{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Nearest Neighbor Regression\n",
    "### M2 Modélisation\n",
    "\n",
    "October 2026"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import math, random\n",
    "import numpy as np\n",
    "import scipy.stats as sps\n",
    "from time import perf_counter"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 1. Theory"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Given a sample $\\mathcal{D}_n=\\{(\\mathbf{X}_1,Y_1), \\ldots, (\\mathbf{X}_n,Y_n)\\}$ of i.i.d. random couples with the same distribution as, and independent of, a generic pair $(\\mathbf{X}, Y)\\in\\mathbb{R}^d\\times\\mathbb{R}$, the $k$-nearest neighbor estimate of the regression function is defined for any $\\mathbf{x}\\in\\mathbb{R}^d$ by\n",
    "$$\n",
    "r_n(\\mathbf{x}):=\\frac{1}{k}\\sum_{i=1}^{k}Y_{(i)}(\\mathbf{x})=\\sum_{i=1}^{n}W_i(\\mathbf{x})Y_{i},\n",
    "$$\n",
    "where $W_i(\\mathbf{x})=1/k$ if $\\mathbf{X}_i$ belongs to the $k$ nearest neighbors of $\\mathbf{x}$, and 0 otherwise.\n",
    "\n",
    "Under the following set of assumptions :\n",
    "- The random variable $\\mathbf{X}$ is bounded, namely $\\mathbf{X}\\in[0,1]^d$,\n",
    "- The regression function $r$ is $L$-Lipschitz with respect to the supremum norm,\n",
    "- For $\\mu$-almost every $\\mathbf{x}$, we have $\\sigma^2(\\mathbf{x})\\leq\\sigma^2<\\infty$,\n",
    "\n",
    "we know that:\n",
    "- For $d=1$, there exists a sequence $(k_n)$ with $k_n\\sim\\frac{\\sigma\\sqrt{n}}{\\sqrt{2}L}$ and a universal constant $c_1=2\\sqrt{2}$ such that \n",
    "$$\n",
    "\\mathbb{E}\\left[(r_n(\\mathbf{X})-r(\\mathbf{X}))^2\\right]\\leq c_1\\frac{\\sigma L}{\\sqrt{n}}.\\hspace{1cm}(1)\n",
    "$$\n",
    "- For $d\\geq 2$, there exists a sequence $(k_n)$ with $k_n\\sim\\left(\\frac{\\sigma^2}{4L^2}\\right)^{\\frac{d}{d+2}}n^{\\frac{2}{d+2}}$ and a universal constant $c_d=2^{\\frac{3d+2}{d+2}}$ such that\n",
    "$$\n",
    "\\mathbb{E}\\left[(r_n(\\mathbf{X})-r(\\mathbf{X}))^2\\right]\\leq c_d\\left(\\frac{\\sigma^2 L^d}{n}\\right)^{\\frac{2}{d+2}}.\\hspace{1cm}(2)\n",
    "$$\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 2. Simulations"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In order to illustrate these rates of convergence, we consider the following framework: \n",
    "$$Y=r(\\mathbf{X})+\\varepsilon,$$\n",
    "where $\\mathbf{X}\\sim\\mathcal{U}_{[0,1]^d}$ is independent of $\\varepsilon\\sim\\mathcal{N}(0,\\sigma^2)$, and  \n",
    "$$\n",
    "r(x):=L \\sin\\|x\\|_\\infty.\n",
    "$$"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2.1. Rate of convergence when everything is known"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For a fixed value of $n$, we simulate $N$ samples $\\mathcal{D}_n^{(j)}=\\{(\\mathbf{X}_1^{(j)},Y_1^{(j)}), \\ldots, (\\mathbf{X}_n^{(j)},Y_n^{(j)})\\}$ and associate the corresponding regression estimate $r_n^{(j)}$. Then, to estimate the $L_2$ risk, we simulate an independent new point $\\mathbf{X}^{(j)}$, and consider the empirical mean\n",
    "$$\n",
    "\\frac{1}{N}\\sum_{j=1}^{N}\\left(r_n^{(j)}(\\mathbf{X}^{(j)})-r(\\mathbf{X}^{(j)})\\right)^2\\xrightarrow[N\\to\\infty]{a.s.}\\mathbb{E}\\left[(r_n(\\mathbf{X})-r(\\mathbf{X}))^2\\right],\n",
    "$$\n",
    "where the almost sure convergence comes from the law of large numbers. \n",
    "\n",
    "The number $k=k_n$ of nearest neighbors is defined above and depends on $n$, $d$, $L$ and $\\sigma^2$. Thus, depending on dimension $d$, we take \n",
    "$$k_n=\\left\\lceil\\frac{\\sigma\\sqrt{n}}{\\sqrt{2}L}\\right\\rceil\\hspace{1cm}\\mathrm{or}\\hspace{1cm}\\left\\lceil\\left(\\frac{\\sigma^2}{4L^2}\\right)^{\\frac{d}{d+2}}n^{\\frac{2}{d+2}}\\right\\rceil.$$\n",
    "The idea is to do this for several values of $n$ and compare the empirical mean\n",
    "$$\n",
    "\\frac{1}{N}\\sum_{j=1}^{N}\\left(r_n^{(j)}(\\mathbf{X}^{(j)})-r(\\mathbf{X}^{(j)})\\right)^2\n",
    "$$\n",
    "to the theoretical upper bound given by (1) or (2), depending on dimension $d$. \n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    ">**Prerequisite.**\n",
    "Considering two sets of vectors $\\mathcal X^{(\\ell)}:=(\\mathbf{X}_1^{(\\ell)},\\dots,\\mathbf{X}_{n_\\ell}^{(\\ell)})\\in\\mathbb{R}^{n_\\ell\\times d}$ and $\\mathcal X^{(t)}:=(\\mathbf{X}_1^{(t)},\\dots,\\mathbf{X}_{n_t}^{(t)})\\in\\mathbb{R}^{n_t\\times d}$ as arguments, the following function returns a matrix $O\\in\\mathbb{R}^{n_t\\times n_\\ell}$ such that $O[i,j]$ is the index of the $j$-th nearest neighbor (with respect to the supremum norm) of $\\mathbf{X}_i^{(t)}$ among the set $\\mathcal X^{(\\ell)}$.\n",
    ">"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "def order(Xl, Xt):\n",
    "    unl = np.ones(Xl.shape[0])\n",
    "    unt = np.ones(Xt.shape[0])\n",
    "    dist = np.zeros((Xt.shape[0], Xl.shape[0]))\n",
    "    for k in range(Xt.shape[1]):\n",
    "        dist = np.maximum(dist, np.abs(np.outer(Xt[:, k], unl) - np.outer(unt, Xl[:, k])))\n",
    "    order = np.argsort(dist, axis=1)\n",
    "    return order"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    ">**Question 1.**\n",
    "Write a code to compare the empirical risk\n",
    "$$\n",
    "\\frac{1}{N}\\sum_{j=1}^{N}\\left(r_n^{(j)}(\\mathbf{X}^{(j)})-r(\\mathbf{X}^{(j)})\\right)^2\n",
    "$$\n",
    "to the theoretical upper bound given by (1) or (2), depending on dimension $d$. \n",
    ">"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "def r(x, L):\n",
    "    return L*np.sin(np.max(np.abs(x), axis=1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "def Kopt(n, d, L, sigma):\n",
    "    if d==1:\n",
    "        return math.ceil(sigma * np.sqrt(n) / (np.sqrt(2) * L))\n",
    "    return math.ceil((sigma**2/(4 * L**2))**(d/(d+2)) * n**(2/(d+2)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "def L2theo(n_values, d, L, sigma):\n",
    "    if d==1:\n",
    "        return 2 * np.sqrt(2) * sigma * L / np.sqrt(n_values)\n",
