/cf02v//mZJ4XPBXPmsqxLVJH0jXjsvnTwTqv62Xnkds9oQCGFFzZRrA6Fblc86cyqE3GQmBasm0arv7zshqVat1rRCKIhOLt1cWxVCIwVaxtBdFS9Irqz1zJaxaqBCi0Oes66jahBopuJrMB/eRzOOkT+n6meqrgJvsuQ1XHXkvgcAq8NND51I3m5WUcIMof4ykbJlrFcqnVcbtoxhRSUbY/Ktt4zlM67KXhgIhfMknKRCiCcBtOrG6MVFXF3QdIZQEJ1Quvna2y0YzmdoZ0RXVfxA+YyjILC9Ewh5vvLZ1h7zs1E10Gy5vkLo5KHSU3UB0HTJa3iRyWBpAFgdHpsoyxjp//vZJ+qJ20dO+vjm4ZyOT5ebfCV6TcULGip4mw+Vri0TwvKjQggrKpkh1EbLWCEbpv/jsxWtH8xGt4UnGTxgoBUTxap+8uhkw23rmlYIRU9YmVxbQ6WH8swQQndV/SAJzHtlGGO5GrS8Bj7rOvKDcIbQQFTFl22ydj6uCNo8lNN02WsoQ6dCCABWhwNjs7rwtI36nRec2TRA2LF+QIfHS02+Er2m/kKO1Lziy5hwmUSvvI5ZbQiEsKLiaonBNoZKxxVCx6bK2jqcj26jQgite2yiJDunCnV4sQqhdodKc1KKLqp6VtloI1evlFqXPD8J79OKX/BPFKsayMYtY468ORVCk1HgumNDQVOlKi1jALAKHRoravfGwXk/vmN9QUenSic9R6D3xM/b8VjBZi1j8e0EQp1BIIQVFQ+EHl7CUOnj02VtGYkCoWw8Q4hACOnFpaq7Nw1ICn8fXafJwNs4EMrk2hoqPZLPqOwFPKGha6p+oGzGhC+keuSFcXsVQuH9cbJUTS4gZJyTQ654htDO9QOaLnkNQ6VnaRkDgL5X9QMdnihq18aBeT9nx/oBBVY6OkXbWK+Ln6fji/vNKr6kKBDyeR7vBAIhrKiZupXxrVpXyGp0ptK0QoiWMbSiGP0e7toQXm1qOj9IqrWJtVkhFLdGUqmAbqlELWO5aEtXLyh77bWMSWHgk1QIuScPlY5fWO5YX9DUnKHScXXe0amS/vG7+1hDDwB96PB4SYGVdi1UIbShEH7uRLFbh4U2xbP+dmwIA775AqGsa7ig2iEEQugqa63+9UcHkwqe+MS4naHSuzYO6NB4UcenK9oykpMUBkLGSOUerxD66p2HNTZTWenDQKRYDX8P46tNTecHSXUzhPJtzRCKK+FoG0O3VLwgbBnLnDyAeaWUqu23jE2XPRXiCiHXkRfMHSrtaTif0fqBrKYrnsZn6yuEwvvdJ7+zX+/9yl36/C0HlvK/AQBYAQfGZiVJuzbNXyG0c334sUeZI9Tz4kBo5/owxFuoZaxXWt9XGwIhdNX9R6f1e1+4TV+98zFJ0mzZkzFKrvi2YvemQZWqgSp+kFQIGWPCVZM9nCBPFKt682d+pC/+8OBKHwoixUr4+xJfbZq/QihuGWuvQij+vgRC6JaqH66dz7rttYwdmSzp1gPjy3pM7VQI1Q+ZjJ8vsk1bxqoaKWQ0XMjI2nA+2MZo6cBMvMQgCpQ++q0HTqowAjrpxgdP6ASbj4AlORgFQgvOEKJCqC8cnSrpQ1+/T8P5jM48ZViSlHObjGyQGCrdQQRC6KrRqCrm8ESY2E+XfQ3lMjKm+Z1/IfW9w1ujGUJSNGy6hyqEyp7fcEIVhwGjs1QI9Yq4ciD+nVowEDJu+F+bW8YkAiF0T9WPh0o7qnithx9/960H9YZ/+P6yHlNbFUKZ2nNErWXs5KHSUyVPI4WMRqIqv0PjRW2PrhTHFanxzLADo0U9cHS6vf8JoEUTs1W97hM368P/ef9KHwrQ1+JziO1RRUkz6wpZDeczVAj1uI9e/6AOjhX1D294mk5ZF/57zj9DyGUkSIcQCKGr4vkORybDB+jZitdWu5jU2DscVwhJ8Tr63nnA+NKPDunlf/td7Ts+I0kqRuHDRLG60Jehi+IWxlogNF/LmCc5GclxJdv67xgtY+i2ih8o6xrlXNNWy9jYbEVjs1WVveUL2dupEMo4dRVC9UOlm6ydj08EpDAQ2jKcU851kpl19cOlZyrcF9EdN+07IT+w+t6DJ1b6UIC+Nh3NkpsvOIjtWF+gQqiH+YHVv99xWC940lY97fRNSfXuglvGeqT1fbUhEEJXxasFj06FgdBMJawQasd8FUKFrKvSMp68LNX+E2Fp6/cePC6p1rYwMUsg1CviioFdm8KQcd3AAhVCTkYyTlsVQknLWImTUHRH2DLmRr33rb+Qin9XR5dx5lnZ85XPtDdDSFJSXVTIuqr6QcNWybhCaDi6r02VPK0fyGow7yYVQvVD3dlIiW65MQqC7j86rQeOTieVqQBaM1Pxksf4hWxbV2DLWA+7Zf+ojk6V9Qt7d0qqVdHPF/TlXUeVHjq/W00IhPqc5/fXCutahVD4AD1T9jTYZoXQUD6jTUPhMOktw/UtY47KPVQhFF+diF8MxlenqRDqHfG/ybaRvEYKGW1fN08ZcuDXVQi13zI2RSCELqn6gXKuUdZtMxCKwpMT08sXCJWqgQrZFmcINWkZe9L2EVkr3fvYVPKxcIZQVuvqThbO2DqsoVwmqQaqrwoiEEK33PTQieS1yov+8tv6g3+9Y4WPCOhPUyVPI/nFA6HBnJtskUXv+fZ9x5R1jS560imSlBQIzLtlLMOWsU4hEOpzf/7v9+jST9y80oeRWrwCOG4Zmyl7bVcISdLujQPKukbrB2otPr1WIXQ46l++6aFRWWuTjVYEQr2jWPWVyzjKuI6ufdtz9es/vaf5JwZeGAYZtzZgugXx9rKpEv/26I5ky5jrqNLGdo4kEFrGCqFi1Ve+zS1jkjSQC98+f9d6SdLthyaSj03GFUL52nPCE7eNaCjvarYcb7f0k0Cql9qLsXpNzFZ1z2NTuvSZpyWvV25b5mHtwFoxXfaSC2wLyWeZOdPL4q2g8b9lfJF/vir9nMuWsU4hEOpzDxyb1sGx/umPjUOQo5NlWWs1W/FTPajP57TNQzplpCDHqV09LmTcnrrq++hEUTnX0fHpsh46PkOFUA8qVvyk6uBxmwc1OF9IWT9DqI2WscGcK9cxSTAKdFpDINTGC+NahdDylN3PVjxVvEAbB3MtfV3DDKHovnrqhgFtGsrpjoPjkiRrraZKVa0byDYMhn/CtmEN5esqhMqeNg+FlRq99FyB1eveI2EV297d63XVbz9LT9w20tDqDiC9mbKXzIlbSCHj9NSSGTSqeEHDvKBzdq7Tv7z5WXrWGZubfn4uw5axTiEQ6nPjs5Vk/kk/mCyGL8grfqDx2WrYMpZrr2VMkv7XxU/U373uqQ235bNOz1wRCAKrI5MlXbB7gyTpkdHZ5Cr1OFvGekax4qf7PUxmCLXXMmZMWM0W3w+ATpsseVo3kNFArr2gPJ63s1wzhOLWs81DrQVC9S1j8QwhY4zOP3W9bj8YVgiVvUBV3zbMEJKk07cMhS1j8Qyhiq/Nw+HPp0II3XBfFAg9YduIzjxlRLs2DjQMNweQ3lQp3QyhfNZRqUfOB3Cyih+cNED6aadvmnfzdC7jMlS6QwiE+tz4bLWv+mPrKyOOTJU0lTLln8/uTYPau2tDw235jNszL/KPz5RV9a2ectoGSWH7WDxIcqrsKQgofewFxWqtQmhB9TOEgvZ+x9YVMlQIoSuCwGp8tqINAzkNZt2GYcrzedUVN+oLPziQvB/Puzq+TDOE4tazOJBJq6FlrO6+unfXet1/dFrFip/cr0YKWQ3XVfllXUeDOTcZ6D9b8ZJAqp8uqKB/3X9kSsP5jHZGa7ILOZffPaBNM5W0FUK91TGARq0umMi1WemMxREI9bmx2YpKni9r+yNYmCxWlYte2N9+YELHpso685ThZf0ZhWxvlIhee/thvfqKmyRJF+zaIMdIj00UNRsdm7UMF+4VxYqfrLJeUDJDyGmrQkiS1g1kaRdEV0yVPQVW2jAYbtma7+JB2fP1+k9+X9fc/qi+v29Uv3/V7ZLCgdRxteXozPK0jMWtZ5uHW2uXaZwhVLuvPvW0jfIDqx8+PJZU3q0rZBraiCVpy0hex6Zrywzin8/JArrh3iNTOvOU4eTK90DWVamPLuYBvWS6lDIQYoZQT6t4QXJOmMZAzmE7Y4cQCPUxzw80VfJkrfrmAW+iWNUZW4ckSVf96KAk6Znz9Iq2K5/pjSeA937lJ9p3fEZSWMm0dSSvRydKDSdlBAO9IX2F0NJmCEnhYOlJ/t3RBROz4e/ZhsGcBnOuZqu1iwcHx2b19n/+sUpVX/c+NqVv33cs2Xq0Z0v4GF1fUbRcW8aSCqEWW8ay7slbxqSwvNx1jG566ETShrshmk/05794vv7td54jSdq5vqDRmYpKVV8zZV8bB7MyRj1x8QCr3/1HpvWEbbWLX/H9EUDrpsspW8YyjvzAtrVhE51X9k5uGVvIUD6TVPpieREI9bH6MKFfrnJOlTw9afuITt0woJv3jWr9QFbn7Fi3rD+jkHVUik58PvDVe5Le/W7bsb7Q8PaO9QM6PFFseDAbLzJHqBfMtlQhlAn/a7tCKKNJKsPQBWNRQLJxMKvBXEZ+YJOw/HsPntDVtz2qex6b0n1HpiXVKhZP3TDQ8L60fFvGkhlCS2gZK9QFQsP5jM4/db1ufOhE0tYWh03//RmPS+a3bV8f/j8dGi+qWA2XGeQzzJfA8vnGXUf0Tzc/fNLtJ6bLOjFT0RO2jSS3DWRZhw20o+z5qvo2dYWQ1D/nSGvN3KHSixnOZVTxA9rGOoBAqI+NNwRC/XHnmCxWtXEop7dedKYk6Rl7Np1U2r9U+ahn+OhUWR/91oN646duWdbvn9aj4yWdtnlQv/xTu7RpKKedGwo6PF5K1s5LVAj1ilKrM4RM+xVC4VBp/t3RefFzxIbBbPL7HZ+ExmHPkcnSSaF5vI0r3jBWyDo6sUwtY6MzZRWyzvyb/OYxX8uYJD3r8Zt124FxHRybldQ8bIpntzx0LKzaHMplOCnHsvqnmx/WR69/MHn/6GRJ7//qPXow+p2LK++k8He47AXMEQRaNB09d6ULhMLnjV7oGsDJKn6gfIsVQpJSzUNEawiE+lj9lqp+GE7oB1ZTZU/rClm98qd26UVnb9Orn7Z72X9OIdoyFp90u8scOKVRrPg6Pl3WL//ULv3fX36yjDFRhVCpYbMIgVBvKFZbnCHkuJJtd6h0lqHS6Ir6FqqhfPj7HbepTEW/g0ejQOgJ24b1s+dsUyHrJC+44xddj9s0qBPTFfnLcPJ6YrqSrHyfV7Ukfe8jkl970TffUGlJevKuDfICq5seGpUkbWrSjrYjqnq6/2gYfg3lMypkGTiK5TM2W9XRqVIS8rzuEzfr7771oP79jsOSpB1RlZpU+x3uh9duQC+JL1SkCYTigcU8zvemcrW1QCj+N58mEFp2BEJ9bHy2dlLZ61c5Z8qe/uG7+ySFQ3WzrqO/f/2FeuHZ25b9ZxWyrrzA6uhUeEU7PhHqpkPj4ZXqXRsHk9t2rC+oWPV1eLyUtDQQCPWG2UqLM4SM0/4MoYGsStVAZa+377Pof2NRm9eGgawGooqcYlT9Ew9gPjJZ1v1HpvWk7ev0sV+9UL+wd2fyYmsq+vOnz9yq2Yqv//XF25Z8TMdnKtqyWLvYvhuk6/5QOviD5KbcPC1jUq3F7a5HJzRSyDTdWhK38D5wNGyPG8q7YSDElWMsk4liVVXfamy2ov3HZ5JWzB8+PCZJ2rmh1kYeX4AgEAJaEz8/DaUJhKgQ6mnN1s4vJKkQYrD0siMQ6mNjdYFQqcdPLr9866P6s2vvlhS2zHRSnDYfGitKSncVYbkdGA1/9u5NtSuCO6OTlgePTWtH9MKwPtTDyim1PEPIbX+GUDQIMT4hBzolbhlbP5DVYBSixDPM4gqhh45P69B4UU/cHs43Gc5nkhfccaXQa56+W7/+nD361x8d0tHJ0pKOaXSm3LSCp0E1bLFRudbKlomGSucyzklVn/Hj6aMTpXmHVReyrjYOZvVgHAjl+qtC6MePjOnIEv/u0VnxzK4jk2V9/a4jye0/eXRCA1m34bXP3BZOAOnEz0sjKYZKM0Oot7W6ZSy+wE/L2PIjEOpj9S1jvb6+dL4NMZ0QPwEcGl+5QCieZVFfIRQHQidmKto4mNNIIaODUWgVG5up6Mu3HtLND53o3sGucdZazbY1Q6i9J6R10UkBbWPotPHZqkYKGWVcR4Nxy9icGULffSB8rIkH3o4UwkDIWpu86BrKZ/Tk3evDr1viC7ET05XFV85Xo+CjPJncFLeMNbufbh7KJVcZF/reO9YPJBVCg3k3WUDQS/zA6pb9oyfd/hufukV/960Hm3wFekEQ2KTi98hUSQfGZjVSyGjLcF6BDUPLeOW8RIUQ0K7WWsbC54V+mbO61rQ8VDppGeNxc7kRCPWx8T6qEKr6tdkTp28ZXOAzly4eIvdoFAjFrRLHpsoaXaZNOYs5MFZULuNoa93Jya6NtWqhwZyrp52+6aTg5/JvP6h3fO5WvebjN9FO1iVV38oPrAZbnSG0hJYxSQyWRseNz4bhs6RkiHM81H6qHP7+xY8ze3eFgc9QPiNrw+AoeeFdyGgot/RhjtbaKBBapELIi4Lyugqh+KJCs0AonNEWVgktVH20c0NBM1EgNpTLqJDpvQqhP7v2Lr3y8hsbBn37gdXoTKXhIhB6y2SpKhu9zDk6WdKB0Vnt3jiYPO/Xbx2VlDzfUCEEtKb+eWkx8QViWvR7UzhUOn2RAEOlO4dAqI+N1Q+VrvR2+l2JHoz/6/dfoHN3ru/oz4ofXB6dCE8q/CD8u/ndz/9Yf/ilOzr6s2MHRme1a8NAwwa1zUO5JKwazGX0rDM266HjMw1tAPHbgZX2H5/pyrGudfEV2rlzSZpKZggtbai0JFbPo+PGZqvaMBj+vg2ZskY0m7SM1bcsbh3Ja9u68IR1uO4FVzKrIZdJKoxmlnBlbrLkqeIH2pJmqLQkVaaTm4wxyrpm3tbO+IR7oflE9RWb4VBpp+euHP/rjw5JagyM47cZpNm76i/QHZks6+BYUbs2Dmj3pvB3rn6gtFR7vqFCCGhNOxVC5R57nEeoXPXbrBDiuXC5EQj1sfFiNem97PUXFRU/fDDeuNjsiGVQqxAKTyriJ4Ijk+WuzWDYd3xGp9etmJXCE5r4hGQg5+pZj98sSbrxwVqV0OhsNbkCvv8EgVA3xFdoW54h1Pba+XiGEBVC6Kzx2Yo2RBVCO773J7oi+5d1a+drv397T62F9PFchqmyp+mSp8GcK9cxyQux2SUMc4xbaU/dOLDwJzapEJKkjOPMG9zujE64F9pgdtGTTkneHsq7Gsi5PfXcaW2t7ai+QjSeBTVFiNyz6i/QPTZZ0sGxonZvqlUI7ZxTIcQMIaA9ra2dZ4ZQL2t7qDSB0LIjEOpjE7NVnbIufPFbqvrJqtNeVIkm/LcyPKxd+XiGUDSfJ94uMFt3xbuTrLV6+MSsTt88dNLHdkcvDgezrs7esU4jhYxuebg2L2JsppLM6thHhVBXxCeE6VrG/NqWsbaHSocVG7QEotPGi1VtiFoU87OHtc2MJYHOVMlLrp6ev6sWCMWtYdOl8PEyftEdt5wt5TE0npm2e+MibcPJDKHGQCjrGg1kmz+HxIOlF2oZe3YUwkvh/0+vtYw9eKxWEVVfcRK3irFZpXeN1z2e3314UsWqr10bB2otYxsaQ9BaC2fv/P4B/WCm7MmYdK/Zai1jVAj1miCwqvq2raHSswTpy45AqI9NlT2dMhIGQh//r4d0xrv/vWdT0zgQqh8u3SnxSU5clRT3Dk9HV7w77chkWcWqrz1NZiXFV8YH8xm5jtGpGwZ0dLKcfHx0pqKd6we0c32BlrEuiU+Q06+dX54ZQgRC6KQgCOf1bIxaxly/rLyparbqy1qryVI1GST95N0bkq+L5zLMlD1NlqrJ+8vxQuzAaDxsf7EKoeaBUC7jLNAyFlUILdAylnGdpFJjKOcqn3V7qmXs+/vGkrebVQh14/kL7YlDu8dtGtSPHxmXFAaf8YWhuSEoFUJAeyZLnoZzmYYh7fOpDZXmftZr4nO0ViqE8hlXWdfQMtYBBEJ9bKbsJRtVHj4RvtD+5j1HV/KQ5lWOygLTPIAv1dwT+1I1CDdJ1Q1J7aS4smduy5gkbR0OT0biuUYbB3MNpeZjsxVtHMppz9Yh7Yv+TdFZ8Qlu6zOE2nuBUci6GslndGyqvPgnA2267+iUpsuezo3awYxfUkEVzZZ9lb1AVd/q4nO36YpLf0rPP2tr8nVxRdBU2dNDx2b0uGgGynKUah8cK2oo5yZzjeY1TyCUdZ15g9s4bN86svB8on9/x3P1T298hjKuo0LWUbmHThT2n5hRJpo799Dxaf30+7+pux6d1MQsM4R6XVzR9aTtI8ltuzYN6FlnbNbHLv2phuo0SSrkwpffsz30+wf0g9GZijYttpggEr+u+4fv7teeP7hWz/zz/0weT7Gy4qqtfAuBkBS+FunV4od+RiDUx2bKnjYMZJMXkJJ07e2HV/CI5lfxAuW70C4mSefsXKctddu9yp6vih/IC2yyTrmT4tk/zVrG4iexeNvZpqGcxqInp1LV12zF16ahnE7fPESFUJfEf8/x8M8F1c8QsoHU5u/StvWFrs2zwtp0UzSb7FlnhCeiplpSXlXNVnxNRvOD1g/mdPG52xuG38eB0ESxqoeOzSRVRIPZpQ+VjueqLHphoNp8hlBYIdR8bsRzz9yiD7xyr56xZ3PTj8c2DOb0nDO3SAovHvRSy86B0Vk9bvOgRgoZfX/fqA6OFXXvkcmk+oRAqHeNzVZljPT2F56V3LZr46Acx+hn59zHpFrLWIkKIaAlx6bKDRt8FxLPFL33yJRyrqPHJkv6xHce6uThIaVKu4FQLsNzYQcQCPWx6bIXbUqpXTG9/t6jPRkkVLzWBoctRT7j6teefVryftkLkpOYwKrjLQL7j88o5zraueHktojt0SYfzw+DhA2DWY1F4VB8hXHjYE57tgxpolhNPobOuf/otHKuo9NSBULRDCEnOiltc9PY9nUFPUYghA668aETDVuO5BWVN1UVq14ynHhdk7W9cYvYXY9OquIHSSCUcR3lM86Sh0ov2i4m1SqE6raMSdKfvOwc/dbzzmj6JRnX0asu3C3XSV+FWsi68gIrz++NtrFwM9Wg1g9kk0rT2YqftIyVqkHPHCsaTcxWtK6Q1Xmnrtd/vOO5+rOXn7fg0Nu40o1ZGEBrjk2XF60EjdWvND/zlGG99Pzt+uR39xMo9IB2Wsak8KIVFULLj0CoT1lrNRMN/IwDocdvHdJQPqNXf+zGniuJ7GYgJEm/+bzH620XnannPWGrytWg4cFjqtzZv5v9J2b0uM2DTU9MLnrSKXr7C8/Su17yJElxhVBFQWDrqoayyZPdiRnaijrtviNTOmPrkDJpKtjiGUIm+tw25widsi7fMDsKWE7WWt28bzSpDpIkVUvKydNsqZpsuBtpFghFJ7E/eiScZ/OEbcMNH2t3sLG1Ngk8FjVPhdBFT9qm8+o2oi1VfPW41CMDRw+MzWr3xgFtGMyqGl00KFb8hgHTS6nQQueMzVaTVsizd6zT65552oKf7zpGuYzTUxVqQD84Pl1u6AJYiOuYZHbp5uG8XvGUXZoue7rvyNQiX4lOS5YNtdwy5vI82AEEQn2qVA0UWEUVQuE/40+dtlEf/pWn6MhkWTftO7HId+iuVlcLLlUu4+h//uwTdfrmQZU9v+EqXKcHc56YriTDvudyHaPfe/ETktlPGwdzCqw0Waoms4Q2DubqNlGRgnfa/Uem9cS6uQ8Lqm8Zk9qeI7R9Xdgy1subAdG/psqexmerOqsuzIlXuVfKpboKoZNn+eQzjjKO0e0HJySFV1Vjg0t4ITY+W9V02WutQqg8vfDnLVEvrSSeKlU1PltNKoRisxW/YcB0py9ooD3jxao2DKabaxIbyPbWljug11W8QOOz1dQVQlKtSmjzUC6Z7dmLnRRrTW37dIr5nXWG8rSMdQKBUJ+K7wzDeTcpPd40lNeFp2+U6xjdEb2Y7xUVL+jKyvm58hlHZS9oePDodLI8U/GTAayLiVckj85UGuYKrRuIBruWePHfSVOlqg6NF5O2mEXVD5WO32/DtnUFeYHV6CwtgVh+cUXkSH3gE61y96qzSSA00iQQMsYkbWOP2zSYzDqRwt79EzMVvf+r97TcOvZwtGHscWlaM+O186Vx6fq/kCYfbelnpRUHQv/nmrtWfMj7wbEwsNu9aaAhEJqpeMkMIYk5Qr3qxHRZm4daD4SW0oIJrDVx1XzaCiGpVgm6eSinx20alGMIhHpBvAG61RlCtIx1BoFQn4rvDPUzhLYM51TIunrCthHdfqgHA6EuVgjF8hlXZS9oeNHV6SussxVPg/OsRp5rY/QCcmy2UqsQGsolJ2pTrBnuiAOjs/rCDw7o/qNhBcJZdVUQC0pmCLm199uwLZol9dgEc4Sw/OIqyCSYtlbywxfSfrmUPNY0axmTavPMXnj2KQ23D+Uz+s79x/R333pQNz802tIxxS/A9zTZvngSL24Zm5S+fZl0z7Ut/ay04ufOL9/6qK6767GO/Iy04kBoboVQMZohFHcg80K4Nx2ZLCeP62kN5lwVOzzTEFhN4uC+nQqhTcM55TKOTt04wBbfHtB+yxiBUCcQCPWp+CrhYC5TVyEUhgvnn7pOdx6a6Pg2rVZ0u2Usls848gOrybrWq063jM2U/Yar6gvZNBhXCFWTCqENA9nkRI1AqDM+94NH9PtX3a4fPRzOSTkzdSAUzxCKW8baezG/bV34YqYfNo09eGxa333g+EofBlowFVcIxYGQV/s9OzE+ob/6xn3asb4w71XWV1+4W6/8qV36w5ee3XD7YM5V3OU4mbJ6cabs6bM3P6IHj03LmJTb/Kpz7hfF8VQ/q1WFuuekOJBZKQeiCqrdGwe0fqBWaTJb8TUxW9WO9WGrHc8JvafqBzoxU04e19MqZF0VGSoNpNZWIBRVCG0ZCr/m9M1D2ne8s+3IWFy7gdBwPqPR2Yq+8IMDnTisNYtAqE/NJC1jmeTBLp5Lc/6uDRqdqejQ+Mq+wK1XXqmWsejvZrRuOHO7Q1HTmq14GkpdIRReCR6bqWhspqL1A1llXKeuQoiWsU6IBzp/695jch2T7iRVOnmGUJsVQtvXh1eSj/TBYOm//eYDets//3ilDwPzmChW9Z37GwO75PkhrgCq1p4LZmZn5Bijz7zxGfO+EHv/K/fqg7/85JMGrQ/VBd2TxXSPTX97/QN695fu0D9//xHtXD/QsBVzXt6c567SeKqf1aqg7qJJHMislGPTZWVdo01DuZMqhCaKVZ0aba2kZaz3HJ8uy1rplJHWKoQGcq6KVf49gbSOT8ctY+nbMwvxDKHoa/ZsGdL+47M9ddF8LSq3uWXsnJ3rVPEC/f5VtzfM18PSEAj1qTjUGKqbIRT3rz8pGpD7wNHeScBXsmVMkk7UrW/vZIVQEFgVq74GW50hNFvR6Gw1eX8o58ox6a/C97ogsLrpoRM98wQcv6j4/r5R7d44oGzasDKZIRR9fptDpbcO52WM+mL1/LHpskZnKpyI9qj3/Nudet0nbm4INJKWsdzJFUK7Rxx9/reepcdvTVkVV6d+NtpkisfR0ZmK/vF7+yVJx6cr6drFpLBCyK17wd+hCqEzTwmfKwey7opXCBUrvgayrowxDYHQdNnTeLGaDOOmVL73xMF+qxVCgzmXtfNAC+IKoXZmCMWvr0/fPKTpsqfj08xwXEm1odKtnRu+6sLd+vNfPF8Sz4fLiUCoT01Hg5GH8xkN5BrT73hzTC+t5at4QcuDw5ZD/DPH6gKhqQ4+gJQ8X9YqdYXQQNZVPuMkFULx2lpjjEYK2VXTHvD/vn6vfuVjNyWrrFfasSgQqvhBsnViUdaGAdAyVAhlXEdbhvM62geBUPyi6eAYPfe96ET0u3z34cnktqRlrEmF0N+/9vz0wcwcQ/na41qaK3PX/eQxzVZ87Ywq4k7fskglnleWTjwYVggNba3dXuzM48aZpwxr31+8VJdcsHPFf79LVT+pnoqfB6TwBMgPrE7dSMtYr4pbf1udITSUYxYG0Irj0xWNFDLpKk0j8YXhOESKt2/ecWh82Y8P6ZWjQCgO7FoRX3RnKP/yIRDqUw1DpTONM4Tigcadbo1qxYrNEIoeaE7MVGSMlHFMR1+AxSFc2qHSxoQtAvGWsU11a2tHCplV8+L/0zc+LCldVUE3HJ+qBYSnb055chyHP/VbxtqsEJLC1fP9UCEUt1seHO2dFlTUnBb9/t5Zt0igvqVYUkOFkOu336Y42GLLWPz7/fKnnCopxX3tx5+RPvqsMADK1VUwdahlTAofg3dtHNDx6UrX57lMlqpJ5V2p6icXd+JNbDvXF3R4Irzf0TLWu+Jg/5QWK4TC4ai9c+EO6HXHpsva2kJ1kFS3ZSy6aP70PZs0Usjo2ttXdpHAWtfu2nmpdtGdCsvlQyDUh0pVP7k6O5TPaOtIXtvW5ZMUPC7rn+2hF44rt3Y+/DsZm61oKJfRUD7T0ZaxOK1OO1RakjYO5nRiJtwytnGoPhDKrooZQmMzlSTYSjt3pJOCwCYtY1LKrUdSbcW84y65QkgK2wt6fcuYtVYnogqhA1QI9aS4DbN+s+RJW8bqhzR77QdCw3UVQmnaWY9Pl7VxMJtsKzt7x7qFv2DqsWQbmtafWru9Qy1jsXiGWLerhPa+9zpd/KEbJEnFqp9c3Dnv1PX6/rtfqKectlFj0ca3TUM5DWRdzZQ9VbxApSovhHvFkcmyXMdo81BrJ6rDeZeAD2jBxGxV6+sqKNPIZ1wVsk7yujyfcfWz52zXdXc9lqw+R/e1O1Raqp1jEagvHwKhPnTJR76ry/7jHklhSvrbP/N4fektz0k+Hpf1z/RQcrpSM4QKyVDpqgZzrobzmaTdrhPiB6f61orFnLIur6NTpbBCqC4QWlfINGxH61ffvu9Y8nYvDICbKFblBVZZN9zjnLplLAmEMkveMiaF7QVHp3p7qPRk0ZMXrZVa6RkraC4+oazfLDld9pTLOLXH3PohzV77IWRjhdDij03HpsraOpLXT522Sdf9j+fp2Y/fvPAXVOrm3p3+09LvfF+64HUdaxmLxfN5uvk7Hgdq8fKHUjVQoa6y9JR1BQ3WtUVsGMxpuJDRdNnTn1z9E/3qJ7/ftWPFwo5MlrR1OC/XMS19Xbw+uVdm6wG9bqJYbZixlsbGoax2RlsaYz+/d7umSp5u2d8bYwzWGj+wSRjXXiAUVwj1/zlSr0hfxoAVVaz4+un3f1N/9vLzdN/RKUlh2JFxw//qh33mXKfjrVGtWrm18+GDxuhMWcP5jLKuo+ly50KJeGNIKxVC20YK+uH+MZW9QBsHGyuEVnquxXK4P/p9lcKrOystnh/05F0bdMvDY3r81jYCoWWoENq+rqDRmYrKnp/8nvaaE3Xb+VbD7+JqFJdMH5+u6MhkWdvXFzRd9mor56VlrBAKv6dJOfD++HQlmdvwhG0ji/+Acu2xQpkBaesTpYENHW0Zk6