ch05_Machine_Learning.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Chapter 5: Supervised learning\n",
    "\n",
    "This is the notebook companion to chapter 7. We use data from the Lending Club to develop our understanding of machine learning concepts. The Lending Club is a peer-to-peer lending company. It offers loans which are funded by other people. In this sense, the Lending Club acts as a hub connecting borrowers with investors. The client applies for a loan of a certain amount, and the company assesses the risk of the operation. If the application is accepted, it may or may not be fully covered. We will focus on the prediction of whether the loan will be fully funded, based on the scoring of and information related to the application. The data come from the following URL:\n",
    "\n",
    "https://www.lendingclub.com/info/download-data.action\n",
    "\n",
    "We will clean and reprocess the partial dataset from 2007-2011. Framing the problem a little bit more, based on the information supplied by the customer asking for a loan, we want to predict whether an accepted loan will be fully funded or not."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "import matplotlib.pylab as plt\n",
    "\n",
    "%matplotlib inline \n",
    "plt.style.use('seaborn-whitegrid')\n",
    "plt.rc('text', usetex=True)\n",
    "plt.rc('font', family='times')\n",
    "plt.rc('xtick', labelsize=10) \n",
    "plt.rc('ytick', labelsize=10) \n",
    "plt.rc('font', size=12) \n",
    "plt.rc('figure', figsize = (12, 5))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "import pickle\n",
    "ofname = open('files/ch05/dataset_small.pkl','rb') \n",
    "(x,y) = pickle.load(ofname)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "and check the shapes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dims: 15, samples: 4140\n"
     ]
    }
   ],
   "source": [
    "dims = x.shape[1]\n",
    "N = x.shape[0]\n",
    "print 'dims: ' + str(dims)+', samples: '+ str(N)"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "Let us fit a simple classifier."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Predicted value: -1.0 , real target: -1.0\n"
     ]
    }
   ],
   "source": [
    "from sklearn import neighbors\n",
    "from sklearn import datasets\n",
    "#Create an instance of K-nearest neighbor classifier\n",
    "knn = neighbors.KNeighborsClassifier(n_neighbors=11)\n",
    "#Train the classifier\n",
    "knn.fit(x,y)\n",
    "#Compute the prediction according to the model\n",
    "yhat = knn.predict(x)\n",
    "print 'Predicted value: ' + str(yhat[-1]), ', real target: ' + str(y[-1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "And check its score/accuracy."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.83164251207729467"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "knn.score(x,y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The distribution of the labels to predict is shown in the following pie chart."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYkAAAFdCAYAAAD2cpGEAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nMwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10ec7f510>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#%matplotlib inline\n",
    "#import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "plt.pie(np.c_[np.sum(np.where(y==1,1,0)),np.sum(np.where(y==-1,1,0))][0],\n",
    "        labels=['Not fully funded','Full amount'],\n",
    "        colors=['g','r'],\n",
    "        shadow=False,\n",
    "        autopct ='%.2f' )\n",
    "plt.gcf().set_size_inches((6,6))\n",
    "plt.savefig(\"pie.png\",dpi=300, bbox_inches='tight')\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let us compute the elements of a confusion matrix."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TP: 3370 , FP: 690\n",
      "FN: 7 , TN: 73\n"
     ]
    }
   ],
   "source": [
    "yhat = knn.predict(x)\n",
    "TP = np.sum(np.logical_and(yhat==-1,y==-1))\n",
    "TN = np.sum(np.logical_and(yhat==1,y==1))\n",
    "FP = np.sum(np.logical_and(yhat==-1,y==1))\n",
    "FN = np.sum(np.logical_and(yhat==1,y==-1))\n",
    "print 'TP: '+ str(TP), ', FP: '+ str(FP)\n",
    "print 'FN: '+ str(FN), ', TN: '+ str(TN)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[3370,  690],\n",
       "       [   7,   73]])"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn import metrics\n",
    "metrics.confusion_matrix(yhat,y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "classification accuracy: 1.0\n",
      "confusion matrix: \n",
      "[[3377    0]\n",
      " [   0  763]]\n"
     ]
    }
   ],
   "source": [
    "#Train a classifier using .fit()\n",
    "from sklearn import neighbors\n",
    "knn = neighbors.KNeighborsClassifier(n_neighbors=1)\n",
    "knn.fit(x,y)\n",
    "yhat=knn.predict(x)\n",
    "\n",
    "print \"classification accuracy:\", metrics.accuracy_score(yhat, y)\n",
    "print \"confusion matrix: \\n\" + str(metrics.confusion_matrix(yhat, y))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Up to this point we used training data for \"assessing\" the performance of the method, as we will see later this is a bad practice. Let us simulate the exploitation stage by holding out a subset of the training data and assess the performance on that set. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training shape: (2898, 15)  , training targets shape: (2898,)\n",
      "Testing shape: (1242, 15)  , testing targets shape: (1242,)\n"
     ]
    }
   ],
   "source": [
    "# Simulate a real case: Randomize and split data in two subsets PRC*100% for training and \n",
    "# the rest (1-PRC)*100% for testing\n",
    "import numpy as np\n",
    "perm = np.random.permutation(y.size)\n",
    "PRC = 0.7\n",
    "split_point = int(np.ceil(y.shape[0]*PRC))\n",
    "\n",
    "X_train = x[perm[:split_point].ravel(),:]\n",
    "y_train = y[perm[:split_point].ravel()]\n",
    "\n",
    "X_test = x[perm[split_point:].ravel(),:]\n",
    "y_test = y[perm[split_point:].ravel()]\n",
    "\n",
    "print 'Training shape: ' + str(X_train.shape), ' , training targets shape: '+str(y_train.shape)\n",
    "print 'Testing shape: ' + str(X_test.shape), ' , testing targets shape: '+str(y_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "TRAINING STATS:\n",
      "classification accuracy: 1.0\n",
      "confusion matrix: \n",
      "[[2348    0]\n",
      " [   0  550]]\n"
     ]
    }
   ],
   "source": [
    "#Train a classifier on training data\n",
    "from sklearn import neighbors\n",
    "knn = neighbors.KNeighborsClassifier(n_neighbors=1)\n",
    "knn.fit(X_train,y_train)\n",
    "\n",
    "#Check on the training set and visualize performance\n",
    "yhat=knn.predict(X_train)\n",
    "\n",
    "from sklearn import metrics\n",
    "print \"\\nTRAINING STATS:\"\n",
    "print \"classification accuracy:\", metrics.accuracy_score(yhat, y_train)\n",
    "print \"confusion matrix: \\n\"+ str(metrics.confusion_matrix(y_train, yhat))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TESTING STATS:\n",
      "classification accuracy: 0.759259259259\n",
      "confusion matrix: \n",
      "[[884 154]\n",
      " [145  59]]\n"
     ]
    }
   ],
   "source": [
    "#Check on the test set\n",
    "yhat=knn.predict(X_test)\n",
    "print \"TESTING STATS:\"\n",
    "print \"classification accuracy:\", metrics.accuracy_score(yhat, y_test)\n",
    "print \"confusion matrix: \\n\"+ str(metrics.confusion_matrix(yhat,y_test))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can automatize this process with the tools provided in sklearn."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean expected error: 0.751690821256\n"
     ]
    }
   ],
   "source": [
    "#The splitting can be done using the tools provided by sklearn:\n",
    "from sklearn.cross_validation import train_test_split\n",
    "from sklearn import neighbors\n",
    "from sklearn import metrics\n",
    "\n",
    "PRC = 0.3\n",
    "acc = np.zeros((10,))\n",
    "for i in xrange(10):\n",
    "    X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=PRC)\n",
    "    knn = neighbors.KNeighborsClassifier(n_neighbors=1)\n",
    "    knn.fit(X_train,y_train)\n",
    "    yhat = knn.predict(X_test)\n",
    "    acc[i] = metrics.accuracy_score(yhat, y_test)\n",
    "acc.shape=(1,10)\n",
    "print \"Mean expected error: \"+str(np.mean(acc[0]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can use the validation process for model selection."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAtEAAAE3CAYAAABsPB0FAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n0QAAAECRSKIBAACAIv1/XSKzhNc9Ht0AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x112d454d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#The splitting can be done using the tools provided by sklearn:\n",
    "from sklearn.cross_validation import train_test_split\n",
    "from sklearn import neighbors\n",
    "from sklearn import tree\n",
    "from sklearn import svm\n",
    "from sklearn import metrics\n",
    "\n",
    "PRC = 0.1\n",
    "acc_r=np.zeros((10,4))\n",
    "for i in xrange(10):\n",
    "    X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=PRC)\n",
    "    nn1 = neighbors.KNeighborsClassifier(n_neighbors=1)\n",
    "    nn3 = neighbors.KNeighborsClassifier(n_neighbors=3)\n",
    "    svc = svm.SVC()\n",
    "    dt = tree.DecisionTreeClassifier()\n",
    "    \n",
    "    nn1.fit(X_train,y_train)\n",
    "    nn3.fit(X_train,y_train)\n",
    "    svc.fit(X_train,y_train)\n",
    "    dt.fit(X_train,y_train)\n",
    "    \n",
    "    yhat_nn1=nn1.predict(X_test)\n",
    "    yhat_nn3=nn3.predict(X_test)\n",
    "    yhat_svc=svc.predict(X_test)\n",
    "    yhat_dt=dt.predict(X_test)\n",
    "    \n",
    "    acc_r[i][0] = metrics.accuracy_score(yhat_nn1, y_test)\n",
    "    acc_r[i][1] = metrics.accuracy_score(yhat_nn3, y_test)\n",
    "    acc_r[i][2] = metrics.accuracy_score(yhat_svc, y_test)\n",
    "    acc_r[i][3] = metrics.accuracy_score(yhat_dt, y_test)\n",
    "\n",
    "\n",
    "plt.boxplot(acc_r);\n",
    "for i in xrange(4):\n",
    "    xderiv = (i+1)*np.ones(acc_r[:,i].shape)+(np.random.rand(10,)-0.5)*0.1\n",
    "    plt.plot(xderiv,acc_r[:,i],'ro',alpha=0.3)\n",
    "    \n",
    "ax = plt.gca()\n",
    "ax.set_xticklabels(['1-NN','3-NN','SVM','Decission Tree'])\n",
    "plt.ylabel('Accuracy')\n",
    "plt.savefig(\"error_ms_1.png\",dpi=300, bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Learning curves\n",
    "\n",