    "    return 2**((3*d+2)/(d+2)) * (sigma**2 * L**d / n_values)**(2/(d+2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Duration (s) : 2.9\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAlIAAAHLCAYAAAADXnZlAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8pXeV/AAAACXBIWXMAAA9hAAAPYQGoP6dpAAB2/ElEQVR4nO3deVxU9f7H8dcMOyggIAgK7guoqOCSG+K+lKllWZbbNZfUTL1lWfdWt7rXe63UFjRNc6ks01JbXMJcwy3Ffd93xRUUZJ35/THJL8IFcGBY3s/HYx7FmTPf8zmHEd6c85nvMZjNZjMiIiIikmtGWxcgIiIiUlQpSImIiIjkkYKUiIiISB4pSImIiIjkkYKUiIiISB4pSImIiIjkkYKUiIiISB4pSImIiIjkkYKUiIiISB4pSImIiIjkkYKUiIiISB4pSIkIP/zwAwaDgXnz5tm6FBGRIkVBSkSIjY0FoEGDBjauRESkaFGQEhFiY2Nxc3OjZs2ati4FgKSkpDw9Z43xC4Ktty8i1qMgJSLExsZSr149jEbr/kg4fPgwvXv3xtfXFycnJ4KDg4mKisqyzltvvYXBYCA2NpaePXtSpkwZqlatet/nAH777Tfatm1L6dKlcXV1pVmzZvz88885Hv+vFi9ejMFg4Ndff8323NSpUzEYDOzatctq+3fp0iUGDx5MYGAgTk5OlC1blubNm7Ny5crM1/fv359KlSplq+f2uLflZKw7uT3O3r17efrpp/Hw8MDPz4+//e1vxMfH3/O1IgL2ti5ARGzr0qVLnD17lh49elh13H379tGsWTOCgoL44IMPKFeuHCtWrGDkyJFcvnyZN998M8v6jz32GE899RRDhw4lMTHxvs+tXbuW9u3bExoaysyZM3FycmLKlCl07dqVr7/+ml69euV4/NseeeQRfH19mTVrFm3bts3y3OzZswkLCyM0NNRq+9enTx9iY2P597//TY0aNbh+/TqxsbFcuXIll0f7wcd6/PHH6dWrFwMHDmT37t2MGzcOgM8//zzXtYiUKGYRKdGWLVtmBswzZ8684/PJycnm/v37mytUqGAuXbq0uUmTJuaYmJj7jtuxY0dzhQoVzPHx8VmWjxgxwuzs7Gy+evWq2Ww2m998800zYH7jjTeyjXGv5x566CGzr6+v+caNG5nL0tPTzXXq1DFXqFDBbDKZ7jvGnYwZM8bs4uJivn79euayffv2mQHzxx9/bNX9K1WqlHnUqFH3rKdfv37mihUrZlt+e9zcjHUnt8eZMGFCluXDhg0zOzs7Zx5HEbkzXdoTKeFuN5qHhYXd8fn09HQqV65MTEwM169f5/nnn+fRRx+9Z59PcnIyv/76Kz169MDV1ZX09PTMR5cuXUhOTmbTpk1ZXvP444/fdby/PpeYmMjmzZvp2bMnpUqVylxuZ2dHnz59OHPmDAcPHszx+H/2t7/9jVu3bjF//vzMZbNmzcLJyYnevXtbdf8aN27M7Nmzeffdd9m0aRNpaWk5qvFOHnSsRx99NMvXoaGhJCcnExcXl+eaREoCBSmREi42NhYnJydq1659x+fd3Nx44403CAoKwmg00q9fP0wmE4cPH77rmFeuXCE9PZ2PP/4YBweHLI8uXboAcPny5Syv8ff3v+t4f33u2rVrmM3mO74mICAgs4acjv9ntWvXplGjRsyaNQuAjIwMvvzyS7p164aXl5dV92/+/Pn069ePGTNm0LRpU7y8vOjbty8XLlzIUa3WHMvb2zvL105OTgDcunUr17WIlCTqkRIp4WJjY6lTpw4ODg45Wv/AgQPcunXrrg3bAGXKlMk8OzR8+PA7rlO5cuUsX/+5cfqv/vpcmTJlMBqNnD9/Ptu6586dA8DHxyfH4//VgAEDGDZsGPv37+fYsWOcP3+eAQMGZNm+NfbPx8eHyZMnM3nyZE6dOsUPP/zAq6++SlxcHMuXLwfA2dmZlJSUbK/9a1DLyVgiYn0KUiIl2PXr1zl+/Djt2rXL0fpJSUn06dOHf/zjH1kuqf2Vq6srrVu3Zvv27YSGhuLo6GitkgHLWbImTZrw/fff8/777+Pi4gKAyWTiyy+/pEKFCtSoUSPP4z/99NOMGTOG2bNnc+zYMcqXL0+HDh0yn8+P/QsKCmLEiBH8+uuvxMTEZC6vVKkScXFxXLx4ET8/PwBSU1NZsWJFrscSEetTkBIpwW73R2VkZLB48eJsz7ds2TLzkk9aWhpPPvkkISEhvPbaa/cd+8MPP6RFixa0bNmS559/nkqVKnHjxg2OHDnCjz/+yKpVqx6o9vHjx9O+fXtat27NSy+9hKOjI1OmTGHPnj18/fXXuToD9Veenp706NGD2bNnc/36dV566aVsU0M86P7Fx8fTunVrevfuTa1atShdujS///47y5cv57HHHstcr1evXrzxxhs89dRTvPzyyyQnJ/PRRx+RkZGR67FExPoUpERKsO3btwOWj7jf6WPup06dwtvbG5PJRN++fbGzs2PmzJk5CikhISHExsbyzjvv8I9//IO4uDg8PT2pXr16Zh/Rg2jVqhWrVq3izTffpH///phMJurVq8cPP/zAI4888sDjDxgwgK+//hqwzOX0Vw+6f87OzjRp0oQvvviCEydOkJaWRlBQEK+88gpjx47NXK9y5cosWbKE1157jZ49e+Lv78+YMWO4dOkS//rXv3I1lohYn8FsNpttXYSIFG6DBg3i8OHDLF++HGdnZ1uXIyJSaChIicg9nTx5kkqVKuHs7IydnV3m8mXLltGyZUsbViYiYnsKUiIiIiJ5pHmkRERERPJIQUpEREQkjxSkRERERPJIQUpEREQkjzSPVC6ZTCbOnTtH6dKlH2jCPxERESk4ZrOZGzduEBAQkG2C3QehIJVL586dIzAw0NZliIiISB6cPn2aChUqWG08BalcKl26NAAzZsyge/fuOb7R6/2kpaXxyy+/0KFDB6uNKcWf3jfWVZKOZ1Hc18Jasy3rKuht5/f28mv8tLQ0Fi9ezHPPPZf5e9xaFKRy6fblPFdXV9zd3a0apKw9phR/et9YV0k6nkVxXwtrzbasq6C3nd/by6/xb48LWL0tR83mIiIiInmkICUiIiKSR7q0JyIi2WRkZJCWlmbrMrJIS0vD3t6e5ORkMjIybF1OJlvWVdDbzu/tPcj4Dg4OWe4HWlAUpEREJJPZbObChQtcv37d1qVkYzabKVeuHKdPny5U08/Ysq6C3nZ+b+9Bx/f09KRcuXIF+n1QkBIRkUy3Q5Svry+urq6FKrCYTCZu3rxJqVKlrDoP0IOyZV0Fve383l5exzebzSQlJREXFweAv7+/1Wu7GwUpEREBLJfzbocob29vW5eTjclkIjU1FWdn50IXpGxVV0FvO7+39yDju7i4ABAXF4evr2+BXeYrPO9EERGxqds9Ubc/Ji5S1Nx+7xZkf5+ClIiIZFGYLueJ5IYt3rslOkj16NGDMmXK0LNnT1uXIiIiIkVQiQ5SI0eOZO7cubYuQ0RERIqoEh2kWrdubfV77oiIiEjJ8UBBavz48RgMBkaNGmWlcizWrVtH165dCQgIwGAwsHjx4juuN2XKFCpXroyzszPh4eGsX7/eqnUUKLPJ1hWIiIhILuV5+oPff/+d6dOnExoaes/1YmJiaNy4cbabDx44cCBz4qy/SkxMpF69egwYMIDHH3/8juPOnz+fUaNGMWXKFJo3b860adPo3Lkz+/btIygoCIDw8HBSUlKyvfaXX34hICAgp7t6V9b8VIBh8fOEnztLxikfCGpstXGleLv9HixsM1AXVSXpeN5pX9PS0jCbzZhMJkymwvfHndlszvxvYaqvdevWhISE8Mknn9y1rjZt2lCvXj0mTZpklW3eHm/ixIlAwR2T/P4ePOj4JpMJs9lMWlpalukP8vPftMF8u+pcuHnzJmFhYUyZMoV3332X+vXrM3ny5GzrmUwmwsLCqF69Ot98803mTh06dIhWrVoxevRoxo4de+8CDQYWLVpE9+7dsyxv0qQJYWFhTJ06NXNZcHAw3bt3Z/z48TnelzVr1vDJJ5+wcOHCe64XFRVFVFQUGRkZHDp0iHnz5lntI8KOaQl03DMSI5Y3zRW36hz17cR5j3AwlOirryJSgOzt7SlXrhyBgYE4OjraupxcGTZsGF9//XW25W3btr3vz/cHde3aNezt7e/ZKpKTdXLjkUceoW7duvf8fdehQwdCQkLu+Ps5N2x5bHMrNTWV06dPc+HCBdLT07M8l5SURO/evYmPj8fd3d1q28zTGanhw4fz8MMP065dO9599927rmc0Glm6dCkRERH07duXL774guPHj9OmTRseffTR+4aou0lNTWXbtm28+uqrWZZ36NCBDRs25GnM+xk+fDjDhw8nISEBDw8PANq3b5/tTFtepYTX5NKPbxF4fQveiYfxPn4Ys2clTI0HYwp9GpzUyyXZpaWlER0dbdX3YklWko7nnfY1OTmZ06dPU6pUKZydnW1cYXZms5kbN25QunTpbB9zd3BwoGPHjnz++edZljs5OVn1l+adlC5d+q51paam4ujoaPUa7O3tcXR0vOu2TSYT+/bto3///g+87T8fW7PZnDnzuLOz813Hvr3fOV1+292+x/d73W3Jycm4uLgQERGR5T2clpbGkiVL7vv6vMh1kPrmm2+IjY3l999/z9H6AQEBrFq1ioiICHr37s3GjRtp27Ytn376aa6Lve3y5ctkZGTg5+eXZbmfnx8XLlzI8TgdO3YkNjaWxMREKlSowKJFi2jUqFGOX+/g4GC9H7YVGrC94hD8n/0Uh+2z4feZGK6fwO6X17Bb+z8I7wdNhoBHBetsT4oVq74XpUQdzz/va0ZGBgaDAaPRiNFoxGw2cyvNNjcHdnGwyxZKbl/quV3jnxkMBpydne/athEZGUndunWxs7Njzpw5ODo68s477/DMM88wYsQIFi5ciK+vL5988gmdO3fO8ro6deoA8OWXX2JnZ8fzzz/PO++8k1lfZGQkwcHBREVF0aZNG+rUqYOjoyNz586ldu3arF27lsjIyCxXb0wmE++99x6fffYZp0+fxs/PjyFDhvD6668DsHz5ct5991327NmDnZ0dTZs25cMPP6Rq1apZ9vl2DX89JgcOHCAxMZGGDRs+8Azkfz62JpOJhIQE3N3ds4x7+zjdab//uvyXX37h5Zdf5ptvviEhIYGGDRsyadIkGjVqlPk9btu27R3Hux+j0YjBYCjQf8O5ClKnT5/mxRdf5JdffsnVXytBQUHMnTuXVq1aUaVKFWbOnGmVSbP+OobZbM7VuCtWrHjgGqyutD+0fQNa/h12fg0bp8DVo7DhI9gYBbV7QNPhUD7M1pWKSDF3Ky2DkDds83Ny39sdcXW07l3M5syZw9ixY9myZQvz58/n+eefZ/HixfTo0YPXXnuNSZMm0adPH06dOpWldWPOnDkMHDiQzZs3s3XrVgYPHkzFihUZNGjQXbfz/PPPExMTw926Z8aNG8dnn33GpEmTaNGiBefPn+fAgQOZzycmJjJmzBjq1q1LYmIib7zxBj169GDHjh05CkaxsbHY29vft4/Zmu62339dPnbsWL777jvmzJlDxYoVmTBhAh07duTIkSN4enred7zCJlcxddu2bcTFxREeHo69vT329vasXbuWjz76CHt7ezIy7vyXy8WLFxk8eDBdu3YlKSmJ0aNHP1DRPj4+2NnZZTv7FBcXl+0sVZHl6AaNnoMRW+Hpb6BSSzBnwJ6F8Flr+Lwz7P8JTLb5a1FEpLD56aefKFWqVJbHO++8k/l8vXr1+Mc//kH16tUZN24cLi4u+Pj4MGjQIKpXr84bb7zBlStX2LVrV5ZxAwMDmTRpEjVr1uSZZ57hhRdeuGfTeLVq1ZgwYQI1a9akVq1a2Z6/ceMGH374IRMmTKBfv35UrVqVFi1a8Nxzz2Wu8/jjj/PYY49RvXp16tevz8yZM9m9ezf79u3L0bGIjY0lJCTkjic9Tp8+TWRkJCEhIYSGhrJgwYL7jnf72Lq7u1OhQgXc3d2zHNt77feflwcGBjJ16lTee+89OnfuTEhICJ999hkuLi7MnDkzR+MVNrmK+23btmX37t1Zlg0YMIBatWrxyiuv3PEGgZcvX6Zt27YEBwezYMECDh8+TGRkJE5OTrz//vt5KtrR0ZHw8HCio6Pp0aNH5vLo6Gi6deuWpzELLaMRana2PM7tgE1TYM93cGqD5VGmMjw0DOr3BqdStq5WRIoRFwc79r3d0Wbbzq3WrVtn+QASgJeXV+b///nsjJ2dHd7e3tStWzdz2e0/xOPi4rKM8dBDD2W52tG0aVM++OADMjIy7vh7r2HDhvesc//+/aSkpNC2bdu7rnP06FH++c9/smnTJi5fvpx5yevUqVOZlxrvJTY2lrCwO1+5sLe3Z/LkydSvX5+4uDjCwsLo0qULbm5udx3v9rE1mUyZPVI+Pj5Z1rnbfv95+dGjR0lLS6N58+aZyxwcHGjcuDH79+/P0XiFTa6CVOnSpbN9A93c3PD29r7jN9ZkMtGpUycqVqzI/Pnzsbe3Jzg4mJUrV9K6dWvKly9/x7NTN2/e5MiRI5lfHz9+nB07duDl5ZU5tcGYMWPo06cPDRs2pGnTpkyfPp1Tp04xdOjQ3OxS0RJQHx6bDu3egi3TYevncO04LHsZVr8L4QOg8WDwKG/rSkWkGDAYDFa/vJaf3NzcqFat2l2f/2vPzO1emj9/DTzwx/rvFUgAXFxc7jtG165dCQwM5LPPPsvsTapTpw6pqan3fa3ZbGbHjh13vf2Zv78//v7+APj6+uLl5cXVq1fvWfftY3u3Hqnb69zttX+uDXLWmnO/41hY5Otn641GI+PHj+e7777L0m1ft25dVq5ceddv8tatW2nQoAENGjQALKGpQYMGvPHGG5nr9OrVi8mTJ/P2229Tv3591q1bx9KlS6lYsWJ+7lLh4B5gCVOj90GX98GrCiTHQ8xk+DAUvhtkOXslIiIPbNOmTdm+rl69+h3PRuVE9erVcXFx4ddff73j81euXGH//v384x//yLyic+3atRyPf/ToUeLj4+96RurPtm7dislkIjAwMMfjP4hq1arh6OjIb7/9lrksLS2NrVu3EhwcXCA1WNsD/6mxZs2aez7fvn37Oy6vX7/+XV8TGRmZo8ayYcOGMWzYsPuuV2w5lYLGg6Dh3+DQcksz+skY2P2t5VGxhaUxvUYnyyVCEZFiLCUlJVvvrL29fbZLULl1+vRpxowZw5AhQ4iNjeXjjz/mgw8+yPN4zs7OvPLKK4wdOxZHR0eaN2/OpUuX2Lt3LwMHDqRMmTJ4e3szffp0/P39OXXqVLbpfu4lNjYWsFy+3LNnT+ZyBwcHatasmfn1lStX6Nu3LzNmzLjvmLePrclk4saNGyQlJeHo6JjrY+vm5sbzzz/Pyy+/nHmVacKECSQlJTFw4MBcjVVYFJ1ztnJ3Rjuo9bDlcTbW0ke1dxGc/M3y8KoKDz1v6aNyLBqnSkVEcmv58uWZl6xuq1mzZpZPw+VF3759uXXrFo0bN8bOzo4XXniBwYMHP9CY//znP7G3t+eNN97g3Llz+Pv7Z7amGI1GvvnmG0aOHEmdOnWoWbMmH330EZGRkTkae/v27YClt+vPHnroITZu3AhYglGPHj0YN24czZo1u++Y1jy2//3vfzGZTPTp04cbN27QsGFDVqxYQZkyZQrVjPU5laeZzUuy2xNyzps3j549e1ptnoq0tDSWLl1Kly5drDNm/Jk/+qhmQ0q8ZZmzp+XsVePB4O5/r1dLEWH1900JV5KO5532NTk5mePHj2few7SwuVd