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": {},
   "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": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "import pickle\n",
    "ofname = open('files/ch05/dataset_small.pkl','rb') \n",
    "(x,y) = pickle.load(ofname, encoding=\"latin1\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "and check the shapes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "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": {
    "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": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.8316425120772947"
      ]
     },
     "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": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "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": {
    "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": {
    "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": {
    "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": {
    "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": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "TRAINING STATS:\n",
      "classification accuracy: 1.0\n",
      "confusion matrix: \n",
      "[[2375    0]\n",
      " [   0  523]]\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": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "TESTING STATS:\n",
      "classification accuracy: 0.749597423510467\n",
      "confusion matrix: \n",
      "[[852 161]\n",
      " [150  79]]\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": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean expected error: 0.7531400966183575\n"
     ]
    }
   ],
   "source": [
    "#The splitting can be done using the tools provided by sklearn:\n",
    "from sklearn.model_selection 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 range(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": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x360 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#The splitting can be done using the tools provided by sklearn:\n",
    "from sklearn import tree\n",
    "from sklearn import svm\n",
    "\n",
    "PRC = 0.1\n",
    "acc_r=np.zeros((10,4))\n",
    "for i in range(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 range(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": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAW0AAAEqCAYAAADeXAH2AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuNCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8QVMy6AAAACXBIWXMAAAsTAAALEwEAmpwYAAC9IklEQVR4nOz9eXBk55neif5yQyZyP8iDfalCAQXUxtpYxVVFikuJlChKbLbAdsu0mzMT6tvTwTuOuDEe34kJzx27w3a4o2fGms221O6mZui2WyVLpMluUSpKFFlciqwq1L5gRxUS+0lkJhK5Ibf7x4uDTCQSKKCA2qh8IhAAMs/yne+c83zv93zvYsjlcpRRRhlllHF/wHi3G1BGGWWUUcbaUSbtMsooo4z7CGXSLqOMMsq4j1Am7TLKKKOM+whl0i6jjDLKuI9QJu0yyiijjPsI5tt58DNnzpT9Ccsoo4wybgEPPvigodTnt5W0F068ru2vXr3Kzp07b1NrNg/ldm4uyu3cXJTbubm40+08c+bMit+V5ZEyyiijjPsIZdIuo4wyyriPUCbtMsooo4z7CGXSLqOMMsq4j1Am7TLKKKOM+whl0i6jjDLKuI9QJu0yyiijjPsIZdIuo4wyyriPUCbtMsq40wgGoa9PfpdRxjpx2yMiyyijjAIEg3DsGGQyYDJBVxcoyt1uVRn3EcqWdhll3ElomhB2U5P81rS73aIy7jOUSbuM307cLYlCVcXC9vvlt6re2fOXcd+jLI+U8duHTZQogkExllV1jYdQFDnfunYqo4w8yqRdxm8fCiUKv1/+vwXyvGXuV5Q1n2/dg8Ltwj3TkDLKpF3Gbx82SaLYJO5fEYuDQiSKKTZH16tWlFbv5p2g8ESrEbLekEgEYjF49VVobd38dpSxJpRJu4zfPtyCRKHzWjicXwa63fK0pglhNw2ewB+oRHtzHOX1527TyLDKdEHThLAHByEQgDffhNdfL1vcdwll0i7jtxPrlCh0XpucdNHZmd/9tsjTCyOEaq7GFEviD1Ri8nlR7f03N+cX9jWGw2s711qmC6oqFnYgAD4f2O2bP60oY80ok3YZZdwEhbw2MWFYwlfr4P61oWCEUEwmul58Do1xVHs/iiuzujlfsK9rcpLF0WU1rGW6oCjw4ot50na5yl4vdxFl0i6jjJugkNcMhtzt5asiy1dxzKO8+jD090N7++okXLCvYWJibdbwWqYLwSB88QU0N4vFffTomqz98qLl7UGZtMso4yYo5LVAILKUhzaboIotX7MZjh8XMh4aWt1FpWDfnMGwdmv4ZtMFfTDo6JB2pdPyealrL0d83naUSbuMMtYAndfS6Wz+w9tBUMWW73pcVPR9BwZIDA1trB2FKCWhrHTtt9ulpowyaZdRxi1jFYLakAFebPmu10Wlu5vK0VFpxO0YSBRFoklLXXs54vO2o0zaZZRxq1iBoDbVAC8kTLM5n6tkpQMuDCTp2tp8bpPNsHSLBxL92nt7Rec2m5e3t1g2Kevcm4IyaZfxpcMd44cVCGrTFQJ957WMBAtkah4dhcbG22fpKoosSL75prgAHj+eb1MxwZd17k1FOWFUGV8q6Pzw/vvy+7bng1IU2L59CQmVNMA3mqBqrdkBFwaS6COPCDnC7UuMlU7LxXV0rN6mcmbDTUXZ0i7jS4VSVq7++Z2amS8zwNkES3M9WrGikNq6Vf6+nRbuWttU1rk3FWXSLuO+wWqyh/6d2bzcY+5uzMyXKAR9m6CXrDX8sjAi0mxe/3nXoy2ttU3lzIabijJpl3FfYDVZtPi7o0fzM/eb6ct3RP/eLEtzYSQIBkHrK9Hm4ojIP/iD9Z33VhJDrTUkdNNDR397USbtMu4LrEa+xd+l0yIz61iJt+7Y+lgpS7PEaLGWAWTVNhdHRKbT67Nwy4mh7guUSbuM+wKrGaurfbfazPxOxYEIGSuoqiLHL8G8QZQ1DSCrtrlUROR6LNxyYqj7AmXSLuO+wGrkezPJdCXeWlW12CTdpKTikF7OvBrKmgaQVdtc0BGRQODWdPNXX8278d3txFBl3+6SKJN2GfcNVjMab0UyLSZ7EO841RxCOb45uklJxeHVapQi5lVZm/ysEKTr4AwaKmqbZ3mzFjoiq+cHWS9aW0USKUWWd5JEy77dK2LDpN3Z2fkdIARs6+np+cGGW1RGGXcQOtkHg/DGGzA7C+7kPK/5TCgd9RLxd/o0HDp0S6SxTHEgitYfQilcLVUUFG4iPweDMDAAJ06g2GxC+m1dwDrbtBbiLTUC3mkS3Yh2VehKVNDHXxZsiLQ7OzufBQZ7enq6Ozs7n+3s7DzY09PTvUltK6OMO4aBATh1Cjwe6Jl2c6TDy6HeXjh7VjbQM+ytE0sUB6K4es6g2vtgKLOM+FacLeiEOTYmg8iLL0IotH69eSPEe6cTQd2qx02hHnX2LBw4IDLPl8hS36ilfRo409nZ2YVY2u8Xb3D16tV1HTCRSKx7n7uBcjs3F3e7nUNDVsLhKrLZLMGIic/ce6gwR6lpaCBVWYl5dJSx46fxWxsJh3vweCTbXzhsJBg0oSiZxc9K4ehRI3OXJmiMfMpcpYfE6CTRkyfzQTCrwDI8jGN0lIzTiTMcZv6TT0irKpFAYEUZpFR/6sdJ19ZiHh1d8/kBjOEwrslJDBMT5AyGVc99s+OYgkEyikLW41n1vhv37s1vOzEBExM3Pb5+jRgMVM7MEItEMMzOrutaS+FuP5+F2BBp9/T0hDo7O/8tcAxYRtgAO3fuXNcxr169uu597gbK7dxc3O121tWJUTc5CdkszFs8nMw20VWfpc4QJ+ht5czoY4yGZvF66zlyROSOEyfEAB0ZyRviK6oPnTVwrEd2aGwkuONRtLT35rP3ujo5QSYDzz4LR45AW9tSR/WBAfl74fOS/Vl4nMZGeOSRtQfbdHbKz0Y07WBwWYddnZjY3PuuX2MkAlVVeF0usbTXcq2r4E4/n2fOnFnxu82QR97v6en5087Ozn/Z2dn5nZ6enp9s5JhllHG3cOQI3LgB1dV6vn8H2uHfQfFMoYVryJxy4HQGOX1atG+jUZws9NoAAwPQfSJKJjyHyeOk6zXHUp4oWPkMmqs5dtybVyqOhlDS0yu7v6yUOW9gAH7xC7hyBXI5OHwYXnut9AWuNzKxlJxS6AC/XqyUY2AzUXiNL79c1rRL4GBPT8+fLvz9L4BXNni8Msq44yjkpkRCyHhRSm3zgOJBDYKpG65dszI7Cw2eOSaG4owknUAlLhcwO0vm9Hma3BH8fS60I3tRDnmWnmxBuNYK01H3RtHefA9FDa6sNa+UOW98XBZKq6vBapXRZDUyXI+bje764nTK743q2KV06jVIHuvGlzz6cqOk/YPOzs4/BAYpe4+UcZ+i2AA8fFgWJAsNND0T6enTYDUk+cX/PYmBLHubhohV7eHll+14A9N05zL4acKUm0FFAzwlz7mEv2JzqPbo8kW+YJDgwArufYUlwC5ehOlpkQE6OkqT4UqlwVazus1mWcxLp+Xvl1/eWEfrVrAu5ZRxS9iwpg2UibqM+xrFBmChXFyIQAAslhzPPxTkwmwUu6+S/TUa/ooI6bQdpa2KrsMn0GbHMDfBwOwBOA1tvuXSxxKlwmxFOZ5ZVs4r+MbbHDu1lYxhDNOhg0vlFr3RoRA8/jhs2yYkfuCAHLyQtEslZwkERF+22Va27tNp0bFzOTAY8rUh14vCwQGgu1va0t2Nce/e0tt9iS3ljaIcXFPGbz10K3q1guf6Gtr16xWEJivZ5bmO3TCLf9aFqcMpfKQoKK99GwZmeOMXdZz6cweG+SSHjFd47eGrKMbZJYuI+Vm8t2Q5L20yQwYTTVYNf3gOTSsg7ULWLyz+qxcjKEThVOLcOfj+9+XzgQH49rchlZK/i6cXZjP09OQtbbN5/cRaPGAcPLhkWmPS83yXg2nWjDJpl/Fbj2CwoOD5pShdRyZQ2qryGfU0CIfFKP3qV+eIRr08d2QHbW6NgVkV3I78wRQFzaMwmxYOJBQjHDGjGWpQTv96IXrHvdwDpFiHNZtRBz7HNJTBb7RgapxHDQ9AsGr5PqXqNRZeXDgsYn1vrzijZ7PyeSwGb78tnhWlrO50Wix3pxPm5sQ61ztqrcRarD3BkmlNZqWsX7ein/+WWOpl0i7jtx6LfOGdw//uWbTZEZT6OYJHX1n08EgkZNtIxERjI7Qd8AAeuo8tzvQXOUxVhZd7esCQstNhT6NOXoZcjqBnK9qvzqNOvI/S1r0y8aXTsHcvB9urIBikLfMfUU45odu03LdwpQW+Bes1ODmPNppF3deIcvgwXLsmC5jt7TJw7Nwpvo6livS6XHKBRqN4qAwOwr59+eCewnaUiqIcGZGfWEyO1dYmPwv7ZHUZZ6Ppa4tXk4sHxS8RyqRdxm8NVjLEFvmiJ4opl0HtqIJQGK0/RCbjXbJAOT0d5ZFHZL/TpyEyGaWjJow/4lmULxRFvO6OHAGw0ubbhRKoIPgLN8d+VUVmYg+m2t101VxCWcGiDJqrOdazl0w6hylmo60qJMQZiSz4FnYv1agPHpQddaKamBDXwsl5OWckhum6la6v+FB27JDj7N0LtbViTR8/vpwwCxcOf/EL+OQT8Yns7xcdfbUKE3pegFOnYH5eFkifeSZ/A3TXQZ201+uOWAx95PV64d13ZUZTX/+llFnKpF3GbwVWk0wX+WLAhnpiGCUUB5MJtd2LaWjpAqXDkQIWIqUno3z+1jhTapg6zzDqy7sA7+IxDx3Sz+6F1kNobCcTCdFUew7/XAIt5kBZwaLU0l4yBx6kyRnGfz2DdukDlMEzQpQHDuSlhN5eiZHXR562tvxBzGYZeEIWmhqz+K1OtJ1HUB6wSWx9ILDQPO/qKRQ9HrH8q6tlCuF0yoiUTi9tR2GOloEBCf232WT/TAbeeSffzrW4Na4H+nF7emThtKPj1kL97wOUSbuM24p7RWa8mWSqKIhPddu3FxusKN5lXLZgwJLJQINjlmwW4hYvZMeFBFu9K7ZBbfNg2ubBX62Km9+r+0Ap2L6gs1RVweRy4M84MFVOoh5uhZoO0Zbd7ryUEItJhI9+YQsLiuaRERgeRm20YjqXwm9sxWS3oh5oljbqQn4kAj//ueQzcThWaHih3mOABx7IDw4mkxB2YY6Wo0dFIx8fh+vXoaVFOq+wnRsh06Gh/KqxXlmncFZw4oQQ9pe0HmWZtMu4bbiXHALWLJkWWXvFxl84bMRsFtm0Z8pDhWWOR2sGCSUr0VBXzbmXVwAcqGpRtGRR4m3l1Vfp6mpFGwijzo6hnE9B1pTXhX0+Ia5HH4UvvpALSyQWFxTdV67A1q0oj+6gq/I6WrMX9cndKPqgMjAg+vT0tJz71Cl48snSyZWW6j1LteKuLrGwIR8a2t8vFvbLL4u3ysMPryzBrBdDQ/CP/3Heo+VP/mQpcR86tEQz/7JZ2VAm7TJuI253YribWfHF32+0tmwwCG+95cFmA4sFnviaHY+rgVDKjclilGCaYLa0Rq23xRxCDQfQwioUBsyUSLytvPoqSveCtwaIqK5buIvuLkP5opjhsJBvUxO5oSGxwv1+FGcKZY8VvLl8Y06cgMuXxRrWSU+PfNRlDr1deofpZKjLDvrnhw5JO3RCbm+X/0Mh2eeppzbnBoAMCHo9ub4++b+4jmU5IrKMMm4Nm1XPthSGhpYWWCk2Dley8ks5OKyVRwYG4OJFGy0tss713HPw2h870AbSqCd+hnIqnvfuKDjYohE9GSXwyQCOinmqXEUBMwuJt4PjCbTKHWKznx1GG3eidlShjF2WA8HS8PLJSSGu6mqCV8bRenOosSmydntetz5xQsi8uzvvkJ7Nio/2229L8qjRUbh6VX4DXLokv3U3wKNH4a23RB/q6YGtW6Xzv/c9Ic3CSMeVNPK1kGlxIE7xMdrbxcLu65Pf7e2rH+9LiDJpl3HbsJaglVtBMCiEffmyqATbti234tdi5d+KfGNIpSAcIjdvA2zCQ54psMVXPNmiEX0txfBIFUa7je+qFwkVBswoCsEX/z7HTl0kE7eSuOiAcAu2nhimM2m6rD0ivQwNwUMPiYYciwnJzswQ/Ogix7K/S8ZoxvTk4+x/PkKD15sn6MJFS7s9n2v66acJbnsQ7TeXUKPXUXI5aGiQbXM5uXF+v2x/+rRIMGfOiBeJ0SgX+L3vLY2w1P0f15tcqthtD5b7jre2iiRSrGn/FqFM2mXcNiwJWhnaPE1b04R3fD7hirq65Va8buX39so2IyO3Vti30PBr84U4ZDiNIeyjw56mzbfgLVIwpQgmKtHCNajB/LEWq9fEK6nzjhGOpTk/7MFRkcIcDaPnJ9EcW8g8qdLkDHPmqg1DzwDt1fP4h1No2/ahdLRIQ6enhXCjUTlBNouWcJLxeWnKjeBP5wiPJuHz/0O+7+mR3/qiZUeH/G21Emzey7HPW8iM5zA5H6TL8Ocovb2y8Aj5aZLXKyQO0mm5nBwrFpOBIJPZWIGG4hty5owseuqDRuHxWlt/K8laR5m0y7ht2KimvZpftcslFnZdnagApbzHjh6FH/5QZtLDw/mspYVkqhuOFotIwsECsl1miR8M8Hv7zmFuexh1bgglXQ14F/Xa4MAMx07UkTnlwFQUN/P44xAMWqloaaU+EiDbN4996DLH/7mJrn95CKXVi2oOYYol8ePBY4+CNYnf2oapagLVfn25ZgxCnA4HZjNoWo6YrRGXx8mWz96A2esysnV2ik92dbW43Z07J0mmslm05CAZw1Gaaubxz3nRtj+G8kKBd0ihVHH4sMgx8bhY706nnN9uF+u8r0+Iu74+L2HcZMHBGA4vvbH6SDs/L94st0NbK2rD/bZgWSbtMm4bNqJpr82vOoyKhuKtgiK/jWBQZtCZTP68etbS4vczkZBgv8KZvaKUGHRQ8diTNGR7wCUXlK8/oAAKGdtSt+X29rxn3cgItLVVkg0b8M0O0+GI4B+0op2tRfHmUI4fo8tuQos5UF95FP7mAtpkBvWACeW731maG7ogZ3QwkOW4qQL7ZIyYU+XlQ6M43wkIqQYC4idtMsEHHwjBjowIySaTqEkDptEY/lQjJm8F6itPw/4t+c4pCOVXX/ovJPGV2Zz38fb55AJDIdi1S6Irt21bPdy94Oa6JidlUAE5yUMPycDi84n8oi++bjap3kuuTetEmbTL2HQUGjC36jBwU79qgiiLMeRLX7pC77mhIZnJWyz5rKU6BgaEyGtqxBJ3OuVweu4kszlv+MViYPZ5iDz/vBCKqhJEWQz6MxiEs+z2vNtyLAZ/8zeyeU0NmDJJWu0h5qwxYlkr/lAO0/wsqikImg0yGZSOehS/H9LTYE+iVIXBXiK9a8GinpaGTAN0PCS5udOfn8EyMSGjUEsLOBwEf3kK7fIk6rceQ2lGSNfvR4lE6MqOoE3UoLqTKB/sgi2vybEXUsMeO1FHxubAZPLS1eWV07a25m90YdbAyUkZrbLZpeHuK+hShomJpRGemkYQBc25S2YzHs/tIdPCCMqeHmlDPhrqnkaZtMvYVKxU7CQYXH22XIzVrPRgELTTYdTIQsX0IlYvTDUNwi8tLcsrdJ04IQSbTArpzs2JcVeYO0k3/Ox2MR737lVg+3aCQeGmycmFxFBIsrwjR/K5mCYmxKNucBAO74lhHvUzxzQuc5yXnxghfbkHtTqJMpiCLU9Jw/UcHSDEV18vASrf/76wP4geVKDpFufmNleaubj7JfYYAigPbSc4FOTY0B4y45cwvV1B1zM+lFeegR/9CEZHUTQNpcYio9vkZD6nyLFjaONOMr3NNH3zAP6QM9/NpbL32WxCgr/6lXTmpUviR77SgoPfT85gkM8WSDyoZSQdrd2Eybydrper11tvfm3QtbF335XrPnHivslVUibtMjYVK1WUKiZyfduVSHwlt95FrohUYzq7nS76UFxLWb2QxFyuvJtwcTttNvjmN8XQeuIJaG7Ouzp7vULog4NyPP16gkHTEkt+YEDeeatVBgldCj75QYzAcIr6eit1NRkecV/j1We6SW9pwzw1TlrZg+pMo+xrkSrr+sgQi+WLDejm+sCAiPfJJFRVyXevvroolyhA18EZtNkKzA1Bjv/7embm4lyt6qDr72xBO/1rMhPTNO104fd1oG1JocxMC0E7nZJAamRERq5wmAWBHCIRVJcFUzIueVk8BtTwhGQaLLZU29uFBD/7TPTo6mq5Wfpi6Qo3NxIIyGC0MGBpNJI5/DBNNSn8cx60tOP2kLaiyAg7O3vfhbyXSbuMTUUpC7mYyIvzHRWnf9ahKwCFVvrAgHBcZ6eDEA+i7W2R8POiKMabyTI6L8ViIu/qtQN0C/ydd4TDzGbhUv16FCVzc0s+GORVx89509iGfSqBazbFoQeuodw4SbAyKYmgWloxXUnR5RhGIba02KRebED3EElJvhM0TXyqIZ9vJJGQCMpIBGVwkL5tz5HJbsX3gEK65gBaOo3qiGMiiz/swpT0ow5dh8CAHNfjEeKNROTv8XHRlFpb4exZlHSaLoMb7UAT6uDP877oR48utVTTaWlrMikEmMvBli15Qi4cdQuTRunpXhcGLPXJ3Zg+zOKf82ByOVDNIegrKCCxmYuHbW0yk7nPQt7LpF3GpmIlwiwkcsgbab29QsQrpb0oDKIxGoVke3uFxA8dcqAecgDLtZfV4jh0V8RCw7ZwkfPIETE4OzvlfT58OL9vdmYW1TyAKVGH3+/A5RJuXVLURdNorQrz6rcinD0+jbcySahuB1rAQDjdQqallabwZfxmK1rAgPLai/lQ9ELy0GUSj0dGBYMBduzIe2+0t8P58+Kw7nTC4CDq3gSmChNjc17q2xyos2dR0tN0fbsG7eJ11MooSsdu6A3JsWMxqKyUzqivF3fCUGhJLm1lbg7FPiq+6LplHQiIZn3unOz72WfS9vl5OV40Ku2OxWTk0zu+aMplCgbzI2BvL8qHb9Flb5DF2GceRTn+i6XZDPVV3VhsmUy0bmxWlOYdRpm0y9h0FBNm8bsBS61ZtxsefdS47DjFQTQul0gaL74oxH3kiCxIruYFUMowK7SUCw1bgOBQCG7M4knbCZ2PYpqP42uY5/hgG5ksRD4YZ/tTFziaqKR/y4tUb3Mtd5RQVYKJSt76xMbpof0k00YMZ7M8Yp7E2OSA1BX8lSYSVfWEbSGCaRdKMXkEg/l0q3pVcd1r4xe/EOnk8mWZvlitQpRGI4rWR9fWGKe3P8qhr4ZQ3vqNkOHFiyjNzeBW8yurjz8O/8f/IX+HQrJa2toqZO315nNpu1wid3zwgejVFRXShmAQPvxQrGujEfbvl/Y1NoqVHYvlvUv06yvSzjKKItJMQfIrpaNGFmOne5du39+/LNSf11/fmAV+H4a8l0m7jDuC4nej2JoNBk3L9ikOovF45CcUEqOwrY1V3UxW8uoqlEZcrvxAEhwKcewfXyATS8KAn8PuC7RFL6Bd2UXG+RJNj7VwOWtgwLCd7stZMqkYJy+7IB6jxhYhYnBKhON2Be3I7xAeSuF2OwmHDYRHwjibq8nWqBy2nodEghNaJT+f3kLgzxt4+cV56issqGbD6gNROp1f8Bsfl1GvuVmI+2tfg2wWxWdkz7V3UEyDolsfPgzvvy+ShabJPo2NMmpOTIhe7vGI+8v3vifH1r1C9MHi+HGx7pNJid8fG4OpKbkRIDdT18Z27JDPKiuXVnIv1s7MZrG0i89T7I9e+P/Jk/Iw+HzycOiLJm+8ka8KVOiM/yVEmbTLuO0oZQS1tYmWHAqJNBqJGJcEtkDpIBqdT/LHWtnNZKVF0eNvRbHPRYnFnbz8sj1vgfeHyKRzNNVn8PfG8JijKI55QhYDWgBiN2wYjTmYmiRjqKOp04F2Mcqpt8ewG+KYrQFefrER8KC2efC0Qt9pmDeAvcHNnNGFa3qCto4JtKdeYO7X1fzqjIfggIEP3p3jOwcmqPXcoOu/dKGs5O+oqkJa4+OiKQ0MiG+0wyGWc08PeL3Yf/lLgqf6pa88V1AssTzR2mxC5CMj0kGhkIxgFotY07A0fFzvyH37JHx9dFS2zWZFTrHbRZ/+gz8QS13Xt//sz2Q2oFdyL5xyLRC0Y3RU2qEPTMUzjuL/X311adIZfaHj1CkZeHp6xCK4T9z3bgVl0i7jtqIo4+iiDKm/j3pRlOPHXZw7l88/VOwCDELYy2azBdGIGioqHnSPNL004pJF0YEwmdMX6HBH8M+6SAf2Qqv47KntXkzmG/jHTZisZlRrhGC4guPRXdibzMTsPvZ0bQGXQsLlwx9ywIyfw8azOJwGpsIVBC7kaN3vQSHIa0dmOLJPCgf4fDbSQ02Y3/kZWrYK86VzBJK/QyRegWKPEgrlyLm9ZNKTaCEzJCrRuiOo7koUsxlOnyY4a0KbSKEOh1FmZoSg7HaRNaxW6Y8Fx/JQyMAvUkfJmI2YxkN0Ke+jDHwulWqCQSHnLVvgW98Sl8LqatizBy5cEKlj587lFrIu8O/bJw7o3d1C3CYT/OEfwtNP5+9LX9/S+pK6BqXfwIW6luna2kX/7MXvirU1yI+4ra0iiRRbAbrroP77S4wyaZdxW1Ei42ihtxog8uzwcOXiPt/7Xj6grjBvkJ6krjAwECCIwrFuhUgEAj+V9M2Dg7IPFAfVaZhyGfw0kUjOEr4RJriQIlVp9XL0v91H/7kI1dU70ZKPEZ6Mk+lx09Fiojea49en69i1qxbsclxfwwxvfRHn1PQODKl5Tpxz0XYkhHL8GEomwyGTieDRV9DSXsykOT6yg0xFJab5OC//jsbgWDPZpJn4RBrD+CgmVwKzSeGNwDeZDYM7Ba/91Y/h3FmO9e4jM6Fhyil0mWdR7AhJXbokhFpXJ/6NFy4wM3+ZzOQUTYYx/JXtaHV7UGYGpBOmp0XmiEbhlVekQ3/xCyHySEQs6Rs3ZEAotpBVVW7Yxx/nPVusVrlhjz22dEZQqImv4KttHh0VqWYlzw29bFmx9FFI7G1tYlmHw0v9Lr+kKJN2GbcVi8mSAvnYkMLqWO3tYoglEkbcbjHczp7V3fqWJpsrrqxVGG4eiUhm0XPnxJHB5xPpVU9tsegd0lZF1+ETDExOcyLQyKkhH91a3u3w+BceJic9nPopHD7chNMYhdHz+CdSxGJBKrfULKoWHg+0tjVz5MmLhM/G6GyOEWr0ofVPL8obwd4pjr2ZJKOC1uvBPjhFh2MMf7Ka+vqdfP+fhun/eILq9o9wpEKoU1cY+OBpTn1aiWdPMz1DSY7Um/CY68nMZ2lKDuA3tqDFrXJNPp900Le+JcT2zjsQjVIzO4jJfgh/UMU0H0Tt/iUYktIhNTUibwwOwo9/TPDZLrTmp1EjZgmy+eAD8Vax22Xxr3CKEwyKPh6Nyo3LZGSwsNuXlhu7mWfGwvfRkyelGvxKGvRapA9FETK/z7xAbhVl0i7jlrGWBftiGbK4OlYkIoZePG6gokKkzu7uvCfb/v15P+niffUZtT4wTEyItNvQIIPEf/7PYuQtCXZTFJTXvo3ndBhbdTVNHY5F3/FgUNbtrl0TQxPg8LYYX28bxdPqI3p9iB/5PfT2FhiPikLbf/MCDW8mCdmd4lvcnoIh0dm1mELG7qSpCWLDaWLVW/H7tmCaj0velC9+Tqs2BsFeIaPZOFTaMWSzMDuLIW0Eixk1PY4p68FvaMbksKGmE/DQw7IQaDTKKJdMCokbDPjiY3R530YzgdrmRknZQN0j1qrDIYTrdBK8PMaxi34yVjsm44N0Gc+ijI3J97OzsphYmKJR0+QcBw7IVCaZlGRUly/LiHvp0lJreDUCVRRSW7fenGTXIn0Unus+TQS1VpRJu4xlWMszv558O4UypO4goKdM1bOAtrcn2bmzko4OeO89kVinpyUK+sCB5fsWuv8qirgBjo5KDpHZWVm8dLtXSH+hKKiHFExD+XbMzAgHnTolPBSLLRB3ys2rX0nhnRvig8t1OJgkNqry8v/bJ14efRqKqtL1em1Bn3kXrUzVXI3puAwMrjoHLz9xQ+Qd9zyKu4pgxITmfhA1NYoyNQVmM232cQ41VhGeC9DhzNLWmET5zu/TFTGj/fIMamoCxfV1WXhUVemweFxGEqMR5ubIejwoLS6U7DCYjZDMyiJkRYVsV1kJDgfaYIKMJU2TcQh/uhbNn0DxeuX7cFgIvlBzVlU5Rzwu5G02i6U/MiL/37hx84XA4kIHq2G90sd9nAhqrSiTdhlLsNZnfr1pVwsNoaNHxfLOZuVdr6oS7drnk+MZDCKTut3yU2pfPReILmt88YVIo9evy2+bTbhptWC39naZ6WezYii++KLo1P39oh54PLBjh5X0cy+gXblCZuQiuw1jhMYHSD/5MEx+vNhRytGjEM4WlBGTRisUqgQOFL6+SFjBkIFjZ2Nk0jlMht+j64F5lF0aSlMTrz3uRTtxVarWhMStT9m+HWVvs+hHV69Ko3/zGyFOt1su/FvfAreb6OQk9uvXZcTcsUM69sIFIdvz5xez6KmTPZiC9fiDYUyWa6iV74IxLNYzCNFv27a0A7dskSlJTY10vH4Dh4bkxs3OrvkBM2/dKu1fLZ/BeqSPfMjs3Q1Nv43Wfpm0y1iCtZLxRtKu6guJ7e354gSVlclFl77VDCt938KQ+GBQZJaaGjEQIxF5X2020bWL8wDpvDE2Jpb5U09JO/RU0C++mE8F4nJJdr+hSTMjCZVhRy3OeAJz72VwZhbzsAZ/+BOODR8mYygqI0axSpD/R9Mgc+BBmpxh/NczaH/9ryTyMB5Hef55FPMMhML5DtZDOcfH4cwZgiED2kAGdX4ExTIn21RXw3e/K6NOZaXozZWV4q0xMSEX7XIJwadSKE0Oupp70L4YxGwzo1maYC6NEgyK//fXvrY0xl93BfL7xQqPx+Hzz2VapFek+Oyz/D6Fna5pcmMz+X5zv/OO+IevZiGsNQCmMAuYhMzendD022ztl0n7twRrHfjXSsYbiQAu9CA7dEhm0+l0iNbWWmB1w6qwfTMz8NOfCif19IhxFY8Lf+hh8aUye+oLl2azHEPPynnkSJ7gt2zJp279q7+Cy5/vYC4EseksT9bf4K3JRzkS/Jy22BRKLIaWbSDjqaIJP/7CMmLF0G+E2YwazmIy1uHPNmCKXZE8G6oK77xDUMugubeh/tERlCceWOImR0MDwf/nXY6Nf4VMKoOJPeLS19oqxHz2LBmPR4hbzyXi9cqolkrJ58mkWOGJBEp8DmxzHJv7OpmsAZPjUbq+UoFSEc1PdfS0hpHIkso3mM1idVdUiGzS0iLnKM438sYbsl0ymX+oYjFylZWbV/lZzwL21FNS5GHfvrtjZW+0+sdNUCbtLwlWI+X1DPzrIeNbjQAudY6rV7OrHrc4R/fAgBD24GC+OMvu3cJZb70l/DE9LR5txTCbxTgcGpI+SSbhpZeWp7Ho7hbO+/RTcDpdVOzdydxYAMfRRk73uJhte4r6yhRdT06hvvcepukJ/FYXpg5nycEuOBRCe/MXsgB58SOU1la6LDVoB7+G2hJB+bfTcDZAMOvmmOEVMsNBTP8xQ9deg/SHPmINDKDNu8kYzDSZx8SbxNqIoicP//hjWcjctk0IescOsXDHx4Vwv/51sbTb2mTk7O9HGzKT+YsbNFkm8Qcq0S73oPhmpDOvXJGpiM2Wzzx44YKMaCMjkm9ETxo1Py8j5WKY6YJP+AcfSBuiURkdn3sOfD6yP/rR6hbCemQGPYGWnuD8/PnlFv+dwEamoWtAmbS/BLgZKW9Ef9aPv9ny3HoIv9T1eTxC1uFwflbe0iLft7QIV5nN4gK4ZcvSc6XTQtAmk5B7JiPHKCRtvc8aGoT8x8bAaHTisKa4fvYGudkoHTUjhHz70D68zHafka6O8ww8/F1odCzKqYWpRI79METmvAOTyUnXyE9R4nGUgXdQ+k+JsL5nD4yPo5EjMzxC0/wg/ssWtB8Oo/zDl6Q4wMHfQ838HNV1AlPUjD/WgKnGh3pwG5ir5cLHxjC4XPDkk/ncH5OTQrCNjfL3E0/kO+PQIVRlBtNvDPg9BzAN9qFu8cOudplmfPKJkO0f/qGQoJ5XIJsVIp+ZkSCaxkb44z+WYxfKKYODokM5HPJjtS74S7YuKSqxqrWRSCydCq30UOn5EQr1szsdHbmRaegasGHS7uzsPAhsA+jp6fnJhltUxrpxM1JeKdfGWnC7F+PzUufyhFE6ClM361kB29pKh7jrAXaJhBDyyIi0X2+3Hinpdsss/sYNMQ6Lc+DrxtLYmHze0SHqQL0tiCORomaHhVDUgikwjdocXchSN0X3BTOR/oVi551RXMzR9apk88v0DtA010PvhJvT8VYORS6gaGNyQqtV/JWbm1EfUjD9u1n8MTcmsxE1doPg2WGOfTJPxu7ENLCDLpebrro+tJgd9Tt7UBIZycsxMQH79onsEAxK47u6xNKdm5OFhFAonx83EoFAAOXhh+naPYoWsaLuGkVxWKR45vy8VIL45BP46CMh7f374Wc/EzKen5fR7pFH8tkCi2/cvn1irUciQtC1tYsPYdbjWblqe+GN/w//QTq1s1MGhpUewrY2GRA++ODuFje41WnoGrAZlvZ/39PT09XZ2fmHnZ2d23p6egY34ZhlrAOrzcbCYSMnTpROQ7oWbLY8V+ztpQ8Ik5MuOjtXXvRMJJbmuH7uudLRkX/8x+ImeOWKnGv//tLFGOx2cbTQfcFLORocbJ9l1hbGba5iNu1AVeGBzjT2njiHm4bwODKoL25F+SJDsHeKD/qaGLC5qHVAOpbEcOE8Y3EzZ2cmcR9oI4GV3qanOTuRIZbNcXJmF68a/j2tphTEYgQv+tE8bahHW+iq+T8Z6A6JNfuFES3lJhM4KPLF+QBaysL2WC/Kt74FT+6DU/NyQT/5CaRS2Hp6ZFHy5EmxuE+cELng3DmxhGdnhXRHR8Ux/eJFlH37UIxZqKqE0xeFhHM5sZKNxnwBhpkZ+byhQYi4slLcdkZHl/p1Fy5ePPWUkLfbXZpES03n9P0//1zkmPp6uWHNzbJavNKi5X1a3GCt2BBpd3Z2/iFwaoGsf7BJbSpjnVhtNhYMmlZMQ7oWbKY8V6pClT4gTEwYVn23dC+z+npx04tExMoutvwVRYxBt1t4arViDLt2yWAwNlYi7fMbUTKnz2PKZXhp9wkCR14gnfZwttfHzm1NtH17HKWtChSFoOcV/q//PcWJXjdTISvNzUB0nk8DNcxb7JwcquHxuAmnMUSr+QYxxcyEbRcBA7yZMPN67P+BmmaOfVxPprkGU/cgRz0+Tni+wmw4h9uf5KXcF5jaD+EfyWHKpVCf2QfDWenEtjaxmmdnRa5IJETOuH5dCPb48byLjckkkZpxG9rnM6gTAyiKRTqsry+fu3tsTH77fNJRDz+cHwH9frHorVax3vfuFQIHWezU/bq3b88vQMDKFm+p6RzIfvrMwOWSEXtqSkLoE4mVp333aXGDtWKjlrbukDXT2dn5b4F/1NPTEyrc4OrVq+s6YCKRWPc+dwP3ajsnJuRHh92eYnJyjIkJAwZDjkAgQjqdXfkAJbB3r5Fg0ISiZJiYyC45/nowPGxhdNRBbW2a0VEzihJnctLGxISBTCZFINCzrG3hsJH33nMRixk5f76SEycMhMMmKiqSZDLznDw5x9atKcJhaaPJlOPTTx3kcgaSSQMHD8ZoaUkxMZFlZMTMlStuhoZy2O1Z0ukIW7caeecdN5WVOX70oyyPPRblxo0Khi7H2JG5gZ8mxsanmR/u5osvDhKLGZmeTnH4SJZmh3T2pUtWjnf7iERSZLNpKiqSHO7wM/LrcYZCtYzNqXwxZGX3lip2hz9hLreT65MGrE4zcZuX/pZDmKJR4qEwjbFfMTpfw+fGeT6L1uMxzRGOeXgo8wHP5/454437qEucZ/aSjcF0NeaxOLaeHsJbD5D46ApbZk24TSkqJieZzWTI+HyYk0nMsRjm2VmyNhvTpwf4a9eTzHs7caTaeSn5S9TTpyGZxGg2k3E4MEWjpFMpDOk0sYkJMkYjudOnyRkMpH0+lIoKspWVYLEwn82S1TQqg0ESp09DZSXhtjaygQCWkREqT58Gm42cwUDk+edFEllAIpFg8PhxnBcuML91K6ZIhPjx49i7u7FdvAgGA/OqijMex6Rp5DIZZqqqMIyOEj15UiIqS8C4dy+mYJCMopAtfiluAffS+74Z8shAT09PqLOz8wzwh8CfFn65