RdG7tfIfSDfaMN7xervgbmzFKobz1eP5DVSCGj8dmqHjg63fPVhWvJkanWV85LYSDkBVZlL2hpJgqwVk0Uq+lfs0X+588+8aTxC3H78tEpHkdXwps/80N9/a4jktoLhHqx8KHfUSHUJ45OlXRipqJv3H1U8cWk4Xk2WRljkitPvSJsGev+C57aUOmqBvNucoW1U1qdISSFrUPxENhNQ7UrH+tWyQyh/cdndfrmQQ3m3J6oEDoeVeX8zgvO1Kd+/enJyeCi6mcIOUufIRQPID3aw6vn4+18+YyjA8wQ6kn1j/PxvJnpstdwkWC5KoTiIdV7Ng+lui/HFUKp1VcIZaKvG9ggVWeXFGQtJtn2ONO9x6cbHzzR8LPrh0rH6rdVbhjM6sytw7rz0QntOz6jqbK3KlqKV4MjEyVtbXHlvFR7DddLr9VQ88nv7NMvX/69lT4M1BmfrbRcIbRlOH/SaID44utYFx/zEfrB/tEkDJLU1sKh+KJ7L41G6XcEQj3u8ERRP/dXN+juw+GV01sP1MobFxqAOZRzeyo5Xem18xU/0GAuo+F8eIW1U9qZIXRK3WaSjScNle7/J6t9x2d0+pYhrR/IaqJYlR9YveqKGxtayboprhDavWlQz3/C1kU+u079DKG4ZSxo/8loW7R56cgKD5Y+NlXWz/3VDdp3fOakj8Xzg84/dT0VQj1qpuIloUu8EW6m7DVeMGioEGr/9+0FTzpFf/mqJ+vZZ25ONQ/s2FS5pfXAKtcFQtmoxL+wIfyzg3OEsq6jkXxGY7PdW0P8/f1hhVDVD9tOS9Vw7Xy9+pax9QNZ7d21XgdGi0kb56e+t1///eM30XK0gvzAat+JGZ3RYtWCVJvxxRyM3vTl2x7VD/aPrYrXYatBEFhNlb2WA6Fm1g1kZYw03gMXKdeav/vWgw3vZ1pstZXCDY0SQ6WXE4FQj7vnsSnd89iUvvfgcUnSg8dqJ20LBT6D+UzP9FZaa1d8y5gUXo178q71uvvwpI51aHZLOzOE6lfVNswQGshqquwpCPr3xb61VvtPzOj0zWEgNF6sanSmou/vG9V3Hzi+4NfuOz6jV11xY8MA6OUQ/9u3VLkgLXvLWNxisNKbxu46PKl7HpvSTQ+dOOljccvY2TvWabLk1QbZFsekT71MGtvfxSNFM7NlX6dFQ5Hj3+2pkqfhwvJXCBWyrl7x1F1aP5DVZGnh2SczZU/Fqr+ECqHocXFgY/hnh9vG1g9mu1rB+Oh4rZrLWhsOlZ4TCMVbxwpZR4Wsq/N3bWj4+KdvfFjfe/AEg4lX0MMnZlTxAp11yvDinzxHPKSdf7/eM1WqJpsbHz4xqwePTetnP/RtPe8D1+uF/+9bDVsd0R1TJU/WSuvrLpy2y3WM1hWyGu/iRQCEbjswrtM21yrzjWk9EBpghtCyIxDqcXEpcbOr9wsZ6vDw5FZUoiug7ZQFLlX9zxzMufr5vTsVWOmrdx7uyM9bsEKoOC59/lJp+mjDzfWzBxq3jGVkbVgB0I++de9RvfFTt2i24uuMrbUKofik69Hxov7y6/fp2tub/1u884u36/v7RvX9ObM2lurYdFk519G6Qosj1JoOlV7aDCEp3FCzkk5Egdv+BSqEztoWnuyMRi1keuwOad8N0sFbunOQmNd02dPjNjcGQtMLVQjd/3Xp6rdLS6gqWVfIyg+sZiu+/ujf7kwuWNSLj2XJFUIDG8I/O7xpbONgrmsVQtZaTRSrch0ja8OrnMXKyXNk4ueRDQPh88L5p65v+Hg8lL4XWnGX09hMRW++8ofJY1MvstbqnV+8XZ++8WFJKWdkzZFUCM15jr/3sSn9zmd/lKyzR/f9YP+o/Ohi3L7jM7pl/6juOzKtc3eu00PHZ/SNu48s8h2w3OLHueWoEJLCNtxOdgzgZBPFqk7MVHTxuduX9H1yGUdZ1/RUJ0y/IxDqcc0CoTTVdUM5d97eym/cdUR//18PLcvxpZGsFlzBtfOStHPDgJ64fURnnTKsa+YJIZYqLl9sWiF0x79Id18tffv9DTefUjd7YNOcodKS+naO0J9ec5f+854w/IorhCbnBEJ//18P6aofHTzpa621DS0V1lq979q7dPvB8SUf18GxonZuKKS/KnHfddL3PjJPhVD7W8bWD2SVzzgr3jIWV2A1C51HZypaV8gk4VUSCMWhZn1FB1bEbMXXhoGcNgxmk3/Lk1rG6iuE9n1b+tGnpDuvkq5+W1uzedZFL8gfGZ3VlTc93DTUjY+ltQqh+qHS0eNioTsVQvHJwV994z5d+ombdfVtj3bsZxWrvqq+1c4N4f/jTNlTuUnLWPw8siFas7xpKKdTNwxo96YB1T98rbZA6JaHx/TVnzymmx5a3osBy2my5OnztxzQP35vvyTpzDYqhIaazBB6bKKk13z8Jl17+2F9pUOvU7C4mx4aTTaR7j8+o0fHSzJG+vBrnqIztw7rjoNhhdDHb3hoxdrf15rxYvj6Y/kCoe5dBEAovvD4U6dtXPL3GsxlmCG0jAiEelxc5ROvppXCK65/8JIn6a9/5YJ5v24wl5k3OX3jp2/Rn11797Ie50KSQGiFK4T27gqvrv70WVt0R92K5uUU/50Xmm2NilsfZhtbc7YM52RMGPStq3uiiwe49msgtK0u6NpTN0Monj1y1+FJzVZ8HRg9eTbNDx+uzcqaKFZ1eKKkj//XPn3qew8v+bj2RzONUvvsL0vX/WEt/HBcyUS/V0uoEDLGaNu6wooHQnEV0P4TJwdCx6fL2jyc1+bhMKi857EpffBr9yqYiq6O1m+FQtdZazVT8TSUd7V1ON9YIVRoUiHk1pXaX/Ub0o8+LR1t/blgXRRWxwHtgTnr2n/0yJje8293Sgof31JbsEKos+uBNwzmND5b0eXfflD/df9xfe77j3TsZ8UBzq4NYWXXVNReV5izZSwOiOpPgN78/DP0m897vE6pC9pWWyAUb/450MNzy+rbzndvGmgc4p7ScJMZQrc8PKrRmYoyjkkGj6P77jsypSdsG9GO9QXtOz6jwxNFbR3OK+s6On/Xet1+aEJBYPWXX79Pn7156a9LsLhlrxAa6G6bMGqvM+cO+W5Hr83K7XcEQj0uvnJUn11sW1fQbz3/8brkglPn/brhvNt0c8W9j3X/BC5uGVvpGUJ7T90gSXrcpkHNVvxatcMyuPKmh/XA0SnNlj0N5lw5zcq4nOgF42zjVc+M62jLcF4bBnNy675uRzR0+JEmgYkkfe/B4/rmPb1btnxkqqSRfEYvOnubdm4YSK7Ax1d5StXw9+LgWLEhnHv4xIx+68ofanPUPjc+W9V9R8Lf25seOrGkIM9aGwZCm9t4Mnr4xvDPZZohJIVtYyu9Pjoesv3widmT5lUdHCtq+7qCNg+FJ59X3vSwPnL9Axo/FlV1lfu7Qmi24ulvr39gwQH9vaxY9WVteAFgy3A+qcqZKjWpEHIyUq7J730bVV7xC/JbD4RXyecOHP/rb9yve6LnmqVvGYuC9FZbxn74KWk0fSXsxsGsDo0Xk8elTga1cZvCqRsHkve9wJ48VDpuGRusnQBd+qzTdekzT9OO9QPJbROrrO0hbqPt5UH29bPtnnBK6+1iUq1C6NYDY/ryrYck1S4APefMLbp534mkbQnd8c17juiHD4/p8HhJOzcM6PTNQ9p3YkaHJ0rJa7K9p67Xsamybjs4rmLV18ExNnB2Qxze1D8eLsXGwSwVQl227/iMjAnPw/7s5efpN593RtvfazCf6dvXbr2IQKjHNQt16mfOzGe+odLX3l4rg+/WsOJeaRnbvSl8AR2vGV+uJ/EHj03rj/7tTv3Bv96hmYo3/4axeJjr7Mll8NvW5bVxzpPcOTvWyzHSHU2GF86UPf33j9+sX//H3pzhYq3V4fGSXvW03fr7118o1zFaP5BVserr+FTjE3Cx6ifrzSXp6lsf1YmZir7w5mdpONr8c/+R8ETx0HixrfXn333guO48NKFj02XNVPzWrk5senz450PXh38u0wwhSTplXX7FK4TizVRlL9DhumOx1ur+I1N64vYRbYqqPO5+dFKSFEyFLWNeqb8Ha173kyP6v1+7V++9+icrfShtiSsLhvOuto7k9dhkSZ//wSMqe8HJM4QyA7U2rMxA7Xe4NNnyz103EH7vuELo4Fix4fkkrp648LSNSZi4KN9rHHjtRY8J+XXRcbbwuzY7Kn3l7dKNH039JRsGsqr64f/DKSN5He3gbK/4xGbnhvA5KQ4X5hsqHc8QqhefnNZ/v9Uinp3Tyyfa8e/46ZsHdfF57c3DGI5eK3z8v/bpf33xdllrNR0FQj977jZNlTzd9Wjr90+0771X36W//Pq9enSiqB3rCzp9y5D2H48DofD+Gg93j9tKm1U5Y/kt/wyhHDOEOsxaq3+55YBmyp6+ctuj+uY9R7Vz/YAKWVeve+ZpevdLz277ew/m3L6dsdqLCIR6XP32iXgDVf2a8vkM5dymq0xvqhvQW6x2J1ldyZax+oqbeGbMruiq7HK92IznZ/xg/5iu+8mR+TeMVaMXDU1aH57z+C16xhmbG24byLk665QR3TFnbs7ND51I2jF61USxqmLVbzhpWV83d2Su+n+LRyeK2jyU0+O3DmvDYFYTUYVQ3M/fbBvWQsqerzd/5od617/erv3Hw5/dUstYPDvowbpAaBkrhI5Mlld0bfSJ6XISHuw/PqPvPnBcRyZLOjRe1EzF11nbhjWSzyjnOkm138yJ8IXw4SP9PTthMlon/PlbDiRvLyYIrL565+Ge2P5XP8R+y3BeB8eKeudVd0hSY/uRV5SyhVrVzd5flt76g/DtNkK97dH9+u7D4clqxQuSUMMPrB48Nq03PXePvvjbz254DF5QPD/ojJ8J/9xyVvinm5Gyg43zhRYTt8EdvjX1l2yom992we4Nmip7HdtgUmsZC5+L4nChkFt4hlC9Zz9+s576uA0N32+1iEPyXj7Rjv/NvvSW5+hVF+5u63vUv1aoeIHGZ6uaKnsyRnreWVslSbcfGl/ysSKdILA6PFHUXY9Oaqrkacf6AZ11yrDGZqvaf3xGO6KZX+fuXKesa/SVKBCaLHmr7j7Yi+LwZjmHSk+VPHl++7MgV1Kp6vd0l4AkPXB0Wv/ri7frSz8+pLf98491+8EJDebmOUdq0WDO1WSxqn+/ozdej/U7AqEeV18hdMaWIe3ZMqQn71q/wFeEhvIZFav+SeXGB+teYDWrPuqE8goGQrFLn3la8nYcCC3XfIJrbz+sC3Zv0LZ1eZ2YqZxU9p+I53gUT64Q+oOXnq0//8XzT7r9/F3rk3lHP3x4VFOlqn77n36kL/04LC8Pt9T03gPho+Ph/2t9W0M8H+nhut/BuGqs/oX/4YlS8sJrQ1TSe9+RKV142iatH8jqxwfGWzqW79x/XFMlT3cemtS37g0rW/a00jIWnzD7UcXAMlYIbVtXULHqa3IF50Qdny7rmWdskusY/cN39+nST9ysv7zuvqQq6wnbRmSMSQJpSXKLYRDkFfv76nV9ddaXo/vUYm586ITe/Jkf6aZ9Kz/fI75gMJTPnNSadVb91qO4QiieITS0VSpEzyPl1v8NTxkp6KmP26DA1u7D37r3mA6NF/XI6KzKXtD489OoRDOszn2F9N4JaXBT7WP5kdbmVR2LAqHH7pD8dCdq9aHLBVHQ0qkqoYk5LWNxmHZSy1g2DGrXNwmELn3W6brqt5+tjGNW3clorWWs2JPPb1L4b5aJKl/blXGdhjmHR6ZKmipVNZzLaOeGAWVd09NVUqvJLftH9dDxaVV9q7HZuIKvoGecET4OeYHVzuj1TCHr6oLdG5LqWqm32xtXi8liVfmMc1IlZbs2RhcB+vXx85rbD+vX//GWhhmzvSa+j/zokdqF8Je0WVE511Auox89Mq63/NOP9K37ji7+BVgQgVCPq18dv2Ewp+v/v5/Rq5/2uEW/bigqRa6/wlnxAj02WdJp0Yri6TUSCO2/7Of1p5ecm7w/Ushqw2C27SfwILC6M2rjOjxR1L1HpvQLe3foZXt3aptGNejNc8U9rhCqztbaIRaxd9d6HZ+u6Ot3HdEv/d2N+pMv/0SjMxW95+fP1h++9Gz5gV3RMGE+j02GT1BxsCPVrsAfGJ3VSFSRcuHp4XyQg2NFTRSrOjg2q8PjtdLsjYM5jc5Wdf/RaT1x+4hO2zzY8r/btbcf1lB0ReIfv7dfWdck230WFQRhIOTUtd84bli1IIVtLkuwLaq0WKn1wtZanZiu6AnbRvSKp5yqb9x9VIGVvvfQ8WRuUzwfY3PdcODBShiG2D4fKn1ksqzt6wo6s27z4EPHpjW1QLXQ4Wjm08E2WheXW/1Ww7ja4HlP2KoH//ylesETT6l9YlwhFLdkDW6pa8VqL9T7hb07JdXuw79/1e16z5fuqP3etBoIxfOo8k22NeWGWwuEjt4T/umVpGP3pPqS+OQg4xiduzMMyzrVzhmfgJx6UstY43PkpuGcRvIZnbm1+QYrY8JAYrxPT2jmc3SqpIxjVPaChpPuXnJsqqwtw/nm8wJbUN/aeWSyrOlSOBDedYx2bhggEOqCf/juPr3y8hv1B/96R8PtO9YP6Ozt65KweHtdxfOzooruTPTvz79T500Uq8tWHSTVLgKM9Wnb2InoeWO8h+cgxTOabn1kXJL0D294mv7Hi5+wLN97sO6x83sPrPwFun5HINTj6qt4WnkgHIxODurbxg5PFBVY6YnRC/VmLWWdELeM5VdghlBs7orxXRsH9PCJWd16YFz3H2ntpPYT39mnX/ib7+jHj4wlQ7rPP3W9fn7vDl2R+5B+feJvmn9h/XyMqXTrZOPVjP/j87dKkr4clSg/fc+m5AR9bBmHYy+XuEJoZ12F0Ka6QGj3pkFtW5fXM/Zs1sYonPvAV+/RL19+ox6dKGpn9MJr/UBW9z02pdmKrydsG9HujYMtv/D61n3HdPF52/X0PZs0W/F11ikjyqT9XSxPSrLSxtNrtzlRC4vUuM67DfE698dWKBCaKIbDbDcP5/W2i85SLuNo96YBHRgt6vp7j+qUkXxSnbB5OKxAMQq0LhiXJDmttPH0CM+vtTcdmSxp2/qCfmHvDn1//6juPDShn//wd/T7X7x93q+PW0UOT5T02ERpRYe+xs8Pg7lMMhfrV5952sltWl45nB8UhypDW6VMLqwaKrc3B+ql5+9QPuPo4nNrV/vuPjyVPJ6e1eoa7nigdK5JkJQfaW2A+dG7paEoEHv0x6m+JP4937aukDz+HJla/gqhI5MlnZipyDG1E8x4rtrcCqHhfEa3/NGL9OJzts1/3KtsU07VD0Ogc3aGgeUN9x3ryZaO49Pl1gamz2OoIRAqaarkJRtGd20c6Om2udVgfLai/3PNXZLCtv96O9YX5DhGz9gTVgnVX0h65uPDQOipj6td1EJnLX8gFFcI9d5r6DTiNvde3kQcL+95KFo3v3vjwEnnY+0arHu+vLHFURI4GYFQj6sfmNXKA2GyzrTu6+MnrCdtD19wd6tCaCW3jM3nlJGC/uv+43r5335XL/7QDQ1rzqdK1Xn7UWfKni7/9oOSpK/cdrjhavgFuzdohzmhp2bnWUFarXvBMJmuPeXcnev1+z/3RM1UfG0dycsPrLKu0RO3j2hj1MIz2oNXBw5PFOU6puEF856t4QmrF1itH8jqq+94nn77Zx6vXRsHdWCsGK12DV8Q79hQqxCKZ109Yduwdm0c0KE5A2wXMjZT0ehMRefsWKe/f/2Fuvqtz9E/vfEZ6f9HSuPhnycFQlHQVVnai/V4QPxKbRqLr75vGc7pcZsH9V+//wJd8boLJUk3PTSqc6OTMknJ1reNmlZG4X3arZ68qr7X/d/r7tWz/+Kbuv6eozo6Wda2kbx+Ye9OWSv96ie/r2LV13/c+Zh+8mhjUGKt1WSpmgRCdz76/7N33mFyleUbvs/0srO9l2R303tCSAMCAQKhN0GKgoIKqAjYENSfimBXFAuCVBGkiPTeAgTSe6/be5+d3s75/fHNOTOz2U02yW6yyLmvK9dmp+2ZmVO+7/me93ndnPy7pby8aXDH8nCQCJU2sXBcHqt/eDqL+xMPIgGxz6qiijNX/LSlCwfcYeYIffyD0/hiUjluS2+QVdVdlGYdRhtuVazqzyF0OCVj45cIcally8EfT8IhVJRh03L6htq5F4rGmPfL97j/w32k283YzEbMRknr9NdfKYTVZDzgADrDYab3f0gQUo+vBfEJ93f/s4ln1zYcy03ql3ZviNwk1+Thkpyn0dYbxBtKdAg8nAUQnUOjsUcslPYdn0qSEIcBThmfj9EgUZbt0O4/blQWaVYT8yuzSbOadOHuKNDtDw9ZhzEQjQQAun2fzvNnb0DM4bwjWBDqu2BdmLRIfKQ4kjLYtjf3jmin1KeBkTND10lBnfB6Q1HUxd5DORGqna78SS4gtdRmYpGY5B2tDKFjGSo9EGrb8R+cNZHcNAt/enc3IELaTvrNUh5c1n+74ufWNdDpC1OZ5+T1Lc3sbPGQm2Yly2lBkiTyTUGK5FaRh9G3nCiSNGBwD34S+Y1FY3nvu6fwu0unAzCxMB2ryag5bo61Q6g/h0RjvF15slMhzWrSSiQyHWaynBbNkdLQ7ddKcSDRQSd5nx9X4KI020E4JmsTqINR3SkEi/IcJ+k2M9NLMzUhbVCoE+WBHEKRIxWE4hPPYXAiDAbVKZMbd/8UpNuYWOgi32WlIN3K/503WXusKgjlSuIz8StWLLG4ICTHYIRmfSQTkxVeWN9IOCZz4xPrqOvyUxAvGfvxuZPo8oU5Z1ohLquJxz6pSXnuU6vrmf/L96juEKLKx3s6CEdl9rUlRDFZVo5q5olPC5UWA6OUhgPJ559oUDiE5PjAVxWErOnQuh1+Wwl73z3kv5/nsmI0SHxy+2nce8VMAJbt6WBOefaBn9gfmkPoCAUhfxf4OyFvImSWQW/TwZ8DWpfHokw76TYTNrNhyEvGktvDq5ORNKtpwC5jg+F/zSGkfubzKrJ5+aYTcVlN7GxJlDWOlEyhds/QOIT6lox5QlFcNrFvlGbZ6fCG9NbKw4iaE3byOHFOtJgMZDst5KZZtTHr5XPKePvbJ5PvSpxfbWYjb9yykK8vGktZtoO9bYfgYNQ5LFrcQW3MNBSouYhdI9BlPxg0h1Bo5J7/kxesXTZTavfTI0SNRinLtqMosCFelqZzeIycGbqOxpOrajnh1+8Tjsr4QlEtT+VQHEJqZkqyC6i+S7g2xsTzCI5Wu76RKAjdcvo4Xr7pRL6+aAw3nDyGZXs62N7Uy47mXtyBCM+t639F8uVNTUwsdHHL6eNo6Q3y1tYWJhTGJzCRAFIsiIQC/7wAHlyU2oUqEhTZHQA9A7iIBmBMXhpzK7KxGA3MKBP5FiPhYtbjDzP/V+/x05e2prh29rZ7qczbP7h5fIH4rJL35dL4KmizO7ESqu7zqqW3MN1Ght2cCAQf5GpcTdymekhdxZIJ9IifKYKQccgEIZvZSIbdfAwdQqmCEIDBIPHsDQt4/eaFVCZllxRm2JAkGG0VA99aJR+b7Bc5S389Ht796dHd+MNgbU0XbZ4QX5w/ilBUJhCJaS6try6s5OWbTuSez89kelkGu/sM8J9bV48/HGNtvKxAda4liwbfenoDt8bLO48G/qRQ6RSiIbhnIiz/q/g97EsVWpyigxG2DNGJS47CrjcPeztKMu1a6QQk8jUOiQNlCFldg+8ypnZxdOaBq3DQ5bkum1kI1HFLe0G8A+BQkizcqOfANJuJDk//odKDIcNu/p9qnax+5vkuG9NLMynPdVLdkRBdv/ufTXzrqcGVAQ4Xsiyy15LPm4eL02pCkqAi10mbGiqtlYyJ60xjj+4+GS7aPOL8fdJYMTYryhALBGVZCSdD8rg5mbJsB3aLkZPH5bKyqlN3KAwjiqLQ7A5SnDl0DpPCDLFo2V/n208D6vXk0+IQKh5CdxCgmSXUBajhyvz7rDByZug6GutqumnpDbKpoQdfKEZ5rhgUHIqzQZ0gXPngSt7YIgbEDd1+ijJs2kD06JWMiYmT5RhmCPUlwyHcIgDnTC8CYF1dN1viYdF72rxaOZhKU0+AdbXdnD+jmCVTCilMt4m23PHQXU08AGhcK0oVtjwHVR/CvTPB3wGOHDFR6ak75G12WEz8+2vzuOV0EciWCMQ7doOQt7a10O4J8c8VtTy5SohcMVlhb5u331BZ9bZkQagsy044KhOMJHIiNIdQ/HHj4kJSWXyAPFgbfU2HD4MEo5Ks3ofEwUrGIkdu5xet54/NhawzXjKW06f0oTzXqWUGqVw+p4x/f3U+FWliENKo5OJQAiht26GrClb+/ZCcb8NJJCaz+J4PeXNrS8rtr25uxmY28N0zJmgCdbKrZnppJjazkfIcJzUdPn75+g6+9rjo4rE+vvrk6XPeVHNmQtEY7+1o1QLnjwa+pFDp1I3aCr52+Oi3wuUW9oElSRR1xAUbW7oQgwBqlh3RtpRk2jVBQy33OSQOmiE0SEFIdfXZ0iGtEDwtB358HKNB4t9fncdXF1YCkO+yDr1DKEkQCseEgO60mLTvsW+o9GDI/B9zCKkdc1TxvzzXSU1nQhBaua+TXS2HF4Q+VPTEs9eGQhDKsJspSrdRmmXXQqXTbYmVbxCLeTrDgypAnpgkCP36kmn8Nu7IHgznTi8iKiu8vW1ktwBPprEnwPxfvkdV+6fD2dTlCxOKytrYcCgwG8UCQHXnp6/0HdBKhfuOSUYSXUmLFUWDbeYySNRKgUmFouql81Pq9BopjJwZuo6GenJavrcTXzjKcaOy+NPlMznzAMGSfUleMX5/p2jHV9flj2c7qIHTn12HUDLFGTaynRa2NPSwucGNy2rCIMGrm1JLDd7aJiYW50wrwmY28s1TxwBJ3XRU8UDFlgEf3wN1K6G7Gjr3iU4/maMO2SGkcnx5tmZTT7OaMBsluo5h/fOrm5sZneMgz2Vlc4OYiNV3+QlGZM0NlIzaijq9j0NIJdtpwZBUu5/lFI9TP2N1kjDYTmPVnX5KsuyHv+/16xAyCZeQyXbEDiEQncaOlSDU4Q1hkBL5KQfCZTOzYEwO2SZxEW5RsjFLMYI73hIPkGOw8r7h3NxB0+kNs7fNy0d72rXbYrLCG1ubOX1iAVlOC7Pjjpb+LOgVuU7cgQj/XdfAO9tbuW/p3pT7kyNd1JyZTfVughF52FqV94cnGMVkkPYX29Ug5aAb1j0WF4SSRFFj/PizJjKiaN8J3sNv3WowSIwvSKMk056StTFoDpQhdChdxrTXccUdQi3CxTYIji/P1pyXWQ7LkDtvkoWbuvh1Xg0QBrBbDs8h1HuA3LtPG/VdftKsJm3RoCLHQWN3gHBUpjcYockdPOZdgVQ361BMTm9ZPI57r5xFvstGWzxUWi2rUK+N9XpL82GjtTdIjtNCZV4aNrOB4gw7lXlpjM3vR5gegGklGYzKdvDG1sG5EUcCu1s9tPQG2dE88htD3PCvtfz+bRHrUDTELpPyXLH4c869ywasDBipqB2GR3KodLcvrAncQ/3dqZEUYwvScFqM2gKnzuExMmfon3HUMpf3d7WhKELcuWhWySHlC1TmOvnROZMoybRT2+UnJivsavEwocCl1V16QzG2Nbn5wkMrU9rTDzVq8OlIFYQkSWJaSQabG9xsaXAzuzyL+ZU5vLqlOSWvYEuDm8J0m9bN5/I5o/jxuZM4b4ZwGKU4hBw5MO9GMcnqEiHUeFpEqVHm6MNyCPW33VkOy1HNENpY38OV/1iJ2x/B7Y+wfF8n504rojjTrmUAHajtdH8lY+oqKMBPz5/MHy+fqe0rGXZLyvNsZiN5LivrarsHNQGq6fBpeVGHheo2yByVuE1tQW+2H3GoNECBy3rMuox1eMNkOy37d6Xqy6434KmrQFHIMsYFIeIukF2vi88nfxJ0VQ/zFg+OnnjXkOQOgquqOunwhjk37ghUXSz5/eSAqMe4uuL05Ko6Fo7L1QTJ5Dbgqpi3Yp/ocuEJRY+a2L63zcvoHMf+ocNNG0R5qjNfuLfUkrFvrIQr/p14nE0VhOLP37f0iLbnR+dO5jefG/zKegphL0hxobUvVhfEwqIU7mCkCEJFoMSEO/MQyXQMvfMmWWAyxL+z5EwFm+nQBaF0uxlFGdmTgkOhoTtAaVInmvJcJ7IiFrT2tAo3g9sfOaZZQs3xLppFQ1C+MiYvjTnl2RSkW2lyBwlEYqRZxfUxL82KzWygtlMXhIaCpbvauPFf61AUhdbeIJ+/fwUb63u0LLQ/XT6TG04Zc8ivK0kS00szqPkUfU9qnlnPCO+w1eMP89a2Vp5dWw8MjQibTHmOkx3NvWxv7uV7/9k0pK893PR+GkrG/GGmFIuYi+Ih/u6+v2Qid14whUXj88hJs9LlOzZZnP8rjMwZ+mcYtz9Ctz+C3WxkU30P0E8+xCAwGCS+dnIlJ4/PZU+rh6p2L75wjGmlmRgMEg6LEV8oyrvb2/hkb+ewrRLIssKza+upzHWS6zxye/VwMb00g50tHna1epgebyFf1e5L+Vx2tXoYX5gQOSwmA19dWEl6PABSy6445XY470+QLUoPqFspfkZ8YrKTOQp66lPzhQ6TbKflqHYZe2ZNPSuqOrnvw73UdwuhcUZZJsUZNm3VdE88e2VcP4LQlOIMvnvG+JRW1SWZCTfBgsocLpxZov0+vTSDm08by1lTi7TbvjBvFEt3tfPHeBD4QCiKQk2HT5vcHxbBHjFJTXZSGOKTNrNjaErGMmy0e0JHtX35utouvvvsJto9ocGVPTz3Fdj1GnTuI9Mg3nPYIRyLtpZ1UH6y+IxCx7aUQ0XtGrKrxaNNHF/b0ozDYuTUCaId+ZVzR3HL6eP6FS6TM6escXHy1sXjtX1pWqkY4OS7rHT7I4SiMVYmtT09WiHhWxp7tNLXFJo2QvFMIYoEe8W5x+IUot3EcxOPU/fr4pmQPQaW/2XQbpr+mFuRzUnxcNZDprsGMkpS7Vcq6nYOpvW8ug9a04VDCAadI5RMht085JMlVWD6/pIJPH3DfICUnK7DcQipjlE1C+XTTkO3P8U1qh6LNR0+TeANx2T8xzBoWb3WDeUEZ3RO4j2rrjGDQdLKV3WOnBc3NPLmtha6fGFWVnWyuqaLbU29mkv0rKlFTCgcvDMomTyXVcsC+zSgRg2M9PwxNcZBHR8NddlRRVxwBpjQz1hgJKOGSh+t+I/DodsXZlJROrefPZGLjys5+BMOgQy7mS+dUI4kSWQ7LXrJ2BGiC0IjhP+sref59Q1audgFM4q1+9L65kMcAuPyXXT7IyzdJUoBpscnMk6rCV8oyu42McA60IDjtc3NPLHy8Eqc3trWws4WD7csHofhYC6EAyHL8OYd0Lbj8F/jAEwrydD+f/6MYs6aUojRIPHaFlE2puXi5PdTzqCilozNuAImX5BwlSSXh5kd4nY5MuhsiwNxtB1Csfhk8cmVdVquQ47TQlGGcAg9/HE1T6yspSTT3m83AaNB4lunj0vpzmK3GMlNs2I2SvuJE2ajge+cOSHFUXTL6eNYOC6X17c0s2xPO/d9kFrOo+INRfGEolpns8Mi0AP2zNRJquYQcoiJ9hGSn25DVhIBz8PFSxsbeWaNcKY9t66R/65vYEtjz375Qf1SEO82VvMRLimAT7GSliEm/hIKVCykzm+k192lPWV9XTe/en2HJsgEIzHueH7zUSmPc8cn8r3BqCbObGl0c9yoLG3Sneey8u0zxvfrjirLcmiBhb+9dDrfPWM8s0dnaW6z0ybmc82C0VyzQLRcb+sNsbOlV5vUHY332NobpLU3lHLuAoRI2bYDimeJ8iu1DMzSjzBqyxQ/M0fBKT+A1i1C+DsWtO2E/Mn936eWkQ1GcNQcQunCIQSHda7NsJsJRmRC0aETHlRB6MZTxmirpvOTArith+GiVUNWm/oE0//2zZ2sr+s+3E09JiiKojmEVCrix1xNp4/drQlBsOcY5iY1uYOYDNJ+OWtHQvICSlpSGWF5jvNTm3Ey0lDFhYbuADUdCTeP2ljgSMhNs+IJRQlGPh0d4VQhaKQHYatRBIAYIw7xwnLy4k+6feg6YA0nq6u7+M2bO7XcTU9wZIp6wUgMXzhGTpqFG08ZkyL0DzW5aRY69JKxI0IXhEYI932wj8eW12jCzOfnlGkTErXE63BQV7//u64Ru9modUpIs5rwhqLsbokLQgcYcDy2vHq/DI3B8v7ONrKdFs6bXnzwBx+I7mqRT7L95cN/jW0vwu63+73rhLG5XDq7lBe/eSLjClzkpFk5YUwOr20WZWP1XX5CUTnVTRCLwDs/AW88p0QtGbNnip+Zo/f/Q2YbZMVvP8wcoWSG0iGkKAp/fGd3vx28NtR189gn1Vq4pTcU1RwRWU4LxZk2/OEYf3h7FzFZ4Usn9PPeD0Bplp2CdNugRENJkphRmklNp5/7P9zHX97bm1I+EIrG+PUbO7XStSNqDRx0iyyoZJJLxoYoVBqGX0T414pa/ho/jrc09sT/5iAdQmp3vOpluPDjwUFebsIJopSfxJZ2hYAnMQH9/n828cBHVWxrEpP49XXdPLW6ng92HX5OzWBJzhjZ3SpcQtWH4BazmAyUZjkoy7Zz4cwSvnX6OCAxeCzJtPPzC6cyJS7G7Gnz0O2PMK/i6HW72BIfKKsiv4a7UZRJ5YwToog3LoaY+xmMqSVj6aUw7VIwWqBhzTBu9QDEItCxW7SK7w9r/LwbHoRDKKg6hFxJDqHDEITiuVruQIS9bd4BhedDwR2I4LKZUkTIufF9Bti/9G8QqCUUzT2Jc5HbH+G+D/bx5MojL00+mvT4I3hD0RRBKMtpIcNuprrDl9Lo4WguhNR1+rn33T3adaa5J0BBuu3gpbaHwLikxab0JEGoIs9JfZefaEzmpY2NvL/z0xNcPJLwBCNUtYtxbn23P2XMOxStzNVxxtKdbTz88cgonT4QqhB0rPO4DsbWRrcWIVCYMbgx4qFQkRQpMNI/C5X/rK3n7x/s034fqeXCqug4mIzKIyXHqZeMHSm6IDQCCEZi1Hb6aO0NUh3vijS1JF1bQezPaTFY1OyVXa0ephSnawMYp9WIOxDR2rnuavHwt6V7+1Wa67sCNLmDA658KIrCY59U9+sy2tLoZnppxpEPnLqqxE/fEUwmP/odfPzHfu9Ks5r4/WUzmFmWqd127rQiajr9bGvqTeTiJNuJG9bCJ/fCi1+HtY/EJ1ISWOMTNFchGBLOFkCICKpQNAQ5QrlpFtrjIbY9/jAPfLjvsEuPGroD3PveHl7d3MyLGxpTuiU98kkNP391O3vbvdo+pa7WZjssWlicPxzjW6eP4/qTD60O/3PHlXDZ7LJBP35cQRoxWWHFvk4CkVhKl4U11d3c/+E+nlotas4PqxPM1ueheZNwfakuCpUUh9CRZwaogtBwt57v9IWp7wrQ4w+zqyUxscoZzIqb6syoWYYTP17FTnqGCGVuM5fQrOTQHbNijSXOAeoA+bV4l0M1B6OpZ/jFkmQb/K4WD12+MJ5gNGU18GBcMbeMa+aXp9x26oQ8Fk3I04ThApf47lZVCWfUvArh9jgawdKbG90YJJhcnJ56h1q66sgWoognPoG09ONuVEuxMkpEKaQ9O/H8o0lXlXBN5k/q/35VEBpMsHSoV5TmmiyQFm/EsOoB2PPOIW2S6kp0+yM8u7ae3765S7PoHy7uQETrDtn37xwuBek2JAktww0SjSlU4ffTgto9sm8oudppbHerR8ucO5qd1V7Z3MQf392tdUAT7a+HtnTFZUvsB2qGEIgJaySm0NgT4Gcvb+PP7x25MPlZZGtjwl3Y0B3QxtqQ2mnycMmLjzPueWc3d7+2PWW8rCgKDy2romsYRcyYrPD3D/YN2vHTrTmERrYIsrnBzekT87GaDEMeSgxQkmXn7KmFjM1PO6oi85HQN2T+YILQ1kY3r28ZuGz6qdV1/S4EHy5tniAPLavSXO/ZziO7xg2G7DQLnd7wMc2W+7SjC0IjgH3tXmQF2j0hqjt8FGXYsZqMWujp4WQIqeS5rCyozCHHaeH8pDI0p8XEtqZeorKCQYK3t7fyu7d28dLG1M5aoWiM1ng2wUDBhh/t6eBnr2zn7te2p9weCMfYHc/kOWI642q49whWx/ydhxQuumRKISaDxMubmtgZn0Anr+IRja/I7n0HXv02bHterLgb4oeVwQiZfUQOswMySsX/3fWH+040ijLteEJRPMEIr25u5ldv7GRVdefBn9gPqquhtTfIrc9s5Ly/fKzdt6fVo+2js8qEELC71YNBEqGmhUlZCofzfV+9oJxbFo8b9OPVCbmqfbUlOTI2xydBqoPikB1CsSg8dy08cHKiZCwZVRCyOIYmVDpuV2/tDdLYE+C1zcPTqUS9OL+yuZlILHHRzHUNYvVGdV342skKNuDBgTUtE4B10hR2t3rw4sCu+NnW5GZlVaf2N1SXnSoYq/kbw0mPP4zFZKA8x8GLGxupiv/titzBW5a/sWgsXzu5MuW2yrw0Hrt2rnZOVr871S03vTQDu9l4VBxCVe1eyrIdOPo6SAPxsj17lhCBwnERpd+SMdUhVJx4jr9r/8cNN2op8ECCkNqKflAZQp6EgGSK79tt2+DZaw5pkzRBKGnhxD3A5GnZnnb2th1crHIHIv0KQLefPTHl+nwomI0G8tKsKceVeqztbfMetYDzI6G6w8fSnW3aRCfZIQSi09jWxl7aPCHmlotxUbc/TLM7wIsbGrXHKYrILBzqTA21e40qWDW7g8MyOVVJ7jynithvbWuh2y/cavqk59BRxVGb2UB9l3AInTW1kBllmcwtzz7wkweBOs7Y0+ZFUWBbUy/PrqlHURT2tfu4+7UdvJC0rw41mxt6+M2bO/nWUxu021ZWdbK9qf8yW7Xksq+AtGJfp5ZfeqwJhGM09gSYUpzO5XPKOGPS4DstDxajQeLvX5zN2VML6faHPxXdGtXzEIjmBwOd7z7Y1cb6OrFA+n8vbu33Mb5QlDue38LTaw68QB2JyTy1uo5o7OAZg398Zzd3v7aDe9/bA8Co7CPI8RwkOU4LUVmhNzDyr3cjFV0QGgGo7hNZEV2c1MHQudOKGJXtYNThtPCNI0kST10/n3X/dwZfOqFcuz3NatJWK45PuhgmB6OCWM1Xxx7V/TiA1DIjgGifE+n2ZjeyAtP6Cz09VNROXWp51qGiKEIQ8g3++VlOC0umFPLwx9Xc/+E+ppdmpIpz/a2k27NSf0/uTgVi5dpsF64Tz5Fbv9VygRZ3UBtMr9x3aIJQMBLjpY2N2urnnqSJTZsnSCQma1ZrgOllGRgksbKU6RAdqtTVUrNRYmLR8AfzVeY5U1xnrUmODFUIUt/HoAWhna9BJCg6w6n06xAa2lDpnDTR4aS1N8SSP37EN/+9fshzCMJRWVtFen69aK2qOhUGVZMfcmvdn1y+OjyKHUtWCZ220fzbP4ftzb30KnasRPjZCxu47bnNWshtXZefvW1e7fzRPMxOKBD7ZpbDzDdPHcvWxl4e/Eg4DCtyD5ABdhhkOSyk20xsaXQjSTAqx0FBupXWAcJF39rWMmTfbbsn1G+HNO28ZM9KCCPQvyBUMEU4FouPE787slO7JR4t2neCZIDc8f3frzmEBpMh1JsaAj/5QvHTeGi29WRBSBVYBlpN//Yzm/jjO3sO+poDCUI3njKGv1w565C2L5myDBMlze+gXqzVY01WYHvzyAh6PxD3vLObm5/eQFP8GtQ3960816k5gtSyzB5/hD+9s4dbn9nI3ngjg50tHm57bjOvbU5d2Nrb5h1wYjwY1FKE+i4/sqzQ4g4OebgtJNyiyeHi5XERW3W8ekPR/fKidA7OrhYv+S4r4/JdbG3qpccfYVZZFi9988TDDpJOpu844zdv7uS2/25mZ4uH9vj1YDjDwVVRYNmexKLnj17Ywi9f7z93UxWCkrO4FEXh1mc28P3nRka3rSY1vD1eot13gWYoyXJYkBX6dYH6w1He6NN1+FgRjckpY6iSTHu/gtBz6xr48qNruOnJ9TT1BOgJiM7AH+5OnQOp88DWg7iaP97TwR3Pb2Fl1cEXjKzxbpnvbG9ldI6DSUdhTqBmYXbqZWOHjS4IjQCSgxLruvyaXXpGWSYf3XYqWc6hr79UhQ2HxcjpE0XXHUmClVVdKSe9hiRrYn85Q7tbvWyMryY0dgfY0+rRSl/UMLj9Qk8Ph74OoY69olPXwahfI1aVw17RtjjQLRwgg+S3l05nfmU2pVl2Hrzm+NQ7+1tJN/eZdKnlYab4AFfN8XAVHVbnm76oq5RN7qC2arCiavCCUDAS49w/L+OWpzdqq1eb6xOlYm9ubaG200c4aVVgdLaT7LiIkBUXFfJdIk9hQqFLuxgMJ1aTkfKkriy1nX7W1ojvQ93vZAUM0iDrl1u3w9NXwfYXoXmjuM2Rk5ohJMXfV4ogdOQDPKNBIi/NSm2XX7uwD3X5WLJVfUNdD4XpNubHy5sG7RDKEQ4uU8xPyOigoiifD854nWXRKbyzvRUvYl+samimoduP393FDWPFd7GiqlM7fxwNQajbHybTbuHiWSWU5zh4e3srRoO0n/PgSDEYJJZMKURWxMDMajKSn27r1yG0q8XDDf9ad0Dr9qHQ4Q31L3b6kxxCKYJQP2JYVjncujmRa2bPSjiMjhaKAjtfhbxJQizvj0MqGfOkvu/PPw6L7hDi7mDa1sfJjAs33f4ItXE7fXc/5RjBSIwOb0hzoR2Invh+OdSca1zNLZ13Qe1yQFyrXfFr/NOr64e0HGA42NLQgycYpaE7gMVo2E80S87+mhMXhNo9Id7cJrKhVFelunDR1htia6Nbc3T9/NXth9VSut0TYleLR+te09AdoNMXJhyTKRqCMqO+fPdMIYiqpaggSpFKMu0pJU4f72lnx6dA6BtJNLsDlGTZKc2yaw6YQykhPhjZfcbpq6vFeXR3q0dz5x4oq/NISb7Gq4vMnb4wu1r7P2cmh0qvru7CH45S3eGjtTfE7lZvSl7XsaI5Xl4+nG48FfX766+s72uPr+XrT65nX/sgHKrDSF2nn0/2dabEQhRn2vEEIynztrbeID96YQsAoahMiztITFZ4+OMqvvzo6pT3qF7T+huzuP0RrTFBXfwaMpjmJ8kC1XnTiw4rG+9QUaMP9E5jh48uCI0A9rR6UtwOQz1p6Q/VXv6lE8qZW5GN2ShxzfzRdHhDPLeugUC8pasqMpgMEjuae7UVMnUwsmKfWI04d1oRtV1+rn54NT95SVgTN9T1kO+yDkkHB80hpDp8nroc3rjtwM8JeeDRs2DF34Q7SMU/eMHEaTXxxFfm8eYtJyeCB2UZWrYmVtJNNhi7WPzfk7oySclx4MyD3Hg5lDn+Gq7CIekylhwo2hA/YW+s79G+v/7oDUa0CcLb21vZFx9Er4kPYJLzeJ5cWaeVy6ktdkuz7OTG1Xj1Imo0SEwqcnHi2MNsOX0YzK3I1gJ1//jubi69fwWb6ns0pxMk3DcHpSPewr63CZrilmt7dmrJ2Ak3iZ/G+P48RKHSAAUZthShYKhFk74X8etPrtQGwwfNEFIU4brIHavddMascZRlO7SV1Q11PYSN4vXS8CMr8BXpJb7fdAujMsx8srdDKzlt7gkM+0pbTzyrxWQ0cHM8ELosy47ZOPSXvHOni05W6qS1OMPW7wRcnQwk272PhHbPAIHggW5AEs42a5IIZBmE09SedfQzhHa+Ci1b4MSbB36M+j4GEyrdVxACSI+3u+1t2v/xA6CKEjubewlHhSDeExAD762Nbm0fVsXb2k7ffvt1NCazsyUxcXcHoqQfYWYQIJyyXYng2rEG8b7q178ZL1/zMqMsk8o8J/9d38DPX90+0Csdc9z+CDXxc8P25l5y0yz7TSDU7n5Oi1gIcFiMvLalWYR0W028GncEqcdYqyfI5x9Ywd8/FOOGVneQui7/IZ93fvfWTq5+eJXWvaa+26+5qCvyhtZtCHDZ8WXU/PpcMpJypiRJ4tFr55Dvsmo5ez/47xbOvnfZkP/9/2Wa3UGKM+zaOE6SGFLngtlo2E8UAiHOqA6hZKf1UJOcf/Pxng6iMRl3IEK7J9RvNo4qBHR4w1z+jxXc9eqOlMXEVzcN7lxZ3+Uftk5lzZpDaOjF176oC+/Jji4QLvlP9orPpcV97Nwn0ZjMNY+s4iuPpTZ9KMm0E4kphKKJRdu/f7iPqKxw0cxiuvxhzbG8q9WDoqA5MSEhgPWXe/jbt3byub8vZ2+bRzMHDEYQUj8/s1HiwplD22p+INRjTy3vre/yf2oyoUYKuiA0Aqhq9zEjqVPMcLbmU5kRL+O6fmEls0ZlsfmnS/jayZUYDRLff24zVz64kjU1Xexq8WAySJRk2XlpYxMLf7uU59Y1cPa9y6jt9LGiqpPSLDsnjs0lHJVp6Q2ysb4HRVFYWdXJvMqcI1eHo2ERwGx2iAlB+27o3HvwUObeZpCj0Lg2VQQ6hLIxEAOylM4Gu16H+08UIdIWF9xWBef8TtzXdzJ13Jfg29vAGRdKUhxCRy4IFWaIQFHVIVSUYSMSU9jU0DPgc37zxk4uvV+sJNfGV7XNRglfHxHpjrMnsqvVw31L9yFJcNW8UaTbTBRn2jVnQmaS++b5r5/IbUsG6BI0DNx14VSeu/EEXFaTdgFSa5YnF4mSkbzBBkonC46qIORpEUG3asnY4jvhx+2JXJKhKBnzdULIS2WuM2XVRx0IqcJCsztwwNrtmKwcMJtHveiPL0ijJNPOVfNGMb00I95N6yACdNgLiqw5hACtJGdsfhqOeHlDRakQRtIQ2zFZqsUkhzmzTOGtba2EojKVuU584Ri9Q9wVQ5aVFCGwxx/WSuIunFnC+II0LaR/qDlxbC4F6Vbt9acUZ9DsDtLZZ+CUyB85ckEoFBWfYb/7d6BLiJgGQ2rpVH8lY31RM4SOpjV+/eOitHbqpQM/xuwEJFFaFjyIMyLYu39nwAxVEBp8hocq3CSfS3v8YZ5YVcd5f/mYd3eIBgeqeOsPx2jzhPCHo9p3/88VtZz1p2WsqRHO294BSsYOmZdvgj/PhGohChTHxOStaePbXP3warY19TLR6ePVr89h4bjcEe0Q2tqUcKTuaO7t1/WmitfjClxIkkSWw8LeNi8um4mbThvLnjYvjT0BrVRuZ7MHfzhGXZf4vd0bwhuKDipAt8sXxh8W56c9bV7aPCFtMlTX6efP7+1hbH4aJx3FxY/xBS6W334av7xkWkqZ6EgoYfk0oCgKTT1ifDQlHsL/m89NH/KxtrpQlpwBtbvVq02im9yBYWtL3+WPIEli8ba1N4g7ENFO433dPtGYKCFXG9Yoiuhc9fz6RgrSrRw/OouP9hw8b7M3GGHhb5dy4xPrhvz9QOLcmpxROVxkx8ey33hyPXc8v1m7/aUNCWHsaGQD9qWtN0goGuP5DY3UdPq1aI6F48T5Rx2/qZEA3lCUf6+q4+JZJZwwNhdFQRtbqou/yYKQ5hDypL63aEzmja0tKArc+95ebfwyGAdOuyfEaRPz2fTTM1M7Mw8j+XHjQbNbLDhe9dBKvvb4Wv0ceQjogtAIoM0TYmpJBqpuUnYUHEI/OHsiG/7vDE0Vt1uMlGY5WPrdRfzu0ulsbXRz2f0reGx5DSajpNXtAzy5SrRL39niYVV1Fwsqc7Q6d/X9rKzqos0TYkFlzpFvbFeVmJCWxku2tv5X/DxYyZV6f9NGMfFWOYRg6X5RM2Ya14IjS0y0sirEbaNPTH2sJIHJmpigqCURrkLRDlo+eEDbgVADRavavXT6wiyOh+4dyO67rrab1t4QPf4w9d1+8l3W/WrojQaJ606qYFx+GtubexmTl8aNp4zhg++fisVk0Cai2UmCkMVkGNI2vAfDZDRgMRm0CwHA+zvbyHZaOG+GECdyB5sfpJYkepqF+wsSYbyqQ0iSEmIQxEOlfYc/eZZl+F0lPHkpv7x4Gq/cdBLLbjsVEAOh1dVdLPztUu54fjMLfvU+T64aWAB9dXMTp/zugwG7mKh11X+6fBav37wQm9nI2VMLWX77aeQcTDRTJ+DpxYnSx7jQYDMbee+7p/Dqt07imkXTAHBJYuAwziCyik4pSggjp0zIA8T+OZThjS9vauKU3y7VwsVFhlDCvfbfr5/A7y6bPmR/Lxmz0cBbt57Mt88QgtnUeInslqQufZAs7h35oFJ1LPRbMhboFu42OHjJWF8c2RALDZnzbVB0VUHxLDAeoHmCwQBp+bDhCSGGgGhV31+oe78OoXiQ/yE4hIwGCZfNpJVEg1h9/FM8M6+/kPTqDhEee+n9K4DEwPuNLS0EIjHCMXloBKHmePnTc9eBopAfEULXbOM+fnhGOVbC3LbvGhxr/055jvOolGkeLsnHiScY7df1lmE3U5Jp1xyh6md45uRCpscXt6ravdp3si2eF9TsFhl4XUklXwfjuLve4aw/CaFNfT11srW2tps9bV5uXTzuqF7rQFzvjAZJ+wyAIQ/PPhZEYvIBHc1DQY8/QigqU5Rp53PHlbL6R6fz+eMH39V0sKjnY3WynuUwsyfJIaQoDJs42+0Lk2k3U5QpSpaTy1t3t6U6K9U8ruRxu9Egsa62m5PG5lGaZafTFyIak/cLpU8O1n/04xoAtjUOT/lisztAbprlqMQQZCV1wkou/93X7tUWvfqKJsNNMBJj8T0fcv8HVfxrRS1j89MwGSQMEjx4zfGs/78ztOwc9Vzw3g6x+Hb5nDLK+gietXEHZUuSsNXlS3SbSxYrV1R10uULM7konVc3N2nXwS7vwQWhDm+IvDTr/g0vhpG8NCu5aRa2NfVS2+mnvivA2tpulu3pQFGUYT/H/C+gC0LHmEA4hjcUpSDdpg2ESo8gRHqwmI2GfrOJRuU4uOz4Mt75zin88fIZAEwocPHj8ybz1NfmA7ApntHy9rZWevwRFozJ0col1DHSPz4SE2y1U9oRoQ5+x54hfqqCkL/zwJkQqgPH1watWxK3+45QEFKdSf7ORIi0JMG3t8NVz/b/HNVlok6oXUXCvXQI5WsDUZRhY22NcCYdX56Fy2oaUBBSO7+BmLw0dAcozbJrYbum+BdY4LJiNhp48mvz+NdX5vLP6+ZiSrJEqwOf4ci3OlRUC7i67501tVBb+Ru0Q0gVhFq2iAlxchh4X7eBitkOSkxMTAdLsnjUuFb8rFuB3WJkWmkGZdkOMh1mmnoC2mREDRM9UP16baefcFSmuiPxmGTBRbXRlmTatXIESZJSJ18DCVtqkK8tA9KEoJM84S7KsDO1JAN7mjgW8s0him0RSiSxb5+YF+KZ6+fz36+fwHnx8qrL7l/Bo8trBnw/h8qq6i6issLuVtGBp8cfSSm7cNnMBx+cHMFKUqYjMWidWiLEMjXcXF2d0xxCPf0PKhVFSVnN6k8wi8kKsqzQEZ9g9Fsy5u9KnJeSRaBBOYTiQtLRyhGSZZEFp2atHYhr34Cpn4NdbwrR5/274KHT939cyJ3qjIJEFzV3wyFtXobdTCSm4LAYSbOaeG1Ls7ZCWtu1fyZWTYeP1dVdVHf4aPMEtVKzd3a0aOUiOUNxzoz4wWAW17bualy+WsLp5ZiUCNePbub9z9uxRD3QsoXCDBvuQERzvYwkOuMl6ulJjoqBmgD89+sncNtZwoGqTt7Om16kjT1qOnxayVggPrFp7gmmiOQN3YObjNfFy2C6kya/aibThAIX50wtGtTrDAe/u3QGt58tPofOQUzORjq/fkOUpQwXye7R4gwbBoNEvmt4HCelmQ7Ksu3MKM3EZJC4YEYxtV1+GroDmI1igLK1yT0sroUuf5gsp4V8l43W3pA20QfY2uAmEncYy7KiBUmrpZgAr928kH99ZS4/vWAyWU4L3b4I/1hWxZl//Ejb3q2Nbmbd9TbL93UgywqPLhdlq4fcyXWQNPUEj4o7CFIzoFrcQe09V3f4mFSUjstm6resariIxmQ21vfQG4yyvq6bXS0eTpuYzynj8yjLdmAzG8l2WrRMOvXc98qmZgrTbcwelbWf+1vt/NqUNAZJLqtq9wgRsKHbzx/e3k2a1cSdF05BURLXuYOFNsuyQqcvPLhsyiFEkiSmlWSwpcGtlT6m20z830tbOeV3H3Dun/US24OhC0LHGNVKmhfP2jEZJK3TxLGkItfJxbNKWfq9Rdz3xdmk28zMr8zWQoRBdMwBmDUqiwKXjTSribOmFgKwdFc7RRm2lODfw6ZpgyjPKT9J/N65J9H6+0Bt6JMdRHvfS/z/iAWh2sT/7QnnFBklqZkdyfTnEOq7jYdJUYZdU/zLsh2ML3SlBJUno3Z+A3EBqe/2U5rloCL+PalOoaJ4l5d8l42F4/L26/qiTkSznUOw2n2EqILQudPFpO+8aUVa3tGgBypqyVjnXvGzMMlN0rfLmIoaIB4Z5IpfTx38ohAa4kLQ1ufjr58qOBVl2Gl2B/ezJ0cOUDKmlqiookNvMMLsu9/hlXgOQKcvjMkgkW4fQBTxdcKvymDf0v3vC8ZX8G3pkFaQ+H9f4iLRuEyFEzMTQqfB08S8yhxmj85Ksehv7eOgORLUlsLVnT7NiTGoMHGVoBt+PQr2vHPE2+KymanMc7K50c2bW5uZ+fO3aU8qO2kaoGTswWVVLL7nQwDW13Uz6Sdv7reafNn9y/nxS1u1FecBHUKOvg4hKSFGHwhVSDpaOULe1v0F2IHIGQNzrxeP3/UG1K2E9l2pLktF6d8hZE0Tx9khlIyB+C4BTh6XR6bDrHWzGp3joKYj/n32BMiwm7EYDWxpdGvC7dZGt3YM13cF+NZTG0izmjhj8hG2Tg70iP1V7Z62512ksAfL3GuFALj9JUp64ueYrn1a/sZIdAl99z+bqO/y8+crZ2kO6X5FTkTZiFrikpdmJcNu1so17WYjWxrddHjDKc6dNk8w5X3XH0QQSnZDvLM9dWwxa7Q4Nm5dPC61hPwok+W0aNfp/4UA1a2NbrY39+7nRBkqvv7kOi6+7xNg+EuPbjtrAv+6bh7XLCjnpZtOZF5lDooi3BbT4lUA335mE/d9sG/I/mY0JhOTFbp9YbIdlniXy4QQ6rKaeGZtPVc9uJJn19Qz71fv0Ro/JlRBqDDdxtj8NBaOyyPdZibbYcEbirKtqZfGngDt8fHFCxsakRV4f0cbNZ0+evwiw2u4Sqma3YGjEigNYDcnXEj+cExrX17T6aM8x0nBAM0ihoO23iDTfva21iFuRVUn4ZjMuPw0fnPpdB798hztsQvG5FCQbuXvS/fRG4zw0e52zplWhMEgUZRh69fJmOxq7UpykrX2Bvn6k+s56TdL2dro5veXTee4UVkpJZAdBxGhu/1hYrIy+MXYIWRaaSZ72jy8t6ONPJeVx66bS5c3TF2Xn6oO3zEp+fs0oQtCx5g2dWCfZqUw3U5Jlv2oW5EPREWuUxMDJElKaSHvDUVx2UyMznZgMEj86ytz+dkFU3DG7ZU/u2DK0KTLN2+EwmmJVV6AOV8TPw/Uut3bKiZBkhFqPxE/JaPIiVn/L3jglMPbnuTsor5t5gdCE4TUUOn4CuMQ5AiVJK0ClGU5GF+Qxp5WT7+rUKprAWBfm4/mniBl2XYto0G1oxcdZOCkOYQOZdI9TBRm2Fhk3MifWr/Mo1+YyoIxOdpnMqhA82BvPFcqaV8tmpn4v1oy1hdV3BusINS0EaJBqF8lJq47Xha3h30gx+2sm57mH96bae3x0eQOku208OA1x1OZ60wJOuxLhy8RpAewuqqLbn+EdbViYt/pDZHTT1irRvtOUSLXsmX/+9SSMWsGOPPj/++nLjzuyrhyRibH29X9WgJ3YhJekG7j0S/PoSDdOmQW3mAkxq548Hl1u49V8XD07L775iu3wj8W9f8i7kbhhGpY0//9h8j0kgw21ffw0Z4OPMEor29ppqE7gNEg4QlG+y31WFPTzb52H8FIjI/3dBCKyinlNIqisKPZw9Or61hVLQS3fksiA11JJWNx4c7sEGVXB0MVkvrroDgcqOfSrPLBPb50rgiI3vYitO0UDr1k8SriF+XF/e2f6aUp++Jg2BN3U543o0g712XYzcwsy9TyaprdQUoy7YzKcfBmPHMBRLfDVk+IeRXZnDw+j+oOH9eeWH7krkr1M5twNhgtsPU58XvBFJhwDux4Bfa9L27rrNK6YQ3kTDtWtHmCfLS7nRtOrmTRhHzt8x2MiP+9JRN48qvzsJgMSJLE6BwHS3eJbEA1IwZEp8ltSRlFfUvGGnsCzPnFu5o4nXyOfXpNahfTq+eP5tEvz9EWvY4luWpHnUEEvB4LXtnUxBn3fHjA3DsV9TvZ2zaIwPhBIssK5//lY+59dw9Ld7VrzojizOEVF3LSrJTnOrFbjEwpzuC4UYnx4Zi8NB750hzKcxws33f4i5LRmMz3/rOJS/++nNc2NzP+x28w48632dPm1RxC7b0hrWTsL1fNYtGEPDY1uFld00W7J6Qt5qrjpL5OfvUctTe+sFjTIZrJqN38VlZ3atemxZML8MUrHYaa5p6gtrg33PQdG/3oxS1c9LdPaO0NUZHrEELbURIUNje4CURiWsdc1Wk6odBFbpqVyqRAe5vZyDdPHcvqmi7ufHk74ZisRSaYjAaKMmzYzKnX/ne3tzL7rnfY2+ah25cQ0ZvdQVbs6+Sksbk89/UTOGtqEUZDIjLEZjbQ5QujKApn37uMFzakOm4f/riaBb8W155BxzUMIdNLMpAVeHdHKwsqczhuVBYv3nQiPz1/MpDoQKzTP7ogdIxJdgj94KwJ3PP5mcd2gw7C9Hg+xuz4atm0kgxttWzWqCzyXTaevXEBL33zRJZMGYKBkxwTJWPFs8CRFOI48yrx80AOG08zZJRC8UzxuyNbtBL3d0DNx0JoCh3iIESOpba7d2QP/NhkNEFIDZUeOofQtSeW8/0lE/jDZTPIc1kZl++i2x/RVnWS2dzoJt9lpTTLzsqqTqKyIhxCcUFoWkkmwH6OoL6owZZq/fKx5MsnlPPzaZ0Ye2o4tSiCJEkUOWBj0S+5LKf64C/QIYKoKZiauK1wWuL/AzqE4t/lYPNWNBfSPvE3exuhaIYoHfSKgFoa1lAWqcLirqLFLUIwz5hcQEmWnfYDrMyodd3q4Fq1zKo24i5fmOwDdRNTJ5n9Oe60krF0keMC/U+4466