    "Let us try to understand the behavior of machine learning algorithms when the amount of data and the \"complexity\" of the method change. Let us start first by varying the amount of data for a fixed complexity."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXIAAAE3CAYAAAC3scrWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzsvXlwG+eZ7vsD0NjRJMFNpCRqoRbIkmVJli3ZkWx6k21N\nxMSxY0+2M0lmX05NTc2cU+dmpqZSqVO5s99zz51z/5rUmeUkdzzHsZ2JlMi2vMmWLFv7LoJaSIri\nvoBEY0cDuH+8bAEEQYkUKYuU+qliASAa3V9/3f187/e8y2fJ5XKYMGHChIn5C+udboAJEyZMmJgZ\nTCI3YcKEiXkOk8hNmDBhYp7DJHITJkyYmOcwidyECRMm5jmU6f4gEAhsCgaDJwo+/9bY2xXBYPD/\nmLWWmTBhwoSJKWFaFnkgEHgaeK3o875gMPgPQGMgEHhqlttnwoQJEyZugmkReTAYfA+4XPCvRuCZ\nsfdXxj6bMGHChInPEdOWVgoxZokbeBB4dWbNMWHChAkT08WsODsDgcAm4FgwGDw5G/szYcKECRNT\nx2xFrTwdDAa/N0v7MmHChAkT08CtSCuWwg+BQOC3gsHg3469f3pMR7+OY8eOmcVcTJgwYeIWsHnz\nZsvNt5omkQcCgZeAzYFA4MVgMPjGWNTKXwYCgf8C+IGXJ2nMdA5z29Hd3c3ChQvvdDPGYS62CeZm\nu8w2TQ1mm6aOudiuY8eOTXnbaRF5MBh8HXi94PN7QNV09mHChAkTJmYXZmanCRMmTMxzmERuwoQJ\nE/McJpGbMGHCxDyHSeQmTJgwMc9hErkJEyZMzHOYRG7ChAkT8xwmkZswYcLEPIdJ5CZMmDAxz2ES\nuQkTJkzMc5hEbsKECRPzHCaRmzBhwsQ8h0nkJkyYMDHPYRK5CRMmTMxzmERuwoSJW4euQygkrybu\nGGa0ZqcJEybuYeg67N4NsRh4PNDcDIpJKXcCpkVuwoSJW4OmCYmXlcmrpt3pFt2zMInchAkTtwZV\nFUs8HJZXVb3TLbpnYc6DTJgwcWtQFJFTNE1I3JRV7hjMnjdhwsStQ1HA77/TrbjnYUorJkyYMDHP\nYRK5CRMmTMxzmERuwoQJE/Mc0ybyQCCwqejzS4FA4OlAIPCfZ69ZJkyYMGFiqpgWkQcCgaeB1wo+\nbwJywWDwPWAkEAhsnOX2mTBhwoSJm2BaRD5G2JcL/vWrwMjY+yvAM7PULhMmTJgwMUXcikZuKXhf\nAQwXfK6aWXNMmDDxuWO+10uZ7+2fBZhx5CZM3MuY7/VS5nv7Zwm3YpHnCt6HgMqx9xXA0IxbZMKE\nic8Pn1e9lNtlNZv1XoBbs8gLpZX/DWwG3gcagX2lftDd3X0Lh7l90DTNbNMUMRfbFQppjIz04PPl\nJhhfug6RiKXkd7cTc7GfptQmXceZSGAJhci53SQ1DeLx2W2IruPctw9LPE4W6P7Sl2bPap6l9s/F\n6zcdTKs3A4HAS8DmQCDwYjAYfCMYDJ4IBAKbx6JZQsFg8GSp3y1cuHA22jpr6O7uNts0Rcy1duk6\n7N07hMtVNWEmPZ1Ztq7PbomQudZPMI02ffvbt7deSigELhfU1kJHB5WqOrtp/bPQ/rl4/Xp6eqa8\n7bTOOhgMvg68XvS/H01nHyZMzARicFmorZWie5qW54TCWXbxd4UwZdUi3O56KQVVEnNu9+xXSTTr\nvZjOThPzC6oKbneuZOXUqVZVnSrhm5glFFRJTGraPT5q3h6YPWpiXkFRYMeOJKo6cSY91aqqZhnt\nOwDDap5t/d0EYBK5iXmIG82kb/RdoS5ultE2cTfBvIVN3BMopYvPWE4pHBlMmLiDMIncxD2BWdfF\ni0eGzZtnra0mTEwXZhlbE/cEZl0XL0pEsUQis9JOEyZuBaZFbuKewKwvL1k0MuR8vllp57Qx2wHx\nJuYlzCtv4p7BrIYbF48M/f2ztONp4E4FxJuDx5yDeRVMmLhV3OlElDsRED/TwcOouWIOArMKsydN\nmJivuBMB8TMZPMZqruBymSm1swyzF02YmK+YivA/2zLITAYPTcMSj1OyvoKJGcEkchMm5gpuhXRv\nlgE12xr6TLzGqiq1Vm5lEDB1+RvC7BET9zTmDD/MpHTjJCdhiURuj4Z+q74BRSG5Ywcl6yvcCGaV\ns5vC7A0T9yymxQ+3m/FvtXTjzp2wd+/Ek9B1yGTA6ZxbRWVuZRAwq5zdFCaRm7jrMFXOnTI/fB4W\n4a2WbuzpmXgSqgq7d+Ps6oK6Onj2WTmx+WrFmlXObop5emVNmCiN6XDulPlhjDx1bzlafxw1pKHU\nzLJFeKulG+vrJ57EWHtzPh8kk7Kv+UricBuyue4+mD1i4q7CdGbhU+YHVUV3etm9XyWWW4CnUqX5\nhSnwyXTlmKnIDqUaXfx5jOwtXV2waNHdYcHe6Zj9OQ6TyE3cVZjuLHxK/KAoaE27iA0nKKt1E47a\nbi7T3k45prjRpT43N5NsbYXVq00L9h6AeYVN3FWYjVm4kXwIeWlZ9St4anyEo1OUae+0g05RyFVU\nzA6Jz5nQHhOTwbwqJu46TGZlT4WPdB1+9jM4cAAsFti2DV544RYGiLvFQWeG/s0LmFfExG3HXDDo\npspHmgbDw2Czyefh4bwxfbPcm3ElROaag+5WLoKuQ2dnvgPM0L85C5PITdxWzBWDbqpKh6pCZaWE\nYFss8v5mxrSuw759zoklRGbgoJvVwe9WLoLxG02Dc+dg3bp8Io+JOQeTyE3cVtyqVHy7SoSEQkLQ\nbnfp7RRFpJSmJvk8lfBrTYN43HLDEiLTOZ9ZH/xu5SIYv/H7hcQfewwaGu78zMJEScz4qgQCgZeA\nEWB5MBj80cybZOJuwq1IxberRMiOHfDjH0uy4969k+9XUaCmpnS7SmXGu93gducmPcfpns+s+0lv\n5SIU/kZVTRKf45jRlQkEApuAK8Fg8EQgEHg6EAhsDAaDJ2epbSbuAtyKVHw7Aj50Hd58E06fhooK\nWLNm2hVYJ2TG79kjGnplJTz5ZBK/v/Q5Tvd8Zt1PWuoi3GyKMNc0fhM3xGxcnb8CngUag8Hge7Ow\nPxN3GaYrFd+OgA9Ng1xOSHxkROSVaVZgHUfGnZ1w8CBYrZDNwurVFhobp3k+k5DptDk0kZBU/fp6\nqfVdCoUXYapTBDMJZ95gRkQ+ZolfCQQCw8BvzlKbTNzjuB3GoOGnW7NGSPzll6ddgXUcGft8MjCA\nvFosk/NpyfOZhEwLuX1KHJpIwPe/n//RD34wOZkbuNMx7iZmHTOVVsqBEPB/Av8QCASOB4PB9tlo\nmIl7G7NtDN5scJiu0gCwfXteWnE6czfk0wnnY5Cp1yvrfYZC6P6a6fsGenpkX1VVMDQkn5cvB13H\nMjIijYvHx5+YqoqjoKdnamE5JuY8Zmrr/DbwF8FgMBwIBK4AXwX+tnij7u7uGR5mdqFpmtmmKWIu\ntmumbYrHx382wgfjcQtud44dO5IARCIWfL7cBDI1fr9lS36b1tY4vb0RysqydHRYOXo0QmNjdvJG\n6DrOWAzH++9DOk0qm6Vv2xfp6vLi8+Xo6rLQ2pqkoiJ345PJ5fBZLFivXiXr8xHJ5eDKFVx79pCL\nRAh3daGvXk3O55Na4GP6uDMUwjoyQtZiIdnd/flo4LpOtLeXbl2fc5r7XLzPp4OZ9mYOsAAEg8E3\nAoFASXll4cKFMzzM7KK7u9ts0xQxF9s1220KhcR6NsIH3W7J7JxOnXJ9iU5trY+zZyWZqKvLxyOP\n3ISvdu2CY8dE4mhpofzLLxGsKyM2nGBRnYvVq5Ub/96YRvzVX8HAANTXU6Yo8K//Ci0thNNpymw2\nObFsNq/XhEJyYnV1+aiUW02FnSrGpKThri4qFy2aeiz75+RsnfY99Tm0raenZ8rbzlQj/9tAIPCf\nA4HAZaDSDD+cHgrvBROT41aTEqf6m2L9G0pLyMXXSwvpqPv3oCSjlCcS/PEffptXf6qwZIlI1zeV\nnhVFRo0xj6mCTjN70EijYkdhF5M+ookEvPaaCPSqmifGUOi6V9fa2SnSSTo9PplnKt7kGzlEb+WC\nFJbWjcWknUalxlL7hbmRSVYKcyXLrQAzPnowGPyb2WjIvYbie2Hz5jvdoqnj80y5L/XM3MpvbhRx\nV0r/Lua5wn06nbJNcjiB54Kf5iYFSyhEjUtj2TL/9ePedIBWVdiwQbSamhpQFJRkFH/9FFYJeu01\nOHRIRpulS4UYa2rGeXVTdXXw67+eJ3LjpKfiTQ6FxMqvrYVodPxodiMSm6yTC0vr1tXB/v1SK73w\nohbud/t22Y/dLq9zySE7B53Fc2SIu/dQfC9EIpY73aQp4fM2Rko9M0Y7jGScYl9e4W9CIQkVrK8v\nvSKagWJnpMFzdrv83uXK79OY8dbXugmf96D195Nzu1H86o35sdji3LtXrHGfT2QWRbmxpWyExSgK\nRCLyu1OnYHRUSLyouldC02SbUrhZ4Zj9++H8ebhwQSqHXZ+GTHTS4vfnL8ZknVxYWremBt55Z+JF\nLbzQui6lAYz+euWV0m2dCm5XmvAcKohmEvkdwsRwtps4te4wjGdB1z9fY6TUM6NppcuAGLxRmI5/\n7pzsx2IRxcGo/VQ4s4eJz7mhevz5n0tkSkWFODfDYVErAMJRG57tm1CbVtIf18SiJm+4jiuiVTwC\nbt+eT4EPh/Oj0fbtsnOjU42d6LqEGYZCcO2aRKak05Jx+cwz47Ucg6Tj8RuS2KRfaZpYy088IWTd\n1DQx4uXDD6VTy8vFKZBMig5f2Mk3KmhTiggL/6cocmHtdjnPePzmYZWlcLvShOdYstSdb8E9iuJ7\nob//TrdochTLCp/ner6lnplIxEIslp91G6+dnflM8uZm+Qx5/57FIu12OvMz++sySXLic97ZCWfO\nCH90dcFv/qaoAgavSsy4guLyQ3c+FKYkdxRPLWA8cbnd+dHJYoGvfAX27cvv5P77xfLu64OODliy\nBB54QCzuRGLatQFuyG8G0UajYj0XL1rR1CSjW1UVXL0qHVRfLx3idpe+OcYO6OzqgmBQUmOLp1LF\n+paqSgNnUqzrdskgcyxZyiTyO4g5di9MiuJn4dlnJ/qpbieK+8nny+Hx5J/5RAJaWuS7Qsu8oUE+\nh0MZVEuMnV9xE08r6Amdd3bLaj89/VKvtr5+4nPu88l+Uil5rajIW9vFCsKN+kvTwF9shfr944nL\n0IFbWiT1NBqVUcblkv/7fPJZ02TH2axYwy+/nJcmimCJRCbVmTUNYlqGMnucsOZG02z5Pr6Zxen3\ny7Tk44+lHTYbnDghJP6FL8BTT02sNlbs7IzHJ978k+lbM7nR5qAMcjtgErmJm6IUB30eTs7JnmFF\nEYOupwdefFF8cjbbxBm9okDzTh3ttbdQc2GUfW5cO3eiv/cWngt+wuc9VD66CWxKyee8pgb+w3+A\n7m5YuDBv2ZeSl27UX6rK5ORoEJeqiiU+MiIjhqLAyZN5R+Urr8APfyhVvxRFSP3ll+W3b701MXoF\nyCmKhDcmk0L6BTqz6tbxnDtCWMvgUW2orzzMODq4kZVRbJW/8450SGXl5Is9Fzo7p7qO6GxYOnNQ\nBrkduDvPysSs4vN6FgodmHv35lWGl18eL48WW8Q7d+Ytb48lhup2Y9zaSlzDbx3Ns25PD0oySnOT\ngtbfj/rUSvD7J41meemliX68UvJSXBNRXHeraHFFlANNR0VDQZX2FBDThIFKUUROMSxxRYGNG+XE\nDSdnQ4PoO8Y6dJCPXqmoQF91H1rbCGqFDcVtx/3GG2L5+nxSm6BAZ1biGs3r2tDslajpYRRtJcRv\nMM0qrj/g98tINzAgFn919Y2L2BQ6Oz/vdUTny9R3BjCJ3MSUcLMgh5mSfKFmm83K54sXhRsAvv71\n/L4Njdzg5ni8yPLe657g+dRDGqFsBSj1+BUXSn83/spK8KtQ6tzGTkpRVfx+hVDoBvISOs59+9Dt\nbnafW05s3RY8nhzNSIz5uFoqCZ1QZ4T9x30kdSX/lZ6An/5UjutyCam/9Rb09oqmDGKp79yZz1bK\nZsUCttvRu/rYPfoksWMX8SSHac7txh4Ly/dW6/XGXr9WdjdKNoV/pE0s6eJwwMILOVk9l+ZmIfKL\nF+U7p1PaXar+wdjgkzP0qttxE93DMHvMxIxQuJBMKet5qigOGUwm8ypDLjdeSna5chOkC0UrsrwL\n9BV9ZzM/ezXOweMecsdhu72eF1YOT37zl/AEqqpSUl7SEzqhli4yo1G0+kV53XkYNNLjYsJ1t8ru\n7x9hYNDChb5Kmr67gmjMhhbS8f/iVfi3fxPrdvlyeO45Wabo+HGJUnG75f+dnfmOGhyU8MPWVrS0\nj9hSC2X9pwh3jqKlT+Nc5IHGRrHsX35ZTus1jZjuxHP+OM3xkyiZJDz0kBx3smiTUvVcGhrkf4oC\n69fL7xMJIXaXa3yST8EiqM5AQKY5hfrcdAP/TUyA2TsmZoRiHx2Mt55vhkI5pXAdgy99Sb53OuWz\nEdQRi0Ei4eSb3ywKeriBU0uLKwzEVVIZcGTjDIdBW+PHH4tdJ61xnFHCW6n4/RNLeid0dn//CLHR\nNHQ4eemlQZzeCnpG3FRW5nDjINSTQC13QUKn8/Io2miG2lqF8+1x+tuj1DSWoaLJydjt0iHJpMSJ\nDw8LKdrtcOwY+mgUzbEA1edCCYelEYEAZLOoigvPSIRwxILHEkd1pkiU1wuJf/nL0g+vvUXsUAVl\nrjThYR3N68PvsYk31+HIh/Toev6vp4frhdaHhuS1pmZ8GJPHI+9bWsRZUajVa1p+EdRsFscnn8gM\noaZm/DbFTgdVnXPZk3MZZs+YmBHcbnnWjCqAhvU81eXciiPu0um8Hu3xyP+NSLW8xW65HvSQj9dW\nUCYR8t1uuHJJ53JLBsWh8Ij9CuqxT6HcC6+8MjHbfaeKUjAo6G4VLTSxJInWExELvMZFp7aUzsZ1\npJz1RAdt+DKwx/ZFkpkYzsNn4eAFYlkXZ7oqabxwmkcdYZ5Sr+Df2YyijJHjsmXSAU88IVr0lSvy\nl82iL1vJ7vJvEjuWxbNlPc1fzKKoblnd4vJllFyG5udSaCNlqP/2v1FyHrIejwwC+/aBxYKqg6ei\nlvBwFo9XR00Pg2KXC/f443JSBw+K81JR4OhR0ey9Xvjt35bwxzVrpI3FOlMkIjMIq1XkoMJMU2MR\n1DEZaEK26GTJAnMse3IuwyRyE7cMw+loOP9Wrcpbz+OSYSZBKCTPfEdHPrT661/PP8PFuTLGs+52\n5yakzYvTU0HT/RACv6qjxIXU4xqsT59gw+okWtzGU5usKLXbIJ1G1+K89paLQwczVHiSrFnvRIsr\nqDub0XoiuGt87N2rlDQM1XofHtVGaCBNy3Ad2TMLeWufjVWroLUVHnhAYXGFlZ4RC7jKqXWMkFSX\nkejVsXls+M9+jKJtz2dlFi4UqmkiWWzYACMjaJZKYueylFXYCFt9aIoNvwvJCN2wAXw+FL8f/09+\nArYc2Jwoly7JSLh4MSxbhqJpNDdm0Fb7UR1JlNA6IdhMBsaqMBKPyyDS1iajc20tfPaZRMxUVEjC\n0q5dE8vgJhJC/JcuCVlXVoqEoij5c9N1Urt3C4kXzppK1UjQ9c83YWGewyTyuxy3U2Y0CLe6WmrF\nPPbYzVPhC9u1f78svdbWJgpAoRY+QQMveNY1TcrMdnaKnONyyeurr4pknElneNBykq89eBFXhQv1\nke2othgxu486+yj+8ixkLeieMjpHVPRkhoqBi4yMQjYdIfHMRt57TyGZrLjueDXCuQsNQ8Wl0PyD\nh+lsiRJ6P0VFlY1MRpQKu30sNybtlkzQ5Cj9MS+2gT6W65eJ9nnRGuxctzEVJb9QqCFreDxixS5a\nhLpjJ54304RzHgkXVJkYvrN9uxzc64WrV7EaIYEOh4QyVlWhuN34//RP4aOPJMupr09kj23b4PBh\nIWSfDx55RLT5w4dF93Y65YQGBsbHW2Yy8MYb8O67MntwuWDFCrHeC+NAx84t+fzz+QSfUjUSirPP\nnn3284l3necwe+cuxu2ui1JIuKqa939NZUZsZIE/84xwwNKl45/vUiqJ8ayHQlKpNRaDn/9cwpLd\nbpkRALRf1onGy8AW4IuBK/iB5m0htOF+1Eo7yq6vomtxdu9X0Q4ptJyJsco/SLJGxZGNsvunCc63\nuXliS5xQ0s3JM7Zx4dyFUFwKDfeX4zs2RDYrCZiNjcJbu3ZBPD4Wo61phAYzVP79aaLWZXjSo6hb\n18pOC6cvpYjM7Ubp6aT5CR+ay42q6qLjFwe067p0RF0daBrZqioZWRsaZLQMBmX7n/xEjmfEgZ85\nI8dsa5OEnkRCZgPr18tA0dsLH3wgg4TPl9fx6+tlOnXokBB8JCIzgFRq8gUrbhYKWHwDlYpJNzEB\nZg/dxZiMVGfLSi9FuFNNpDO2i8Xg6adl5l1oeBn7Km6nrsOePS7OnxeDMhoFxZph/Yo4ittFIqGQ\nzihUVeQ4fsFFKtFITY2fnc/vQu+JEPL5UFHoibjQxuSbNQ84SR72YE0kODJYz/btDizvXaY/Oozb\nZ2XjA5txeWyTlvxQFNixI3md6AudsLKtAi4/NWqCF7z70BYmUX05lJe+P3H6YlQdNCJEEgn4y7+E\nU6dQHA783/qWWNhGfQFDfnA684uIPvggZDJkR0aE2L/xDfjkE9lXPC6a+apVsh+/X8jXIOhUSqY6\nf/3X+WnFihVSDsCouWIU9xoclClVX59Y4BUV4qV++eXr1RynfbPdI5mYsw2TyO9ilHombtVKv1EJ\n2OJSHFNJHjK2M3JbimfPxYapQfSaJhKMUQpE9WXIXmnHaenlK48NsfX3f4VPDiuk4itpPZei9hkH\n4aiNV38Kp05VkMnI/tauhZbzOusaEyhOB7b1KylTcwSP+RjqjbNtRQ9NDydQ08PsVe4jllVvWPKj\nsB9crrETCBV1QjyOsv4+/EYhqOLg9FBI9KazZ0WmaGwU8j1+XL7P5USDrqsTazgUgkcfFQKFfEXB\njz+GS5ewWa2Skurzwbe+JdOXvr78jZFISDuM4leLF+edrGfOCDkvWcL1wjanTol+5veLB/qf/kmO\nW1Ymv334YfiN38iPdKViU6eCwsJhpjU+JZi9dBejFKkWc8dUggGmS/5TSaQzok0my0MprJa6f7+o\nADU1wh8+X47160WeXd8Yx9nSy1eeHGXfoXJikQTllT62/YqNg3VuognhkHhcjNVUaiyqz55hXfIE\nj2qd+M6e46BlO1HFx/Ztm2h63IX/YEiSeSo8NO90o8WnMYMxYqeNUB6jvKyqonvK0IbTuMs9xF31\nqE6vhBIaK1okk1JmMRqV15ERIVGQBKD6+ry+ZJR2VFXpGI9HHJAdHaBpWNJpkTsMyaWpSX6TyUjn\nGo7Q/n6RXx54QI7Z1SX/M1aXvnRJRr5UCrZuzZfB9Hjk/FIp+d2v//r46Uqp2FTDoTtZv023+Lzx\nu3s83vzePOt7CMWkeisz19mOBDOe14EBKXfd1DQ+Gq2wnQafGBFr8Xhexvj61yGuuVD3D6ANp4nl\naiirdROOCp+88EI+Rn3PHokksdmgskIn0TWMJZ7g+LUa9OBinPfbeDbQjv+plSg1fnhh13VyUBRF\nIkSKT2Iy8giF8jJHNisnWFODjsJudqFlEpw76mJNTEGx7OIrT2ukXWLxKx6PkJ7xWlMD3/1uvii6\nEaTf0iKEXFYmnaRpQnxtbULWdjuWK1fkhH/+c9nf4sWShblwIbS35/9WrhSL3GqV2ixGKGJbm3T+\n4KC8b2gQq9/lyss6htZUVVW6vkph/ZhcTgp5TYbC0duodW44gG92MxVP3e4xQr+3ztbELdVNmS3Z\nsrimeW2trF3Q358POS5uZygkRp9RgkTX5fu8jKGg79qF3hnBWe6TGuEFkS6GL3HbNvlT0HF/8Eve\n/LCCSPsAJwnwRJmVaAJwe9BQUXVQblaT4GaWo2HNGq+M8VRSwV7hY1QT9SIWsXD0kIvNj1hQK2w0\n79iJ8tNX5XfnzklMeXMzvPmm/O+tt2RnsZh0XjIJNht6eRXaU19GbViO8sQTsGQJKYcDp9stkojh\nDU4mRZppbRUPc3u7EO3Ro3K+IyPi5PzTPxUZ5tAhqT1+6VI+yP+ZZ/IzhuFh6eRkcuIIryh5OcXQ\nw3Q9r9UX34Cqmq8xnMuNn81MhlJTt8rKe47Q742zvMdRbDxOt4bQbBTNKjacFEUI/NFHS1c9NY5r\nhFgbMsw770hm57e/XRDksVchFquQII+CfRkKx8GPMuTSSbY/4eCFpzS0SJKc14u6LEc2UkP/o69Q\nWaaz3+0m+U6+/gm6jtYTQa33obgKGldqilJ4rqofbf3jqKGrKA311ztbVcHj1NEGEjjtLmIRC56e\ny8SiOeyeUWKbNqMNJPDH40JkLpfo1D09QnwVFRJ5Ul4uJDymoeuLlrL7Ay+x4RieujKan3ke5R//\nAVtPj1jgq1fD889LhzgcYnVXVEjb02mx0DVNolMGB4WkW1rEadreLudZXy9ZpJlMfuStr5fvjRvD\nCOkpvOGMWcTYBXR+8IFEzRSuBmL0qarmqyoWJw0V7/d6MH/R1K2qSnwEhhZXvP+7lNjvzrMycR0z\nDUEsXOlmJnJKcS2VbFb+b5SfLQxwKF6+zfgzfG6RiOX68128tGRhtJqmwfBABmv7ZdB1hj9MoD2+\nFne5g3OtdrQuO976Pna6jpPYuoNDR5T8CkIDOgf+7yPExsq8Nv/g4TyZl5qixOP5/t4DsVOL8KS9\nNDdEUIwCXG43zbyFZkvz4qNO3hxpQmeQlqFq0vEUqiWGWuOCS5fQz7dKZcLNW1GOH5fokE8+kQ70\n+WDHDnktL0f74Cixy6spcx4m7NiG9pOf49+zB2s4LMStqtLhPp+Mmv398vr660LM6bQ4GD/8UPT1\nri5JDti6Vazzzz4T2aWjA775TbHIDc/zunXynaGHK0reP1BePn6UNkrcGjXSjUWYjSJgRinLmpqJ\nSUM3klAKp25DQyLnGDdE8f7v0lT/u++MTIzDTPRtXYd9+5y4XDNfSL2Q+ywWefYLF3MwSmtoWunl\n29zu/BL43mewAAAgAElEQVSOFouC212wtOTZDOeSKTZtseN2K+OOWemOkU3pZOxOHNkR3MSJP/VF\n1nUOY7/cQsKh8+GJMnKZBOfafdePq49EGBiE2joH0VAKrXMEf7VtfKD7WMiNntAZ7QhTW1mLFtKJ\ntfbizUXoty4gNDBCzWuvXdfLlVwOf20Z9Pfy9eYYmrMX9+AR4pULUZ8JoGgR9GSG3Y4XicUteAYf\nobnyE5S1ayWW2+WSVXk+/ljIas0a1NX1eHqthE934BmNoq4flA4zRsktWyRqZd8+Ibfycvhv/020\nHVWVbKxHHpER8dIlOT+LRaSVqiqJLd++Xaz1J57I69bGRT14UKzh/ftFvzp4UL5vbRVtf9kyCao3\nQhZVNV9nXdfz4ZZGslGp6d9k3m/jBjGmbgMD+boxxiBgDBzFGV13EUwiv8sxE31b0yAet1BbOzEO\nvZREXJhhXSqZx3g+jVoqpUpr2O2iJsTjeYOqpkY+G0s49vfrhgFMMpbhMdcR3j3nJ06MveXrad4F\nSlxK0L7wNTdbUx38/BM/ObudvQdVdj4ParWT2FUnlniCiMVDfa2bdd58duq/v+njxLVa7NeSPH7/\nCOrx/aDLkmr6zmY5v/0HQdPY/fMc/b4aWmr62Rl/E2frKt5vW0x6ST1qtp5n1vXgX+xFCQ8Lgb39\nNtjtKD4f/vPHYHgYV+gDsIyix1K0DZfR25Oh3jZE9Fw72jYv/uywdFwqJR00NCQW8tAQis9Hsy+L\nlkyhZpwol92wbBnpigpcTz4pVrRRdjYUEnnm3DlZMSOZFGnlG9+QbM/hYRl0Hn9cLo7DIcf77DOx\n6A8ezOvWijJ+gYmhIdHYcznZb1cXXL4sGaLbtsH27SR7ekTqicfzXuhz52S7xsa8Ll5MtpN5v4uJ\n+dNPpf1GkR6YvUWc5zBmTOSBQGAT0AgQDAZfn3GLTMwqZqJvS92U3ISieMVWfuHs9UZrYBZq85Ml\nEo2MiFT7xhvjS3YYMdyxmIQfGsZcNhajfyCH02uj3hMmOhBDe+0jKWnr8cDOZj507OCcLU2Fz8Ea\nDeJv7qXZGia0wcf7mSaOn/Zx+l0bjz8uXBUKwaHPFFxrlpEIJ9n2nRjK6TYoK0MPaex+LS7