c+fOdR3s6tWr697nbuB+aSdc5R/8g4Z7Iqq3ri6fDbSxEZ59Vn40DQKBHh55pHPJ9sFg3lJuaxP5NBaT/aNRB263SKmhkKgCugTkcuVnFgcO5AsLnzgBW7fKGtoTT4i0GonIZwuFU/j8cznOyGwCNRvCToakpZUr2mEqKpw4HEHm5jzMzXnQb380Kn2bnZ8nFkmzvRq6vJd4U21hNGahqt6DJZfFYXRy+KEqah518j/+a5VcJs4N+37U9Ad4Zwc5nWlkOuKm0mOhpqMJy2AN5qwFSzCL1eOlxTJEi3uSoGcXx05tIVNVg+nTOo6GBznHUTJTZq6F43S19pFuCuLZskX0YLNZ5I1IBOrriXs6sVe46Wg24rc9SPKBJuzv/2txo9Hd5oxG+enowNHUJJb0rl2iX7/1Vr5m5UMPyecffQT79+MeGIDDh6m5fFk6JxyWm/7ii3KjfL4lGnbPyZNsGx2VGzQxITdCVaUNtbViUedy+dXliQkacjl5gFbLWbLJuNPv+5kzZ1b8bqOkfQpYELrwAqENHq+M24DbuCay7naUknEUhWUWtj5jnpyUMoSNjcI9+/bJ+xwISBR3KLS0uk1dnRB3QZ59+vqEOyIRmWH39opU7PfLulZPDxCLEhsNYrdl6WivgoedtNpbufrFLP9uYB+Bq04uXYJ02r6YcVAvM+bzweMHokzO9WJ0Zfie9VNabSle/U4N/CQIqQmwOnlVvYRinEULj3G4MUvOUoFB9ZE2t6Nk+umq70cbiqA22aCqhkPWOcJTWTqSl2mzjQqpzcygharJpF00RfrwN/8e/b1ZIoMf4/RaiGRsaOY6PJ2doi9brfDAAyIXVEomRbXTh0l9EL+pQtwTaz6E4zkhyHRabsjIiKzUXr4sHdTWJvr4f/lfyg343d8VfTwWk0XC8XEhdLNZJJKJCSHbzk65AXrkUpFUYQoG5Xhf/Sq8/bbcmPPnZWS+dEn+93rz3ikul0g1pao1/5ZgQ6Td09Pzk87Ozv+us7Pz2YX/y7p2GatirQOIrj/X1Mi7un+/eIs9+aR8f+KEcMknn8j/uhea7klSWAglkxFpV+eY8XFZy4rF5PjEY/iufMwzwQ/4YqYdv78V14G9tAQ+4cJ0M+5IiqzaTlOTlWw2xfbtVtLpgoyDiShPVV1gusVP+2GF1llZWGy19/H6g3No2SrUfY0ooXk4fATzaAUXf11DaDSOacTAi495JRQ+6kdptsIf/n3o7OS1WRMD//kSDCZh1CAjj9GIOhvCVPU0/qgX0xcnqbae52zgvyBtd2G2NvCy8gWW3kkZoerrxULNZKQTP/0UJTZKV+7HaNseQw30oPyfP8hHKm7dKiQ5Oiqdff68WNz19dJxuk4cCgmRRqOy4BkOy4/NJjq32SzfjY2JJb5zZ0mf6cVyY2NjoqfryWIOHJBBo6ZGjqPXxdQXLX5LCRs2QR7p6enR5ZD3N3qsMu5j3IIz92p5g/QF0EhE5AqHQ4ystrb8DL6pSYg3FhOytlhkFq77W+sBhE1NYrHrboL/6T+JARkOy1rdjUsp7En4IvEAR7dfJWDPMGtxMxszkba78V+3Epufx2i0YrFkyeWExwCM8SiTv7nMO8kManCOoRkPXY/7UF59mWAgi9ZZgfmzE2i9M+CuRPH5GPqL02jXKwkmxG/5x+c72BJNo8Qm5GJPnJDc1D6F7qExIlNWfj63jVeVf0+rZwblyhW6Un+FlrCjTofQzHUcqLyAs9LM3HwF6Q8/wTZ1Xsh0fl46cXJSOjyZlPD1c+dQ7D8RCziVEs+Q+nopZKAnljp7Vgh8elpGx1RKyFN32xkfh7/8S1kwtFgkuukP/kC+O3EiPyVyOOT8erKYgucj6/HkFyvd7vygoGcJy2RkcHjooXxugRLHWXygbrbo+SVAObimjI3jFpy5i3fZu3epn3ahlKLXtS1V3d3lgmeekfe5uVnyF+nBNIWeL3p9yXBYvE4ee0yO7XZD5x4zDSMZzp928F50K72O/Yz4d8BUFak0qM44X2gWmmojzMYzPPignPdnP4PKbJb0uMrhB5I0NU7jdzajHXkYvB6OLRQOP3v+ZQ5si+By2zl6YYD3PzQzNWsjMZ+jpjJKNmtAM9dBdg4trKBeHkP5/vfRHvoukeFpBkM+Aqka3qz8PV6P/BClpgbFkEJxzUIqA/EALlOcjKMel2JGnRvDmEgIAeojW2Wl/G+3C3HH42LB6iGhmibyyMSENPrv//18KbMvvshXYP/zPxcXwCNH4J/8Exn19NB2VZXOLxxVP/lECHvbtqXudwsjtjEclnMcOpQfkUuVGtN/F9eOKxz133hDnOlzOcn89SWtFVkm7TI2jltw5i4OmPF6LTzyyNJtVpJS1vI+6/t2dYnBGAoJb6TTQt7ZrGTza2iAoSEHp5JP48+mSASMEKnA67PQ0GghMztPRX0Wrs/hjs0weMPNsf8wT2y+gmwW2porqHIkGL2ewV5VhauuEtWXyxuJhiixcJaowQFZO/0jVnyWOfa7h+ger8cUz+JKTGH2OTk2+S0yBjOmMRtdnkuoE39JLPEYAUsdPtM09h1b0JxPo+R6Rbbw+6WQr9VKV/vHaK///1BPvIXyt/2MG6qYyzSj7n8K5dQvpVNyOSHtqipJNjWVRZ2Po5jmpbO2b89HEQWDMmUZHZUgm6kpsbQrKqQie3e3aM6pVP64kYjs296+WPaMgQH5/t138wmchoZEWwKU6WlZcFxe/HPpA6ATvr7SbDYvrWijR0K63bL97OyXzj9bR5m0y9g4bsGZuzhgJh63c/To2t+xYkIvPr0uvUSj8Bd/kde4/+RPxAD74AP48EMxAiMRqK620brHtlhLIBKBObcVm9fKjdE5IjH4eKiJVDKFMZDDoQgX9V23YmY7W9UIVbEoL1eeQDkeh6OvYEpYmPzNZUZ7WmB8FntjFc+87sC1pYqWwHUGTE6a7AmoryPwtT8gs8tH0/DH+CfMaJ42trsneXX+L3kz/jL2ijSund9ArfDAtbRIDlVVclEdHSj796AcVgg6nuf02Vl+GdyN4jBhGn+IrvRnKPoC44MPErTVc+yjWjIzfkzZB+iqeBulvl4IORoVK/XIEenYd9+VhczTp/Oadjye18mTSRkB9WCd6moh5aNHZbScnhbtanBQ9GqAH/5QjhcOYzOZ4H/738Q7RC8mXBw1pWc3BNG/X3tNrP7Cijb79snffX1yY3TL/0uIMmmXsXHcQq6FwjQRnZ1w7VpuXYZRsYReWOR7YCAvqfb0iHG2e7e8z2fPCgeZTPn4kGhUcipNT+cnDG1too8DfPRLIwcdPbx/tZ5Q1Eg0aWduYcbvdEJLiwV3zog1bCRd0wgZkRIO+pLcqByEfVAzfoG5TBOO7hBHX2nhX15XSES8jIXmSI84ebRzKyaLCT9gil1ErTHCwKgUkJj/BVr7I5iPJNDCX4HxcZTqqDROt37r6giaq3njlMJg+vcZSlj4uw9phCpskkyqbqF6u92OlvQsJJsax59tRLM1ocSiIlMEg/lc2j6fkPPsrFjD8/NyLn2xweeTz/Q8JyBTl7Ex6ejubln9nZqSTtWjonp75WYND2OqrZXwVatV3IFK1ZTT0zIOD8tN+/GPxVe0sKKNXj9SH2zKmnYZZdwEt+BX2NYm73goBAZDbs2G0WoS+rFjYhD29sI3vym8MjgohJ1ICJkPDYlV7nTKfgYDfOc78t7rs+22NlAIEhyYob+2jshjO6idz1IZn6e60YLPJ7r4yIjwkz9SSZMthzo3RNBYybFfuMmcvUBicB5j7CoTuUrc9VbM8Qj9QR821xxu0xzZdI5EpgI3s3TF/jPaaD+qN4ASr5RGfPYZSmCE0LydN/2PYXdX4Eo+SVf6BkpjhVi9Dz0ENhsDPz3Hqb+txWZyMRp3cd7cwrahM6j9n8LopcXFQnVPHaaghn++BlM2jpq7DmPzcr5Ll/L+k7t3y/Hr68WC1iMSKyuF4K9e1W+eyDUWi3SGzSYd/umncoMTCdlmcFDceEAPgyVTVycdnkwKyesl0wq1LlWVv7/4QgaI3l4h7eLE64py55ND3QWUSbuMu4ZCCzkQiKyZ8xeLkzTNEfJH0QZs0OZZku65p0d+Ghrgn/5T4YuPP5b9ZmfFqq6oEEPxkUdEvk2nCwy0YJDgG2+jzVZw1Azpr7/AMy96+Iu/mGLLFuGOI0dkANA0qKiwYnfvgSdVNFQy7yawGNKcrdiHJTFDdeI6sWvXeWvYTPYFH2NGDx7TEDm7lUfs52kLO1AiN1Cc4wSNPvoyLaiT11FGRwlm3Lw5fZjLqh1fZw3bUhG0todRWpJSRf3NN4UYTQ9hiP/XWFuraPTO8lBLmKdqLqIknTCKjHDj4yiPpOja2o2WCqJGhlCMUUhbJAdJOi0+lh6PkOONG9I5U1Py+fR03lru7JT/DQYhW6dTOjeVkhulV1Wen5fPq6vlf5dLjm8wkFRVmbKA7OdwiGvOSlJbOi0uhfBbVcy3EGXSLuOuQjfQbxZaX6hR//Sn0HspQd/fjnOoZRKz5QbH7N8mknVw9qxsr8uyOgk7HEKwyWQ+/9F3viOLkc3NEuSnS6pdXcDADMdObSXjqcIUnqHrOY3t+z1897tBzOYaRkdFmo3F5Bjf/jak0w5J+KTCzFyat3/VRjxtxpKz8cdbzuOnkdlUhoNqEA7YaM18RkttgjbDEIrnmzA4SLB3mmOhB8jsfQBTxX66TJ+ioWBPzeILDxO4YaeuySXyyccfC7klk+Bw0JYd4lDyY8KjTbRYYjzVsQ1lOCOdZjQStNahBSpRu4dQEuMoDRkYTUGlRy5clz9GR6UTQyEh5WRSrGG9AKfZLNOcnh6xxA0G0dczGelwv18Gkbo62b6hYWnliro6sd7/yT8hPD5OdXW16NO6dX34cD4nt178oalJ/MzPnRPNWs+VvVKGwC8xyqRdxqai2Pd6NUNoJT/tYr26sMrVhx8KZ1RZE6jeWY48liUdgUx4jo4HpVzN3r0ySy5OT6v7c3s8QuLhsHAEiPX+6KPCB5oG4dlqxiIzdNomCBkq0VDRD3fiF1F+/X4Wf8BGRaWFZFIqvz/9dL7NO1sTfKga6VCSnOuz81FwD52NEezpGP6pClxbqniq2oaSjoB7l1ixDz+M1qaQ6fbS9FAD/i/G0AzVqKFruNjHNsMwdTYbr/73B1GCGvzmZ2K1zszA3ByKLcxL5r+hv/opGkJnUE7tFkL92tekHuX1h8hgxBRoo6smg5KcyOvCVVXiAfL1r4tubLNJaLrLJcTb2SlEq4eV+v1iOdfWys/XvibTGbtdLOdoVNrV2ysjml5Bx+mUAcBqBYdDcofU1Yk0olds9vmWJjdXVWmHrqV961ticRdXx/ktsbrLpF3GpqE4Zz0stV6LCyuU8tMupVfr61DRqJBqZSXMxSvZYsvQlusDtwkTzkW/bV3W7OvLDwhvvSWcMj0N//V/LXq0Hqvhdou7cSwmbXzmGThx3k2v9QH6ppMcesSM2iYDQngkzuzJK1QnHEzOVhFN+di5x0Jbm1j2urSyzRsinFCYHHUSzlZgqvURrajghcMajodmUQ94ULy/D5pG0FzNwJABAtfx2WOYqlXO+X0EIjkeffpFtv+qly77F2iOLZgfhLS2jeDxz1F0v+vt28HnI1hRy/FPG8lMubgWf4iGniGU6V5ZfJwxkjFaaKqYwh9rQuvYhfJ3/p50TjgsI9ngoHS80ShySCQiOnVTk2jIb78thK1b4rlc3hXwzBlZjGxsFNK9cEEc2Z1OWUioqZHj6XJKW5vcnIkJ6bSHHhIXPrtd5J5XX80Tt66j6SvM6fTSHAWFoa9riRO4zwm+TNplbBoK3bXPnJF3s729tOt2sWt3MGha8nlhwQOfTxI5hULi6fXAA1BZaeXVFztRHBJFczTkoL9fzgdLib+9XaTf7m5534eG4E//VLzZhoYkpUZDg1jac3PCXTYbvPi7Nnp7bRx5NIyi9QEqam4as8HJdFqhyh7H7ZmnsdFCbe1CEeKFUccx7uSbu7agedu4OFZJwL6LcEjjnS+CvO55D0XzwNGjBMNG3ngrzalzFgzGr3FoT5wHX0zx7v9ZiTlk58/CT/Mn7RdoNY9AhYVjMw+S+ciC6dMGunIelKlRmVrMzqJdvE4m7KYpdYPhTC3ajRiKwwiahprKYLIY8RtbMCVjqK0uudBYLO8PuWePdHQ2K2S8Y4f83rpVrOwHHhBSDQSEqOfnpaMqKsQq7+tj0dleL8iwsOC4WLXm8cfFcn/uOXkgJiakz955R/TzmZl83m49H4FOroUBOIVEPTIibdZD4Eu5IRXU5lwXwd+DKJN2GZuG4ghEWHk9qdi1W1Eyi58X+m+73fIu5nJC3pWV8u5KviAP4Fl0481khIQL83T7/fIeB6fnyaUNOO1GslkTf/u3Yrk/8ohIs1VVwldWq1jrdrvsV++O0nb+Z9ATJ5io5PL8LjQtQ9v8FQxeB45HazBbYjA9R+i6lYHBCAwq+NoUaqtSxN1mRs+YsKRjqGENKkbQ3j2J8kwLTE6iRRuY/VsvHsc2MBgJD1Zw4UwP2ek9bKubxW9opP/Ia7TWXECL1JFJ7qPJMIY/OIuWMqHMBURUDwRQU05Mxgfw2zswJGOolrBcJKC40nRVfMCAqUPIV62C6+fz4eMXL8pIGwyKn3RFRb6eW2OjVFi/eFFWcsPhvKxSUSHWczYrx5mclM+rq4XcL1wQy113zWlslFwkbW35h0HTpMMrK8Uab2yUz4td//QFEF3n1qWaS5fkAblxQ/TwUnqbPorr59LTQN6HAThl0i5DsAlTxmJ3bVj5kMXbTkxkFz8v9N8OheTHYll0OKClpbTVrlvno6NLg+cObAvzlapBbvRtZ342h9tpBSoYH5f3vK5OjpNMyizeZpM2eDyghidQTsUJelt54z85eaevluBwmnb7KD5rgPS1cR6r6mEs6uWH/99KetlOcuwwj2wd54W9fq74zWy1jBKczWBNhBhNeRkZNaJ+fBnlW0dQa4y4jRF6klYM0xpNo1eYiOWYyu5mIlTDtspR2tM/gNBZ1OoOTBXg37INU6UVdX7BBSYchmQSxWGmK/EzNLZhr0mhvPZ3ZPEwFJKLStrpNv8+mc9P0X1pji5nL4plTrShqSkh31xOiPfll2X08/nEQr52TUY2PXwc5JiKkg+4SSTgG9+Qz/RQdrNZRsKTJ+VG9/QsDy/XNesdO4SId+yQz+321WvonT0rz+30NHz3u9KORY2qAIXTOj1ZzW0qunsnUCbtMja1EGSxu/ZqhynctjBHvc+Xr8/ocomTwPXr+QpShUYaLLXO5+fFzfiBB/Ll1VrTU7z+xCViZg+BqTTO1hoslgpaWmRmHQoJr8zPS1foVWyGh8FcraKYTGi9M4STPtzmGCMZN5eS2zHOzNGWDjFu8KF2VkEiyfS8kayjgc8iPnZ27KF5LkTKMcGNlJvwjJmxVBV/nv0uhy1pXosPA1GO1E2wzzuPu+ICRGY5FW3he7kfciG1j9/xncQ7dJa+jIIauEZX079Fqz2Cemgc5fMURCvyrnWpFMqerSi/9xITN27IRW3ZQtDZjDZbQXjeSOZaL03mhYrrDh/KeF/eao7FZORLJsXaPXBA0q4ODeUz7IFsMzsrn8Vioqlfv76w2DAnssajj0rU48cf53ObPPywWP7F1ZYLR/DCFI1vvSXWv8ezVL9Op+WaDxyQtn/2mfzf0LD84dAfkMJkNaWS2dxHKJN2GZtfCHIV3Myg16WOwpqWra1inOkJ3IpRaJ2bzWLUORzCIek0oKqksbDXM4K3IcV/mm6h74xY2T6fEH5lpRitLS2ynvbP/pmc32738Kf/w++hpjUykXqG+pNYDLM0Z8ewWBM8tLeCz89Zsc1pXIs0oqUtmEwGHGbwJiZw9VxiW2wGy4yFnCPDvMEGlXZmsw4G9nyL7gEvmceNmOJRuuw98NFVuudrSGdyPGC7RGvsCsfCXxOvj1SSLu9pto/8WnTmlhaZWugeG2Yz/N2/C88+S2RoiLqqKoJXxjn2iZWMWyHxyWmY1/DPWzBlYqiOONibhfhmZ2WUMhik0/r6RGpoa5Npjs22WOkdo1H+r67ORy9FozLq+f0SsVhdLURpswmZappYxHrk00K1ZWM4nF8xLnTfCwblt8EgN+Ktt2SQOHtWyFov0mCzyaryapXabyFi915GmbTL2NRCkKuR8moGfThsXCxWUFzTUs+4qYemnzix/B3Vi3B/8onM9D/9VN7tcBiCqoL66vOY3kzSG3VhTdh44gFxhtDXyRobhfuefVa4q79fjLKxMTg76GHbNg+nLoGpYp50pZVdNbNMhyqYS5hw1Tl4+JsKycFqrh03UpGMMD2boOrN/50udxCt2YvZNc1bkWc4PVVFLpGlw5mBU6fJWL9C0z4f/jE7WqiJ7QTpsryFVlmHagmjJarImK0yno5b0Yy1KFMLFufTT8vFa5okgKpqQ+0PoEz9Fa6ZGXC70XpzZHobaGrI4Q9OcdhxBc/8KKozgYKSt1ArKqTTR0fFcV3XlWtrZXTT3XdmZvRFCNmuri5PypomN2FmRm7ekSMy5VEUkVo++EDarVekGBjA9d57co7iB0KXV9rbZQU5HJbF0HRaPFKy2eX+3KvhXqkEsgkok3YZm2aJ3ExlWcmgDwbhvfdci9WlYGnlmcLQ9K9+VeIwdK4pXJ86ckTk2P3784E4p07pJO/l6KsLATbv5/mgpkacL55+WmbZH30kkommifGpuy++8YYYiq6KJI01MfY9U8OTjtMEGh7gRLiFMZuDS31gMSVQXTF89nkGkw04DGZU8zRYExyZ/YhtKSMZu4sDTgveqxN0J434+2swdbShuiU8Uzl7FiXQD4k01LoxBUz4DQ2YvBEZTxv3SaPm5gi6WhiI7+ZErBNbsgLT2yGO7h4ldH0Ma/gEqsWCaeYZ/IlqTPNx2riAYgzB7/xdIUCjUULS5+ZklPvkExm9XC6xovVSQVNTkjskGpVtvd68T/X4uOxTXy8d+utfi9V97pzcbD0drNEoJB8ICNneuIExFis9w9M1r+7ufMVlXaaZm8snV7/bRHwX3AfLpF2GYBMskZupLKUM+mBQpM9o1FgyIE4/ph6afuGCzJg7O8UKPn06H0jj8+Wt81hMjuP1imufnlME8lVu2tuFWxaqcHHlimwzOiqSjL7wabOBzx7FZQItaMVjj7Hf0Uu60kXbCzto8zr44AMxADGZ8Acc2NNzXM02MVnzAIlEDiIzzA37OTW/jz2Ja1z+3Marnqt0PfgJ2mwF6tZHUbqH8vmsrVbI5Qi5t6BOjuOtiHOg4jyKzy0dYzAQfPQbHJt1Mz4+Sq+ljm8O/yVjCS9vTjfhNDlxzNXSZXmLrvk/R7M/gOqaRslFIWMUa9hmk0aPj8vF6n7Reh7bv/orIWD9++FhWaicnxeCbm2Vm3PokFjjBw/my4zpFWlcLvnb7ZZRz+cTYl8IUbVduSKWtstVeoYXi8n5jEZxM+zoEP3qXiHsTVoLWg/KpF3GpuFmKksp7xI90vHqVRt1daUNKL261eHDYvidP59PJAdigHV1CVkfOCA8NHU9CoE4PVE7uZydjg5x6ZuZEYt8eloGgaYmIevmZuEakP137hTCd7vF+0T76XmeqanguiHL3///bOGL3q+SsTsxHXfQ1ZWvn7t/T4bA5TC/29EDZh9NX22n+5qd+PR5RgxObtDMRLqWpmQE0k5et15gu6EPfnZVRo9wWBYPWw9y9jdh/s3AH2Ccj2K+muJPnB+LBT46CtXVaMkzZK410ZHop2cwTk+uHhNZ7ETYkh1iEhUt6WK7aRBl/vJCJYgGgoYqtNNhVGUGZWZAOsRigX/378Q6ttlELz93TnQpnTgzGbkxuVzey+T0aek83fMkEhEy053qLRZp79694l/56KPy3dAQdHSQHB4WIn/yyeXTMt2qvnJFbvjhw3nvlXsheOYOrgUVokzaZWwaSpGyvsZUKq+9Xg6sowMmJxMlw89LKTcHDghXwFJ3W91zLBOJUnvjDEc7rxOI2zmx+3n6+x189JHwyqVL4lWWSsmsf25O+Ozw4byHylMPhpkejNC+30Vruo+uxk/RDu4mPXEJc9Xvc1ltXHxXBwZkILFaITGT4pntN3j0QJJ3TlcRm/Li3mIn2eMmbrZQkc4yn6ukshrs1hxaRkGxWKQRu3dDNEpwKsWxD1Quxh9i0LCVrzZcwj9eQb/3QbwzxxmgGaIOfGOfYkoeIZRxcth5lSPGT/HF/RyveIEbth3Y0wHU9BQYczLybd1K0NfOsU8bySRSmCbn6bIMoAwPi8velSsE5yxolnpUcxxlfl72y2ZFqnC7JQBHr4gTCsl34bB0pt8vBK27/6TTskg4PS1h7i+8IKvMkYiMoCMj2C5fFqs7GFzqCqiq+YKe4QV/895eGUxWw520fjdxLWg9KJN2GZsKnZTX8u4UPvN2e3YZYRcfs/D/Q4fyKVb192WR4E9Po9KH0lFDq9+Pr3OKv/6ilZoa2e7MGeGa8XHhGptNiPy7313wNouGOf5n58mkc1z6WZYjDYO0TQ+w3X+RieZmrO1eEpeg+5MY5mSUK9i4ft0lWnp/muaei3zwN3bsxuvEkrt48hsOhm3bidum2DE/y6XxKnY86MAVUVGNF4SQ+vrkYior0dq/TmbWxt49Nj6/ZOWabT9e21WqK+d4I/n7nEo/jCGQ5lD1dV46cIPA+DwkRmkzakCSgzXjJAMauzyDKIFZMC0s/k1NoU15yBi20JS4ij9Vg1bpRlkQ8YO90xzLvEwGAybXProeaUQ5+2sh4poamYJs3y4jo54qNZkUi9pgEIlEj36srpafWEx+v/CCdG4mI4sRCzfAPDEhI2d/v7gB6lXWFUXc/wKBvIbldJb2+SzEnbR+75JXSpm0y7gtWMu7U/jMDw1Fl8x8b+aFomnLC5zo+yqHPDCUAb+fYKKS4+drmJ2Fa1fSVJrTxNNm2tvNBIPiNFFdLcdJp4WT+o4HmQxXELNWcflqhsmAGUf9g7xa/RtyBxuwer0QixI720PPuIePfxnlasSG22vBkrFC1WH8xioea7xOPJ7jz/5FEvPEHPGkiz96ZIDvPX6NtKUSNTmKkq2AeZ9k4ItUojrMqHVmTLE50gM3+LYtzM52KweeCpC+Xsss+/FknICBsOom4LPSPeMm82gDJwbHoLEC2+gAKUOEXYmr0inptPyORlETPZhCLfgz1ZiScdSapBDrtm1oN6xkKlSaDKP42zvRfv8llMd3yepsS0s+hNRqhX/1r+QG69Ob2lqRMHTPk8pKCVWvqZHFg9ZWuXEmk0xLkkmRTPQKyyMjMggU39h/8A8WS5MBQuSrkeOdtn7vgldKmbTLuC1Y67ujP+///t87GBjIV5sqTg8BIkPMzso7v1IiqsWDLowG2oiNzEdzbHHE2OOZo9ETZXzWQWqulspKO/F4fpFSXxi9MqXw1nk7ybSBcKwCqx0yfRZGJ5/lkXYDLWdhdiJOrTPGpYoGzMk49b4UlYoFc85MtTvHpzccxBMtJCbM5EiQjViZnnfx08udbHlkFkdFGrY9ACf+M8HPezl242EyJgummTRdU29xdN5G/+xO2uvmaE2mYfe3CQ5O4jZHuRhuItnUTtOuGGinyNgbaDKOcyYMhlSU9swNBjMKWs6HYvDnJQ6jESXqp8twjAHDVrBk5ILT03D2LKrBiclhw+/cL94sbR7wHZBcuDduCJEePAj/y/8ii5J6/uuHH4ZXXhH/7EuXJNCmpUW26ewUwV+/L0ePSk7bVAo++ohsZaUsBtTVyYLF2Njy8PXXX1+7Nfsl88kuhTJpl3FbcLN3pzB/T3//Uu+R/v6lVrruo71QVhCbDX73d5fnBlpqncuH6rtvY+rdyuREBmOmhaYtJnYkR9l30Iy72k4kIrNvPdfQ8eNw6ZKHrCfFgdYYZ3rsjFnbIJWiT6vgi79IUF8PlSYX+Osxp6PkjEbmM2bsBrC7K5jduoeWuTiPOf2MzEX42wstzM27SGXN+Csq+R+/b+bJzilc9iBdXUfQWkfJTDtoyvnxJ1QGEo10z3WQyRkYitdx1DJM+m96MfdPsy9zghvhFFa3hv1aGp+pD9O0E/+4BY/dBmo1/uvtGNJhVGYhbsgHoeh68EyO7vBeMjkj3bH9dHEMJZFAaVLpesmM9sTXUQ9US79qab0enOje/+bfiHVsMAjx1tXB974nlvSWLeKL/cUX4l2ilxfbti0/uupW/8JigjmVggcflEFlbEysfpA2RyJyQ7dvX54icjVS/hL5ZJdCmbTLuG1Y6d0pzI999qxwQqH3SHv7Ur0a8oW2rVZZ1zp/Xow83YW3pIauaSi2OEefSvPmT2wcrhuBWQcvPXKD9K4H6IkKn7z7rgwKn3wix6qpAYPFwkzGQ2s72IwJRkdyzM5BRcpEKgW/93tW0rvqeWJXADxufvKLClIpUQmeftrOefsctkEDVU4bL+RucC6yjemwEYvXQTIQwmC3EYklOD2iUt1xAO1cjFjchqsyxmzcxFhKpdMyxFiigTcnnsGe83B26mmaUoNMpRW+GT9JKLmddDRE19j30ebdUhihoQPt0R1YRi6ihNzQa5NOczolijKTQWvcTua8h6bMdfyzHrScF2XuOoTDKFM9KO5JwAko0sGBgBC23S5RS1NT0sl69i6vV25Yf7906PXrQtgjI4t5sxdHVz1kdSF4Jmcyicatp3Q9fBh+8AMZGMxmCYkt9fDcx1n6NooyaZdxx6Hr3U6nGF41NbBr11LvkWIvFL3QdjIpsmkyKVr08eP5bZdp6AtT7PTYFKrajPeh3fT2mxjasxM3bhIJcfvTi3f3X4xy6vMsdk8Fzc1Wnn8eZkajRM9c4/zoVlKpHFmHCbtdju/z2Wk9YicQyA8qQ0MLNWa/V4H25jjR+ATvRFp5ZL+FsxdMzF+fY1qz8MtfGXDV+IjtaKJn+u/R8tVhApev87LpP3Nm5GF653fSl9vFFtMIvu3NOKsspDUXNYkIN9I2ejLbaEjEUI0zKNkAij0C9hpQQKmL0Juup+9GDJVpSJvQZqpQzw6jzGuYK6rQGp8hFm/GlelDzUQhlRWC/fxzuYBt2yTHtZ5uUU/QYjZLxyeT+crJ//P/LJ4d+sLA7/2eLBbOzEiHzM+LFg3y/SOPiLdMLIYxkRALOxCQQWFkREbxmhrZRrfMix+eO+xmdy+hTNpl3HFEo/Ku2+3CAXNzy71Hiq30114TCeP992UmfeOGcIoukZTU0BfYXx2YYSZVx98cd5BOw8l+8URzOuGJJ2RAGOuPEjjZT2PaQEUoh2Pfdtrb7fRPzqF6I+zcOk98NkrQ5GD/ITt2u7T/+HGZGeSSCZJTCcJzlczOWlEOeeHFR/ngjShUVxIbmuZ3HpjjnTErD7QlGdNc7NjpomaLgy8uwiXDDsx11fw0bKd5ywAv3vgNvdl2jrjP0W9oJDJfh3l7K9HQJFvTGZ5wD3Jgdwrl+LBICfPzQnCVlQRR+JtLLdhm3CRiO8FgwEYSk9/L0epzHJ85hN2gEctU8HLlxyhWI0Rd+WRResGCN97I57i22eR3JiM/uZx0QDotUxQ9ujIclmlLVZVMOyorxb8yEJAfvVbkiy/C6CiRRAJbNrs0+ikWk5tcGHBTqKfdBTe7ewll0i7jjmJoCP7sz/K59//b/1bkz+LCvsWypZ7qorlZuGVkRIy7+void78iqTOIwgAKWlQkWD3187VrsqbW3Cxy7Jvfj2Mky8+vbcdTEcMWgOd/B0weJ5G0FVMyRpYKbG4DbrfM5HUfcTdhdqevcbK/GltFjvOf1XGgNYX2zmdEripMnA0TqKgncXYOW9KF2x4jmXNjr/IxNSW8aDCA1+liR7Wd2Fw9ofqd1G9p4kDuFAeaz6A5t/KM8QY/PttOdjbCecshDniuiEWsZ73z+WD7drSLCZido8k1S3e8kVw2S7v1Cv6kk/6AQsbnosMRxO/ZTbpfAW9aXPUsFumkhRzcGAxyI0ZG5O9cTkhz504Z6WIxsb6tVpFEEgkh8qoqSeJSWSn/G41SheLKFTnG7t1S1gzI/PVfyzkCgXzNtuLiB8WSSCm3od8ilEm7jDuK/n5533bvzssd27dLYd+Vios89JCQrZ44LhQSGeXIEeGpQlfBYuJ/4w0ZKIaH8zn6KypEovX7Ja+/1wtqcyW5wRTeihi7G0KEFQ+Dg/DQUw4GfXsYsKRIjxipzk1QBcRiVVLSMBDD57jAo3VjTMY9PGC4RPp6G9pZM2ZS9AWqGIo7qfGasM3HafIZ8TQ66UhP89TjLj4OVnHwoMw24rM5Ykojr7ycJn21CuamYLICZWcdyuR1TgfbGE7U4UmmuXHZzJHEGIeCfiGvSATGxwmOxQm79xN1PYg/acVt0iCXxm9qIVFZhckXI5Gy4J9XMeXSqLtqYPd+6YgHHpCO2b1bjvkP/6F0ot+fr5Le0iJW944d8lkqJeQcj4s/dyYjDvBVVYvEzOysZOmzWuWGX7smQTqNjZjCYbn4/fvlnEeOLK0PCcslET13yW8pyqRdxh2D7qabyeSjIU0m+TwcNnLiRL64CMgs/epV+Ou/lnfUbhfL3OFYHgYfiy0tKwjidXL6tHDFxIRY1smkyBkTE5KL6Pp1acP+/Q4Me/ZgvQ5hxcN0yEpvr8gxTU0upmYi1I2dYiZmgMhFXvznj/LOJ1XYe8/z1tkYsf4ogbkxfhH30NHfSzR6nQ9m9jE252QkZcY/bqbJUcGe2gh7o5+zzTbKB39p4FNzC9dnXFS5kqRHp2A+wE+OZclWP4bPOke3P0fXx6dQzBFwtmAYHQVmMVApF6Yoi4J6MAjHZp4hk/Yyv2Mrhx9N03b1b2BigoFRGydidfQ0H4VYhsPfqaWtIYHy+UI2v3RaFv9cLukU/QboJXyqq+U8DofoVK+8ktebR0fhN7+R/0Mh0b7+2T8T7TqRkBs6PCw/waCMkleuwHPPYUgmZbqklxjz+ZY/OIXaVyKRT/dYXogso4y1Y73pHQpnuPv3iywxMiLvd38/eL2WRWNK0yQ7XzYr77ae916XN/bvl2PqFacGB8X4evNNcenV2zM7m5dQW1okknrbNnEnvnEj79yg5xppabGz92HhhpkZ4azBQfndPxcllTHgUgy8WHcKx3QDaoWFJvcEZ66bMWRNfNX6KW/zHNnEJO98XkPUmaVxq5Wcy0rviIOKZisfTDQTcZlQTSEy8ylU1zUqtjxAcDJFNpdjOunm56dVGquS7KyNs83bjFY9h1Lpp63Tx6HROcIVtXQMdNOW7RO5IR6H/n60RDMZpmnKjaG5HsJzaBvK2DugRfBUu7FdnKapZwi/rQ1PQxvK5Dmxkk+dEs+S0VGJSEylhBD1GP9MRvTplhb4oz+SoJjCqVBtrZB5PC7WuMslo6jBIJ4i2Wxe466oEE+Rq1ehuxuz0ShEHY9LWwqL+hY+ZLr2FQ5Le2/XQuR9UPS3TNplrBvFVddXyz+vo3CGC2JcRaP5d89gyBtTILPnS5fyhcHn5vKR0Tr09BSBgLz3uleHLoPq+UCmp8Xoe+op+b6iQlyKJyfFcWHbNjnWz3+el2CdTjmn2SztTFc6mEgoWOMRPqh8gJe2tWAKOvFPV+BJTIDDwFisBU8sxL6KS4ylG4jP2QhfvM58ZRPORBxGZgiEtzCTyDJjUqhR5umZqWZiJodPMTIzU8GEZiGSshA2VjI+l6QueBU1+LdgCKFEIry2rRbN04a6L4Ki1cAUQpLJJKrFhWnShr+iGXvfNdRj70AiTLBPI2z2kZidx2/2YIpOo549DvOTcnHZrAj0Y2MS/djZKa533/ymDArBoFjX+/cLOQcCMlo6nWI9d3fL51NT0rlGo3Tk1JRYxDabkLTdLlMc3R/bYiG7Y4fsk83Kd/ro++qryyOstm+XtnR3356FyPvEnXDTSLuzs/Nf9vT0/KPNOl4Z9y4KazLqaU/r61d/xou9O4p9sZubUzz7bF7TfvNN4YtMRt55l0vWqByO/DH19BRvvplPAV1Ym1IPwunpyRf/DoXE0WJ+XqzrhgbJOfLhh6IO+HxC4gcOyLmfeUZIPxp1Ea5pJTI6yuz+3QTsVRw8Ajjq8f3lmwRMNcw6nLh3NROa2okrGeOZunMMzniJNwU4020gkKvCl45jdVWgxSpxV6SZTwphqZHrTGVV0hkDdrsB1RqhxhHl8eZZSDfDxBx88QVKdTVKx4xYvJ6X5eLjcRgfR5kL0GX/G7SWx7DOvodyYppgRS3HIs+Tqa4D6ziHvf20TX2G8vMqsawVRQi4uVn+TyTkeC+8IKQ8OysW9v79MkXJZuUG6S5+AwOyr9stx9q6VTpQl09OnJDRMRSSyMlf/UpukNsNkQgZRRELfHp66eh79qyQu14odGAgn6+3eMV5LdbxWra5T9wJN4W0Ozs7nwW2bcaxyrj3oRNwoY9zcXRiMUp5dxQX9i1cSNRzBaVSeamiqiofaq7v19paOspZb2MolC8dODQkHFdZKUbbs88KOesFuvU82x6PFERIp4VbXnpJrP7+fg8aBrY63SKrZqMk3p+GuWexWTKYMvO81HqOdJuVaDjFO9efg/g0Fy5vpzHTT5UpyFNKmN5IAzjtjEZybOmwYroyRVBL02iexFdnIF7VSJPXRLPVQE90C/19cbqip1CqDGL+X7smBPoP/yG8/jrBs8NoEy5U/zmUgdMo85dJhCagpgYt7iGTytKkxvFPGvBoAyhRPwwEReB/4IF8SlW3W6QPv18Sl4+M5H2lf/Qj0a3NZhkd9+6VjhwbkxvU1ycWs9OZLwasj6Czs0LM2ax8lkzK348/TuSBB6g+elRulD76Go1imff2ynF37VquY+ulydZiHa/Vgr5LWfvWiw2Tdmdn5zZgcBPaUsZmIxjEMjwsocabaDHohKuHl4dCa3vGS2Xr0/8vLOwLQsalcgVB6fdvpWRUel1JnRN0a7quTo4B+ZxH27bJ53v35hc/L17MJ5+bngavN0k2K5zT5AzzcdJDHAuP5k4TyjpIb2lHzU3xH+Pf4rLNRaYqQf/lJJbqbWCA5x+bw3UjRdP2KL/40MbsaJRqc4gW02V8hgC2nJNYx9M8fvXf0aP5aDKO4Q9ZGUg34BnWUJ2z0OhC+2gc9eGL8MQTHHs/TOZCDabsU3QlLqNUVpKpqACrFXVuElP1k/izDZhavagzWUjZhKhTKSFHj0f+HhqSjtqyRbToL74Qfenq1Xywi17Md35eBo++PrkRRuOSoJnFkkO6hrZvnwwKzzyTt9irqki1tORvoD766rr1iy8Kce/cKdZ6KQt4LdbxWi3o+yRvyWZY2tt6enre7+zsLPnl1atX13WwRCKx7n3uBu71dhrDYVzvvUfF/Dxjn39O5PnnyW6ym5TDAY8+aiQYNKEoGSYmssvId61YqT+PHs0fP5HIcvKkhdFRB7W1aUZHzZw8GWXr1lTJY4bDRt57z0UuZ2BmxgiAwWCmp8fM+fPQ15fixz/O0tUVZu/e7OJ5RkYshMNVJBI5ensr+Nu/jVFZmcVmy7FnT5TR0QzDwwYmmefKhAPSlfjjT7CvYYz0xCUuxx3MZeawpONM9CVIxu0QD5GwOKmwTmEHApfSHIxP09kUoHbyFFttfnIOB/49X8VrG8Iz+J+4knmZ4UiaREUlx9VvYZycRov6qBzJ4iNE+l/00zpiYfpKmPZoD6OZekZTHowAXi/ali2wBR7bbiB66SzV82MYXG7mA7Kym6qvJ93SQtbrJVdZiaGyEmM8znxVFabxcWyzs5hCIXIWC5mZGQwGA7mKCrJOJ7Oqit3vp6KujpzFgml2lkQuRzaZJJFKkfrNb7COjpJxOnH+5jfMDwyQVlUSu3dTabGQrq3FPDlJZnJy2X03ptO4JicxTEyQMxiIOp04LlxY/D8SCJBd8FwxhsNLti38bvF4a9hmGSYmllgS99L7viHS7uzsfLanp+f91bbZuXPnuo559erVde9zN3DPt7OvD2prGcnlaDAYFgMv7hZuJimutT/r6mTWnsmIE8Ijj6xsEC10AU1NYrDFYiK59vWJV9vwsJ7uuYbvfU+6SFVFRvX7xfA0GuGZZxz094t0MjY2wfbtdRw9Kl4vtTVRGhyz9Ex5eOFru9jZ/ADBaAWn/vkozKVR7SGcKpjmo3iMczxXmwRvEm0gjGrtQQkOQdWcBLa0VNH0gE0a8W6Mv2v7FVrOQNjs48PAYQbTLQwbtmIMZ/hu88eEYglaPvkU7YaDmbALpy1JY6sLTyJMdGYGz/XrYm23TIChH3ILBTi/+lXw+6nw+cQibmiQRUOjUW7S/LxMUZJJWaisrpZFC49HLF6nE9fcnDxPenmxbduozOXEKh8YEMv8kUdkf48HHn9crHjdK2ThBs7W1tJZ6r53di59YPSqNaUeoOJtS6GzMz/tuoVSZXf6fT9z5syK323U0p5Z0LO9wLbOzs6DPT093Rs8ZhmbgQV9zjw6Kux2F/W5zVqUXy2PdikUSpRGo/CG2y069v/6v4oiYDCILPv978t6mssl7Xvttbz8MzYm+n1nJ4yMmGhvl0XY9nb44AMHU3MO6uqg7QCguOH0AGSz2Ou81MWHeemhcdLxFGqjFWX/DjhzBqV2FuZiMmo4nTK6tLWJBhQOw/btknlvh4/gQ8/x839XSSCxjSpDmqmZCs6bDrAtPUJb4jJtbSa0XBi1YhbFGIZcDlM8Ln6NZrOQZmenEPPEBEFLDVrHPtQmG0rPSekUr1ekED36MZsVotYjLU0m2S4YlOOYzaIj/Q//g2xz8qQ4teslxlKpvAuhxZLX0Nra5Gdhxdl07ZqMjoGA3DSfL39zC42M1TL3rTWrX3e3PITd3fesZ8hasCHS1gm6s7PzDxHiLuNewYI+Fz15cnVz9A5grZLiatb40JCkYc5mhd8KK1OthGLtvacnHwW9fz+8/bZsd+qU6NmplPzWs4EeOiT8opc2a2iAkyctnDgh7oSxmFjwyWQ+bQaAhorNOka724/f0Ej6awfYfsApLmx65eGeHrkYk4ngg8+iWRtRX/smitct2z35pBDZyy+jnDnDq63d/G/nDFxhF1ZPBqPHyVHPZZSMBj39KAB1W2DODJkMpmhUzqO7zMzMwMwMQZPKsfMdZIxmTN1ZulodKHpeWocjT7gWi+yrV4nYtUsWHi9cyMsGDz8sn4M41Hu90nmxmFjagYA4uuu13PSKM5omFvg77+CamoK//Mt83UmjUY6rj56b9dzeJ54ha8GmeI/09PT8APjBZhzrnsTddLjfyLkVhdTWrXf94VzLonxhRGSxNR4MCmH/8pfiXFBTI77hhw7d/NyKIpxhsy2Ngj5wQBwmKitlsdHtls/r6sSI1Im6rU23qIVnUykDHR1C1oFA3m88nc7zgNrmwXToIP7wHKYOJ+YDDvrSoB59BYaG0M7/GtWkolRlOdf4Dd4Y+Qa+agO1P87R9ewwih6BpJPL5CReZ5pKG1hyJmp21+DzmUm7O6B1m1i+6bRY1rOzYDZjiMelI+fmxHI+cgT++q/RJtNkgrM0PbsD/6QJrWEeRbeC9+4Vb5LxcclxradF1ZOt6MULKirkdzyev5l68I3NJt/t2ycdrLvseTz51WCQkdJsxhaNShvr6uS4uv+3Hhq7Wc/ufeIZshaUg2tuhrvpcH+fOPvfDCsmcyoYj4JB0+L7que+L3QQyGaFsHM5sWx1BIM3lypVVRwYuruFkE+ckONpWr6wSmenGKOqKkXJh4YWchtti0IsRl+/m9mklXRatOxkUnhocDDv9qjzgKJA12sONM2xJI9KIuGFER+2oVZM0Wp2xAb4nwa+QSDpQElN84zrNNrkNMq2zKIEERyNoX08Sbh/Gp8lyJZEP+ODJkaGJjBv6YGTX8iz0dIienNVFYyPk7VapTF6BsD6ejh8GPX0AKYx8J+dwrSzE/WPvgOOeemYv/or8Y9OJGSfqqp8+CmIj6Q+woXDYhHrVvOWLTJN0fNiBwKSJKq3V7bXne8vX85X0jGZsNy4IW2Mx2VUtduFxAud7m/nQ3gfokzaN8PdnFZ9iaZ0pZI5FY5HPp+Bs2cXCuuaxcgrTCBVWysWdjIpak9bm3z/f/1fIqdaraJZryab5HLCLzabkCyIcfnqq0LS//pfC8fcuCGfe2wJJj8ZgByouSnUtjai6RSjo0L0Dgf8V/+VcFLxgKFfb9/pMJmxBE2dDrrPG8ld8VMXGeV8oIEPTC8xY6khOxclOGsmkDagfvI2WA9ATQ1BLcOxgU4yFS0kcv0Y7TbqsgFGDGZ8thjHLd+gyzeDopqE9CKRxYs06v6MNpuQaSQCo6Mo8TG6HrWieW2of8eHsn+LbNfXJ6SvJ3UaHRXh/3/6n/IJXVpbZdoRjcrfly5JGOmJE/K/1wt/8if5v+12IXfI58v2+aSDYzFoaiJdWwu/8zti3T/0kEyBblcGvy9JRZsyad8Md3Na9SWa0hWjeDwKBs0cOJAPHw8ElkYxv/SSzPAhT5CnTwthh0KiAExOruyma7MJ3+heJH6/GHN6Du/+fjlPa6sQ+PXr0OJNstsl3g9XrprITSWx+HJompxvYEBciA8cWIELgkHUE1LuzN9nwl2zhViFgXf4JpFsgrSvHke2kuhMEBchnnKeIpR2ogWqUI88gvb5AJnhGzQ5QvitcHhPkOB0Brt9jIbIND3zexiw7uLQXquQbFOTkOIvf0m2r0+s2URCpgN/8RdCxhMTKDYbys4dcGBrfmSMRqUDpqelQ+12mXroCV1AMvXpSZvSaXG/GR3NV2CvqJCObG3NT4/q61nMP2s05t1/HnkEgPj+/bgrKmRg0Suxl7EqyqR9M9zNadWXaEpXjOLxqKVlnuFhIWmXS7YpJPV0ermGPTubz/2fSgknlBrXCs/lcokVX2zMtbeLsfrhh3nHiO88BwcGe2BujrMmF6GHn2MsM8+nv0kwO5qkp9/ORx9Z0LQVlKuFcmddLyYYOB+BXT5m0YgkK3BWpzmXa6WzIUNifhL1+nn651s4Nv8SBwJZXP/JyFGXhqlxL37lAUzVYdr+rgO2beOL/9vOX517DKslh+fQHtoeH0H57G9FeohEoKWF7OSkEPjkpBDi9LQQ6/bt8rk+AuppEk+ckE5Kp2WE0/MU6AldQKxmPaKxt1es42Awr6dXVUlHgnTi2bMyQo6OymdGY959b0Gzj6pqPi3rl+j5vp0ok/ZacDenVV+SKV0xloexp5e44kLpvECFksn58+CyJsGUYtd+I9/7nn3Frjp4UH6vpHt7vcJt4bC4/rW1QfMuD8reJwn+8Cf0z3WQ+biPMafKttglPumvxRSfZy7gIjKZQTs9jXLIQxBlsX3pcA1qohL6++m+spWMo5oZxxYGjRmMJiupjJWD28ZpYpiemkaM/VHSeHHO95Pxp0nvd9Ll/DkDNUdgqx0OfAW8XqI2yFaCpw6yigstM4dSVSURhOfPw+wsGadTLOLqanF7mZ6WCEY9oZOeiDyTkdHqzJl8WbGvfCVfJqxwJLRY5IboerfNJoRfVSUd+Pf+Xl5KSadlCqIXVHA4ZGqk5zhZSBaTam6+q/ED9yPKpF3GbcVqzi/FYezF45MewNLevryAiaaBnSi/W/MZ55MKD+U0vDxMsedpsXaue50Vb3P6tHDP7t0izcRiC1ylpdGsjWSq62jCTyBiYFdThDmTwuhogplRI7HwAKr9IsFLJo7RRSTr4OxZOHDAgyvxAgdH3yZjttI0cYZJ1xNs2+empgY+PR7hyofTXOs14DAGscWMmM0J5npHcRljqNtnYPt2umMPkvFtofu4g/Z2aZvdLu2OBWKopqAQdCgkJOt2E/7Wt3BEoyJDVFbKxqmUrJ7u3ZufaphMomfr2nMsJm59DoecZHpatKKZGemYtjYhd6dTBgj9hukBOsHggguNmp8y6Rn/DAaxqmtqhNxbWvI5ue+1lKj3WnsKUCbt+xl34MHayCk24vwSDOY17aGh5cV7YzGIjcQZm6nEn6yheiqB9maSrteXnuNma7mFleH1AJq6OlmclO1UVPc8iYuznElWM9/oYps6zk/PpcnkzKTTBl7cM4TSUUNfd4RMbg5nvdSidDohMg03so0klHr8gQAezxyeajuz12fgci9aLsRcxMW2XdU8V3mClxPvk3ZZUR1xlKCFPl8nmeYtNHU46O2VOruffir8W+tL8mL436P8x1PSwc8/L+R65Qr2cBgefVTIcnJSdO3aWtkR8tZze7vIKtmsLGJaLLIYWVMjHT89LTJHNiukvX+/CPmtrTKdeOghWVj46lfF8taz8ZnN+enNyy/LvnqiGqNRBoLJSVyTk9Ku4jSsd5MoSz249xDKpH2/4g64A270FBtxfim17zJt+jUL/W8EAeioj+C3O5ecQyriiBG60lqufp5Cb5LCAsMoiqyCTs+TiDkJROe4vvNrNE2mqdliI5cDBznw+1HdlZhwEpkTzpqagotXXUwlanFbIxzZo9H2vQrwwsBf9ZJ0jjGY3Yov0Y9vPognMUVrbpDgjAmtYg+4q1FffgLTNQd+vwxUFRWSazyRgJbcOI4P3oX5y0LGurbz1FPMnz0r5Hr5shDp1JRcz4MP5jNv6a43mpZfSNTL3NfVCdEv5OrGYpHvJybEvU9PjxiJiDSSy+WryszNiR/24cNCyF1dQvJ6JGRBIQPDxIRMp1Z6UO6GxVvq4buHUCbt+xWb4Q54kxdio6fYiPNLqX2Xr8t68P6DrzD0ZhK/3YnJ5ViifR87BpHJKIHRBM9+vYIDT7iWtb94IFhC2Ho/pL1kfRAJw/Bwin/zpgenE25ERU4JP/oNgu5plLYqjoYkT8kzzwjnXbhgY9a3C/90kuee3Y3SKqGTh75Ri++X/5E3B+axNyRwtdShznsJ1n2LY29XkJm0YspW0XXiBF3fbUZLezGb8w4cALW2MCpavojA4KCQ6qlTMgpt2wY/+EG+xuPOneJrnU6L9XzixKKXR9DVIsUVnAk4dAgto6BGzCjtdbI6qyc37+gQ7by1NX8z9JXdcFi2vXZNdGuXa2llCh16WLzfT85gWJ5cvfgm3mkLvNTDd6uZ0G4DyqRdjDs1sm/0PBt1B1zDC3HLp1i4NkVV6epSbukyV3KcWZbetdVL1+vLt9M0IezBXw0RiFTwyViIA3s7QPEuO8/Ro9B/NkK7V0PBSzC4tM1ms3ipjY+D1ZqlogIe3R/lRm+Skb5KPrS56Xa5OepbKum0t4uBitVGzmVjFpGPVRWU1lZa/9U/4PWzw2jedtSqLMqf5Th9Mcp4ooYOV4hQNIl2Pcv29DTKdmn3a68tOH7MhmmLhFEmPXAjLLq1Hpk0N0c8ZqXvVzdQoxUoeiXjcBh+/Wu5oMlJuaBolGCykmPN/w2ZPftI1EpUoy01h+nIK3TtvCSLnAaDkHxdnaRs3bKltPP9z38u53K55Le+OKAnM9erVSwkkIkEAksHgOKbeDfiFEo9fGXSvkdxp0b2zTjPRt0B1/BC3NIpiq5N6epC2X5rfbhWx5lS26kqxAJxApEKfHUW7OYUWn8IpdW7rLnH34qSOX2OoVyGo7s/47j922RsjsU8JcePi3E4OAgNDWlioQTRLy4zOKRiNgcxmRvZttO2bJbvdoslPjEhRu/583kvOCmD2Iri9aIsdPDQH/4LfvrPJui1x+mJJDhsu4jqiAvJLrC9oii0+UJoP38P7FGJKHrmGYLTabTLk6jT3dC4lZ8NH8b6H6ZJTH6VI65ztGV6UeJxIdXGRsmCZbEQbDvEaW0rkb2P0/HCdnEiSUP7gyLJaC0HUbQ+2d7nk/D0lSpeKEpeetGh/1+czHxhhdkUDOYXL1ebBt3pOIV72GurTNqFuFMj+2adZyMP1hpfiHWf4h6J4lQUePU1C28OhrCbU7jsGdR2b+nmhudockfw00T/ZIRM1RxN7UJahVWv7HbweGI8s22O/rcmweliwp8mMJ6kbottcZavB/CYY2Hs0SSqw0kya2dubmkZxNdfDaEclwEumKjkzRvfYpBteJtuoFrSHNlhRfmDZ5Ys0gUfeo5jb0TJDLhI2Jo54ozju3GO4+GHyCScmFKNHIyNkJvP4K228M7Mo4QjLhrMO+kyjaKMfyRW9twcQWUbx+KPEmnexdmZZuiVNUQAf28UU2wO1WctXfEiGoWf/lT07MLoolKlhPr68tGQ4+MybQH40Y9w1tfL/6UMly9xnMJGUCbtQtypkX0zpI2NPsi364W4h6I4W/d7eP37HWj9IdR27zIrG8SI1eZdxDQFl2WG9t0RhuxO/L1REoEoV0cd9A466OsTvXvPngStnTV4zwcY0lzYq63UdVgWC4gfPbqgAsQ13vknF7FXwsEtSXrrjjAScOTLIBJF+/AySiQCHR1on0xhv3IKX6KBQNbGlv0dtP2jg5Cezg+Cvb1ob7xLZmAr3iuf8W7yKLNmB0brE9ibq+lIXsTf+DAYcmSooXfUgcFmo9M+RWgmi/b5gCSjcjrBYkHbeohMro6Or22FnG1xEZZQCO2HP0PNTKK8ZRVNRk95qEdP/vN/LiOUwQDf/jb88R+vrGEVlgayWKQDamrg8mWyDsfqyaHuYYv3bqFM2oW4UyP7Rs5zK9JKKZK/Xdr9Ta7tTjsDKK15si4+t+5WaPfZibGPl5+dpPXAPrpCKQZ++CtuaA6Ggh5e/NoD9I45OXIEHI4sKArKa9+m68gMGipqWz6oJ50G1R6l6dr79AZSxCq89OIkZonz8ssOPvwQ7PEArlOfocavwo3zAKjGLC5XM9sajdRNz/LqKzZpdzCXz3aVTKL6GjBNO+mp2E3OodDRWslYX5TYRBi/vRpTlYc2dwbbzjCJGQPuyXlCMzsxuWdRXR6YaZHnZn4e1TaHiQr8US+u2oJF2IF+lOvvi77jn82nVNQJ9Phx0ccrK8VrZaX8AYs3oeCZ0DNoRSJgNmPUw+fv9OB+D/th3wxl0i7GnRrZb/U865QfjOEwy3Kewu3V7kstUGkaQXM1x45775mEiYXufn6/nXRzKygQOjvEzy82gdtNz3Uj9Meob3PS1lawHqUoKIcUipuvqmCKzdE7W0usIsKTztN8GP8qdl8l167Bqy+GSb/xH1HN3Shhm+gue/eitLdz9K8+o38ySXtbhNYDHUsPnMuBw4FiT9G14yIDgQlOVNQScjbi+kYtL9eeJ924G3XoU5QrFzCEw3ifeYa2155A+/EV1Eu/QfH35lMafuUrKJcv01U5isavUI8+j1KwSBtM2tGmPKhzQZTZ2aVtqa6WztOTsLjd6ysQWuB1Mnft2vJ877ebUO/z7Jll0r4duJ0P3TrlB5Ne2qnY5/RO6c4FL4imKWTsL9LU4bhjcrd+K8LhEn7f5hAmLYk/lncXDAbhzfdruTyRxjeboLMuzt6HGzj01M2dCPRzPfSkjXdOObDXwIexr2J/eC8d++Wa09NBtjcnpLiunsD70CGCKBy315CpmmPI7qQLhwwIAwNiydbUCHEfOYLi8XDoVTNtgSwDozb4/CRe4yzKwC9EO/7mN5n/+GPZdv8WFM9X4dg0/MIvXh96Z1itKO0+lFBQZJhgLj+4pl4iMziMybKVrs+uobS25qMoHQ7xXY/FxEPk6aeX+jKXevaL34mF71KJxPLt3nhDEsu43WurdrFe6KO11ysRVQMDa0vOfo+gTNqbjds9iq9TWskoiiz0FJP8HdKdgwMzaONO1I4q1Ngoptgcfr/jjsyIC29FYqE84uIlm0Mox39Ml92EFnOgviyWZl+fyCW+fc2ywLjHwqGnVs5pUupcmubBfng3HTVheqc8xHDkz9vuhSFXvvT7Quil1gcZm2NxAVTTQCEoOal//Wvxw25tFZ9oPb+HF7p/PklmUKU73E5X3QmUWBBCIdLV1fn8tcePS6RkJCL7zc/LCuvUlOTH7ugQnfrdd2FyEi1UTab+WZrsYfyV29EiF1DefDNvMBw9KkEzeseePy/kp3eyzbb02V/POzEwIH7mHo8cs1S1i81wl00k5HpzOZmJ3kLdyLuFMmlvNu6E98Q6pJWsx7Oc5IPBm2dQ2gQEg3DsRB2Z3mZMPRm6DpvoetWKll7b+7bRd7P4Vhw+LFygqqBossCndNSj+P0EAwH60l6iUvMAT40Ni9PGi68snHuhMUY9R8cq54rFIBZz4M86cNXCy0tqWnrz9yMaFdc3QFVbl4+jmiaa8WIIZEGuDv2cdidNzhD+4Vk0jwPley/k/Z91z41IRCx1vSyPquYTrbz9tgwIP/6x5McOhVBnTZjUZvzeWkyGKOZklD4aUOs8KKOXZIagJ4YxmYRcm5pEd8/lxEG98Nlf7zthMMjvVEqyBxY+o5vlLnvkiFjzHR0ruzDeoyiT9mbjHvKeWEQhya8lg9J6sQK7atqC9fjNA/h7omhHDrC91bNMB17pkBt9N4tvxdLxKf9lMFHJsRN1RLJigLa0CH8dPrwQR+LJu+a5JidFEy5qzFrSvy5CUYQo/uzPFqs+KH/yJ3R1tRZ1oyqjjF6qp7Z2yfOkqmAygj/kxGSMiU+31wuKQlYndz1Fqp6/dn5ekjv19OQrxOzbJ9by3Bzkcij2JF3qB2gHv4a5s43jl36fzNkLmH49RFfLKIplQYax2WQwicVEwrFY5PPiZ1+3bM+cyY+aK6GtTSxr3TdyaIgluW83yyhqa5Nc37oL473wnq4RZdLebNzrvqWbPRNYhV0XiSzkxNTgRF3H+DAwkPePLmUIreYQYzbnCbPwVkBBROLCfQoOzHD6Rg2RIQdOZz6dtN0u0u34OAycneXQQp8ZJiY2JxCpv19Otn27NKq/H+Voa34//WJKVX8oPOeRCbTZSdSOKpSQYXnejv5+6USHQ6SW6em8tb17t1jfoZAMCI8/LtLE+DhK7+co6Wn6Qi+Q8b1C02Mt+BNxtMe+hRLpFl1cryoRjUqSKbtd2qt3Pkj6xNlZIXbdgi64vmUzF0URHfv0ack2KKvE+evaLKPoXn9PV0GZtG8HSskX94qL0WbPBFZh11t9L4JBkRl7e1n0jy5s5kpJ2PRsfZIWNV/Qe/v2lfZRONatLO7T2Sl8ZjCIYvDpp8JFbrOPNnslip4rYzMCkdrb85GOZnO+eECpC1wI+S55zrYqlPo5CIWX3M9FryE9fWFTk3RIZaVY1XV1sihZTLIffCBRkwsLlWpmUtYh7A2YvOOoub586Xk9i5XPlydXfSDSFxR10rbZZPoRCskz090NmQye69fl+gsHJEWRm14qH8lmku196gNeJu07gXvBxajQDN0sPftm7MqtvReaJu/4iy/KoY8cWXqMlZKw6XEjelrUwpiN1fYprheZTovR+dFH+ljkQDv8OyieqbxWvFG0tko9RT1huL64WHyBvb35nB35+Pf8tiuQ2KLXkH5xra1iTU9OCnF+61tCpv39S7NkPfWUpE09dQqSSZQOK10vWdHSDtSX96Kk6/P3WH+e3norXzU5HBay7e8XicPtlpFveloGj4aGfMd7vdjefVcGkvr6pe/FauR8n5LtZqFM2ncCmylJ3IrFXpg0utAM3aiefTN2vcU265OBUEje5eJmrjRZMJkWYzaWFfRebZ9SGf5UVXhnUfJs84DiyWvFt3htS9DaupSAS12gXpFheFhkCL1m401IbInXkMslZAxi5dbWCmGfPSufXbok904fxBezUgFtbSiKd2EdwstikYlgMH/8QED08qEhmaLoicn1kvb64uQTT8izB0LyPT0yrVlpMfC3nJxXQpm07wQ2S5K4mcWuWzgrWW4rmaEbva6V2LWwzZFIaUuxeNs1ZAdcyQgrzhRanB1wtX1KGXM3nYXfrhmUfvKBAXHVu3hRxPWdO4UgP/jgpkVwS3oNQT4c/fRp+b+hQSorhMPyt34Nq/ktB4OSi/vtt8VP22KRTjcahaDTafFWefhhCcS5dk0klP7+fJ6SheuLx+N4b9di4L0iSW4yyqR9J6Aoy2tn6VjPg7