hXFOIc/I6iRltorteH1fGqRPzGZOX1u/+eTjsavFoA/5P9nbw9SfWMaHAtf/5Z92jwm3YX46MWiKVLPa+eQe89aPD2qbZ5dm0eUK8vU18nk+uqsUbimqT1eae/fcZtUSwrTekDVyqk7IMfOEYgUgMWYHH4uV2uf0df4GehFCtOhYHUy4GR98hpH7eg3EIgRC1xi6Gve+I0jBI3WeDSftqXzJKUt2dg/lz8UnCaRPztZDy8lwn5TlOLSC2qSdAcaaN86YXaY6NPJeVLQ1u2nuDlGU7+MfVs7nn8zP45qljB/xbg0Z9DzljhAhUv0r8nl0JUy4W313jOiEKRnyUmcRnMpAz7Vjx5tYWZAXOmyEWetQV5cEIQqVZDi2rC8TCVbsnhNEgce601HKuzfViP0m3mfYThFbs66TdE+KjPUJMSu6go4rpZdniWlicaePUiflHpYXywciOH/fPr2/k4vs+0SaMI4Xl+zq0oO8DEYnJmlvhYG3Ob316Aw9/PIjrOcJhuaXRzV+X7kn5bAZynw0XhUku+TyXlVMn5nPC2Fw2Nxx+2djv397Nc+saWFvbzS9f34HZaMAbitLuCcUdQjY8oajmSJ1XkcOZkwsJR2VW7BPjgn+vrsNlNXHxrBLuvWImv7x4WsrfUMunVLdjTYePjQ09tPQGGZefxramXpbt6cBmNnBCXExqSxJL3t7WwvSfvcVJv3k/JXPoUOjwhvCEoppb/WjwzPXz+f1lIirjtS3NWm5Oea6Tgngp3tFgd1viWFDH5iCaePTH5XPKKMqw8d/1DZRk2plVlqndV5bloDTLoZW8ghhLdPrC/Hd9I12+sPY3Vld34Q1FuWBmMTOTXmN+PAt2anEGnd4Q3f4IO5p7WVOTOk54d3urdrwdC4dQcr6aeq0dk5fGFXNGYTRIbDlAsx0dXRA65iRb/8cVuDShZaRyzYLR/OqSaZwyXmSJTEs6AFWmFGcwI+lkckR07Barvmro6MUPwE3rDuyw8XXAw0tEBxZXIZQvFLebbKIFvK8jMRHZ+y78ZTbct0B0JTsYaucplcE6hFSXieoqUUtv1I5WR0BploNvnjqWz80WwamqnVx1TSSzpcHN9NIMKnKdrI0PeEuzRM37z86fzIUzi/nT5TP54vwDZ3rMrcjmJ+dNPqpt5geiIN3GKCW+H6iTw+4aMru3klb15sFfYPcbIBlg8gXi97RCcCWVdQyUIaS28Q4P0Ep224vwzBcT2TRqTlHXPqj5SPx/xpXipyqaxLe/PLSbfe0+zS6dl2bVcmP6Q63rVksi1IFfTZKL4YDZJeokU+3A9+q3RfcnSCoZy0jst9Z+PhOTVTgWgm5cde9grDgJMsr6zW3Jc1lTVuMf+biaE3/9Pl8/jG4lm+MrldNLM9jV6iEUlXnk2jkpGUIpbH9p/9tUAaQ7STDY8QqseUiUIO1+C569ZtA5Q2fEPub35vvp8IZIs5q0Es455UJAbupTviPLCrVxd1erJ6iVwNUkCULq9//ztP9yLS+TYTdj3f0q/Gk6PHK2yLKKRYSA17dkbNCC0FHOEOqpET8HKwgBVJwMsSRx1NeW+H+HCHzeL0MIYNR8aNsOK/4G/7pY5M95WuCRswZ0Dj3/jRN46JrjcVhMWkfFihwhoCuKEOwae0RZw3UnVYgujBk2Fo7NZVNDD22eEAXpVmxmI5ccV4otqSzhsNFEtNGJzn+TLhCC0Lgz4dQfwQk3wxl3ApAXFcffSHMIvb6lmXH5aVoXGjVz4nAm7arD9YQxOYnXi7+Oen6YOSprvxJMdYKgOmfVc9JXT6rQHjMmvhp/tMWEA6Gey9/a3sKGup4U4XgkoG5P1QDbdcfzW3hmTR3NPUGthH3PARxCgXCMVzY38+HuwXWIfTXuZInEFJIN98fCfT8p3vFUFVmml2TgCUb5+hPr+dvSvdrjVlZ1cvXDqw5YGg6wvrabiYUuzEaJxp4AV88fjcUopnJZTovmit7V4sFuNmK3GLXwaFWgi8QUzphcgM1s5MKZJdgtqecl1a2ndrSq7vRpx8jNp49DUeC5dQ1MKc7QFg+TxZInVtXRG4zS0B1gW7ObYCTGFx9axaakgP6D8cjH1UgSnD4xf9DPOVLmVeZw4lghfiRf6stznOSn22jzBIck/+kXr23nsU8GFjf3tHopTLdx10VT+cXFYqGyLNs+YA6i1WTUBJBzpxeliNa3nTWBOy+YQmY83iH5PPbq5ia6fGEqc51YjAatTHZ6n3nd5XPK+On5k1k0IQ9fOKaFUycvbMmykhIDMFy5UgciP93GvVfM5L3vnpLSKMduMTIuP027Fuj0jy4IHUVae4N89Z9rUlah1P9nj4Bw3sGQn27jyrmjEiVGcUfJsNG+K/6HJ4mfM66A3LHC6WMwCYfNrjfg5ZuhdRs8/QWRF1S/UjiBXIVQEReE3PXgzBGTXnVQve0FkRvTtl2sqh4M9Xn5woI4aEFo7Bmw6A4oiK/EmCww64vCtbDuscG9xiBR24m/t6Mt5XZfKMredi/TSjJTwgTLskTJ35dPrMBpNXHRrBLKDhJsbjIauO6kiqPS/WFQqO6btu3w1FXQHG8bejDBTVHEPlB+UsIhlDkq1RV0oFBpEO6eF76eKNNQWf5nISqoYk9XlfjZWSXEyvSSRC6WKprEnULTDVXUdfm1/I88l5V2b2jAwUhyF51uX5gdLb04LEbquwOsru5iW1PvgcW7ZIdQoBvWPgo7XhW3qYKQNb3fUOkUrOlQ/aF4vSmXCFdGP7kteWnWlPPgK5ubaOwJ8MbWlkNuy76loYdsp4UTxoj3N6c8e3+Hm6IIQRjE990Xf5JD6LXvice4G0SJ3+63xO/bX0p8FgehoOENLjV+hIMg318ygSvnjuLq+aO5NC7a9p2YNrkD2sralga3NrhWHV5A3FGlcIX0DleZPhDuoN1vCTGvbrkQHFVhS3MIqYLQIDqMJT/vaDqEnPmJY2kwqMeMiuqu620Sol1GGVT0Uw485WLx860fimN1/eNQ/RHUrRD/+mFqSQaL45k/mfYkh1D8+vfrN3biCUZZOC6XdJuZP14+kx+fN5njRmfR4Q0TlRUt42xIeO8usf2WNPFdLbodFt8Jlz4iGhsYTXDKbXDmXdpCiKWnmtw0yyEfV8PNzhYPc5O6lx6KQ6gvFfHr2XnTi7Tnzx6dCYhW9mlWExMK0qjr8qeEtasTBNWRp56TbjhlDPd/8TjuvWKm1qlnJJRHq9jMIuR8oLbiA/HJ3g5++EI/ZcFDxPPrG/j9W7u0fK2afgQhfzjKM2vqeGJlneZikaT934M/HOVrj69lS4PIGIrJSr/Oyr7EZIXXtzSzoDIHi9HAlOIMnvzqPP5y5awheIeHjuoKDcXP7+oi6pvbWnhhQ+La+Ny6Bpbt6dAm2yAm2d95ZiPL9yZKzJrcASYWujh5nLgWX3xcCZOKxHk+22nWzjc7mj3avKIyN3H+VztmqWVF/dF3PlLT4WN3q4d0m4mzpxYyN76wMS4/jfz432uLd+Dq8oX5ZG8HF8SdfzUdfva0evl4bwfv70wdkw6EOxDhseU1nDe9mHFHqW25Sr4rkbujllqV5zopSLcSiSkpQfOHg6IoPL26PuW778vuVg8TCl1cPX808ypysJgMjM8/8Ofw+ePLuPm0sVx3YkXK7bNGZXHi2FwtfDojniM5oyyT+q4Ae9q85KRZOHFsDi29QWxmA2PzUscLLpuZa0+s0MSkHc3iWFXz2R79pJo7nt+CJ6ls8FiUjAFcOLNEE/GTmV6awcb6nsN2rH0WOHo94XR4b0cb7+5o4+1trVw1T6yItntCZDstmI2fLm3ulHF5XD1/tNZGethQJ/rZlam3GwzCyeFpgWV/ENkfbdvFz+RJW1oBlM1P/J5dCVtfSLQUTxYMBlNOoD5m1ALx95JDpQ+ELV0M3pM5/8/QuF50TZv95cG9ziBw2cwsGp/H61ua+cl5k7WSvm1NvSiKODFmOS209AbJd1kPKv6MeORYQmzZ+ryYHKu5Pi2bIRYduKV1yxYhCC64KVEOlTU6aUKdAYYBRC+1ZOzJz4mfvQ0w5jTx/+6ahMDYtFGULqoOIXcdBHtE3kd6ibjt4z+KfTk+uZ1mEKtHqkMoN81KOCrTG4zu17Y6Jit0+cJIkgi4fWubyDG5dHYpj6+o5Y7nN5PttHDNggO4vjRBqB1qlwNKQsgJ9Qrx1WyHCecKQStngNIXq0uUeBrMMPFcIar42sVE/cL7tBKmXJcVfziGLxTFYTGyp9XLnPIs1tR0s7Kqk4tnlQ68rX3Y3OBmWkkGlbmJSeF+hDxC3DGYxfcSCSbyvCDhiOltgDUPwr73xGcAYp/qjQt2vY2pmVKKAm//GCZfBGWJoEepU+yP46QGTpt4Nl86oRwQ39WobAdPra7jyrmjuPfd3VTmpaWs2r23U6zSVeY52d7Uy23PbeIbi8bS7gkxSmrDEumlXOrlphPyYGetCDeP+MT5qGCKeBF1/1WDzwfrEDLbxH59tDKEumsPzR0EQuTPHS9cPRFfQhCqXy2Oqy/8JyFcJpNdKbLBmjeCq1hcN6bGj91BnPvVhgoVuU4mF6UzqSidD3e3M6U4XQuKPn2S+JncEXDQHY3qVsHOV4WYMxDLfi9+hr1iFp0zBk66tf/HZpSJ/b1rH1+wVpPZEAOm9//YYaKpJ8Bf3t/DT8+fkuKO8gQj9PgjKdce9RjotwzyIJw2KZ9rFozmvOnFWov48QUuPtnbiTcUJTfNQnmuk1BUprlXZD5FYjLbm4Rw3tgToNMbot0TwiCJCfFZ8U5ilblpFKRbsZhG1hgt22nRclsGKwi9vqWZf6+q47YlEzTH25Hy+7d2MXt0FqdOzOdfK2vZ2ujWSnhVQSgYifGzl7fxzVPH0tob1LKdtjWJcsbjRmXt52h+Z3sr72xvxR2IcHY8t2kwwehra7po84T48XmTOXtaIcUZ9mPqZL72xAraPCGumivOc+MLXFhMBsJRmZoOH6FoDKvJqLl6qzv8jI1P/ne1enh+QyOdvjAnjM1FlhVae4MUZdr58pRCJha5mFyUzrTSDDY1uMlyJBxCjT0BreOlGrweiMS4+fRxtPYGWThu4LF7Vp9mIdUdPjq9YSYUujAZDTx67RzufW8Pnz++TPt7ar7Of9c1EJMVvrawkje3tVDT6SMtHkqcvMBxIDY39OAPx7hiTtmgHj+UGA0SBS4rTe4gv7pkGnWdAdKsJk1o++o/1+CymfnxuZMOS6xqdgfxhKLsbvUiy8p+AfUxWWFvm1crxTMaJG4/a2KK46U/LCYD3zlzwoD3qyXPPzhrIsv3dfLNU8dy2u8/wBOKkuWwcMWcUSzd1c6U4gxMA8xHc+Ln6O3NYo7VFBdo73xlu/aYP1w2g+oOX0qJ2kjg0tllvLihiasfWcXzXz9hwPf4WUb/RI4iahnAsj3t/PqNndR3+enwhg5rAHSsyXCYueuiqVrXj2Gjs0oIP/05ElyF0LA6EQSr/qxJai8Y7BGT0FN/DJc/ISYEIbcIHgUxEciuBIsrNT9kINq2i7KY0SeI3wfrEOoPg1GUMTRtSu2UMwScN6OYNk9IW2EDcZEFseo9syyTB685nl9cPO3o2Kh9HfDmDw8tsynoFhkufZ+z9hHhjFBxNyRKSFSBr26l+BkNisDkvjRthPfvhm3Pi6DxSRckBKHMUfGSEwnsA7iDICEIqXTVwJbnYPOzsOlpcZtkENsU8oK3BfLiTrdQL0w6L7H/NG+EN3+QJAjVYiQmHEJyjJMb7qdMau03WLrHH0ZWYHy+i0hM4ZFPqinLtnPhTLFCt6/dx42nVOI80LGqlkp5W4V7CRKupWC8jbckQXoRnPXLgQW27rgNevaXhHAy4SwomyfcNTtfFSVN7/yECkSJX4c3RJM7KOrWZxSTYTdrA+PBEAjH2NPmZXppBqdMyOP8GcXaymQKqmgwfonIbGrd1ueF+jhiVIGxaKZwPLXHS5H6lhZ17oMVf4XVDwjHyZ53xbEcf/6to/ZRuu63Wmi40SBx8+nj2NbUy/l/+Zg/v7+XD19+FGXLs9pLrqrqwiCJbnm+cIxn1zbw5rYWOrwhpktV2uMuLoiXvo49Xexn7TsTXfJUkcVgEOc2yyGIvunFh5y1c1jIMbHfq+7PQ2Hhd+HUO8BoTZSJqgLmQGIlwCk/gPnfgLN/I9ylm58Rtw/i3J8Rn0CX5zixmAz8+6vzuOS4Eu6+aOp+uTKVuU7NqZI/mGB7gA3/Eq7C7pr+749F0YLvz/n9wV/PaBLCWet2ro7+l8u6HqRt01vwyb3aQwLhGHe/uh23LwRLf5nIUxsi3trWwlOr69na6CYmK/zqjR1Ud/i0LJ/ktsjnTC/i2hPLD2tMkZtm5ecXTsVpNZHvsvKFeaNYMqWQM6cIga43GE1pTw9CRAlFZS6cKUT5W5/ZyId7Osh2WlOuidNKM7jptHGH9wEMI8n5fYMVhFRBZahKzFZWdfLXpXv5x0dVmsCmikEA1Z2JEuan19TzxtZmzY0lK/Df9Q0YDRLnTCui2R3UQtwhUfa1urqLP78n9ktvKEpv8MAr/K9ubsZmNnD6xHyuWVCuOfyOFU6riZ9fOFULajYbDXxz0VjOnFxAVFao7vBR3+XXyrmSXVXqtfCTvR10+8J0eENEYgrFGTZmlmXy/SUTkSRJc+pnOy2aYwcSrjY1eB3glPF5/PT8KQdchE52w5Vk2qnp9LG7zaMJIE6riR+eM4mx+WmkWU04LEZae0O8tLGRX72xgxPG5DC1JJ3R2Q6qO3zae+rPMdYfqjh4MBFkuCjKtOOymbhoZgm3LBbH/vGjs1g4LheDJPHRnnbe2Hp4DWHUYzUQifWbsVXX5ScUlVPEputOqjhiUVMVgKeWZPCzC6aQ57Ly+bjgFo7KzCjL5ObTx2mLV/2hOse2x4Xc3mA0pTug1WTgwpnFfG/JhBGRtZbM3IpsfnjORDY3uKnp9OMPR7nr1e1657EkRpaE9z+OeiFUTyRpViPtntAxqbX81NC1T6yC9seY0+Cj34r/501MnfiXLxSr4vO+Ln4/5fviZ9PG/V8nvUR0I+uqhqW/EqVmc2+AvPHi/j3vwK7XRSvypg1iFb7iZBhzOpTMPrL3VzRTCBzd1QO/z8Ng8aR85ldms7Kqk/97aSsvfOMENtT1UJRhOzb727rHYOXfIH8iHHfN4J6z5x0x2S6aIcqPVt0vhLg3fiDKuG7ZJL5j1UUGEA0kfhpMYvLfsBoKkwKjFUVk5DStF2VElaeIUsJYhmjlPOEcMZG2ZQxcLgapZS6L74R3fwovfyvuTpJgfDybZPuLifDwcWdAu2glyoRzoe9FM+KD/ClY27YxVmqkKOMkaNrAhN0PcInhc3R4LxTBgrvfiv+NM7VysbOnFVLb5WN3q5cbTxmjlQXmplm5en75wO8jFhWTaYMJ/J1QFW89H+wRTpFdbyacJwfjtB+LFvZLfiF+L5oB170Ff5omnDaxMHxyL9PGtvI5QyaBvWaaM2eKj6MwnXkV2Vog9mBQSwmmlWRQkG4TZQFyTDiuZn4hIfL5kgShna+K77406dgdyBEz+8vw6q2J33sbhItq7SMw5yuJLKh9S2H7yyLrLH+ith+e2vYvaJVh4jlQNheAi2YW8+bWFq004CvyfxmzrZ108/2kpzkIdrdwQ+YqCvNnYCfIl41v09xgIz2vlGmGahSDCUmOCgG8txFmXC6E6rYdQuiWjKkB6VbX4B1CIM6lbf2IqENNyxYhNqoZb4fCjCvEz1UPJMQ+d6MQaQ8k0k88R/wLdANSIjNrEILQiWNzWDypQJukZDktAzaBkFq3clvWR3zfM2/wJWNt8fNC9bL+u6752gEFzr1H7HuDoXgW7HiF3JAbJDC/8EUgDDO/CCYrDa/dy7Orx7HAsofTl/9GnC9P/8ngXnsQqBPA6g4fkgQPfFjF7hYPV80TbkW1HAuESyS5K9PhYjBI/CIekvvrS6ZjNRkYl+/SBKHqDh8njs3lqdV1mI0SN5xcSVW7l1XVXYSjspb5MtLJcaqdPs3siWeUybLCP1fUcNrEfEbn7H/Mqyv6NZ0+Zg3is97e1MvG+h6unFtGtz/CU6vruGbBaFw24Tb407tCKF9X183WRrdWFgVCFFW/f/WcvqvFi6wo5KZZ8ASj7GzxUJJp5/zpRdz92nZe3dzMt89w0eUL8+Hudr4wbxSrq7tS8oWae4JUR3zsafNy6exS2nqD3PfBPi1757UtzZw2Mf/ACyDHmFsWj2N7Uy9vb29ld6sXT1zkMkgir0dlRVUnDosRfzjGm9tatH2zsE8r9kUT8jh5fB7TSzNxWU3MKMtkU30PsaTyyIpcJztbPClxAQNhNhpw2Ux4glHOnlrIQx9XE4zIjO8n1FiSJPJdVt7Y0syjn1QztyKbh750PJIkUR7fB1S3SHWHD0VReGVzM8UZNo4v799hv6fVS7bTcsxyuy6aVUKHJ5QiauSn2/jXV+YBcPJvl7Kr1cMTK2vZ0dzLhTNLUspf+6IoCg9/XM2504u0YxWEONTXob+hTixOJXdLHArUkufkcsBbFo+jpsPHRbOEKP6dM8Yf8DVUAX9rY692myou57usXDSrZEQ7b6bHs21rOny8vb2Fhz+uxmEx8t0DOKs+S4zcM+b/ALtbPdR3+Tl9UoHWGlm0LhSr/LtavbR7Q8wegkHQ/yyd+8Qkrj8W3SEmXv4uMZH/5F6RrbDnLSidA4t/uv9z8icLh08sLFbVFVmU81hdIntoT9x5EuyFzz0Im56BF28UkywlJoSjGZeLiebVzx/5+yuO17Wv/DtMuwxGzTvy1wQcFhNPX7+Ap1bXccfzW7jxiXW4dyzl2imDL8U5bBrXiRKdykWJ29Tclm0v7C8I1XwsShv6vnfVpVH1Iex8TQgrFpf47nztQiAauxhW/UM8zpmXmOCBECNCXnj7J5A/Rewr1R8J0aNpvXAXRIOJbBGjCT7/eOL59qyBO4yBKMkoPg5O+5HYp0CIQaNPFBlXFz8Ar9wMW/4jJpxZFUJgqF0unqO2AT/pO7DpqYRoNP5MaNvGObmtYvK5VogO4wwNtHtCrNm+j+NeuA6jMwfGb8bbtIvLjB8wt3wO8yrm8tu3dnLFnDKynRZOGZ/HhTOL9wuNTMFdL/btwpnCsdG+E3InQMcu4RrwNMHF9w/8/GRO/v7+t0mSOD5XPQCtWwEoqH2FP1h8xN56hKppfwByGF+QxsJxuby9vZX1dd3saO7lklml+2370p1trK3tYl5FDu2121liWM300tNFt7eNT4qcrnd/JsSG3AnQmeR6KD5OfDfNG1O3MdAt9q2wRwTWe5rF4yaelyoItWwV76N9p9gf1ZBtfzzjoWVzIthYPb+A2O/igpDJaOChLx0PwBMrahj1ZjP2WIBLs3az2T6fub0f8LXgM3Q5r+DZtHuYFt3K+7WtvGP/OSeba5CKZgjH3c5XxfeWOSou4uwQ+1/+pFRH0MRzD82Fkz9J5LLtfVfsq6VzhdNrqFGdnBWHIQippOUnxL7eBuFuGszKpD1LdBJsiWeNJYeJD8DEwnTtezsgigIvfZPL2jaxteQeCgYjwCtKIi+vZhkcd/X+j1HPD2pDhcFQPBM2PqH9akWIx907PiBL8jJu8295wlKBvCM+IO7c18+L9Le5Cs+sqeesqYWsrOpkQmF6SjccFdUhUtPpoy3ubly6q10TFJIdQsOBxWTgV5eIMjlZVrCZDSzf10GLO8gza+r5/PFllOc6eeaGBfzwhS38e1XdiCt1GAg1WPqsqUU8s6aOvy3dy6xRmdz5ynb+/sE+nrvxBEblpE42Ew4h/36v99LGRuaUZ6e0Zv/5q9tYWdXFzpZeIjGZp1bX8+6OVp786jya3UFWVnXFF566eHyFOIbUMe4pE/L45/IagpGY5nTZ0+bBF4oysyyLgnQrb21r4YzJBeSn25hbns1/1tYTlWWW7mwHBb4wbzS3Lh7Pd/+zicpcJ48tr6HZHeCJlbW8t7ON6aUZPPhRFf9d36BNdG0mI1+cd+CmGCOByjwnRoPEf9c1sKami4mFLqxmoyaixWSFVVWdnD+9mA313Tz6STW3nC4m7EV9WrHnp9t4/Lq52u+PfXkOX39yneYSBlgypRCDQTrwWCCJbKcQ7a6aN4qP9rSzu9WrBbb35bSJBby8qZEzJxfyx8tnan+jItfJh7vbccSPqd5glM0Nbr7zzEYWjMnRBBaATm+I93a0cdnxpexu8zC+YJC5d8PA1QdpqjK+II2NdT28saUZWRHZjXMr5g74+G1Nvdz92g46vGE6vSFNbNvd6qU4085rm5spzrRz1bxRrKzqJMNuZlLh0ApCJ47Npc0TTCndTbeZefjLcw7wrFQK0m3ku6zauRxERz+A28+eyCXHHYX5xRGgZs1tbXJrnVpf29zMd84YP+IcTceCT8eV71PE8r0d5KdbGZvv4qcvbWNDfTebfnomO1s8RGWFG08Zw0PLRHr+5oYemnuClMw4em0VP1UEe8VgfyDnjMEAZ96d+H38EpELseetgSdAJotYPW/eJFpit+8UDiFbBsTiJ7nxZwtH0NpHhZOk/CQ47f/gkTOFg0MVcYYCdTvXPCi2+9Y+gY+xqJh4TrpAbPshcunsUp5YWcvqHdV8YvsT9p5S4Eqxqt6yWQgqQ4miwH+/KtxWF/4NZn1BlNu0bhW5HVUfismsM8n++vK3hICkOn5U1MnJ5qfFynXZPNFiOWOUcPy8fzd89HshBORPFpPBve8mnp9RCkt+Bf88P95VKO4ekgxCLDrpO/DenTDp/P7fS8XJB558WRxwveqmiedW2bPhmpfAGK/Bn36FCLi+9OFEK/uvvZf6Oot/KkSpB+IT49EnwuoHuWV8N1S9JoRKYILUwEeeEK3v/YI5ihd6vLDnHSa/eT2/M3fRu6KZ9Cln8sINV2p//5/XDTxIAUSHtBe/IT6TCWcnhJIZV4jPZt2jwgVXcfKBX+dgzPyCKKkKdMPcGzCufoBexYHbXMwZm77Nhc4fkOk4lyVTC/npy9v4ymNr6PZHeG1zM498eY42iOnwhvjGk+sJRGI8vryW31sf4j7LuxgMX4fNb8Fr3xUB7iBcaX1LwdIKxPHb1ykY6IaCyaLz1Nzr4eWbIHuMyKLJmyRcXbZMWP/PhIviw98JZ07J7EReVMQvHFUgcsZqPxHOq5plcPL39vtYJmeESZfEfnmxeTWt6YuYZhBCaPaWR8mObqXZMZ5TfB+wtGMnU5W9UHKN2N4t/xEvkjk6LuK8LkSDKRel/pFzB1FelEzeRCE0Pf0FIZha0+F7e1Izlw4VWRblmZMvTAStb35WfMbp/ZT4DRZnvhA0QYRKq5lcg6F8oTgHZleK19j0jBD2Jl0AmQNkV0SCQoibcklC0E2meZMoF23eBJKRO9NfAuMg3DzuBiFGGswi8HrDkzD98tTSTLU0zlU4+PdYfFzi/8d/BX/LbpT61TRvfJusAjthycIEqR6rWurZNThBaE+bl9uf38L6um6eW9fAedOL+XNSaG+zO9CnVMSPOxChMtdJuzfEy5uacFiMR7WZhsEgUZ7j5PUtwqGdm2bVOvOAKNH896o6tjZ9OjrRzKnIpqbTx4Uzi3lxQyO/e2sXE+PutW5/mCdX1XLHOYlxkD8cxR0QTpS+pTst7iC3PL2RS2eX8vvLZrCtyU1bb4hV1V2U5zg0sWf26CzW1Xbz8LJqNWGNn184lbP+9BEvbGjEZTNx06ljeXNbC6dOyOfRT2q4/l/r2NbkxmyU2NnsIRyT+dzsUr6xaKzm5AL4wvzRfO8/m3jgwypcNhMPXDObyXGXxOPXzaWxJxAXhILxtu1w+383s6nBzZdPqOAn508ero96WLCZjYzOcfDh7nbG5Dl5/Lq5/OqNnayKu6l2NPfSG4yyYEwOJ43L5VtPbeCReGeq4oO0Ys9yWnj6+gUpt100q0RzggyGLIeF2k4/Bek27jhnEj97eRtTSvp3TP/k/Mn9fv7lOU7CUZlN9cKd3uwO8u1nNxKVlRSnDMBjy2v4y/t7GVuQxp5WL5ccdwjn8qPMuAIX78abtuQ4LQfNRlIF0RVVnaAoTC/NYF+bj9e2NPG3pXu1LLDKPCcrqjqZV5G9X7bQkXLW1ELOmnoI144BmF6awbs72qjMdVLV4WN9vFtxUcbIn8dmOS1k2M38a0UtPf4IlxxXwvPrG9ne3MuU4v737caeAC9tbORLC8pHtOtwKBi53q5PIU+truOqh1Zx8d+W89a2FlZVdxKMyHy0u4MfvbAFi9HAudOK+OT20zh3ehG1nX6issK8ipxjvekji6AbalckHCLZh1BKNeZ0Ie6UHWAiPP4skd2jTvYzShKZG3kTYd71IrTz1VvFRPiqZ8XrqS1+h1IQMprFBMSaIVwkfVeqNzwO//2KmEzVLhciWcsWMfkZBGajgdduXsiGM/bikL1I7TtFyct7d8ITl4rynqGkZbP43mzpohuOtx3+82VRmnXBn8VEc8criccHusXjfe2w5uHU11InJ3JUOD2ufkHkgxx3DVzyoPhusirguzvhGytE1hQk3DrppeK7/fJr4vutOAXuaICfdMINH4mJ880bBi4xueDPIqdkMNgyhJAz64sJMQhg3GK4aXVCDBqI5NB0V6EoJVz3T/jPl6D2YxTJSLnUwrL1W7lCfo0dlAMQe/oLyAr8M3oGrn2vCnHtP1+O5430Q08dbHgikbG0/K8ihPtzD0HlqeI2szMhkslRIbAc6epJwWS4ox5+1Axn/BwlvZQ/Ry/mPPdt7JFLuIv7IOwj32VjXkUO3f4I4wvSWL6vk1c2NbF8XwfeUJQHPtxHKBrj0QtyyAvXkR+swoiMtOMV4cIB2PuO+BnoFt/L9MsT2+HIFo6Xtu2JUiP1sc48uOFDIWJWnJzoZjXlYuE4zB0vPo+scpFh84X/iOecfJv4ftWA4q3Pif19xhVCvJz1RREYHO2T/9S8mfERUSZUJ+cxuWcpFXavFijO9pcBiXVz/0QUE5c1/RY7QbFdky5IvE7mKHE+U2Rx3jrS85PaQTEahGmfF3lX+9478HMOxt53xXns3Z/Bc9eKIO6WzQOLsYMlLT+1ZCzjEFYoJ18g9o+pnxOuwxeuF+esh88QYmJXP22BP/6jeB+739z/vmhYiGgr7xOf4Wk/Ss2fOhBqufPsL4tz4UvfEMf+5v+I14XDcwgVTBGCZMYoOO8eHF99lZ3myaQ1r4CmDWw3TuSGyHfoVlwo+ZNFZp+iiA6D6jmiH9SMj2fXNiArYqKjdkCs7vBx0d8+4QsPrdK6SO1u9bC2touTx+exZIo4T5dm2Y/6qqw68fr24vGs/fHilIm1WvKRLBKNZC6dXcozNyxgfmUO2+5cQkG6lZ0tHiYVpTO5KJ0tfVosN/UksjL6TmBXVAmH41vbWnD7I3zpkdVc+9gaFAX+cc3xfPPUMRSkW/nrVbNYPCmfB5dV8fz6BuaUZzG+wKV1eFw4LperF5Tz5Ffnc/L4PH56/mSW7+1AkiQunV1GOF7WdfbU/ffhC2YUs/vus9n7y3PY8JMzOXVCarvxApcVgwSb6nto84SoyHWyvq6HNKuJGxdV