/Tn756/xPhaWuB9m22Wymd+/e1KxZk7feeivft1eQ49/tPZyWlsbChQvzZWZzXe8prjwqQPu3Ycw+6DwBylSC5Ovw20SYXBe+HwLnd9q6ShERKWAxMTHMnz+fxYsXU79+ferXr5/tE/mSc7q0V9w5lbLMiN7oOTi41NJHdWoj7PrG8qjUEpqOgOod1EclIlICtGjRokheQiusFKRKCqMdBHe1PM5us8yYvncRnFhveXhXs8xHVe9pcLTOzZhFRIq6+32gSkSnIEqi8uHQcya8uBOajQQnD7hyBH4eA5NC4Nd34EbO71koIiJSUilIlWSegdDhHRizFzr9Dzwrwq1rsP59mFQHFj0PF3TdXERE5G4UpAScSsNDQ2HkdnjyCwh8CExpsHMefNoC5jwKh34BXVMXERHJQj1S8v+MdhDyqOVxZqulMX3fEji+1vLwqfFHH9VT4HD/WxyIiIgUdzojJXdWoSE8MQte3GH5VJ+TO1w+BD+NgokhsOpduHHR1lWKiIjYlIKU3JtnEHT8N4zeCx3Hg0cQ3LoK696DyXVg8TC4sOf+44iIiBRDClKSM87u0HSYpY/qiTlQoTFkpMKOr+DT5jC3GxyOVh+ViIiUKOqRktyxs4fa3S2P01ssfVT7f4BjaywPn5qWwBXaS31UIiJS7OmMlORdYGN4cg6M3AEPDQfH0nD5IPz4IkyqDav/AzfjbF2liJRwJ06cwGAwsGPHDluXkiuVKlW65z3+cisyMpJRo0ZZbTxbsPYxsQYFKXlwZSpCp/9Y7uvX8T+WPqqkK7D2f5ZAtWQ4XNxn6ypFpBgyGAz3fPTv39/WJd7X7Nmz8fT0zLb8999/Z/DgwQVax5+PXalSpQgPD+f7778vsBqKIl3aE+txdoemw6HxEDjwI2z4BM5uhe1fWh5V21ier9oWDAZbVysixcD58+cz/3/+/Pm88cYbHDx4MHOZi4sL165ds0VpZGRkYDAYMObxPqZly5a1ckX35+7unnn8bty4waxZs3jyySfZu3cvNWvWLPB6igKdkRLrs7OH2j1g0K8wMBpCuoHBCEdXwZePw5SmEDsX0pJtXamI3IvZDKmJtnmYzTkqsVy5cpkPDw8PDAZDtmW3HTt2jNatW+Pq6kq9evXYuHFjlrE2bNhAREQELi4uBAYGMnLkSBITEzOfv3btGn379qVMmTK4urrSuXNnDh8+nPn87TNLP/30EyEhITg5OXHy5ElSU1MZO3Ys5cuXx83NjSZNmmTew2/NmjUMGDCA+Pj4zDNBb731FpD9Mtb169cZPHgwfn5+ODs7U6dOHX766ScArly5wtNPP02FChVwdXWlbt26fP3117n5bgNkOX7Vq1fn3XffxWg0smvXrix19OvX767H4a233qJ+/fpZxp08eTKVKlXK/Lp///50796d999/H39/f7y9vRk+fDhpaWmZ68TFxdG1a1dcXFyoXLkyX331Va73pyDojJTkr8DGEDgXrp2AzdMsAerSfvjhBVj5L2g8CBoOhFIF/5eXiNxHWhL8J8A2237tHDi6WXXI119/nffff5/q1avz+uuv8/TTT3PkyBHs7e3ZvXs3HTt25J133mHmzJlcunSJESNGMGLECGbNmgVYfvkfPnyYH374AXd3d1555RW6dOnCnj3/PwVMUlIS48ePZ8aMGXh7e+Pr68uAAQM4ceIE33zzDQEBASxatIhOnTqxe/dumjVrxuTJk7OcSStVqlS22k0mE507d+bGjRt8+eWXVK1alX379mH44+x+cnIy4eHhvPLKK7i7u/Pzzz/Tp08fqlSpQpMmTfJ0vDIyMpg7dy4AYWFhmcuHDRvGiRMnsh2Hffv24eDgkOPxV69ejb+/P6tXr+bIkSP06tWL0NBQevXqBViO9+nTp1m1ahWOjo6MHDmSuLjC13erICUFo0wl6DQeIl+1hKlNn0LCGVgzHtZPhHq9LA3rvrVsXamIFFMvvfQSDz/8MAD/+te/qF27NkeOHKFWrVq899579O7dO7MZu3r16nz00Ue0atWKqVOncvr0aX744QdiYmJo1qwZAF999RWBgYEsXryYjh07ApCWlsaUKVOoV68eAEePHuXrr7/mzJkzBAQEZNaxfPlyZs2axX/+858sZ9LuZuXKlWzZsoX9+/dTo0YNAKpUqYLJZCIhIYHy5cvz0ksvZa7/wgsvsHz5chYsWJCrIBUfH58Z5G7duoWDgwPTp0+natWqABw+fJhly5axfv16WrRoke04PPHEEzneVpkyZfjkk0+ws7OjVq1aPPzww6xatYpevXpx6NAhli1bxqZNmzLrnzlzJsHBwTkev6AoSEnBcvaAZi9Ak6GWaRM2fALnYi3hKnYuVGtn6aOq0lp9VCK25uBqOTNkq21bWWhoaOb/+/v7A5bLR7Vq1WLbtm0cOXIky+Ujs9mMyWTi+PHjHD58GHt7+yyhxNvbm5o1a3LgwIHMIOXo6JhlO7GxsZjN5szwc1tKSgre3t45rn3Hjh1UqFAh2zi3ZWRkMH78eObPn8/Zs2dJSUkhJSUFN7fcndUrXbo0sbGxgOXs2sqVKxkyZAje3t507dqV/fv33/U47N+/P1fbql27NnZ2dplf+/v7s3v3boDM7TRs2DDz+Vq1at2xKd/WFKTENuwcoM7jUPsxOL0ZNn4C+3+CIystD98QS6Cq+wTYO9m6WpGSyWCw+uU1W/rzZafbl8RMf0wibDKZGDJkCCNHjsz2uqCgIA4dOnTHMc1mc+ZYYGlu//PXJpMJOzs7tm3bliU0wJ0v4d2Ni8u95+WbOHEikyZNYvLkydStWxc3NzdGjRpFampqjrcBYDQaqVatWubXoaGh/PLLL/zvf/+ja9eumO/Su/bn42A0GrOt9+fep9v+ehnQYDBkfj9uv95QBP6gVpAS2zIYIOghy+PqsT/6qL6AuH2WaRMy+6j+Bm4+tq5WRIqpsLAw9u7dmyVE/FlISAjp6els3rw589LelStXOHToELVq3b0loUGDBmRkZBAXF0fLli3vuI6joyMZGRn3rC80NJQzZ85w6NChO56VWr9+Pd26dePZZ58FLAHu8OHDVrkUZmdnx61bt4Csx+H2pb3bx+H2tsqWLcuFCxeyhKvczuEVHBxMeno6W7dupXHjxgAcPHiQ69evP/D+WJs+tSeFh1cV6Pw/y3xU7d8G9/KQGAer/22Zj+rHF+HSwfuPIyKSS6+88gobN25k+PDh7NixI7Op/IUXXgAsPVPdunVj0KBB/Pbbb+zcuZNnn32W8uXL061bt7uOW6NGDZ555hn69u3L999/z/Hjx/n999/53//+x9KlSwHLp/Nu3rzJr7/+yuXLl0lKSso2TqtWrYiIiODxxx8nOjqa48ePs2zZMpYvXw5AtWrViI6OZsOGDezfv58hQ4Zw4cKFXB8Hs9nMhQsXuHDhAsePH2f69OmsWLEicx+rV69Oly5dGDJkyF2PQ2RkJJcuXWLChAkcPXqUqKgoli1blqs6atasSadOnRg0aBCbN29m27ZtPPfcc/c9M2cLClJS+Lh4QvMX4cWd8PhM8K8P6cmwbTZENYavnrDcjiaHH48WEbmf0NBQ1q5dy+HDh2nZsiUNGjTgn//8Z2YvFcCsWbMIDw/nkUceoWnTppjNZpYuXXrfT6rNmjWLvn378ve//52aNWvy6KOPsnnzZgIDAwFo1qwZQ4cOpVevXpQtW5YJEybccZzvvvuORo0a8fTTTxMSEsLYsWMzz2T94x//ICwsjI4dOxIZGUm5cuXo3r17ro9DQkIC/v7++Pv7ExwczAcffMDbb7/N66+/nrlOVFQUYWFhdz0OwcHBTJkyhaioKOrVq8eWLVuyNMLn1KxZswgMDKRVq1Y89thjDB48GF9f31yPk98M5rtd8JQ7SkhIwMPDg3nz5tGzZ89cfdTzXtLS0li6dCldunSx2pjFhtkMpzZa7ut34Gfgj7esXx1LH1Wdx0tsH5XeN9ZVko7nnfY1OTmZ48ePU7lyZZydnW1cYXa3P6Hm7u6e50ku84Mt6yrobef39h50/Lu9h9PS0li4cCG9e/cmPj4ed3d3q9VceN6JIndjMEDFZvDUV/DCNmg82PKJnot7YPHzMLkurHsPEq/YulIRESlhFKSkaPGuCl3es/RRtXsLSvvDzYuw6l1LH9VPo+Hy4fsOIyIiYg0KUlI0uZSBFqPhxV3w2GfgXw/Sb8HWz+GThjCvFxxbqz4qERHJVwpSUrTZO0LokzB4LfT/GWp2AQxwaDnMfRSmtYSd30B67uZSERERyQkFKSkeDAao1AKe/hpGbIVGz4G9C1zYDYuGWPqo1n8ASVdtXalIoafPIElRZYv3roKUFD8+1eDhDyx9VG3fgFLl4OYF+PVtSx/Vz3+Hy0dsXaVIoXP703t3msdIpCi4/d4tyE/damZzKb5cvaDl36HpC7D3e8ttaC7sht9nwO8zoWZny/QJFZvrvn4iWGaw9vT0JC4uDgBXV9dCdYsOk8lEamoqycnJhW76A1vVVdDbzu/t5XV8s9lMUlIScXFxeHp6ZrsdT35SkJLiz94R6j0Fob3gxHrYOAUOLYODSy0P/3rQdASEdLesK1KClStXDiAzTBUmZrOZW7duZbufna3Zsq6C3nZ+b+9Bx/f09Mx8DxcUBSkpOQwGqBxheVw+DJumwo55cH4nfD8Iot+EJoMhvL/lU4EiJZDBYMDf3x9fX9873mjWltLS0li3bh0RERGFasJUW9ZV0NvO7+09yPgODg4FeibqNgUpKZl8qsMjE6HNPyxTJmyZDjfOwcq3YO0EaPAsNBlqmbdKpASys7OzyS+le7GzsyM9PR1nZ+dCFaRsWVdBbzu/t1dYv8f3UnguMovYgqsXRLwEo3ZD908tt51JS7IEq4/D4Ztn4ESM5qMSEZE7UpASAcu9+uo/DUN/g74/QPWOgBkO/ASzu8D0SNi1ADIK16UOERGxLQUpkT8zGKBKK3jmWxj+O4QPAHtnOL8Dvn8OPqwHv02GW9dsXamIiBQCClIid1O2BnSdDKP3Qet/gJsvJJyFlW/CxNqwdCxcPWbrKkVExIZKdJDq0aMHZcqUoWfPnrYuRQozN29o9TKM3gPdpoBvbUhLhC3T4KMwSx/VyY3qoxIRKYFKdJAaOXIkc+fOtXUZUlTYO0GDZ+D5GOizGKq1J7OPalYn+KwN7F6oPioRkRKkRAep1q1bU7p0aVuXIUWNwQBVW8OzC2HYZgjrB3ZOcC4WvhsIH9aHmI/g1nVbVyoiIvks10Fq6tSphIaG4u7ujru7O02bNmXZsmVWLWrdunV07dqVgIAADAYDixcvvuN6U6ZMoXLlyjg7OxMeHs769eutWofIffnWgkc/gtF7IfI1cCsLCWcg+p+W+/otexWunbB1lSIikk9yPSFnhQoV+O9//0u1atUAmDNnDt26dWP79u3Url072/oxMTE0btw428RaBw4cuOtU7omJidSrV48BAwbw+OOP37GO+fPnM2rUKKZMmULz5s2ZNm0anTt3Zt++fQQFBQEQHh5OSkpKttf+8ssvBAQE5HbXs7HmrL+3xypsMwlLDjl5QvMx0GQYhj3fYbdlKoZLB2DzVMxbpmGu+TCmJs9jLt/Iqvf10/vGukrS8SyK+1pYa7ZlXQW97fzeXn6Nn5/Hx2A2P3iHrJeXF++99x4DBw7MstxkMhEWFkb16tX55ptvMmfJPXToEK1atWL06NGMHTv23gUaDCxatIju3btnWd6kSRPCwsKYOnVq5rLg4GC6d+/O+PHjc1z7mjVr+OSTT1i4cOE914uKiiIqKoqMjAwOHTrEvHnzcHV1zfF2pIQxmyl7Yw9V45bjd2N35uKrrlU56tuJ854NMRsK16zRIiLFWVJSEr179yY+Ph53d3erjftAt4jJyMhgwYIFJCYm0rRp02zPG41Gli5dSkREBH379uWLL77g+PHjtGnThkcfffS+IepuUlNT2bZtG6+++mqW5R06dGDDhg15GvN+hg8fzvDhw0lISMDDwwOA9u3bW20K+7S0NKKjo606ptjaw8ArpF06gN2WTzHsXoBX0lG8TkRhdq+AqfFgTPWeBee8/4PW+8a6StLxLIr7WlhrtmVdBb3t/N5efo2flpbGkiVLrDben+UpSO3evZumTZuSnJxMqVKlWLRoESEhIXdcNyAggFWrVhEREUHv3r3ZuHEjbdu25dNPP81z0ZcvXyYjIwM/P78sy/38/Lhw4UKOx+nYsSOxsbEkJiZSoUIFFi1aRKNGjXL8egcHB6u/kfJjTLGxgLrQPQravQVbZ8KWzzAknMFu5RvYrXsPwvpCkyFQpmKeN6H3jXWVpONZFPe1sNZsy7oKetv5vb3C+j2+kzwFqZo1a7Jjxw6uX7/Od999R79+/Vi7du1dw1RQUBBz586lVatWVKlShZkzZ2KwQp/IX8cwm825GnfFihUPXINIjpUqC5GvQvNRsPtb2BgFlw7ApijYPBWCH4WmIyAw52FeRERsK0/THzg6OlKtWjUaNmzI+PHjqVevHh9++OFd17948SKDBw+ma9euJCUlMXr06DwXDODj44OdnV22s09xcXHZzlKJFDoOzpazUMM2wbPfQdU2YDbBvsUwsx3MaA97F0NGuq0rFRGR+7DKPFJms/mOn44Dy2W4tm3bEhwczPfff8+qVav49ttveemll/K8PUdHR8LDw4mOjs6yPDo6mmbNmuV5XJECZTBAtXbQZxE8vwEaPAt2jnBmCyzoBx83gI1TIDnB1pWKiMhd5PrS3muvvUbnzp0JDAzkxo0bfPPNN6xZs4bly5dnW9dkMtGpUycqVqzI/Pnzsbe3Jzg4mJUrV9K6dWvKly9/x7NTN2/e5MiRI5lfHz9+nB07duDl5ZU5tcGYMWPo06cPDRs2pGnTpkyfPp1Tp04xdOjQ3O6SiO351YZuUdD2Tfh9huVx/RSsGAdrxv9/H5VnkK0rFRGRP8l1kLp48SJ9+vTh/PnzeHh4EBoayvLly2