WaxT40BP/4H8sLk83CH/1R3qrRyXUlM3Qj13UzZtM0Oe/gYEGmpNeXbVscuXmz7ICljLCbGWbr3afY6UaCdIxLr20dM6iVbnXJzxVFPC2qqiS3x09+Ijmqe3rkXNev3zzw5GYXZzTKDMlgWHnFtxQ0TWQWq1Xkj4kJaY++IFD4jO3aJTOE4j5aGBhm02nqfb7NJ9Z7QZK8TSiT9p1AqdpZirL+B2sliz0YlOTzsZgEbfzmN/Czn0F/P8a9e2XRSdexn3lG9MXiweNWcTOmVFVp12KmpBJJ8VkhcnOjevsmWVmFt2ly0pX3+FvHDGqlW73s86MhlPS0HEs/fiol9296enGBkNnZtfVRqYQr+glBZBC3e32ub6oqEksyKa58bW3w+78vM6gFF8bF34HA6qWDbhfukWyTtwNl0r4TWOkBWu+DVcqy1V/CyUmxxKJRsaAWirdaRkbEdaowLNBmyw8eevvWkkzqVohQUUQSKUyAX+LFXTFy81awmVZWMIh2OkwmUk1Th4OJCUP+Nq3Dk2FNj0BvFO3N91DUYL7dhYmW3npLZI1EQnTgm/XRaglX9IY0N+dLh610vwt9KPVB+o//WGpQhkKyf7HkVTw4HD68fOE7GMT13nsyAGy2NXwvxktsEsqkfSew0gNkNstLEYutXa4otmz1l3D//vz3ExPihud258tENTVJcINe3snvX+J6tehaphdg3UgocjFK5Vguwooa7K1gs6yshWs2T1aineokFtuNwZBbepvWuFi20iOw5PPYHKq9SErYvj1/fH2BUM/LcTOU6odSDSm+hmAwnz+7uN6j7n6oqpJzZK3n9niW95OmYcjl8l4yp0+XrvdWCjczINYxoN5vKJP2ncBKFvLx43lXrpdfvrUHq/Al1F+qt96ShSUg4/WKNaS/OJB/YWG5a1k2K39/85tLNc6NEmEpcit88Vba5lawWVaWphGMmDg+sRe7eY5YIM5j34jeUhNX4pAln5utKMczBVUUzMsP0taWHzy7u1cfPFci6NXITB+cx8dlAdRgyNd7BHlG9AjHQo+TtZy7xDY5g0GuV/dkKZQPS6FwQCk2LEp1+peIrHWUSftOYSULWQ9KCARWiXu+yXELX0LdDW/BmjZkMsvDAgv/7u7OB0nY7fkQZ92n9mY+c7eKIsvduHfvxo5XiPVaWStZbaqKFnOQCYToqI/jb64kkwmtuRmlEtutzi1egg89h/bGu6g+K8rx48sJaS2DZzCIZXhY1jLWUmizEIXP5alToqfr9R4DAZFTvF5JtjQ7u9y/uvCibnYPFIXI88/nE1k1NAiBr5R1T39mxsZkuxdfvO/yhmwGyqR9J1H4FheSYCKxNsthJRS/hAXkminFFoV/FyenD4VkKlxsRW32dLOIfEzB4MaOV4y1WlmryT6Kgvrq85gKqr0rSmZNp7/VWhXH3rGRmdiDKeWly35a6kcWDSSl8ngsPlrmEMrxYzhGR/NlvLZvX1ObCQZlhpZIyHPw6KOyRlJZKd+/+KIkY+npEdntZsmW1nAPsh6PeK5cuiSuh3p+7lIWvP7MdHZKMFdvrwwaXyK9ei3YEGl3dnZ6gW0LP4d7enr+0WY06p7DZngilHqLdRIMh8Wq2YyV7iJyzRYmgC51HYUv1hoso0VpJ5/Eo/S13qy/iiz3TIntboeb7bJj3sRyLaz2bjbDtWumxTq0q+FW1CRNW8ja54vjH59HsyRQotHSfV2Qx2PJo6Ul6bKbSOupU9f6LBVWS08kJMT9+efFqi7ssC1blteL3AzX0SNH5D1YzfWwMC7g0KHV5ZkvMTZqab8C0NPT84POzs7DnZ2df9jT0/ODTWjXvYPN8kQo9Rbri0zBoMgUK2mZ60UhEeukvVI+i+KQ5JvhZv2x1v5abXBZx2EKt78ZwZc85gIRBHun0GIKqrl6WRZC/XjHjsHoqGPFOrSFWDImJaKo4QkIVq26k6qCyeXAX/cgppHPUe0xyQRYmEilSP5C09BQ8o9WzIkWc+CcvSFV19dCqMFgvlq6wyHkmUrJuYotdV1X17FW0rzZDWprE3lktYHgS7y4uB5siB2KCHobcHxjzbkHsVmeCKtpwopSUBHWLjLFZgcDFF6HvuioT7VBFiCLMy2VOn9xfwwMFCSpVm6enk9H8UtcRNrr6fa1jCP6hGbZMbcrBI++wrE3k2TsTkzHHSUvXW9PbW16TQbsIr8MhFFP/AzlVBy6Vx99Fvc5HUW1T6I43TCYXhrBqj9HBQO86i14tFxS0CFwzbO8jNdK0LR8tfThYXEZXS294npT+65lBF4rIX9JFxfXg03RtDs7O7cBMz09Pe8Xf3f16tV1HSuRSKx7n43CGA5jCgbJKIpobEXfuSYnMUxMkDMYiAQCZNPpW2qnce/e/HkmJpYQlWV4GEc6TbqyEvPoKNGTJ0lt3brha9PbWXgdxpkZcjYbKZ8P2+XLGBMJ5hsbqZyZIR6JwOzsiucvPI4hmSQ3PAw2GzmDgehjj+H81a+ovHwZTp0i8cADhBf6a9kx3nsPQy4nffr88ySs1iX9GQ4bmZx0MTFhwGDIEQhESKezJa9xeNjC6KiD2to0o6NmTp6MsnVravE4773nIpczkEwayOVyTEyw5JjDwxZG0w5qK2cYHZ1dsn9xe+bnc1RUjK3ankJYpodJhIYYqa1d832N5OaYHI6iolE3N0diYICs3S7JqgIBrGYzjp4eMh4P2R/9iMjzz7N3r0IwaEJRMkwksiTq6rha9IytBGM4jCsSwehwYGxoIGe3k712bcnzvng9w8M4RkdJL1xP/Phxsi5XyXdnpX0K+6Dke3STdq/2vt4u3A1eWgmbtRD5nZ6env9XqS927ty5rgNdvXp13ftsCHqmukyGFee9nZ3LLIBNb2ddneRomJoSHXGtVtJNsKSd+nVEo7LoEwiIB0AuJxayw4HX5RJLe7Xz68cp1uINBti6VXzGe3vhhReoK+UF0Ncn7on6fj4fV9NpdtbV5ft5p0JnbQitP4Ta7kVpbVjxGuvq5NZlMqIIFDa9+FRLqtcoN9+/+LJPnuznkUda1n5r1nrwBSw+jjVbMQWm6frvuqirt+VnZro7XjAIO3bIvfP5li02rvv5LHzGYWWLt/B6vF5xB7TZVn53btIH627nWt7X24A7zUtnzpxZ8bsNk3ZnZ+d3enp6/nTh74M9PT3dGz3mHcVa5uEbmZIV+yKXijArRGGi+NsBPceH3S4P/QMPiD7a0yNZ2Jqbby4MF3rA6C6DJpMkBzp5UqbtKyWQgpJSkbGnZ0nuEY4eRTl+HCWTgaE1ygqlvfZWqV5z8/2Lt9u6NbW+R2GdOqymQSYSpWniDP5AJdqH4yivPyf79fVJ/zQ0wN/8jfSzooiPfxHCYeOqa8Ul27mSh9FK17OeBfTNSgf8JQ5PXys26j3yLPAvOzs7//uFj+4/75HbGe5aqrLsStpxiQWmZQ/jzRZzVvu+lI8riLXd2yvmp56BbS3XUyrMeq3BQiWIbFnukf7+db2cK42rmymVBoMixdTVrZEnCu/HGt3uVFUiI/2BSkw+L6q9P3/t+rM6NiY10R57LJ9Vr+i0773nui3R4cBSL6LCQbvUu7PZ5e2+xOHpa8VGFyLfBzZ4F+4ybueKdKFVoIeQ19WVTo9aYoFpCTbqtaFpEsTgdkudwN5e+dvhEE+B9V5PsQeMbgUWVjDR27VS/oqC9i3LPdLeXrpyyS1gM9au9O5dq/fIetxfisfarletaG+Oo9r7UVyZpRGjespWtzvv11/UN9pAGMP0DE2dbvwh5+0zRtfy7ujPjNcrs7mVAmc285xfcmyWpn1/43atSBdaBfqCydycENjUlPyvk/PNPEiKvT+K8zSsMG00hsNCqNFovsCr0ZjP7rYe//DVrJziYKGREfk5f35Nniklc4/cQy/ner1H1jqNL8ntrV6RREpdu6Lkc2KX+j4YRD3xNpXXq/CHgpgOHURVHbcvt/TN3h3dQ+ndd2VWcOLExiWS33IPkjJpbxSrvQzFxAPy9zPPyEJgMTnrlmipF73QEi9VcaQUoRZmUdM0WWyqqZGBQ9euC6e3hTULS2nuqxFpoRX4i1/An/+5WPZWq0zj11J8oZSuqk/D1yXQbj707h0dNa/N/XmN0/gVuf1m117QV0seQU1DscX5+lensUXnUI80oFB193JLK0o+ydXNIijLWBPKpL0RrNX/tPhl01/C4jd1BeJdfCO7ulavOFJMqH19+SxqIyOy0g9C4sXWrK5Jl6qMuwJZLIOiyIxCzyXhdsvLOjV168UX7pFk9npXnTwZXZtjzxpnCqsG4azh2pcVBnqxhlaTCTVyg/rGRmiruvuLd21tsjB9uyrU3C4UOxHcIyiT9kZwqy/DSlZYKcu8OIpRUVauOFKYxnPhPDmDAc6dg1/9SqzsU6fgn/7T/HY6CeuadGFJskhkfekyQch5YED0aKMRnntOKq/4fLeWEOtuE04B1u09soZp/OItLxWEs4Zr15cqFgsD4eH1V18heu2zpe6Fd3Px7h6TutaE25nQbIMok/ZGcKsr2TeTGfT/dSItrsodCEj9PbNZFnc8nry0oR9vwUqIPvaYeBs0NsLu3bLN9PTK16KXJJuaktScU1Miw9ystJWOdFrI4tFHZd9vf3tjC0+3knP8PoOigOKZAlv85rmvi6CqJQoDpb2kt25dOjDfbdLcrNXgO3UNRQPmpic02wDKpL0RbORlWIteW6hjj4zIGzkxIW8oiOeHwSBv7VtvyYJfLCbufB98AOEw7lBIisDqpK5b7cVZjwqv5eWXRSK5cEHkDb8/r53fLABDVYVcMxkpfbYRF6/Nyjl+B3Ezr8sVH5VSBL2G50tRCgoDEcUVm0M1W5lIl9hwM/vuThfNvdMyWdH9KJXQ7G6hTNobxUZehps9iIUeJT6fSBtms+iD2axYtQcPijvh5KRYyYGAaNeBALhcuPRgl+Zm0bIjkf9/e98W3MZ5nv0CBAgQJEiCB1GkRMk8Qkfaokhbik0fxUZN7cTjWv6nrS58kUMnk97+nl60nWmukslMM9NOZupO3bijaRs7cawqbqRKkzhibMuWRMkxJXkpUiRFUiRFgAB4AEGCAHrx4NO3WC5OBCgtxfeZ4eC0++23S+D53n3eE6zzgYHUerXXiy7bRPD6z85K8dTrxYJRUaFbzjTvHWi0YYQ5Yr34JtW/My3nJLtuGXy/GhqIvnfcT54Tp6nKsUCusxG6u5638w/Cz3C/ZTLN/0Nb0OxBgkn7QUIIkkI/1vsier2yhrB4XV8PvZhIhhMuL8t7ZKsVYVbRKBX4/bDOr15FevnCAjpkp/PiNzUh51t4/UtLpXg6PAwC/epXQejacfJl1a1DIsV68k0qXsmIc3K4bq6VafSWVN/Or9fq9CD8DA8iqUb9/2DS3oTQS2dXx05bLInpyKKt0vvvI9a5r4/oySdxL6wuqXrlCgj44EF0ZHc4QOglJUT9/RRxOkHi0Siy6K5fX92VRjs/8WV9/fXEOf/619LK7usjOn0azk2dNOq8EMY6aLHryTeZhrGvC+doDhArKEi/Oq3lf6RulJDuZPK5aBhBlzcImLTvB3w+op/+VNYGjURgEQeDibHT6izCd9+FVfub3yAqZHaW6NFHE7teX71K9OMfg/AdDqJvf5toaQk1QH77W6LGRgpPToLAy8uJbt+Gzvziizi+xyPH0vuBay2/48eJ/uVfENtts6GjSVUViFw9L/X5lpVl7sTUQ5612PUmz2QlNtadczQHMF24kHp1WnNrnTQd1nMZP5Nz3MRkLcCkfT8wOIjQOZsNj5WVsHR37gTxRaOJkRFCNllaApmHQpAn1B24BTGOjyM1Phgk+td/Rfq3kEmam0HeHg8qw4muHyImW/yg2tsTZZrBQXkc9Y+yoQFWdTQKIrh2DdufPJm4rTjf0lI4P7u6cosgySPWizwzKbGx7pyjOsCqsgDa1cmDpsWekmaqmhta3dZMD9rbFL0O68m25YSavIFJO99IdksYi8miUYEALOe5OcRMFxcnbm+xoFrewABI2+8nev55ELGINBHEWlQE4l5YwLZffgmru6CAKBYj28gI5rO4iFC8hQVY6Oof1Pg4pBWLRRL7rVtYNFpbib71LWlJNzXBWg+F5B3DmTOoX9LYCEYU56t+NBDWgzyNxlG6ZQFU302fpZrevdJCkZUYFVha6Ngrqzv2rEI2tylc2GndwKSdTyS7JRROvakpopkZaMylpaiHvLICEg8EpKXq9eKxsRFOw8pKxD0L6zgUgmU9PY0Ijpoa7D8zAyfhzp2w3M1mMpnNINRgEES8bRsckp99JuuEfPoprOdwGFJNIIC7gtFRkP2JE0Tf+55ku2PHYMGPjUm9fMsWmaKudWLmWtltA8CQHKVenTTfTU/7/6PIgYO0vSRAY/Nl5FkpTk/a2dymsAa9bmDSzie05pa6Fdfrr+P14cMgycpKOAzPnIHMYDKBoA8fJvrkE3irh4fRVKChAQ5Enw+EKyoGvvginIqNjUhpj0aJCgtB4tPTeB6N4vgNDZAonnoKpL5zJ+YbCMDBKCz6QACEOz0Nom9piWdsqExHlwux3yMjWIjCYXknMTqKMV5+eW0ZkBsUhucozXezavYWFQTraIzK0KIs00Umm9sU1qDXBUza+YS20l1Pjyyf2d0NB14kAkdhVxf2+eADbOPxIBpjZATkXFaGeOySEkRqXL8OGWNxEfsVF4Nk6+qI2tpQRc1qxdhFRbCoX3iBlt55h2zNzSBSRSG6cQPx3i+/LAtNEcEyDwZxXJuN6JvflIuLXiaiOrrEYsHdwZkz0NVjMRB/Lg7IDQhDc5Tmu+n6/EM65iglTxA9JV2u8tX73O8EGkZGYNLOB7RFnYQFK8qe9vdDM56bk4kiwgK3WKAxT07iR1VbC1khGIQjMRzGWO3teD4+DpIOhWBRNzZC9hDk/dlnkF7Gx4mGhmixo4NKn3mG6Px5hAcuLxP993/DChYa9PHjMpmmthZEvW2bDOVLFiGgZqmVFcyvtBRa+NBQ7rWTGfmD+lYg/t10bd9CrrExopVpIipP3N4ghboYq8GknSv0vtwtLbKrx9WrIO99+xByRyQtV5cLCSojI4ju8HpB1nv2wFqurMQ4t26BjCMR7FNaCglFyBOxGEi/okJKIjMzRJ9/TrG2Nhxragpj2WyQZdQadFUV0de/TnTunJRtTp7Ej7umhui735Vx40SSxLWtx6xWzHV6GgtCprWTfT6yDg9T5i1hGGuCWGQz6ThjNM8q4x6YtDNBqtvEVF/u5mb8OCwWEOCOHSDFZ56R2zQ1Ee3eDZItLYUl/ld/hbhqUZva4cDnBQWQN/7wB4wXi4EkGxowdnExyH50FAtDQwNZZmaIfvQjEHkkgsVjYQH7ihraYtGprkYsuKJAtikuxrYVFYhk6e3FuM89B3lFHTbY3Y157twJq/vAARwz3Y89vugVj4+vX6NWvs1PRCYCvCE9qwwiJu30SFdQQi87TMRQDw2hwl51NSzp0VGQtrYrzaOPwgqPxaBvq489NwepwmoFYW/diup5RUXY1mwGoTocOM5XvgJ5paCAKBQik8i23LsXVvgzz+BPOAnVqfTz87C2BwdhLTudsP5/9StIKzYbxmpsXN3D8coVHHfPHtxdnD+P7V94IfX1jS96KzU1qRskrMf/bzMjlQAvFrnu7k3lTN4oYNJOhySWtFl0NdfLDrt0SSbTTE6C5JxOEHZrK0jt5EkZejcxIXXuyUlkHVZXw1K9dQuPIrJjbg6PgnBFc1eTCUQpwvCeeILo2DHy+3xU+ctfIkJlZQXHVGcvjozIGO3ZWWjeDgeOOTeHOe/YgVR6qxV6tShmLxarUAgLSn8/xqiokJX+Tp2C9Z3sRx+36Czj40TbtpHPUk2efDap4dv87MCLnOHBpJ0OSW4T73UPFw1LiVYn04j32tsRLXL2LLToDz4AEb73Hsi+tRUkHwzCEejxgDSffBJjVFcjeuTOHVi+hYWwjHfuJDp6FD+ysTH84FpasK/VSlRfT8u1tVhMfvpTEPBnn0kS9flAqqK9mMeDRUOkvX/96zhHQd5C796+XZ5zZyceL15ECOLVq1hU/H7o5tpwQS3it+oLFy6Qb9dhevds+dr5Qk8G4dv87MCLnOHBpJ0OQv9Tp3ZTPE34yy/1G5Y2NUGOuHBBSgxEMq3cZJIas5A2Dh3CdsJZOToKHdvpxDYffYTXi4s4hiCgbdugH+/aBdKansZnpaX4fHIS2nR9fWKaelkZyNXhgGX86acY227HMV0uPPb1YduXXsLiIxYp8aMWCIVA1E1NRI8/TvTOOzIRKBBYXb9bc43DjzxCnpXytfNFMgvR8AHUBgMvcoYHk3amEDHWvb1Ex44hTThZw1IRFbKygs/u3IHkMT8PcikpQUheOIzkGa8XqeJeL6zwsjKif/93EKzYPhgEIU5NgaDr64leew1Wc28vyPfIEejoZWVSUiGCJX3lCsaYnQVpNzXBajab8cOsq8N+N2/ieAUFRG+9hTnU18Oi378f2w4M4Ec9M4O7BWGBu91YLMrKcHcwNQXL227HoiYaKSQhzpz4IpWFaOgAaoOBFznDg0k7E+gRApEkqzt3ViegqJuZer3Qe202kHVHByxQnw9W9/CwLPK0tASNORDADyYWA4lOToJAd+3CQrB9u5Q6urtByv39OPbQEPY/cYLM3d0g7R07kEk5OgoyXV5G1Mozz2CfmhoQfXMzjvH732ORMZnkOYjWX93deP3ee9DcAwE4SD/6SGrtDge08+FhjHHpEhaM2tqkukdOfMEWYv7Ai5yhwaSdCXQIwawosB5TtcJqbpZZi319MkLEZkPI3DvvYDuTCYTW0wOrua8PMsfUFBaFQ4eQOHPggNST1VEbH30Egh0dxfYi4sThgPZeU4MxRfeX4mJYy1u2SMu3shLvNTdDAhkdhZVstWL/V16R3dqDQejtlZUgbJGYU1+PeQWD+COS/SZjsdV3JDpYM1+whcjYJGDSzgQ6hHDPEanXCkuE/F28CEJuaJBEVl4OQt22DRZ3ICA7w0xOguBCIexnNoPgFxeJ/uZvMI7PlyhPvPMO9qmowD7btsGK3rWLyOmE9r6yAoehzQYdPhjEYrO8DKlEhCFGIiD3ri6iP/9zxHSPjUEW8XjgDJ2fx3Fv3cK5bN2Kv5dekkWonE6Q/MoKHr1eLEgi6kSEReabYNlCZGwCMGlnCg0hpKxX7PHAci4rw+tAAARYUyNrhzQ1rW6UOzsLqWV5GYQnnII1NXJRcLmkHHLzJgh7Zgaf7doFkqyslM5PEWdbUwOLvaQEYX61tfea/1J