Classification NN.ipynb

{
    "metadata": {
        "language_info": {
            "file_extension": ".py", 
            "mimetype": "text/x-python", 
            "version": "2.7.11", 
            "pygments_lexer": "ipython2", 
            "name": "python", 
            "codemirror_mode": {
                "version": 2, 
                "name": "ipython"
            }, 
            "nbconvert_exporter": "python"
        }, 
        "kernelspec": {
            "display_name": "Python 2 with Spark 2.0", 
            "language": "python", 
            "name": "python2-spark20"
        }
    }, 
    "cells": [
        {
            "metadata": {
                "collapsed": false
            }, 
            "cell_type": "code", 
            "outputs": [
                {
                    "metadata": {}, 
                    "execution_count": 1, 
                    "data": {
                        "text/plain": "   Material  Usura_mat_g  Durezza_Disco  Comprex_LP  Comprex_LR  Grindo_LP  \\\n0         3          7.3            228         187         180          0   \n1         4         10.4            195         229         236          0   \n2         3          6.9            179         194         191        676   \n3         3          7.7            226         161         179          0   \n4         4         10.0            183         232         226          0   \n\n   Grindo_LR  CodPinza_0  CodPinza_1  CodPinza_3         ...          \\\n0          0           0           0           0         ...           \n1          0           0           0           0         ...           \n2        718           0           1           0         ...           \n3          0           0           1           0         ...           \n4          0           0           1           0         ...           \n\n   PistoneCpx_54  PistoneCpx_57  Inerzia_45.0  Inerzia_51.0  Inerzia_65.0  \\\n0              1              0             0             0             1   \n1              1              0             0             0             1   \n2              1              0             0             0             1   \n3              1              0             0             0             1   \n4              1              0             0             0             1   \n\n   Inerzia_72.5  Raggio_medio_104.0  Raggio_medio_113.0  Raggio_medio_117.7  \\\n0             0                   1                   0                   0   \n1             0                   1                   0                   0   \n2             0                   1                   0                   0   \n3             0                   1                   0                   0   \n4             0                   0                   0                   0   \n\n   Raggio_medio_118.0  \n0                   0  \n1                   0  \n2                   0  \n3                   0  \n4                   1  \n\n[5 rows x 50 columns]", 
                        "text/html": "<div>\n<table border=\"1\" class=\"dataframe\">\n  <thead>\n    <tr style=\"text-align: right;\">\n      <th></th>\n      <th>Material</th>\n      <th>Usura_mat_g</th>\n      <th>Durezza_Disco</th>\n      <th>Comprex_LP</th>\n      <th>Comprex_LR</th>\n      <th>Grindo_LP</th>\n      <th>Grindo_LR</th>\n      <th>CodPinza_0</th>\n      <th>CodPinza_1</th>\n      <th>CodPinza_3</th>\n      <th>...</th>\n      <th>PistoneCpx_54</th>\n      <th>PistoneCpx_57</th>\n      <th>Inerzia_45.0</th>\n      <th>Inerzia_51.0</th>\n      <th>Inerzia_65.0</th>\n      <th>Inerzia_72.5</th>\n      <th>Raggio_medio_104.0</th>\n      <th>Raggio_medio_113.0</th>\n      <th>Raggio_medio_117.7</th>\n      <th>Raggio_medio_118.0</th>\n    </tr>\n  </thead>\n  <tbody>\n    <tr>\n      <th>0</th>\n      <td>3</td>\n      <td>7.3</td>\n      <td>228</td>\n      <td>187</td>\n      <td>180</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>...</td>\n      <td>1</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>1</th>\n      <td>4</td>\n      <td>10.4</td>\n      <td>195</td>\n      <td>229</td>\n      <td>236</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>...</td>\n      <td>1</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>2</th>\n      <td>3</td>\n      <td>6.9</td>\n      <td>179</td>\n      <td>194</td>\n      <td>191</td>\n      <td>676</td>\n      <td>718</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>...</td>\n      <td>1</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>3</th>\n      <td>3</td>\n      <td>7.7</td>\n      <td>226</td>\n      <td>161</td>\n      <td>179</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>...</td>\n      <td>1</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n    </tr>\n    <tr>\n      <th>4</th>\n      <td>4</td>\n      <td>10.0</td>\n      <td>183</td>\n      <td>232</td>\n      <td>226</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>...</td>\n      <td>1</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>0</td>\n      <td>1</td>\n    </tr>\n  </tbody>\n</table>\n<p>5 rows \u00d7 50 columns</p>\n</div>"
                    }, 
                    "output_type": "execute_result"
                }
            ], 
            "execution_count": 1, 
            "source": "# The code was removed by DSX for sharing."
        }, 
        {
            "metadata": {}, 
            "cell_type": "markdown", 
            "source": "NB: in the whole code we have to replace cross_entropy with accuracy !"
        }, 
        {
            "metadata": {
                "collapsed": false
            }, 
            "cell_type": "code", 
            "outputs": [
                {
                    "output_type": "stream", 
                    "text": "The autoreload extension is already loaded. To reload it, use:\n  %reload_ext autoreload\n", 
                    "name": "stdout"
                }
            ], 
            "execution_count": 12, 
            "source": "%matplotlib inline\nplt.rcParams['figure.figsize'] = (10.0, 5.0) # set default size of plots\nplt.rcParams['image.interpolation'] = 'nearest'\nplt.rcParams['image.cmap'] = 'gray'\n\n# for auto-reloading extenrnal modules\n# see http://stackoverflow.com/questions/1907993/autoreload-of-modules-in-ipython\n%load_ext autoreload\n%autoreload 2"
        }, 
        {
            "metadata": {}, 
            "cell_type": "markdown", 
            "source": "# Create reusable functions"
        }, 
        {
            "metadata": {
                "collapsed": true
            }, 
            "cell_type": "code", 
            "outputs": [], 
            "execution_count": 13, 
            "source": "# Given a set of model parameters (learning rate, regularization penalty coefficient), find the optimal parameters with cross validation\n\ndef cross_validation(set_of_learning_rates, set_of_regs, X_t, y_t):\n    n_folds = 3\n    skf = StratifiedKFold(y_t, n_folds=n_folds) # Get train-validation set stratified (keeps the same distribution) splitter\n    set_of_params = [(x,y) for x in set_of_learning_rates for y in set_of_regs]\n    \n    print \"Number of folds: \" + str(n_folds)\n    print \"Number of parameters combinations: \" + str(len(set_of_params))\n    \n    cross_entropy = np.zeros(len(set_of_params))\n\n    # Repeat for every train-validate combination \n    for cv_train_index, cv_val_index in skf:\n        X_cv_train, X_cv_val = X_t[cv_train_index], X_t[cv_val_index]\n        y_cv_train, y_cv_val = y_t[cv_train_index], y_t[cv_val_index]\n\n        for param_idx, param in enumerate(set_of_params):\n            print \"Combination \" + str(param_idx) + \": Parameters: \" +  str(param)\n            print \"Combination \" + str(param_idx) + \": Train model\" \n            W, b, W2, b2 = train_nn(X_cv_train, y_cv_train, param[0], param[1]) # Train\n            print \"Combination \" + str(param_idx) + \": Predict\" \n            _, cross_entropy_cv = predict(X_cv_val, y_cv_val, W, b, W2, b2) # Predict for valiation set\n            cross_entropy[param_idx] = cross_entropy[param_idx] + cross_entropy_cv # Calculate error (cross entropy)\n\n    cross_entropy = cross_entropy / n_folds # Calculate average error\n    \n\n    print \"BEST LEARNING RATE = \", set_of_params[np.argmax(cross_entropy)][1]\n    print \"BEST REG (CROSS-ENTROPY) = \", set_of_params[np.argmax(cross_entropy)][0]\n    \n    # Plot the results\n    fig = plt.figure()\n    ax = fig.add_subplot(111, projection='3d')\n    Y = map(lambda x: x[0], set_of_params)\n    X = map(lambda x: x[1], set_of_params)\n    for i in np.arange(len(X)):\n        ax.scatter(X[i], Y[i], cross_entropy[i], color = \"r\")\n        ax.plot_surface(X[i], Y[i], cross_entropy[i], color = \"b\")\n    ax.set_xlabel('alpha')\n    ax.set_ylabel('L1')\n    ax.set_zlabel('R2')\n    plt.show()\n    \n    return set_of_params[np.argmax(cross_entropy)]"