nfvn22dY1GI+PHj6dly5Y4OjpmLq9bty4rV67E29v7jtvYunUrrVu3zvx6zJgxAPTr14/Zs2cD0KtXL65cucLbb7/N+fPnqVOnDkuXLqVixbw37IrYXClfaP0atBgNu+Zb+qguH4KNn8CmqRDSzdJHVSHc1pWKiAh5CFIzZ87M1fp3ClgA9evXv+trIiMjycn0VsOGDWPYsGG5qkekSHBwgfD+0KAvHP3VEqSOrYG931segQ9B0+FQtYOtKxURKdEeaB4pEclnRiNUb295XNgDm6bArm/h9CY4vQl7z4pUcWsJKS3BwcvW1YqIlDgl+qbFIkVKuTrQfQqM3gMRL4OLF4brJ6l79kvsP64Hv/wD4s/YukoRkRJFQUqkqCldDtr8A0bvJaPz+9xw8seQkgAbPobJobDwb3B2m62rFBEpERSkRIoqR1dMYf1ZFTye9CfnQaWWYM6APd/BZ23g806w/0cwZdi6UhGRYks9UiJFncGIuXoHCHkYzu+y9FHtXginNloeZSrBQ8Og/jPgVMrW1YqIFCs6IyVSnPiHQo9PYdRuaPl3cCkD107AsrEwKQSi34D4s7auUkSk2FCQEimO3P2h7Rswei88/AF4VYXkeIj5ED4Mhe+eg3PbbV2liEiRpyAlUpw5ukGj52DEVnj6G0sflSkddi+A6ZEwqwsc+Fl9VCIieaQeKZGSwGiEmp0tj/M7Lffw27MQTsZYHl5V/uij6m0JXyIikiM6IyVS0vjXg8emWfqoWowBZ0+4egyWvgQTQ2DlW5BwztZViogUCQpSIiWVewC0exPG7IMu71vOSiVfh98mweS68P1gOLfD1lWKiBRqClIiJZ2jGzQeZOmjeuprqNjC0ke1az5MbwWzH4GDy8BksnWlIiKFjnqkRMTCaAe1ulge57Zb+qj2fg8n1lseXlWh6TCo97T6qERE/qAzUiKSXUADePwzeHEXNH8RnDzg6lH4+e8wqTb8+jYknLd1lSIiNqcgJSJ351Ee2r9t6aPqPMEyS/qta7D+A0sf1aKhltnURURKKAUpEbk/p1LQZAi8EAu9voSgpmBKg51fw7SWMKcrHFqhPioRKXHUIyUiOWe0g+CulsfZbX/0US2C4+ssD+/qlj6q0KfA0dXW1YqI5DudkRKRvCkfDj1nwos7odlISx/VlcPw02hLH9Wqd+HGRVtXKSKSrxSkROTBeAZCh3dgzF7o9D/wrAi3rsK692ByHVg8DC7ssXWVIiL5QkFKRKzDqTQ8NBRGbocnv4DAhyAjFXZ8BZ82hzmPwqFf1EclIsWKeqRExLqMdhDyqOVxZitsjIJ9S+D4WsvDp4blvn71ngIHF1tXKyLyQHRGSkTyT4WG8MQseHEHNB0BTu5w+RD8NOqPPqp/q49KRIo0BSkRyX+eQdDx3zB6L3QcDx5BkHQF1k34o49qOFzca+sqRURyTUFKRAqOs7tleoSR2+GJOVCh8R99VF/C1GYwtzscXglms60rFRHJEQUpESl4dvZQuzs8Fw0DoyGkOxiMcGw1fPU4THkIts2BtGRbVyoick8KUiJiW4GN4ck5MHIHPDQcHEvDpQPw40hLH9Xq8XDzkq2rFBG5IwUpESkcylSETv+xzEfV4d/gEQhJl2Htfy2BaskIiNtv6ypFRLJQkBKRwsXZA5qNsJyh6jnLMoN6Rgps/8Jyye+Lx+DIr+qjEpFCQUFKRAonO3uo8xg89yv87RcIftTSR3X0V/jyMZjSFGK/UB+ViNiUgpSIFG4GAwQ1gV5fwAux0OR5cCwFl/bDDyMs0yes+R8kXrZ1pSJSAilIiUjR4VUZOv/XMh9V+3fAvQIkXoI1/4GJIfDDSIg7YOsqRaQEUZASkaLHxROaj7TMmP74TAhoYOmjip0DU5rAlz3h6Gr1UYlIvlOQEpGiy84B6vaEQathwHKo9QhggCPR8EV3mNoctn8F6Sm2rlREiikFKREp+gwGqNgUnvoKRsZC4yHg4AZxe2HJMJhUB9a+B4lXbF2piBQzClIiUrx4VYEuEyzzUbV/G9zLQ2IcrH4XJoXAj6Pg0iFbVykixYSClIgUTy5loPmL8OJOSx+Vf31IT4ZtsyCqEXz1JBxbqz4qEXkgClIiUrzd7qMavAYGLPv/PqrDK2Duo/BpS9jxNaSn2rpSESmC7G1dgIhIgTAYoGIzy+PKUdj8KWz/Ei7uhsVDYeWbGMMH4pBe3taVikgRoiAlIiWPd1Xo8h5EjrNMmbB5Gtw4j93a/9DB4IjB4XfLbWp8qtu6UhEp5HRpT0RKLlcvaDEaXtwFj32G2a8u9uZU7GJnwycNYV4vOL5OfVQiclcKUiIi9o4Q+iTpA1fxW7XXMFXvBBjg0HKY0xWmtYSd36iPSkSyUZASEbnNYOBK6VpkPPkljNgKjZ4Dexe4sBsWDYHJdWH9B5B01daVikghoSAlInInPtXg4Q9gzD5o+waUKgc3L8Cvb8Ok2vDz3y1N6yJSoilIiYjci6sXtPw7jNoNPaZBubqQlgS/z4CPw+Hrp+HEb+qjEimhFKRERHLC3hHqPQVD1kO/H6FGJ8AMB5fC7IdheivY9a36qERKGAUpEZHcMBigcgT0nm/po2r4N0sf1fmd8P0g+LAe/DYJbl2zdaUiUgAUpERE8sqnOjwyCUbvhTb/gFJ+cOMcrHwLJobA0pfVRyVSzClIiYg8KDdviHjZ0kfVfSr41bH0UW2Zbumj+uYZOBGjPiqRYkhBSkTEWuydoH5vGPob9F0C1TsAZjjwE8zuAtMjYdcCyEizdaUiYiUKUiIi1mYwQJVIeGYBDN8C4f3B3hnO74Dvn/ujj2qy+qhEigEFKRGR/FS2JnT90NJH1fp1cCsLCWdh5ZswsTYsHQtXj9m6ShHJIwUpEZGC4OYDrcbCqD3QLQp8QyAtEbZMg4/CLH1UJzeqj0qkiFGQEhEpSA7O0OBZeH4D9FkE1dqT2Uc1qxN81gZ2L1QflUgRoSAlImILBgNUbQPPLoRhmyGsH9g5wblY+G4gfFgfYj6CW9dtXamI3IOClIiIrfnWgkc/svRRRb72Rx/VGYj+p+W+fstehWsnbF2liNyBgpSISGFRqixEvmLpo3r0EygbDKk3YfNU+KgBzO8Dpzarj0qkEFGQEhEpbBycIawPDNsIz34PVduC2QT7f4DPO8CMdrDne8hIt3WlIiWeva0LEBGRuzAYoFpby+PiPtg0BXbNh7NbYeEA8AiEJkMtocvZw9bVipRIOiMlIlIU+IVAt08sfVStXgVXb4g/Db+8bpmPavlrcO2krasUKXEUpAqJa0mppJtsXYWIFHqlfKH1OEug6voR+NSE1BuwKQo+qg/f9oPTv9u6SpESo0QHqR49elCmTBl69uxp0zrMZjMvf7eH93bZsf3UdZvWIiJFhIMLhPeDYZvgme+gSmtLH9W+xTCzHcxoD3sXq49KJJ+V6CA1cuRI5s6da+syuJCQzJ6zCVy4ZaDXjC28uWQPN1P0w09EcsBohOrtoO9iyySf9Z8FO0c4swUW9IOPG8DGKEhOsHWlIsVSiQ5SrVu3pnTp0rYuA38PF5aPbE6TsibMZpiz8STtJ67l1/0XbV2aiBQlfrWhe5Rl+oSIseDiBddPwYrXYGIIrHjd8rWIWE2ug9T48eNp1KgRpUuXxtfXl+7du3Pw4EGrFrVu3Tq6du1KQEAABoOBxYsX33G9KVOmULlyZZydnQkPD2f9+vVWraMgebo60Luaidn9wwnycuV8fDID52xlxLxYLt1IsXV5IlKUlPaDNq/DmH3wyGTwqWHpo9r4iWXG9AX9MZzdZusqRYqFXE9/sHbtWoYPH06jRo1IT0/n9ddfp0OHDuzbtw83N7ds68fExNC4cWMcHByyLD9w4ACenp6UK1cu22sSExOpV68eAwYM4PHHH79jHfPnz2fUqFFMmTKF5s2bM23aNDp37sy+ffsICgoCIDw8nJSU7CHkl19+ISAgILe7nk1amvXuhXV7rMZB7vw0vCkfrT7K5zEn+GnXedYfvsS4TjV5rIElWIrcdvt9Y833YklW/I6nPdR7FkJ7Yzj6K8bNUzGeWAd7F2G/dxEt3KpjCkohLeRRMNrZutj7KqzfH1vWVdDbzu/t5df4+Xl8DGbzg02Re+nSJXx9fVm7di0RERFZnjOZTISFhVG9enW++eYb7Ows/1APHTpEq1atGD16NGPHjr13gQYDixYtonv37lmWN2nShLCwMKZOnZq5LDg4mO7duzN+/Pgc179mzRo++eQTFi5ceM/1oqKiiIqKIiMjg0OHDjFv3jxcXV1zvJ28OH0Tvjlmx5lES3iq4WGiVxUTPs75ulkRKcbck05R9dIyKlzbhNGcAUCiY1mOle3AKe8I0u1cbFyhSP5ISkqid+/exMfH4+7ubrVxH3hCzvj4eAC8vLyyPWc0Glm6dCkRERH07duXL774guPHj9OmTRseffTR+4aou0lNTWXbtm28+uqrWZZ36NCBDRs25GnM+xk+fDjDhw8nISEBDw/LxHft27fPdqYtr9LS0oiOjs425sAME7M2nuTDX49yKB7e22PPi22q0b9pEPZ2JbrFTbj7+0bypuQcz6EkXzvD6e/eoEb8etySL1H37FfUufwDpgZ9MTUcBB4VbF1kNoX1+2PLugp62/m9vfwaPy0tjSVLllhtvD97oCBlNpsZM2YMLVq0oE6dOndcJyAggFWrVhEREUHv3r3ZuHEjbdu25dNPP83zdi9fvkxGRgZ+fn5Zlvv5+XHhwoUcj9OxY0diY2NJTEykQoUKLFq0iEaNGuX49Q4ODlZ/I/11TAcHGNa6Bl3qlue1RbvZcPQK/1txiJ/3XOC/j4VSp7xmM5b8eS+WZCXieJapwIGAnlTp9wkOexfCpikYrhzBblMUdps/hdrdoelwKB9u60qzKazfH1vWVdDbzu/tFdbv8Z080CmNESNGsGvXLr7++ut7rhcUFMTcuXOZP38+9vb2zJw50yq9Pn8dw2w252rcFStWcOnSJZKSkjhz5kyuQlRBq+TjxlfPNWFCz1A8XBzYczaBblExjF+2n1upGbYuT0SKKgdXaDQQhv8OT8+HSi3BnAF7voPP2sDnnWD/j2DSzxmRO8lzkHrhhRf44YcfWL16NRUq3PsU8MWLFxk8eDBdu3YlKSmJ0aNH53WzAPj4+GBnZ5ft7FNcXFy2s1TFicFg4MmGgUSPieDhUH8yTGamrT1Gpw/XseHIZVuXJyJFmdEINTtB/59gyDoIfQqM9nBqI8x/Fj4Og83TIOWmrSsVKVRyHaTMZjMjRozg+++/Z9WqVVSuXPme61++fJm2bdsSHByc+Zpvv/2Wl156Kc9FOzo6Eh4eTnR0dJbl0dHRNGvWLM/jFhW+pZ2J6h3GjL4N8fdw5uSVJHrP2MzYhTu5npRq6/JEpKjzrwePTYNRu6HFGHD2hGsnYNlYy3xU0W9A/FlbVylSKOQ6SA0fPpwvv/ySefPmUbp0aS5cuMCFCxe4detWtnVNJhOdOnWiYsWKmZf1goODWblyJbNnz2bSpEl33MbNmzfZsWMHO3bsAOD48ePs2LGDU6f+fyK5MWPGMGPGDD7//HP279/P6NGjOXXqFEOHDs3tLhVZ7UL8+GV0BH2bVsRggG+3nqHdxLX8tOscD/hhTBERcA+Adm9a5qN6+APwqgop8RDzIXwYCt89B2djbV2liE3lutn89nQDkZGRWZbPmjWL/v37Z1lmNBoZP348LVu2xNHRMXN53bp1WblyJd7e3nfcxtatW2ndunXm12PGjAGgX79+zJ49G4BevXpx5coV3n77bc6fP0+dOnVYunQpFStWzO0uFWmlnR14u1sdHq0XwKvf7+ZI3E1GzNvO4uCzvNO9Dv4e+iiziDwgRzdo9ByE/w0Or7DccubEeti9wPKo2NzSmF6jU5GYj0rEmnIdpHJ7pqN9+/Z3XF6/fv27viYyMjJH2xk2bBjDhg3LVT3FVcNKXvw8sgVTVh9lypojrNwfx6Zj63ilU02eaVIRo1ETeYrIAzIaoWZny+PcDtg0xdKUfjLG8vCqAg8Ng/q9LeFLpATQRETFiJO9HaPb1+DnkS0JC/LkZko6/1yylyembeTwxRu2Lk9EipOA+vDY9D/6qEaDswdcPQZLX7L0Ua18CxLO2bpKkXynIFUM1fArzcKhzXi7W23cHO3YdvIaD3/0G5NXHiIlXR9hFhErcg+Adm/B6H3Q5X3LWank6/DbJJhcF74fbDl7JVJMKUgVU0ajgb5NKxE9phVta/mSmmFi8srDPPLRb2w7ec3W5YlIceNUChoPghFb4al5lr4pUzrsmg/TW8HsR+DgMjCZbF2piFUpSBVzAZ4uzOjXkI+fboBPKUcOx92k56cbeHPJHm6mpNu6PBEpbox2UOthGLAUBq2Guk9Y5qM6sR6+fgo+aQi/z4DURFtXKmIVClIlgMFgoGu9AFaOacUT4RUwm2HOxpO0n7iWX/dftHV5IlJclQ+Dx2fAizuh+Yvg5AFXj8LPf4dJteHXtyHhvK2rFHkgClIliKerI+89UY8vBzYhyMuV8/HJDJyzlRHzYrl0I8XW5YlIceVRAdq/bZmPqvMEKFMJbl2D9R9Y+qgWDYXzu2xdpUieKEiVQC2q+7BiVARDIqpgNMBPu87TbuJaFmw9rYk8RST/OJWCJkPghVjo9SUENQVTGuz8Gqa1hDld4dAK9VFJkaIgVUK5ONoxrkswP4xoQe0Ad+JvpfHywl08O3MzJ6+od0FE8pHRDoK7wt+Ww6BVUKcnGOzg+DqY9yRENYatn0Nqkq0rFbkvBakSrk55D5YMb864zrVwsjcSc+QKHSevY/q6o6Rn6K9CEcln5cOh50xLH1WzkZY+qiuH4afRlj6qVe/CDfVySuGlICXY2xkZ0qoqK0ZF0KyqN8lpJv6z9ADdp8Sw52y8rcsTkZLAMxA6vANj9kKn/4FnRbh1Fda9ZwlUi56Hi3ttXaVINgpSkqmSjxtfPdeECT1D8