1NfYpLUXThakpbE8EyeQf/xGyjaLgsaxMZlUePChLt4oiVGoi9vkwB3UBKaLVTkMGg5ERmLTXiFX1iolkreuqKhCgosjqfQ4H5INgELU+1E4yIpBbUREs5spKWKV2uyRDcQzRoXxiAha3ySRjxZ94QhJ2vFiVc2oKDkIx1+Zmop/8BPsLwvf7sZjcvAlrXxA2EcafmADZLy7KWuD19ZiTzZa4oKgtXRHRISQVcd6iK7zWR8BgMNKCSTsXqGs5aC3H11+XlfxEPelUcoDHA2njK1+B9BCJwBq2WJA2LvYRTtG6OiToLCyARGdmUGFPEP3oKNFzz5HF40ks3NTXhzFmZmDRm83Qow8cAAFv3QoJxOnEPG7fxvZmM45TXo59Z2dBvnv2JJc7RLblrVtYSERdbj2noYEapzIYRgaTdj6gF13S0iKde34/iDSV3lpVBaLcvRuWq8MBIr15E05JosS0+ZERKVEUFEC2uHsX0kVNDbTr//gPshHJsDxRiW/XLjgZl5cR3fHUU7JyYHU1rOK2NiwIFovs3n79Ou4GVlYQAlhaitBGIv0Y6aoqjCUiY4gQRdLRIS3/eESKORBYr/8Og/FQgUk7H9CzHIeG4DwUtT++//3E9HEthLPzyhUQ9unTIGyLBUR66RIkj2gUC8H16yDd8XFYxqJuR20tQv8WFohMJrL4/SDwEycgT5SVQe6YmEAUydgYZI/PPpO1JhYWiN5+G6npMzMyDFD0rbRaYenX1EhdXi9G2uXCMU+cwDkqCs5NODsrK9FtPhCgslAIMg47EhmMlMiZtN1u96tE5CeiRkVR3sx5RhsReuFmly6BAFtaQL4DA6lJmwjW7FtvSY34a19D8s2pUyDSgQHIEwMDINNnnyX6+GN85nTCUrbbITVYrUQlJWS6c0eOt7IC2SYcBmm7XCD9oiJZGrapCZr3mTOwkm02kHsshnMrKpKdb0pKMOdUMdINDZBELl0CYdfVIYt0dhYLQH8/UXU12W/f5vrbDEYGMOeyc5ywSVGUc/HXR/IxKcPB5wPx+nzJt3G5QNDqsqwWi7SWm5vTH2dgAMS6fTsIeHISFve1a5BI5uZArtu3y9rVW7dCIllZgbUrLOHxcSKLhUyxGCzyixdB7kSwaOvqQPAlJSgG1dcHS/78eZkIs7AAoi0pQb2T2locY2oKGZ5vvYW7Cb8fi1ZnJ85zcDDxWrlcIGOnE9a2iNleWMBCsbQEh2em15rB2MTI1dLuJKKfxZ/fIqJ2IjqX45jGgtrJGArJVOx0aGiAJCISVoSVnSo+ubkZUsUHH0DyCIdBdCKGe/t2yCNlZbDCW1vhcIxGQdItLcimnJq61419/oknqGhxEST89ts47rVrsJ4rKzGvyUmMOT9P9MtfgkjNZoxRUQGr2uHAPsEg6nlPTcmQxoEBkHJPDwhfr92YuoZLTw9CGycmcH7T0xRqasJ8uMIcg5ESuZJ2ueZ1pXaDGzduZDVgKBTKep/1hHV4mIrHxynidFLJhx/S8sAArVRXU/i550g7S3MgQAU+H2K4ie49j8ZLl5oDAXKePk2mWIxiJhPNHT2K0MH4vtbRUSorKiK7xUJUXk5hh4MKpqbINDREZDZTZGaGyG4nUyxGi489RgtFReQIBCjmcJCptJSC9fVU6nKRfXKSwvX1tFxdTaaJCVr48ksiu52iFy9SuLYWhOhwUCwcpqjdTubSUrKMjFDszh0yzc5Sgc1G5lCIlnfvpvlnn6XFgweJiKjsV78iy/w82QMBMlksZPZ4KFxYSDOKQqbr18l+7RoVxMMPVwYHafbCBQo/8kjiRSouJvPhw2Tv6yPbzp20smULFQ4Pk//xxyn04YfkPH+ellpbyRSJ0ILe/g8YRvt+JgPPM78w0jxzJW0/EVWk2mD37t1ZDXjjxo2s91lXbN0Ka3ZiApblU08R+f0UDAapWT1Pn0/W1/7yS7xnt2Nf4eCzWOC8Ew67ykrZmkxYn8Eg6m/fvg1L22aTUoFwCjY0UGl5OfYXmrXFgrkWF2MfIqLGRrrtdtOWpSVprVssctzdu2FBz89DPqmrQxr9nj1EHg859uyh8vJyRJi0t2P7tjZErhQXE0UiZGtro5JPPoHEIeZRWEjU1ERbDh1Kbim73dKqfvRRmrPb6ZH334cDt7+f6BvfQDKQwSxtw30/k4DnmV/c73levnw56We5kvZFktZ2IxGdzXE840F9W19aei8VO6JOihHNUkUERW8vJILmZhDQiROQJUIh7CMaB4iSpSKe2WLBZ1VVIOAjRyA3jIzIZr9LS9heSBNuN973+Yj+7d+gTVuteB0OU6SiAvP4/HPMvbgY8s5LL0Ea8XoRibKwgO08HiwGO3fC6ShqmRDBQg+HQai7d8Mpev681NIjEZSk7ehAuGI6wm1vx2NTE1l+8QuM/+yziFrZvdtwhM1gGAE5kbaiKD93u93/P+6ALBcOyYcOwpGmSjuPTk6u1ruJQLqlpfK5qPMhrGvRNKCnB9Ztby/qT1+5Ii3qp58G6Xu9sK5DIVivxcUgUyJ8fuoUrPrJSZBuZSUIOxwG0Z86RVvv3gU5itKuVVUg/NJSWQDqwgWMNzCAaoAvvggt+0c/wrxWVtA2TBsh4/NhMbh2TdYb9/kS49LVGj6RjM1W95dsaqLlHTvkouVygfQZDMYq5BzypyjKD+NPH07CVkOdpj05uTo+ubMThKhHUMK6FohGYcWKmtM7diA1vKAAY5rNIOFXX4VzcnYW+xQWgpSHhvD+zAys3KIiRJjU1WH/sjKipSUq8HiwMBQWJkZuEGE7UTt73z40QrBaQcQ7d2JOIyM4zqlTCN1raUm8Hq+/jm1Pn8bCMTqK/UVcuCBnce52uyzdKopteTy0Ul+v77hlMBgJ4OSaTJFJKyttgX/xXB01cfEiiPbqVZDZ+DhS1z/+GGnsdjsiVBoaQHii52JDA45hMqFl2cgIkm5mZ2EtC6J2ubBtvF+kKRzG3A8eBLnu20f0J38CaaO3V8oy8/OwvltaUEhqZARRIsXF0OEXFpLHUe/YgbEvXIBF39ICUlY3/718GXNvbpYVD7Vp7A0NTNYMRhowaWcCvdoiRJnXcHa5QKh2OwhsZETWsBafb9uGprvFxSDPlRVZze+TT0CCY2OIld67F1azxQLrPBSCTBKJILPxxg28V1BAy1u2kDUYlCVVl5ZkN3ixmDQ3Q1+3WGAlz81hrg4H3jeZQLJnzmCRKC2VYY/iujgcRN/8JtGHH8KqN5uxqAgiFxr82Bg+e/JJOY7LxbVHGIwMwaSdCZJ1riHKvIazsMr7+5EUE4uBRC0WkKKIgw6FZK0QUZZ1aQl/djvmIWSOkhJov59+ivcXFiBlFBbiWNEoxYqKQPLl5SB8lSRBVVXS+g+HkYQjuu1Eo7CerVY8378fpH3+PMY/dAh1R9TNjZ1OWN3Cafrb38omEcePYztxx6Eo8u6EwWBkDCbtTKAm3GBQdi9PBa2cIqzyS5fweV0diOvpp2EtWyz6LbyGhxGed/cuLHHhcLTZ4GDs6JAp4YuLkFuExm23k3l2FgtBMAgCvnxZ6u7qut9+P445N4fXCws4fkkJ5nvnDs7J58OxL1wgOnwYY4vmxsvLmFt7O6SXyUmELwaDMgFHfcfB3b4ZjKzBpJ0JROOBEydgOZ49S+a2tuTbp+oM3tEBJ6JoQ6YOjVtZ0W/h9ad/CoJva0OkRiwGp2NfH6rtiRDBWAyLwNQUyHligqyRCIi/sxPEWVMjpYqqKhmxYTKB7EdGoCs//jisaaFhX7kCK9vrBWlXV0PeUDc3FnHmY2MYd2gIKenj4xhjaAjXkXs5MhhrBpN2phBdV+IWYkGq2hjpOoOnap7gdMI63boVTXd/9zuQYVkZSG58HNts2wb54ne/A9Hv24e47gMHiP7rv0Doovb13buwhmtrQdoOByx+UQ+lthbnt7gI6/jaNVT16+iQ866vRwx1RQVIet8+KW0ISUXdZiwQkJEi8/P3knFoZWX1+V+6RLahIZxzKqt7PVqUMRgbDEzamUITKRJRk4aWTNTbqpNotETj9yeGxHV1JZZHPXUK24lxPvpItiLbtUuOEwpJnd3vh/Y8N0c0M0MxkwkWuJB0PB5IG5WVRP/zP3Ba1tXBCi4shE6tl9yiblkmNGoi2Z1dLGouF8a6fVsWf/L7ER3z5JOJcpGql2ZFIIDzVNcrUSPZ3UsyMMEzHlIwaWcKjYV8L7lGONbs9kQy0Yb59fbKqBNBPiJeWV2utLYWJHjqFCxeqxXEt7yMY+zahap+hw5Bovi7v0P4IBFI79lnQaDPPUd05QpFZ2dxDJ8PURsWC9Gjj+KY//u/sOIXF6GBm0ypk1tUGYwJ56E+76EhojfeALGHQrDIjx7FuF1diQSq0tSj0SgWt2Qad6q7Fy2yJXgGYwOBSTsbqCJFzIoCQp6YgIPyxRdhUaobAOg53Ygk+Yh4ZXXSy8AA0c9+BiItLka4X7xQFEUikCCamkBGoiN6dTW2vX0bi4OQRiwWWnjySbIXFGCei4sg++lpRJwQwaIWnWpeew2fVVeD+IeGZD2R99+X4X6iI8/cXGI0issF7fvWLYzh9YK4w2EsEtpIEVUvTbNwgCbTuFPV7NYiG4JnMDYYmLTXiAKfD8TQ2grSVRQQk5pMkhGNeE9owKIR78AALG6XC87DHTtAxnY7wu6Kioj+6I+w709+IuOno1EQu9kMS72pCaRZVERRqxX1Sfr6QNpiDrdvg5BDIRC26KgurNSpKdwhdHZiv3izAvriCxB7ZSUImgjnIcYtL8c85udxl9DVhTuD5ubVxCkyKru6aGZoiMqPHEkd655JTHyq685gPARg0l4jIi4XUrb9fhCbqLOtrbehdbp5PIjMmJ6W6doNDdj35EkQ3Pw8HIKibklBAYiwogJEdPMmrNzt20GSL76IcX/9a2Qx9vfD+VhURLH2dhD5yZOwnD/66F44IJWUwIo+ckRmIoriVX/4Ayxmmw1Oz6UlfL60hEVCpMK3tSU6LA8cQEEpUaP77l1Y6END+jJFPKJmqbg4vTWcaUx8NgTPYGwwMGmvEdGyMn1i0NNTRfnVd98FIV65AnJTE5nLhWiR994DMVutIFLRyeaJJ0CARIiyGBvDmK2t0M4//hhzsVqxfWkpUUkJmRcW4Jj0ekHSJhO2sdkwv8JCWeCKCOfi9cLqX17G2Nu3Q1YJh/FcZGc6nYmETYTn3/2urHx48SL26e+XTX3vB4lmSvAMxgYDk3Yu0COGZHrq4CCcfiJFvaREOiPFGA0NRH//90Q//jG06WAQVm5bG6zjmhpsV12NfaNROAfffRdWsTnePa64GJa810s2ux3ZjCJBp6wM1nwkAvI+dChRa3a5YHl/8QXmOjuLMq4HDqy+Y0hmxaqjQ3p7QdhCSklmcTMYjIxgXNI2ashWfF7mQED/cz09VTQ56O+HtWoyQQJRa8ECjz1G9A//AJLz+0GW5eXyWvj9sozr0JAMGbRaQdpWK9Hzz0Or3rGDFm/cIOf4OIi6oIDoL/4C4YTT07CE9+1LOC+qqsIxn39eOh61kSTZyhQiC1TrtGQwGFnDmKRt1JAt1bycU1Nw8OlptN3dssSoywUN2m6HxdrfD/27vj71glRfL7MlRSKPiOu+cweW9f79KBBlNoPYFxeRmFNcjFjrYJBMy8sg5pERyBrDw3Aizs/DOenx4NHhSAxbfP31RMs61f9DhD4S6Vc6FFmgmTgGjbpYMxgGgTFJ26ghW6p5mUQ9be28fL7E5gLHj0vr2++HU1BNxiIbMpkm3t0trWkRMlhaimiR//xPhApGo0hfDwahdzudGNfjoUKbDSQdicCRSYSsyokJEHV1NZyGVVVYZMT1VneWv3kzsTCUukSrKkGGTCa8r02QSZUFqiV/Iy7WDIaBYEzSNmrIlmpeMZNJf16Dg7CCp6chQ5w4geYBel1f9AhKu2Cpa1J7PCBHUTmvshKW7ZkzqDVSWwsL/eRJHHvbNszpscdgRff3Y+EoK5Np8aGQ1LTVxaTUFq9olSYKQ/X0SItaXXSKKHmCjFrnTkbMRl2sGQwDwZikbdSQLdW85rxefSu7pweZjBMTyDx0OFZbrkTJCUq7YDU3S2mBCOGFW7ZA7ujrQ9ghEZyUVit04/l5aOfT01Rgs8GRuWcP0c9/Div7889xnKYmRKW0taGM6uwsxlKn1xcUIESxsBBW+d69kGcGB7G/qMMt6m63tsrFTE/qSHLeZjFOKGS8xZrBMBCMSdpExg3ZEvPyelff4ns8sGi/8Q1Yu9XV+s5GotR3E+p0cfUCpu6t+MgjyGD8/e9lLPXsLMhQZCJev05Bt5tKPvsMMktxMUh7eRn73L4tFwCHA8fVWvdXrxL97d/i2MPDuItwOvFa6OCiAYK6qUEyizqJo9Z5+rSMjunsXK2NMxgMIjIyaRsZapLRI6RwGNEX6oQbLfTuJrREJ2QL9QKm3aesDJavyM5saIAlXVhIVFNDy62tsrreM8+gAqAo/drWtrqkqta693pB0Hv3QnIpLyf64z+G1BIISB28vj6xf2SyOwm98755k0yxmNxWtE1jMBirsPlIOx/RCR4PSEbrmMtW1tHeTWSi6ar3EU7PykqQ7ssvg5ztdjQo+MUvqOjzz9EfsqoKlrKIeLl1C6VdRUnVF15IzNIU53H4MLqyX72K6JQtW7CPuuO8npShtagtlsQ7E00FwZgoVsWyCIOREpuLtPMVnVBVRaalJX3HXC6yTrYOWJGw43aDSEV51IICvG8yUbSoSG7f3AwpQzgp//IvIZmoZRdtliYR0Z/9GTqll5VBSvna12TsdrIFKpmso3fdXS6aO3oUi4+RfBgMhgGxuUg7X9EJLhcF29uRpdjamljdLxcIohMxz2JcvdC4wUFEjfT3w4Lt6JAk394OvZqIQqLkqnCGfv/7MoZc1BsRIX1610VY8+EwyDcSwZ/4PNU5C+JPNX4c0bKyRHmFwWDoYnORdh5DCcM7doBU/f783NIL2cZiQeq3ukaJ0ymtU3G3cOcOCPu55/C8qwvjiDuJUIjIbCbL1BSsarU80d2deOxU18XjkU2GJycxVnW1fpx1MmRw3c2BQHZjMhibFJuLtPMYSnivYJSwinOBSFAJBGRzXJMJC4LJlFijRNwtuN0guTt3ZDlW7Z1EZyctTE+jcUIaeSLpdREda154AU7JV15BFmY2ElO6657MsctgMFZhc5E2Uf5DCXt7QWCiM81axh4cRH2O0lI4A+vrEX99965s0yUsZYtFZld2dKyOUBFtycxmoqNHKVxcDK07Gwen9n014Xo8iEPPVmJSh0WqX8ffS4ge4aQaBiMpNh9p5xO5aOTqKBYiODSJEKp34ACcfl/9KizpRx9NtJS1PRnV8HrRlcbpJHr/fTIfPgyrPJ0slCqqRp3NGAjoJ8Cki8pJ5QS2WNBUQrRNs/DXksFIBv515IJ0Wm0yItOrL7J3LzTjvXthPYdCIObaWljg6sVBELbWah0cBPGJbuhTU7JrvDZhRzvPdFE16m2IEhNgMtk/2QIXd3TGiHBX4XajtVmyBr8MxiYHk3Yu0EZ7qJFNjQ2vF6F4VVV4LC9fXWCptzcx5jnZ2HY7nJgLC0RmM5mCQaJ/+ieMK/pLaud56ZJ+v0c1tHNWJ8BkcseRLG47ECCKRChWVARpqLwcj11dsigVg8G4BybtdMgkGUdP105FZFoCIwLZbt2KZJ0rV2TyijiuVlfWG7upCfr31BQ07ddeI+dbb0EbLymBg1NboU/dTYdIpqhrIzlS3VVkEpWTLG47FCIiogKvF3MuKrr3HoPBWA0m7VTI5bY/FZHplSrt6UGyztISJA6LBfp2Z6eUCtTH1hvb5Uqsg+3xUNRmQyGpTz9F4+GtWxMr9In0d1E18LHHEglV7ehMFgGSaVROsrjtzk6a27aNqkV7NVHDhMFgrEJOpO12u8uJqDH+16koyhv5mJRhsJbbfjWBpiIyLQl3daHgk9NJ9JvfwOIsK8N7QrtWNxpIRaBacp+exvPqatTe1i4s/f2w8B0OolOn8FhXh0Vkdha6ujZLUotU0SFaaK9ZUxMa+3Z3G6+yI4NhMORqab9GRKQoyptut7vT7XZ/W1GUN/MwL2Mg29v+dASaCk1NIMe5OZBmOAy9t7UVVrdeowGRQZhMwhGZm0RIillYgEWtrSh4/To+r6uTxaMUBREtmWZ8ZlMiQO+aTU4mv17czYbBuIecSFtD0I1EdDa36RgM2dz2E2VmZWZyrFdegXNS9GgUz9WNBgYH8TpNXY97mZvV1SDj48fxwaVLkGSiUUgn0SgWp44ObOP14vNMMz6zDX/MdEHjbjYMRgJMMREfnAPcbncjEb2hKMp31O9fvnw55nA4shorFAqR3W7PeU7rDfU8zYEAOU+fRoJIKESLHR0Urq9H1uQaoR7TtLREtLhI9sFBoliMlpqbKWq3E9ntZJ6ZoZjdTitbtlDh8DDNP/ssLYlmvfF5OpaWqMDno0ic7JynT5PF46HCoSEK7dtH9i++oND+/WReWKD5p5+mpf37yRwIkHV0lCgWo/COHWnPRT3fmMlEc0ePZnX+yf7v1uFhKr5wgVZqasgyNUULhw5R+JFHMh4339iI308jg+epj2AwSAcPHjTpfZbW0na73a8SUYXm7VuKopxTvX5VS9gCu3fvzniiREQ3btzIep8HgYR53ryJFOzycmjCRUWwEHOxCsWYKmddAi5exGf9/bCKr14lisVoy9gY0ZEj945748YNcqu7qYtx3W70mTSZ0DuypISopIS2dHbiPHp6EuPIMzkPt3vNMkbS//vWrcgOjUQg4Rw69EAt7Q35/TQweJ76uHz5ctLP0pK2oig/T/W52+1+VVGUH8aftyuK0pv1DDc61A49k0mWSs0lHVvHWZdQeU/EbTudyKDs6clMf1Y3Ge7sRLbl3BwkkspKSC3t7VLq6O+HlNLcnDwLU60557tSXx7rxTAYDwNyjR45QkQ/cLvdfx1/6+GKHskU6iSb0tL8VP7LJsSOCOVWMzmueq6zs+hyEwigQcL+/RiDSC5CV65AC3/77dUVB4nuj+ac73oxDMYGRq6OyHNExAG1RCCVjg5ZbS8fVmG6EDv1Z9lao729aD4syrvevAkCLy6GxX3sGCxsIkgn167hUV1xkChvNcq5NCuDkRnMD3oCDx1crtWd1x/EcX0+ops3QYZaqJNqYjGQ7Z49yEh0OCCREGERcjrR3d1iweNasiHTQZRmPXcOVruol8JgMFaBMyIfRqgkC+fUlOwLKaDVtUUDBeHcFBZzS0tiGKKepp0PzZlLszIYGYNJ26jIJaFEJVmYJidXk6Ae0aqdm9rMznTHz1Vz5sa+DEbGYNI2InJ17qkki5jJlDyTM98W81rhWmNjX86UZGxCMGkbEbk691QEPOf1Zr7vekRpZEisWTf25UxJxiYFk7YRkQ/nXpyAoysra59HrpbsehJrnqJWGIyNBiZtI8IICSX5INzBQRSgykeykRb5WNgYjA0IJm2j4kEnlORqyfp8yNLs70f8dUdHfonVCAsbg/EAwKS9mZCN3JGrJevxoBvPSy+BuLu6OFOSwcgDmLQ3C7KVO9Tp7muBOha8tpY70TAYeQKT9kZBrk7BdHKHdnyfD4Td0wOLWd3/MtM5snzBYOQdTNobAflwCqaSO7Tjd3cjlf3OHUgbL72U3pGYbI5M1gxGXsG1RzYC1FayKNiULYTle+TIatLXjj8wgEe3G6Vm+/vT69r5mCODwUgLtrQ3AvIV3pbM8tWO39xMNDQE67qjI7Ej+3rPkcFgpAST9kbAeuvDeuNnezzWsBmM+wIm7Y2C9daH9WqRZHu8bJr1MrkzGGsCkzbj/oJrhjAYOYEdkQ8L4k0PDN9AgB2WDEZOYEv7YcCDsl7XInOww5LByAlM2g8DHkTFu7UuFOywZDByApP2wwBhvfb3o3O65T78W3NZKDjphsFYM1jTfhjgciGLMRiUjXnXW9tmmYPBeCBgS/thgWi6e78kEpY5GIwHAibthwUPwvJlmYPBuO9g0n5YwJYvg7EpwKT9MIEtXwbjoQc7IhkMBmMDgUmbwWAwNhCYtBkMBmMDgUmbwWAwNhCYtBkMBmMDIW+k7Xa7f5CvsRgMBoOhj7yQttvtPkJEjfkYi8FgMBjJkTNpu93uRiK6lYe5MBgMBiMNTLFYLKcB3G73EUVRzrnd7ncVRTmm/uzy5cu5Dc5gMBibFAcPHjTpvZ82I9Ltdr9KRBWat2/FifqIoijnsj0og8FgMNaGnCxtt9vdTiD0ciL6ayL6lqIovfmZGoPBYDC0yEnTVhSlN25pC+JmMBgMxjoiZ007X3C73T4iukREZxVF+aHO568SkZ+IGhVFefM+T0/MoZwQJdNIRJ2Koryhs03K81jHuaW8Pnz9ssMG+T62E9G78XkQEZ3TXtMHfT3j1+k7iqJ0a97zk4G+q9p5Gvm7aqTkmmOKonSn+IGQ0M/jIYYPAq8RUYeiKD+Pz+PbOtskPY/1Qrrrw9dvTdgI38cKRVGaFEU5SETfIqJ/1tnmgV5P8b8WMOp3VTtPMvB31UikXR4PH9RDJ8mwwltE1H5/ppQIRVHeVK38yUIdU53HeiHd9eHrlz02wvdRHQTQqCiKka+nAH9Xc4SRSLuCiGbcbreetVCueV25/tNJjvg/aiZJ5Eyq81gvlGtea69Pus/vKwx4/fSwkb6P39axFAWMcj0FyjWv+buaJe5bE4RUoYNEWNni2/ndbvermi+hX2ffBzLPOF5VFOU7evunOY/1gp9SX590n99vGO36ZTsPPxnrenYTka72a5TrqYKf+LuaE+4baac6obhedClFuOBFkitwIxGdze/sJNJd+Pg/54fx5+3qOWdwHuuFdNfnvl2/dDDo9UuAkb6P6RB3mCX7zBDXUwP+ruYIo8gj7xAlOCGE+H9W9box7pQoT5XQs56IH/8Hbrf7stvtvkxxi0DMk5Kcx3oj2fXh67dmbIjvYxwVRDSjfsNI1zN+jTo0czDcd1U7TyN/Vw0T8sdgMBiM9DCKpc1gMBiMDMCkzWAwGBsITNoMBoOxgcCkzWAwGBsITNoMBoOxgcCkzWAwGBsITNoMBoOxgcCkzWAwGBsI/wfrk9hq8Gt/1AAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 432x288 with 0 Axes>"
      ]
     },
     "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",
    "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.show()\n",
    "plt.savefig(\"toy_problem.png\",dpi=300, bbox_inches='tight')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "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 range(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 range(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": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "C=1\n",
    "MAXN=1000\n",
    "\n",
    "#Repeat ten times to get smooth curves\n",
    "for i in range(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 range(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": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "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": 19,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/santisegui/opt/anaconda3/lib/python3.8/site-packages/IPython/html.py:12: ShimWarning: The `IPython.html` package has been deprecated since IPython 4.0. You should import from `notebook` instead. `IPython.html.widgets` has moved to `ipywidgets`.\n",
      "  warn(\"The `IPython.html` package has been deprecated since IPython 4.0. \"\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "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 range(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 range(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": 20,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean accuracy: [0.82536232 0.8263285  0.83188406 0.83429952 0.8352657  0.8326087\n",
      " 0.82971014 0.82560386 0.82198068 0.81956522 0.81256039 0.81086957\n",
      " 0.80483092 0.80169082 0.80072464 0.78599034 0.77995169 0.78091787]\n",
      "Selected model index: 4\n",
      "Complexity: 6\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "%reset -f\n",
    "%matplotlib inline\n",
    "\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn import metrics\n",
    "from sklearn import tree\n",
    "from sklearn.model_selection import KFold\n",
    "import pickle\n",
    "\n",
    "ofname = open('files/ch05/dataset_small.pkl','rb') \n",
    "(X,y) = pickle.load(ofname, encoding=\"latin1\")\n",
    "\n",
    "#Create a 10-fold cross validation set\n",
    "kf = KFold(n_splits=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.split(y):\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 range(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": 21,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Mean accuracy: [0.8254832  0.83031158 0.83091854 0.83423816 0.8339406  0.83243002\n",
      " 0.82759892 0.820653   0.81883577 0.818237   0.80887144 0.80373822\n",
      " 0.79891166 0.79528628 0.79559113 0.78230262 0.7841144  0.78351199]\n",
      "Selected model index: 3\n",
      "Complexity: 5\n",
      "Test accuracy: 0.8260869565217391\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "%reset -f\n",
    "%matplotlib inline\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn import metrics\n",
    "from sklearn import tree\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.model_selection import KFold\n",
    "\n",
    "import pickle\n",
    "\n",
    "ofname = open('files/ch05/dataset_small.pkl','rb') \n",
    "(X,y) = pickle.load(ofname, encoding='latin1')\n",
    "\n",
    "#Train_test split\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.20, random_state=42)\n",
    "\n",
    "#Create a 10-fold cross validation set\n",
    "kf=KFold( n_splits=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.split(X_train):\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 range(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": 22,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.856280193236715\n",
      "[[3372  590]\n",
      " [   5  173]]\n"
     ]
    }
   ],
   "source": [
    "%reset -f\n",
    "import numpy as np\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn import svm\n",
    "from sklearn import linear_model\n",
    "from sklearn.model_selection import KFold,GridSearchCV\n",
    "from sklearn import metrics\n",
    "\n",
    "import pickle\n",
    "ofname = open('files/ch05/dataset_small.pkl','rb') \n",
    "(X,y) = pickle.load(ofname, encoding='latin1')\n",
    "\n",
    "parameters = {'C':[1e4,1e5,1e6],'gamma':[1e-5,1e-4,1e-3]}\n",
    "\n",
    "N_folds = 5\n",
    "\n",
    "kf=KFold( n_splits=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.split(X):\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 = 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": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'C': [10000.0, 100000.0, 1000000.0], 'gamma': [1e-05, 0.0001, 0.001], 'class_weight': [{1: 0.25}]}\n",
      "{'C': 100000.0, 'class_weight': {1: 0.25}, 'gamma': 0.001}\n",
      "{'C': 100000.0, 'class_weight': {1: 0.25}, 'gamma': 0.001}\n",
      "{'C': 100000.0, 'class_weight': {1: 0.25}, 'gamma': 0.001}\n",
      "{'C': 100000.0, 'class_weight': {1: 0.25}, 'gamma': 0.001}\n",
      "{'C': 100000.0, 'class_weight': {1: 0.25}, 'gamma': 0.001}\n",
      "Mean accuracy: 0.8555555555555555\n",
      "{'C': [10000.0, 100000.0, 1000000.0], 'gamma': [1e-05, 0.0001, 0.001], 'class_weight': [{1: 0.5}]}\n",
      "{'C': 100000.0, 'class_weight': {1: 0.5}, 'gamma': 0.001}\n",
      "{'C': 1000000.0, 'class_weight': {1: 0.5}, 'gamma': 0.0001}\n",
      "{'C': 1000000.0, 'class_weight': {1: 0.5}, 'gamma': 0.0001}\n",
      "{'C': 1000000.0, 'class_weight': {1: 0.5}, 'gamma': 0.0001}\n",
      "{'C': 1000000.0, 'class_weight': {1: 0.5}, 'gamma': 0.0001}\n",
      "Mean accuracy: 0.8557971014492753\n",
      "{'C': [10000.0, 100000.0, 1000000.0], 'gamma': [1e-05, 0.0001, 0.001], 'class_weight': [{1: 1}]}\n",
      "{'C': 10000.0, 'class_weight': {1: 1}, 'gamma': 0.001}\n",
      "{'C': 100000.0, 'class_weight': {1: 1}, 'gamma': 0.001}\n",
      "{'C': 10000.0, 'class_weight': {1: 1}, 'gamma': 0.001}\n",
      "{'C': 1000000.0, 'class_weight': {1: 1}, 'gamma': 0.0001}\n",
      "{'C': 1000000.0, 'class_weight': {1: 1}, 'gamma': 0.0001}\n",
      "Mean accuracy: 0.8562801932367149\n",
      "{'C': [10000.0, 100000.0, 1000000.0], 'gamma': [1e-05, 0.0001, 0.001], 'class_weight': [{1: 2}]}\n",
      "{'C': 100000.0, 'class_weight': {1: 2}, 'gamma': 0.0001}\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn import svm\n",
    "from sklearn.model_selection import KFold,GridSearchCV\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=KFold(n_splits=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.split(X):\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 = 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": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "%reset -f\n",
    "from sklearn.model_selection import KFold\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn import ensemble\n",
    "from sklearn import metrics\n",
    "import numpy as np\n",
    "\n",
    "import pickle\n",
    "ofname = open('files/ch05/dataset_small.pkl','rb') \n",
    "(X,y) = pickle.load(ofname, encoding = 'latin1')\n",
    "\n",
    "dvals = [{1:0.25},{1:0.5},{1:1},{1:2},{1:4},{1:8},{1:16}]\n",
    "\n",
    "kf=KFold(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.split(X):\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": {
    "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": {
    "scrolled": false
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}