7vd6nwYWVOYwsdDF09cvID/dRkWukyZ3MMVVNb8yh3OnFTG+II11td1YTQYt4H44yXZacFiMOK0mTp2Qz4ffP3W/hhYHY0ZZYpJ9yngRYl3V7iM3TTQ1cQcitHmCbGlwa+/3oWVVeEPRo95d7FCYEN82i8nAxbNKaOgOaCWL/aGWTG5p6GFbUy/TSzNZMCaHrY29FGbYWPq9ReS5rPzfi1up7wqwYMzInRNOi+dVqa3p18UdQmpH3JFOea6TTl8Yl83EHWdPQpLg3e2JseDeNi972xJZZn9bupc/vbNHE9P/l9EFoSGixx/mN2/uZOG4XLLTLNzwr3Wo5bvffmYju1s9PHDNbC3wTT2hmAwSs0frJWMpvPY9ePRssRILQqQZLMUz4fa6/jMYVBb9AK59PVHukV4qVtlBrBqXx4WGsYvhqmeEE0SSxMTOmX9o2zMYPv84XBefYCQHYkdD8NEfxP83PAGPngMvfh0eOQte76c0ZyC2vyS60+TGywL2vR8XZRSo/fjAz+1t7l98kmUxAd/5mnBEROLum20viPK6RT8UGTTPXStKj658WnyeOWOFuNW0AXa+npjE27NFKY4si9dq3yUCcssXilXz0/9PuIduWivyoKxpcPWLcOPHiRVztbPQqHhXuYz4ClN6EXx9uRCUDiVL5VC59vUDdwg6ENa0hKDlzBf7MYrYj41WpJlXYpZifLH9DzgIYbn0IXqldIyxEH9TLuMe8/XIdzQKx9zOV4VbpC+KIrp9vfRNeGQJfPAbWPE3kWU09XOpn19y56cjKenpD7MN6dtbeS/r8+TmFVD6hb+SHuvWBMHLji/FbjbywBePY3FGA/e9v4urHlzF9/+ziX+trOWSGQWcuvp6nnT9lQmGePD1thdSj53SOUJkXfhdUTKlYjAKEUKRxXGh4u9KFQaveSlRcrroB/DVdxP7k5p5U7FQOGcmnCX2w0seir9Wp3AhHncN3LoZJp4v3GmvfjsRHr/pGfjHKaS9cgMAP4tdh0GOcFnn/ZRK8fKzaAByx1FcOZkP5BlMJ176NvokkUVlSRPOroxScWyd/H1AOrxMnmRyxgoRwWgR4cv2bPH5RsP9Z7ANRNgnzg11K4UwAmJ/QxJC7B0NsPhnR7atGaVCQPF3idD2Q3EIjZovrhVlSaWqX3lXiH6v3AIv3JD6Xna+LsQeEOewvmz4l3AaXfGU2B9mXAVIiXLZA9EW79Jy6g/hh0I0Zeer8PxXxeuC6EIoGYQIOVjMNuFUqzxFuylSfgqjYrXIzVtYHR7Nbtd8ZoXup2XsFcIB62kh+vpttL/xK+q79i8tgv1DjNs9IZbtEaVYtz23CX8ohqKI4OAsh5k9bV6CEZmF43I5N94FMDk/6Ghx6WwhGF57Uvl+95mMBqp/dc6nRhBKxhAPZgaYX5nNtNIMtjS6qe/y0+MXgmJLvFxsfEEa1e0+zYkCCQeDJxjlxifW0eENMybPyXGjMhlf4OL7Syay8o7TKcqwc+vi8fjCMWo6/VqJyOPXzWXrnUv465XHJW8W155YwdY7l7DtziVcOlscm1OK+y8vPBgmo4F8l423tgmH1+8unc62O5ew5keLB9/Jb4Rx90VTef3mhVquo9oV7K5Xt7N8XwcVuU4KM2wYDBK3LhblYsWZR0dIHVeQpjnODpcpxRnceYHIHzx3ehEuq4kTx+bwy7gzbE+rh5++tI3P3b+cTfGmJ6qDb2Zp5hH97eFkXLyc7bhRmUwodBGTFRq6A7j9Ed7Z3srWRjfRmMzSXW28va2FNdVdVOY5kRVxrH5pQTn3fH4GW+9cwlu3nkxFrpObTxvLnjYvBglOHj/M3ZuPAFUIKs1yUJJpp75LjP0LMz4dx2BFvIx2XkU2eS4rkwrTWRk/F66q6uSCv37M+X/5hBX7Oqnv8vPsmnqumFt2UFfe/wL/2/6no0imw8J/bljAqBwH3b4IVz24kjSbiWhMYXtzL1fPH52y4qHW4s4oy/yft6EdEu274qUQiliNzRkrSruGAzVwNr1YBARLBuEeMprgm6vA1Cf74YSbYd6NqQ6QoUCSRMmHI1cEik6/XEy8Vt4ncjEyRyUmu6pIVrNMhOcO1BJdZdsL8NxXRFnLVc/An6bDh79OlDjVfCzKOPrS2yxcPk9fKSZIN8aFo1hUiDnr/5mYqIDo3DPpAlj9oOh6NPkC0TWrZpn4TMfE3SdTLhZlXv9YJH43xS8ip/xAPH7f+yIovH6VuH3cmcKhpWaiJA+EJCn1d1Xgm3ge1HySWjbYX3nHSCNnjCiRdOYKF8jK+4QTqmimENU2PMFi4wbeNp7MmVPnEN6xhJYdy3gitJAHrz0eo9UpwtCX/lIIbZMvSH393W+K727Jr8QE/YNfin1+0e3ifleRyNEZf5bY9515opSvZBC5KYeKJPHGLQuxGA3CGj3mNPjkT3D8dVw8q4SzpuTjeOu7PBR6nOdjJ/Fv6TTat+1kukHiRxnrYGct2hpzVrkozQJxrPTUCbfTNS/130K9YLIQdbe9CHO/JoSyQPfBOwaqYkNy+Vzy/meInz92vynyx9T7xy2GU24Xx92oBeKYffEbYt+PBohhIDRqIVJhI6M3xPNe0gpEiVDxLMYXuHhSWsAS1tLhGEuuM756OOUiaFyfOB+d9mPhVjqM0tIUTBZRUpk5SjiqJl8gBCyTTRzzX30PSuP7RG+TEDD76zr30e/EORyEqASAIgSt7CFa0S+bJ15z2wvxPLjDKDPIjQdpLvohlM2BWzYLAX3ZH0QplyVNlJw2rRdi96gFImMpEs+vcxWJ73nF34QQOeHsREe+0SeI61nFyeJzLT5OCDvOPPGZ+btEedmah8WihCP+OZ14iyizfPRs8d7mfEU4hAb6rA/E1S+gdScDjr/4VgJ/eBC77GNjtILLji/j3vf2sNKdxcVAw0ePUupp5Fed5/P2vcv453VzmD06m5issLmhB4vJwO5WkYfoCUY5f3oxz6yt55pHVuO0GPGFY9x14RSeXFXHzhYPJ43L45VNTWTYzSyakI+sKBSkW5lYdPQdALecPo4bTxmTkqORzKc5R+LiWSU8tryGUyfk0+IO8sTKOs7444dkOyz86nPT2dEsgmBPm1jA/R/u4/J/rORfX5nLwnF5rKzq4tQJeWxpdLOiqpNFE/J49MtzUjqGqZ/N1JIM1vxoMeGorLUcNxikAbvDqZ/1+AIXDovxiHJGKvOcLI+LV5OL03FYPt1jZ0mSUi4hp07I5+uLxvD3D4Q7+sq5iYWZs6YUMq0k46hNvH+wZCKyohz8gQfhSyeU8/njy7BbjKz+0WJsZoPWZXBjfQ/v72wjHA8jv3LuKJ5aXcdPz5/MtNIDNPQ4xozJS8NlM3HaxHxN3Fyxr5MHPtpHbacfk0HiirllPLEy0azg7tOn8uMXt/L548u07zD5mPni/NGcMbkQi8lwVEtpD5UZZZnYzAbG5qexeFI+/1xRi91sxGoaXDbVsaY8/n3NrxTjqAVjcvjXylqW7mrjG0+spyTLjgR889/rOXFsLgaDxDcWffoWCQ6HYTmbSpJ0FnAvYAQeUhTl18Pxd0YaqsWxMMPIm7eeTDgm89f397Kv3cs3Tk0texrrimI3GzQbJSBWYLuqhOPBnjnwHwr2ionOoQ4M+6IowsXRdyLk6xCr3DljxeQlGhKiwHC5K0LeROvwl74p/o6rUDhEplxy5OUqA5FdKSY5maOE+PHd3QmXRF8xCMSEzzBMF2NJEiHMO14R+0DDanH7GXeJbXzj+2Li4m4UAdjeFlE6VjxTPE5RROaOnFQq1LBGrHSXzRXlLVaXmJjsek1M0AqmQNUHQvxJT7JwB3rg7wtEiZzaxr19t+i69skfRXYPwIm3wtRLRJj3modFiVN6MVzwF/H9ZY8RZV9TLkm89tTPiYnWhHPF6v6q+8VEaOZV8M5PhAClBvGC2AcH2zI7uxKQhHtiyiWpOUWfBgqnxbt9GYWw8J2d4IqLXPFyxW7SWTnph5wJWC76CznnBFmKU2tpi8kinA/JbhkQ+8fSXwjn29yvCWGzfYc41lQ3ndkOt2xMTN5zx4vbjiQ75gCkTMoW/RAeXgxrHkQ64WYcb3wbNj6Br2gBlzR/zCXGJCfbKhKCCcD598LK+0Wp2Gk/Ea6KUfMS5ytVQFEFQxD7xwe/Em22zQ6RH6ZOxgcid5xwqx3IgXPpI+IxfUWZRbfDnrdFqVTQLRwbJ30HHr8AKaucf3x5ATBTZGy560XA/Iq/QvEs0qwmvv+tW4j941GypyeFO5/zh0TumcqRikEqV7+YEJoWfEvkP6kC8Hs/F/lgYR/cf5IQe8+4SxyrRhPEIkIs2fpfUUbZsUcInZMvEq6saZcOzTaCOLcZLSK3Bw7NIaSSNVo4vdQFAotDiDHL/iBEw+ZN4lx78QPi2OquhccvgH9fLoTXk74txLmufULISb5eTf0cvPYdeDT+vZ38ffjkz2L/XPg9UX7maxdutqv7OImcuWI//fA34GkVQtKh5Aep9FnAMDmzMJzybeSlv2SjPJZvTS1kW1Mv/9jWycWAY819hCUTn//ijWx4vZbrH1/H+6/h7wAAJJtJREFUR7edyr9X1fGL10WJo8Vk4LQJ+dx98VSyHBZWVXciKxCNyWQ6LHx+Thm+cIx73t7NWVMKeWVTE7edNQGjQcKIxNu3njLocNuhRJKkAcWgTzvTSzNZ9cPTyXfZtLbQwYiMJxTlS4+s1h536+JxXDizmK88toZ73tnN7lYvdV1+vrqwgt9eOoPW3iCVeU4kScJi6n/sdTgTVpfNzEe3nUq24/DPUb+8eBqLfv8B4wvSPvViUH9IksRtSyZglCT+unQviyYk5gcGg8TT188ftuFwXwwGCQND88fUY139WZJpx2Ex8uCyKkJRmRmlGexu9fLjcyfxg7MmaC3SRyo2s5GPvn8q6XYz3XEH3g9f2ILLauLeK2Zy23ObeWJlHSeOzeGOsydhNQkB5aSxuQO+N0mSPhUum2ynheW3n06m3UxJlp1/rqglEIkd680aNJOK0pEkWDhOHFsLKnN4+ONqrntsDePy03jyq/Np6PZz8X3LeWVTE18+ofxT8b0MBZIyBApwygtKkhHYDZwBNABrgCsVRdk+0HOOP/54Ze3atUO6HccEbxsoCkoshhIKIYfDhDDQLRspNESIuT1IRiNS82qkd35Iz/jzsJ34HRw52RDoRXnuqxi6t4MlA+WSR1AyKzFYLUhmMaBTFAWpay+xx69CyRyLdOl9YLSiGA0EpRjmsIwxFMGYkY5kMhGMhQj5vbgcmRiM4kTsjwYwSkYMKPje/AEZu9+Fix6A0rkosgx1q1Be/CbRaBhj2Xw47acY37wZKepFvvxJFCkd2edDslox2KxIViuS0YgcDoOiIFksEIshB0MokQjEYmK7JQkMBiSzCaPLhRKLEe3uQQq0I714PXhbURRAsiJd9g8MvXuJvn438qXPYhx7PMbMzJTVOyUSEa9nNKLIMnJvL5LdLv4+iNKM5L9tNIrHx19DURQUby+xpj0ozkLMhYVIptQBhqIoRNvbIRIBkwklEACTCclkQjIaxWsqCgaXCyUUIubuxZSVicHp1J6vbUf8H7KMweHAJwdxmp0o4TChvXuRTGYsWSakpy9F6qpCOf3nKCULMFTOQW7YivTgiUhn/VKUv8gxlD8fB4vuQJl5NVGPG+XVO2HPu1jSo9qAQVEQK9RXPiNKkiCxIp1VLsSnd+NlMWf/DmXc2UQamzFWPQ8rH8BgEueGaNAIs7+M6dQbkB5ZTCRtGrHp12E78QIirW1InkYM/z4XKbuc6DkPopgzMOXlYvjgpyjr/43hB7tRzE6UYBA5HMYY66bT7iQrCoa/zkEefSrSpQ8gPf8VolveQTnr95jblsHWZ4le+Q7GUVNpbd5HXnohhowMDJb9L6hKOEzM68UY7RBOm/hnryiKOObi322st5eY243BbkeyWjG6XMh+P9G2NiS7HYMzDdnvE/ebTOI4DoWRLGZtH1RkWbyGxUK0uwdLaQkxrw8l4BffczSGMSMdoyt1BVzbF9V9q60NYjEMaWkYnE6kaAAl2AuOPJRIBCUaRYlEiAUCWIqKUNr20BywkWa2YJcUlEgYY0YGRpeLaEe8zEiShEtr2T0Ez32UUJoLV34xls6NxJ79BrGFP8cw4xIUBQxWC4qsYMrJJlxfj+z1YS4sAJOJsM+D1QpGZxqywUEs4MeYmYlBMiS2v6UFQ1oaMXcvBocdY2YmsU6xemvMzkYJBsW5S5KQg0GUcBhjejqS2Zz4DG02DHY7SjiM/MQVGFtWQ/lJ9O56E+f876DM+QYNW1ZQkmWgMwYurwez2YSx4jh44GSIeJG+vRVFshOt24FscGG2hzCOPR7Z6yVcVQVGEyanAclqQzGnYXA4kQKdhH51Asb0DAyhJoxmGenCP6NMvwqMRiSDATkUQvb7kcwWDBazKLDuqUe2FWJwOvZzEygxMSiS4udaJRxG9ovPTYnFiG14CeMr1yKNPR2ueBI5FCP2pwWQMwH5zN8gmUwYM9OR/J0oLZvhuevgmlegeCZKNIox2oVsSifmC2NwCpFU9vkwZmSALIufcWS/H0WWxb5vsSAZjUTb21HCYUzFxcQ8HuSAH1NWNgaLBTkcRna7MWZlif1eUZC9XuRAAIPFgvTm92Hbf5GOuxJpU9zFlFEmrnlxUUouPB7O/QPSqzdBZzVKyIvhkvtQAm7kV3+IfOlTmIrLCfssGFwu7e9KZjPGtDRkv59ISwvGtDRMxcUQjYpjOj2d4PbtSFYrltGjE+d4RYFoFOnJi1FqPhbnvZvWpjhKlUhE+z6VSISYR5Q6GbOy+nWDKIqCEt8m6cFTkbubUXrbMSz6FtKZPxePkWV48/9g1V+Jhh0YrWCYeTFsfobYDeuJ+mUMTicGhxPF7yW89k3slcXwwV0o9ZuRLCYkYuJ1nIUoZ/8W8qehOPJQolHtegESJjqR/j5fZDm1bhXi+VVP77fNRCLinKGAMc1JzOvF7e8magSD3Y7JYMJusmONL3Qo0SiRtp2sC9uYXVjM7r2tXPivzWyyfg1bJEx49OnYr3uKDS0+LrlvOd9YNIan19QzLj+Nlt4gtZ1+bj5tLN+Jt+tt94RwWo1EZYVwVCY3zUo0JlPfHaA8x8G+dh9j+7SsVseeXcEuzEYz6Zah6aoTk2PU9NaQZcvCF/aRZcsizTL0HYuSz+cjjUhMZtbP3+GEMTncffFUtjS4+co/xRi75teihPbfq+r44QuiicXpE/O574vHDdkqv6zI2rViqPGGooSj8oh2URwpW9q34JIqKM89dp22QHyPoVgIu2noy2U+9/flrKvtpjjDxvvfW0Rrb5DROcNY0j/ERGIRTAYxrqy4Q5To33XhFK5eUM5dr27n4Y+reeEbJzBrBHeRDsVCWI0Dd8GMyTGMB6hCkGWFyh++TkmmnU9uP+2Af2uknC8VRaG6w0dlnji2eoMRjr/7XSpznTzx1XnkponP49pHV7N8XyfLbjtVi3r5X0CSpHWKovRr+x8OQWgB8DNFUZbEf78DQFGUXw30nP8VQeiNa8ZhqTGiSBAxQdQEBeEY2SGFHS4TARvYYwoz/WH2mO2EZJmwCcb3RrGjIBkl/LYC6iNdxAwKigTmCBhlUJKOozKDhNsYpRMjKGCQwRyDkAXCJrAHwaCIfzETSDEojUaxKAo+g4FWU/x5gA2JwmiYKosZWYJxkTANdjtukwklFkKRwBGQyJYVHHKUvU4TNpPMuGCEnQYrhihICsTES2KUoTAWw2WIIRkUiFf1KApssloxRsEWAtkAPrt4vEGWUMxWFAPIRiOmmIQlLOOzQ9BhweGLYQ/EMABPXj+WKWs7mL6xl0JrLhISrf4WAg4TpoiMKaJgNpgoyyxHMhiEq0eWxeA7FsMb9tIWaCPDmklWZiGbg/uQDRIZPRGk+LEgJR0SXpeJqMmAUVYIWwwYYgoGBZySjYkZ49jVvRtzIEzYBB6rgtMbxRwRDpcsWxZptgxqPLXIRgnZIKEAlrCMURHCgmyUaC+wYYgpZHWHMUVkUBRiJgNWi4MyUx6S1YLic9MW9eKP+il2FhPsqqJTklEkCNjAZY4xNr2CXc2i/los7EgoBiOj0kfhi/joCnaRa8/FbDDT5G0CFBQ5hkGOISkxYhJMsgSpjVrospgxh2UkScLrlECOUdYrkxOR2VSYSchuprApSE+2ECttgSiSAn6nmahZwtUbxRKKETNAusFJgauIukgrvbIfWyCGRTFSnFFGOOqjQ/GRZcnCIsM+OggZFFy9EYyRGN50C+ZwjJjLwdTMidDrpcvTRncotXuUbJAwp2cwWsqlpreGqBJDkUCRJAyygiEe6BW1mgil2zAEIxhDEaxBmbDVgCfdjDkiM9MxnoZYB57eDiQFoiaJqMlAsTmH3KgNKb5iZshIR4lEMGVlU3LPH+j+97/xLl+OZBRiYbSrCzngR0JCQUFCora3FhmFiowKugJdNNsCjM2dSMDdSVdnA5ICCgoxo4RslIgZJGSTRMxkZLpchGSx0Kr00hLppDJ3PJjN1NdtwRqUMRcUUJk5hnUtayEWQQn7CDgUrFGY1hPBHJIwOGx4xs1hd/sOcW6Jis/E6YnSlWMhZDOQ3hNBAmJmIzZvGHMUsl35ONOyqG/aiWSQAAkzRkaPmUV1yw56zBGmuMbhaWuk0SBWpu2BGGGzhDGmIClgtDsYmzeRmMeDEo3QHuig3RRgjLWESMBHe7CDiFkixxcgOxqh2uHAnJZJRcFEdvbuJeLpxRaM0ZNlQVLAFoxRFIlij4SotjuJGSCQbkFyOJlLOdHOTowuF1WZIRrd9aT1RqhwlGKy2Gls2Y05JhHLMzKxrZk9WFHCEgajDcVkodRZjFEyYDBbMDidKJEIoaCX+q4a7GY7xbkVNHfUEIgG4m62uMgnSRQ4C3AaxeA5Qox9wQYyI2YKXcXsjjVh9gSQDEYMGIhZjHisMaIGiajFgDGm4PRExTkTA7IS1UpCDUYTzogBnxTB5zBgC8kUO4pwZOWxr26TOE8HxYArpsSImA1gMmKUJcqsBVgkE167RJW/ntKgg6zcErZ6dpMVMlNqL0Qym6mR25G7esSqsCQRsELELLbLHIM0g5ViSwE13XuRlBijQgHhurKm0xXooCvmRTaAOSo+CxcyeVkTkQwm9oTqCFoMuLwxOrOMYvEiqqCYjZgUA9ZAlJBRoTvLjMMfI8MdJWaUCFkNOHxRWoptSDLMCOcTCvnpDnSSZc8hw5FNrLueulAXNpONosyEE7fN34Yn5kOSFYySEdko4coqoMCRT6S7mxZfCzE5igKMSh9Fs6+FQDRA1ATmmIQJCSXmJWgEk2zDaU6jwFlATW8tUUnGGI3hSTcxpbMHYwwUazrenBJ2mTooM+VRJGWCyYS5tIRdq94kFosQkUJYZBMmo5VCUxpBs5n2iJuctHycNhe73PuQTEYikoxBhjRPBJMsUxr0MeqkLj4yHgdrOpEkCYNkJCbHcJqdFGQUs8tThaLIpEUMeE0xJLMFcwxM4RgKClm2LFwWl7avVrursUgmwsQI2A1UxIroCbbSbQ3hsLlYfVoJG0bLeDtmsWNvJfayR6nMcxCKyjR2BynLtg86ZDbHnsPfF/+dO1fcydaOrQDcfeLd1Hvq+c/u/xCIBghEA9r5FSDDmsGDZz7IL1b+gk3tmzAZTNqkQlEULh53MYvKFnHr0luZkjOFH83/ET9c9kN2de+iMqOSnlAPTrOTdn87N864kYWl/Tv7vvPBd2jwNGi/GyQDt8+9nYgc4eEtItfMaDBikkz0hnuJKTHCsTAROYJBMmCUjNx94t00+5r5y4a/oKCQbknHExbi4+j00fzulN/x8xU/p9pdTUSOEI6FE9cuRSbTlokn7KG/8bjZYOavp/+Vjxs/5vk9z+OL9N/RSEHBZrThMDv4y2l/4S9rH2d5y1sY492KorJCTFawmgwcV3Actx3/A97fKcJU13kfp8q9l7tOvIsNbRt4tepVQtEQUSWKrIiFFVmRtfd23+L7tO8l3ZrO7XNuxx12889t/6TJ24RBMjCvaB7fmf0dbl56MxUZFQQiATqDnXQGOkmzpFHoKKQr2EWTr0n7XmcXzOa2Obdx03s3oaBQ5CzSvpt0SzptgTb8ET9fnPxF5hTM4a6Vd9Ed7CbDmoHD7OCeRfdon8fDWx7mrZq3KHQWkm3LZmHJQqbmTuVHn/yI3lCv9jijZCSmxHhoyUOkW9J5Yc8LvLzvZfLseSwevZiZ+TP5zgffISJHUBTx/h9a8hC/Wf0bfBEfZoOZVn8rFqMFm9FGV6gLi8FCRUYFRsnIzcfdzD82/4OdXTv5zcm/4dldz7K9czudgU7CchiX2cWpo07lorEXie9RUfj16l+zo2sHjyx5RBMd3q19l6d3Pk0wFsRlcWE2mAlEA4RjYcpcZdx90t3cvVLshz3BHqJKVJvsR+Uo0WT3OOKc9/tTfs+PP/4xzb5m7ll0D6/se4WX970MgMlgIhgLYjPakCSJ8vRyuoPd5DnykJD40fwf8eT2J3m//n0yrBmYDWZ6Qj0oisKS8iVcOv5Srn79aqbkTuHH83/MnSvupNHTyC8X/pJlDct4o+p9egJBzKYoRa4cZEUm25bNj+f/mBvfvZFwLExMFsfa7075HWta1vDUzqcAsBgtZNmycJgclLnKuGnWTdzy/i04zA7uOvEuHtv2GB83fiwEfpTETxRQ0PZl9b4nznkCi9HCS3tf4t3ad+kOdWOQDOTac/GGvaRZ0si0ZlLbW0tEjjApexIrmlfgMruYkD2Bf6/diWT08YvF13HJ+Avp8PVw3VtfI81qRkKKn68N3Hf6fZo4/V7tezyw+QEkSRLXJ0VGjrvjbSYbdpMdu8mOzWTDHXJz08yb6Ah0cN+m+7T91mww47K4yLZl8/MTf86PP/4xu7tFE4PfnPwbAtEA79W9x+6u3do+ql4zZGRicoxHz3qUBzc/yLt17/KLk35BMBrkpb0vsbNrJ2eWn8kl4y7hK299BbPRjMVgwWK0YDaYMRlMdAY68YVDOMx20q1CzPNGvMzKn8W3Z3+bG9+9kUAkgD/q1/btP576Rza0beDpnU8jSRJROcpZ5WfxhUlf4EtvfIk8Rx5/Pu3PAPx1w1/Z0rGF3pA47xolI9dMuYapuVO5/aPbmZE/A5Nk4jvHf2e/8+ELe17gpX0vcd3U6zi5NFHmH5Nj3PHxHTR6GqnMrOTOE+7kL2seQzb4aPDWUO+px2K0kG3NJxqTCONmas5Ubp19a7/n3U8bR1sQuhQ4S1GUr8Z/vxqYpyjKTX0edz1wPcCoUaNm19bW7vdanzrWPy4mY6qwMJAaarYL+3ztJyLvQmXUfFHG420TobByPzY8gxEmnCPKCg4l4LM/csYQK5yBcddriQ44JqvIgrGlQ/1qYZUffSKKyYZU9X7cdiIYEsVXkkQXo6Q2v4qiEFNi2oVQJRwLI0mSdiGUkLCZDk+5HSq1WlZkantrcVlc5NhyDuk1w7EwJoNp0CtpETkCihicGlq3Qf3qxPvIrhA5LH22TX1tf8SP2WDG3E/+USjUi7zjFewREToZBUyTLgBPE0r84iVVfSAenDlKlGQdBru6djEua1zK+93RuYOIHGFyzmRCsRDesJcCZ8EBXiXBQY+zflD/ht1kx2HuvxRNTi5ZS2IoVzwbvY24LC7SLenEZDFhM0gGbfCgcrAVGujneFEU2P7i/tk55QtF2V/S89T32t/fUF9XQtLuD8VC2jaaDWJf8kf8WI1WYkpMu62/4zeZmBwT7zc+uYP49+hth9YtMOY0eoI9ZNoyicrR/l+rpx7F3wVF0wEIy+H9VrtCsRBGyag9X1EUbfJpVyQcjeuoy6sko2E9VCwiZjSSYcnY7/OIyTFkZO397e7eTWlaKTaTTXyOyEJklwwDfl+KohBVooRjYUKxEOFYmCxb1n7brCgKYTmMxWDRPp/ecC8tvhZK0kpIs6QRlcVEzWwwa/tLMBokEA2Qac0EIBANEJEjOM3OAb+LvvuXrMiEY2HCchinyYnRYCQiRwhFQykuCyUWE9eDipMT5bWN60WJaekcUX7V2yiyywBfxIfFYMEb8ZJhzcAgGbT3GYyK806aOU3blqgcRULS9pG+hGNhFBSsRitK2E+g5iPCZfPITCp/ThYPekI9AGRahctPVmR2de0iz5FHljUr5e8aJSOBaIDuUDfh7mrSYzI55Sdr59OIHMEkmRLbtW+pKPMduxg5fh3rexzHNwj2vE1k1Hy6YkEybcJxF4wGMRvM2Ew2YnJMW5FXt73Z10xR0zak3nqUcUuQMkoIx8J4wh7SLekp5/SYHKPR20hJWslBzxsDoSgKm9o34bK4GJM5htbeIO/uaKXAZWPx5AJkWWHprjZOHp+H2Xj0stkicrzLiwJIQihRj6mYHMNmstEd7CbLtv+K/IGu9+p3rt4fk2NIkkREjmjiizqhTrekYzKYMBqM2rlAnWiHY2HMRjMGDHgjXtIt6dp3KEkSnrAHg2TAZDBpx7b6vrxh8fj+vjN1+9whN0aDURP1+sMf8ROMBcmyZuGP+rGb7Ptdt9TzfvLf8kV8hGNh0i3peCNeAtEANqNNG5+oE9rkc7963u4N92Iz2QhGg/giPvId+Snnm45ABzu7duI0O8m15ZJtz8YX8dHsaybDksHo9NH7fTfesBdfxEeTr4nStFKMBvH+s23ZZFiFEzISi9Ad6ibLmoU77MZpdqY4WdTtq/fU4wl7KHWV4jA78IQ9ZNuyB/wsANwhN53BTrKsWWRYM7RrUPJnGZWj7OjcgYJCviOfUCxEKBYi25aNP+KnpreG8VnjKXQWsqV9C5m2TErSSsT78jaRac3Utsdlce33vfoj/pQxSndQiBQ2kw1P2EMkFsFusmMxWrTb3SE3oViITGsmJoOJUNy9aTKYUs9ZJM79ETmCP+LHZXGlOoAVsd8ZJAP+iJ86Tx1Z1iw6gh1ISEzMnkhEjiArshCg5CiZtkxAXHdz7blE5SjNvmbKXGXaedhlcaGg0OprxSAZcJqdmpBU5irDaBBCtz/qxygZsRgt+12/gtEgPaEebbvzHHmEY2HqPfWMyRxDMBrUxi6SJPX/s89t6mfeG+4l3ZKOrMh0B7tJs6ThCXvoCnZRkVGBQTKwrnUdcwvnEoqFaPA00OU1kGnJYXJRjnZsKCiamKr+32q07revy4qsbasqegdjQW2cEowGSbekk2PPSfl+YkpMG89ajVYybZkp1zwFhdreWup666jMrKQ0rZSwHMYoGfFH/cTkmHau9Ia9OMwODJKBJm8T7YF2xmaOxWl2atsYlaOaEK7+zLZlYzFatLEHQJolTRMR3SE3JoMJh8mhXXdBjJMisYg2porKUcwG837jRl/ERygW0gTQ5HFYd7CbHV07GJs5lnxHakdCEOfVSCzS77yn2l1NkbOIek8947LG0e5vZ2P7Rka5RjE+azyhWIhWfyvhWFh7bfW882lnRApCyfyvOIR0dHR0dHR0dHR0dHR0dHR0RgoHEoSGY3mnEShL+r00fpuOjo6Ojo6Ojo6Ojo6Ojo6OzghgOAShNcA4SZIqJEmyAFcALw/D39HR0dHR0dHR0dHR0dHR0dHROQyGvG+joihRSZJuAt5CtJ1/RFGUbUP9d3R0dHR0dHR0dHR0dHR0dHR0Do8hF4QAFEV5HXh9OF5bR0dHR0dHR0dHR0dHR0dHR+fIOHotInR0dHR0dHR0dHR0dHR0dHR0RgS6IKSjo6Ojo6Ojo6Ojo6Ojo6PzGUMXhHR0dHR0dHR0dHR0dHR0dHQ+Y+iCkI6Ojo6Ojo6Ojo6Ojo6Ojs5nDF0Q0tHR0dHR0dHR0dHR0dHR0fmMoQtCOjo6Ojo6Ojo6Ojo6Ojo6Op8xdEFIR0dHR0dHR0dHR0dHR0dH5zOGLgjp6Ojo6Ojo6Ojo6Ojo6OjofMbQBSEdHR0dHR0dHR0dHR0dHR2dzxi6IKSjo6Ojo6Ojo6Ojo6Ojo6PzGUMXhHR0dHR0dHR0dHR0dHR0dHQ+Y+iCkI6Ojo6Ojo6Ojo6Ojo6Ojs5nDF0Q0tHR0dHR0dHR0dHR0dHR0fmMoQtCOjo6Ojo6Ojo6Ojo6Ojo6Op8xdEFIR0dHR0dHR0dHR0dHR0dH5zOGLgjp6Ojo6Ojo6Ojo6Ojo6OjofMbQBSEdHR0dHR0dHR0dHR0dHR2dzxi6IKSjo6Ojo6Ojo6Ojo6Ojo6PzGUMXhHR0dHR0dHR0dHR0dHR0dHQ+Y0iKohzrbUCSpHag9lhvxxCRC3Qc643Q0fkfRj/GdHSGF/0Y09EZXvRjTEdneNGPMR2dVEYripLX3x0jQhD6X0KSpLWKohx/rLdDR+d/Ff0Y09EZXvRjTEdneNGPMR2d4UU/xnR0Bo9eMqajo6Ojo6Ojo6Ojo6Ojo6PzGUMXhHR0dHR0dHR0dHR0dHR0dHQ+Y+iC0NDzj2O9ATo6/+Pox5iOzvCiH2M6OsOLfozp6Awv+jGmozNI9AwhHR0dHR0dHR0dHR0dHR0dnc8YukNIR0dHR0dHR0dHR0dHR0dH5zOGLgjp6Ojo6Ojo6Ojo6Ojo6OjofMb4TAlCkiTZJUn6UJIkoyRJMyVJWiFJ0jZJkjZLknR50uMqJElaJUnSXkmSnpEkyRK/3Rr/fW/8/vKk59wRv32XJElLBrEtT8Yfu1WSpEckSTIn3bdIkqSN8W37MH6bRZKkjyRJMg3ph6KjM4QM5zEWv3+UJEleSZK+N4htuSn+OookSblJt2dJkvRCfJtWS5I0NX67fozpjHiG6xiTJMksSdI/JUnaIknSDkmS7hjEtgx0jE2Mb1co+VjVjzEdHR0dHR0dnZHFZ0oQAq4DnlcUJQb4gWsURZkCnAX8SZKkzPjjfgP8UVGUsUA38JX47V8BuuO3/zH+OCRJmgxcAaivdZ8kScaDbMuTwERgGmAHvhp/rUzgPuCC+LZdBqAoShh4D7i8vxfT0RkhDMsxlsQ9wBuD3JZPgMVAbZ/bfwhsVBRlOnANcC/ox5jOp4bhOsYuA6yKokwDZgM39BVk+2GgY6wLuBn4ffKN+jGmo6Ojo6OjozOy+KwJQl8AXgJQFGW3oih74v9vAtqAPEmSJOA04Ln4c/4JXBT//4Xx34nff3r88RcCTyuKElIUpRrYC8w90IYoivK6EgdYDZTG77oKMdiviz+uLelpL8bfg47OSGW4jjEkSboIqAa2DWZDFEXZoChKTT93TQbejz9mJ1AuSVJB/L4X0Y8xnZHNcB1jCuCMu3fsQBjoPdCGDHSMKYrSpijKGiDSz9NeRD/GdHR0dHR0dHRGBJ8ZQShul6/sb/AqSdJcwALsA3KAHkVRovG7G4CS+P9LgHqA+P3u+OO12/t5zsG2ywxcDbwZv2k8kCVJ0geSJK2TJOmapIdvBeYM5nV1dI42w3mMSZKUBvwAuHMINnUTcEnSdo0mIcjqx5jOiGWYr2PPAT6gGagDfq8oStcwvA39GNPR0dHR0dHRGSF8ZgQhIBfo6XujJElFwL+AaxVFkY/2RiHKwz5SFGVZ/HcTwq5/LrAE+D9JksYDxEsEwpIkuY7BduroHIzhPMZ+hih/8R721iX4NZApSdJG4FvABiAG+jGmM+IZzmNsLuI4KAYqgO9KklR5mK81IPoxpqOjo6Ojo6MzcvgsBTsGAFvyDZIkpQOvAT9SFOX/27u/UL/nOI7jz5dtEcpqCUVNSDKRKaYR7qRcaGmUsJQLRc6NXYjdUJYL5U8uNtpyJZGlDLFMoqRmf8zWUnQWyVLaMpa9XXw/h2N2zs50fvv91vf5qFPf3+f7+X5/79/Fu/P7vb+f7/v7eRveR/djcW67eno+sLft2wtcAIy3ZfVntfkT4xMmHzOlJE8CZwMPThoeB/ZV1QHgQJLNwJXA7rb/VODgjD6xdGINMseuBZYlWQ3MBw4nOVhVLxxvkFX1K3B/iy90t6F9O2mKOaZRNcgcuxvYWFWHgJ+SfApcw79zY7aYY5IkSSOgNyuEquoXYE6S0+DvpfdvAeur6o1J8wrYBCxrQ/fS+jUAG9pr2v6P2vwNwPL29JYLgUvo+gKR5MMk/7l9LMkDdCuA7jriiu7bwNIkc5OcTvdDeGc7ZgHwc/vCLo2UQeZYVd1QVQuraiHwHPD0RDFoqhybSpL5E09comvmvrkVicwxjbQB/x/7nq7vEEnOAK4DvmmvjyvHpmOOSZIkjY7eFISa94GlbftO4EbgvnSPeN+S5Kq27zFgLMkeut4Ka9v4Wrp+JnuAMWAlQFXtAF4HvqbrBfRQVf2Z5BTgYronrhzpZeAc4LP23k+0c+1s59hKV1RaU1Xb2zE3010JlkbVQHJsKtPlWJKHk4zTrY7YmmRN23UZsD3JLuBW4JFJh5ljGnWDyrEXgTOT7AC+AF6tqq3/J8eSnNvGx4DHk4y3lUxgjkmSJI2MdBcG+yHJ1cCjVXXPCXq/RcCKqhqbpfO9Caysqt3HnCwNgTkmDZY5JkmSpNnSq4IQQJIVwLrW2PKk0W4NWF5V64cdizQdc0waLHNMkiRJs6F3BSFJkiRJkqS+61sPIUmSJEmSpN6zICRJkiRJktQzFoQkSZIkSZJ6xoKQJEnqtSQLkmxpfz8m2du29yd5adjxSZIkDYJNpSVJkpokq4D9VfXssGORJEkaJFcISZIkHUWSm5K807ZXJVmX5JMk3yW5I8nqJNuSbEwyr81bnOTjJF8meS/JecP9FJIkSUdnQUiSJGlmLgJuAW4HXgM2VdUVwG/Aba0o9DywrKoWA68ATw0rWEmSpOnMHXYAkiRJJ4l3q+pQkm3AHGBjG98GLAQuBRYBHyShzflhCHFKkiQdkwUhSZKkmfkdoKoOJzlU/zRiPEz3nSrAjqpaMqwAJUmSZspbxiRJkmbHLuDsJEsAksxLcvmQY5IkSToqC0KSJEmzoKr+AJYBzyT5CtgCXD/UoCRJkqbgY+clSZIkSZJ6xhVCkiRJkiRJPWNBSJIkSZIkqWcsCEmSJEmSJPWMBSFJkiRJkqSesSAkSZIkSZLUMxaEJEmSJEmSesaCkCRJkiRJUs/8BbgyB6aOhCkIAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 1440x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# load the provided data\n",
    "train_features = pd.read_csv('./data-processed/dengue_features_train.csv',\n",
    "                             index_col=[0,1,2])\n",
    "\n",
    "train_labels = pd.read_csv('./data-processed/dengue_labels_train.csv',\n",
    "                           index_col=[0,1,2])\n",
    "\n",
    "test_features= pd.read_csv('./data-processed/dengue_features_test.csv',index_col=[0,1,2])\n",
    "\n",
    "# Seperate data for San Juan\n",
    "sj_train_features = train_features.loc['sj']\n",
    "sj_train_labels = train_labels.loc['sj']\n",
    "sj_test_features = test_features.loc['sj']\n",
    "\n",
    "# Separate data for Iquitos\n",
    "iq_train_features = train_features.loc['iq']\n",
    "iq_train_labels = train_labels.loc['iq']\n",
    "iq_test_features = test_features.loc['iq']\n",
    "\n",
    "print('San Juan')\n",
    "print('features: ', sj_train_features.shape)\n",
    "print('labels  : ', sj_train_labels.shape)\n",
    "print('test: ', sj_test_features.shape)\n",
    "print('\\nIquitos')\n",
    "print('features: ', iq_train_features.shape)\n",
    "print('labels  : ', iq_train_labels.shape)\n",
    "print('test: ', iq_test_features.shape)\n",
    "\n",
    "sj_years = sj_train_features.index.get_level_values(0)\n",
    "print('\\n San Juan Unique index values')\n",
    "print(sj_years.unique())\n",
    "iq_years = iq_train_features.index.get_level_values(0)\n",
    "print('\\n Iquitos Unique index values')\n",
    "print(iq_years.unique())\n",
    "\n",
    "fig, ax = plt.subplots(figsize=(20,10))\n",
    "\n",
    "sj_train_labels.plot(ax=ax)\n",
    "print('\\n San Juan Unique index values')\n",
    "print(sj_train_labels.describe())\n",
    "\n",
    "\n",
    "iq_train_labels.plot(ax=ax)\n",
    "print('\\n Iquitos unique index values')\n",
    "print(iq_train_labels.describe())\n",
    "\n",
    "\n",
    "\n",
    "(sj_train_features\n",
    "     .ndvi_ne\n",
    "     .plot\n",
    "     .line(lw=0.8))\n",
    "(iq_train_features\n",
    "     .ndvi_ne\n",
    "     .plot\n",
    "     .line(lw=0.8))\n",
    "plt.title('Vegetation Index over Time')\n",
    "plt.xlabel('Time')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7dd613b4",
   "metadata": {},
   "source": [
    "Remove `week_start_date` string.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 184,
   "id": "ba52a048",
   "metadata": {},
   "outputs": [],
   "source": [
    "sj_train_features.drop('week_start_date', axis=1, inplace=True)\n",
    "iq_train_features.drop('week_start_date', axis=1, inplace=True)\n",
    "\n",
    "sj_test_features.drop('week_start_date', axis=1, inplace=True)\n",
    "iq_test_features.drop('week_start_date', axis=1, inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3ed71841",
   "metadata": {},
   "source": [
    "Checking for nulls and Nans."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 185,
   "id": "f3ea7bf8",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      " San Juan any nulls?\n",
      "ndvi_ne                                  True\n",
      "ndvi_nw                                  True\n",
      "ndvi_se                                  True\n",
      "ndvi_sw                                  True\n",
      "precipitation_amt_mm                     True\n",
      "reanalysis_air_temp_k                    True\n",
      "reanalysis_avg_temp_k                    True\n",
      "reanalysis_dew_point_temp_k              True\n",
      "reanalysis_max_air_temp_k                True\n",
      "reanalysis_min_air_temp_k                True\n",
      "reanalysis_precip_amt_kg_per_m2          True\n",
      "reanalysis_relative_humidity_percent     True\n",
      "reanalysis_sat_precip_amt_mm             True\n",
      "reanalysis_specific_humidity_g_per_kg    True\n",
      "reanalysis_tdtr_k                        True\n",
      "station_avg_temp_c                       True\n",
      "station_diur_temp_rng_c                  True\n",
      "station_max_temp_c                       True\n",
      "station_min_temp_c                       True\n",
      "station_precip_mm                        True\n",
      "dtype: bool\n",
      "ndvi_ne                                  True\n",
      "ndvi_nw                                  True\n",
      "ndvi_se                                  True\n",
      "ndvi_sw                                  True\n",
      "precipitation_amt_mm                     True\n",
      "reanalysis_air_temp_k                    True\n",
      "reanalysis_avg_temp_k                    True\n",
      "reanalysis_dew_point_temp_k              True\n",
      "reanalysis_max_air_temp_k                True\n",
      "reanalysis_min_air_temp_k                True\n",
      "reanalysis_precip_amt_kg_per_m2          True\n",
      "reanalysis_relative_humidity_percent     True\n",
      "reanalysis_sat_precip_amt_mm             True\n",
      "reanalysis_specific_humidity_g_per_kg    True\n",
      "reanalysis_tdtr_k                        True\n",
      "station_avg_temp_c                       True\n",
      "station_diur_temp_rng_c                  True\n",
      "station_max_temp_c                       True\n",
      "station_min_temp_c                       True\n",
      "station_precip_mm                        True\n",
      "dtype: bool\n",
      "total_cases    False\n",
      "dtype: bool\n",
      "\n",
      " Iquitos any nulls?\n",
      "ndvi_ne                                  True\n",
      "ndvi_nw                                  True\n",
      "ndvi_se                                  True\n",
      "ndvi_sw                                  True\n",
      "precipitation_amt_mm                     True\n",
      "reanalysis_air_temp_k                    True\n",
      "reanalysis_avg_temp_k                    True\n",
      "reanalysis_dew_point_temp_k              True\n",
      "reanalysis_max_air_temp_k                True\n",
      "reanalysis_min_air_temp_k                True\n",
      "reanalysis_precip_amt_kg_per_m2          True\n",
      "reanalysis_relative_humidity_percent     True\n",
      "reanalysis_sat_precip_amt_mm             True\n",
      "reanalysis_specific_humidity_g_per_kg    True\n",
      "reanalysis_tdtr_k                        True\n",
      "station_avg_temp_c                       True\n",
      "station_diur_temp_rng_c                  True\n",
      "station_max_temp_c                       True\n",
      "station_min_temp_c                       True\n",
      "station_precip_mm                        True\n",
      "dtype: bool\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "ndvi_ne                                  True\n",
       "ndvi_nw                                  True\n",
       "ndvi_se                                  True\n",
       "ndvi_sw                                  True\n",
       "precipitation_amt_mm                     True\n",
       "reanalysis_air_temp_k                    True\n",
       "reanalysis_avg_temp_k                    True\n",
       "reanalysis_dew_point_temp_k              True\n",
       "reanalysis_max_air_temp_k                True\n",
       "reanalysis_min_air_temp_k                True\n",
       "reanalysis_precip_amt_kg_per_m2          True\n",
       "reanalysis_relative_humidity_percent     True\n",
       "reanalysis_sat_precip_amt_mm             True\n",
       "reanalysis_specific_humidity_g_per_kg    True\n",
       "reanalysis_tdtr_k                        True\n",
       "station_avg_temp_c                       True\n",
       "station_diur_temp_rng_c                  True\n",
       "station_max_temp_c                       True\n",
       "station_min_temp_c                       True\n",
       "station_precip_mm                        True\n",
       "dtype: bool"
      ]
     },
     "execution_count": 185,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "print('\\n San Juan any nulls?')\n",
    "print(pd.isnull(sj_train_features).any())\n",
    "print(pd.isna(sj_train_features).any())\n",
    "print(pd.isnull(sj_train_labels).any())\n",
    "print('\\n Iquitos any nulls?')\n",
    "print(pd.isnull(sj_train_features).any())\n",
    "pd.isna(sj_test_features).any()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5cd64b5e",
   "metadata": {},
   "source": [
    "<b>Preprocessing:</b>\n",
    "\n",
    "meanRegressor is an EnsembleClass to get the mean prediction.  \n",
    "Kmeans is used for train data set clustering. \n",
    "Gradient Boosting Regressor is used for......"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 186,
   "id": "3bed3657",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.base import BaseEstimator, ClassifierMixin\n",
    "import copy\n",
    "\n",
    "class MeanRegressor(BaseEstimator,ClassifierMixin):\n",
    "    def __init__(self, \n",
    "                 kMeans_n_clusters,\n",
    "                 kMeans_n_init,\n",
    "                 kMeans_max_iter,\n",
    "                 gbr_learning_rate,\n",
    "                 gbr_n_estimators):\n",
    "        self.kMeans_n_clusters = kMeans_n_clusters\n",
    "        self.kMeans = KMeans(random_state=rng,\n",
    "                             n_clusters=kMeans_n_clusters,\n",
    "                             n_init=kMeans_n_init,\n",
    "                             max_iter=kMeans_max_iter)\n",
    "        self.gbr_learning_rate = gbr_learning_rate\n",
    "        self.gbr_n_estimators = gbr_n_estimators\n",
    "        return\n",
    "        \n",
    "    def fit(self, X, Y):\n",
    "        self.regressors = []\n",
    "        self.kMeans.fit(X)\n",
    "        clusters = self.kMeans.labels_\n",
    "        X = np.c_[X,clusters]\n",
    "        Y[\"cluster\"]= clusters\n",
    "        for clusterNo in range(self.kMeans_n_clusters):\n",
    "                cluster_x = X[np.where(X[:,-1] == clusterNo)]\n",
    "                cluster_y = Y[Y.cluster == clusterNo]\n",