        }, 
        {
            "metadata": {
                "collapsed": true
            }, 
            "cell_type": "code", 
            "outputs": [], 
            "execution_count": 4, 
            "source": "# Classify the X input set using 2-layer network with weight and bias parameters for each layer as parameters (W,b and W2, b2)\ndef predict(X, y, W, b, W2, b2):\n    hidden_layer = np.maximum(0, np.dot(X, W) + b)\n    scores = np.dot(hidden_layer, W2) + b2\n    predicted_class = np.argmax(scores, axis=1)\n    print ' > Prediction accuracy: %.2f' % (np.mean(predicted_class == y))\n    return predicted_class, np.mean(predicted_class == y)"
        }, 
        {
            "metadata": {
                "collapsed": false
            }, 
            "cell_type": "code", 
            "outputs": [], 
            "execution_count": 5, 
            "source": "# Train the 2-layer network (X_i input features, y_i input labels)\ndef train_nn(X_i, y_i, learning_rate, reg):\n    \n    # Parameter initilization\n    W = 0.01 * np.random.randn(D,h)\n    b = np.zeros((1,h))\n    W2 = 0.01 * np.random.randn(h,K)\n    b2 = np.zeros((1,K))\n    num_examples = X_i.shape[0]\n    \n    # Epochs\n    for i in xrange(50000):\n\n        hidden_layer = np.maximum(0, np.dot(X_i, W) + b)\n        scores = np.dot(hidden_layer, W2) + b2 # Logits\n\n        exp_scores = np.exp(scores)\n        probs = exp_scores / np.sum(exp_scores, axis=1, keepdims=True) # Turn logits into probabilities\n\n        corect_logprobs = -np.log(probs[range(num_examples),y_i])\n        data_loss = np.sum(corect_logprobs)/num_examples\n        reg_loss = 0.5*reg*np.sum(W*W) + 0.5*reg*np.sum(W2*W2) # Apply regularization\n        loss = data_loss + reg_loss # Calculate loss\n\n        # Backpropagation\n        dscores = probs\n        dscores[range(num_examples),y_i] -= 1\n        dscores /= num_examples\n\n        \n        dW2 = np.dot(hidden_layer.T, dscores)\n        db2 = np.sum(dscores, axis=0, keepdims=True)\n\n        dhidden = np.dot(dscores, W2.T)\n\n        dhidden[hidden_layer <= 0] = 0\n\n        dW = np.dot(X_i.T, dhidden)\n        db = np.sum(dhidden, axis=0, keepdims=True)\n\n        dW2 += reg * W2\n        dW += reg * W\n\n        # Update parameters\n        W += - learning_rate * dW\n        b += - learning_rate * db\n        W2 += - learning_rate * dW2\n        b2 += - learning_rate * db2\n\n    print \"iteration %d: loss %f\" % (i, loss)\n    return W, b, W2, b2"
        }, 
        {
            "metadata": {
                "collapsed": true
            }, 
            "cell_type": "code", 
            "outputs": [], 
            "execution_count": 6, 
            "source": "# Calculate the final score for the model using test set\ndef score_model(X_tr, Y_tr, X_ts, Y_ts, best_learning_rate, best_reg):\n    print \"BEST LEARNING RATE (CROSS-ENTROPY) = \" + str(best_learning_rate)\n    print \"BEST REG (CROSS-ENTROPY) = \" + str(best_reg)\n\n    # Train the model with the best hyperparameters\n    W, b, W2, b2 = train_nn(X_tr, Y_tr, best_learning_rate, best_reg) \n    \n    # Predict test data\n    predicted_class, cross_entropy_test = predict(X_ts, Y_ts, W, b, W2, b2)\n    print \"CROSS-ENTROPY = \" +str(cross_entropy_test)\n    print predicted_class\n    print Y_ts\n    \n    # Plot the results\n    plt.scatter(np.arange(0, len(Y_ts)), predicted_class, color='b')\n    plt.scatter(np.arange(0, len(Y_ts)), Y_ts, color='r')\n\n    lines = []\n    ax = plt.axes()\n\n    for i in range(predicted_class.shape[0]):\n        lines.append([(i, predicted_class[i]), (i, Y_ts[i])])\n\n    lc = mc.LineCollection(lines, colors=\"g\", linewidths=1)\n\n    ax.add_collection(lc)\n    ax.autoscale()\n    ax.margins(0.1)\n    plt.grid(True)\n\n    plt.show()"
        }, 
        {
            "metadata": {}, 
            "cell_type": "markdown", 
            "source": "# Perfom the evaluation over the data"
        }, 
        {
            "metadata": {
                "collapsed": false
            }, 
            "cell_type": "code", 
            "outputs": [], 
            "execution_count": 7, 
            "source": "# Prepare the dataset & configurations for the network\nX = df_data_1.drop(\"Material\",1).values\nD =  X.shape[1]\nnum_examples = X.shape[0]\nK = len(df_data_1[\"Material\"].unique())\ny = df_data_1[\"Material\"].values\n\nstep_size = 1e-04\nreg = 1e-03\n\nh = 30 # size of hidden layer"
        }, 
        {
            "metadata": {
                "collapsed": false
            }, 
            "cell_type": "code", 
            "outputs": [
                {
                    "output_type": "stream", 
                    "text": "Input data shape:  (651, 49)\nTarget data shape (651,)\nNumber of features:  49\nNumber of classes:  5\nNumber of examples:  651\nNumber of hidden units:  30\n", 
                    "name": "stdout"
                }
            ], 
            "execution_count": 8, 
            "source": "print \"Input data shape: \",X.shape\nprint \"Target data shape\",y.shape\nprint \"Number of features: \",D\nprint \"Number of classes: \",K\nprint \"Number of examples: \",num_examples\nprint \"Number of hidden units: \",h"
        }, 
        {
            "metadata": {
                "collapsed": false
            }, 
            "cell_type": "code", 
            "outputs": [
                {
                    "output_type": "stream", 
                    "text": "Train shape: (554, 49)\nTest shape: (97, 49)\n", 
                    "name": "stdout"
                }
            ], 
            "execution_count": 9, 
            "source": "# Split the train and test datasets\n\nfrom sklearn.cross_validation import StratifiedShuffleSplit\nsss = StratifiedShuffleSplit(y, 1, test_size=.15, random_state=0)\n\nfor train_index, test_index in sss:\n    X_train, X_test = X[train_index], X[test_index]\n    y_train, y_test = y[train_index], y[test_index]\n\n    print \"Train shape: \" + str(X_train.shape)\n    print \"Test shape: \" + str(X_test.shape)"
        }, 
        {
            "metadata": {
                "scrolled": false, 
                "collapsed": false
            }, 
            "cell_type": "code", 
            "outputs": [
                {
                    "output_type": "stream", 