W8gU/\n29dkiGlh1Bqd2NHzaOnLrFMz7LPfh6ta5SftqxjJnKGudYTmDUMQS6F1x1FXqCg9PXLCVVXQ0YF+\n9CQ/u3w/H3U/xpVwFY21UR63pyUDtGKLdMalS0Ku2axYzF4vhEIolZX4u1vA1gBlS+F3fodoXx/q\nokX5C2o4Dd55R7Rzq1UuSGOjaN11dfK+t1es6Pp6Sbs9e1Ys3VWrJlq1fr9Y+O++KxesvFwcH52d\n+fU802m5oKqKI52WUEZDF4vF5PPIiIzUBRUpS97IhoRiTO3c7vw2pYr0wOws4jzHMRsW+feCweAr\ngUDgPwUCgY3BYPDkLOzTxCxius7Nwt/t2JHE7c47Gg0Zs9DKh/yst61NnpPFiyeuVj+VRKLOTjG0\nzp0Trrp2LW+VF9da2bsXck4Pbp+dR73dRHMePO4cai58XYzXWrrI5RZTscDN8DDEhjXclRpKtYrS\nkyCZseHx2ejrk8FoZER4Utchl7NhVz0o1Y7ri1N0xqrQnB78tRA+74G0jke1MTCSpdJjY//wFxho\nt9M94qJhUEMfVVBfvI/Y8aOEehIc+J+XiGXq8ZxO0PyffCgVFUKSFgtaSGfYWo190QJW1OZY4Hex\nzfIhypGx+gUvvigdEAhI0k5fn5Bxf79IH8YyalYr+sHPyPzrHnRHEmV1I/zRH8EvfgE/+pFcmFRK\n+iiXE8v87Fnp8EhE9rF4sThZQyEhQa9XOieZnDiti8dlf+Xl8r1hCTc0iLV+/rzIOJWVWBYsyBO1\noZeFQuJcdTrl/5NZzYaEsmtXvlzlnj15vXyy6afHk49muUszRWdE5IFA4CXgMEAwGPzbWWmRiTkF\nYwadTI5fkcdYM7O+Xr43osaMzEmvV56f/fvzuSc3WnTCsPTr6+VZvXpV+KWzUzjly18uCA7RdbTO\nEWKair/aRvjhTTz+sEZC8Ul79rmvJ8y4U1YsZwI0rnuQaNSGs8LD3pZlNK9rQ630YIm4CIeFgy5c\nkN1byJBLJElknVRW21D9Cokdzbz24wRJi5szJ21s3AgV2zfhb1pJ83fcXDh2GdcpB/uPr8aqJWgY\n9LGiMYO3b4jsqA3VlQWHg1haoazWSTjhQ7v/UfwOq+jdioI62EalMkza5uKK0khOjXDQEuGFsihK\ndExvqqmRKcrFiyJHGHHe772HHk2iRR24N2xi7yE/IwMPoSbDNHUcRD35u8R1O+poFMWCkPPDD8vF\n2rRJHJ9+vwwMf/Inst9f/lI6JhiUKJf77hOtHfIecE2T3yxYkF/qzXBCvvSSXECHQwh7eFis/kL9\nbdkymQVAvj03s5qN7C6vVxy0hXr5hKLx+swegHmCmVrkDwO5MXnlmWAw+Dez0CYTcwyFho7TKWHA\n772Xl0927pRKqgMDeQfmY4+JTLF3780XnSjW3B9/XCxxQ0W4dk0i1s6elcqGq2NX+M4XevFcXEp4\n3RacHhsHz1Xk5ZydzSg9negZC3svriaXTKJrMTZuVKmuthGybqFz7WoaFqR4mRDgJxxV5HhKhkRL\nO+vqBqiqspIIbKaz08bBgwqfnvRdz3CPxeDrX1dQXLIoxaetdcQza/jx6Qw+1UpZvJfvbThMTZ0N\n7Qsr4b0IatKKp66csLsSzyIH7lOfEnrvE9RoP4o1h1JXwwv3j7LhySrePqpQoarEjvnQ+gfx13jG\nOwcPHBCpoLVVJB+7m91VrxDLLCOb3EpOOYvHEudgaB0DTi/t8ZWss5xHjS4U5+nyBbIyUF+fEGEy\nmS8s/2d/JtZ0Z6do26tWyQXeskUItjB6ZNs26fQ1a/KjdaGWVl8vbRzbPvHMM3lHyfvvSzSOzSYD\nhrGKyM2s5kK9vDBCxbD0xxWN1/IFvu7iuuazIa0MBYPBE4FA4JlAIPCSqZPffSiUJ/fvl+f4/HkJ\nYAiH87Pc9nYxkoxKiCDP2fBwPhek1HNUquDd0qUi07S0yHdWq3BI+2WdoVA15eUqLz/QSvyx+9B9\nFdfLjITDoMUV/A0NaIqf2EgKf6UNKj1YxkoUnDufI3PkMpaLQZ5ecplntm6GL+2kulphtDtBeaYL\nm89FbCRFy5kkibSHlhYZVC5fFuUhGJR2u1wSqnj44xw5BWIpBw+tg6xej970NMoaLwf2KsSsz+Px\nxNi5O0O8tRO338Xe77xKbLAeZ8RJk/MQfn8fSiRCw2/9Hp1vwvlRUJU1vPKwD6p9QuAffSQE+Omn\n4oRMpaCmBm0wQ4xByjxOQq0XsSg2ulZ+gVwaVGsH2oAVa62bAfUBQt5eamwJseqbmmTU7egQq9lm\nE8fmZ58JUba2ioZfUSEXcsMGOfGysrw1XFkp4ZDFlQqNFYbWrLm+iLTr3Xdlf+m01JBxucTC/sM/\nlJF/Ko6cUtlik3ny75EiXDMl8iHgytj7EeAhYAKRd3d3z/AwswtN08w2TRGF7RoZsdDb68TtzhGN\nOmht1XG5cuRyUF6eY+FCC6tWpVi0KEt/v/x+61b48EMfAwNWjh7N0tQUue6DMqDrEI066eqyUlGR\nJR5P8uCDEAy66e+3Y7dnicUsjIxAOG5lsZpAGxymS9Nx28IktBhDQy4GB8URqmlJ4nHQtz1AMmSh\nI+fCnR3hyW1J+vqsDF3VOXVY5/TlDfyvlo2sbEmwKdzD1qcVbKt1FngvkYkkaHXWMeTVcKYHiUUq\nWFCfpbLCwcKqKPGog56eGOl4Gu31j+DKamw2OzZLgMHBLFVVWQbTMa4diXP1qoPy8hxdIZ3u//kG\nfssIPZrCSCSLx53jw+GH6HXVU63keN51ib4PP2Nhzf24h87A+Tb6/mAP9nQc9/JqFIcVC2CPxchG\nk4QdVfhGozgsMWzuJP3dGk77ZZ58Mkm05xM+y9UxlCpHqXRzwNuEbXQULyGa3QewJJPoHR0kly8n\n+93vov7FX2CJxXCMjJBTVSwOBymfD1IpMqdPY+3pQQ8GsXV3k1m8GCUaJWWzYenqyo/mFguJXbtA\nUXDu24clEkFpaSGzaJF8BwyVlWHt6UExqjNms0RCIXI2GxNujpthyxYskYgs6mzcdMXYvPmm28zV\n52+qmCmR/xR4aex9BXCk1EYLFy6c4WFmF93d3WabpojCdtXW5ktaP/ecGHOqmq/EumSJyJyQlymN\nWk1G0IDfXzbBItd1mVkbIc5GomFDg/xvZESs/6efFklndHgp9TUBlr7sRkfhtdfyhlyhDq/r0Px1\neW/ItkuWwKWgzmlSWEiiWLJYvRWcurgQa5mNmhpY/Y1fY++bcbQyJ72/uELVgiGeq+xm229sZH38\nJKMjOSorYe3qh1G0ELo9wmPekyQSXpq2h9n0+9s4eVbhJz8pQ9fBktN5cKXGonNHWdL2M4mL99ej\nxOoZjttRsLC4zkY0biEX0lkx0smSTisxPYfTmeHIuTqSWQVPr43mp6Ii/Tq9/DLbRCznx+Ox09xw\niq/2vo0WcaKGVJTTywi5E7gzKdypOBvKL5L60jbqB08TPb+MjH4av64LsbW0yMITfr9IEMZSbm43\njkRCLPVDh6QDKysl47OpSVL/k8l8Wd2zZ+VilZXBF7+Yr8Fy8aI4Zz0ebJkMZQ0NItc0NsqsorIS\ndfXqvBNkuqn1s1A0ay4+fz09PVPedkZEHgwG2wKBwMiY07PS1MjvbkwWyli8Stf45dfymnIp+VPX\nRUI5eFDIvqNDkgGNekurVo0v2bFtG2gjCR5YJgd/7TXhmIoKmcEbfrJSse7G8978gsLhI0uwWnSG\nQhYaanU8BUtJ9gwoxKwqfm+EdQsGeWxzggZnPwoLeGnDFcm4TAygdFaLPHDmDE+Er2JftxFlVRaS\ny9BGFmOx2Gi/rFM2fIXYycN83fbPKNcuy8I+6RXkPC7cjYt5dOga0VUP4nFnUR+ph+pqtq/ql7U6\ntbPs1suotQwQjXrRIhb89y8iNAADVU9Su8RF1FFBKJ1ECWdRqzIojUtg+XJGu+MknQr15WlCVbU4\n/ArRqgfxJNOorRFoH5BIl7o6OfGyMnGA/Oqvih4+Oiqd2NeX16CPHBH55YtfNFbOkAv3ve+J7mZk\nj23dKpp3f78MDE4n2O1imT/2WF57M/Tyf/93kWrsdnGS3KzGinHzDAxIXL7VOj6D7B7DjM84GAz+\naOytqY3PY0zVqCkVylj4v+IyufH45HHsBtn29gqZr1wp5G2sYJbLjSfx738ftNEMSscQf/3KATRL\nGaOjz+Nw2BgezocVhwZ09JEIMU2lzG+bEOuezcKDDyls3WJh9MBpnlp1lXMDC4iGH8aj2qiv0fGk\no4SjWTxeKz49BH4P1NejeBz4ey/J1CQ6it56hd3eX2Uo9BEXDz3AusNBPK2tlLu7iQ2sp+tcHEsu\nygWrj84F1TRkO+mx1qMpFfiTvYQHbDzVlEH5/U2oFTb4+GN2v20nZl+Gc/M6Ml6VC74I52Nxtpef\nRe06hX7hMPsTz3Fed3LBGeCRX1vN/sgDJLt247lyhubB/Si9vZSXVeJxLCdcvhh1RS077f9OPJpB\njR9EyaXlAl27Jk5Nv1+qLFqtEm3ywQdy0VIpsbh7e+UiGNUW//t/l7K3L74oOviqVeKdzmbFkfDh\nh2K9P/qohB6eOgVeL9lAQEjciCgxRvL335d2KIpEx9zMKZlIyNp+hw7J6L9xo8Sj36XOzJvh3hu6\nTExAqfr/t5IzkUhIJJyiTFxvc7Ll3mIxmc0vXy6yh/GMG1nVRg7HyIj8r8qXoitioTNVy/Ezdt66\nmCaVsxEIyICxd49O7MAJnLkYToeT8HohZxhf60VRwBKPs9AXZtVqG6vqrqA9tkYKZO35GTuPfUbb\nSBmfeHawJ/0cSsTDyzq4Uik4dAi9q49Q3MdIvAwtF8VaW4U2WIZdT5Bs7+HZxRdYW3uev3MGUD0Z\n2q5U8HZyDZ2RbazxdNKSWMS6pRWo1U78f/wllLNHZDm4I5eIxbZR5gvR88vPoDtE00M6/ddSNCkt\nEInQGS4j5i3jicVX6X/qSR7cbOXwWznKGqsJ62vQFibx54axZ1M0bw+hDfWhRj9Bea0T14YN0H1V\nnJjRaL44l7Ee55IlcjPkcnLRjJoFJ05IjPm1a+IIdbvz5XCrq0U+Wb48n55bVSUk294uNWLKyyGT\nIfXgg3LxBwbyTttoNE/iqZTs+0ZOSV2XqdjHH8ugYrdfj8W/W52ZN4NJ5HMQn3ed/FL1/40wwcKi\nWTdqSyIxZjFrErnyx3+cX+1rHApOTlWV6yv9PP54Pq9D10WWHR2VGfmLLwrZqyoMjTrw+XL40iFG\n9UZWrbGT0sUYC4VAG0hgTyeJOVR2Nl5BeWwNaoOsoFO4vujOnWO1zCtD6KE4Pckq6mt8KHENfSDE\nnra1vNe5isvperxdDhYsVshoo3zx2AnU3hH2XAhw8OJGMlW1OB99kCULLajhOGmlDGdcY3BE4UyP\nk7JEP6N6FZXVVjz1C9E6F+BanGXdcIjHHkjQUBVDQb8emqOGu/D0txP+tJdKiyykHE2r1KxegOpY\nzO6zj6BZ4pyLrWCdPU3NwHnqjw6SPVtO6HALanYU1XoRPA6s4TDKtWv4ly3LhxkePiwkvWKFWNnR\nqBDg6Kjo2MlkXs8aHBQpJZUSq9tmk++NbMqODimva7PJYqUPPCDfnTkjFr3NNv436TTW7m4JeYrH\nJfTQ5ZIbZskSSd2vqoKvfU32X3zjGfdOIiGDSlWVSENLl4rlXxj6eI/h3jzrOYzbvWByKZSo/399\nIJmsaFYxenrGLOaq/FKTJUm84OSU5maam5WS9ZHWrJHZ+MgI/PSnkofygx9AT4+NXLqSGv/TVNao\n8KkNu10GjZoaOHfFhXa5HtUe55VH7LgafDA2IBWvIOZyKSSe38X3/yyFlnSittn48++pDGQXMRCL\n4LKksTnshMJ2Ghxw6IhC/OpmfEPVRGwJrFV+rPXVLFuhsHLtOr7b/TFapJb/6521vHPNRm93hi/d\nd4nXOpaSyVp5vcPHGvtl0p5y1BofDY+mZBELn08uwP79KF0dNHcfQYuBaotBREfzLED1+dFWbiK2\n4gH8a5Osy5Xz6Bcs+CJtvPVRLbmr17AMDrBTfRsl6oKGMYdBMilkbNQDz2RE/hgczKfgPvSQdLoR\nYphMwle+ItJFJiPEHI3Kth6PSCjGQsx+v1jTuZzIMrmcODm6usQyaG3NryQ0Oor31VfFet60SUg+\nkZDR+qmn8s6RUGii7g35aeOZM9JGEPL+lV+ZxGqYAQwLpvCGmcOY2627BzHbq/FMBcX1/w2n5GRF\ns0rhusU8JK/19SU2Gjs53VtOqCcObRH8yytK1keyWESGVhSZoT/3XF6C6e5WUGr8vPASND0lv7le\nVXW9gn3DMhIjcXoeXEMDCtxgHYjOHoXBiEJdnbT9x68qeDzPceW+BMm4haqYHcWSIqo56Opz4a/Y\nQGBxFe6ebrLWejKKk/Y+J7rTi1b3CssW9jP6+hCqK82xqJ+9vRsZinl5pvYMPRkH33noAov/7Duo\nlXaxxN97TxaLOHpUrMyVK1HCYfzZMORsYMvhX1YOVlAjPXgGrxK2+nDaNA4eW0ncupbWdidN3R8T\nDY8SHx3CFZWLakmn82vblZVJB6RS0qkOh1Q8hHw1w5UrZfTMZKQmi90u2/l8MkIvXZon/eFhIdtk\nUv4yGUmNXb9eRnWvV6Jb7Hax1u+/H86exbp7txDkyZPiVN2+XQYQow0/+5n0SVvbeN0b8rKNpslA\nkM3KYs83WkUokcinIE9VLzRq0hw4IDfitm1Tc77eQczdlt2jmO38hanKNEb9/+JCVkbRLI9nzJE4\niczichkW8w2eGVVFd3r52ftlHLiyGssFlW2TBChs3SrP0bVrIrO+/bZY5cX1Wgqf4fwAZKOl04ft\nEKhnhSuKB0e3GzrbdD7bH6e3x0N7u401a8bKfVTbWPOglwMf67SfHMJuTdO4IMEDm5cTjqxG2djA\ny19TeK4nQdeojzNnLeRiEQ585KX79AFOnVqDI6djc2b5FfUAb8a2cPxKOe4qF2f666h7czcs9sKZ\nE0KCLpeMWjabhPClUmMrS6vo3nK0pAdViaNYMjSvaiF0vodfpHdwOFxG2aNrySxO0t9RQY2lAzUx\nCFE32O3k0mmRP2IxuRGqqkTXNhYvTaXEMq+pEcJcv14I1miDEf7j9QphOp1yYY4ckf3098vF7ugQ\nZ8aFC2KBL18u8aIWizg9QXR0n0+2W7RIImVyOdG5jYqHoVA+pd/hmKh7G1XV+vpkcQwjzVjXS9/c\nhXqfqsoNeiMyL1zCanhYrgfI+znuRDWJfI5hNlbjMTDTRZONollGPLgRLz7ZvlwueYYna4umKejb\ndjHcmcRmc4HVNuEZMdo8MiK1VmIx4R9D7jE09IGB/PNrzHyNvmtry0fTRaNSLzyrxQmlPKgVIsV8\n/891Bo930jukEGjoordmFcuX28jlhE+SSWi/pBOPW4haPeiZGMtro/hWWnj5W24Ul8JHp110dmTo\nOXCRelcPloEhKocvscEBAf0cbVVbYXiIlxd9xnB8kCULnRzqWEzkPY2a8jTN/jBKhU9kiGhUyDGV\nEmKzWtEf3MJu51eJNQTw9LfRrP8M5Zc/R0l5sWaXUbGwgZFTLrY01/FcshflyBUYLIO6amhoIO10\nSkfZbGLJ1tWJFp3LSXihrstIaVQSVBQhY6NU7Nat4rzQNHk9fVqI+coV0b4GB4WY9++X/eu6WPVG\nLWIjBt3lkjb83u+RCIXwGOVxT52SbbPZfB3ztjbZ/7JlYv0X6t5GVbVMRgabCxfghz+UhS5eeEG2\nKXxwivW+np4b36CFpQfGnLNYLPOi2JZJ5HMQt1qtsBilZBpDMpnqIGG0pTissJBUb7a/8c+IQvlC\nhUwHWLITnxGjzS6X8FouJxxgzBB0Hd56y3m9hLXxvH/ta3lj69QpMQwvXID1a3X2/uVJLOkUitvO\nzh9uomdAQRtOU+eN0D5QS+9gknhWZ98+G8uWid9s57M6v/jREKMxKbq0PTDEC1Ut+O0RlH1ueh5q\n5sf/YiEdSZAd9LLzxVGqlRTpS3H8llFqsgMsrTxDU2If6tr17HWsZGDdE2TsWZyuc/ReGiZUG6am\nq0OIs7cX2tvR7W40hx01kkDrChNbV0XZylrC3V1o4Rz+VAo1EcOT6qchHmT1Ei9f/XI1+65+mVjt\nCjz2azRv6kKxZHEeOCAXK5mUIPmNG8Uhqety8Ww26TQjdvPv/1507VxOrPRkUixmu13I7cyZPGFv\n2CDbHTok+3C5ZBDKZGS/DgccPYp+rRct7ULtG0IJBGQ2YJDumTPymsuJrJRMik5+331CzsuXT7yp\nfD7Zx+CgHMPtlvaXWgloSnqfwBKJjL/Bn31W2gKmRm7i9uNGRFos07jdt+5ILSX5GFKiUW5jMhmx\neEC50TNSWJd8aEj4xO3Oz1J0HYaGrGiaGLEXL4pPz2bLS0PJpMgpb78NBz9M03e+jg2rY6ytHiA+\nEKG+vgJ3mZ2O1jLuXzDIiiUJTujLyYTk2PE4JAajvLCpg6G1Kq7kCM9/cyFqW7skA2nDjHSMEmlP\nE4laiUes/Mu7y3ly7QjK6kb+vOpj4t3D0LwZf9salHiI5qVnGFixgOCFAK+2bcUWDaGmFbYOv0tD\ncgCX14vuLWd3cgcx+0I8ARc7n8vhqQwQGs5iaWvHHWmRk07nIJPGFhrC132RuOVhYik7ZR6dcKoG\nreMM/v4gtqEhqUrocMjCEUbZ2tpaKS+5axf8j/8hnb13r5B5KpWXFowCOitWSMihxZIn6kWLZKQ0\nViFyuWS7//gf5ZiA/slhdseeIpa24wGag5dxaCG5oN/6llwk4+ZpaBArOxqVmUMxiRfGyBpW/vCw\ntKeyUrYpZWncVO8T5Hy+8Tf4PCDvQsyflpqYgJtJJ8UyzUwcqaUkH6MIn80mz1NT03jNunANzuk8\nI488ImpDJiPP3siIDBhWq8zkL160cfiw7K+iIm8cGm3zePKKQnmNg2GnhYFByNY70V0+EgmIJRRi\nVYspq0zxjR/aKXvLxvHjcsxLl6DlnI9MdzW5dAaLsoC/eW0Jzmtp1tcP4lHLSUQUcrkk0YyLZYvi\nZGtrsW+7Dw03bWqaY4NOrBeXoTqX0Ky8jmLJ4Hrj/2Nl5e+S8NdgDbfxv07dzwfWGiqdEf7Y+f+i\naBArq6IMjbBeTrytnZ3uN3ntQA25RIq92naaHecI5TwMWxdQ6xglmgR6+/Akhwlf6seTCuHOnGGg\ncgnptE5tSEOprhDy27FDOuUf/zFftCqXk5uir0/+53KJ1VtZKTHk587JRbHbZfqTSMh3fn/eQ75l\ni4ye5eUS3lheDo8/jrb+C8Q+uUpZYoBwyoV26jxOX1oGBZdLplGFi7PeSFMsXEIqHhcNXtPylruu\ni0RjOHGMad5kel/xArEwe5rmHcD8aq2JcZgKMRfKNFNxpN7Iwi8l+Vgs418L91Ocqm+s7jXZ/guD\nBbLZfEJQOi384/MJ37S0KNTWiuG2cKE8p4XP7vbt+fothw7ZWPpwHauWxHAsamTvPoUDByRHJZm0\ncdHl5t9eh9/4Dck6HxnUefcXCVo73UTrVxALp/HX2MlgY7hyFfbNixlIujl9CtYv6+dcO+zYHKLV\nvpxPj9poO17Pm/0eLLkcm1xZ1j5dT+iXP0K51oFb66em9lMcsc2MaDmwKtQ6RzhR+SSvPvs4DaHT\nOD0ewpc68SytRo19hvavp7B2PUCZHiIctRLKlLE/vonzlpVcCLvZRgL/AgfN6zvQDn6AW02z59oO\nDnY9QtoCT5Rf5oX6Eyjt7UKyv/3b4ngcHZVImVxOOs2osQJiwdvt4vi0WsXSBiF2o3KhQcDPPCOy\niE3Q9ukAACAASURBVMsFf/d3ciNcvAgDA6heL55HNxBuvYZHtaH6e0m3X5Kp1vHj0uGFI3+pG6zY\nGjBunmw2b7mDLDARich2O3femIiLrft163Cm0/Dtb89ph+aNYBL5PMZ0I1xuZvRMthbnZPD7xUdl\nzI6Ly0AXp+qr6o1nEEZUm5FHYgRYLFkiK4kNDAg5K4oFp1M4qKlJwohBvj94ML+QxVe+Ij66t9+2\nEY+rHDkOmzdmSIZTWHAQi9nwevPWvIJOw+lfUNnu5+KnK0iW12FVXLjLwe3MUOlLkrY4cGRjZHMe\nlFWN3N+QYtcf1RJJDPDhXoWMb4S2awly2RzXWrOs3ppgf98akn3VePRRdqmv0/TwMIkT5/l/wt+l\nX1+IUl/Lkq0KiZ98wrOWj1HCZ1CjtSihAdwOB1k9Q8hWheoYAG81yVQZT7hO0K9X0hS7gHKoCr7x\nCv69uwkN6Ax7FmN112Cx2xhOD6O1D+GvUfKr2KuqRJ4kEvJ+YCC/YtDv/q6MpFeuSEdWV0uCUEMD\n2GzoWhwtpqCORFDeeUc0cotFwguzWZnSyLQJJZ2mecuDaLZ2VEsExa6SYJlIL8YAciNMZg288sp4\nS9648GMLXF8v2DUZCq37sRRiyzyvVW4S+TzGrUS43MiRWsrCN1DKUlcU0cVLHb/UIHOzGYSqyoBg\nBAssXpzPGXniCXl2ZV3fNKtWudF1WTnoww+Fe5JJ2d7jEaMvk5Fn2m4XCSZ4IYN28BTVo04W2irQ\nfXX4/TbKyyXwIjmcwHNW5fH7BhkIWWlVPLR0lZFOwYbcCX7tvhOkT3TxfqYJV1st8aWreXKHm+Wr\noLs7R8MKJ21OK7acTtZiweNIkYvHiGUc+JUYYaWSeNVi/Gvr0VQ33w+9S09PjuPufhL/YsWTCeN3\nJ0AFLenAnXOyN7uDXLSNrDPHI8pRVGsSjzVJNKNS4wzjL8uKQ3LTJlixAtUfpjKRJWvzkY1EqFzs\nRHVUycLRhqPy135NLkBnp0SJQH5R1t5eCRe6eFE6cNEiCeK/7z70hM7u731CLHgVz7/vprnuKIp/\nLOh/aEgI32IRQk0mobISRU/gD9RC3QOQTpMaGZFwxuKR/2Y3ZCgkWrdRw6GYqI2BYSoDhHFzGjdu\nOk3uZmUB5jhMIp/nmK0IFyjtHO3psVBZOXnooWHglSL5yRZYnmwGYQwMTU3y2TgvYx+JhOjXGzZE\n+Od/LiMel8Vs1q0bH4Bx4YJwxYkTwkFGEb5HN8Z4KnYRfb2N1z+qobW8gtG0l74+UQyqqxyEr/Sh\njB7Fd76BYWct9nScoWw5hzvTeC/ZeTzTTtK9nCdzZ+lJhHhw6X2gl0vbX1Jo+sIKBv/lF3x4qpL6\nVAfhzhgWq4Vw3So8uRjudY3svvoAMd2BxzpE84ZDLE8dRzvfierW4WKS3ecaiTn9ZG2LyK1fQFlV\nF++HNhByLKAh08OuzSeJd55HdSRRnMvyBJ3JSKJUw3GaqsJEYlEaEr0olVVCytEo/MEfSOevWiWd\nG48LSSYSMvV57DGRG4wl0jKZ67Gd2kCEWCxHmVsnfDWKFu3Hb78ixH3hguxz40ZZKu7DD4VUjXT8\nixdhyxYShpwyFcvDuGHGlu0D5PO2bePjTv3+8Y7Tmz0QhTfnmHWf1LR5p4sXYv623MSso/D+NuLG\nu7qcHDuWr6FUbEnfyOFaKi69cP+lLPniJB+QfSQS8F//q1EXRsXpFJnXsLxttvzKY0ePyv6NAAyD\nw8qrnfjbjpM4cYFkzxeJ1S0k4m+ktdWG2w2b16RQGxegBtbivqbjzUFvX4bB3ggkcrwxEiBREeJq\nwsqaiiTtV7o4ZE1wtt7L5t9ZJG1vcOH/ky/T8k+DRI9bUAeusrPuIPHapahb7kOz30fsYhVlFz4j\nHLOhDXVLLRi3GwIL0a6NEnNUUKZECaVVLH19XItVENTqsOQydOgVNCmfUFORA69fOmvLFtG/bTZI\npVC+0kxNKoUSCqG81wpxTazs/n7pYJ9PRj8jdV7TRBf3+yUxyXB82u35ERJQ6314KhyEL+p4UqOo\nC20wOtbBbrcQ+tGj8rm8XDJBDct/bEFUZzYrVRbh5kV8JJFB9HRD+nn/fcn8dLnkYhuldI3XW5ma\nGqsUzWOYRG5iHIrjxn2+nLHwS0lL2pj9er3CE8aK94UoLltxM63c2L4w4acwt2N01HJ9m4oK4aAV\nK+S7r35VOMQobf3OO/m8k/aWDGurN/Avw08zavNwtddB3YIUmYybwMoMjz2cpKF8kNA1C6O2Sp5e\n1UN9ucrlXg++RI62Pi++BV6W9fWz3nsFW1bHX11LWMsQ7dHAL8SkuBSav1VBSGuD5DmUVRvwL1kI\nTz+NeuIMngtHCQe78VR7cK+qZPf6/0LM3YMnGWLn+hN4Tg4STnlRU0PscB3nn8t3QcRFW24Jy7KX\nJIzHjhBZY6OQ9JUr/z97bx7dxpXeif4KdbEVUSAAEhQXURS1kda+WZYt2ZRly5YisdtutzztdE+6\nM0mfM/POO+m8zpmZ1++dmT7OzCQvmZ4tZ3KSOYkzSbo7cbfcsboltyzLli1bsizJWqydEkmRoriI\nIAEQBaCwFFDvj4+XVQABEhSpzeJ3Dg9JoJZbt+793e/+vo06Y+5ccoT/67+G4+hRokrcbupYngsl\nnaZAoO5u6liLxSgqcfIkxgpI8MIQv/418Du/A8YYWtf1QwlcgSxcBItnqbMBAkJFMcL5jxwhwO3u\nJhCWJEAQIHZ3k3Z+6ZJR9LWYH2w0SlsuVSUe/LnnDM8agK795pv0cu9VYqIHVB7Np56VSYXvant7\nBdTVGXamQly43U7zVtfH+5NzT5Rjx+j7zZuJOinGlfPjP/6YsGnhQtrtb99uxHZ4PFm8/jrws5/R\ndbq7yX6WzdI837qVtPWKCnKD5tx5xy0b9oZX4ELIgjX+PiiODE6ft0LXM7j4cRC/KR4COr7Akciz\nuGhpwI3rjWhYJKIq04lGtwDJl8FpPAlLeQg+mx1SvR2RuAhJtsB3/lOgSzKMcu++i6Pd8xAfsEBy\n1aK1+zDYqVNg2Sxan41BEW5A1iNQEvMRT1jgXt2IyM/boIZvorXsBJQ16yCHe6AsWgO5zYk1VQIC\nVyJYa71GvLi3glayfftI8+VO9pwmef99CDy3SjoNOJ3Q5s6HYquAbE2AZbPEf/t8pIFXVdFCANDL\nKyujl1JWRivhnj3A9u1gI8PwyhqwcjQHit1OP6tXk/b86afko55M0rnNzUR59PUBR4/CAdD9JAl4\n9tncoslm0TTgJz8hwPeO7jw2bqTfPAKUR2Fu3UrX4fmJH0L3wenKo/W0s1KycBrk2rUkliwZb18y\nGz9bWmiu5hczBwxPFIuF/g8G6bfdTkqh02lkRdU02oUHAkaSPh6bkk4bsR26HoXH40ZlJYH4zZt0\nnCwTlni9hE8ffkhKHVdUMxkRTU8uxWfDaZxQF0JNi9CzOhY3alC6U2gf9qLithVx2Y0NjUFEQg1Y\n0ZiCbXEVtjxdCUdlGfb9yWVUsRRi0WZs/X8fB9PTkBMBKPuHgKol1PieHijBNOJ2D9xLBQyVL8BF\nxNCsJ+FIhsEsWXhrnUDPEORgN6Q9f4eQswpCaBhOexAsm4JXjACPLYa8cglk5wIsrWuEMPABdt/+\ne7CACog6NMEKZSAGWbkKtnKpwWsfPAj090Pkhr8nnoC2fiP26bsQP/4FpEsn0Rp9FyyTJCf8hgay\nJus60RjcJ9xqpRc6PEzXjkaJOunspJf28su0wjocuUCczeZmWmxuzk0sz42hg4MGX55vTeeLRHk5\nAXRtrZF8iwcoVVcbOcl9vlGL9SRa/pdUHp0nnZUpC2OAx6NP6qa4YwfNx0LFzLknSjaL0SLNdH4i\nQbv4sjLaje/aRZy8ohBOcOcHvsvn85uyH9LcPn6c8CGRIMr3wgXS5qurSUH88z+nNty6BXzrNQ0X\nnQkkEk689M8cOPZJBs7bnXjvRhWGkYDbqeNSt4T44Bp0DrvQVB/HSNttnD5/E05bFrsrw3DsboV/\ny3LEPjoBe5kGHDkC2ZkA++wo7BcvkuvMwoVAeTnkcgYpG8VQWsavTtfhxM0+lCOM19mfweF1Qkvr\nUKoWQvYy7NBPYU/qK9BTGg6cq0ZrxQ2wQIAegjFs/r8eh/azfwJL7wfLpgArgxaMYN/Ak1BsPghC\nOXbPFeGYN4eMi5zrSiYJjB0OKO46xIdtcA91ItIbhqIPwzvHZnDg585RtsKFC7mTvZH5j7v1/ehH\n9NIsFnJLjMeJw+L5UAIBeik8Quzb3wb+8i8pYvPTTwFZhkVR6FrPP0++oTzpfb41XZZpBd61i9rC\nM6Zx/++bN+kYvsUDiEe7l2lDHyCZBfJHXO6kiEUhH/FibpCM0VxctYrw4sQJAtu9ewkneJ3Onh66\nptdLSfiefJIAnl/DHBcSDguQZaPur9VKbtG9vTT3HQ7g5z830mbbbBn8xf/Thc0Lb0NwWrH999fi\n0pkURDGNbWuGUOOKwL16IX7+yROYN1+EmElhedUHiNjacW3Yi5jswU8OVOA7zyto3QqEggM40t2A\n9z7NQtKB1nI7MrW11KDVq4GREcq1vtWBi90yTvzPNPyWMgwPOdGfqkO9VcW+4fWIZ/yQAils1m/D\nEhyCu1ZCZKASijQH3o7z0EQ79olfhTIQxKWfV2JZ6lnImUVotb4PxV0HJVSOq9oihOO1QFULXmP7\nKZ/JpUuALENPpca0WBkKpCufI3TyGoR4DE7cJmPpjRu0CsbjtEr29hogLgjEwXs8ZFA9epReAM9L\n0tFBRsxYDPjOd+hlmSPDDh+mlzA8TC9PkqA1NMD2e79HHi4cvPkqn29Nzx9U3HDDKxfx2p98gMxk\n2tCHTGaB/BGWYh4nZnAvJIXcFCcKMuKatqIY+c653LpFCa7OnKGdNK/gwwP2zAF4ixeT1j13rh2D\ng0bxmdZWymoqSZSgb+5cwhge0l/py0JN6BjJulGujoAloli93gU1m4AtE0ZKdMJaVYab/SISGcAh\nZnAmVYlf9Tbh9rAVrngGyVsiMntk/M5vaWBCFslwAu4KGyIhkViAvj56gLfeInApLwd75RUscjLY\nZSDQ50K5HEFNNgElLiKetsHtZQhFrAjL9bAPRhHR50KKd0OWqHyZcjuOuB6C9dplKJkyWCs9iGc0\nKM65kMU4BKcDYXczPPUy9JOnoCxX4S3TKELq4kVkP/2UtitXr4Jt3IgdTTewR7RAt9pwILsLremj\nYENDpN0ODNCL4GXTLBZ6sWVl1LGXL9PKbbMZmRIvXaLPzp0jDn33buKx+/ro80yGXujNm2NujJl5\n82jLpKqGNjA0ZBSKMIfo5rs9mQeeLBsgziW/csgjJI/W085KjhTLjmgG93Xrxp9XyE1xomhNRSEc\nOHOGMGDxYlIU58whhe2FFwhvtm7NtVVxBYxnRf38c0rOFwxaUVZm5Bl/+mnymlu0iP5ft46OW7iQ\nFoqVKy24ctKHnqE0PLIH6SMuWJkA57KFsFmBsxdcuPSmOIZP1bVWHO+oRkgVoUBCOGVDMiTh1i8E\nONsu4jfXpCFZNUTiTkhWFXKlDGXuM5DmzQP+4R8IBD/7DNozW3HoMz9WrxYQr/LgK/gYzPsbkM+e\nhVS7EKFbIVyKNQBDQUi32/HCyD54hTBYxWKgvAxyRoBUMReK1Ql5nhdp1zLI0hXIZVVgWgK7tdPA\nvC3Q+7ohn+mA3HcKaKij9LOMIVtZSZkE164FGINqK4fF44Y72I2I7oVS2Qjv9ieokxIJavdoHnS4\n3QS2n35KUVWKQmDMq4wsX06dzimQaJRemCgafJmqUp7zZJK0bpsN6aVLDXDmPuKXLxNPn81OHF5f\nLAKuENf3kOZMuVOZsadsamr6121tbf95pq43K3dfSom+jEaFgucWSm/L6wKYFSVe7ScQINCfP59+\nvvENOubIEaPaFz/efC4PwLNaCZxjMaC7W8SqVTTvfT7CkY0bySja2Um7/Z4euu6cOYAuiEhJHkTF\nLEKqBR9+BDSmr6PeHYZqtUF0rIKmkSbv9wMLFom4Eq2G0BGFAAeyKSAUBEQhgxNXZLQ+6Ufr0g5a\nyI69D+WCCpGpwMs+aDYJCryQMwko4QziAxF4O77A+YMJHMhG4J8vodWXQmvFp+jpGQbUYXhFBRFX\nBZjTCeYUyJtk9WowQUCrEIVSHsaru9ZB/dlFyGfawU73AfX1cDQ04LVveKD85TuQOw+AfTEC9PcQ\nyH7/+8j88R9TR3Z0AD/4AeS9+yE9/xQil2ogLZ0PWZKApiaiT6JRo5AxN0bKMoH8oUPEV+k6vTzO\nWS1bRqtyeztp87w8W00NDZ4XX6T7O53EX+s6HB98QBkSPR4jefzp09TGRMIoBVWM8ysUAZcfAbpn\nD7XjToye97pg7gzJjLS0qanpOQDPA5gF8gdZ8gZpKdGXLpde7HQAhYPvZNmYP4zRjjuTIY3cYiGg\n5yUWeWGYI0dorpvnnrl9zz4L/NEf0c4+FMrg3/wb+v7gQcIZSSL72fHjdI9o1KiXoGmAoANqHGB2\nwOdK4na7iIV1dojJOJLhJKo8FviqrIBFRG8vEIza4SkfgVvvg9WZRW+2Bv4qF2CxQhseASu3Qo6F\nsa99MeJSJSyOLF5e+xQO3FiNeCAKye/CjrOfQjpfhsFLAehiHapsEcRuKFDCAXhtXahfvw7ycBMi\nuhPSUA/kzBfAkgVkUBAE4IsvwNJpeDc4ADUEhzUKvPgcIOpUqHj+fLCFDfCWpQC3BGRGc6mcPw84\nnaT9lpfTtbq6wCxZtO7UodgGIC+3g7nWEHB6vWS8bGsjMF+6lIyat26Rq6DbTd4pFy9Sxy5YQCkB\nfD76myfb4h4jfPD4/WSknDePVmFNg9jXB7zxBvC973FrOi0QvPDzwYO0yk9WxcQs5kErCHS9OzF6\n3o+CuTMkD0crZ2X6UmSQThR9Kcs0ryc4fez4nh46rlD0p8NB83nnTuMYc/QnY4bbcX5QEW+fLFM8\nCLkvpuD3k+J18iRhQXMz/ZZlonutVvrRdSCqZKArEVgyWdS7Ekikq8CsFrj0CHZuDkI8eQFJWwYj\nWhnYsuXo6hHRMF/E3NUudB8NImt3wHZbQU21DQ53PY6138RLyzugXO6BgkpYE3FEyuagX6hBPNUN\ntyONSDQLdSSF1meBUPQMfANRRNJ1EOQsnCuXAhdOgV0+jx2Lsuhf8gxqWp4B84xyTJ98Qisbj6JS\nFFoFeTaxLVsISPkDb9hAYNrZibFK1KkUrJcv08p3+zZd5/p1sOZmeNeNauCxDPlk79pFx7nd1Hm1\ntQTm7e30sm/epMXh8cepLX/0R3S/H/0IeOcdymjGU98eOEBtNK/2GzbQ/11d0MvKaHHhA8TrJeqH\n5yTWdfIxnUq+5UJc350YPaeT5/k+y7SBvKmpaU1bW9sHTU1N/3YmGjQrd0mmMEgn273mn84Yadmy\nPHEelWI1cvmcP3iQcMjnGx9xbU7QpShJKAphkdtNrsmCYDg6hEJGdGcqBYhaEqxvEILNhnhUQzKu\nwb9mPpKNc5DemMBu7RD2nF8CXyyKix1JXOuSkM0CbIETG1bEUSaEMNggITq/Fs3zVCSOW6B80Qln\n5yVccnwfirUCViGNb554B9KJQYQq5iOdVjBQb4fz8nH4vRp2bbRiz8jz0K934MClBFqXpQBFwQG2\nC8qveyGcu4XdFR/BsaCWIimzWYpmAoD6emiBEJSXfwtyOgh2+zbw3/+7AezLl5Nh4Ac/oEiqs2fJ\nAJlOG1kJeXBOPE7fnz5NXFJjI+0AFMVIovXee9Sh0Shdv7KSgL2rizp7/3661sWLub7hqkruQ9u2\n5WqyLhfw058Cb7yBeDYLR0VFrkHz61+n9oyMEF++e/fUPVDMg3ayFJ+lVmJ5iDxfZkIjfziWrEdd\npjlIS0l4NZ28/KpKoMuVNTPNaY4q5bz8O++Mpe/Ahg25pR39fsqFtG0bzduPP7Ths38QEI+nYXeI\nKK+3ITAsIrvUBbmGIRQXoYaSqKoWsajejppGcqxIpxmYsAZpNYFb7Q4kAyKGAk5sFHVo0QSUsmos\nE3phbRQwrGYRSHuwzb4HP1fn4p+612FPvA4rXAvww+/0IPDu59BDx+EVwojoDigJK+CugzKSxdXw\nHAS1esQ6l+I78+NwpNPERY2mndUsNuwLPIH4CR+kwQhalT1g0TA9rMtFAFtXR2C9YQPw2WeAKML6\n+efUibzckssFTRehZMshg4HF4wTWjBHwj7otYv580u7nzqVzv/51uvb167RynjlDW6fhYaPm5sgI\n/X7jDTrn+edzt14eD/C97yF57RrGIszML18Q6GUnkxP7s5YixTLJFUuLa15UHtLiEtNq6ag2fnj0\n3xLyR87KfZNpDtJSTi82fyYTRSG8mTOHMIB7ovHqQPmgvn+/A1evEobNmUM0b35WU/MO4JVXGZ55\negF+/baKs1clXGkTMXcu4LRr0N45iCNX/LjSX47LriV4wifCOepk4fEAO3Yw9Pe7gE8Iy27cEBGT\nn8R75yywN6QhiRbEVzyBa29dRtlPLyI5vBkRWww6BDiSEQwpGv7ubRcsR4DObAQrLFchv7oD8lIN\naGmB8GE1gr3DCIRFnEs3Ys/5NF5bsATsyAdjkZqKNAfxtU/DvWYhIj+/CWUkC28kRHSI32/kGz55\nkjTqkyfpRaRSRMFcuwZEItA8ldh3vQHx6zqktITW+Edguk7+3itXkvbNgZ0nzJIk6tyXXjIq1nd0\nkEvQxo3UKZEIgSOPAv3bv6UXV11tJLUfHTi6x1N48FgsdM9MJncw8bQDMwGsExlFuXtW/iB+SIyf\n023ZgqampkYAFQAqmpqaVre1tZ3LP6ivr2+at5lZURTl0W7TFDK9FWrXZKdTNLcAl2t8VGix49Np\nO+bMEVBdDTzzTAJ//ucuRKMWMJbF8uUavF4dvb0Czp5NIR7PQhAiOHXKivr6DNzuJF58MTnhvYaG\nMggGLVi0MImhoAPLl6cw1BPHlXgYfaoHte7bGBS8OHdOgiTpcLt1bN6cGCt0EYvZ8e67dsRigKc8\ni5a19QirAp55Xkcs2IO+ZRZc6mlE2GKDqCjIiioiKTcqZOBMcA7k7AJoFhvWWU6i2s4QuZRAWAGe\nkR3offpFxM8B1jlNCNXH0DH/N1D/xWmIdjv0RAKZ5jok6psRHlbh1UYgunQkEjbolZXQqqog9vcj\nGwiAXb2KtMOFeJ8KVywJQchA+fhjiCMjyKTTCF24heHelXBnwggPqxiKxeCJnUY6m0V69Wqgvh7C\nF5cQDmgoHx6BzeFAtqEBWiQC/P3fE3+t69AXLoSlowPZZcuQ2rgRttpaOA4fhnj9OrLZLARFgZpI\nAL29SF6+DNuJE7CEQsh6vVCeegp9mgYhGqU6maOBC/Zly8aOSaoq+aInEnDs30/3dLmQzKdspiqa\nBnsiASEUGgtC0svLIfT2IlZXh74CNIz90CEIqgrd6Zz+/e+iTKtVbW1tvwCApqam7wIoL3ZcbW3t\ndG4z49LX1zfbphJlqu0y717tdoqeLiU+49vfzq0tun79qLeJQNexWIg9WLMGuHx5GOXl7rFI70TC\nNUa7FGxTQsOn/+1z9F50I2thqG9egPJyF+QaNxpVH374q/m4NliOIVWG0wlUeDN4arMIWfbB7zey\nMF64AIiWDG4pYWzQB1DriWPx1WEoI1mc7pNx0+JDldiLBXOH8dsLulFuTcDlAv5Lz26IbjeQyqBq\nxXyU/+ZvYt9fDyA+UAGpM4D/s/lXeLtqKXRmhxzWsFAdBJszh3gmtxuO3/s/8O2uW1BOXYUc/znR\nKk4n5RYPBMasutqJ0zjQtYaKHSczeG5FN9w8G6HbDWd6BBUshbi9Bh6hB5VODSybhSMWA156iQpH\nHJQQPxOGlGlE69IOsJUr6QXwpDXpNL0Um42MGStW0M+rr9KLP3UK6OuD88YNCsH3+6njMhng1i2k\n1q9HzenT463mv/3buZqvpgH/+I/k7eJ2k5+501nc0FKq8IFmDoCoq0Oiunr8OA+FaDfCfecnGmR3\nQfr7+0s+dkaWl7a2tr8C8Fczca1ZebjFnNb2o4/Iy8Tvn9iTK3/36nTS/OWf/bt/R/jBv9+2LTk2\npxOJApx93gWV/iiSMQ1b1kYxOJBF6+9WwVHtgSwz3Li+A0pZFhaXiEwUiIeT0GMZXP0kBi1RCYBB\nUYgCtloBIZuF26biCXYKzeHL2PM366E7JLhCV7Bh+WPQxUpcDq5A2qXAb1ewa6uKzT/5DMENDfA5\nYvD+0X+B8svDUNrTsA6fh5LOIn39JF5bdhlK5QLIlXZo3nr0PPVN1DyzGI5Fc4HvfhcsFIJ3eJi4\npDkV9OC8IobNBrS3Q3H4EdescGdCiCS9UDQn3G43rX6pFFgzQ6v3GpRBFfK1g2Bpi0Fb/NmfQfn2\n7yN+Kwj3SA8ZtMuC8B47RpFX+/aRRr54MRk+6+pyM6T5/eQjmk4bkVotLfQuOjtJA7ZYqKRaIat5\nPqXBw4DNnLzPB7zyyvS04mJGUe6eZZaHyPj5YO4TZuWO5X5SeppGP7wijyAUzoiYnwIg361RVSnW\nxFx8Od9F0u8vUmbOtCXQ7GUIbdoFjblgL2OIhVLwV4rwN7rARjl1l4fBKQPCMGXqs4saKj0ZNFcH\nwRIOAJ6xaPCaGiARt2C1PYRmsQNvB1pwfMAPT7QXDXoIz1j2YGTLV3HauRDtLhvaOm6gpe84Xlp0\nGcrj1ZDTEbD9e+H84jIuhXZCGfFDdmp4deRXYB9dg7euDonH1uCHB56CkimDHGjA669chENRqCN5\nKDtjlNOFu+udPg14vZBtSUiWBCLMC8llgcsjEui+8golt1FVsLUr4e3tBfob6MUMDZG2/dFHkHe0\nQrKoiGgSpMwAZC0EdETJyCnLdK+REXJFvH6dAJtnLty3j3jx99+nxaa8nL4LhQj0R3MM6zx3aQS4\nMAAAIABJREFU+GTgyEP1a2uN7GrHj1P4byGt/E4G/mRGnYfI+PngtmxWpiz3M54hn1JpbSUFKj8j\nYn4bufKWnyaAUyyCYKS55ecPDQmwWg035BwZ3RJoZeXYe9iNox+kIDgkbHx8PbauiQIuF0IKg9fU\nL4sWaAi1B+GvTqGB9WDVKsBXKUKucUFLaFD6o/jqTheeeYYhGhVR718M5adPIfpZHdzOKIJBF8Kp\nJbjUvx7JIwtxy6ZjyUoLhAULgd/wgZ13wDvSS5qz3Q5VqsCy5FlYMyGkRwSoMuCYVwc89hj6vUuh\nSNWoWOTF8AhDj2MRKu3VcA4OQZ23CvJ/fR1M1KE5XFAOHIXMUmCXLgF1dWDVlWitGIDSfhuyR0Qq\nOaqt791L+Q10nTpwwQKyFKfTxFlZLFQm7uPDaPWfguJ1Qg50gjlcBPKJBK2u6TQtJBs2kLGzpSU3\nlwLPz7JyJf2+cYNC/AWBKBe/H2U/+Qnwp3+au8XiL9YMmIyRZfuNN+jv7m7yppls8M30wL9TC/49\nllkg/xLJ/Yhn4PMvkSC6lmvg3NEhX5nJryhUKGkdn8N79hD2HDhgpM84fBh4/33XWK4VcxELAGPb\nYWVQRTBdC9FhByxAMCjg4GERX7QJyIBcp10uUkjPn8vAwdIoKxOxdglgf2wRLHU+7H9HBz4/hWQs\nA3sZQ8v316OxkQFw4Ejl19DGosj6bmGFrw3JzkH0ls2BmAhiLhMxbyiK+p2r4W30AJdEapvTCYgi\n5EVzIJ9PQ6mpgZDR4KwZBGTiuWtGrkKOrsbwyWGUrVyEM20uJL/6V7h0fATLniyHfMWJHTuAA/s1\nxC/Ph6QOodViAyQ3FKsf8nNr4W27BDz2GBLnz9N9IxHg+nXKX54tg7ymAmzePCOgiHtxdHWBldnh\nLXMCX/8ugfDnn9NL8vlIy3a5qNBEZaXhxcJf2k9/apSUW76c3CC7uymy0+0GVqyAZXiYBgrPisYH\nUSEg5q6Ba9fSOWvXFh7QD3Egz0zJLJB/ieReU3p8/ikK2bNSKYrn4DvuQspMoYpC27fTPK2pMUBZ\nVUnBM3uKqSrFsgDkSRIMFpizo9thOaTAVy4jc1yEkM7A2XYG0YiGRL8ft6yNCAZFlJVRW4bCDFbV\nCasYA5NskOp88FaK6G+PAUGgqtqGj864EHwzAV+9C6tWAfEkw5ZdMgbfT6G1OYODZ5vQk12PTN8A\nnrBdxA52Gv7UNTDlRdJiq6oI4FpbwRjDDtcH2PN3Mei6gANzvoPW32sEG74Nx7lzeN36C/T3ZeFo\nXIXD8a/CaslAsfthlYB4IIr+HgfiSQb3lrWI9McQWpfA0bMS4ovdkKoWoTX4I7CODrCLFwl4HQ5o\njYuxr3sF4rBDEneidaUFTAlRuxYvJq+Unh6K4nzsMer8gQEja9ncuQTkTicFHR04QFGhr71GDvsr\nV9K1HA7i7gcGiGLhGc9cpN3rnBcza+DFgJhvzZYuJa3eHCyQP6geRC77HvKcs0D+JZJ7Tenx+ccL\nyTz1lGHjmiiBXUsLATefw2+/PT7HUaH0GVVVxr0yGaPgRMGbeL1o2QpsegZgUQXOIx14/Z21uN7r\nxJAlg6bHxLHMjIsXi5Bd5XhipQ3f/m4NDn0oIhQCbC4rnI4UBnutEESGijonjh6jtnd1AcuWifA/\ntxL+lkZsbVXworUcB//wCvTrKt4PrsVuLU0TLH/l2rULalqEJZNGmT2DwYgNocNn4beGgS++gENR\n0Di3GpoUg3TuGJSkDfKAD+nPIpCtCdSUhyHZdyISESC5BGDLDsQzKbhlHRFFgNK4Et7Lx6DLMnXY\n0qVQbqcQV+fDPUdCJJSC0t5GJdtu3SIXIbud+OiuLgLsEydo4bHbaYVeuZLolBs3yMDKE8