XBxYM/ZBLpFxTB+2X5upWoiTxEpAE6l4aGhMHI7PPkFBD70Rx/VPBxmtKLZ4f9iOBKtPiopNBSkJAuDwcCTDQOJHhPBw6H+ZJjMTFt7jE4frmPDkcu2Lk9ESgqjHYQ8CgNXwHO/Qu0emA12lL25D/v5T8OUJrB1FqTdsnWlUsIpSMkd+ZZ2Jqp3GDP6NsTfw5mTV5LoPWMzYxfu5HpSqq3LE5GSpEJDeGI26cO3csS3M2an0nD5EPw06o8+qn+rj0psRkFK7qldiB+/jI6gb9OKGAzw7dYztJu4lp92ndNUCSJSsDwC2Vv+adJf2AUdx4NHECRdgXUTYHIdWDxcfVRS4BSk5L5KOzvwdrc6LBjSlGq+pbh8M5UR87YzaO5WzsfrtLqIFDCn0tB0mKWP6ok5UKExZKTCji9hajOY2x0OrwT9sScFQEFKcqxhJS9+HtmCF9tWx8HOwMr9cbSfuI4vNp7AZNIPLBEpYHb2ULs7PBcNA6MhpDsYjHBsNXz1OEx5CLbNgbRkW1cqxZiClOSKk70do9vX4OeRLQkL8uRmSjr/XLKXJ6Zt5PDFG7YuT0RKqsDG8OQcGLkDHhoOjqXh0gH4caSlj2r1eLh5ydZVSjGkICV5UsOvNAuHNuPtbrVxc7Rj28lrPPzRb0xeeYiUdE2VICI2UqYidPqPZT6qDv8Gj0BIugxr/2sJVEtGQNx+W1cpxYiClOSZ0Wigb9NKRI9pRdtavqRmmJi88jCPfPQb205es3V5IlKSOXtAsxGWM1Q9Z1lmUM9Ige1fWC75ffEYHPlVfVTywBSk5IEFeLowo19DPn66AT6lHDkcd5Oen27gzSV7uJmSbuvyRKQks7OHOo9Z5qL62y8Q/Kilj+ror/DlYzClKcR+oT4qyTMFKbEKg8FA13oBrBzTiifCK2A2w5yNJ2k/cS2/7tf8LiJiYwYDBDWBXl/AC7HQ5HlwLAWX9sMPIyzTJ6z5HyRq4mHJHQUpsSpPV0fee6IeXw5sQpCXK+fjkxk4Zysj5sVy6UaKrcsTEQGvytD5vzB6L7R/B9wrQOIlWPMfSx/VDyPh0kFbVylFhIKU5IsW1X1YMSqCIRFVMBrgp13naTdxLQu2ntZEniJSOLh4QvOR8OIOeHwmBDSA9GSInQNRjeHLnnB0tfqo5J4UpCTfuDjaMa5LMD+MaEHtAHfib6Xx8sJdPDtzMyevJNq6PBERCzsHqNsTBq2GAcuh1iOAAY5EwxfdYWpz2P4VpOusumSnICX5rk55D5YMb864zrVwsjcSc+QKHSevY/q6o6Rn6A7uIlJIGAxQsSk89RWMjIXGQ8DBDeL2wpJhMKkOrH0PEq/YulIpRBSkpEDY2xkZ0qoqK0ZF0KyqN8lpJv6z9ADdp8Sw52y8rcsTEcnKqwp0mWCZj6r92+BeHhLjYPW7MCkEfhwFlw7ZukopBBSkpEBV8nHjq+eaMKFnKB4uDuw5m0C3qBjGL9vPrVRN5CkihYxLGWj+Iry409JH5V/f0ke1bRZENYKvnsRwYp36qEowBSkpcAaDgScbBhI9JoKHQ/3JMJmZtvYYHSevI+aIPnosIoXQ7T6qwWtgwLL/76M6vAL7rx4j8uA/MeyaD+mptq5UCpiClNiMb2lnonqHMaNvQ/w9nDl1NYlnZmzm5QU7uZ6kH0YiUggZDFCxmaWP6oVt0HgwZgdXPG6dwv7H4TC5Lqx7H5Ku2rpSKSAKUmJz7UL8+GV0BH2bVsRggAXbztBu4lp+2nVOUyWISOHlXRW6vEf6CzvZG/Ak5lLl4OYFWPUOTAyBn8bA5SO2rlLymYKUFAqlnR14u1sdFgxpSjXfUly+mcqIedsZNHcr567fsnV5IiJ351KGI36PkD4iFh77DMqFQvot2DoTPgmHeU/B8fXqoyqmFKSkUGlYyYufR7bgxbbVcbAzsHJ/HO0nrmXuxhOYTPohJCKFmJ0jhD4JQ9ZB/5+hZhfAAIeWwZxHYFoE7FQfVXGjICWFjpO9HaPb1+DnkS0JC/IkMTWDN5bs5YlpGzl88YatyxMRuTeDASq1gKe/hhFbodFzYO8CF3bBosHwYSis/0B9VMWEgpQUWjX8SrNwaDPe7lYbN0c7tp28RpeP1jN55SFS0jVVgogUAT7V4OEPYMw+aPsGlCoHN87Dr29b7uv389/hylFbVykPQEFKCjWj0UDfppWIHtOKtrV8ScswM3nlYR756De2ndRfcyJSRLh6Qcu/w6jd0GMalKsLaUnw+wz4OBy+fhpO/KY+qiJIQUqKhABPF2b0a8jHTzfAp5Qjh+Nu0vPTjbyxZA83ktNsXZ6ISM7YO0K9p2DIeuj3I9ToBJjh4FKY/TBMbwW7voUM/VwrKhSkpMgwGAx0rRfAyjGteCK8AmYzzN14kg6T1vHr/ou2Lk9EJOcMBqgcAb3nW/qoGv7N0kd1fid8Pwgmh8Jvk+DWNVtXKvehICVFjqerI+89UY8vBzYhyMuV8/HJDJyzlRHzYrl0Q3dnF5Eixqc6PDIJRu+FNv+AUn5w4xysfMsyH9XSl9VHVYgpSEmR1aK6DytGRTAkogpGA/y06zztJq7l262nNZGniBQ9bt4Q8bKlj6r7VPCrY+mj2jLd0kf1zTNwcoP6qAoZBSkp0lwc7RjXJZgfRrSgdoA78bfSGLtwF8/O3MzJK4m2Lk9EJPfsnaB+bxj6G/RdAtU7AGY48BPM6gyftYbdC9VHVUgoSEmxUKe8B0uGN2dc51o42RuJOXKFjpPXMW3tUdIzTLYuT0Qk9wwGqBIJzyyA4VsgvD/YO8O57fDdQPiwHsaNH2Ofrj8abUlBSooNezsjQ1pVZcWoCJpV9SY5zcT4ZQfoFhXDnrPxti5PRCTvytaErh9a+qhavw5uZSHhLHar/kXHvaMwrhgHV4/ZusoSSUFKip1KPm589VwTJvQMxcPFgb3nEugWFcP4pfu5laqJPEWkCHPzgVZjYdQe6BaFuWww9qYU7LZ+Bh+F/dFHtVF9VAVIQUqKJYPBwJMNA4keE8HDof5kmMxMW3eMjpPXEXPksq3LExF5MA7O0OBZ0getY0PVsZiqtOH/+6g6wWdt1EdVQBSkpFjzLe1MVO8wZvRtiL+HM6euJvHMjM28vGAn15N041ARKeIMBi651yHj6W9h2CYI6wt2TnAu9o8+qvoQ8xEkq70hvyhISYnQLsSPX0ZH0LdpRQwGWLDtDO0mruWnXec0VYKIFA++wfDox5Y+qshx4OoDCWcg+p+W+aiWj4NrJ2xdZbGjICUlRmlnB97uVocFQ5pSzbcUl2+mMmLedgbN3cq567dsXZ6IiHWUKguRr1oC1aMfQ9lakHoTNk2BjxrAt33h9BZbV1lsKEhJidOwkhc/j2zBi22r42BnYOX+ONpPXMvcjScwmXR2SkSKCQdny6W+YZvg2e+gahswm2DfEpjZHj5rC3u+h4x0W1dapClISYnkZG/H6PY1+HlkS8KCPElMzeCNJXt5YtpGDl+8YevyRESsx2CAau2gzyJ4fiM0eBbsHOHsVlg4wHKWasMn6qPKIwUpKdFq+JVm4dBmvN2tNm6Odmw7eY0uH61n8spDpKRrqgQRKWb8QqBblOWyX6tXwNUb4k/BL6/DxNqw/DW4dtLWVRYpClJS4hmNBvo2rUT0mFa0reVLWoaZySsP88hHv7Ht5FVblyciYn2lfKH1a5ZA1fVD8KkJqTdgUxR8VB++7Qenf7d1lUWCgpTIHwI8XZjRryEfP90An1KOHI67Sc9PN/LGkj3cSNZcLCJSDDm4WG49M2wTPLPQcksaswn2LYaZ7WBGe9i7WH1U96AgJfInBoOBrvUCWDmmFU+EV8BshrkbT9Jh0jp+3X/R1uWJiOQPoxGqt7fcJHloDNT/o4/qzBZY0A8+bgAbp0CKekj/SkFK5A48XR1574l6fDmwCUFerpyPT2bgnK2MmBfLpRspti5PRCT/lKsD3aMst6GJGAsuXnD9FKwYh/3HodQ++zXEn7F1lYWGgpTIPbSo7sOKUREMiaiC0QA/7TpPu4lr+XbraU3kKSLFW2k/aPO6pY/qkcngUwNDyg2qxS3DPiocFgyAM9tsXaXNKUiJ3IeLox3jugTzw4gW1A5wJ/5WGmMX7uLZmZs5eSXR1uWJiOQvR1doOACGbSb9yXlcKhWCwZwBe7+HGW1gZkfY9wOYSuYnnRWkRHKoTnkPlgxvzrjOtXCyNxJz5AodJ69j2tqjpGeYbF2eiEj+MhoxV+/AhuqvkvbcGqjXG4wOcHoTfNvHMh/Vpk9LXB+VgpRILtjbGRnSqiorRkXQrKo3yWkmxi87QLeoGPac1WR2IlJC+NWBHlNh9B5o+RK4lIHrJ2H5K5b5qH75Z4npo1KQEsmDSj5ufPVcEyb0DMXDxYG95xLoFhXD+KX7uZVaMk9vi0gJVLoctP0njN4HD08E72qQEg8bPoLJobDwb3C2ePdRKUiJ5JHBYODJhoFEj4ng4VB/Mkxmpq07RsfJ64g5ctnW5YmIFBxHV2g0EIb/Dk/Ph0otwZwBe76Dz9rA551g/4/Fso9KQUrkAfmWdiaqdxgz+jbE38OZU1eTeGbGZl5esJPrSam2Lk9EpOAYjVCzE/T/CYasg9CnwGgPpzbC/Gfh4zDYPA1Sbtq6UqtRkBKxknYhfvwyOoK+TStiMMCCbWdoN3EtP+06p6kSRKTk8a8Hj02DUbuhxRhw9oRrJ2DZWJgUAtFvQPxZW1f5wBSkRKyotLMDb3erw4IhTanmW4rLN1MZMW87g+Zu5dz1W7YuT0Sk4LkHQLs3Ycw+6PI+eFWB5HiI+RA+DIXvnoNz221dZZ4pSInkg4aVvPh5ZAtebFsdBzsDK/fH0X7iWuZuPIHJpLNTIlICObpB40EwYhs89TVUbAGmdNi9AKZHwqwuGA4utdzrrwhRkBLJJ072doxuX4OfR7YkLMiTxNQM3liylyembeTwxZI1z4qISCajEWp1gQE/w+A1UPdJSx/VyRjsF/al7f5X4OJeW1eZYwpSIvmshl9pFg5txtvdauPmaMe2k9fo8tF6Jq88REp68fsEi4hIjgU0gMc/gxd3QfNRmJ09cEq/AWUq2rqyHFOQEikARqOBvk0rET2mFW1r+ZKWYWbyysM88tFvbDt51dbliYjYlkd5aP8v0l/YyaYqY8CxlK0ryjEFKZECFODpwox+DfmkdwN8SjlyOO4mPT/dyBtL9nAjOc3W5YmI2JZjKa6WqmHrKnJFQUqkgBkMBh4JDWDlmFY8EV4BsxnmbjxJh0nr+HX/RVuXJyIiuaAgJWIjnq6OvPdEPb56rglBXq6cj09m4JytjJgXy6UbKbYuT0REckBBSsTGmlfzYcWoCIa0qoKd0cBPu87TbuJavt16WhN5iogUcgpSIoWAi6Md4zoHs2R4c2oHuBN/K42xC3fx7MzNnLySaOvyRETkLhSkRAqROuU9WDK8OeM618LJ3kjMkSt0nLyOaWuPkp5RtCapExEpCRSkRAoZezsjQ1pV5ZfRETSv5k1ymonxyw7QLSqGPWfjbV2eiIj8iYKUSCFV0duNLwc24b2eoXi4OLD3XALdomIYv3Q/t1I1kaeISGGgICVSiBkMBp5oGMjKMa14JNSfDJOZaeuO0XHyOmKOXLZ1eSIiJZ6ClEgRULa0E5/0DmNG34b4ezhz6moSz8zYzKuL9pCoeTxFRGzG3tYFiEjOtQvxo0kVL95fcZC5m07yXew5ltvbsTF1F2EVvWgQVIYQf3cc7fU3kohIQVCQEiliSjs78K9udXi0fgBjF+7i6KVEftx1gR93XQDA0d5I7QB3GgSWoUGQJ/UDPalQxgWDwWDjykVEih8FKZEiKryiFz8Ma8on367A0b8Gu84msOP0da4lpbH91HW2n7oOMZZ1fUo5ZYaqBkGehFbwpJST/vmLiDwo/SQVKcIc7Y3U8jTTpXVVHBwcMJvNnLySxPbT19hx6jrbT19n37kELt9MIXrfRaL3We7lZzRADb/SfwpXZahWthRGo85aiYjkhoKUSDFiMBio5ONGJR83ejSoAEByWgZ7z8VbzlKdvs6OU9c5e/0WBy7c4MCFG3y95TQApZ3sCQ30oEFgGeoHelI/yBOfUk623B0RkUJPQUqkmHN2sCO8ohfhFb0yl8UlJLP99PU/LgFeY/fZeG6kpBNz5AoxR65krhfk5Zp5ObBBUBmC/UvjZG9ni90QESmUFKRESiBfd2c61i5Hx9rlAEjPMHHo4k12nLYEq+2nr3Mk7ianriZx6moSP+w8B4CjnZHa5S2N7PWDPGmgRnYRKeEUpEQEezsjIQHuhAS407tJEADxt9LYdcZy1up2wLpzI7sj9f/4hGCDQE9CA9XILiIlh37aicgdebg40LJ6WVpWLwuQ2ch+O1TtOH2dvecSuHwzlZX7L7Jyv6WR3WCAmn6ls1wSrFq2FHZqZBeRYkhBSkRy5M+N7N0blAfu38j+ze+WRvZSTvbUC/SwhKs/LguqkV1EigMFKRHJs/s1su84fY1dZ+K5eYdG9kAvlyyThoYEuKuRXUSKHAUpEbGqOzWyH467mfkJwR2nr3M47ianr97i9NVbWRrZQwLcMy8HqpFdRIoCBSkRyVf2dkaC/d0J9s/eyH570tAdp69zNTGVHX/8/6yYE4Aa2UWk8NNPJBEpcHdqZD91NSnLJwT3nb9zI3sN37/MyO6rRnYRsR0FKRGxOYPBQEVvNyp6/7WRPSFzXqvbjewHL97g4MWsjeyhFTz+OGulRnYRKVgKUiJSKFka2csQXrFM5rK4G8mZlwO3n/r/RvYNR6+w4Wj2RvbbUzCokV1E8ouClIgUGb6lnelQuxwd/mhkzzCZOXTxRuYnBLefus6RS/duZK8f6ElYUBk1souIVShIiUiRZWc0ZGtkT0hOY9fp+MxPCG7/SyP7bZZGdkufVf1AT0IreOCsk1YikksKUiJSrLg7O9Ciug8tqvsA/9/IvuNPN2n+/0b2OFbujwMsjezVy5ailrOBlsnpeDk42HI3RKSIUJASkWLtz43s3epnb2S/HbDOXr/FobibHMKO1R+so2/TigxoXlmN6yJyTwpSIlLi3K2RfeXeC3y4Yg8Xb6UTtfooM9Yf58mGgQyOqEKgl6sNKxaRwkpBSkQESyP7E+HlcbmwE6fKDZn22wl2nr7OF5tOMm/LKbqG+jM0siq1yrnbulQRKUQUpERE/sRogPYhvnQODWDjsStMXXOU9Ycvs3jHORbvOEebWr4Mi6xKw0pe9x9MRIo9BSkRkTswGAw0q+pDs6o+7Dkbz9Q1R1m65zyrDsSx6kAcjSqV4fnIqrSu6atpFERKMAUpEZH7qFPeg6hnwjh+OZHp647y3baz/H7iGr/P3kqtcqUZ2qoqj4T6Y29ntHWpIlLA9K9eRCSHKvu4Mf6xUNa/0pohEVVwc7TjwIUbjJq/g8j31/DFxhMkp2XYukwRKUAKUiIiueTn7sy4LsFseLUtL3WogbebI2eu3eKfS/bS/L+riFp9hPhbabYuU0QKgIKUiEgeebg6MKJNdX57pQ1vd6tNeU8XriSm8t6KgzT/7yrGL9tPXEKyrcsUkXykICUi8oBcHO3o27QSa16OZFKvetTwK8XNlHSmrT1Gi/+tZtz3uzlxOdHWZYpIPlCQEhGxEgc7Iz0aVGD5ixHM7NeQ8IplSM0w8fWWU7T5YA0j5sWy52y8rcsUESvSp/ZERKzMaDTQNtiPtsF+bDl+lalrjrD64CV+2nWen3adJ6JGWZ5vVZWHqnhp6gSRIk5BSkQkHzWu7EXjyo3Zdy6BaeuO8uPOc6w7dIl1hy7RIMiT51tVpV2wH0ajApVIUaRLeyIiBSAkwJ0Pn2rAmpda8+xDQTjaG9l+6jqDv9hGh8nrWLjtDGkZJluXKSK5pCAlIlKAgrxdebd7XWJeacOwyKqUdrLnSNxNXlqwk1YTVjMr5jhJqem2LlNEckhBSkTEBsqWdmJsp1rEjGvDK51q4VPKiXPxyfzrx300/+8qPlx5mOtJqbYuU0Tuo0QGqR49elCmTBl69uxp61JEpIRzd3bg+ciq/PZKa/7dow5BXq5cS0pj0spDNPvvKt79aR/n42/ZukwRuYsSGaRGjhzJ3LlzbV2GiEgmZwc7nmlSkVV/b8VHTzcg2N+dpNQMZvx2nIgJqxm7cCdHL920dZki8hclMki1bt2a0qVL27oMEZFs7O2MPFovgKUjWzB7QCOaVPYiLcPMt1vP0G7iWoZ+sY2dp6/bukwR+UOhC1Lr1q2ja9euBAQEYDAYWLx4cbZ1pkyZQuXKlXF2diY8PJz169cXfKEiIvnIYDAQWdOX+UOa8t3zzWgX7IfZDMv3XqBbVAzPzNjEb4cvYzabbV2qSIlW6OaRSkxMpF69egwYMIDHH3882/Pz589n1KhRTJkyhebNmzNt2jQ6d+7Mvn37CAoKAiA8PJyUlJRsr/3ll18ICAjI930QEbGm8IplmNGvIYcu3uDTtUdZsuMcMUeuEHPkCqEVPHi+VVU61C6HneaiEilwhS5Ide7cmc6dO9/1+YkTJzJw4ECee+45ACZPnsyKFSuYOnUq48ePB2Dbtm1WqyclJSVLKEtISMj8/7Q0693d/fZY1hxTij+9b6yrsB/Pyl7O/K9HbUa2rsLMmJMs2HaGXWfief6rWCp7uzKoZSUerReAk/39LzYU9n29k8Jasy3rKuht5/f28mv8/Dw+BnMhPi9sMBhYtGgR3bt3ByA1NRVXV1cWLFhAjx49Mtd78cUX2bFjB2vXrs3x2GvWrOGTTz5h4cKF91zvrbfe4l//+le25fPmzcPV1TXH2xMRsbababDuvJF1FwzcyrCcjfJwNNPa30RTPzPOdjYuUKQQSUpKonfv3sTHx+Pu7m61cQvdGal7uXz5MhkZGfj5+WVZ7ufnx4ULF3I8TseOHYmNjSUxMZEKFSqwaNEiGjVqdMd1x40bx5gxYzK/TkhIIDAwEID27dvj4OCQhz3JLi0tjejoaKuOKcWf3jfWVRSP55PAzZR05m89w6yYk1y8kcLik3asjrPn2SZB9HkoCG83x2yvK4r7WlhrtmVdBb3t/N5efo2flpbGkiVLrDbenxWpIHXbX2/yaTabc3XjzxUrVuR4XScnJ5ycnO74nIODg9XfSPkxphR/et9YV1E7nmUcHBgaWZ0BLaqwePtZPl17jOOXE4lac4yZMSd4qlEQgyKqUN7TJdtri9q+QuGt2ZZ1FfS283t7hfV7fCeF7lN79+Lj44OdnV22s09xcXHZzlKJiJQ0TvZ29GoUxMoxrZjyTBh1y3uQnGZi9oYTtJqwmjHf7uDQxRu2LlOkWClSQcrR0ZHw8HCio6OzLI+OjqZZs2Y2qkpEpHCxMxroUtefH0Y058uBTWhezZt0k5nvY8/SYdI6Bs3dynbNRSViFYXu0t7Nmzc5cuRI5tfHjx9nx44deHl5ERQUxJgxY+jTpw8NGzakadOmTJ8+nVOnTjF06FAbVi0iUvgYDAZaVPehRXUfdpy+zqdrjrJi3wWi910ket9Fgj2N1GyUSK0AT1uXKlJkFbogtXXrVlq3bp359e1G7379+jF79mx69erFlStXePvttzl//jx16tRh6dKlVKxY0VYli4gUevUDPfm0TzhH4m4yfd1Rvo89y/7rRh7+ZAN9HqrIqHbV8XTN3pQuIvdW6IJUZGTkfWfqHTZsGMOGDSugikREio9qvqWY0LMezzWvyJi569lzzcjsDSdYtP0so9tV55mHKuJgV6S6PkRsSv9aRERKoMo+bgyqZWJ2/3Bq+pUm/lYab/24j06T17H6YJytyxMpMhSkRERKsOZVvfl5ZAv+3aMOXm6OHL2UyIBZv9P38y0c1if8RO5LQUpEpISztzPyTJOKrHk5ksERVXCwM7Du0CU6fbieN5bs4Wpiqq1LFCm0FKRERAQAd2cHXusSTPToVnQI8SPDZGbuxpNEvreaz387TlqGydYlihQ6ClIiIpJFJR83pvdtyLznmlCrXGkSktN5+6d9dJy8jlUHLt73A0EiJYmClIiI3FGzaj78PLIl4x+ri7ebI8cuJfK32Vvp+/kWzZAu8gcFKRERuSs7o4GnGwex+uVIhrSqgqOdkfWHL9Np8jr+uVj9UyIKUiIicl/uzg6M6xxM9JgIOtUuh8kMX2w6Sav3VjNj/TFS09U/JSWTgpSIiORYRW83Pu0TzteDHiLE350byem8+/N+Ok5ex8p96p+SkkdBSkREcq1pVW9+fKEF/3u8Lj6lHDl+OZHn5m6lz8wtHLiQYOvyRApMobtFTFGSlpZm9bGsOaYUf3rfWFdJOp7W2tfH6vvTvlZZpq07zucbTvDbkct0+XA9vRpW4MW21fB2s979+wrr98eWdRX0tvN7e/k1fn4eH4NZ52FzJCoqiqioKDIyMjh06BDz5s3D1dXV1mWJiBQal5Phx5NGdly1XOxwtjPTsYKJiHJm7HX9Q2wsKSmJ3r17Ex8fj7u7u9XGVZDKpYSEBDw8PJg3bx7dunXDwcHBKuOmpaURHR1N+/btrTamFH9631hXSTqe+bmvW05c5d9LD7LvvGWKhIperrzaqQZta5XFYDDkedzC+v2xZV0Fve383l5+jZ+WlsaSJUvyJUjp0t4DcHBwsPobKT/GlOJP7xvrKknHMz/2tXl1P358wZfvYs/w3oqDnLyaxPPzdtCsqjf/fCSEYP8H+yVWWL8/tqyroLed39srrN/jO9HJVhERsTo7o4EnGway+qVIhreuiqO9kQ1Hr/DwR+sZ9/1uLt9MsXWJIlahICUiIvmmlJM9L3esxa9jWvFwqD8mM3y95RSR761h2tqjpKRn2LpEkQeiICUiIvku0MuVqN5hLBjalLrlPbiZks74ZQdoP3Edy/dc0PxTUmQpSImISIFpVMmLJcOb8/4T9fAt7cSpq0kM/XIbT3+2ib3n4m1dnkiuKUiJiEiBMhoN9AyvwOqXInmhTTWc7I1sOnaVRz7+jVe/28WlG+qfkqJDQUpERGzCzcmev3eoya9/b0XXegGYzfDN76dp/f4apq45SnKa+qek8FOQEhERm6pQxpWPn27AwqFNqVfB0j/1v+UHaD9pLct2n1f/lBRqClIiIlIoNKzkxaJhzZn4ZD383J04ffUWz38VS6/pm9hzVv1TUjgpSImISKFhNBp4LMzSPzWybXWc7I1sOX6Vrp/8xrhFe0lItXWFIlkpSImISKHj6mjPmPY1WPVSJN3qW/qnFsae5Z3tdoz8ZicLt53RpJ5SKOgWMSIiUmiV93Thw6ca0LdpJf714x52nUlg2d6LLNt7EYMBQst70LqWL21q+VInwAOjMe/38hPJCwUpEREp9MIrlmHh4CZM/XYZqT41WHv4MnvOJrDzTDw7z8QzeeVhfEo5EVmzLG1q+dKiug/uzkXjXm1StClIiYhIkWAwGKhUGrq0rcbLnYKJS0hmzcFLrDoQx29HLnP5ZgoLt51h4bYz2BsNNKxUhja1fGld05dqvqUwGHS2SqxPQUpERIokX3dnnmwUyJONAklNN7H1xFVWHYhj1cE4jl1KZNOxq2w6dpX/LD1AhTIumaGqaVVvnB3sbF2+FBMKUiIiUuQ52htpVs2HZtV8+McjIZy8ksjqA3GsOniJTceucObaLeZuPMncjSdxdjDSrKoPrWv50rpmWSqUcbV1+VKEKUiJiEixU9Hbjf7NK9O/eWWSUtPZcOQKqw7GsfpAHOfjky1nrg7EAVDDr9QfocqX8IplcLDTB9ol5xSkRESkWHN1tKddiB/tQvwwm80cvHiDVQfiWHPgEttOXePQxZscuniTaWuPUdrZnogaZWld05fImmXxKeVk6/KlkFOQegBpaWlWH8uaY0rxp/eNdZWk41kU99VaNVf1dqFq84oMal6R+FtprD98mbWHLrP28GWuJaXx867z/LzrPAYD1C3vTmSNskTW8KG2v/sdp1ew5bEs6G3n9/bya/z8PD4Gs25ilCNRUVFERUWRkZHBoUOHmDdvHq6uuq4uIlJcmMxw6ibsu2Zk73UDZxKzhiZ3BzPBnmZCypip6WHGRaciipSkpCR69+5NfHw87u7uVhtXQSqXEhIS8PDwYN68eXTr1g0HB+vMU5KWlkZ0dDTt27e32phS/Ol9Y10l6XgWxX0t6JovJiSz7vBl1hy6TMyRKySmZmQ+Z2800LCiJ61qlKVFFU+Obt9Ahw4FfywL+pjk9/bya/y0tDSWLFmSL0FKefoBODg4WP2NlB9jSvGn9411laTjWRT3taBqruDtQG/v0vR+qDKp6SZ+P3H1j08C/jG9wvFrbDp+jf8BAa52pATE0T0sECf7gp9aoaC/j/m9vaL0vlSQEhERuQ9HeyPNq/nQ/C7TK5xLMvHK93v5IPoI/ZpV4pkmQXi6Otq6bCkA+oyniIhILt2eXmHu3xqzYWwrugZl4FfaibgbKby34iBNx6/ijSV7OHE50dalSj5TkBIREXkAHi4OtCtvZtWYlkzqVY8Qf3dupWUwd+NJWn+whsFzt7Ll+FXUklw86dKeiIiIFTjaG+nRoALd65dn49ErzPjtOKsOxPHLvov8su8i9Sp48FzLKnSuUw57TfpZbChIiYiIWJHBYMi8Xc2RuBvM/O0438WeZeeZeF74ejvlPV0Y0LwSvRoFUtq5aDRUy90pEouIiOSTar6lGf9YKBtebcOodtXxdnPk7PVbvPvzfpqNX8W/f97H2eu3bF2mPAAFKRERkXzmU8qJUe1qEPNqG/77WF2q+ZbiRko6n60/TsSE1bzw9XZ2nblu6zIlD3RpT0REpIA4O9jxVOMgnmwYyNpDl/hs/TE2HL3CjzvP8ePOczSu5MVzLSvTLtjvjrejkcJHQUpERKSAGY0GWtfypXUtX/aei2fm+uP8sPMcW05cZcuJq1T2ceNvLSrTM6wCLo4FP8Gn5Jwu7YmIiNhQ7QAPJvaqz2+vtOH5yKq4O9tz/HIi/1y8h6b//ZX3Vxwk7kayrcuUu1CQEhERKQTKeTjzSqdabBzXlre6hhDk5cr1pDQ+WX2EFv9dzUsLdnLgQoKty5S/0KU9ERGRQsTNyZ7+zSvTp2klovdd4LP1x9l28hoLt51h4bYztKzuw3MtqxBR3cfWpQoKUiIiIoWSndFApzr+dKrjT+ypa8xcf5xle86z/vBl1h++TE2/0vRvFoSjydaVlmwKUiIiIoVcWFAZwp4pw+mrScyKOcH8309x8OINxi3ai7uDHR41LtGudoCtyyyR1CMlIiJSRAR6ufJG1xA2jGvLuM61KOfuREKagaFf7WDx9rO2Lq9EUpASEREpYjxcHBjSqiq/jm5JuI+JdJOZUfN38Plvx21dWomjICUiIlJEOdobebaaiX5NgwB4+6d9vL/iIGaz2caVlRwKUiIiIkWY0QCvd67JSx1qAPDJ6iO8vngPGSaFqYKgICUiIlLEGQwGRrSpzr971MFggHmbTzFiXiwp6Rm2Lq3YU5ASEREpJp5pUpGo3mE42hlZtucCA2b9zs2UdFuXVawpSImIiBQjXer6M3tAI9wc7dhw9ApPT9/E5Zspti6r2NI8Ug8gLS3N6mNZc0wp/vS+sa6SdDyL4r4W1pptWdfdtt2oogdf/K0hA+fGsvtsPE9M3cCs/uGU93TJl+1ZS36Nn5/fG4NZrf05EhUVRVRUFBkZGRw6dIh58+bh6upq67JERETuKu4WTNlnx7VUAx4OZp4PycC/hP7qSkpKonfv3sTHx+Pu7m61cRWkcikhIQEPDw/mzZtHt27dcHBwsMq4aWlpREdH0759e6uNKcWf3jfWVZKOZ1Hc18Jasy3rysm2LyQkM2D2No5cSsTDxZ7Png2jQZBnvm3vQeTX+GlpaSxZsiRfgpQu7T0ABwcHq7+R8mNMKf70vrGuknQ8i+K+FtaabVnXvbYd6O3AwuebMWD272w/dZ1+s7cx9dkwImv65sv2HsTGY1dYdc5Al0L6Pb4TNZuLiIgUc56ujnz1XBNa1SjLrbQMnpuzlSU7Cs8tZY5duslzc7bSd9Y2fjxp5HDcTVuXlGMKUiIiIiWAq6M9n/VtSLf6AaSbzLz4zQ5mxdj2ljLXk1J5+8d9dJi0jpX7L2JnNNC8nBmfUo42rSs3dGlPRESkhHC0NzLpyfqUcXVk9oYT/OvHfVxNTGVM+xoYDIYCqyMtw8SXm07y4a+HuZ5k+URd65plGduhOoe2rqOMq4KUiIiIFEJGo4E3u4bg7ebIB9GH+HjVEa4kpvJOtzrYGfM3TJnNZlYdiOPfS/dz7FIiADX8SvGPh0OIqFGWtLQ0DuVrBdanICUiIlLCGAwGXmhbnTJujvxzyR7mbT7F9aRUJvWqj5O9Xb5s88CFBN79aT+/HbkMgLebI2M61KBXw0Ds7Ypup5GClIiISAn17EMVKePqyKj521m6+wLxt35nWp+GlHKyXjy4dCOFidGHmP/7KUxmcLQzMqBFJYa3roa7c9H4ZN69KEiJiIiUYA+H+uPh4sDgL7YSc+QKvT/bxKz+jfAu5fRA4yanZTAr5gRRq49k3u+vS91yvNopmCDv4jMraNE9lyYiIiJW0aK6D18PeggvN0d2nYnniU83cuZaUp7GMpvN/LTrHO0mruV/yw9wMyWduuU9+HZIU6Y8E16sQhQoSImIiAhQL9CTb4c0pbynC8cuJ9Jz6kYOXbyRqzF2nr7OE59uZMS87Zy5dgs/dyc+eKIeS4Y3p3Flr3yq3LYUpERERASAar6lWPh8U6r7luJCQjJPfLqR2FPX7vu68/G3GD1/B92iYth68houDnaMaled1S9F8nh4BYz5/GlAW1KQEhERkUz+Hi58O6QpDYI8ib+VxjOfbWbNwbg7rpuUms7E6EO0fn8Ni7ZbZkp/LKw8q1+KZFS7Grg6Fv9WbAUpERERyaKMm+WWMhF3uaWMyWRm4bYztH5/DR/9epjkNBONKpXhhxHNmfhkfcp5ONuw+oJV/KOiiIiI5Jqroz0z+jbkpQU7+WHnOUbN38HlLrW4ngCPTdvE3nOW/qlALxfGdQ6mc51yBTo7emGhICUiIiJ35GhvZHKv+pRxdWDOxpO88/MBLNHhBqWc7BnRphr9m1XC2SF/JvEsChSkRERE5K6MRgNvPVobLzcnJq08hAEzvRoF8lLHWvg84FxTxYGClIiIiNyTwWDgxXbVaVrZk52/b6D/oyE4OBT9WcmtQc3mIiIikiMNgjzxdbF1FYWLgpSIiIhIHilIiYiIiOSRgpSIiIhIHpXYIHX69GkiIyMJCQkhNDSUBQsW2LokERERKWJK7Kf27O3tmTx5MvXr1ycuLo6wsDC6dOmCm5ubrUsTERGRIqLEBil/f3/8/f0B8PX1xcvLi6tXrypIiYiISI7l6dLe2bNnefbZZ/H29sbV1ZX69euzbds2qxW1bt06unbtSkBAAAaDgcWLF99xvSlTplC5cmWcnZ0JDw9n/fr1edre1q1bMZlMBAYGPkDVIiIiUtLk+ozUtWvXaN68Oa1bt2bZsmX4+vpy9OhRPD0977h+TEwMjRs3zjZx14EDB/D09KRcuXLZXpOYmEi9evUYMGAAjz/++B3HnT9/PqNGjWLKlCk0b96cadOm0blzZ/bt20dQUBAA4eHhpKSkZHvtL7/8QkBAAABXrlyhb9++zJgxIzeHAYC0tLRcv+Z+Y1lzTCn+9L6xrpJ0PIvivhbWmm1ZV0FvO7+3l1/j5+fxMZjNZnNuXvDqq68SExOTo7M/JpOJsLAwqlevzjfffIOdneVePIcOHaJVq1aMHj2asWPH3rtAg4FFixbRvXv3LMubNGlCWFgYU6dOzVwWHBxM9+7dGT9+fI72JSUlhfbt2zNo0CD69Olzz3WjoqKIiooiIyODQ4cOMW/ePFxdXXO0HREREbGtpKQkevfuTXx8PO7u7lYbN9dnpH744Qc6duzIE088wdq1aylfvjzDhg1j0KBB2dY1Go0sXbqUiIgI+vbtyxdffMHx48dp06YNjz766H1D1N2kpqaybds2Xn311SzLO3TowIYNG3I0htlspn///rRp0+a+IQpg+PDhDB8+nISEBDw8PABo37691abIT0tLIzo62qpjSvGn9411laTjWRT3tbDWbMu6Cnrb+b29/Bo/LS2NJUuWWG28P8t1kDp27BhTp05lzJgxvPbaa2zZsoWRI0fi5ORE3759s60fEBDAqlWriIiIoHfv3mzcuJG2bdvy6aef5rnoy5cvk5GRgZ+fX5blfn5+XLhwIUdjxMTEMH/+fEJDQzN7sL744gvq1q2b4zocHBys/kbKjzGl+NP7xrpK0vEsivtaWGu2ZV0Fve383l5h/R7fSa6DlMlkomHDhvznP/8BoEGDBuzdu5epU6feMUgBBAUFMXfuXFq1akWVKlWYOXMmBoPhwSqHbGOYzeYcj9uiRQtMJtMD1yAiIiIlV64/tefv709ISEiWZcHBwZw6dequr7l48SKDBw+ma9euJCUlMXr06NxX+ic+Pj7Y2dllO/sUFxeX7SyViIiISH7J9Rmp5s2bc/DgwSzLDh06RMWKFe+4/uXLl2nbti3BwcEsWLCAw4cPExkZiZOTE++//36einZ0dCQ8PJzo6Gh69OiRuTw6Oppu3brlacycut2bn5SUREJCglV7pKw9phR/et9YV0k6nkVxXwtrzbasq6C3nd/by6/xb48L//973GrMubRlyxazvb29+d///rf58OHD5q+++srs6upq/vLLL7Otm5GRYQ4PDzd36dLFnJKSkrl8165dZm9vb/PEiRPvuI0bN26Yt2/fbt6+fbsZME+cONG8fft288mTJzPX+eabb8wODg7mmTNnmvft22ceNWqU2c3NzXzixInc7lKunD592gzooYceeuihhx5F8HH69Gmr5oJcT38A8NNPPzFu3DgOHz5M5cqVGTNmzB0/tQeWs0QtW7bE2dk5y/IdO3bg7e19x0kw16xZQ+vWrbMt79evH7Nnz878esqUKUyYMIHz589Tp04dJk2aRERERG53J1dMJhPnzp2jTZs2bN261WrjJiQkEBgYyOnTp636sUwp/ho1asTvv/9u6zKKjZJ0PIvivhbWmm1ZV0FvO7+3lx/j3/4du2/fPmrWrInRaL1bDefpFjGPPPIIjzzySI7Wbd++/R2X169f/66viYyMzNGpt2HDhjFs2LAc1WEtRqORChUqYG9vny+Bx93dXUFKcsXOzk7vGSsqScezKO5rYa3ZlnUV9Lbze3v5OX758uWtGqIgj7eIEcu8UiKFgd6L1lWSjmdR3NfCWrMt6yrobef39grr9/hu8nRpT6zv9kSf1p5xVUREpKTLz9+xOiNVSDg5OfHmm2/i5ORk61JERESKlfz8HaszUiIiIiJ5pDNSIiIiInmkICUiIiKSRwpSIiIiInmkICUiIiKSRwpSIiIiInmkIFUEnD59msjISEJCQggNDWXBggW2LklERKTIu3HjBo0aNaJ+/frUrVuXzz77LNdjaPqDIuD8+fNcvHiR+vXrExcXR1hYGAcPHsTNzc3WpYmIiBRZGRkZpKSk4OrqSlJSEnXq1OH333/H29s7x2Pk6V57UrD8/f3x9/cHwNfXFy8vL65evaogJSIi8gDs7OxwdXUFIDk5mYyMjBzd6/fPdGmvAKxbt46uXbsSEBCAwWBg8eLF2daZMmUKlStXxtnZmfDwcNavX3/HsbZu3YrJZCIwMDCfqxYRESncrPH79fr169SrV48KFSowduxYfHx8clWDglQBSExMpF69enzyySd3fH7+/PmMGjWK119/ne3bt9OyZUs6d+7MqVOnsqx35coV+vbty/Tp0wuibBERkULNGr9fPT092blzJ8ePH2fevHlcvHgxVzWoR6qAGQwGFi1aRPfu3TOXNWnShLCwMKZOnZq5LDg4mO7duzN+/HgAUlJSaN++PYMGDaJPnz4FXbaIiEihltffr3/2/PPP06ZNG5544okcb1dnpGwsNTWVbdu20aFDhyzLO3TowIYNGwAwm83079+fNm3aKESJiIjkQE5+v168eJGEhAQAEhISWLduHTVr1szVdtRsbmOXL18mIyMDPz+/LMv9/Py4cOECADExMcyfP5/Q0NDM679ffPEFdevWLehyRUREioSc/H49c+YMAwcOxGw2YzabGTFiBKGhobnajoJUIWEwGLJ8bTabM5e1aNECk8lki7JERESKtHv9fg0PD2fHjh0PNL4u7dmYj48PdnZ2men4tri4uGwpWkRERHKmoH6/KkjZmKOjI+Hh4URHR2dZHh0dTbNmzWxUlYiISNFWUL9fdWmvANy8eZMjR45kfn38+HF27NiBl5cXQUFBjBkzhj59+tCwYUOaNm3K9OnTOXXqFEOHDrVh1SIiIoVbYfj9qukPCsCaNWto3bp1tuX9+vVj9uzZgGXCsAkTJnD+/Hnq1KnDpEmTiIiIKOBKRUREio7C8PtVQUpEREQkj9QjJSIiIpJHClIiIiIieaQgJSIiIpJHClIiIiIieaQgJSIiIpJHClIiIiIieaQgJSIiIpJHClIiIiIieaQgJSIiIpJHClIiIiIieaQgJSIiIpJHClIiIiIieWRv6wJERGwpMjKS0NBQnJ2dmTFjBo6OjgwdOpS33nrL1qWJSBGgM1IiUuLNmTMHNzc3Nm/ezIQJE3j77beJjo62dVkiUgQYzGaz2dZFiIjYSmRkJBkZGaxfvz5zWePGjWnTpg3//e9/bViZiBQFOiMlIiVeaGholq/9/f2Ji4uzUTUiUpQoSIlIiefg4JDla4PBgMlkslE1IlKUKEiJiIiI5JGClIiIiEgeKUiJiIiI5JE+tSciIiKSRzojJSIiIpJHClIiIiIieaQgJSIiIpJHClIiIiIieaQgJSIiIpJHClIiIiIieaQgJSIiIpJHClIiIiIieaQgJSIiIpJHClIiIiIieaQgJSIiIpJH/wd1FmSq5ftVvAAAAABJRU5ErkJggg==",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sigma = 1\n",
    "L = 1\n",
    "d = 2\n",
    "n_values = np.arange(100, 1000, 100)\n",
    "N = 5000\n",
    "\n",
    "t0 = perf_counter()\n",
    "L2emp = []\n",
    "\n",
    "for n in n_values:\n",
    "    k = Kopt(n, d, L, sigma)\n",
    "    W = np.zeros(n)\n",
    "    W[:k] = 1/k\n",
    "    Xnew = np.random.rand(N,d)\n",
    "    error = 0\n",
    "    \n",
    "    for j in range(N):\n",
    "        X = np.random.rand(n,d)\n",
    "        Xtest = Xnew[j,:][np.newaxis,:]\n",
    "        ordre = order(X, Xtest)\n",
    "        Y = r(X, L) + sigma * np.random.randn(n)\n",
    "        rn = np.sum(W*Y[ordre]) \n",
    "        error += (rn-r(Xtest, L))**2\n",
    "        \n",
    "    L2emp.append(error/N)\n",
    "\n",
    "\n",
    "print(f\"Duration (s) : {perf_counter()-t0:1.1f}\")\n",
    "    \n",
    "plt.plot(n_values, L2emp, label=\"Empirical $L_2$ Error\")\n",
    "plt.plot(n_values, L2theo(n_values, d, L, sigma), label=\"Theoretical Bound\")\n",
    "plt.xlabel('n')\n",
    "plt.title('$L_2$ error versus n')\n",
    "plt.legend(loc=\"best\")\n",
    "plt.xscale(\"log\")\n",
    "plt.yscale(\"log\")\n",
    "plt.grid(which=\"both\");"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2.2. $L_2$ risk as a fonction of $k$"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In this section the size sample $n$ is held fixed (but large enough) and we let $k$ vary between $1$ and $n$. We know that:\n",
    "- For $d=1$,\n",
    "$$\n",
    "\\mathbb{E}\\left[(r_n(\\mathbf{X})-r(\\mathbf{X}))^2\\right]\\leq\\frac{\\sigma^2}{k}+2L^2\\left(\\frac{k}{n}\\right).\\hspace{1cm}(3)\n",
    "$$\n",
    "- For $d\\geq 2$,\n",
    "$$\n",
    "\\mathbb{E}\\left[(r_n(\\mathbf{X})-r(\\mathbf{X}))^2\\right]\\leq\\frac{\\sigma^2}{k}+4L^2\\left(\\frac{k}{n}\\right)^{\\frac{2}{d}}.\\hspace{1cm}(4)\n",
    "$$"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    ">**Question 2.**\n",
    "Represent on the same plot the empirical risk\n",