                    "text": "Cross validaton\nNumber of folds: 3\nNumber of parameters combinations: 4\nCombination 0: Parameters: (0.01, 0.01)\nCombination 0: Train model\niteration 49999: loss 0.552327\nCombination 0: Predict\n > Prediction accuracy: 0.70\nCombination 1: Parameters: (0.01, 0.02)\nCombination 1: Train model\niteration 49999: loss 0.563649\nCombination 1: Predict\n > Prediction accuracy: 0.75\nCombination 2: Parameters: (0.02, 0.01)\nCombination 2: Train model\niteration 49999: loss 1.501722\nCombination 2: Predict\n > Prediction accuracy: 0.35\nCombination 3: Parameters: (0.02, 0.02)\nCombination 3: Train model\niteration 49999: loss 1.501722\nCombination 3: Predict\n > Prediction accuracy: 0.35\nCombination 0: Parameters: (0.01, 0.01)\nCombination 0: Train model\niteration 49999: loss 1.506124\nCombination 0: Predict\n > Prediction accuracy: 0.36\nCombination 1: Parameters: (0.01, 0.02)\nCombination 1: Train model\niteration 49999: loss 0.760513\nCombination 1: Predict\n > Prediction accuracy: 0.64\nCombination 2: Parameters: (0.02, 0.01)\nCombination 2: Train model\niteration 49999: loss 1.505455\nCombination 2: Predict\n > Prediction accuracy: 0.36\nCombination 3: Parameters: (0.02, 0.02)\nCombination 3: Train model\niteration 49999: loss 1.505455\nCombination 3: Predict\n > Prediction accuracy: 0.36\nCombination 0: Parameters: (0.01, 0.01)\nCombination 0: Train model\niteration 49999: loss 0.613270\nCombination 0: Predict\n > Prediction accuracy: 0.73\nCombination 1: Parameters: (0.01, 0.02)\nCombination 1: Train model\niteration 49999: loss 1.505843\nCombination 1: Predict\n > Prediction accuracy: 0.36\nCombination 2: Parameters: (0.02, 0.01)\nCombination 2: Train model\niteration 49999: loss 1.505455\nCombination 2: Predict\n > Prediction accuracy: 0.36\nCombination 3: Parameters: (0.02, 0.02)\nCombination 3: Train model\niteration 49999: loss 1.505455\nCombination 3: Predict\n > Prediction accuracy: 0.36\nBEST LEARNING RATE =  0.01\nBEST REG (CROSS-ENTROPY) =  0.01\n", 
                    "name": "stdout"
                }, 
                {
                    "metadata": {}, 
                    "data": {
                        "text/plain": "<matplotlib.figure.Figure at 0x7f25fba1fb50>", 
                        "image/png": "iVBORw0KGgoAAAANSUhEUgAAAjwAAAElCAYAAADgPjhUAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzsvXuUXFWZ/v/UvaqrL0lAhjQdciEdkiAgweayJo7AYAed\nQHCWQkZGdHkXgkbXYETBjDKDgIoaYvyiI0NAJi3DOHbEZUTiKKD4azV4GWBJmku33YLQufSluqrr\ncs7vj573ZNfuc6+zT+2q3p+1spJOV53adW77Oe/7vO+O6LoOhUKhUCgUimYmWu8BKBQKhUKhUIhG\nCR6FQqFQKBRNjxI8CoVCoVAomh4leBQKhUKhUDQ9SvAoFAqFQqFoepTgUSgUCoVC0fTEHX6vatYV\nCoVCoVA0ChGrX6gIj0KhUCgUiqZHCR6FQqFQKBRNjxI8CoVCoVAomh4leBQKhUKhUDQ9SvAoFAqF\nQqFoepTgUSgUCoVC0fQowaNQKBQKhaLpUYJHoVAoFApF06MEj0KhUCgUiqZHCR6FQqFQKBRNjxI8\nCoVCoVAomh4leBQKhUKhUDQ9SvAoFAqFQqFoepTgUSgUCoVC0fQowaNQKBQKhaLpUYJHoVAoFApF\n0xOv9wAUCoV7dF2Hrusol8uIRqOIRCLG3wqFQqGwRgkehaIBIJFTqVSg6zoKhcIckaPrOpLJJGKx\nGKLRqBJECoVCwaAEj0IhKbquo1KpoFKpoFAooFKpoKWlBZFIBLFYzBAxFPXJ5/OIRqMolUpztkXC\nJxqNKkGkUCjmJUrwKBQSoes6NE0zhA6AKmGjaRpmZmZQKpWqBAvB/pvdJgBju3aCiBVDShApFIpm\nIkI3Qwtsf6lQKGqHIjQkcuiajEQiiEQi0HUduVzOECqJRML4fxJB9DeAOULFSbjQ59H2zLASRLRN\nJYgUCoUkWN6MlOBRKOoE78sBqkVOqVRCsVisiua0tbUBAIrFYpXIIFHU0tJiKoTCFkSUclOCSKFQ\nhIzlzUaltBSKEGEjOSRASBBEIhGUy2UUi0XMzMwgGo0ilUqhpaUF5XIZMzMzhhjiIUHBp7j4zwZQ\nJYAozUX/R+OwE0V226U0nKZpKJVKSKVSAI6JLCWIFApFvVCCR6EQjJUvh/6QL6dYLBqVVu3t7YjF\nYoGOg0SF1XYpgsNGhSj65FUQRSIRFItFxGKxOYKIjWYRfJpMCSKFQhE0SvAoFAJw48spFosoFoso\nl8tIJBJoaWlBPB43ndytIjtBwgoLM1HkRRCxaTm3EaJKpYJyuWxsh0UJIoVCUStK8CgUAUIigCZu\n4FiaiTw7VGUVj8eRTCbR2traEBO2F0FEQo9Nl1lFiFjREoQg4svulSBSKBSAEjwKRc1QhIMMyBMT\nE1iwYIEx0VYqFSNlFYlEkEwm0dLSYum18QtFVeo1ubPCIhqNolgsIp1OG783M1NTmTwriKzSZbUK\nImBWqMXjcSWIFIp5iBI8CoUP2CUeeF8O/Y7Mx5qmIZVKobW1FfG4v0uOT2k14gRtJ1oAc0FE5ucg\nBFGhUDCiP1YRolgsZpoua8T9rVAoqlGCR6FwiRtfDvXKGR8fd/TlKKoRLYh0XUc0Gp0jOt1GiOwq\nzNTxVSjkRwkehcIBN74cMiCTt6WjoyPQlBUf4aln6qpeeBFEZqZqAKZdqt2Yquk4WxnHrQQRW7Wm\nUCjqixI8CoUJrC/HrF8O68sBgFQqZZSSHz58WE1wdcBJEE1PTxv+HRJEFB3y25QR8CaI2JSaEkQK\nRbgowaNQ/B9syqpcLgOY2y+HIjmVSsWosGIX8qT3iI7AmH1GGKXrjQyJDCsflVl36qAFEb+OWalU\nQiKRqIoOqXXMFAoxKMGjmNe49eXMzMwY/XLS6bSxnpUZIoQHu001AYohEonYNmUE4EsQ2fUKKhQK\nSCQSVX4ks3Gple4VitpRgkcxL7Fa4oF8OWzKKhaLIZlMIpvNBl5KrmgM7PoPAeZNGd0s20HvcVrH\nTAkihaJ2lOBRzBtYX06pVEI+n0dbW1uVL4dKyYFqX44XRKWWVLpKXrw0ZWQN1cCskdrNsh1Ogoh6\nGpmNTQkihUIJHkWTY+fLocgOiRzy5WSzWelKyd2ORYkia+q5b6wEUS6XQyaTqerfVOs6ZixeBBEr\nhpxScQpFI6IEj6LpcOPLoeqro0ePIh6PO/pyvFAv87CamJyRbR/x648B7iJEfLqMF0RemzKyK93T\n+mdk7uYFkVrHTNGoKMGjaBrsfDkAjH45MzMzxv8F3S9HNGxl1nzsxTNfcSuIguhSXS6XDQO3mSAy\nqw5UgkjRCCjBo2honPrlaJqGQqGAYrEIXdeRTCbR3t6OaDSKI0eOCBE7oqq0gNmnbyqNB6qrgmhS\nUt6M+YdTDyIvgkjTNESj0aoGjV4iRLwgom0oQaSoN0rwKBoOM18OUJ2yIlFApeT8Eg90o26EKAlF\npgAgn88jlUohmUxWpTNI7PERLrP0hppo5CDMtKcXQUTnmtuFXb14iKyW7WD/KEGkEIUSPIqGgW6a\ndr4cWjogHo8bjQHNbpgib6JBRHho4mEXHwWAtrY24+mbn2Q0TUMsFkMikbCsCrIywZqJoSD2USMI\nynojw/5hzwW6fhKJhPF7s6aM/LIdTk0Z7QSR0zpmbJRICSKFX5TgUUiNky+H7ZcTiUSQTCbR0tLi\nKlVFwkSWGyY1sisWi0YH3kwmY5ipqVzeDSI8H/zEZbffZNmnCn/wx4/EhhW1dqm2izwBs9d5oVBA\nPB6fMw4+OsRXmKlzUUEowaOQDtaXMz09jUQiYaSjyJczMzNTFf1obW21XDLA6bOCxmuEhzdTp1Ip\nU9EWpDfIS4ojqK7CiuYlLEFEER52u4C7CBFfdq8E0fxDCR6FFFj5csiDA2DOEg+8L8cr9bzRmaWs\n/DQ5FIVbo6pVmTTt20KhYJkym+/I2jNJRNTTThCZnU8kiOjfRLFYtPSl2W3bSRDZGarVudo8KMGj\nqCtOvhxd15HP51GpVBCLxYxoThA3IRHVVHbbdUpZef0Mt58bNHapMuCYeJ2enjZeY9Yzxs5MPV8m\nmfnyPe3wcj7F4/HAmjLStgEYYkgJouZGCR5F6Hjx5WiahkQiISz6IUogsNt1m7JyIixBUyvsJGAm\n5uz8Q7WsSq4IBpl8bUD1+WQW0fVq0PcqiKggwu7ao2IBszScTPtyvqMEjyIU6OZDQgeY2y+HRA4t\n8dDa2op8Po9EIiFE7Ii6EdH3KRQKQlNWsk1MbnHjH3Lj9/Babt+o+0th/2Di1qBfiyCyO6cozU73\nNfZ3fHWZmd9NER5K8CiEwfpyrFJWvC+HX+JBRGNAIuiICZuyItHmN2Vl9xnNjpPfw2+5Pb1PFpQA\n846f/eVWEFmtcu+UgtV1fY6Zmt82LVjM/86uwkwJouBRgkcROFa+HJp02JRVLBYzFuw0Ezci0zhB\nbZtPWdHNq7W1NYBRHkPd/GortwdmGzdaVZapFMQsMgoxkWNyE3F0auFAVaVePGmsWZs9R/mxKUEU\nHErwKALBjS+HRAEA1ykeWX0rdlVW1PwwbGTdV2FiN3nlcjmk02kj5ajK7RuHeoowJ0FEZmp6oPOz\njpmTIHJa6V4JIncowaPwDV3cJGLYi5cmX+qXQymebDbrqZScJicReBUIQVZZ+YEdr7qR+YM3lLK4\nLbf34/VoFNT55Q+rlBbgzqRvF3W0EltuBFGlUjH6mLHps2KxiJaWlgD3QGOgBI/CE2a+nMnJSbS2\nthqrK7OiIB6Pz/HleEGGlJbXKisVaWlM7FJlgLV/yG+5vYypI1mReV85jS0ok77duWUliKgPFvvQ\nqGkaLr/8cjz66KM1fvPGQwkehSvsfDmRSKQqZVVL6bUZ9RAPsjcGVIRPLf4hs4mLfXiQJf0gs7CQ\nlVr3mZ1Jn7bvVxCRoZod39GjR9HR0eF7vI2MEjwKS6x8OfSHSq8rlQqmp6eFiAKRN18+EhNUykpU\nhMfNdlVkqX54fZKnPySqAbW6vRWyCrEwrjenqkXAuks1MOtdi0aj+M53voM//vGPWLhwITRNw1NP\nPYVly5Yhm826Hsu+ffuwdetWaJqG9773vdi2bduc1zzwwAP47Gc/i2g0ijPPPBPf/va3Acw+JJx5\n5pnQdR1Lly7F9773Pa+7omaU4FFU4dQvR9d1FItFFItFo5Q8FoshnU4bK3oHieiUFlVXiIpOhYmM\nE4LiGPzERUuoZDIZ371iVMdfOainoRowjzhqmobp6Wlks1lomobu7m4cOnQIzzzzDJ599lm87W1v\nw4svvoi2tjYsW7YMd999N1772tdafpamadiyZQv279+Pzs5O9PT0YNOmTVi9erXxmsHBQdx22214\n4okn0N7ejrGxMeN32WwWBw4cCPDbe0cJHoWrfjnlctmoPorH40ZjwEgkgqmpKWFjEyV4KNVAYifI\n6FQ9IzyKxoGNWvhJl5FR1co/5KfcXuZIihqXN1iRHIvFsH79eqxfvx579+7F61//enziE5+Apml4\n5ZVX8OKLL+Lkk0+23d7AwAC6u7uxdOlSAMDmzZvR399fJXi++c1v4tprr0V7ezsA4Pjjj68aT71R\ngmceQ0LGzJcDVC/xEIlEkEwmha/izRPktinUS40OqZSzra1N2puWGXRc6Duw+XrZkGlCkGksXnGT\nLmNTZX6MrzIj67GTdVyA9djGxsbwmte8BsDsOXHiiSfixBNPdNze6OgolixZYvzc1dWFgYGBqtc8\n++yzAID169dD0zRs374dGzZsAADMzMzgnHPOQTwex7Zt27Bp0ybf380vSvDMM9z4cqiUnMy6ra2t\niMetTxXZBY9ZyiqbzRpRq6BvWCL2Bx23UqlkLLdBn8NPbNPT05blrWHenGWdCJoRP/4hs3J7em2p\nVFL+IRfILHgA82vw0KFDWLFihedtmd3T+O2Xy2UMDg7i0UcfxfDwMN7whjfgqaeeQnt7O4aHh3Hi\niSfihRdewEUXXYQzzjgDy5cv9zyOWlCCZx7gxpfDL/HQ0tLiul+O6OiCn203S5UViTKKslEHZ/JS\nsRNcqVRCqVRCKpWy9YFYNdYDlEhpVpyMr/SnXC4bUV+ZVrfXdb3hPHX1xkqMHTp0CCeccILn7XV1\ndWF4eNj4eWRkBJ2dnXNec/755yMajWLZsmU49dRTcfDgQZx99tlGFGn58uW44IIL8OSTTyrBowgG\n1pdTKBSgaRqSyeQcXw4ZkGOxmBHN8XrzosiQCNinTqdxmQk3uyorWb02rFjTdR3JZBLt7e1GNMpq\n+2y+3osPxCrtoaqExCJLdIA9tpqmGUUIhJtye6tzphHSZbUgyzE0w07wsN4at/T09GBwcBBDQ0NY\nvHgx+vr6sGfPnqrXXH755dizZw+uvvpqjI2N4eDBg1ixYgWOHj2KlpYWJJNJjI2N4Re/+IVphZdo\nlOBpMsx8OfS0Fo1Gq3w5gPslHuwQndJywipl5fREKJPvhTeGW0XZahmv17SH20U5VZVQc+P1vHHy\nD/kR0rIKC1nHBdh7ePxEeGKxGHbu3Ine3l6jLH3NmjXYvn07enp6sHHjRmzYsAEPP/wwTjvtNMTj\ncXzxi1/EwoUL8cQTT+CDH/wgYrEYNE3DDTfcUGV2DouIww1UjtlAYQvdWMrlctUTF/1dKBSM6AAt\n8ZBKpeY0pPILiY22traat2XGkSNH0NHRUXXDNUtZ0XdyS7lcRi6XC7wJl67rOHLkCBYtWuT4WrZh\nYyQSQSqVQjKZNJ1c8vk8dF1HS0uL8YTNHj9akTmTyQT6fQDzp3z2aR+Ya4otFotIp9OBnWe1MjU1\nhWw2K8VYgNnrRtd1Ie0c/FIqlVCpVKoiPLVgVm7P/80KKl4U0R/q8yVbSpqu22QyWe+hzKFQKCAW\niyGRSFT9/yWXXILHHnusmVOElhe4ivA0KF58OTQxZrNZIes+iY6U0Pa9pqzcblcUVk9Y5L9h1xhz\nMoZ7+UwR+DHFAjDScio61BgEHbFgj20t