K//z4B\ntc1GIN/VRUDt85HGbjjkA7EY0qpKqX6dTiPv+csvFwbiyQa0GSgfNC77HvOcD8ATz8pMynQovakq\nEFy7DgSIKslmaQ5Pdv/RtBsYGSFse/xxUt7ynRjy02fEYuTRtmBBFPX1clG3RvNOQRCA3S+7oNpd\nWDEngMdqIziiVKOxkbAmGNBgRwJ2twO7v+GCw0EK5t/9jQb9/EU4bTpa1/bCV7kew2ERuk47h7Iy\nXteAgTEvWFqFw8lgeWIDrg64EY5ZgQtRvPYtL1h+LbyeHsiffwS7UoOPepZDSDD4mAMvvVBOATq6\nDoRClORq9S0oDj9eHb4CNSnA6bZBbQtjx2v9UI+dgawOAp+UQ+qci0gsC6nMArlZA7JZCr5JpYD9\n+yHX1EMKOBEpWwZp5DzkW1eApoVGAM7GjYZGLoq0ANTU0PfBIIG7KFKu81u3KChIEEgD7+42wvTn\nzyft++mnSZN+7z36bHgYsFjAurspJJcP1EiE7r97d25OBvNgLCVKs7V1Av/T++ABcI/pnlkgv9/C\nBxmvt3gfm1HI/7sUEUVS6rZuNc6ZKBhPVSlVxxdfEEZ88QWBubmmADCRp5g+tjsvdB+uaV+9SvgC\nMOx+eTukjIqjZyR4vCIcDsCia5Dbz0JIp7B7yxAcbCc0jeGtt4B9v8zAFqrE/Do3tq4YwkvbYgjB\nNmbALVTXQJYBwW5DuG4ZPFIS+go7FFWEl7vnccuvwwEGDS2VlxFMuVC1sgwxsRzKzT54RZFWteFh\nQNfB3E54k8OAXwY7fgr7fplBPOuA9NHbaG26DibZgHgcrctWQamYT7lYtr0A2CxILFkCSdcJQOf4\n0Zo5BaVOg9xcB3ZkPnmT3L5NhkhJIu+Wjg7qtDlzqOpPZycBfHU1VQj64AMjb/hLL9FO4+RJOj8e\nJ8f+eJxW3GPH6Nxbt2ihUFVYIhFaIEIhurfHQy9SVXPdmibTZvOBsljh5fvlAXCP6Z5ZIL+fYhpk\n9kSCghXu07bQbIQ8csTw/y5UWMJ8TjJppJ/mTd+7l873+QoYI2HkKG9rI8WvvJxKuzU2TkzJ5EdO\nF0ubwa8fDhseemqaoWWnjGCKmIGbN4FUKIm51hgiVhlKMA21JwrN5YGqAjaJQQtZkY6rgNMJ5pXh\nZ7kGXMBYSHgbd+8GABG6LuUuTLx6jSxT0im7Hd75FvibliO27hma6yt9wNE0GQn9fuqcTZuIonA4\noLQNIi4pcKeGELkxCEVth3fDEkAUwYQMvOlBur7fD7z2GpKXL1On/dmfAcPDYJUeeBdVAIkoAe2q\nVeTSxxPYHDhAQC6K9PK4H/jQEFWy1jTivjMZapvLRSlnRZG+u3rV2JbxlJZbthBFs3UrIMtI8SyH\njNG9VZXOM7smlaLNmvMnZ7NkBeeWczNY3y9D6D3mOWeB/H6KaZAJodB9tbbzeTE4aPDRsRiF2hfb\nmRZSOkIhUsR4vd6WlvFuv4xRfpTPPqP5q+vG3C5VJqslwHOgq6rRNk0jKvfjj+medqsdZeky2LMx\nvHN5PiwOGZJMGDa/UUS6tgpbnnDB+42FdNNQCEyW4fWycQsJX/AcDrL/jfUX8g7kFTBaWsAGB9G6\npQGKJwv52LtgpyIEaoyR25+mUWNHnfPlBX5IbBiRgTQkFoAcHQBCfupgSaLVi1fY0TTYTp2iBq1Z\nQ4mqmpvpf/NKdPFiriHC6STXwEyGFhCrlUC4v5/+7+ig1V4UaaX0eo0Hdjpzk1BJEg0iv58+U1Uk\ndu0ycqhomhEsoKrUbn6dybRZs2uTqlLSry1bxgct3E9D6D10XZwF8vsppkGmO53TG2RT4QELHMsV\niFAo1//b4dCL7kyLKR36aPq0TMaY6/lN8vtpF84196mO90LG0HxfdF5tjLtCHzpkzPPt24FEguHJ\nx9fg2GEVJy9I8FwX0bw4gx2bFWx9hvKAOJ0yeno01Jx5Bw4tOtYJisJyKymNZMbUc8aY8TyBkOGX\nGYlQrt7Ll2mVs9vByo7DK3xGD2C1UoOfeopWuWCQ3IE2byaNurUVrcuWQfnD/wY5dBNMqqH6dn4/\nuQPyStiMAT09EEZG6O9jx8gg0dND4GeWzZsNWu/jjykUPpkkrqymJheMUynycBltO44fp4XAzE/z\n7GWalrsLGQ2HtycSwDe/SS+BZzuUZQLvYlusYuOZuzbV1FBisMFBaqfTmcu5PWiG0LsgX86nmq7M\npHFkomuZBllSUe78XpPxgJOFUprAPD9i8to1YcKdab7S4fXS/A0EiF41z3XepESCFLzt28fHhZQq\nhYyhkQgg2TXImgIlJCMeZ+jpocWEl5Pr6aF7Hz5MEeVgDLokw+MDwsEMhHNn4BWvA5KEwBO78Pqf\nMsSCaciDC/H6d7rhiIcBRYHT6UU2O4oXkoaK44eAMvv4KtZHjpBx7+JFKqsmSaQ53rhBoMdpDUEw\nOiMeNzIPcs8Qux3w+cC2b4d3fjkQTAO6lYJ7zMYCDojDw5RwiqeY9Xjoftw5n4NxOEwvasUKut+/\n/tfEg1sstPJxQNU0Mlzu2JHb9kKDIpEw7iPLOVFfwtAQ8OabRs7iF16gc/MpEDNnXkz4ah6LEQXV\n0pKzaOS8iy+5X/kskOfLTBpHSrkWH2RTyEg4TibiAUsJpcwb5OZx73LpU9qZslEeuaeHCt1wnOK1\nPDUN+OEPjYXi9den3r1m3MoxhoY0OA+/A+XtFJzlNgjZnQiHGTwewqxkkp5j1Spyi1ZVst+dO0c4\nxurieF6/Ca2sHAeOutB1LYkLFxjWrbEh1OtE/800GudL0Jwy9u8nTd/pBHZsUhD/RRyo8eduCXp6\nqK83byZNPBol0Fy0iIyJqRQQCkGT3FA2bYfMVLCvfQ24cQOJX72H/lMB1LjjcLRuI2MjL/yQyZBH\niKbRQ1ksdJ+vfY3uPTAAtLVBiMVoW1JbS8ZTnh/4+nVazfr6yDBx/Tr5jgeDBp2SD6icf47F6EUC\nhQdFIkFpbM+do61WczN9zgdRNku5U3jKWnPF7TspJpGvbZuLyD5CwUGzQJ4vM2kcuVeGFu4H2N8/\n3rk6vw3AlCbMnexMGTMqBuXX8ly+nK5VUUHY0t+fG+SXL/kbGk0DDh2yw+EYX3LOmlDwtwfnwC7b\n4bFE8PL3FUD0jimG27aRzzrHP57PKZkkTJW9Thw+OQ/ZSylooh2+agdEERgYFFG5ogE13/ABfhmh\nEMuxA+x8UYbF6TT6lGvFikIPX19PD+p0Uvm0OXPoxkuXQstasDf9GwjuFeGzJ7HL8R5CVwfwXz9v\nRcxRCfl2DK9H++H4y78kLb293SjCEI8T0Pb1kWb91ltkBDh1Crh5k9wPdZ1WrSefpAIRb79Nx7rd\n5E3S1UXGzH/8RwLz8nIaS5P5dPOxlR9mv2cPgXggQJ+ZSzcpCpJdXWRp5t9NZ6Dx88xz6kENDrrL\nMgvk+TKTA2EigL0TuRPKJ/95TJOqVD6dKSF4p8h/8HmZX8vT5aLbDg/T75qa4tfI36HzZquqMEY5\n87UxGgX+4A9lXD/ZCK+UwK4NGaQdcq7hcdSrZM8euv65cwTiPh9hykiUoebpVRhqD+PcTS/Sn4to\nbgZ+93eBxkYGh8MADG4H0HV62OS2bYbLDF88vV7K/rV0KXG4mkZaaHk5AajDgVCY4ejxNMT+LqQV\nFVGripBvDS6qaawt60FIK0P/pzfQ2NdH10skiKKRR62yvNSa203g2d5O34kikk89BamhgSzL3JOE\n3GqoLdksddySJdTmDRvoc7OWbH7nhXg0s/CMhT4f/b96dW40ptcLXVGIBilkHJkJCuQR4cTz5dF4\nyqnIgzoQJqJp8v0AzZp/secpZcKMJtYfp/6WKGbN3FxY/fXXaW2rqRlf2cf8uHv2EMfOiyvzR3A6\n9Zx1VtOAn/wEuN7BoMqVgFVDcmkdZO/44tLcPlbpzWD1EhVJOCDJDJIEIKMh8skXYMkUVos2WDes\nhRIXUVmZ205uBzBjUZ/KEIIXMgBmdo0TBGr8M88Q0EoSac2yDC2hIZytQFYXIGpppG1lUFNWzJNH\nIPrrMGDJotIeQk3kKt2EGyyffJIUhJ076R6KQr6cw8MUnPPYY0BTE5Jbt9Kx5ndmdqvZvdvYonDX\nQUmi44qNfa5MmD1U+HF8IWtuNkLqC5Vueuklo7D03ZBHgBPPlwcEpR4wmamBMBHAcjEHBE2kcU9E\n00y2i7jT51EUCKqKcepvicIfJ98BgbGJ6ZTRW4+5JfLiyvz8bduSY5jBaVG7nTdNxOLFIr71W4Vx\nSJbJIBr56Cw8Qhw7Noahbt0J2cuAkAJlpAvOCgn7P3bh2Gcp6HYnjhwh7OHtkuVcg7CmAXv3OsZy\nULW2MjBT1i/t3fehbNpFHLhMAKhZndj3dhqK3Q67NY0FtQn4nEOApQ7DlYvwm7+n4/GTe1DvUeA4\nFSJwFkUC5jlz6KHfeouiqeJxAuHycqORL76IbEND4U4wjwcO6lYr0N8P7dRZKG9/BNlnBXtp13ij\n+d69hiV7xQpjq8RfbKlK0NGjpRnnHxRF6gGX2V66mzIZwJoDgvhWPJksPLgLXcusHXFXr1LK8Uyh\n/bqZ+50CNTRdm3Eh5c7s9phPi3o8wK5d1H3f+hZROIWEMaC1RYES7IJc5QSLReFgCsC8gFeG18+A\neBgta4CgakNVDdn3AgGyV6qq4dnDbYB79gCnT9tQV2fsHLwg1V8rK8e+IzLiQQ2S30v94HVACQFx\n3Qpv50msyGTw9JoIan5jDfb/zxtAVydsjgQaKyJgdpF8z7/zHbr4oUOGzyU3kFss9OChEGUIbGgg\neiSdnryjOYWybx+0gSHse8eC+OKVkBBHa4sC5jd1NA8SSKXIn3zVKnrB+TvAyRb7qRjn72Uu5odY\nZnvobspkGoppQFt6ezGWyKOYn1++sclsUFu2zNCOZrD9OdzvFCbUdO28U1HuJjt2XKprDtixkaIG\nPa9Thv+AiFiMtP1f/5oA22aj1CE80InvHNzubM7OASD7iHJjCPGMH2UVTgwGCAt5fIwkxBEJZyD7\nbKiXw1BGUkiqWdTU2hAJiFBWbIDXGiX+ZvXq3Ae1WoGf/Yy4d+7FwqtxcHceXvWmEB3CX5KJ01fk\nWsQzQbhTw4hYK6BAxrhXpuvUCTx4p7p66raffKXE7PdtDjEeHDQ6bFYmlFkgv9tSKLacTyDTgM7y\nKgmTRbOZk5mQ5Y9yVqxYMV47uhvtL0U08uWW7DIiETapMl9sJz2VW5uPNeOWopArd+5GZwLkH70Q\n0zS0bg5BgQwNbMwhhApFG4fz17hokQafb3TngFH/yEwGsiMNuzWDj44AgpibtqB1txMKwpD1CJjs\nhLxoDiS5D5HhFKRyK+Rvfw1I5/HQZg1aTUOpWAT58WYwl4O2JIyNAaMwNEQ3PHCA+PNTpyipjcdD\n14rHc7IPysoIpOULEVmwHJJfIrrJLF4vUTt9fWQYff75O9sBFgoCMAcC2e1UJ1AQiud5mJUcme2d\nuy2TBeOYA4Jqa0vnBp1Oivrjmad4xjnOtd8vznF0a8zicbTay6C8sGvM6DjB4RPvpKfQfk0D9u3V\noAQSuNTpwPwFDNeukQadE71dZJXQNPJHl4/sB0vG4JUkaDta4fczzJ9PIL5ly3hb8rVrSSxZQhcI\n/eO7kKP9wNWrUB7fik2NfQimVVTNcyI2rEIJOeD1MzAHg/e17WPPxhhD6+uPQ+mPQq5xgTkYgALW\nYEWBpqjYd3Ml4ucVSL0ZtK69BvbaaP/s3QscPQoXD5TRNOCXvyReu6+PvFhEEWPRUgCwezeYqqL1\nVRmKyop3tSiStsxLNN3peOL9n+/3rar0sszZIh8RX/DpyCyQ303hxiHu3tDSUjSBvsArZhczhubP\nLFWlqL543HAPUVWKvsuPLryXnKOJU2GRCLycf5788MIUTF4NTqVl14SJIpWQhvjRs7Cmk1A6aiAv\nnw9dF8eit8ftDEz9q4HRrQIJSFe8aG1hYLERMFXBSy95x3JIFYrp8njIH3HfHhXx4x7YXX7guohk\n1AG7zwffWhtin5yBJMQh+0IANyTmLSjMwaiqUL7k7eSUbBnifWG462REGpZDWVAJL+fMg0ECXIuF\nxkQ8ToBYXk6/ef28nKxiFPjDAHiLeBKVZLwvJPnKjFmK2ZHM2SLN9qB8v/VZoyiAWSC/u5KfQWrT\nply3NKdzDKjsvb3kQmYG2olAWJZpsNtsNBErK42SOuZJdq+CkrhM0Q9/0sNH228YDRNIiPaiiSJl\nKJCEOBSbDNmqIhtXsXmzaywtb845ef2rbG5FPM7grnIictEB5cYNeOtdo9oynT/RmqgoQFyX4PaI\n6L+tA3NXouaZxYhkXXhhvQKmcANriVqmmSPi9IPdDmzaRIWWy2yIhGRILAXZ7zA6z+cj3pxnInz+\neQrCUSmT41jniaLhpF8Kz12K8b4Q2BbKLsalGM1VyB5k7vhCnz3CYP7oPvlU5U5Xf3PkCGPji1GO\nhszrLtd4jnsiww/3x21pof9l2ZRwxDTJ7nWk2xT98Cc9fLT9yqCKuD4H7ionQrdSRXGQeWW0bgpB\nCQ7i1Y02qFubIRdjAPIWORkKJMlLXWVNQc6EgYyz2OHj2uB0AllBRGjxBviWxAGnExGNQZIBb70L\nrJCBtZiYATCbNbKCffQR0N0N1t6O1ue2QRkE5PVusIbHjA4dHRfR/n7IS5YQgP+n/2TU0+O+3fnR\nUpPJRC+rmNIRGk0aVlEBBAKUyKvQdfNfZr49KL/jgUcyFL+YTBvIm5qavjv658K2trb/e7rXeyAl\nnyIp1fhSKHJEUXI1ZwCQJAi9vZTo3zzBeWSouUyY+d6M5Vr087UY7gkwjQCnYsUbJpQpGkgnPJwx\naDtaofVEYS93IRITYbPpBU0B/Hj20i54R5/XMVGjZZnomv4EZF8ZIMuUDHAgBHbsINCbBHq7yZXP\n758wUFfTaB3VdUBgInbtpi954BNzlLDAFatjyXdwg4ME6rdvAzdugGUPwutyAddOkHa9caNR3cPv\nh66qhRNIldTxRdpV7JxCq5wsU3ayzz+ndACLFsEGkE/8VH1RCykjj2AofjGZFpA3NTU9B+BQW1tb\nV1NT08+bmpq2trW1HZ6htt09mSo6hUIUwCCKtGUtlGS7kHDtyDwJ8gflaMh88to1CpXOn2hTMfzw\nSVYotv0OtJVieU1KPbeUtWOy4zQN2LufIRj0oLyccGr//vGmgBwpEaA0MOzDLsSRgD3jAPZTatpL\np5xYdqMJsiON1vkXS5ok4bAwlq2Wv6bxMS8TtKtQpQzzONm0ifoq7YJ84VOw1atpDOo6rRbZLGnr\nIyP0eWsrhGh0+lprMU07/8UVAttQiNLxMkbc+rp1EBRl5nxRH8QI7Psk0336BQAaAfw1gM7R/x9s\nIC8Udg5MPiB4gh9zop8S7jXuukUGpV6sskJ+mbBStuSFYtvvAMjJuzEvr4k8OUKXal8t5bj8NXTt\nWiCZFFBdPUM5zZIM7hoXuK1ZlgEl7YB16WLElRiUtRVUqg3FbX2aBpw4YcPly5SxdtMmulbJGKpp\n5EHCD+LeGzxB/JEj0A59iH0X5iNe/wIkazVam3opSjSTIW03mSRPEtNKortcU9NauYIDGAaFYpp2\noRdXLDm93U68UzwO3W4v7Fk1mUxGvzziMi0gb2tr+yvTv2sBvDnli9xry3N+2DlHionQxOs1Ev3w\nMOhEYuLE9xOh1FQdpKeieRSLbb8DkeW8vCbO0hC6VPtqqceZ11CXa3yulVJlIiWS53mKxwG5XES6\n+XHILA55t3PsGYvt8AnghbGqZi0ttOBJWRWRkARJFou3k++eeL4THtg1Fr1E2qxi8SB+oR3ueAQR\n0Qtl8w7Du2XrVjr/2LHcBZ8vBman+gm3PnvpGrpOlOCuXWPViXIeuhi451/fTC1u3Ej5Zvbtm2Q7\ndZflS+rpMiNP0tTUtAbA6ba2tnNTOvF+hOPKeWHnwORowimSUe0IBw6Mj6bMb/dkKJWv/eRLKbxk\nkeeDLBeObS927SLCWF5ekxKRt1T7ainHmddQn482J/m5ViYT3tXjA4MKZ2b92teAQEBETY0MZnLF\nK7am8gWPF9LxyhrYgX1o1VUoghvyju1EqxRqWP7u6emnc6s6j3aSPNAHyWpBxFYBCXHIHtE4htN8\n+TQeb3QxDdosikKdbLHQ/4GAUT/PbjcKQBSiUfKr+5hzr5jbpCgQUqmpuy/OlHyJw/9n6imea2tr\n+0GxL/v6+gp+LoTDsPf2Qne5IPT2InntGlEMdyKaBiEape3kJC9H2bgRKUGgY1UV9kQCQigE3elE\nMhSC0NUFIRZDtrY2J3ubEA7DPjAACAJsAwNIzZsHRCK57ebtcDhyr8ujMEePsR88CNvJkwCA1OOP\nQ9m0CWO9NEr/CKpK527bNrUBt26d0RfB4Lh+msq1VVUBY33UdE0r/kzjm4BoVIDLpY9Fit/pcRs2\n5B6T06ZJhPP8w8MWXL/OsHFjCqGQgGvXkmO+3/ScucerqgCnU8e2bcmC3aMqueNt40YFgpCCy6Uj\n0BmGeLMPCvNC1roR6Cw8roVwGPZgEEwUYentRaq6Ggldh3DtWu44XrcOwsgInsJniCnnUOYRMag2\nAn1FOmD0YRRFQV9fX2nzTNNgB2AbNcCnR5O06+XlEEIhJAMB6OaQVtMYEzo7J7/+aJWhLAB0dxtz\nrb+/pDk7EzJRP/C+elhlRrxW2trafjT693NtbW0f5B9TW1tb+OSqKvKdjsfJYyPf2FeqTHGl7QNQ\nU1VlaArf/rax/fzlL4Ef/5i4x+XLgf/wHwyu0Oej9iqKkWNClo1257fjm98sTL+EQoY72aikBAE1\nvJ9CIeI7JYm23skktW0mEmKFQrQ4cWpJlifUivr6+nLfH++rSThyHqh6N+bnuDZNIPxxlyyhoMZM\npvBQMyehnLR7Rt+zpoxq3Lu3AwBqaym5upaowr6+EcSVDCS5Ea0LloxGaeYJH/9erxEqf+hQ8XH8\n2GNTogXG+qnUefYv/gW1ATDcWSc7R9NIe3e5SHufZB73feUr8NEWZmKPmrshE/TDVMZKZDXcAAAg\nAElEQVTUvZJ+brgpQWbCa+X/a2pq+rcAvAB2T+kCU+V/i8lUg14KuRPycOG+PqOeYTBIRqiLF8cX\nhaX996hvGRvfjlCIvAnM22QuspwbtGGzQTfnbXY6iboZGaGyXadP0zU2b55a3olCFMp0/cpHaR4K\nZS/dnTi/SYXSWU9XJnrcWIy6r1BgkLnNdnvhAjk5N9A0CpG/ugjxcAoSVKxrMQ5TVIb4sg1wW1VE\n0k4oqmhES+Y30jz+JxvHd2rcK3WeTeTOWuwl7t9PvDq3RO/YMfEL5c9g9g031wK8m2A+U3jzAMp0\njZ0fAKiYXgtmwPI8RXASwmEqKKlppAmtWkUJsmWZ1EhBoJe9cCFpGvm5IIpxjrwd+fXNCvnvvvQS\nkb8HDwIWC+wffgjMm0ffqSrx78kk8Omn1E5VJVA3u0pMZrwqVJW8UMrbKRqAJgLribCIn6cok5sY\npirF2lTK3M1v8wsvGK9z7Pg8tFeyZYiHU3B7RER0CdFodOx6sgxIsohI3AVJNg3HYo0sNa88pmGr\nm8z1sdBFC52TH6gUjZImbrFQeltVLV4txCylzpWZli+pp8uXY0ma6kqbyVA+5UyGtOrmZtpqtbYC\nX/0qacK3bxOQyzKMcunyxJoTb0d+fbNi4Otw0ASwWiHwbIajuVfGJrHHA3z8MWVrGh6mvNSlUEn5\nuwPuV56PoHdgAJoIrCfCIn6e1Up/W60zl7BxojZNNndz2mzX4IVCVX7M/ZB3A/nFFkiiAxGdvFJc\nLn3s/TJZRmsrGz8cS9G4JxjHd8VWN9WL5o8rp9OIPJ1KOcPJ5sqsTEkefiA3g+NUXPp4wqlMhsCS\nI4qmkTOwxULBDKkU/S0IxraRD95QiAY/94sF6Bo1NdSe/CIQ+ROGUyiKAsZzr/D28Qn9+ON0H5+P\nzg0EjPZORCXJshGGaLONLRhQFPo7MJocm7EpB41MBNYTYRE/j3+XTueuWdOR6TBGY23mWQ/fixXO\nbcNvYLfTObuJNpFlYLAv9/2y1lZ489PAltLICVadqTKIOVJM686/KB8TxRQi8zPIMs2JnTvpu6na\ncBgbXwtwJgbDIygPN5DfoYqil5cbNRS9XgJljijcEAkQ0qgqAfPgIA16wPD7FQRaCA4coGvwJFic\nyuDHA/Q3V0P53wAtKMPD0HimuvxtaU0N+a7H4/Sbc/JTQS2nkyJqeN6WU6eMXM+7dk0ZASfbABXD\nIvN5r746sxz5dOlPxkCZGpOx4mq9KUAH770HJknwjo65kqIoR6+hhRQokCGDTWkC3vFiNVnyNdMC\nVdBHcyJen+8s71S+xLz1vZSHu9fuVEUx+7fmW93MQQw8+IcnuXe5aCJfuEAgyFPH9vQQFcMY+QFz\noDYHGj37LBktk0m67quvkkfKG28AkQgkh4PKtfPt6oEDBpm8YgVNpm99y5g0pRC/5jBEXhl92zaa\n1DwC0Bw0MoWJNB27Gz9vOvN/Jts0hlNOGWwipOSkezI5bsyVGkWpgWHfUe8d0SN3jHmT8U78oppG\nwTqlRHHe7+Ils5IjDx+Q5+VlnnIIMh+wxRDFDPJcQx8Zoext779PwMe16SVL6Hc4TN9fugR8+CGB\n/Jo1ZJzkQPrmm4YbYXMzXaejg367XJSD4m/+hrh6vhPgfHxZGWn86XRhKqkU7xSzYTM/5P9hmEgT\n5bSe5mUNnGJo3dEKpk6AlMXGXIkoOy16BHf4qiabJ/yifDc5URTnZNTLrNwXeXjeRLHwvFJVFNOM\ntScSKJrQGsidLTzXSSBAtEh5OXHOixcDX/86HSuKZL1PJKi24vHjwP/4H+TKWFdHGrUkEcCHw8ZE\nmDOHrpVIQBdF2tqWldFikErR/crK6LqM0f/5bpPA5O4a+eHZnCJ4WGSynNbTkHHAqrKx3CoFpVj/\nlehOMiGm3q3w8am4H+YfVwr1Miv3XR6Ot8AnciAAXL5MtbbMmQBLUVFMMzbHQ6QU2byZ2uDz0Xnn\nztHfhw7RYH7tNZrYhw8TDdPRYVRoaWuja3g8VCCXMSNs3u8nrv7WLaTcbjiqq4mLFwQqBjA4SPc+\nfZp4+7feopSguk7Xb2kpbKx0Oo38qflbY87dP0wTMg9tBZOr33TljnlnM222Y8f4RGxF+rMopk7X\nJWWyRaBUVT7/uMmolwd9J/eIyAM8e03CJ3JVFXmUmOt2larFmGas7nSWNmPzJ9euXQSQopjrLsXv\nzQ2ofX1AezsBbmMjnTtnDoFyY6NBcfT0kIvjunXQhoeJxwZokYjFgK4u+v/aNTq3o4P+ZozAfmjI\nuD5HIqsV+OEPjXb9/u+Pd0NUVerHcYUs76FMRfvMQ1vd5Sp6makqtXfEO+er8f39uYnYJunPgpg6\nHc7lbucQmYh6uVfyJU12NVPycPQIn8i8mCwPzwNKH8CmGZtUFOO4iQZI/uRS1fHuUoIA/K//RaDt\n95NBtKaGgnvicfq+o4O+P3OGNHNZpoi4jz+mgrgLF0JftszgsXftokx4mQx9duUK8fPZLF2vshLo\n7aVnX7CAjuecOk+HWlFBfufhsOEqKQjUjqoq2tncyYI4EzJV4MlH29GkLIVinu4k6nvKvHO+Gl9T\nk5uI7U4Abjr+k9Ml3kuV++Vh8iVOdjVT8nD0RrEBVKgE1GT8ptebmyFpogGSP7k418yjJK1W4Ac/\nIJrCbqfsTv/yX9K5PGtcQwPRITU1tB0fGaHrRKN0/uLFwNy5SC9bZrSJe6xcvUrfz5tHx9bUAP/7\nfwPXr9PO4MgRykXNC13wRCFlZQanf/YsfZ5MAt/4BvHv+fHqwH0r0Fwy8BRA2wKK8b2p/lVgPCa3\nbTMM8HfSd9MByeksAnfSznu9e7tXC9VDLA8HkAOFB9BkA3giLbNQMv+JIu0KJfnh5zscZJC8coXA\n2ucDVq4Ezp+nz65fp+txbTgSoevxXCt9fbB//DEhUUuLEbRTXQ2cOGGU+XrqKWDuXMPrJJmkn3CY\nnpG3b80awx/+xAm6VyRCAL9793gH7qkuiCXIhAr+DAFPAcX43u38C3HJ0wWX6fh0FjJsf1nkXi5U\nD6k8PEBeSCbSYibStosl/Zgo0q4Q2HEqRZIIJJubCag/+YR48q4umljJJAHzk08SAJsj4oaGgL/4\nC7CuLvJFX7eOAnba2qidFgtFd3Z10WLhcpEWzvNG2+2UsdHnM9ra30/aeEUFcPIk0TDz59MugLfb\nLDM8USbdCc/QFr3QZWZ65//QULPcw+QueffcV5kNGppUHv4eKabFTLAdG4vC83oJxPOT+RcSDnac\naxYE8iLJZon6WLmSXA0HBwm46+oIjKNRMnS6XEZQDh+MjJE2zSWbpYk4PEzXf+wxMpr29BBgL1tG\n16+tpYUjHCaqpL2djv/a14xyN5kM8MEH5GGTyRAXn0gQiJs5dd6OHTsMTxfAyC2TLyUgW0k74Rna\not8NxZjLQ0fN3kXvnvsuD0Osw32UB3lYjpdpeDqYQSknCk+WCcSBiQsyc7Dbs4d8vP/gD8ifm3Pl\nV67QQLPZ6OfoUfIpT6UMbd+ci8K8K7DbodXUkDZ/9ix9NjhIC8Qrr5Amf+ECAXx5OYFyLGbkdxn1\nRcdTTxl50gMB8jnPZKh9djv5s8fjRuUXjk68Uo2u02eZDHH5Ph/x/ub+LwHZpqTgP8Aq70NHzU7g\n3TMleYDfyawUlofnLU3X08F8bP53wPhAm0JeLapKYHfuHPmQ6zr9NDSQ9pxKkWviM8+Qv+3Zs8Sf\nr1ljJNwyR5fyXcGKFUguXkz0x4ULwKJFpMWvWEEgfPIkaeuZDAF2QwN5vyQSBNDxOP3PPWL4fQIB\nIzHYsmUE+NxzxRyp9+67lFPa6aQ2XL1K7Y7HYamoMCpElIhsJe+EH3CV96GjZot490xJHvB3MiuF\n5eF4Q6UYJgtJ/nasWIh+IJBbqt3sBZLv35ZOAzduGLm9dZ2ALpEgbTmVoglksRhpalXVoC24QdJc\nwUCWka2ro3vyPC82G137yhUyWHKuPB6ndjmd5Ir5x39M1z5zhgKSGKPAow8+INBevJi09NdfJ6Dn\nRltzzVIO+iMjhjfNKKVTFovRTuCll6aEbAyUDhaQUXCY3ek7vYfyUFKz06UgHrRtyOzuoCR58Htm\nKoZJfvwkxs9xIfrcPc/hMEq2A4XzTFy+TL+zWTre4SCAH61xCIuFwPLpp0mTTqeJCvnkE7o2dxFU\nFALtUS1auH6dztu+nUDV7ycO/vRpAljGyMhZXk6Jr4aHacFxucivPJkk4P/xj6nNIyNEGTkc420A\n+R4OgkD3a2igRSCVoucF6JhAwKjgUgjZCkXlTFYmaCrv9D7KI0fNPkjbkNndQcny4PcKB9NSDJMl\nlq7JCdHXNKIVrFbSqrdsKVwlwW4HurvJy6SykgDS5SIA1HX6m3uZxGLAc8+RBq1p5IXi9dK5584R\n6A4M0HE+H5BIwHXsGBkwOZcty3SN48epDYkEGVTnzTOMlrydPK95Tw9w8ybRI5EIBSJVVRFVYs6X\nzkGT95XNRtd2Omlh2baN7nnmDLWxs5NSDzBm1HScqM8n0+qm8k5nUGaVuxLkQdqGPGi7gwdYHvzh\nbAZTbpgsNrgmevHFQvQVhYB161aiRLhnCWAM6lCIOPE33iAgTqeJw66ooN/pNFUW+vf/nup7Op1E\nTaxaRfc0F3hYuZK05fZ2Aspbt4DKSohDQ8SNx2JGzvJEgtoTCJAHDE+ylV94UlWNjIqMEaD7fLSo\nbNxI98uvPbp5M/3tcNCzLV9uGHQdDkqZu3MnYlevQm5vJ1/4YJCCm9ato/7LB21ee3HUoVsLKVCE\nckoPe6fvdIZkVrmbgjwo25AHaXfwgMuDP5SnoiFM9OLzQ/QBI/c3D//3+ehzzqHz8wDSSG/cIDC1\nWo0gm+pqOv7HP6aQe+7tUV1t1DJLp+lYl4sAcWiIrpXJEEVSXU2gnU4bbeFBRJIEvPgiASXnkrnL\noPm5OaWxfTvtGDZtotQBgQAtOA4H/c3zkAO0wLz/Pu00PB66By9uwRjg9yOrquTDHg5TW3itNl5R\nyeyWaaq9qG3bgX1vpxHXJUgHxFzgvA9a36xy9xDKg7Q7eMDl4eiZUjWEyV48v46iFM8I+N57hQOI\nbt0iAO7rI1qjqorykUejxGOn03S81Wr8zb1GOjtJU00mKUnW+vUEnnPnEkjeuEGG1qYmAlRNo2tz\nwOTeKPml43gqVa+X2hsI0GIhCMCf/Ak9k9UK/Mf/SMbcy5fJeLppE52zaRPdn19bEMZrPYxRNCjv\nh6tXc2u08T7Pq72oBBKIWzzFgfMea32zyt1DKg/K7uABl4cDyKciJbz4cWW5OCVRoPLLmNTVkRYe\nj9PfjBGF0N1N51y/Tpqw10s0x/r1VEwiFqNFwO2mc+bNIxdDp5O+mz8fWL0awsmTRNvMm0eJsM6f\np/bY7RToU8hd8tgx4uc3bCBXx5ERiuJct44AurmZFp/Tp6ndW7YQfdTSQtc4doyOz2ToGhyw+/tp\ngeL+9Q4HperlBlJzMBEnnvPqlMo1rsmB8x6S1rPK3ax8mWXaw7mpqekVAGEAa9va2v7z9JtUokwD\nBIqW5Sr0maYRZdLXZ4TXp1J0z9pa4o0ZI377n/9zoiD8ftLsh4eNwszZrPH7scdIc+7sHEtHq/Mk\nW0NDBOYrVhBVoarkvbJzp8GLh0J0Xx6m395O4FtZaVBENhsZVJNJWkg6O8mwyI2k3DbQ0kLgvnMn\nXfsXvyCaKJMh3vxf/Su6h3mB5BWVNC3X/94UMcoYGw+c+VV+7jFpPavczcqXVaY1c5qamtYA0Nva\n2j5oampa0NTUtLqtre3cDLWtuEzFclUI8M1GTLNs3EhA7PHQ/4kEUQmBAAG1ohDA1daS1rx7NwGm\nqpJb4JtvEvhaLMSlZzJEmVRW0vnRKN370iXyZOGLhsMBy/Aw+YvX1AB/+IdkgEwmqS2ff05tra8n\nsNQ0uh8PEuLAfuECXb+9Hfje92jn0NlJbeEukdywKMu0cAwOEghzcO/ro+va7UAwCMvt2/SshSQU\nKux/D6Obx4Az/51xY+ssaT0rszJtma4K9M8AvDf6dyeA5wHcfSAv1XLFNcZAgIDsG9/Irc/Jq7zY\n7QREx46N5QfHE08QJTE8bGjWiQTdz+MhDxCrlagNQSAN+oMP6Fo9PaRxr1xJWvCcOZQTJRiktqxd\nC3zxBbW/txeIRGDhnjDJJGnFX/kKtSkSAd55h+5x9aqR1bC8HPj+9ynR1oULlA7g0CHycuFGT6+X\n6JZSvUNkmRYpHsQkSchWVBh9WWgHxP3uzf73pbwzYJa0npVZmSGZLpB7AARN/1dM83qlSamWq1CI\naJGuLqJDAHKrA3KBpb+fKAmeVlbTSKs9d44AeWSEtM2qKuCf/omupeukYff1kRYvirRIhMMExokE\nAX5ZGd2LMTpe1ylQyOczjJwWC4R43EiIxbV6m43ao+vAp5/S9Ts6aNHIZAhsLRYC+OZmg7vn5e/M\nxHB+3U5OrfDi0HwxfOUVytmydy8gCLB/9BFRPfkpfLlw90au1RcD/Px3xg20s6T1rMzKtOXhnD1T\nsVzxyvN2OwEfD9HnLob9/aTdlpcTSGYyRi5wm42060yGuOvhYdLCUynS2EWRtGyHgzTsJUsIwOvr\nSfP2eMhoGYsZWvGCBXTehg3A3/89fR4IQJdlaoMskwFVkojjdjrpuh9+SIvKwADtGLjWPHcuaf1P\nPklaOAdyM43Eg394FOfu3RMvhokEGW+tVtgiEeDZZwnEy8posQkEgM8+M3YzL788eYGKYu9slk6Z\nlVmZtkwXyEMARp2v4QEwXOigvr6+ad5mAuHVfgqJpsG+ciVssRhgsSCdTCIRCCD73nsIZjJgV68i\n09CArCAg2dICR38/hJoa6JKExJYtcCgKbMeOIVtdjUx1NayXLkGfOxdieztS1dXIVFXBlkpBCAZh\nCYWQWLQINk1Dav162M6cgWa1wtbZCeg6hEQC2ty50ObNgzgwgOzNm7BnMhC8XrBIBLqmIZHNQl2x\nAtYrVyB+9hl0hwPQdaRWr4ZNkiBoGoRsFv9/e2cbHFV59vH/STbZ3ZAl2YSARFFCmRzUasEAtoMW\nK6LgS4fWYktnbGc60s7Yz20fP3Q6TqePPp223/yknbEdWxnxpR1p8WVAUGprLUiLIAchoDZsSCAn\nZJPdbLLJeT78z819drObbLKb7Fm4fjMM+3L2nHs3Z//n2v99Xdc92teHEXfhZeOTT+AEg8Bbb8EY\nGYETDiN1++2ounAB483NMNyLVvDTTxE4eRJVAwMYsW0Mb9kCdHTAGBzk5G9PD7d7801U9fQgeOwY\nxtraMDIygtjgIGoTCdTu2wcAGHU/T6ehAYZtI9XbC2d0FEZ/P4JdXXDq62F0dSF14gQcNd9Q6N+s\nAOLx+OyeUzNAxlQYfhwT4N9xFUqxQv4CgA4AewEsA/Bmro1aW1uLPEweCslc2b6dDZ9ef51R7Hvv\noW9sDE1qIecFCxh519Rwom7pUopaNAr88Ie0U1SnwpYWWh0XL6Lu7FlG6V/4AiPyI0dQFwjQqnEc\nluffdhtth08+4WPNzTzekiWMci9cuBRppxsbEWptRSga5XMHDjCiTyZRd+YM3+fQELByJUJ33AE8\n9BDfdyzG9/7yy/rXxFNP0TY5d45eu4r2x8Zov6i8dBUNe5uJhULMqonFgLY2XKivx+KbbmLkr1Y4\nsm3d26Whgb8yQiE+Z1k6RbO9fVYsk7Nnz87eOTVDZEyF4ccxAf4cVywWK3jbor5llmV9YJpmh2ma\nGwDYc5KxovA2XlJ2gXciUxEI8PGaGors+fNc8XxoiEL3r39RjL7+ddoE+/dToFQ7261b2atb5Zkv\nWUJrpLtb799xaIEEg7Ryfvc7iv6ZM5zwTCZphyQStG8SCdozjz/OIqAPPsD4qVNsWOVmsaChgdud\nPMn0wbExLcDz5/N/tbbnwYP8LEZHOe6ODv5/8iRfl0qxOrS6WvccV/YSoO0Q1ZFxaIiteNevZxWs\nuli2tNCKUWuJHj7MSdbdu7WNIr63IMw5RX/TLMt6phQDmTZKYI4f16vsbNuWWzy8ZeTHj8NZtEhn\ndlRV8TWjo5zQ7OvTZey2zQnKVIo9u8+fZ4Td0aG98muu0ZN8qg3s+Dij/P5+pvyFw3z+0CFG4QMD\nFMF584Dvfx/47DOkXnwRde3tFH/HYTR/+LBueDUwwIvIhg30pGMxvbbnmTMUf5U/nkxqT766mu+9\npYWfj23rCla16lB3Nyc9h4aY9aKaciWTCO7Zw9erClgVJagKVrdv+aXJUknWFoQ5pzJDJpUKGI/r\nYhTHyZ+GmFVG7oyOUuwsixFrJEKxdfuLYGiIt//yF51l8vHHFHTbZqbG9dfr9rSpFItnVKQ+OMh9\nNDYCbW2MXm1b9zYZGGCqoboQLV/ONL/9+xmxOw7HvWgRo/FQiEL661/zOLt2ccLx5EkeO5HQF5BI\nhNG0YTDd0nEm5tAnErwA7NvHlMWuLtoj69YxBTORYObM0qWoPXSIxUJDQ/zMlizh/lTU7S3XFwSh\nLFSekKvc8L/9jSJSU0OhnEpMAgGKUF0dal9/nfcNgx63imJVSpxtU8T/+U+K8fXX00+/eJFifPPN\njJgNgwJvGHxu9Wru7+232dvbNHWDK2XBbNzIPijJJH37L34RWLAAI9/+NsXSMHTnw1tuYRRfX087\n5eqr+dzbb/NzGBzk2Gpr+dqGBmbbNDfz18Do6MSFllXL2/PneZFob+d+Fyzg+zp0SHdfVJ9nTw/3\n4+2tYtt8L9lZMupvJPaKIMwZlfctU1G4Kmlfvpw9RApphRoIAOvXI93ZyUnN3/+eEW1TE20WlZIY\nCHD/jY0UxPZ2WiBvvcXIuLWVeduqpW1dHcW2vp77i0RoWVy4QCtieJhRtSoWUqsPXbzI/PD58xFq\naeEFac8ebVmMj/NYixfThlG9TU6d4n7Hxym6amHlVIoXiN/+lmOurZ1YBKVa3o6M8LWWRaG++mp+\nlmrpuEgEGB/HyJo1Or3Q+/mqYqrsvHLpFysIc07lfcMiEQqvKpRpaZleP+tolDZGfz/thI4OPVnY\n1ESfXEX3K1boidRAYKLYf+tb3GcyyXFs2kShNgw9abprly7c2bYNuOsuWiL/+Q+tmNFRFgSlUrRh\nNm3iBSSV4sXgzBlaMaGQtjRaWxmBz5vHSdSxMRYmdXVxfdDRUY5h3z4K6ve+x9crS+rYMb6+upr5\n593dFN3BQYp2IMCsmGSSk53qwqMWp56sslb6xQrCnFN5Qh4IMJtEdfDLjhQLYGTNGopTUxOj++pq\n2hEHDvB+Swsn9rK7/AG0dFIpPfl33318PBKhaCuPGaDN8u9/UzRHR/kLIJmkF718OSdQXd++9t13\nGaGHwxToEye4bSKhPW+1TmhtLS8MjkPxX7qU63UCjKhra3ncgQGKe10dLzq7d9NbT6VoA/3977wA\nxGK8IKjqTPV5hkI8Zq61S/MVE0m/WEGYcypPyAE9KTldXEEKdnXRSrj/fkaMTU20QVSetJrY8+ZZ\nv/oqRfCjj3gRsW3g2We1yN93n7Z81NJs4+O0W/r6KMSGoSPfTZsolI4DHDiAMSXMS5boKPvkSeag\nV1UxW6WmBnjuOQpxezv/rV7NYwwOUmB7ejj+fft0S4CLFynWiQTfH8CLVizGcv+VK4G77+axc3nb\ntp25KEUymb/0P19DMkEQZo3KFHIv05lYc3/2O/X1FDUl1uvXcz9NTRSq7EhS2QULF9KWiMWYbTI0\nxIlDNYG4dq1eQWfdOlozW7YwXzyR4MXivfcYMdfX83VjY0BDA6o//FB76IDOklmxglH3jTeyne2h\nQzyW2jYa1XbT8DD3OzbG/igffUQv/ORJXvjU6kOf/zwj8aYmiv38+RTx7H4q6iL17rvc14cfcgJW\nZfh41/3M9sOzPXTxyQVh1qjsb1chE2teoXd/9hsqIg+H8/bTztnwaWiI2SqqmlO1ol24kKIbjzPz\nIxDgQg/vvkvb5oYbGFkHAtzH6tXsVdLezsnG4WEYIyO0W06f5sVidJTHTiQ4AWrbjLbr6jjutjbd\nqCuVYkQdCPC5Xbv0Is3t7Yy4R0f5+cRiwJNP8kIRCLBJ1tat2sbJ8rYNlUd/221MnUwmdQFQPj98\nOj65ZLgIQtFU9jdnKsHIJfQPPIDUsWPMBJmin/YlsisWAXrQpsnJSNXvOxjkcbq6OJl61VWsHD1/\nnlWkK1dSuPfu1YVFg4NAVxeMgQGKe3U1J1/nzaMY33wz0yAHB3nBefhh2iU33UQx9uZxq5WOtmxh\nU69f/pLPnzjB8QFMq7QsTqjW1fGipKLrHN72pUU4env5/lThkHe9zmw/vFCfXDJcBKEkVPa3ZirB\nyCX0kQhq//EPPl9dTR+7ujqzn3Z2lJjrPkBx3LaN2S+1tcDzz1OgGxoolN3d3PfixYym+/oYSa9Z\no5dfc7scGmNj3PeaNbRtQiFGz/v3c4wDA9w+lQJ+8QudQx+Lcds//1m3k92yhRbLmjXcZniY29XX\nU4wbGjiWZJK9zD/77NJFLqfvrTzvbOspX0l+oaX6kuEiCCWhsoV8KsHIJfS2jdr336d4jI4yIh0b\ny+ynnZ2hke0dq17eCxfSbnjuOe6rv19H2Y89ppeJO3iQonrNNRTTQOBSLxNs2gR8/DHGh4Z0tks4\nzP3V1zMyj0aZOx4IMJp/+GH+ctixg565Ws6tvZ3R+vr12jtX1aOqVD8SoYXU18eLzYIFmSKa7Xt3\ndOjJ5S1bcot2vmraqURZMlwEoST4S8hn4pdOJhhTCX1VFXDvvTpHOxCgheDN0FDZHipqVNkYapk0\nVY0ZDDLyPXWK9996C/jOdziJWFurm1XdeisFX9kZALBiBdIDA4yeV6zgxeO11/iaxYtp4YyMUNx7\neoBXXuF43nmHY21t5bFHRngs73vPWt0+w0vfvXuiiGZFycbgYGGf9UyQJluCUOtoiHEAABFNSURB\nVBL8882ZLb80W3yiUYysXas7HLa06OOk07QyPvyQonnHHRRSFTWq7oiJBO2OjRsp2n/6E1+7ZAmj\n43nzmJ2yeTNfr5pV7d0LPPEEJ0FbWuiZ/+xnwJe/zGrT665j5Ky6FapComCQ2SJHj3LM1dW6CKmr\ni+9l5UraNvX1WpRVWwLP6vYZeeK50gSzomSnvj7336pU4itNtgShaPwj5HPllwYCSN1zj54g9AqR\nGoOa3FOThypqTKcZxR4/Tmvi4EFGy1u2MKsjFGJGyH//y5S/Xbv4nBLPvj69mLNhcBI0FgPWr8ew\naermWoODulpVCeZ999HCCQa1UM+bxwhevZ9XXqGo796dWdCUPVGrKjSBzDRB9Rrva3t6Mj+/6WYK\nSZQtCLOOf75lc+mXeqPA7PREw9DdCauq9AVF+eeGQWskFGKE7C5SfEmw165llkhNDYU1nWb2yubN\nPO65c4z21RJxhw4B6TRXMTp2jJkrqo3t/ffr0vhoFHjkkUxB9oqlbXO88+bx4rBjB8egxDYapf2y\nc6fuiOhdyd62+ZwqaMr3i2gmmUIi5oIwq/jnG1YOv1R1UlR55GohCUCLnRJNlfmhGkSp1EVVzagy\nWdauBf74R26vLkqqX3dVFS8QqhHWo4+y58r8+ajq6qLfXlXF/fT2UowPH+bFY906fbFQeG9HItr6\nGR6miN95Z2aq4M6dLMtvbGQkD+iLp2HwPbuLb+DwYebaq/flPc50M4XEOhGEWcU/Qg7MvV+aL498\n27bMC8rwML1s9dhPf0pRP3GCrxsb4zbPP8+LQn8/t3Mc3r72Wh6vpoZC39vLBSfa2iiYto3xxkam\nBZ45Q4/cMDhR6TgcX28v7+drEOZ2dkRfH/vGvPNOZvvZeJz7Uh0dDSNzJXs1+dnTA7z0EvCHPwDh\nMIL33AN885vaHppJppAgCLOKv4S8HKj8cW8eeTZqNZ7mZpbZ9/ZShG+/nT1NamuB3/xGt59VF6OF\nCyl0GzdqC+aGG3isr32Nizo4DmAYSN11FyN1tbrPX//K27ZN8VeLWEQiExt6KaJR3Tfm1ltZQaqE\nP19HR/U6gPt96ilG8cPDQHU1gu+8w/fX0qJtkmIyhQRBKDlX3rcsnYbR30+RjUZpWShrJVce+QMP\n0AaJRCiQau3KeJy54BcvcrsDB7SNsmwZUwGvuoqPvfKKzj7ZvFkvmqxsB9tG1blzfFz52okEqzpt\nG/jGNziBGo3S9nj2WR3t5vKgx8ZYUdrfr/O/CxHYZJKfQ1MT0yjnzYMTDutUzEJtEslEEYQ55coS\ncm/3w2PHaEVk91ex7dwe7+OPU3xbWhhJq8WKm5r0QgxLl3KyccsWXgzcaBuOw8d7enTmSzzOydIV\nK4Djx1Fr20wlDAZ5EYhEdKOutWt5wbBtWjFqMYwVKyb2AlcLWxw9yqi6tjZ/64FsIhFaL1/9Ko99\n770Yee01jrupSWwSQfApV5aQq+6H4TD7hqve496oNtvjDYd1ul5b20Shv/tuvi4YpNjW19Ne2bRJ\n72/XLlowhqF7iUej7Gh4881AdTWc8fHM5ldbt+qy+lBIF/eMjbHHivK5c/UC7+7mPlQhEjB5Nok3\ncyc7VbG6etb/LIIgFMeVJeThMDA+jqqensze496o1mtBqAnAeFz7ytlCr7r9qUg4Hp+YxqcmIRcu\n1BcCQHvWp0/rjoxqUjGdZmGSV3iXLNEZJ9k+t3fsqi+K6r3i2jfo7aXP39vL+2rln2yBV5+FbbMr\n43XX5c9AkZxxQSg7RX/zTNPc7t78nGVZ/1Ps/mYNZWk4DpxQiEuc5eo9DmiP17Z1r5L+fj63bVtm\nRaRaPu3IEb18mmmyr8rQkN6uoYERdmennvDcvPlStJ06cYK9UpQY5kvjy+VzZ4tpSwvb03oja1Wx\nevo0f1modMvJ0gUjEf56yZeBIjnjguALivrWmaa5AcCblmWdMU3zBdM077Qsa2+JxlY60mnaEkqk\nBgaYY62yOfKJjyoQ6u9nJOw4WhwPHNA+99KljMhvvZU9yo8eZWXnl76kS/qPHOHkZyrFMahfAq4/\n7zQ2Zo7D/fWA8+d1bxT1Xvr79eIOucQUmFgslErRax8a4v+qf/pk6YKBAFIbN+auglXH8BYUTZYe\nKQjCrFHsN24ZgDYAzwDodO/7S8iV0CnRvfFGRpmFrPWpvGogs0BICVhNjRbD6mraJzU1wFe+wgyX\nW25hL/GaGgrookXMPlGTh/v36/U/Ozoyx7x7N597/30K7+7dTGP8+c/1MR9/fOKCECo3PpcdFI+z\n2+H4eKY4T5bNMlkGitqvbfOzVY9JZC4Ic0pR3zbLsp723L0FwI7ihjMLKNFVk4u3345UdXXhQhMK\nTSwQUtGyyi4ZH2ca47p1nEQdGuI2LS25t1MLR7/xRu4ug/G4vvC4aYBYuZJRvjefPRZjBBwM8nZT\nE1/f38+K0WSS95UdFI8DDz00MQd9pumC+TosSjWnIMwpJQmbTNNcBeCgZVmHS7G/kuK1DiIRCl92\nI6ipyO7N4nrtCARY5alW6QkEmM64cye3e+IJTkzm2i6dzt9lUFk6ySS99USC95cvz8xnX7x44ljD\nYaYoHj3K1yr/XPWLAXhxKhW5OixKmqIgzClTCrk7menmsMFwb3dmeeEbLMt6LN8+zp49W