    "                cluster_y.drop(\"cluster\",axis=\"columns\",inplace=True)\n",
    "                cluster_x = cluster_x[:,:-1]\n",
    "                gradientBoostingRegressor = \\\n",
    "                    GradientBoostingRegressor(random_state=rng,\n",
    "                                              max_features = 'auto',\n",
    "                                              learning_rate=self.gbr_learning_rate,\n",
    "                                              n_estimators=self.gbr_n_estimators,\n",
    "                                             )\n",
    "                gradientBoostingRegressor.fit(cluster_x,cluster_y)\n",
    "                self.regressors.append(copy.deepcopy(gradientBoostingRegressor))\n",
    "        return self\n",
    "    \n",
    "    def predict(self, X, **fit_params):\n",
    "        all_test_predictions = np.zeros(X.shape[0])\n",
    "        for regressor in self.regressors:\n",
    "            test_prediction = regressor.predict(X)\n",
    "            all_test_predictions = np.add(all_test_predictions,test_prediction)\n",
    "        average_test_prediction = all_test_predictions/self.kMeans_n_clusters\n",
    "        return average_test_prediction"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0fcf32ea",
   "metadata": {},
   "source": [
    "<b>Hyperparameters tuning for sj</b>\n",
    "\n",
    "The parameterScope is derived from trail and error, by looking at mae and tweaking the bounds and values accordingly. But the problem is that this is just obtaining a local minimums and there maybe be more lower minima(s) out there. I suppose in general the suppose lower minima(s) may have small intercept or even mutually exclusive set of hyperparameters to tune. I was suggested to go with a simipler set of hyper parameters to tune because of Occam's razor."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 187,
   "id": "417b8a57",
   "metadata": {},
   "outputs": [],
   "source": [
    "#gbr = gradientBoostingRegression \n",
    "parameterScope=[\n",
    "        {\"name\":\"knnImputer_n_clusters\" ,\"type\": \"range\", \"bounds\": [15,20]},\n",
    "\n",
    "        {\"name\":\"pca_n_components\" ,\"type\": \"range\", \"bounds\": [4,10]},\n",
    "        \n",
    "        {\"name\":\"kMeans_n_clusters\",\"type\":\"range\",\"bounds\":[2,5]},\n",
    "        {\"name\":\"kMeans_n_init\" ,\"type\": \"range\", \"bounds\": [60,90]},\n",
    "        {\"name\":\"kMeans_max_iter\" ,\"type\": \"range\", \"bounds\": [200,250]},\n",
    "        \n",
    "        {\"name\":\"gbr_learning_rate\" ,\"type\": \"choice\",\"value_type\":\"float\", \n",
    "         \"values\": [0.07,0.06,0.05]},\n",
    "        {\"name\":\"gbr_n_estimators\" ,\"type\": \"range\", \"bounds\": [150,200]},\n",
    "]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "766b98e8",
   "metadata": {},
   "source": [
    "I used KNNImputer to impute missing values of train and test features. I chose KNNImputer as there are more contextural imputation that simple imputer with using simple metric like median.\n",
    "\n",
    "Standard scaler is used for removing mean and scaling to unit variance. Just like standard normal distribution.\n",
    "\n",
    "PCA is used for dimensionality reduction so that there will be as much varianceas possible in the data that can be explained.\n",
    "\n",
    "\n",
    "gbr builts estimators are built sequentially and one tries to reduce the bias of the combined estimator. The motivation is to combine several weak models to produce a useful ensemble.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 188,
   "id": "804dc380",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import accuracy_score\n",
    "def train(estimator,x_train,y_train):\n",
    "    estimator.fit(x_train,y_train)\n",
    "    return estimator\n",
    "\n",
    "def evaluate(predictions,y_test):\n",
    "    return mean_absolute_error(y_test,predictions)\n",
    "\n",
    "def train_evaluate(parameters):\n",
    "    predictions = train_predict_before_evalutate(parameters)\n",
    "    return evaluate(predictions,y_test) \n",
    "\n",
    "def train_predict_before_evalutate(parameters):\n",
    "     preprocessing = Pipeline(steps=[\n",
    "       (\"knnImputer\", KNNImputer(n_neighbors=parameters[\"knnImputer_n_clusters\"], \n",
    "                                 weights=\"uniform\",  \n",
    "                                 copy=False)),\n",
    "     ])\n",
    "\n",
    "     estimator = Pipeline(steps=[\n",
    "            ('proc', preprocessing),   \n",
    "            (\"standardScaler\",StandardScaler()),\n",
    "            (\"pca\",PCA(random_state=rng,\n",
    "                      n_components = parameters[\"pca_n_components\"])),\n",
    "            (\"meanRegressor\",MeanRegressor(kMeans_n_clusters=parameters[\"kMeans_n_clusters\"],\n",
    "                                           kMeans_n_init=parameters[\"kMeans_n_init\"],\n",
    "                                           kMeans_max_iter =parameters[\"kMeans_max_iter\"],\n",
    "                                           gbr_learning_rate=parameters[\"gbr_learning_rate\"], \n",
    "                                           gbr_n_estimators=parameters[\"gbr_n_estimators\"]\n",
    "                                          )\n",
    "            )\n",
    "     ]) \n",
    "     train(estimator,x_train,y_train)\n",
    "     return estimator.predict(x_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1cbc50df",
   "metadata": {},
   "source": [
    "train_predict is only used for predicting data from the test csv\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 189,
   "id": "678ac689",
   "metadata": {},
   "outputs": [],
   "source": [
    "def train_predict(parameters, x_train,y_train,x_test):\n",
    "     preprocessing = Pipeline(steps=[\n",
    "       (\"knnImputer\", KNNImputer(n_neighbors=parameters[\"knnImputer_n_clusters\"], \n",
    "                                 weights=\"uniform\",  \n",
    "                                 copy=False)),\n",
    "     ])\n",
    "\n",
    "     estimator = Pipeline(steps=[\n",
    "            ('proc', preprocessing),   \n",
    "            (\"standardScaler\",StandardScaler()),\n",
    "            (\"pca\",PCA(random_state=rng,\n",
    "                      n_components = parameters[\"pca_n_components\"])),\n",
    "            (\"meanRegressor\",MeanRegressor(kMeans_n_clusters=parameters[\"kMeans_n_clusters\"],\n",
    "                                           kMeans_n_init=parameters[\"kMeans_n_init\"],\n",
    "                                           kMeans_max_iter =parameters[\"kMeans_max_iter\"],\n",
    "                                           gbr_learning_rate=parameters[\"gbr_learning_rate\"], \n",
    "                                           gbr_n_estimators=parameters[\"gbr_n_estimators\"]\n",
    "                                          )\n",
    "            )\n",
    "     ]) \n",
    "     train(estimator,x_train,y_train)\n",
    "     return estimator.predict(x_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 190,
   "id": "ef87b70b",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[INFO 12-03 19:55:37] ax.service.utils.instantiation: Inferred value type of ParameterType.INT for parameter knnImputer_n_clusters. If that is not the expected value type, you can explicity specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
      "[INFO 12-03 19:55:37] ax.service.utils.instantiation: Inferred value type of ParameterType.INT for parameter pca_n_components. If that is not the expected value type, you can explicity specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
      "[INFO 12-03 19:55:37] ax.service.utils.instantiation: Inferred value type of ParameterType.INT for parameter kMeans_n_clusters. If that is not the expected value type, you can explicity specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
      "[INFO 12-03 19:55:37] ax.service.utils.instantiation: Inferred value type of ParameterType.INT for parameter kMeans_n_init. If that is not the expected value type, you can explicity specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
      "[INFO 12-03 19:55:37] ax.service.utils.instantiation: Inferred value type of ParameterType.INT for parameter kMeans_max_iter. If that is not the expected value type, you can explicity specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
      "[WARNING 12-03 19:55:37] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:37] ax.service.utils.instantiation: Inferred value type of ParameterType.INT for parameter gbr_n_estimators. If that is not the expected value type, you can explicity specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
      "[INFO 12-03 19:55:37] ax.modelbridge.dispatch_utils: Using Bayesian optimization since there are more ordered parameters than there are categories for the unordered categorical parameters.\n",
      "[INFO 12-03 19:55:37] ax.modelbridge.dispatch_utils: Using Bayesian Optimization generation strategy: GenerationStrategy(name='Sobol+GPEI', steps=[Sobol for 14 trials, GPEI for subsequent trials]). Iterations after 14 will take longer to generate due to  model-fitting.\n",
      "[INFO 12-03 19:55:37] ax.service.managed_loop: Started full optimization with 16 steps.\n",
      "[INFO 12-03 19:55:37] ax.service.managed_loop: Running optimization trial 1...\n",
      "[WARNING 12-03 19:55:37] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:37] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:37] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:37] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:37] ax.service.managed_loop: Running optimization trial 2...\n",
      "[WARNING 12-03 19:55:37] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:37] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:37] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:37] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:38] ax.service.managed_loop: Running optimization trial 3...\n",
      "[WARNING 12-03 19:55:38] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:38] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:38] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:38] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:38] ax.service.managed_loop: Running optimization trial 4...\n",
      "[WARNING 12-03 19:55:38] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:38] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:38] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:38] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:39] ax.service.managed_loop: Running optimization trial 5...\n",
      "[WARNING 12-03 19:55:39] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:39] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:39] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[WARNING 12-03 19:55:39] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:39] ax.service.managed_loop: Running optimization trial 6...\n",
      "[WARNING 12-03 19:55:39] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:39] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:39] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:39] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:40] ax.service.managed_loop: Running optimization trial 7...\n",
      "[WARNING 12-03 19:55:40] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:40] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:40] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:40] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:40] ax.service.managed_loop: Running optimization trial 8...\n",
      "[WARNING 12-03 19:55:40] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:40] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:40] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:40] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:41] ax.service.managed_loop: Running optimization trial 9...\n",
      "[WARNING 12-03 19:55:41] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:41] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:41] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:41] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:41] ax.service.managed_loop: Running optimization trial 10...\n",
      "[WARNING 12-03 19:55:41] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:41] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:41] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:41] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:41] ax.service.managed_loop: Running optimization trial 11...\n",
      "[WARNING 12-03 19:55:41] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:41] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:41] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:42] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:42] ax.service.managed_loop: Running optimization trial 12...\n",
      "[WARNING 12-03 19:55:42] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:42] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[WARNING 12-03 19:55:42] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:42] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:42] ax.service.managed_loop: Running optimization trial 13...\n",
      "[WARNING 12-03 19:55:42] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:42] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:42] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:42] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:43] ax.service.managed_loop: Running optimization trial 14...\n",