    "$$\n",
    "\\frac{1}{N}\\sum_{j=1}^{N}\\left(r_n^{(j)}(\\mathbf{X}^{(j)})-r(\\mathbf{X}^{(j)})\\right)^2\n",
    "$$\n",
    "as a function of $k$ between $1$ and $n$, and the theoretical upper bound given by (3) or (4), depending on dimension $d$. \n",
    ">"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [],
   "source": [
    "def BV(n, k_values, d, L, sigma):\n",
    "    if d==1:\n",
    "        return sigma**2/k_values + 2 * L**2 * k_values/n\n",
    "    return sigma**2/k_values + 4 * L**2 * (k_values/n)**(2/d)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Duration (s) : 15.7\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "sigma = 1\n",
    "L = 1\n",
    "d = 2\n",
    "n = 500\n",
    "k_values = np.arange(1, n, 10)\n",
    "N = 5000\n",
    "\n",
    "t0 = perf_counter()\n",
    "L2emp = []\n",
    "\n",
    "for k in k_values:\n",
    "    W = np.zeros(n)\n",
    "    W[:k] = 1/k\n",
    "    Xnew = np.random.rand(N,d)\n",
    "    error = 0\n",
    "    \n",
    "    for j in range(N):\n",
    "        X = np.random.rand(n,d)\n",
    "        Xtest = Xnew[j,:][np.newaxis,:]\n",
    "        ordre = order(X, Xtest)\n",
    "        Y = r(X, L) + sigma * np.random.randn(n)\n",
    "        rn = np.sum(W*Y[ordre])\n",
    "        error += (rn-r(Xtest, L))**2\n",
    "        \n",
    "    L2emp.append(error/N)\n",
    "\n",
    "print(f\"Duration (s) : {perf_counter()-t0:1.1f}\")\n",
    "\n",
    "plt.plot(k_values, L2emp, label=\"Empirical Risk\")\n",
    "k_values = np.arange(1, n)\n",
    "plt.plot(k_values, BV(n, k_values, d, L, sigma), label=\"Theoretical Bound\")\n",
    "plt.xlabel('k')\n",
    "plt.title('Bias-Variance Tradeoff')\n",
    "plt.legend(loc=\"best\")\n",
    "plt.xscale(\"log\")\n",
    "plt.yscale(\"log\")\n",
    "plt.grid(which=\"both\");"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2.3. Choice of $k$ by data splitting"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "To choose $k$ automatically, we split the sample $\\mathcal D_n=\\{(\\mathbf{X}_1,Y_1), \\ldots, (\\mathbf{X}_n,Y_n)\\}$ in two parts of size $N:=n/2 $. The first half is denoted by $\\mathcal D_{n}^{\\ell}$ (learning set) and is used to construct the $k$-NN estimates $r_{N}(\\mathbf{x}, \\mathcal D_n^{\\ell})=r_{k,N}(\\mathbf{x},\\mathcal D_n^{\\ell})$ for each $k \\in \\{1, \\ldots, N\\}$. The second half of the sample, denoted by $\\mathcal D_{n}^{t}$ (testing set), is used to choose $k$ by picking $\\hat k_n\\in \\{1, \\ldots, N \\}$ to minimize the empirical risk, that is\n",
    "$$\\hat k_n:=\\arg\\min_{\\{1, \\ldots, N \\}}\\frac{1}{N} \\sum_{i=N+1}^{n}\\left (Y_i-r_{k,N}(\\mathbf{X}_i)\\right)^2.$$\n",
    "We then define the estimate\n",
    "$$r_n(\\mathbf{x})=r_{\\hat k_n,N}(\\mathbf{x}, \\mathcal D_n^{\\ell}).$$\n",
    "Let $d\\geq 2$, suppose that $\\mathbf{X}$ has bounded support, $|Y|\\leq M$, and that $r$ is $L$-Lipschitz. Let $r_n$ be the $k$-NN estimate with $k \\in \\{1, \\ldots, N\\}$ chosen by data-splitting. If $(\\ln n)^{(d+2)/(2d)} n^{-1/2} \\leq L$, then \n",
    "$$\\mathbb E \\left[\\left(r_{n}(\\mathbf{X})-r(\\mathbf{X})\\right)^2\\right]\\leq \\Lambda\\left(\\frac{L^d}{n}\\right)^{\\frac{2}{d+2}},$$\n",
    "for some positive constant $\\Lambda$ which depends only on $M$, $d$, and the diameter of the support of $\\mathbf{X}$. \n",
    "\n",
    "To illustrate this result, we consider\n",
    "$$Y=r(\\mathbf{X})+\\varepsilon,$$\n",
    "where $\\mathbf{X}\\sim\\mathcal{U}_{[0,1]^d}$ is independent of $\\varepsilon\\sim\\mathcal{U}_{[-M,M]}$, and  \n",
    "$$\n",
    "r(x)=L \\sin\\|x\\|_\\infty.\n",
    "$$\n",
    "In order to compare with the optimal choice of $k$ when everything is known, we thus have $\\sigma=M/\\sqrt{3}$."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    ">**Question 3.**\n",
    "Represent on the same plot the empirical risk where $k$ is chosen by data splitting and the \n",
    "empirical risk where $k_n\\sim\\frac{\\sigma\\sqrt{n}}{\\sqrt{2}L}$ or $k_n\\sim\\left(\\frac{\\sigma^2}{4L^2}\\right)^{\\frac{d}{d+2}}n^{\\frac{2}{d+2}}$, depending on dimension $d$. \n",
    ">"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Duration (s) : 34.3\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "L = 1\n",
    "M = 1\n",
    "sigma = M/np.sqrt(3)\n",
    "d = 2\n",
    "max_of_n_over_2 = 500\n",
    "n_values = np.arange(50, max_of_n_over_2, 50)\n",
    "n_tot = 2 * n_values\n",
    "Nds = 10 # Number of data sets\n",
    "N = 10000 # Number of simulations to evaluate the empirical risk\n",
    "\n",
    "t0 = perf_counter()\n",
    "L2splitting = []\n",
    "\n",
    "for n in n_values:\n",
    "    \n",
    "    error = 0\n",
    "    \n",
    "    for m in range(Nds):\n",
    "        Xl = np.random.rand(n,d)\n",
    "        Xt = np.random.rand(n,d)\n",
    "        ordre = order(Xl, Xt) \n",
    "        Yl = r(Xl, L) + M *(2 * np.random.rand(n) - 1)\n",
    "        Yt = r(Xt, L) + M *(2 * np.random.rand(n) - 1)\n",
    "        mc = np.zeros(n)\n",
    "\n",
    "        for k in np.arange(1, n+1):\n",
    "            W = np.zeros(n)\n",
    "            W[:k] = 1/k         \n",
    "            for j in range(n): \n",
    "                mc[k-1] += (Yt[j] - np.sum(W * Yl[ordre[j,]]))**2\n",
    "\n",
    "        khat = 1 + np.argmin(mc)\n",
    "        What = np.zeros(n) \n",
    "        What[:khat] = 1/khat\n",
    "        Xnew = np.random.rand(N,d)\n",
    "        ordre = order(Xl, Xnew)\n",
    "        rnhat = np.sum(What*Yl[ordre], axis=1) \n",
    "        error += np.sum((rnhat-r(Xnew, L))**2)\n",
    "        \n",
    "    L2splitting.append(error/(Nds * N))\n",
    "    \n",
    "L2emp = []\n",
    "\n",
    "for n in n_tot:\n",
    "    \n",
    "    k = Kopt(n, d, L, sigma)\n",
    "    W = np.zeros(n)\n",
    "    W[:k] = 1/k\n",
    "    Xnew = np.random.rand(N,d)\n",
    "    error = 0\n",
    "    \n",
    "    for j in range(N):\n",
    "        X = np.random.rand(n,d)\n",
    "        Y = r(X, L) +  M * (2 * np.random.rand(n) - 1)\n",
    "        Xtest = Xnew[j,:][np.newaxis,:]\n",
    "        ordre = order(X, Xtest)\n",
    "        rn = np.sum(W*Y[ordre]) \n",
    "        error += (rn-r(Xtest, L))**2\n",
    "        \n",
    "    L2emp.append(error/N)\n",
    "    \n",
    "print(f\"Duration (s) : {perf_counter()-t0:1.1f}\")\n",
    "\n",
    "plt.plot(n_tot, L2splitting, label=\"Data splitting\")\n",
    "plt.plot(n_tot, L2emp, label=\"Theoretical choice of $k$\")\n",
    "plt.plot(n_tot, L2theo(n_tot, d, L, sigma), label=\"Theoretical bound\")\n",
    "plt.xlabel('n')\n",
    "plt.title('$L_2$ Error')\n",
    "plt.legend(loc=\"best\")\n",
    "plt.xscale(\"log\")\n",
    "plt.yscale(\"log\")\n",
    "plt.grid(which=\"both\");         "
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python [conda env:base] *",
   "language": "python",
   "name": "conda-base-py"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}