9Qgi6aj\nA8bgIMWypwfxeDxzTOm0fn6KSN3o70ewqwtOXR2qzp/H8OnTcJqbL0W/Rn8/ghcvAoaB2u5ujFx7\nLeA4SH36KX1wFSWn0zCWL+fka2Mj4slk5pjWrUPItinEgQCG162jUP7gB6g6dw7jixYBfX08Xnc3\nnPp6GN3dGDlyBLVLl6ImmYSRTGIkHsew8uK9FPieJ3xWBX7Gs0nBY5pDZEyF4ccxAf4dV6FMKeRZ\n9skETNPcblnWr9zbGyzL2pO9TWtr68xHOAucPXtWj2m6mRcLF7KY6MABRsmWpSsn1fOWxSj4qqu0\nF+3NSMmxClGssxOLFy7k8/E49/Poo7m962XLMsdjWdzX1VcDq1Yxj7y5WXvk2RH4NN5zxmflE2RM\nhSFjKhw/jisWixW8bSmyVp40TfMnAKIAthazv7Kg/Gg1cTmVv+ttTuXNQ1eToN6imlx+tDpmVvvY\n4MWLzGxRi0mo/UyVtx0IMI0xFqPVooqHVB686qme7/jiaQtCxVPsZOceAM0lGkt5CIfpJ8fjnFTc\nuPGSnZGXaJQTmSoPPRzOX1TjjYaVCIfDE9rHOuEwqz9HRpjdkl1+r16fK9Uwe01RlVaZvfaouqhI\nh0JBuKyQHLFkktksVVVc1GHHDk7eee2SbLJT9gqJcLNFeONGVli6vVuqPv6YvwoaGnKX3wO5jwPk\nPnaOqD+j2lQ6FArCZYN8g1WE2t0NfPIJxfTTTzMbTeUqQ/dmshQS4SphVSv4vPii7nS4eTOGP/c5\n9jdXGSoqrU/ZJZMdJ9dj3m29i0Z4xb5YO0XK8wXBF8i3T0XXp08DH33EqHVsTD+faxWhXEUzU0W4\nagWffftY6ONdwSeZhLNgAUv3lV+fXfgTCuU/Tq7HcvWMKaWVIuX5guAb5JsHUIDa2thf3NubHMi/\nilCufRQ6SdrczH2qCs90Wvvy0SgvKtmFP21t+Y+T79jex0ttpciEqSD4hitDyAuxAAIB3UQqe7tC\nVhEqBDVJmkhkVoK+8QaCw8PAd7/L42YvZJGr8Ge6lHqxB5kwFQTfcPkLeb5Mj1wosVMr10ciuVcR\nmim5JklTKWD+fBjnz+umU6GQXsjC65H7CVnSTRB8w+X/7cuX6ZGPXMKfL1KfCbkmSW0bgRMn9Jqc\nDzxA8VZ2il+RJd0EwRdc/kKeywLILlf3ks/7nQ3B8jSdStu2NJ0SBGFGXP5CPl0LYDrebynS79ym\nU059vfjNgiDMiMtfyIHpWQCFCn8p0++mWsBBEARhEqrKPQBfooTf2+RKdSJUeC2YRGJq7326xxQE\nQSgQUY2pyBd5S/qdIAg+QYR8KvJNfkr6nSAIPkHUZyqmWJxYsksEQSg3IuRTIZG3IAg+R1SpECTy\nFgTBx0jWiiAIQoUjQi4IglDhiJALgiBUOCLkgiAIFY4IuSAIQoVTMiE3TfNHpdqXIAiCUDglEXLT\nNDcAuKsU+xIEQRCmh1grgiAIFU7RQm6a5irLsvYAKHJBS0EQBGEmlCIil5JHQRCEMjJlib5pmtsB\nOO5dw73daVnWXjca3+s+5+TcgSAIgjCrGI4zc/01TfNBUMCbAfwAwCOWZR32bnPw4EFn8eLFRQ2y\n1MTjcUR81j/cj2MC/DkuGVNhyJgKx4/jisVi6OjoKMiyLqpplmVZLwGXovaGfNu1trYWc5iSc/bs\nWRlTgfhxXDKmwpAxFY4fxxWLxQretiTdDy3LehrA06XYlyAIgjA9JP1QEAShwhEhFwRBqHBEyL2k\n04Bt839BEIQKQVYIUqTTwKuvcqHlujou7ybLugmCUAFIRK6Ixyni8+fz/3i83CMSBEEoCBFyRSTC\nSHxggP/7LKdUEAQhH+IdKAIB2inxOEVcbBVBECoEUSsvgQAQldYxgiBUFmKtCIIgVDgi5IIgCBWO\nCLkgCEKFI0IuCIJQ4YiQC4IgVDgi5IIgCBWOCLkgCEKFI0IuCIJQ4YiQC4IgVDgi5IIgCBWOCLkg\nCEKFI0IuCIJQ4YiQC4IgVDgi5IIgCBWOCLkgCEKFU3Q/ctM0VwFYBgCWZb1U9IgEQRCEaVGKiPwx\nV8DbTNNcWYL9CYIgCNOgqIjcNM0HAfwTACzL+lVJRiQIgiBMi2Ij8jUAmk3TXGWa5o9KMSBBEARh\nepTCWrlgWdYHwKUIXRAEQZhDprRWTNPcDsBx7xru7U7LsvYCuACg032uH8BqABMmPA8ePFiSwZaS\nWCxW7iFMwI9jAvw5LhlTYciYCsev4yqEKYXcsqynJ3n6RQAqCm8E8H72Bh0dHcbMhiYIgiAUQlHW\nimVZpwH0u5ZKk2VZL5dmWIKQH5mPqSzcFGXv/QdN09xQ7r9jjnFtd/896ZcxeR6f9LMq2iO3LOsZ\ny7JesizrsUJfU+4/oN/wy4ntxQ8ndS5M09wA4K5yj0PhTvQ/6Lf5Ic859UiZx7EBwE7P/VUAHMuy\n9oBBYFlSlnOMawOAN10HYplpmneWe0xZj096zs95ZacPv4hlFSy/nNhZYyr7SV1B+K6Owj2nOt1z\n6nQ5x+WO4ZTnoW+C82kA59fKogU5xrXMM5ZO9365x1QwV3SJvk8EyxcndhZlP6lzYZrmKvdk98W8\nS3YdhWVZh8s8JC//5/6/zAfj8v69GgH0ee43z/FYcmJZ1tOWZT3j3r0FwL/KOR5Foef8nAq5376I\n8Idg+e7E9utJDSBa7gFk4cs6CjcduNM0zT4ws0woEPfXzEEfXPwUBZ3zcx2R++qL6GPB8gV+Oqnd\nIGCve9eZdOO5xXd1FKZpNgCwAfwvgKdN01xa3hFl/L1sAE3u7Ub470KzYTrzfbPJdM75optmZR04\nb865j7+I5RYsP5/YvjmpQeurDfzF0mya5kofXGAKqqMoA98H8IRlWQOmaXYC+AaAcrbQ8P4CfwFA\nB4C94C/gN8syIpLhDJimuV21GjFNc4PrHpRzTAWf8yUV8ilyzsvyRZyioElRTsHy04l9CZ+c1JdQ\nnTXdv2dDOcfiYco6ijLhwBUEy7JeLmfmivsrpcM0za9blvWyZVkfmKbZ4c5P2eW6GGePyx3Pk6Zp\n/gR0DraWe0zTOecNx5nb4Ngd1I8BbPVBRKUE62n3dlkEy/2inQbQ5rF6yoZ7Ur8A/lqIgn+rvZO/\n6srE/dvZAFb76NeLSvE9BdZ3lP2cEmaXORdyPyGCJQjC5cAVLeSCIAiXA1d0HrkgCMLlgAi5IAhC\nhSNCLgiCUOGIkAuCIFQ4IuSCIAgVjgi5IAhChSNCLgiCUOH8PxNzWtg+ONc1AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x119798150>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%matplotlib inline\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "MAXN=700\n",
    "\n",
    "fig = plt.figure()\n",
    "fig.set_size_inches(6,5)\n",
    "\n",
    "plt.plot(1.25*np.random.randn(MAXN,1),1.25*np.random.randn(MAXN,1),'r.',alpha = 0.3)\n",
    "fig.hold('on')\n",
    "plt.plot(8+1.5*np.random.randn(MAXN,2),5+1.5*np.random.randn(MAXN,2),'r.', alpha = 0.3)\n",
    "plt.plot(5+1.5*np.random.randn(MAXN,1),5+1.5*np.random.randn(MAXN,1),'b.',alpha = 0.3)\n",
    "plt.savefig(\"toy_problem.png\",dpi=300, bbox_inches='tight')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAtcAAAFFCAYAAAAnyMGIAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAASAnhGgAAAEgJ4RoAAABIyf8B+NLU71N+CAkAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10a514410>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import numpy as np\n",
    "from sklearn import metrics\n",
    "from sklearn import tree\n",
    "\n",
    "C=5\n",
    "MAXN=1000\n",
    "\n",
    "yhat_test=np.zeros((10,299,2))\n",
    "yhat_train=np.zeros((10,299,2))\n",
    "#Repeat ten times to get smooth curves\n",
    "for i in xrange(10):\n",
    "    X = np.concatenate([1.25*np.random.randn(MAXN,2),5+1.5*np.random.randn(MAXN,2)]) \n",
    "    X = np.concatenate([X,[8,5]+1.5*np.random.randn(MAXN,2)])\n",
    "    y = np.concatenate([np.ones((MAXN,1)),-np.ones((MAXN,1))])\n",
    "    y = np.concatenate([y,np.ones((MAXN,1))])\n",
    "    perm = np.random.permutation(y.size)\n",
    "    X = X[perm,:]\n",
    "    y = y[perm]\n",
    "\n",
    "    X_test = np.concatenate([1.25*np.random.randn(MAXN,2),5+1.5*np.random.randn(MAXN,2)]) \n",
    "    X_test = np.concatenate([X_test,[8,5]+1.5*np.random.randn(MAXN,2)])\n",
    "    y_test = np.concatenate([np.ones((MAXN,1)),-np.ones((MAXN,1))])\n",
    "    y_test = np.concatenate([y_test,np.ones((MAXN,1))])\n",
    "    j=0\n",
    "    for N in xrange(10,3000,10):\n",
    "        Xr=X[:N,:]\n",
    "        yr=y[:N]\n",
    "        #Evaluate the model\n",
    "        clf = tree.DecisionTreeClassifier(min_samples_leaf=1, max_depth=C)\n",
    "        clf.fit(Xr,yr.ravel())\n",
    "        yhat_test[i,j,0] = 1. - metrics.accuracy_score(clf.predict(X_test), y_test.ravel())\n",
    "        yhat_train[i,j,0] = 1. - metrics.accuracy_score(clf.predict(Xr), yr.ravel())\n",
    "        j=j+1\n",
    "\n",
    "p1,=plt.plot(np.mean(yhat_test[:,:,0].T,axis=1),'pink')\n",
    "p2,=plt.plot(np.mean(yhat_train[:,:,0].T,axis=1),'c')\n",
    "fig = plt.gcf()\n",
    "fig.set_size_inches(12,5)\n",
    "plt.xlabel('Number of samples x10')\n",
    "plt.ylabel('Error rate')\n",
    "plt.legend([p1,p2],[\"Test C = 5\",\"Train C = 5\"])\n",
    "plt.savefig(\"learning_curve_1.png\",dpi=300, bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let us repeat the process with a simpler model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAtcAAAFFCAYAAAAnyMGIAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nrgEAAIAE+f/gVCDToKuX5wAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x115a08310>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "C=1\n",
    "MAXN=1000\n",
    "\n",
    "#Repeat ten times to get smooth curves\n",
    "for i in xrange(10):\n",
    "    X = np.concatenate([1.25*np.random.randn(MAXN,2),5+1.5*np.random.randn(MAXN,2)]) \n",
    "    X = np.concatenate([X,[8,5]+1.5*np.random.randn(MAXN,2)])\n",
    "    y = np.concatenate([np.ones((MAXN,1)),-np.ones((MAXN,1))])\n",
    "    y = np.concatenate([y,np.ones((MAXN,1))])\n",
    "    perm = np.random.permutation(y.size)\n",
    "    X = X[perm,:]\n",
    "    y = y[perm]\n",
    "\n",
    "    X_test = np.concatenate([1.25*np.random.randn(MAXN,2),5+1.5*np.random.randn(MAXN,2)]) \n",
    "    X_test = np.concatenate([X_test,[8,5]+1.5*np.random.randn(MAXN,2)])\n",
    "    y_test = np.concatenate([np.ones((MAXN,1)),-np.ones((MAXN,1))])\n",
    "    y_test = np.concatenate([y_test,np.ones((MAXN,1))])\n",
    "    j=0\n",
    "    for N in xrange(10,3000,10):\n",
    "        Xr=X[:N,:]\n",
    "        yr=y[:N]\n",
    "        #Evaluate the model\n",
    "        clf = tree.DecisionTreeClassifier(min_samples_leaf=1, max_depth=C)\n",
    "        clf.fit(Xr,yr.ravel())\n",
    "        yhat_test[i,j,1] = 1. - metrics.accuracy_score(clf.predict(X_test), y_test.ravel())\n",
    "        yhat_train[i,j,1] = 1. - metrics.accuracy_score(clf.predict(Xr), yr.ravel())\n",
    "        j=j+1\n",
    "\n",
    "p3,=plt.plot(np.mean(yhat_test[:,:,1].T,axis=1),'r')\n",
    "p4,=plt.plot(np.mean(yhat_train[:,:,1].T,axis=1),'b')\n",
    "fig = plt.gcf()\n",
    "fig.set_size_inches(12,5)\n",
    "plt.xlabel('Number of samples x10')\n",
    "plt.ylabel('Error rate')\n",
    "plt.legend([p3,p4],[\"Test C = 1\",\"Train C = 1\"])\n",
    "plt.savefig(\"learning_curve_2.png\",dpi=300, bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "and join both to see the differences."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAtcAAAFFCAYAAAAnyMGIAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x103ca9350>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "p1,=plt.plot(np.mean(yhat_test[:,:,0].T,axis=1),color='pink')\n",
    "p2,=plt.plot(np.mean(yhat_train[:,:,0].T,axis=1),'c')\n",
    "p3,=plt.plot(np.mean(yhat_test[:,:,1].T,axis=1),'r')\n",
    "p4,=plt.plot(np.mean(yhat_train[:,:,1].T,axis=1),'b')\n",
    "fig = plt.gcf()\n",
    "fig.set_size_inches(12,5)\n",
    "plt.xlabel('Number of samples x10')\n",
    "plt.ylabel('Error rate')\n",
    "plt.legend([p1,p2,p3,p4],[\"Test C = 5\",\"Train C = 5\",\"Test C = 1\",\"Train C = 1\"])\n",
    "fig = plt.gcf()\n",
    "fig.set_size_inches(12,5)\n",
    "plt.savefig(\"learning_curve_3.png\",dpi=300, bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let us check now what happens when we fix the amount of data and change the complexity of the technique."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAtQAAAFFCAYAAADM/3qLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x112981d90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%reset -f\n",
    "%matplotlib inline\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from IPython.html.widgets import interact\n",
    "from sklearn import metrics\n",
    "from sklearn import tree\n",
    "\n",
    "MAXC=20\n",
    "N=1000\n",
    "NTEST=4000\n",
    "ITERS=3\n",
    "\n",
    "yhat_test=np.zeros((ITERS,MAXC,2))\n",
    "yhat_train=np.zeros((ITERS,MAXC,2))\n",
    "#Repeat ten times to get smooth curves\n",
    "for i in xrange(ITERS):\n",
    "    X = np.concatenate([1.25*np.random.randn(N,2),5+1.5*np.random.randn(N,2)]) \n",
    "    X = np.concatenate([X,[8,5]+1.5*np.random.randn(N,2)])\n",
    "    y = np.concatenate([np.ones((N,1)),-np.ones((N,1))])\n",
    "    y = np.concatenate([y,np.ones((N,1))])\n",
    "    perm = np.random.permutation(y.size)\n",
    "    X = X[perm,:]\n",
    "    y = y[perm]\n",
    "\n",
    "    X_test = np.concatenate([1.25*np.random.randn(NTEST,2),5+1.5*np.random.randn(NTEST,2)]) \n",
    "    X_test = np.concatenate([X_test,[8,5]+1.5*np.random.randn(NTEST,2)])\n",
    "    y_test = np.concatenate([np.ones((NTEST,1)),-np.ones((NTEST,1))])\n",
    "    y_test = np.concatenate([y_test,np.ones((NTEST,1))])\n",
    "\n",
    "    j=0\n",
    "    for C in xrange(1,MAXC+1):\n",
    "        #Evaluate the model\n",
    "        clf = tree.DecisionTreeClassifier(min_samples_leaf=1, max_depth=C)\n",
    "        clf.fit(X,y.ravel())\n",
    "        yhat_test[i,j,0] = 1. - metrics.accuracy_score(clf.predict(X_test), y_test.ravel())\n",
    "        yhat_train[i,j,0] = 1. - metrics.accuracy_score(clf.predict(X), y.ravel())\n",
    "        j=j+1\n",
    "\n",
    "p1, = plt.plot(np.mean(yhat_test[:,:,0].T,axis=1),'r')\n",
    "p2, = plt.plot(np.mean(yhat_train[:,:,0].T,axis=1),'b')\n",
    "fig = plt.gcf()\n",
    "fig.set_size_inches(12,5)\n",
    "plt.xlabel('Complexity')\n",
    "plt.ylabel('Error rate')\n",
    "plt.legend([p1, p2], [\"Testing error\", \"Training error\"])\n",