      "[WARNING 12-03 19:55:43] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:43] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:43] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:43] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:43] ax.service.managed_loop: Running optimization trial 15...\n",
      "[WARNING 12-03 19:55:43] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:43] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:43] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:43] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:47] ax.service.managed_loop: Running optimization trial 16...\n",
      "[WARNING 12-03 19:55:47] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:47] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:47] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:47] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n"
     ]
    }
   ],
   "source": [
    "timeSeriesSplit = TimeSeriesSplit()\n",
    "fold_indices = []\n",
    "\n",
    "for rest, pick in timeSeriesSplit.split(sj_train_features):\n",
    "    fold_indices.append(pick)\n",
    "trainIndices = fold_indices[0]\n",
    "np.append(trainIndices, fold_indices[1])\n",
    "np.append(trainIndices, fold_indices[2])\n",
    "testIndices = fold_indices[3]\n",
    "np.append(testIndices,fold_indices[4])\n",
    "\n",
    "x_train = sj_train_features.iloc[trainIndices,:]\n",
    "y_train = sj_train_labels.iloc[trainIndices,:]\n",
    "x_test = sj_train_features.iloc[testIndices,:]\n",
    "y_test = sj_train_labels.iloc[testIndices,:]\n",
    "\n",
    "best_parameters, values, experiment, model = optimize(\n",
    "    parameters=parameterScope,\n",
    "    evaluation_function=train_evaluate,\n",
    "    objective_name='accuracy',\n",
    "    total_trials=16,\n",
    "    minimize=True\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 191,
   "id": "80bb2ae5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best Parameters: {'knnImputer_n_clusters': 17, 'pca_n_components': 4, 'kMeans_n_clusters': 2, 'kMeans_n_init': 75, 'kMeans_max_iter': 230, 'gbr_n_estimators': 168, 'gbr_learning_rate': 0.06}\n",
      "\n",
      "Mean Absolute Value: 32.38818198504621\n"
     ]
    }
   ],
   "source": [
    "print(\"Best Parameters: {}\".format(best_parameters))\n",
    "print(\"\\nMean Absolute Value: {}\".format(train_evaluate(best_parameters)))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "442573e6",
   "metadata": {},
   "source": [
    "<b>Train and predict sj data</b>\n",
    "\n",
    "Proper formatting for resulting csv file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 192,
   "id": "95fb4ab1",
   "metadata": {},
   "outputs": [],
   "source": [
    "sj_results = train_predict(best_parameters,sj_train_features,sj_train_labels,sj_test_features)\n",
    "sj_results = pd.DataFrame(data=sj_results,columns=[\"total_cases\"],index= sj_test_features.index)\n",
    "sj_results[\"city\"] = \"sj\"\n",
    "sj_results.reset_index(inplace=True)\n",
    "sj_results.set_index([\"city\",\"year\",\"weekofyear\"],inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f577105a",
   "metadata": {},
   "source": [
    "<b>Hyperparameter tuning for iq dataset</b>\n",
    "\n",
    "Unlike sj, the mae from this scope gives a single digit. I am happy about this so there is no rebounding and values switching in my parameterScope."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 193,
   "id": "079a7c35",
   "metadata": {},
   "outputs": [],
   "source": [
    "#gbr = gradientBoostingRegression \n",
    "parameterScope=[\n",
    "        {\"name\":\"knnImputer_n_clusters\" ,\"type\": \"range\", \"bounds\": [2,10]},\n",
    "\n",
    "        {\"name\":\"pca_n_components\" ,\"type\": \"range\", \"bounds\": [4,10]},\n",
    "        \n",
    "        {\"name\":\"kMeans_n_clusters\",\"type\":\"range\",\"bounds\":[2,10]},\n",
    "        {\"name\":\"kMeans_n_init\" ,\"type\": \"range\", \"bounds\": [5,100]},\n",
    "        {\"name\":\"kMeans_max_iter\" ,\"type\": \"range\", \"bounds\": [200,250]},\n",
    "        \n",
    "        {\"name\":\"gbr_learning_rate\" ,\"type\": \"choice\",\"value_type\":\"float\", \n",
    "         \"values\": [0.01,0.1,1]},\n",
    "        {\"name\":\"gbr_n_estimators\" ,\"type\": \"range\", \"bounds\": [100,200]},\n",
    "]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 194,
   "id": "b6262c70",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[INFO 12-03 19:55:55] ax.service.utils.instantiation: Inferred value type of ParameterType.INT for parameter knnImputer_n_clusters. If that is not the expected value type, you can explicity specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
      "[INFO 12-03 19:55:55] ax.service.utils.instantiation: Inferred value type of ParameterType.INT for parameter pca_n_components. If that is not the expected value type, you can explicity specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
      "[INFO 12-03 19:55:55] ax.service.utils.instantiation: Inferred value type of ParameterType.INT for parameter kMeans_n_clusters. If that is not the expected value type, you can explicity specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
      "[INFO 12-03 19:55:55] ax.service.utils.instantiation: Inferred value type of ParameterType.INT for parameter kMeans_n_init. If that is not the expected value type, you can explicity specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
      "[INFO 12-03 19:55:55] ax.service.utils.instantiation: Inferred value type of ParameterType.INT for parameter kMeans_max_iter. If that is not the expected value type, you can explicity specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
      "[WARNING 12-03 19:55:55] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:55] ax.service.utils.instantiation: Inferred value type of ParameterType.INT for parameter gbr_n_estimators. If that is not the expected value type, you can explicity specify 'value_type' ('int', 'float', 'bool' or 'str') in parameter dict.\n",
      "[INFO 12-03 19:55:55] ax.modelbridge.dispatch_utils: Using Bayesian optimization since there are more ordered parameters than there are categories for the unordered categorical parameters.\n",
      "[INFO 12-03 19:55:55] ax.modelbridge.dispatch_utils: Using Bayesian Optimization generation strategy: GenerationStrategy(name='Sobol+GPEI', steps=[Sobol for 14 trials, GPEI for subsequent trials]). Iterations after 14 will take longer to generate due to  model-fitting.\n",
      "[INFO 12-03 19:55:55] ax.service.managed_loop: Started full optimization with 10 steps.\n",
      "[INFO 12-03 19:55:55] ax.service.managed_loop: Running optimization trial 1...\n",
      "[WARNING 12-03 19:55:55] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:55] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:55] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:55] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:55] ax.service.managed_loop: Running optimization trial 2...\n",
      "[WARNING 12-03 19:55:55] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:55] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:55] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:55] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:55] ax.service.managed_loop: Running optimization trial 3...\n",
      "[WARNING 12-03 19:55:55] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:55] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:55] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:55] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:56] ax.service.managed_loop: Running optimization trial 4...\n",
      "[WARNING 12-03 19:55:56] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:56] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:56] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:56] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:56] ax.service.managed_loop: Running optimization trial 5...\n",
      "[WARNING 12-03 19:55:56] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:56] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:56] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "[WARNING 12-03 19:55:56] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:57] ax.service.managed_loop: Running optimization trial 6...\n",
      "[WARNING 12-03 19:55:57] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:57] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:57] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:57] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:57] ax.service.managed_loop: Running optimization trial 7...\n",
      "[WARNING 12-03 19:55:57] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:57] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:57] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:57] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:58] ax.service.managed_loop: Running optimization trial 8...\n",
      "[WARNING 12-03 19:55:58] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:58] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:58] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:58] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:58] ax.service.managed_loop: Running optimization trial 9...\n",
      "[WARNING 12-03 19:55:58] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:58] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:58] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:58] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[INFO 12-03 19:55:59] ax.service.managed_loop: Running optimization trial 10...\n",
      "[WARNING 12-03 19:55:59] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:59] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:59] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n",
      "[WARNING 12-03 19:55:59] ax.core.parameter: `sort_values` is not specified for `ChoiceParameter` \"gbr_learning_rate\". Defaulting to `True` for parameters of `ParameterType` FLOAT. To override this behavior (or avoid this warning), specify `sort_values` during `ChoiceParameter` construction.\n"
     ]
    }
   ],
   "source": [
    "timeSeriesSplit = TimeSeriesSplit()\n",
    "fold_indices = []\n",
    "\n",
    "for rest, pick in timeSeriesSplit.split(iq_train_features):\n",
    "    fold_indices.append(pick)\n",
    "trainIndices = fold_indices[0]\n",
    "np.append(trainIndices, fold_indices[1])\n",
    "np.append(trainIndices, fold_indices[2])\n",
    "testIndices = fold_indices[3]\n",
    "np.append(testIndices,fold_indices[4])\n",
    "\n",
    "x_train = iq_train_features.iloc[trainIndices,:]\n",
    "y_train = iq_train_labels.iloc[trainIndices,:]\n",
    "x_test = iq_train_features.iloc[testIndices,:]\n",
    "y_test = iq_train_labels.iloc[testIndices,:]\n",
    "\n",
    "best_parameters, values, experiment, model = optimize(\n",
    "   parameters=parameterScope,\n",
    "    evaluation_function=train_evaluate,\n",
    "    objective_name='accuracy',\n",
    "    total_trials=10,\n",
    "    minimize=True\n",
    ")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 195,
   "id": "cb8504d7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best Parameters: {'knnImputer_n_clusters': 5, 'pca_n_components': 4, 'kMeans_n_clusters': 2, 'kMeans_n_init': 21, 'kMeans_max_iter': 234, 'gbr_n_estimators': 126, 'gbr_learning_rate': 0.01}\n",
      "\n",
      "Mean Absolute Value: 8.855224274075985\n"
     ]
    }
   ],
   "source": [
    "print(\"Best Parameters: {}\".format(best_parameters))\n",
    "print(\"\\nMean Absolute Value: {}\".format(train_evaluate(best_parameters)))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fada328a",
   "metadata": {},
   "source": [
    "<b>Train and predict iq data</b>\n",
    "\n",
    "Proper formatting for resulting csv file\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 196,
   "id": "e6bba42a",
   "metadata": {},
   "outputs": [],
   "source": [
    "iq_results = train_predict(best_parameters,iq_train_features,iq_train_labels,iq_test_features)\n",
    "iq_results = pd.DataFrame(data=iq_results,columns=[\"total_cases\"],index= iq_test_features.index)\n",
    "iq_results[\"city\"] = \"iq\"\n",
    "iq_results.reset_index(inplace=True)\n",
    "iq_results.set_index([\"city\",\"year\",\"weekofyear\"],inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 210,
   "id": "3f5cdc66",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                      total_cases\n",
      "city year weekofyear             \n",
      "sj   2008 18                    6\n",
      "          19                   24\n",
      "          20                   18\n",
      "          21                   19\n",
      "          22                   28\n",
      "...                           ...\n",
      "iq   2013 22                    7\n",
      "          23                    6\n",
      "          24                    6\n",
      "          25                    5\n",
      "          26                    6\n",
      "\n",
      "[416 rows x 1 columns]\n"
     ]
    }
   ],
   "source": [
    "results = pd.concat([sj_results,iq_results])\n",
    "\n",
    "results = results.apply(np.ceil)\n",
    "results = results.astype(dtype='int64')\n",
    "print(results)\n",
    "results.to_csv(\"results.csv\")"
   ]
  }
 ],
 "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": "3.8.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}