    "plt.savefig(\"learning_curve_4.png\",dpi=300, bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We may use all these concepts to understand and select the complexity of a model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean accuracy: [ 0.82536232  0.8263285   0.83188406  0.83429952  0.83550725  0.83309179\n",
      "  0.8294686   0.82512077  0.82198068  0.82028986  0.81376812  0.80700483\n",
      "  0.80555556  0.79782609  0.79589372  0.78888889  0.78164251  0.7826087 ]\n",
      "Selected model index: 4\n",
      "Complexity: 6\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAs8AAAFFCAYAAAATsQCUAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x102802190>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%reset -f\n",
    "%matplotlib inline\n",
    "import pickle\n",
    "ofname = open('files/ch05/dataset_small.pkl','rb') \n",
    "(X,y) = pickle.load(ofname)\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn import metrics\n",
    "from sklearn import tree\n",
    "from sklearn import cross_validation\n",
    "\n",
    "#\n",
    "\n",
    "#Create a 10-fold cross validation set\n",
    "kf=cross_validation.KFold(n=y.shape[0], n_folds=10, shuffle=True, random_state=0)\n",
    "      \n",
    "#Search the parameter among the following\n",
    "C=np.arange(2,20,)\n",
    "\n",
    "acc = np.zeros((10,18))\n",
    "i=0\n",
    "for train_index, val_index in kf:\n",
    "    X_train, X_val = X[train_index], X[val_index]\n",
    "    y_train, y_val = y[train_index], y[val_index]\n",
    "    j=0\n",
    "    for c in C:\n",
    "        dt = tree.DecisionTreeClassifier(min_samples_leaf=1, max_depth=c)\n",
    "        dt.fit(X_train,y_train)\n",
    "        yhat = dt.predict(X_val)\n",
    "        acc[i][j] = metrics.accuracy_score(yhat, y_val)\n",
    "        j=j+1\n",
    "    i=i+1\n",
    "    \n",
    "plt.boxplot(acc);\n",
    "for i in xrange(18):\n",
    "    xderiv = (i+1)*np.ones(acc[:,i].shape)+(np.random.rand(10,)-0.5)*0.1\n",
    "    plt.plot(xderiv,acc[:,i],'ro',alpha=0.3)\n",
    "\n",
    "print 'Mean accuracy: ' + str(np.mean(acc,axis = 0))\n",
    "print 'Selected model index: ' + str(np.argmax(np.mean(acc,axis = 0)))\n",
    "print 'Complexity: ' + str(C[np.argmax(np.mean(acc,axis = 0))])\n",
    "plt.ylim((0.7,1.))\n",
    "fig = plt.gcf()\n",
    "fig.set_size_inches(12,5)\n",
    "plt.xlabel('Complexity')\n",
    "plt.ylabel('Accuracy')\n",
    "plt.savefig(\"model_selection.png\",dpi=300, bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "%reset -f\n",
    "%matplotlib inline\n",
    "import pickle\n",
    "ofname = open('files/ch05/dataset_small.pkl','rb') \n",
    "(X,y) = pickle.load(ofname)\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn import metrics\n",
    "from sklearn import tree\n",
    "from sklearn import cross_validation\n",
    "\n",
    "#Train_test split\n",
    "X_train, X_test, y_train, y_test = cross_validation.train_test_split(X, y, test_size=0.20, random_state=42)\n",
    "\n",
    "#Create a 10-fold cross validation set\n",
    "kf=cross_validation.KFold(n=y_train.shape[0], n_folds=10, shuffle=True, random_state=0)     \n",
    "#Search the parameter among the following\n",
    "C=np.arange(2,20,)\n",
    "acc = np.zeros((10,18))\n",
    "i=0\n",
    "for train_index, val_index in kf:\n",
    "    X_t, X_val = X_train[train_index], X_train[val_index]\n",
    "    y_t, y_val = y_train[train_index], y_train[val_index]\n",
    "    j=0\n",
    "    for c in C:\n",
    "        dt = tree.DecisionTreeClassifier(min_samples_leaf=1, max_depth=c)\n",
    "        dt.fit(X_t,y_t)\n",
    "        yhat = dt.predict(X_val)\n",
    "        acc[i][j] = metrics.accuracy_score(yhat, y_val)\n",
    "        j=j+1\n",
    "    i=i+1\n",
    "\n",
    "print 'Mean accuracy: ' + str(np.mean(acc,axis = 0))\n",
    "print 'Selected model index: ' + str(np.argmax(np.mean(acc,axis = 0)))\n",
    "print 'Complexity: ' + str(C[np.argmax(np.mean(acc,axis = 0))])\n",
    "\n",
    "\n",
    "#Train the model with the complete training set with the selected complexity\n",
    "dt = tree.DecisionTreeClassifier(min_samples_leaf=1, max_depth=C[np.argmax(np.mean(acc,axis = 0))])\n",
    "dt.fit(X_train,y_train)\n",
    "#Test the model with the test set \n",
    "yhat = dt.predict(X_test)\n",
    "print 'Test accuracy: ' + str(metrics.accuracy_score(yhat, y_test))\n",
    "\n",
    "#Train the model for handling to the client\n",
    "dt = tree.DecisionTreeClassifier(min_samples_leaf=1, max_depth=C[np.argmax(np.mean(acc,axis = 0))])\n",
    "dt.fit(X,y)\n",
    "\n",
    "plt.boxplot(acc);\n",
    "for i in xrange(18):\n",
    "    xderiv = (i+1)*np.ones(acc[:,i].shape)+(np.random.rand(10,)-0.5)*0.1\n",
    "    plt.plot(xderiv,acc[:,i],'ro',alpha=0.3)\n",
    "\n",
    "\n",
    "plt.ylim((0.7,1.))\n",
    "fig = plt.gcf()\n",
    "fig.set_size_inches(12,5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "WARNING: The following cell takes a long time to execute."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "%reset -f\n",
    "\n",
    "import pickle\n",
    "ofname = open('files/ch05/dataset_small.pkl','rb') \n",
    "(X,y) = pickle.load(ofname)\n",
    "\n",
    "import numpy as np\n",
    "from sklearn import grid_search\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn import svm\n",
    "from sklearn import linear_model\n",
    "from sklearn import cross_validation\n",
    "from sklearn import metrics\n",
    "\n",
    "parameters = {'C':[1e4,1e5,1e6],'gamma':[1e-5,1e-4,1e-3]}\n",
    "\n",
    "N_folds = 5\n",
    "\n",
    "kf=cross_validation.KFold(n=y.shape[0], n_folds=N_folds,  shuffle=True, random_state=0)\n",
    "\n",
    "acc = np.zeros((N_folds,))\n",
    "i=0\n",
    "#We will build the predicted y from the partial predictions on the test of each of the folds\n",
    "yhat = y.copy()\n",
    "for train_index, test_index in kf:\n",
    "    X_train, X_test = X[train_index,:], X[test_index,:]\n",
    "    y_train, y_test = y[train_index], y[test_index]\n",
    "    scaler = StandardScaler()\n",
    "    X_train = scaler.fit_transform(X_train)\n",
    "    clf = svm.SVC(kernel='rbf')\n",
    "    clf = grid_search.GridSearchCV(clf, parameters, cv = 3) #This line does a cross-validation on the \n",
    "    clf.fit(X_train,y_train.ravel())\n",
    "    X_test = scaler.transform(X_test)\n",
    "    yhat[test_index] = clf.predict(X_test)\n",
    "    \n",
    "print metrics.accuracy_score(yhat, y)\n",
    "print metrics.confusion_matrix(yhat, y)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "WARNING: The following cell takes a long time to execute."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from sklearn import grid_search\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn import svm\n",
    "from sklearn import cross_validation\n",
    "from sklearn import metrics\n",
    "\n",
    "dvals = [{1:0.25},{1:0.5},{1:1},{1:2},{1:4},{1:8},{1:16}]\n",
    "opoint = []\n",
    "\n",
    "for cw in dvals:\n",
    "\n",
    "    parameters = {'C':[1e4,1e5,1e6],'gamma':[1e-5,1e-4,1e-3],'class_weight':[cw]}\n",
    "  \n",
    "    print parameters\n",
    "\n",
    "    N_folds = 5\n",
    "\n",
    "    kf=cross_validation.KFold(n=y.shape[0], n_folds=N_folds,  shuffle=True, random_state=0)\n",
    "\n",
    "    acc = np.zeros((N_folds,))\n",
    "    mat = np.zeros((2,2,N_folds))\n",
    "    i=0\n",
    "    yhat = y.copy()\n",
    "    for train_index, test_index in kf:\n",
    "        X_train, X_test = X[train_index,:], X[test_index,:]\n",
    "        y_train, y_test = y[train_index], y[test_index]\n",
    "        scaler = StandardScaler()\n",
    "        X_train = scaler.fit_transform(X_train)\n",
    "        clf = svm.SVC(kernel='rbf')\n",
    "        clf = grid_search.GridSearchCV(clf, parameters, cv = 3)\n",
    "        clf.fit(X_train,y_train.ravel())\n",
    "        X_test = scaler.transform(X_test)\n",
    "        yhat[test_index] = clf.predict(X_test)\n",
    "        acc[i] = metrics.accuracy_score(yhat[test_index], y_test)\n",
    "        mat[:,:,i] = metrics.confusion_matrix(yhat[test_index], y_test)\n",
    "        print str(clf.best_params_)\n",
    "        i=i+1\n",
    "    print 'Mean accuracy: '+ str(np.mean(acc))\n",
    "    opoint.append((np.mean(acc),np.sum(mat,axis=2)))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "%reset -f\n",
    "\n",
    "import pickle\n",
    "ofname = open('files/ch05/dataset_small.pkl','rb') \n",
    "(X,y) = pickle.load(ofname)\n",
    "\n",
    "\n",
    "from sklearn import cross_validation\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn import ensemble\n",
    "from sklearn import metrics\n",
    "import numpy as np\n",
    "\n",
    "dvals = [{1:0.25},{1:0.5},{1:1},{1:2},{1:4},{1:8},{1:16}]\n",
    "\n",
    "kf=cross_validation.KFold(n=y.shape[0], n_folds=5, shuffle=True, random_state=0)\n",
    "\n",
    "acc = np.zeros((5,))\n",
    "i=0\n",
    "yhat = y.copy()\n",
    "for cw in dvals:\n",
    "    i=0\n",
    "    for train_index, test_index in kf:\n",
    "        X_train, X_test = X[train_index], X[test_index]\n",
    "        y_train, y_test = y[train_index], y[test_index]\n",
    "        dt = ensemble.RandomForestClassifier(n_estimators=51, class_weight=cw)\n",
    "        dt.fit(X_train,y_train)\n",
    "        yhat[test_index]=dt.predict(X_test)\n",
    "        acc[i] = metrics.accuracy_score(yhat[test_index], y_test)\n",
    "        i=i+1\n",
    "    #Run here the code from the next cell for checking each of the performance plots"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": false,
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "#You may run this code for each iteration in the former cell to get the surface pltos.\n",
    "#The prediction of a configuration is given in yhat\n",
    "\n",
    "M=metrics.confusion_matrix(yhat, y)\n",
    "\n",
    "from matplotlib import cm\n",
    "ccampaing = [10,20,30,40,50,60,70,80,90]\n",
    "retention = [0.1,0.2,0.3,0.4,0.5,0.6,0.7,0.8,0.9,1.0]\n",
    "TP=M[0,0]\n",
    "FN=M[1,0]\n",
    "FP=M[0,1]\n",
    "TN=M[1,1]\n",
    "campaing=TN+FN\n",
    "profit=TP+FN\n",
    "\n",
    "[xx,yy]=np.meshgrid(ccampaing,retention)\n",
    "cost = 100*profit-xx*campaing + yy*TN*100\n",
    "\n",
    "from mpl_toolkits.mplot3d import Axes3D\n",
    "fig = plt.figure(facecolor='white')\n",
    "ax = fig.add_subplot(111, projection='3d')\n",
    "cost_no_campaign = 100*profit+0*xx\n",
    "ax.w_xaxis.set_pane_color((1.0, 1.0, 1.0, 1.0))\n",
    "ax.w_yaxis.set_pane_color((1.0, 1.0, 1.0, 1.0))\n",
    "ax.w_zaxis.set_pane_color((1.0, 1.0, 1.0, 1.0))\n",
    "\n",
    "Z = np.where(cost>cost_no_campaign,cost,cost_no_campaign)\n",
    "ax.plot_surface(xx,yy,Z,cmap=cm.coolwarm,alpha=0.3,linewidth=0.1,rstride=1,cstride=1)\n",
    "ax.plot_wireframe(xx, yy, Z, rstride=1, cstride=1, color=[0.5,0.5,0.5],alpha=0.5)\n",
    "fig.set_size_inches((12,8))\n",
    "ax.set_xlabel('campaign cost',size=16)\n",
    "ax.set_ylabel('retention rate',size=16)\n",
    "ax.set_zlabel('profit',size=16)\n",
    "fig.savefig('rf_cost.png',dpi=100,format='PNG')\n",
    "\n",
    "print 'Max profit: ' + str(100*(np.max(Z)-np.min(Z))/np.min(Z))\n",
    "print 'Max profit for retention rate: '  + str(100*(np.max(Z[5])-np.min(Z))/np.min(Z))\n",
    "print 'Campaign cost:'\n",
    "print 'Accuracy: ' + str((TP+TN)/(TP+FN+FP+FN*1.))\n",
    "print 'Confusion: ' + str(M)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true,
    "scrolled": false
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python [Root]",
   "language": "python",
   "name": "Python [Root]"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}