5fZ0zZB4Vr4zZ8yOJW9Qn28owdNAsL4cq345bMoqFosZ\n0Zx8Pi/sKSSM1FA+n0epVPKUsnJCpIfHDNaATJ4p8lW53S7rleJvaPW8iZmZYnO5HDKZjLGfzTwg\ndE43csm0X2ROh4SFGyGt6zqmp6eN5TjsfGdhnzsym6ntzq/5et4pwdMA0EVeLperRA7bL4ciBsBc\nXw4rjkQgQjjwKSsADVVlxUalWANyUNVijfSU5tcDojoMh49sIow9xolEYs455OfcCdKIL9v+YjEb\nWz6fDyxd2YgowSMpVr4cNppD/XIoLZLNZk1LycNKOdWKVXm8qPy9qP1CQieXy/kq859vmEWHCN4D\nYtdheD5Fh+YbVsLCy7lDD45Blts3muB59dVXfVVoNQtK8EgEm7KidXbMfDm0uGU8Hkc6nXb0sMgu\neNgqK0rDsSkrCmWLIqibFl8BF4vFAkm9Efx+lvlmGxRuPCB2FUL8xFUulwN9wq8FGY+fjCkav9e+\nH/9Qs5TbW+0zvyXpzYISPHXGyZcDmJdde1nckrYj+gbrZfteGgOG7bXxgpUBeWpqCqlUSujkEXYk\nT0bcPuHT9eX0hB/2hCbzpCkbQe8rv6lW1j8EwGgAKlPfKitzMrusxHxECZ46wUZy8vm8IWTo4tM0\nDYVCwShdpeZzftM6NBmKuAjZCJTd9q1SVk6pHpETuZ/9wpvDzSJtIsY8HwVNLbA3/Gg0apTJE3yq\nzGxCs3u6V2IlHOoVCXMS08CsMT8ej88xU9fbe2a1z1SERxEaVr4cAFVlvMViMXDvRz3TWk4pq1q2\nXStets13QG7U5SoUs9g94fvxfzRKusMOGdNsMsLeN8zuz2bpMqfV7dnziD7DL3aCZ+XKlb632+go\nwSMYN74cNvIRj8eNtEiQN55IJGK0jRcBLxyCXsuqXoKH903JEJVSiMeP/8Nrd2ElLtwh635i7Qc8\nbtJlZoLI6vzx6h+y2mdjY2M4//zzvX7VpkEJHgG48eXwBtdIJIL29nZhng/Rxl8SVCRygoxQ1UM8\nsMcnEol49k2JwM1+UCIrHJyiQwCqUmVm6Q72dUE+3deCjOJCxjERfsfl1z/kttxepbTMkcuO3+Do\nul6VvqGIDt3IdF1HoVDA+Pg4JiYmAACtra3GRCra4CpqMqQUXS6XQ6FQQCKRQEdHB1pbWwPp7BxW\nSkvXZ0v9JyYmMDExAV3X0draio6ODqTTaU/Hpx4iTbZJYb5GuWgyorb+yWQS6XQamUwG2WwW2WwW\nmUymanKippS5XA65XA7T09PI5/NGp3S6xubj/pQVkUKMzh+K+KdSqTnnD7s8EKXL8vk8crmcYYso\nFAo4evQo/uM//gO/+MUvcPToUV+m5X379mH16tVYtWoVbrvtNtPXPPDAAzjttNNw+umn4x//8R+N\n/9+9ezdWrVqFU089Fffee6/vfRIEKsJTI1a+HDZszUc9+OUQ2PeKIujJh09ZRSIR46YeNKInTnbV\neLel/vVgvgqIZoO9P8Tj8TlLitilOvin+6CX6ZAxmiLjmID6mqmd0q3UhDYWi2FiYgL79u3D0NAQ\nBgcHsXz5cixbtgzLly/H8uXL8bd/+7d461vfavl5mqZhy5Yt2L9/Pzo7O9HT04NNmzZVLf45ODiI\n2267DU888QTa29sxNjYGYHYdxM997nM4cOAAdF3H2WefjU2bNgW+fqFblODxgRtfDjuJ0jICVr6c\nMCayID7Drspqenpa6NNO0PuHBButLJ9Op+csUFoLYYg0WiONjSAqGgerCdOpOogXRHzvGCsjrOjK\noPmGrEKMFUOJRAInn3wy7r33Xui6jje/+c3Yt28fXnzxRbzwwgt44YUXjDnMioGBAXR3d2Pp0qUA\ngM2bN6O/v79K8Hzzm9/Etddei/b2dgAw0mY/+tGP0Nvbawic3t5e7Nu3D1deeWXg39sNSvC4xI8v\nx61RV3bB46bKSpZKKjvMBBuFjBuh3TpNcOPj41UGdHayA2C0OVBl1M2HV+8HRYjcRodkRFZhITN2\n+6ytrQ2nn346Tj/9dFfbGh0dxZIlS4yfu7q6MDAwUPWaZ599FgCwfv16aJqG7du3Y8OGDXPee9JJ\nJ2F0dNTr1wkMJXgcYEUO330TmI0UkMhhG89RbtUNbLWWyCiJl7SZ1yor0aKtlm2TEGUbN1K0LZfL\nBTjKYwQp0tjxA0Amk0E8Hp/zZKbrumNfkGYro3bLfJk0/UaH6HcAjOvEa2WQKGQ9drKOCzAfW6VS\n8VUla3Yf47ddLpcxODiIRx99FMPDw3jDG96Ap556ytV7w0QJHhNYX06pVEKhUDAmSCtfTi2+D3a7\nok6GaDSKUqlk+xq7lJXTuLwKKi/42SfsMapUKkilUmhra5vjl5AVs54/bW1tyOVySCaTtjcSp74g\ndkswqAU6mxc30SFWMLupDJrP50ijCZ7Dhw9j0aJFnrfV1dWF4eFh4+eRkRF0dnbOec3555+PaDSK\nZcuW4dRTT8XBgwfR1dWFn/70p1XvvfDCCz2PISga4+4fMuVyucqbU6lUAFSnrPw00LNDdITEbvu1\nNgZ02n6tuN02Gw1xa0AWNW4/22Wrdcx6/tB5yL7eSxTRarKjyY1Pg1j5QholFSI7sk6YiURizv/Z\nnSNsdFrEulO6Lt/6XoDc4zI7tw4dOoTjjjvO8/Z6enowODiIoaEhLF68GH19fdizZ0/Vay6//HLs\n2bMHV199NcbGxnDw4EGsWLECK1aswKc//WkjDf/jH/8Yt956a03frxaU4DGBLRGni3t8fByAe1+O\nVyhC0qiNAespeCitSCmfVCrl2oAsctxut8uOPxKx7vkjUpw5VX04VQ2ZCSJlom4evJ4jKoJYf/h9\n+uqrr/oqSY/FYti5cyd6e3uhaRre+973Ys2aNdi+fTt6enqwceNGbNiwAQ8//DBOO+00xONxfPGL\nX8TChQsBADfddBNe//rXIxKJYPv27ViwYEEg388PEYeb0ry8Y1UqFeRyOSMdouu6kQ4RdWFOTk4i\nlUohmUwK2X6lUsHExASy2WxVyiqVSgXyvagHBLn0g0TXdRw5cqQqHGuWfvPzXfL5PHRdR0tLS6Bj\ndtouP37qtWHn/SLhvXDhwqoJhX391NQUstlsaBOIlS+E/g3MXY+qHhOdpmnI5/PIZrOhfJ4bcrmc\n0Y9HBkTtIz46xJ8nTv4y6u0lWzo6n89LOS6r4/i9730Pf/nLX/BP//RPdRpZaFjeVOQ6UpJA/g9K\nhxw9etSTCdkPIp+IKU2i67OND4NMxRFhPNHTjZKiIU7l/m4Q5T2y2q5ZB2e342f3sVOKLiwhYZcq\ny+VySKVSxr6glKNTGkSlypoLN9EhO38ZAMNTaZYqU2bqaqzGNTY2hhNOOKEOI5IHJXhMiMViaGtr\nM36ORqPQNE14J+QgJ16zlBUwW5Io4iINQ/BMTEwEkn5jCUuo8dGc1tZW6Z4Mg4Ymo1gs5ikN4lRC\nXe/lF4JC1gkzbJz8ZdPT00bkm0SzDIu4ynr8rMZ16NChqt4585HmvuP6hD9ZwpgUg2gc51RlVSgU\nhF2kQe8j3sALAOl0GslkUsqbDA8J2Onp6cCiUYSsN1qvkBgywywFYre4oixP/Y2MjOcVGxmyM9zX\nsohrs2EnePyYlpsJJXgsYCdwivCI/jy/n+G2ykpkN96gBI+ZAbmlpQUTExNCPFQihFqxWEQ+nzdM\n6EFGo+YLTk/9wNzFOc2e+tmnfKvqFUXj4uc8sTPcu40iynwu2aW0/uqv/qoOI5IHJXhcEEaEx6vg\n8VNlFVb6xutNwCzlk81mqwSOyAqlILZL0Sgqh08mkyiXy4GboZ2YD9VR7FO/W08IlfTncjnLVFmY\nRmoZj1GjTeBOeDlPvEYR+c+QCav9dfjwYV9VWs2EEjwWhB3hcRN9cUpZOSFyMqSJwsvNie0gLMN6\nY17Rdd0Yv67rSCaThuikzttBE7YpuRExe+qnypWWlpY5gshuLSrRKRB1HO0Rec3bRYfos+3aMQCz\nlVpWRup6mqnNvlOhUBCyuHMjoQSPC0RV8vCfYXVxB9EY0OkzgsDN9kkkkCCo93pjXrdL0YJCoYBS\nqYR4PI5MJjOnuWFYUUHZRCCPLONjTdBOFUN+eg41w/IcshwrM+qxb+08ZhTRTSQSVeJZhkVczR6I\nZD62YaIEjwXsCeNmWYYgPo8VVUE3BqTPqIfg4Q3Ibjogh4XbfWK21INZc0BFY+PWSG1VPi1Dz6Fa\nkG2MskYz6XhaVVrywtnrIq61fGerezCNez6jBI8LwniSppQWv0aXl5SVE2ELHisDsh+RUI9oBi/U\nvK4tVq+IlCxPc812c3UyyPKCiO85ZBYNYrsQKxoHJyHmVjjz0SGrc8VrdIh/zdTUFFpbW719ySZE\nCR4L+AiPyJQWTarAbE44lUoF3hgQEP89KErFijY/q8dbbTssAeF2qQcZURNnffCaKqPrkDpy21UL\nhXVMZYymyDgmoLZxefUOUSTR7SKuZmNTJemzKMFjgWg/Bp8iocZabW1twiZWkV6kcrlsLMkRj8cD\n6zlDiBY8ZpVitQg1WTxHCjngn/gpApTNZl1XC1ktyimjIAgKWc91kULMTXTILpIIADMzM4hGo/j5\nz3+OXC6HSqVirG3llX379mHr1q3GOlrbtm2r+v3u3btx/fXXo6urCwCwZcsWvOc97wEwK/7PPPNM\n6LqOpUuX4nvf+56vMQSFEjwuYCfFWnOrdlVWpOJFCp4gbyC8zygajSKdTgspxRY10VMl1dGjRxGN\nRgMXaqKgCjcKf7O9QxSNhVOqDPDWc8ivkVrWaIqMY6oXTpFEanaaSCSgaRqef/55/PCHP8QLL7yA\noaEh/Nd//ZexivmKFStw/fXXV60qYLa9LVu2YP/+/ejs7ERPTw82bdo0p2Pz5s2bsWPHjjnvz2az\nOHDgQI3fOjiU4LGAj/BYhQrd4LbKSoYqKifMDMhUpUQlmiIIMjrFC08AaG1tRSKRCGT7gJhjSRNc\nLpeDpmnGeMvlclU0gF93qNGMs/MBt/cSdnLzsg5VsyzPIasIk3VcAKrM1O973/vwvve9D/fffz80\nTcNb3/qX27gRAAAgAElEQVRWPP/888Yfp3vewMAAuru7sXTpUgCzwqa/v3+O4LG618kWoVOCxyVe\n19MyS1nVuzFgLdsnX4u+bx8q/+9+RCNRLPzYe4ENGwLZfhhY9f2hxWGDJqh9wY5b13Wk02lkMhlU\nKhVUKhXjxqvrs4vD0oTGhrvna5v9ZseNkdrMHGt2PlQqFeO8UeeDPVa9buqNlRAbGxvD6tWrsXjx\nYixevBh//dd/7Wp7o6OjWLJkifFzV1cXBgYG5rzuu9/9Lh577DGsWrUKd9xxh5HempmZwTnnnIN4\nPI5t27Zh06ZNPr9ZMCjBYwF/0ritjqmlMWAYpmKvPWfY75N9/HFo79mKrfnbAAA7/r8PIPPAv0F/\n05uM7Ysav18xxVa+2fX9CVqo1TpZmHmK2trajHA1K3Lo3+zkZ/bkZuYTcUqNKDHUuLg1x7LnQKVS\nsTRS16PnkKyRlEYb16FDh3ytlG52X+S3f9lll+Ed73gHEokE7rrrLrzrXe/C/v37AQDDw8M48cQT\n8cILL+Ciiy7CGWecgeXLl3seR1AowWMDO8naiZFmagwIVC+TwEZCCv/vfnw0fxvuxbtmX5gHvvqV\nbyHDCB5R4/e6bX6pB7u+PyJvXF5vjBTNKRaLiEQiSKfTgXmK3Jgh7VIj9Z78FMFC5wOdE7T2GzXT\nMzsf6P+cKoXUOVEf7ASPn2Ulurq6MDw8bPw8MjKCzs7OqtewZuj3v//9VabmE088EQCwfPlyXHDB\nBXjyySeV4GkE+AnXT8rK62cEjZ1oC+L71Fvw6Hr1Ug/17OLs1SxqViFm1tRMZPWXl9SI1eTHToBs\n1Ui9ke2JXLbxANbRQrPXOVUKBdVlWMb9BDTeuPwKnp6eHgwODmJoaAiLFy9GX18f9uzZU/Wal19+\n2RA2/f39WLt2LYDZQpCWlhYkk0mMjY3hF7/4xZwKr7BRgscGPsJTqVSENgYMo6MzcOyi4A3IiUTC\ndJkEomXre7HjF+8D8rM/78hsQ2brv4GmtHoIHjsTtdcba9A4Gd35fj9BRnOCxOvkx/pE6LioSEDz\n4FQpZOYb4rsMu40WNppXpt5YjWt8fNxXWXosFsPOnTvR29trlKWvWbMG27dvR09PDzZu3IgdO3Zg\n7969SCQSWLRoEe655x4AwDPPPIMPfvCDiMVi0DQNN9xwwxyzc9hEHG70cjye1Qm6aZfLZUxPT6Nc\nLhsrYSeTycAvRBJTdmWCtXLkyBG0tbUZEYVIZLaxntvvE/nxjzH9lW8BmBVA5N8BZtNIuVwOHR0d\ngY+7UqlgcnISCxYsAGC+1EMqlfJ1TCYmJgyRFCRHjhxBR0dH1ZjMojmpVMqyRT3P1NQUEokEUqmU\n8VTNbp+OKfV1qifkN6LIoplnJCwTdblcRqlUkmbxRNnGA8wer1QqJcTAT1gZqdkqQ/ZcqFQqRppN\nJoE8NTWFbDYrzXiIQqFg7C9C13W8+c1vxuOPPy7deAVh+SVVhMeGYrGIXC4HXdcRi8UQj8fR3t4u\n7PNERkjIvKvrOiYnJ3031tPf9CbDs8OPVHRKjhULXpd6sENkmoi2a9a9WcZoTpDQBGVXUu1moU5l\nom4evEQL6d/lctlovSDDOSFLmtYMu8iTumaU4LElGo0aE2q5XEY+nxf6eSKqnHgDciQSCbznDCFK\nOBgl8bpuPIUGudSDqHGTQKPooJ03Zz4yn03UsqZE6olZqqxSqSCRSCAej88RyGbnhJUgou0HPV7Z\nMDuvKNWvUILHlmQyaTSmE10yTp8RxMRrZ0CemJgIYKTmBCkczEr8AaC9vV3KGw0LK9AKhULglVYy\nP2EGRVAmat5ArYSGNTLuG95I7WdBzqB7UMm4nwizsR0+fFito/V/KMFjA3vihDHR0Gf4uaDMDMhm\n6R4RUSSzsfi9IfBl2Wzq5/DhwwGPdJYgji3t/0KhYERzotFo6BGd+SCI/FQQUbosl8up5osNhNt7\nidM5AQS3PEejCZ6xsTEcf/zxdRqRXCjB45JaxIiXzwC8XVBs5RgJBLt0T1BRJDPopuN1H1E0p1Ao\noFKpWKZ+/Gzb7bj97hM7b87ExISwVFkjUA/xZVVBRCbhdDrteeKbL2JI5om8FtjzodblOdh7tGwd\nqa2utUOHDinB838owWMDHxkRNeGyuBEkZECm7sFevCGiJyEv2zdb6iGZTNqa7mSY7M2iObWsrO4F\n/pxUeMOLibqWcmonmlVcBE0Y+8lPxJD8hIA8y7XQvdEswuOnB08zogSPBygdJLIvhN2kbtYB2U4g\neN1+EDhtnxdrbpsDisTtPjGL5th11JZFoCncUauJ2i46JDsynqcyjMksYlgqlVCpVJBOp11XGoZl\nrjfb5tjYmNEYcL6jBI8N/MkjMh3EfibrsTHrN1OLQAjLw8PjZakHK0QJCLt9YuaNCiuaY4YSUf6o\nNVLg1UTNTnwA5kx29P+yRXpkGgsh25hqMVKz5npg7nlRS1NOq3Pp0KFDOP300z1+y+ZECR4H2Akm\nDLFAN0MR/WaAcCM8uu5vqQc32xaNWTTHayl8vTwsos9RRTW1mKjL5bI0KRHZkE0MEkEZqc3OC3Z5\nDq9+MjvBozw8syjB4wHRER66EVL33KD7zQDhTMK0mGotSz2YITLCQzcZPpqTzWYDWzokKFSEp3Gw\nMlGTkE4kElKYqGUVF82K1XlBuPGT8eeD1YOO35XSmxH5FimRDN4kGvTTM0VBJiYmjB45iUQCHR0d\nSKfTgfuFRPUT0jTNqLIiE3JHRwfa2to8+4ysECnWyuUyxsfHMT09jXg8jo6ODqNBo9+xixgvbbNY\nLGJqagr5fB75fN5IF1LkQImixoBSIolEAslkEul0Gi0tLchms8hms1WpX4r85vN55HI55HI55PN5\nFAoFFItFlMtl4/g3A7KKsLCM1NTdn86LTCZjnBeZTAbJZBKxWMx4UKPjPzU1heHhYbzzne/EjTfe\niCNHjuDpp5/G0NAQKpWK6zHs27cPq1evxqpVq3DbbbfN+f3u3btxwgknYN26dVi3bh3uvvvuqt+t\nWrUKp556Ku69995A9kkQqAiPB2gB0VqhJznW00IGZDLziiLISdgsIhKPxxGLxYSsESRq7NTzhyrd\ngn5yDgqa8Mg0mUwmkUgkqp4G2X4zYUUHFGII00Qtq7iQkXrvK6tU2czMDIDZhrmVSgV/93d/hxde\neAGvvPIKbrnlFjz33HMYGxvD0qVL8ZGPfATXXnut5WdomoYtW7Zg//796OzsRE9PDzZt2jRn8c/N\nmzdjx44dVf935MgRfO5zn8OBAweg6zrOPvtsbNq0Scgai15RgseBICM8bgzI9a6icoOdv4VKNWWF\njkGhUAAApNNpJJNJFAqFwJfbCOKmyItKeupra2szIj3s+ROLxYx+M04TIj8RKjHUOARtomY9JbIc\nf5nGwiLzuKiYoqOjA5s3b4au6/jZz36Gn/70p4hEIsjn83jhhReQTqdttzUwMIDu7m4sXboUwKyw\n6e/vnyN4zOaSH/3oR+jt7TUETm9vL/bt24crr7zS1feIxWI488wzUSqVsGLFCtx3331ob2/H7373\nO3z4wx/G5OQkYrEYPvWpT+GKK65wtU1CCR4H2BPbj4eHJhp2iQQ7A3JYgsfrRct/D6veMyJNs373\njZM3hxYmlGW8gLnhu6WlxYjwsFUiVp/tZkJsxnWqGgFd14W1t6jFRC1TJ2pZhYWsmO0vtuAGADKZ\nDNauXeu4rdHRUSxZssT4uaurCwMDA3Ne993vfhePPfYYVq1ahS9/+cs46aST5ryX/s8t2WwWBw4c\nAAC8+93vxte+9jXccMMNaGlpwX333YdTTjkFL730Es4++2xccsklnhb0VoLHA14mMLapHhmQ7Xq2\nEKLX7PJ6A+G/h9O6UCJvUF4FhFlELWgTeNDw5ftBVeexeI0OKDHUPFiZZflO1GxkSHWirkZWIWY2\nromJCU+CgN0WD7/tyy67DO94xzuQSCRw11134eqrr8b+/ftdvdct559/Pv7whz8AALq7u43/X7x4\nMU444QS8+uqrSvAECR/hsRMjlGJgOyC3tbV5WkspjCoq+gyrk5Ce+NmlHtx+D5HjdxM9MovmOIkG\nkdVfbsQrf96k02l0dHSYCpIwIoBexRD9m/eN0P/JOkEo5lKrbyhIMSzredNI4/Jbkt7V1YXh4WHj\n55GREXR2dla9ZuHChca/3//+9+OTn/yk8d6f/vSnVe+98MILXX823d8qlQr279+P973vfXNeMzAw\ngFKphFNOOcX1dgEleDxhlg4yMyD7aarHf4ZIrD7D61IPXrYdBHbbriWaE8Y+N4Pf37WcN2HgVQyR\n+KQUnBcTbbMj26TpZjxBimH237TtRkDmCjizY+h34dCenh4MDg5iaGgIixcvRl9fH/bs2VP1mpdf\nftno4Nzf3481a9YAADZs2IBPf/rTGB8fh6Zp+PGPf4xbb73V9Wfn83msW7cOIyMjWLt2Ld70pjdV\n/f6ll17C1Vdfjfvuu8/z91KCxwHen8KKHn6CtXoq9/p5os2D9PRNJY1BLvUQpoeHxGahUKi5QWNY\nHh7eC+V1fzuJs3qJN7PJkEyU8Xjc1DfiZjIM6hqQTWA0G159Q2yqFDDvOEznhozINi6ra97vOlqx\nWAw7d+5Eb28vNE3De9/7XqxZswbbt29HT08PNm7ciB07dmDv3r1IJBJYtGgR7rnnHgCzkZ+bbroJ\nr3/96xGJRLB9+3YsWLDA9We3tLTgwIEDKBQK2LBhA3bu3InrrrsOADA5OYmNGzfilltuQU9Pj+fv\nFXG4OcorZ0OCBAH9++jRo4jFYqhUKkgkEkilUoF7LI4cORKIeLJicnIS8XjciIpQWXwQ0YVyuYxc\nLiekBJGEWTabnSM2U6mU7/2l6zqOHDmCRYsWBTpeSqu1trbOqWyj6jCv+5v6sLS3txvnJrsNihq1\ntLQE+l38UCgUjB4zVphNhuy/gWOTYS1iqFgsGueKDOTzeaONgwzUa//Q/GN27NnjL4OJmsaZz+eR\nzWZD/VwnrMa1e/duJJNJfOADH6jTyLzT1taGyclJAMBvf/tbbNq0Cc8//zw0TcMll1yCTZs24SMf\n+YjdJixPCjmuNomJRCJVqQd6am1tbRVaZSHiKZ0qf6hJVTqdDnzhTpFj1zTNaBAY9HIb9BlB30Ar\nlQomJyerKttkmeTCwOlcsDLR0nv9RAYaLU0yn2GPPX/8C4UCotGoEYm2W5wzLBO1rJFCq3EdPny4\n4dbRYr/H6173Orzuda9DX18fAODxxx/HkSNH8O///u+IRCK45557cMYZZ7je9vy58/qESjWp9wk9\nmYms9KG0UBBChDfxxuNx44/szQGB6vJsmuRERb+CuplR5CyfzxtPzXaVbV6oV8rKD0EIUTsxBKBK\nCNGyLGZiiP6PTaHVE9kmTtnGAxwr3ZfFRE2fKdt+AqzH5TelVU9oxQGiv7/f+PdVV11V07aV4HEg\nEolUlb2JLhunz6h1UrMz8YpsDhiUB8msPDsSiWBqakqI2AniJlYul6v8ROl0GsVi0bHJV5A0kiCq\nBbvIAIA5EyH9oXPfLiog44SmmEtQJmov0UGZBY8Zhw8fbjjBIxIleFzAipx6VlE54bYkW3QllV/Y\naI6ZEZxuXiKoZZ9T52ZeWFKTQEX4RCKROWKIjg8vhvhVqmXyjMxnahEXXk3UZtFBs3NA5ocJqwiP\nWjj0GErweCQaDWY9LTu8VjrZLfVgtf0w1utye7PiozlWq6uLFmpetk3VYXZjFjHe+RLBEYmZGCLM\nxBAbJWpWMUTfbT5glyoFnM8BYNZ0LlPzRav77eTkpK/Gg82KEjwuCCs6Qrh5kuDLm62WevC7/Vpw\ns4+cojlO763HjSXoEn6FfLgRQ6yB1q8YkjU1IhP12kd25wCtwUdVrvx5IMI35AazfUX34PkiZN2g\nBI8L2BMpDA+PXYTH61IPVtuvl+BxG82x2q4o7MbstyFjGBEeNWmGB02EZpiJIatqInrgkMVATeOX\nYRwsMo4JgNFbiseNidquqqwW7CJ0Mu7DeqEEj0fC8vCwgocuHD9LPVhtP0zBY+Zz8Vtp5TVd5nW7\nhFkErRGiOSrlVR+8iiEARnTTbhKcr5NVI57DXk3UTivYezHSm90Ti8WibQ+s+YgSPC4IO8JDT4BB\nLPVgRliCp5ZojtO2RWHmh/JbUh7GWIvFYtVNUmFOPSdQMzGUy+WQyWSMc4Qvrw+7z4ys0RTZxuR3\nP4kyUdN2zcY1NjaG4447zt8XbVKU4PEInVyibhAUWSiXy5iYmBDiExE5EdNFS+W/QY9fxNjpeObz\neWiahkQi4doPZYeo/azrutHMkLp+sxMkcKxpm4oWHCPI7x975BHM3HkPACB13btRufhiT+9nUxxu\nogJ2fWZUZCg8REWXazFR032mVCqhUqngxRdfBDBrWPZTkr5v3z5s3brVWFJi27Ztpq978MEHccUV\nV+DXv/411q1bh6GhIaxZswarV68GAJx33nnYtWuX588XiRI8LuBNyyLSKmw0hCbaBQsWCLlxiRBt\n7PgjkQji8Tiy2azw1FMtsMZpTdOM5pIyRkporGSaTCQSyGazRtk7GymgChI1QYoh9sgjKF31YWzN\n3wYA2PHLDyNx/9c9ix4nghBDYZtng0LWiFM9xuXGSE/XvKZp6O/vx7333ouXXnoJ2WwWg4ODWLly\nJVauXIlTTjkFGzdutEx1aZqGLVu2YP/+/ejs7ERPTw82bdpkiBhiamoKd955J84777yq/1+5ciUO\nHDgQ3JcPGCV4fBDUpGtWqdTe3o5oNIojR44Iu7CCEm1WVUuUDpLxhgWYNzWcmZlBPB6XroMzW/5O\nfZWmpqaQSqWqjiOAqsmRTx2qaEFwzNx5D7bmb8O9eNfsf+SBr9x5D+IBCx47ghZDotP0XpFV8MiG\n2TV//fXX4/rrr8eePXswPDyM9evX47nnnsPg4CAee+wxXHrppZbbGxgYQHd3N5YuXQoA2Lx5M/r7\n++cInptuugnbtm3DF77whar/l917pQSPC/gLj24QftI0rLeFlnrgvS100oi86GsRbXwPmnQ6XTV+\nkSkzv9vmxVk6na4yTlPjuaDH6gfepO7F5G31mV4mSDsfSTP1nqkXYRQ9eDHPUpQwn89XvbcRI0Mi\nEWllqAU2Pcpy9OhRnHbaabjkkktcb2t0dBRLliwxfu7q6sLAwEDVa377299iZGQEb3nLW+YInhdf\nfBFnn3022tvbcfPNN2P9+vU+vpE4lOBxCf8k7fWmZbfUg9lniUib8Z/h5TtYRXPMRF8k4q1xohe8\njps3fvPizO923eLlOJJhmlYZD8qk7maMbsWQU7n1fJgcU9e9Gzt++WEgP/vzjsw2JK77Ory28qzH\nPrI61lNTU2hpaQGAOWKI/s2LIf7fQSKjsCBkG5fVvjp06BDOOussz9vi4SPFH/vYx7B79+4571m8\neDGGh4excOFCHDhwAJdffjmefvpptLa2ehqDSJTg8YHbEDAfzfGywjd9hig/idsJ3imaU8u2/eBm\n22YREllLyvlzxE3LAdFimP8sOzFk14iPHavXMluZqVx8MRL3fx1f+T/TcuK64P079cBrZIgVv2GK\noXohqwizEzxeTctdXV0YHh42fh4ZGUFnZ6fx8+TkJJ566ilccMEF0HUdL7/8MjZt2oS9e/di3bp1\nSCaTAIB169bhlFNOwbPPPot169b5/GbBowSPS7xEeLwu9eD0eSKw276XaI7XbdeK3bbZ/R6NRj1F\nSERFpazGy/u30um053Ok3pCZ0gz6fvQ6p/WqGk0MVS6+2PDsiF1oRjxurlU3YsgqEgjANApoJ4Zk\nFBey+lOCFDw9PT0YHBzE0NAQFi9ejL6+PuzZs8f4fXt7O1555RXj5wsvvBB33HEHzjrrLIyNjWHR\nokWIRqN4/vnnMTg4iBUrVvj/YgJQgscl7AkVjc5dT4uiCn6WerD6vLAFj9v0j59tBwW/bYqQFAqF\nqv3utSmj6P1N8BEztxE/FqexhhkBshsDfT499RFmZbZ8zxEzQUTbVYjB775ljzUvgM3EEKXK7I63\nbCZqQsbzz+paP3z4MI4//nhP24rFYti5cyd6e3uNsvQ1a9Zg+/bt6OnpwcaNG6tez96LHn30UXzm\nM59BIpFALBbDXXfdhQULFvj/YgKIONzk5ZS0daBcLld1SC2VSmhtbTVd6iEI30UulzNEhwhyuZwx\nXj6ak0qlakr/lMtl5HI5dHR0BDjiWfL5vBERoe7NFEWjyiU/sMc0SMbHx9HS0gJd16tSbLXs46NH\nj6KtrQ2xWMyYONjvTU3t6h0tKhaLhl/NLfzEyP4MWIshp+NO0VZefNUDKiPOZrP1HgqA2X2ey+VC\n91qYiSE3x7tekcByuYxSqYRMJhPq5zphdW5fcsklePzxx6UUaYKx/MIqwuMS9qSJRCJGY8Aglnqw\n+jzREQcSDLVEc8wQNXbWSDkzM2P0ovEaITFDxJjpxj01NRWoGG5mKE1mFikAUCWGrCJDzeohEUW9\nooF2kSF6kOEX6XSKDIkUQ/WOmlphNi5Z02/1RgkeD7DRHF3XkclkhE1gIiZg1ptDXXpFmHmDHjvr\ndyEjN/UrCoogx0wpNir3zWQygUbqghprrd2Cw4SdGN2kTXhDLfmGgOo2/TJOYIpjYsip4Z5dWjRo\nj1gjCh4Zx1tPlOBxSbFYNJZ6aG1tNZq/iSIajRoTZq1YrclFoidoaIKp9QbB+10ymYwh2kQ1CKzl\nvfwCqdQkMOzqMDeCKKxuwWHgxkNCYX/y2llVF4UhhmSbOGUbD2C/+jfgrvswL4asDPPs8ZZtP7iF\nH/fRo0el88/IgBI8Lkkmk8ZSD+wTpagLpNaneKfSbPJXiKCWfeJUISaiQSDgf8ysmDRrIhmWGdor\nMnQLDgP2qT4Wi1W11DcrtXbqSNzoE2OjUMu91Y0YsmqlYCeGZBSGgPm+OnTokGfD8nxACR6XsE98\ndMOTUfC4XWE9rCowt/un3uP2sl2+Ii/sPj+yiqhGg43u8MwnMTSfziUSQ2Y4iSFgdh7ghVG9j7nZ\nffbVV19VgscEJXh8EoZgcFua6RTNsdp+2GXvPGal/PVsEOg0XuqWTYbKdDqN1tZW25ud6P3st4Q3\nqG7BzYpbMdQs65LJNq56RFOcxFA+nzfajNh1HA/zmFtlGvz04JkPKMHjErNIg9/1tNxATxJ2uI2K\nmOFFUPnBbqLnGzO6EQ5utlsLdp/NLjbqtTJMxHjZbfrddrN2Cw6DWsUQHT/yoskuhhTVJmq+GpeP\nDNmtRSdKDPHbGRsbU4LHBCV4PMBONG4ESRCfxav3oKIiYY2f0PW5y2zI1CDQbLzkJdI0zdMCnmFC\nN1e2waXbfdRM3YJlwY0YKpVKKJfL0izSKqM3pZHG5BQZEhkNtBrToUOHjBXPFcdQgscD7EQSRoQE\nOHZCm0Vz3EZF7BB1Y2EFW5BLKIhOEQXVbZoQMV4Sj5TCpP+jG6amaSiXywCgogcSwUYJNE2ralVQ\ni5m2GY9tIwkeO4JIjdr5xKzGdPjwYZxwwgn+vmgTowSPT0RHSIDZi6VYLBpPhEE2OBRtvNb12c7C\nuVzO9xIKTtsPclskECYmJuruJbKCUoEUHaBuypVKxTiW9Br+5tnoa1c1O7WYaWs9tjKKCxkJej8F\nIYboteVyGZFIBDMzM2hpacHY2JgyLZugBI8H2JNdZISHogwUHfHicfFC0NEHSgNR5CGRSASeBgpS\nqLHRJyJooRPEPmb7ESUSCcTjccOvRWIYqG62FovFjFbzZhOm6lBcP7yeu27EUDOtSyZj1VjYY3Ir\nhujaL5VKmJycxGtf+1osWrQIra2t+Jd/+ResXbsWK1euRHd3N0455RTH5Uz27duHrVu3Gutobdu2\nzfR1Dz74IK644gr8+te/NlZD//znP4+7774b8XgcX/3qV9Hb21vjXggeJXh8Qk/WQWHmzYnFYkY3\nZxEEJXjMetGUSqWqm6tMWC3gOT4+Xu+hGfCVd+l0Gh0dHYhEIsjlcigWi0bjPPLvUBVZpVKZ07iS\nbqDxeLzqmLATJr+wI3vTDaspnyiaNYpBx99vN2JKgVLaXKZjK8s4WGQYE3td0rFLpVLIZDL485//\njJGREVxzzTU499xz8fzzz+OJJ57AwYMH8fa3vx3//M//bLldTdOwZcsW7N+/H52dnejp6cGmTZuw\nevXqqtdNTU3hzjvvxHnnnWf83zPPPIMHHngAzzzzDEZGRnDxxRfj4MGDUuwvFiV4PCAiwmNXsTQ1\nNVX30nErnErhqWxTBH7GbTbeMEzIXs8Ts/OBBC9NXrFYDOVy2RA97L6IRqNIJpNG+pCeBmnyA1Al\n1FlfSSwWqwqT85GhRi+9nm+4FUN0TMmgD9Q36iejOJVxTMDcccXjcSxduhSapuHDH/6wpzEPDAyg\nu7vbMDtv3rwZ/f39cwTPTTfdhG3btuELX/iC8X/9/f3YvHkz4vE4li1bhu7ubgwMDODcc8+t8RsG\nixI8PqnFw2MWzTGrWBJt0PXTw4UmZFp01K5BoKiUn5f9wgqIaDRqW7oven/bwZe+0/nATkw0xmQy\niUQigVKpZER0ksmkcU7SdyZvj1WExq0YohSZmRiy8xfQviyXy0oMSQYrhihKkEwm55wTTlE/EWJI\nRnEh45gA+yU4vI53dHQUS5YsMX7u6urCwMBA1Wt++9vfYmRkBG95y1uqBM/o6CjOP/984+eTTjoJ\no6Ojnj4/DJTg8UCtER6v/WdET8But89WBbk1T4scu9O2+RJ4K0HpdbtBj9UubUXfga0KpHOOhBEJ\nOCszOD9xVSoVw/BKkxcrZnihYiWGWKMs7VMzMUSTJC1jQu+zMtnOF2SbPNlJkxWlfhZpbZYUKI+M\nviLA/FwqFAq+1nk0+47stnVdx8c+9jHs3r3b83tlQQken9DE4HTzchvNsfqMIH1CPE4TPC/QvJTC\n1/WWLVgAACAASURBVEPwsCZkXa+9BF4UVlEn+h3fpI78FWwPo2w262iudmN8ZAUQO4GxkxYbnaFx\nsUIIMBdD9DeVX7v1lCjztJy4EUN8CpQVQ3al9fy2ZDzmMo7JbF/5rdDq6urC8PCw8fPIyAg6OzuN\nnycnJ/HUU0/hggsugK7rePnll3HZZZdh7969ju+VBSV4PMBHeOyoRSywnxF2hIcmokKhUDW5ei0p\nD1PwWJmQvd6gREd4vKStSFBT2oqaH6bT6UAEnJMY4kUQK1TYqBAvhtjqEd48zfqFgOqoEABT4WUV\nOVApsuAIQmC4EddOfjD2vHfzMBkmMo2FJUjB09PTg8HBQQwNDWHx4sXo6+vDnj17jN+3t7fjlVde\nMX6+8MILcccdd+Css85COp3GVVddhY9//OMYHR3F4OAgzjnnHP9fTBBK8NQAeWDoBl5LNMeMMAQP\nTbJ8dCSVSkndIFDXj3VCDmoBTxFjphv8xMREVcfmoNJWInBjdiURxHYMpvHRTTgejxtVYWZ+Ieod\nwn53ZZ5uPtyKITbyp2kacrmcNMdXZsHD43dZiVgshp07d6K3t9coS1+zZg22b9+Onp4ebNy4ser1\n7P1y7dq1uOKKK7B27VokEgns2rVLyv0VcbjBy5m4rCNsz5bx8XFks1lEo9E50ZxUKlXzAS+Xy8jl\ncujo6Kh12KawkypFR1KpVE2dhYlKpYLJyUksWLAgoNHOomkaJicnoWkaotGoUcUUxMU1NTWFRCLh\nK/9tNk4yd+u6jtbWViQSCeN3/BOsWdoqlUpJ1/zQDE3TjIggK5TYiI1bv5BZxJGf4KzM06wo4lNr\n5XLZ6GFUbzFE9wlR7Sa8Mj09bVT2yQAb0eTFEJ/yDUsMyXbMgNnzP5fLIZvNVn3fvr4+zMzMYMuW\nLXUcXV2xPPhynOENBBu6B4BcLgdN02qO5th9VtBQdCSfz0PTZlvcy9hwj4VM06VSybjxODXR8koQ\nY2bHmUgkkMlkMDMzg0QiYZm2olQXnUdtbW3S+Y546EmcqsESiYTluP36hQh2v7EikTdPm/UXYiNK\npVLJ8JUo83Q1sn1v1r/mJjLERh15McT7hWT7rrXCf59Dhw5h+fLldRqN3CjB4xFd15HP543UTyKR\nQHt7u5CLKGjRYLYeF7UiD5ogxk7CjKIklGajfR80fsfMptfoybSlpQWRSMRIv+RyOSPCEYvFjBRi\nvdJWfmF9RbquI5lMGt/VCpF+IQCmVWTsBBePxw1Rxnum6DOtzNPNVGnUKLhJH7k5p9j0q1Xkz60Y\novfJhNV+OnTokJT+GRlQgscjuVzOSFFQlY2oGyF7Y/f7GWa+IormkAASAe/n8IJZ52Y2zcZ6j+oJ\n35OIFhql39EYU6mU4XehcD1wzCtDUUFK08k4sVJ5edACrRa/kJUo4f1C9B6KqDn5hdj3OVUaNUPU\nQFZ/il9YYzyPmRhiU8xWx1fTNGlSfoTVcVPraFkj1xFsANra2qpC5KJNxXQD93pDctPzJwxTtNux\nU2qH7fVjlWYTNW6322XTVpSC4n0rtD2aUCltRVGRRCJRNbFSVROf4jGLboRJpVIxhI7bcvigsBJD\nZukMEkOsMAGORX9YjwpvnvbaeZomSCvztFPUQMZogUyI3D9uxBDfNoGOPT00mvmF6iEYre6taqV0\na5Tg8Qh7grEltyI/z+3kzvpByENi5ysKIoJkh5uxU+SgUCgYhm8RC6W6wS5yZJVeo+/ILqVBN0D6\nbhQJJKFj9934qAQ7keu6XiWAeBNwUNB5xHZxlslXZJfOYAUa7R8ARhrOi1+Itgc4+4X442aXQmEj\nSTJEVmQZByHyIcwOu2hjLperWozX7SKtIverneA57rjjhH1uI6MEj0fCjJDQZ7Cl72awE6uuu2+4\nJ/omZ7d/+J40Xnr9hLHfCT5tRek1+h37hE9/+Gor6gvkBvamS5/DjoX9Q6lKsyooPzddNv0JoErU\nyQ67z6mAwCwyFIZfyMw8zUYN6A8vwPjIwXxGxu/PRvxYvHjCgjzGVoJHxvSbLKi9UgPRqPe1qPx8\nhlvR4Kfhnt+UmZdtE6y5l11KwWvkIIyUlpe0VVjVVnSj5DGrgvKSImP9OSTqaAV2mfEaiRLtF6Jz\nxypFBswew3w+j0QiYaxjxYtYErB2nxvkPpQN2SJOgP2YnFKvvCAKSgyZjUnG4ykTSvB4pB4RHjPR\nwKZWaln1m0SbiHQFjV3TZvu0UHUYmXv93tRE7vdKpYLx8XHXaSvK7XtJWwWNXYrHKUVG3yEWixlj\nlz3CICISVatfiO8vxL6X9QvRe8vl8pzPNfML0d+imy3KfLzrjd97DXs8RKxLZuZ1krnwQQaU4KkB\nEgsin0joCdypcqmW7YsQD3QxU68fOxOyV4IeMyvIACCbzbpKW7FmXi9pqzDhU2RsJIpKtSlCxYbi\ng0iRBQ0fiUqn08JL+Z3EJB9ZY//w+4z2bTweN46FmXma9/2IMk/LiowRHiBYUehGDPGClxdD7MPX\n5OQkUqkUpqensXDhwsDG2WzId4eWHD7CIxK2OoBWwJW9QSAb8aCbe9B9ioIYM2/wTiaTyGQyhoCp\nV9pKFGw60SkSVWuKLGjqWSlmh5MYoogan6aiKI+VX4jeb2aeZj/XqdmiWck1HxWi7TUDsUcewcyd\n9wAAUte9G5WLLw5ku2ELMDcim7yFJHi//e1v43Of+xyOO+44JBIJfOhDH0J3d7fxZ9WqVbbXzL59\n+7B161Zo2uySEtu2bav6/V133YWvfe1riMViaGtrwze+8Q2sXr0aQ0NDWLNmDVavXg0AOO+887Br\n165gd0iAqKUlPEITB3H06NEqf0cQsBEHOvHb2tqEXHRBLafAL+CZTqcNU2YmkwlotLNomobx8XFf\nTzJ8SpBdmqJUKiGXyxlRDzbF4bXaShboxlgqlRCPx2teQoAPwbMRDUqNmQkiP/uKjUQlk0kkk8mG\nEZd8yo09X3hBwu5Pr34hFjYCCcBUDLGCiBX1Mpmnp6am5iyX4ETskUdQuurD+Ej+NgDAjsw2JO7/\neiCih6LrIhq01kIul0MmkzGOc7lcxkMPPYQf/OAHeMMb3oCDBw8af37zm9+gra3NdDuapmHVqlXY\nv38/Ojs70dPTg76+PkPEALPHpLW1FQDw/e9/H7t27cIPf/hDDA0N4dJLL8Xvf/978V/YPWppiaAw\nCycGESGh6AHbh6atrc1IM4hMmfkdP93YC4UCKpXKHD8RuzBmkLA3f7f7hRWRbEqQfkc3fjdNAvnj\nLxsUXWDTVq2trYGIBT5FxuK1ioyM0ew+ZKNouu6uk7Ms8FE0q5RbWH4hJ/M0cExU0nlv5iXhRZDI\naJ7f+8XMnfdga/423It3zf5HHvjKnfcgHoDgkTXFxo+LzrV169bh2muvdb2dgYEBdHd3Y+nSpQCA\nzZs3o7+/v0rwkNgBZsUPL6YbBSV4aqTWSi16AmcXHmX70PCh7aDxI3jYMUejs113zRbwJP9R0Li9\n+Zilrdrb242LlU9bsU0C2aUT+CaBrGfCLLVTrwonPrIQtliotYqMXkPnFJ+ykRXeW+TXzxWUX4gV\nKGyKzEwMsaJIBvM0uy9kQcYJ3WpMr776queV0kdHR7FkyRLj566uLgwMDMx53a5du3DHHXegVCrh\nJz/5ifH/L774Is4++2y0t7fj5ptvxvr16z19fpgoweODWiM8dONiF5m06kMjylTsdftm4sFpsVSR\nY6dtW3lQ+LQVLTRK+549frQtKvGPRCKuxAKfmqBJnH0yNjP+Bn0zZydcu8hCvbCbyNn+OdFo1Fjr\nio4dLyTrbZxm4dOFIr1FTmKIT42x6SuzfUfXCEUAzfxCvHmajeiZRaNYz5BdVMjpmvJzbFPXvRs7\nfvlhID/7847MNiSu+zoq9m9zjQznGwtbRMFy+PBhdHd3e94Wj9n3veaaa3DNNdegr68PN998M+65\n5x4sXrwYw8PDWLhwIQ4cOIDLL78cTz/9dFVESCaU4KkRLxEe1tCr68fKnu2eYkULnmjUvlu0mXhw\n09QQCEfwsFhVsgGoMm+yN14yxNKk5aUHjZvUBC+GgjT98g0OZTHzusGpUaDZhEopRi8pMpFjp9Rz\nUOlCv1idh8BcUU4ih2DfQ/9mU8bA3GaL9Bpe0PCfyX626K7ElYsvRuL+r+Mr/2daTlwXjH8HkDOl\nZXVfPXTokOcIT1dXF4aHh42fR0ZG0NnZafn6K6+8Eh/60IcAwPDWAcC6detwyimn4Nlnn8W6des8\njSEslODxAXvyu5nU2QaBXkvKa02ZOWE1/iDK4MMQPGbeJ7u0FVttxTasC3LSYicDPgLGm37tGtyZ\nRTTMxt5IlWJux+41vWMmKM0ia34nLjbCqWkaUqkUMpmMdBMhD+1D1g9ktsitk6D04xeyE0PsNcBX\nsdFDFm/adqJy8cWGZyeoyA6NWbbjbDUmP4Knp6cHg4ODGBoawuLFi9HX14c9e/ZUvWZwcBArV64E\nADz00ENYtWoVgNmFShctWoRoNIrnn38eg4ODWLFihc9vJR4leGrEKkLCGhjJ0FtLSbmoi44VJawH\npFwu1zxmkYKH9i/1p6DFUel3TmkroD5LJ1iZfvmIBkWe2AmIPFGRSASJRALpdFqa9I4dvLeo1v3u\nNr3jR1C6GXujVOjxBnCzsYfpF2JTXVZ+ISrjp/1eL/M0vz9kfKAISvDEYjHs3LkTvb29Rln6mjVr\nsH37dvT09GDjxo3YuXMnHnnkESSTSSxcuBC7d+8GADz66KP4zGc+Y3QOv+uuu7BgwYJAvp8IVFm6\nD+hGChwrV6aSPzYyEkRXYQA4cuRITd2U7SiXy5iamkIqlapaWd3MhOyVSqWCycnJQC8A2r+0thVr\nEOXTVgBM01bJZLIhlk4AjvlE6ImXPC6iSsGDHjtr5qWS+HoZuu28Lqxpnf5QNCoWixkm6nrvUzew\nIi0SiQQ6djd+If4cZKOT9Dr6mf0dRaIqlUpVKwv+2PH/NhNBQYohutfwVYn1pFQqGcvzsGzYsAGP\nP/64lAItRFRZuiii0agxobJ5/SAbBNKNIOiTuFwuG52QK5WKownZK0FFePi0Fd3A6e96pK1EQ2Mv\nl8tGWbmVR4OddPyUggeNjI0C3abI2Mga+3s6j8KOLniBjSqLWhPNi1/IKbpG22PfVy6Xq6LmrKCJ\nRNytVM8++JhFhej9bvdLo6S02Ii2whwleHxAJxQ9fWuahnw+b6RVgj7hgkwN8SbkZDKJUqkkxFVf\n67hZkzcAo2QfmO0FQRMTeyOTIW3lFz8+EZoEzLZlJoZEldPTZ7GNAhvFW0SpQhKZZMSMxWKBp8hE\nwEfS6iUw3Zj4zfxC7PvZpp/se1kTtR+/EJ2ffhbubBTBA0A6ES4bSvD4oFKpYGpqyqiOAYCOjg5h\nnxeE4OFTbWwFE4kfUReK123zhmmztFU8Hq9qDkifQZN5IpFAPB5viNQV+2QejUYDSUG48bkEUU7P\n+0QaqVEgMLebMy/S3HquzEy/ZqnGICckeuAKoyy+FszORfbBi923rPC08wvRNtyYp/lmi2z1GStm\nWb8Qex2YpcrrjZmvaHp6Wrpu0LKhBI8PqOMu3dipSZ2oi4EuOq9YVTDxN0USCiIiU263bZa2squ2\nooaAbOqHwt3805zdU3g9b2C8tyisBUjdPIk7ldPTOSlr7x87zCJpXkWa2xSZ2+gaH52wgy2LD7KL\ndhjwaTerVC29tla/EHvftDJPWwlZVkiRoOcjQfW6l5jdU1999VUcf/zxoY2hEVGCxwdk7CXo5i+y\n6ZiXCA9NRIVCwTAt2qXagkyZed02paD4sdLvaq224m+abKt+ALZiSARmqR9ZJix2Ercqp6cqGnY9\nOToe1DxQtgaBRFgVV26ia35SZNSsVLbzxg1s2s1tNCoovxArSNjUmFkVGesXouugVCoZkRM+TVav\nlerNBM+hQ4eU4HFACR4fsBcP/SxKMHjZPtvvx657s9/t+8Fq2/xiozRWwLpJoKZpRndqt6ZMt74C\nvgzcauLxO8Hwk22jpX4ofUJRBWo+6bQf+dSO2dO4aPiUYT2jUW5TZPx+JKhijE3PyHwO+RE6bvDi\nFyJRYmbkt0qR8amycrnsyi/ECzHWPG1moK7FM8e/d2xsTAkeB5TgCQCajEVun81Ns7A3cypT9FrC\nHpbgYSd9N2krNj0V9EKYQTyFu/G48IbSRk/98CZqt/uR9h+f2rESQ0EgarIVAb8f2X1PKVy6HoJO\nkYmA9ReFmXbzmmpk/7D7jfY/MBsJJAO7k1/ITgyZXQcA5hwzNjJkhZXg8dqDZ76hBI9P2ImcLhBR\nmG2foh1B9PsRLXhorJS2kr1JoJencDZFxt+0yP8iS2m2W4JK/TilJESV05stW9EoqR9231OK104g\ne0nthJFqrJfQcYOTGCIhQiXx7H3IjV/ITAy58Qux9xQ3zRat8LOO1nxDCZ4ACCPCQxcGv4BnW1tb\nzUZXUYKH+prkcjkkk8k5aSs2mlNL2iosvNww2ScwmoTcRoXqBR+NErnvI5Hgy+mdKq5kxm/azW/K\nlk81ehWVPLKtL+YFXmRmMpmqfV+LX4h9P4A5kXp6rRvzNHsdALPVtdFoFHv37sWCBQvwl7/8Beed\nd57n779v3z5s3boVmjbbZXnbtm1Vv7/rrrvwta99DbFYDG1tbfjGN76B1atXAwA+//nP4+6770Y8\nHsdXv/pV9Pb2ev78MFGdln1CT+8AkM/noeu6sJLAUqmEqakpI9KTTqeRSqUCm4hyuRyi0WhVd1O/\n8GmraHS2RJy2bSZ0zNJWrE9BdmiyYqNRFBFhb1xmoXQZ0hF8o0AK4cuG2X6kfxNUQSmrcZqHjWSy\nHZ1FfyZ/HvIrnbs9J3mhE+R9STReo2lm7zdLV/nxC/GwD4FA9RIgJOxTqRQ0TcONN96IP/zhD/jj\nH/+IqakprFy5Et3d3Vi1ahUuvvhiXGyziKqmaVi1ahX279+Pzs5O9PT0oK+vzxA0wGzPM4rIf//7\n38euXbvwwx/+EE8//TSuuuoq/OpXv8LIyAguvvhiHDx4UIbjrzoti8TOY1MLrLGXBJUI/0cQKTmK\nDvBpq+npaQCwTFvRUy3QuEZeu2gUGxXyu5CoiKiQWbWY7BERNppBkxXtKxLItD+pMzI7gZuldep1\nrrHnTtj+IqdUo5toBgDjNalUCm1tbQ1z3ZpFdPxEMoPyC9mJIbPIECumotEobr/9dgDAP/zDP+DO\nO+/ExMQEnn32WTz77LMYGxuz/Q4DAwPo7u7G0qVLAQCbN29Gf39/leBhm9LSgzcA7N27F5s3b0Y8\nHseyZcvQ3d2NgYEBnHvuuZ72Y5goweMT9uII0sNDFyOVn5Kxd3x8XJjZtZaUHCvK2Mow4NiFSV4j\nNmxOE7xsaSs3mC374EcouPEK2fXD8dMlmc4vEtGNJjJZkezm3LHzXAHhtyVgey/J5nEB7FNkdB7y\nTQHZCFVQKTIRBCV03ODVzE//Nru+2XR/qVQyeo6RCKJU+nPPPYeuri4kEgmceeaZrsY5OjqKJUuW\nGD93dXVhYGBgzut27dqFO+64A6VSCT/5yU+M955//vnGa0466SSMjo562k9howRPAATh4aEnPrbb\nbhALeLrBq4fHTJR1dHQYY2UnFGrTTxcr5aHpIibBFOak4wdWKFhVLAWFiKgQgCqPSCMthgn4j4h4\nMaCLLKdvdI+L3arrZikymarI2PsVpe7DaPBphdcIG5u2Je/bf//3fyORSGDFihV48skncf/99+O8\n887zfG5apdN4rrnmGlxzzTXo6+vDzTffjHvuucf1e2VCCR6fBBHhMTMhWy3gSeF6EWFvt4KHFWWU\ntkomk8Z3sUpb8f1nWCHHh31F9sLxA1tBJINQ8BsVovfG4/GGEjqiugp7fQL3W07PGqlFimQROAkd\nIogUmYiHHV7ohNXJvBb4tC2d+7RYMjB7T/rLX/6Cn/3sZxgcHMSf//xnvOY1r0FbWxu2bNmCCy64\nAJs3b3b1eV1dXRgeHjZ+HhkZQWdnp+Xrr7zySnzoQx8y3vunP/3J9XtlQO6j3yB4jfBQWJ7WsKKS\na7sbuahKKjfb5tNWbDt4u2orNtRtVXViV63DTjph+lvoO8u24rcdfFSIhAIAJBIJJBKJqrQEv3aW\n3xJwEZj5i8KMiDhN4E7l9KzHpVHThrWYeVnsUmSiqsj4tGcjCB0Wdvx8NLNSqWDv3r34wQ9+gN7e\nXjz44INob2/HyMiI4dvx4ift6enB4OAghoaGsHjxYvT19WHPnj1VrxkcHMTKlSsBAA899BBWrVoF\nALjssstw1VVX4WMf+xhGR0cxODiIc845J6C9IIbGOQskgzemAs5dT/lFMWkBTzc3krAFj5e0FRvN\n4aut/AoFNzdKN/4WL83sGtHIy8I+kbtJu4n0CtU6flmFgpNAJ38O7T8AlstuyOZb44WOyCaZXiJs\nblNkkUikaYWOpml46KGHsGPHDvzN3/wNfvCDH+C4444z3nvyySfj5JNPtq3IMiMWi2Hnzp3o7e2F\nps2Wpa9Zswbbt29HT08PNm7ciJ07d+KRRx5BMpnEwoULsXv3bgDA2rVrccUVV2Dt2rVIJBLYtWuX\nVOezGaos3Sd0chJHjhwxXZiTL9NOpVK+yn6npqaMkuGg0TQN4+PjWLhwoWPaim46gPu0VVjYlS3b\neTIAhLLGkij4/R/E+J32ZZBRoSAjCvWAbUtgNn43+1JktNLt+Eko0P1Jxv3vpjUBtSWQ1Q/Iw+//\ndDpdJXQefvhhfPnLX8Y555yDbdu24YQTTqjziKVHlaUHDTvZA3N9PGbCwW4BTzefJzrCk8vlqtJW\n8Xi86mmLXus1bRUWTlEhq2Z29F7yt8icumKhiIKIRoFe9iUfFbJKRfCQRyGMRociMKsYM4soeI1W\nOnnYgko3NmLqh/e4ULVkLBYzlt8IOkUmCn7/8xGd//mf/8GXvvQlnHHGGXjggQewePHiuoyzmZD7\n7G4gSACUy2WjUzDrd6n1ohIheNi0FX0Gm7biqwPYtBVblt1I/hYAc9JubBk968mQwTRtRj39RW5S\nEWYVJmxUKBI51pZA9jWuzGCFZi3j531XLHbl9Lqum0YrvaRuSWg24v63EwpWr/eaIhOduqX7Li80\ndV3HY489httvvx2rVq3CfffdV1U2rqgNldKqAfaimZiYMAQJpa2CnBjJS5PNZmvellkJfC6XM8SO\nzGkrv/D+Fhq/1THiJxz232ZpCNFPi2b+IrvxywRFMtiyfjbd46XqqZ6wpfH16gZuZvZlJ3K71C2b\neqOqn0YWOmzqp5Ztmu1HEcUR7D2Urfik3/3yl7/Ebbfdhq6uLtx0001Yvnx5Td9tHmN5cJTgqYFC\noYB8Pl9VDZPNZoXcqOlGy3a99ApN+BQZoBSUrutGY0MKF9PFzqdNkslkQ/srkslkTf4WszQE/VvE\n5M0KBTLyNprQ5IUau/xAmF4hv/A9dGQVmmbl9GxUCIARTWI9LrKfSyKEjpvP9Css+f1pZQan3/3m\nN7/BrbfeiuOOOw6f+cxn1AKgtaMEjwjGx8dRqVSQTqerln8QAV3wbW1tnt7HV1vx63CxIV7+BklP\nhdTzJR6PN8QNEpjbqC6ZTAr3h/g1TVv5Wxq5UaDbHi527+e9Qvx6T6KjQvxipI20ThQw9xpgWxPY\nrfkkg78FqI/QcTsuPkVmdm7S/bNcLiMSqe7srOs6fv/73+Pzn/88WlpasH37dqxevbqhzi+JUYJH\nBOwCokGmnKw+i9JObuDTVul02mhU51RtRWmHRCJhmLHZG2TYOW+3mJXFy7AIKT9588KS92BQo7tE\nImFE1BoFp4qloD7DTAj5EZZm2+ZL+xspogbMbdZodw24EZZhX+u8x6iRUm+0L6mrPHDM33nLLbdg\nYGAAS5YswXPPPYd0Oo3rr78eb3nLW5BOp+s88qZCCR4RkJkQ8B+B8fJZk5OTWLBgge3reNM0m7bi\nq63oby9pK7Mwr1UKwutk45dGTvuwlU7kCSPMTNMyCEsz+HMojFW/eWqJCtUakZKBoFNvYfpb6PMa\nVegAx8QyFYGwVauapuFXv/oV/u3f/g3FYhHpdBrj4+P44x//iKGhIZx66qn43e9+V9Pn79u3D1u3\nboWmzfbT2bZtW9Xvi8Uirr76avzmN7/B8ccfj+985zs4+eST8cgjj+CTn/yk0e3/9ttvx4UXXoh8\nPo+3v/3teO655xCPx3HppZfilltuqWmMIaHK0kUjsmzcafts6JeeStmGbWbVVsCxRTDpRHdTrcGW\nhZotbWBV+s2LoCD6Y7CTbKOmfWiS4qMJVlU67GQjQwqCT5vUs+LHqYLMrkMypXAp9dNIEy11Qmeb\nZYbRmoCvyOMjwG7L6ZuhaowVy3x7jueeew633norjh49iptuugnnnntu1X4olUp46aWXahqDpmnY\nsmUL9u/fj87OTvT09GDTpk1Vq55/61vfwqJFi3Dw4EF85zvfwSc+8Qn09fXhNa95DR566CGceOKJ\neOqpp7BhwwaMjIwAAK6//nq88Y1vRLlcxkUXXYQf/ehH2LBhQ01jrSdK8NQAe9KSwVfkZ9GNhvXf\nzMzMGOWNfNrKam0rEgk0yQaxvo/dZMM/HZrdHM2iQlY02rIPPHzFm1k0wUlY8pONKNO0FY22GCaJ\nGYKuA1pniW1NQEKITzfKVkHGr9MVVlfqoMrpo9GoEdls1OuYxKaZ0BkaGsLtt9+OP//5z7jxxhux\nfv160+OTSCRw8skn1zSWgYEBdHd3Y+nSpQCAzZs3o7+/v0rw9Pf347Of/SwA4G1vexu2bNkCAFUr\nq5922mnGA0wmk8Eb3/hGALPNHNetW2cIoUZFCZ4a4Cco0REetg8Opa3oic5rk8AwoyF0c+Mxm7jZ\np27+5kiTlKZpDbfsAxBMo0A3kw3tTz6KUYu3hbZPUcFGXAwTmNtDh10XjscpKlTr/vQDe/2bBh5x\newAAIABJREFURXPrjVuhTkUSRLlcNvUM1ds4bQWlrszSn6Ojo7j99tvx/PPP41Of+hQuuugi4d9h\ndHS0ql9PV1cXBgYGLF8Ti8WwYMECHD58GIsWLTJe8+CDD+Kss86qOnYAcPToUXz/+9/H1q1bBX4L\n8SjBExBmEZggITE1OTlpXGSi0lZh4RQyr1QqpjfHWCxmTEBBpcdEQtEQ0U+ydvuTj7KZpRvNUhD0\nfj4iJdMk6wY/qTc+KkTUuj/90OgeI9YnRWKTBL/T/gwraukGO6Hz8ssv40tf+hL+93//FzfccAN6\ne3tDeyAze9g2K4/nf2Zf89RTT+GGG27Aj3/846rXVSoVvOMd78DWrVuxbNmy4AZdB5TgqQE+wsN6\nAYKCTVvpuo5EIoFMJgMg/LRVWLCRLIqGZLNZ4+ZR73SOG8z6z9QrImUlhIC56Ua2HT+7vyORiNQ9\naKzgK5aCSL253Z8UTaqlgow3wjaa0AHMo2rsMbDbn05RtrC8bGz6kKwD9BmvvvoqvvKVr+BXv/oV\ntm3bhjvvvDP0a6SrqwvDw8PGzyMjI+js7Kx6zZIlS/CnP/0JnZ2dqFQqmJiYwMKFC43X//3f/z3u\nu+++OaLmAx/4AE499VRcd911wr+HaJTgqRE2lUU+niBOdrbaiibLfD5vRDec0laNaOIFjkWk7Jat\nqCWdw6fJgqbRoiFm+4E/jyi8TTd9mUzTVvBG3rA8Rk7pW7dRoUgkYgjQSETsyuWicBI6brCLsvFi\niH/4CaKcnj2P+Gv58OHD+OpXv4qf//zn+PjHP44vfelLdXsY6OnpweDgIIaGhrB48WL09fVhz549\nVa+59NJLsXv3bpx77rn4z//8T1x00UUAZtNVGzduxK233orzzjuv6j033ngjJiYm8K1vfSu07yIS\nVZZeI1QKDQATExPIZDJz8p9uYY2UFDJlm51NTU2hUqkYT0P0hw3X01O4LGkrN4jsfWI20YiodmIr\n5RpVbLJmcKvzyMw0HXZDQCv4qFqj9NCxMvUDsBTrMolLHvZ+VI9eWEGU0/NChz2PxsfHsXPnTvzk\nJz/BRz/6Ubz97W+X4n67b98+fPSjHzXK0j/5yU9i+/bt6OnpwcaNGzEzM4N3vvOdePLJJ3Hccceh\nr68Py5Ytw7/+67/i1ltvRXd3t5GhePjhhzEzM4MlS5ZgzZo1xvffsmUL3vOe99T7qzqh+vCIgu2d\nMjU1ZVzgXtA0DYVCoaqjqFmTQD5cTkY/AMYFzLeMl/WmCNRXJHiZuO18GLw3pNEaBQLVUbVa0lYi\nGwI6fW4j+1uAuV2FqQeNmVgX6RWqhXoLnf+/vTMPi6ps//h3BhCEADfMBMGFAQZDlBLRV+vNzMS1\n3NPMfG13bVN/r7ulkGVuhEuKiSVgdimmueEWlILlDggoaLJdGIgg6yzP7w/fczpzmH3mzALP57q4\nymFmeJ5zZs7zPc99399bF+q+89wf5tgxx7dVq1ZoaGiAh4cHxGIxqqursXXrVvzyyy+YNWsWpkyZ\nYnff9RYCFTxCwRU8NTU1rGDRBfcizezMMNbp+poEisVi9i5c10WRb1xnzcWA2UmwVZGgbeHmLirM\n7xiDR1u6uOtCyF01dX9L08INQK87bk3vy+1RZI+7auqEjj7fBU3Hk/sZtdSukK0LHX1gvgtyuVzF\nLHD48OG4desWfHx8UFpaikGDBmHSpEno2bMnJBIJdUi2TajgEQpue4m6ujoQor2fFj9s5eLiorLQ\nqMvPAZpW+uhyIdV0N6MuwdcSd4fqknjtLQGWyQ1iElH5Sb3qjqm1xSUffTyALDkWvm+LuoWbv5MB\nQGVnkLlRsKXjrAu+2Z65XIX1FZfm+N43B6HD95PiphDU1dUhLi4OycnJGDx4MHx8fHD79m3k5OQg\nJycHnTt3xqlTp0z6++Z2RwaAS5cu4c0330R9fT2GDx+ODRs2mDRGO4QKHqHQt58WN2zFXOD06W3F\nvwtn+luZAncHQ9ciY2oOBn+Btae2DwzcqjemWokrEtQlUHLDY7YQerC3HCN1QohvveDg4MA2tLWF\nnUt94CfyWqp9gr7iUp/E/uYgdJjrsTqh09DQgPj4eHz//fd4/fXX8e6776rdyTG1IlepVCIgIEDF\nHTkxMVHFLHDLli24fv06YmNjkZSUhAMHDiAxMRFXr17Fk08+qdYduV+/fti8eTPCw8MxfPhwzJs3\nz67dkY2AtpYQCu4HXiwWsw3jAPVhKw8PD41hK0tVW+nyF9FlBKhPpRP3ws6E+Wx5gVUHfw6ajAKZ\n86apOoe7YPPLvs1RSaLvHBwdHeHq6mpT4UNNcEuVHRwc2PwcJgQK/LMbqsmzxRbyWhj4IsHSztS6\nPqPqkqb5Sb7MtcmcJf6Whnse+G04Ghsb8cMPP+C7777DpEmTcO7cOa279aZ+poRwRy4vL0d1dTXC\nw8MBAG+88QYOHjzY0gSPRmz/ymfj8L14mDspftiKu+ujySTQUgZ12uaij3Gdtr5OwONdL21l5bYO\n9zyYataoS1xyFxluE1ZTE3z5FVf2eh6YXC99Flj+Lps6cWnpaidrCx194H7v1bkjc921mePFmIHq\nuytkbfhCh3se5HI5EhMTsWPHDrzyyis4ffq0YE2guQjhjlxUVAQfHx+V9ywqKhJ4JvYDFTxmRi6X\no7KyknUSZS4gzB2ourAV1zPGFlsmaBNCzCLDzIEbImUuMoYmo1oDSxsF6lpkuLtChjRh5Rqk2eJn\nSR+M7dOlj7jUJNiNSZoWYg62BJNnxIg1V1dXdg767gpZ+7uvTXAqFAr89NNP2LJlCyIjI3Hy5El4\nenpabGxCuCPr854tGSp4zIBMJmPjwQD0DlvZU06FJpgkXgAqDqT8xElbdkS2JaNAbaEHbceUm0/E\nLE72lsjLb4ZpLndwXYJd2+fU0JAjX+iYq3O5JeHnt6gTa7oEOz80ru2YCrErpE3oKJVKJCcnIyYm\nBoMHD8bRo0dVdkwshRDuyD4+Prh3757W92zJUMFjIkqlErW1tXB2doabmxsqKyshEolsNmxlDrgX\nE025LZouYtockS2R08IdB7/Sx9YFJ/+YMmKNaTvAjJ8Qgrq6OqNyryyNOg8da3T9VjcudYu2poo8\nbgjN3lq5MOgjdHTBPab6dlI3564QN2dNndA5evQoNmzYgAEDBuDQoUPw8vIyaH7mRAh35E6dOsHD\nwwMZGRno27cv4uPjMXfuXIvOy5ahVVpmgLlYK5VKVFVVsV9a5ocRQNywlb2VZANN2z6YszpDW1WO\nppwWY/42V6xZskrGnGgyqVO3Ha6uRFnTAmPJSidbKo83BH7Ikbt7AaDJ8bTlMC4Df1fK0tcmfSw0\ndN0Iccv8+dcmpVKJU6dOYd26dQgLC8OiRYvQqVMni81PG+Z2R+7QoQP+/PNPlbL0jRs3WnualoaW\npQtFdXU1srKy0L17d7Z0kfnS1tXVqVw4mLse/gJjyzALE/fu1ZJl5epyWjQlTGtbtIUUa5bCXGLN\nEI8mc4cc+SX+9hjK1eTsDKDJ8TSm7NtSaPOgsRU0OSNzv//Mc5g0gtatW8PR0RGEEJw7dw5ffvkl\npFIpFi9eDG9vb2tPiSI8tCxdKB4+fIidO3ciLy8PDQ0NaNeuHTp06IDy8nKkpqYiISEBAwYMgFgs\n1lmRYysXQsB2fFs0hR3ULdrqknuBfxYhZ2dnu8ypMHfHb11hB03dqdUt2vqKdv6uFLMo2ROahA53\n/raSNK0Ne8oz0pR/xbfucHJyAiEEu3btwooVK+Dr64vGxkZ4enpi2rRpGDRoEJ544gkrzYJiK9Ad\nHjNBCEFycjLWrVuHrKwsvPDCC+jYsSNycnJQW1sLDw8PBAUFqfwwpY/67F5YyrnX1ts+6AN3JwRQ\ntQtQt2DbasiBX3FlzTCoJtM6XTtt3B0dew4hcsNv5sr3MjSUY+pn1R52dHTBFc78zxMhBBcvXsTa\ntWvRqVMn9O7dG5WVlbh58ybrjpyWlobevXubNAZj3ZErKiowfvx4XLx4ETNmzMCmTZvY1yQkJCAq\nKgpisRidO3fG999/b5VE6mYCDWlZgrfffhsvvPACxo8fz5qjAY+/iJWVlcjMzERWVhaysrKQnZ2N\n6upquLq6IjAwEIGBgQgKCoJUKmWz8NVtjwNNew+ZmnvRHNo+AE2NAvm5LfzFRVfIwRohR/4ugq07\nU6tbtLku08A/oVzGFdlWBSYfrtCxdPhNkxAyxqepuQkdxsiUK3SuXLmCqKgoeHh4YNmyZQgMDFSb\n08YISWNRKo13R66trcWVK1dw48YN3LhxgxU8CoUCnTt3xs2bN9G2bVssXLgQbm5uWLZsmdHjbOHQ\nkJYl+Pbbb9U+LhKJ0LZtWwwcOBADBw5kHyeEsDlAWVlZOHnyJDZt2oSKigo4OzsjICAAgYGBkEql\nCAoKQocOHdiSb21hHH2FkK2VZBuLvkaB+hgraqoeEXqnzZqLqynww2P85Hwml8IaLtPGwg/nanLY\nFhJ9vK+4YUf+NYD5jDImoPYazuULHe53mxCCzMxMREVFwdHREdHR0ejZs6fGOZrjO2uKO7KrqysG\nDBiAvLy8JnMEHueDtmnTBlVVVZBIJCaNk6IeKnisiEgkgoeHByIiIlRKCwkhqK2tRU5ODjIzM/Hr\nr79i27ZtKCsrg5OTE3r06KESGuvcubNKKbw23xvmwi2TydiycnsoyeZjTqNAY0wAAfPstHGrS7S1\nr7B11BkeahPa6rxvNJV8WyKUy4yLu7jaYhsOfXyamAR9rj8T1xnZlnIFNcG1XFAndG7evIno6GjI\n5XIsX74coaGhFvnOmMsdmYujoyNiY2MREhKCJ554AhKJBLGxscJNogVjW99mCoDHFzU3NzeEhYUh\nLCyMfZxZHPPy8pCZmYk//vgDe/bsQVFRERwcHNCtWzc2NBYUFARfX18VE8DS0lIVJ1FmIWGEkS3n\nszBYclfKkIRpbQJTXRiHX3Flj35M3PAbkxSuz7nQV2Dq2sE0V3iMLzrt8VwA/8yDMW5kQqGG+t9Y\n0p5A3Ry4QocrOgkhuHXrFr744gtUV1dj6dKl6Nu3r0XHaQ53ZD5yuRxbtmzB1atX0bVrV8yZMwdr\n1qzB4sWLzTNoCgsVPHaESCSCi4sLQkJCEBISwj7OXCRu377N5gnt378ff/31Fwh53HCxuroapaWl\niI+PR3h4OFvCqY8BoC0IIVsyCtS3ykldGIe7ANm70GEMD83loaPP7oWmHUxjPq/cnC97PRfA45Bu\nfX09K3T4olOXwNRHuAt9HeCHdPm7awUFBVi7di3KysqwZMkSDBgwwCrffVPdkdVx5coViEQidP2f\nW/LEiRPxxRdfCDL+lg4VPM0AkUiEVq1aQSqVQiqVAnh8MY+JicGmTZvg6uqKyMhIeHp6Ij4+HqtW\nrYJCoYC3t7fKjpC/vz+bc8FdWMxRlmws9rYTIhI17efE3QlRKBRwcHCAo6MjlEolHj16ZLPWBHz4\ni5IlRaem46FuwWY+r5qSexnxbC+fKU3oEjq60CXcDT2uxuwK8T9TfLuCwsJCrF27Fnfu3MGSJUvw\n/PPPW/XGyxR3ZC7cXSBvb29kZWWhvLwc7du3x8mTJ9nrOMW8UMHTTBGLxSgrK8Pu3bvV3g0plUrc\nu3cPWVlZyMzMxOnTp3Hr1i3I5XI8+eSTbOWYVCqFRCKBs7NzEyEkpJcQVyDYapdpXejj26KPR4s1\n8ln48+B76NhKnpE+iej83TbmdYwJKPMZtlYYx1D4PceECOnqmzStqbmttnAu8z7MLiEjdLifqZKS\nEnz55ZfIycnBf//7XwwZMsQmzo2DgwNiYmIwdOhQtixdKpWquCPPnDkT06ZNg0QiYd2RGbp164bq\n6mo0NjYiOTkZJ06cQFBQEJYvX45BgwahVatW8PPzw3fffWe9STZjaFk6RQWlUomSkhK2ciwrKwu5\nubmsqWJQUBArhAIDA9G6dWsA5vESUpcTYssl2Zowh5uwunwWbrm3UPksXPghH3v00AFUe0Qx3lJA\nU9sHS4dxDIUvdGztu8EXmNx/82+CZDKZ2l3CsrIyfP3117h8+TIWLVqEyMhIu7vRoVgd6sNDMQ1C\nCO7fv8/uCGVlZbGmiu7u7jpNFdV5CXEXFuYOXCwWo1WrVnbRV4mPLh8gc/4d/jE1JGFan/e3935j\ngOE9ogz1abJU2NHWhY4uGOHOVIYyn1OlUok9e/bg4MGD6N69OyoqKnDnzh3MnTsXb731ls1VyFHs\nBip4KMKgy1SR2RFidoWY5D2lUokHDx6goaGBFUcADPYSsgVsRSDos2Br873ht7Cwx35jgGpuizmM\n9ox1mTb1M2vvQgdoGtbl7ugQQpCfn49vv/0W+fn5bJVZTk4OysvLIZFIkJCQgODgYJPGIIQzskwm\nw+zZs3H27Fk4ODhg9erVePXVV00aJ8VsUMFDsSxcU8Xs7Gz2vxUVFWzLh7y8PMyaNQtvv/02OnTo\nAACC7lyYG3sRCOoWbL7vDRMqc3Jysst2IoDlBYIul2ljP7NMbotSqYSLi4td7nYC/8xDXf5aVVUV\nYmNjcfz4ccyZMwevvfaayk1CdXU1cnNzERQUBDc3N6PHIIQzMgCsWLECSqUSq1atAgCtPjsUi0MF\njzaMvQMAgKioKMTFxcHR0REbN27E0KFDAQDr16/Hzp07IRaLERISgl27dqm0m2iJpKenY/ny5bhy\n5QrGjRuH3r17Iz8/H9nZ2bh//z4cHR31MlXk5wZYUgjxDQ+bw523QqFg77otsXNhTpjzwQgEWzkf\nmvKvtO22cefRXIXOo0ePsH37dhw6dAjvv/8+Xn/9dZVSeXNz4cIFrFy5EkePHgUAREdHQyQSqVzj\nhw0bhpUrV6Jfv35QKBTo1KkT7t+/z/5+9+7d+PPPP1UEj6+vL3JyctgcRopNQVtLaEKpVGL27Nkq\ndwBjxoxRuQPYuXMn2rVrh7y8PCQlJWHBggVITExEVlYW9u3bh+zsbBQWFmLIkCHIy8tDSUkJNm/e\njJs3b6JVq1aYNGkSEhMT8cYbb1hxptanqqoKkyZNQnJyMpydnVV+p4+pIhMe45oqchcWba0LTE0+\nbS5tOPStHOPvXBhTiWPteVgTdfYEgHaXaeCfvmPM59oaVXnGwhXQfMFWV1eHHTt2YP/+/XjrrbeQ\nlpZmkRtAIZyRHz58CABYsmQJzp49C39/f8TExMDLy0ugWVDMRYsXPMb0RpkzZw4A4NChQ5g8eTIc\nHR3RtWtXSCQSZGRkoEuXLlAoFKipqYFYLEZtbW0Tc6qWyEsvvaTxd/qYKjI5QoypIvD4Tou7I9St\nWzfWa8UcXkK2ZHhoCob06jLUWNGSFU5CmR5aCq6Fg0gkYufBfK64u5e2JjI1wRc63BuB+vp67N69\nG3v37sX06dORlpbW5GZHSIRyRi4sLMSgQYOwbt06rF+/Hh9//DHi4+PNM2iKYLR4wWPMHYCnpycq\nKipQVFSE/v37s8/z9vZGUVER+vXrh48//hi+vr5wdXXF0KFDMWTIEMtMqJnBN1UcN24cgH/ulAsK\nCtiE6cOHD+POnTtQKpXo3LmzXqaKmryEALCVJbbaV0kfzN0IU9fOhTbnblMMK61pemhOdO1MqUt2\n1+QnpG8yulBoMz5sbGzEnj17EB8fj9deew2//vqrVcI/Qjgjt2/fHm5ubnjllVcAABMmTEBcXJww\nE6CYFfu7gpsZY+4AmOdoeryyshLJycm4e/cuPD09MX78eOzduxdTpkwx38BbOIxA8ff3h7+/P8aM\nGcP+Th9TRUYIMaaKwOMLOJNLxDzGnGem+aK1zP8Mhd8fSmjBpo8BoDGd6FuK0NGGIS7T6gS8uU0r\ntQkdmUyGhIQExMXFYdy4cThz5gyeeOIJk/6eKQjhjMy85syZM3jhhReQkpJiciUZxTK0eMFjzB3A\nw4cP0bZtW/j4+ODevXtNXpuSkoLu3buzMeCxY8fi999/p4LHQojFYvj5+cHPzw+RkZHs44ypYnZ2\nNjIzM7Fr1y7WVNHBwQEymQy5ublYvXo1xo0bB1dXVwCqXkL8bum2ltTLL5G3dtsErhAytBM9ADaP\nhQnB2WIVnDaEzDUy1Q3ZUNNKblUiX+jI5XLs378f27Ztw8iRI5GSkgIPDw+T52gqQjkjR0dHY9q0\nafjwww/h5eWFXbt2WXGWFH1p8VVaCoUCgYGBOHXqFJ566imEh4cjISFBpZdJbGwsbty4gdjYWCQm\nJuLgwYNs0vLUqVORnp6OoqIivPTSS8jLy8PFixcxc+ZMXLx4Ec7OzpgxYwb69u2LWbNm6TUmIXwj\nLl26hDfffBP19fUYPnw4NmzYYJ4DaOecO3cOn332GW7evIlXX30V/v7+yM3NRW5uLmuqyN0R4poq\naipHtoaXEBPqkMlkNl0irwsm+ZwJwXGFj63aE6jDVpOq9TWt5P4/I6IZocOtglMoFDh48CC++eYb\nvPTSS/joo4+0hoMoFAtAq7Q0YcodQHBwMCZOnIjg4GA4OTkhNjYWIpEI4eHhGD9+PPr06QMnJyf0\n6dMH77zzjl7jMaVqzMXFBZ9//jnrG8Hl/fffx44dOxAeHo7hw4fj+PHjePnll813IO2UO3fuYPr0\n6Zg8eXKT8li+qWJycjKio6ObmCoyQqhdu3Yaq3CEWqz5bsL22HMMUA3BOTo64oknnlC7c6GtE705\nq/JMmYcmoz1bQN/wGDcHi3mdk5MTfvrpJ3Tp0gUBAQFIT0/Hpk2b8Nxzz+HIkSNo3769padDoRhE\ni9/hsTWE8I0oLS3F4MGDkZWVBQBITEzEuXPnsGXLFgvOrPmgzVTR2dkZAQEBrLN0UFAQvLy8zO4l\n1BxceAHz9OviL9Z8Y0VNO27mhF89ZmtCxxCUyn96jzk5OcHJyYkVm3PnzsWNGzeQl5cHQghCQ0MR\nEhICqVSKsLAwPP/882YZgxC73AyjR4/GnTt3cO3aNbOMlWJz0B0ee0EI34iioiL4+PiovGdRUZEA\no28ZiEQieHh4ICIiAhEREezjhBDU1tYiJycHmZmZ+PXXX7Ft2zbcv38fTk5O6NGjh8qOEGOqyF2g\ndVXgMM9hwiT26AUEqOYaOTk5mbQzpU/CNJPUa0jCtD40N6HDnBP+bqFSqURqaipbjn3kyBGIxWJk\nZ2fj5s2byM7ORnFxsVkEj1C73ABw4MABg3OL3N3dUV1drfJYamoq5s+fj2vXriEpKQljx441brIU\ni0IFj40hhG+EPu9JMR2RSAQ3NzeEhYUhLCyMfVxfU0Xmp0uXLmx1GBNeqKqqUqkcY5x5AQi2ayEE\n5hQ6utA3qVeT5422pN7mKnT454QQgtTUVKxduxYSiQTx8fGsyzwAPPXUU2qrmkzBGG+02bNnAwBc\nXV0xYMAA5OXlNXnfmpoarF+/Htu3b8fEiRP1Ho+6c+rn54fdu3fjq6++MmhuFOtCBY+NIYRvhKZq\nMopl0NdUMTMzEz/++CN7/r29vaFUKpGRkYFBgwZh8+bNbJ6RPl5CtiSEbCnXiGusyIef1MvPweIm\nUTMGjraQjGwMuoROeno6oqOj4e3tjR07dqB79+4WGZcQu9wAsHTpUnzyySdm8QNiRJ89nveWDBU8\nNoYQvhGdOnWCh4cHMjIy0LdvX8THx2Pu3Ll6j8nc8fS6ujpMmDABt2/fhqOjI0aNGoU1a9YYcpia\nBZpMFR8+fIjY2FisX78efn5+mDJlCoqKijBmzBgolUp4e3uzHei5porAPyX0QoRvjMGWhI4+aErq\n5eYaMc8jhKCuro6tKLOG+Z8x6BI6ly5dQnR0NNq2bYuYmBgEBARYdHxC7HJfvXoVt27dwtdff407\nd+6o/RuU5g8VPDaGUL4RsbGxKmXpw4YN02s8QsXTP/30Uzz//POQy+UYPHgwrRrjkJ+fjxs3biAl\nJQW9evVS+Z2+popSqRQSicRqXkJ8c7rWrVvbrADQBt/4sHXr1iqhK229sdTtuGkSVJaAK9rUCZ3r\n169jzZo1aN26Nb788ktIpVKrnDMhdrnPnz+PS5cuoXv37pDJZCgrK8PgwYNx+vRpweZBsT1olRZF\nK0J1G+Yyf/58hISEYObMmcJOphnDN1XMyspiTRXbtWvHltBLpVIEBgay2/rm9hJqLtVjhvQe0/R6\nfhm9ph03ob2a+CX/Li4uKkInOzsbUVFRAIBly5ahV69eVj1npnijMezevRt//PEHNm/e3OT97969\ni1GjRuldpeXh4YGqqiq1v5sxYwZGjRpFk5ZtC1qlRTEOoeLpDJWVlfj5558xf/588w68hSEWi+Ht\n7Q1vb2+Vvm2EENy/f5/dEUpISFAxVWQSpZmdIQ8PD4O9hJhKs/r6eiiVyiYNJO0JdTs6xoSnNOUJ\nWbITPV/ocF23CSHIzc1FdHQ06uvrsXTpUjzzzDM2cc6E2uU2Fl3hLxoesx+o4KFoRYh4OoNCocCU\nKVMwf/58dO3a1aRxUtQjEonQsWNHdOzYEf/+97/Zx40xVQSgslPB7OQwODg4sDs6TFm9LSyg+mAu\noaMLS3Si5zaMVddeJD8/H9HR0Xjw4AGWLFmCiIgImztPw4YNQ05OjspjTFUWADg7O2Pfvn1qX1tQ\nUKD1vf38/Azy4Kmrq4Ovry97Xfvoo48wcOBAvPrqq6isrMThw4exYsUKXL9+Xe/3pFgHKngoWhEi\nns7wzjvvIDAwEHPmzDH7uCnaEYlEaNu2LQYOHIiBAweyjzOmioyh4okTJ7Bx40YVU0WJRIKamhrs\n27cPU6dOxQcffMB6BOnyErKWA7ImNOXoWANTO9FzfZqcnJyaCJ2//voLa9euRVFRERYvXoxBgwbZ\nzHmwZeRyudrHuZWvFPuACh6KVoTqNrxkyRJUVVVh586dRo2LOrEKA2Oq2K9fP/Tr1499nBCCqqoq\nbN68GevXr4eDgwMGDBiAw4cP48SJE+jRo4fKrhDXVJEbGlO3UFujhJ4ROvX19RCLxVbXC+mYAAAR\neUlEQVQVOrrQZazINAdlKsiAx4v03Llz4ebmBl9fX2RkZODBgwdYsWIFXnzxRSp0KC0SmrRM0cmx\nY8cwb948VlwsWrRIJZ7e0NCAadOm4fLly2w8nQlRcePpbdq0wYkTJ+Du7o4uXbpAKpWyIZDZs2fj\nP//5j17jUSqVCAgIUKkcS0xMVInTb9myBdevX0dsbCySkpJw4MABJCYmora2FleuXGErx/iC58CB\nA/jpp59w7dq1Fil4tLF582b89NNPWLJkCbto8k0VmTYb+poq8pN6hfYS4oZ7HBwc2GRke4Q/FxcX\nFzg4OLB5QgcPHsQvv/yCu3fvorKyEqWlpSCEQCqVok+fPoiJiTHLOKhtBcXG0HixoIKHYncIVTlW\nU1ODyMhI1omVCh5VFAqF3n2u+KaKzA8TBujSpYuKEOrWrZuKqZ+uyiZDvYSam9BhdqfUzeX+/fvY\nuHEjMjIysGDBAowcOZL1Dbp//z6ys7NRUlKCyZMnmzwWIW4+6urqkJGRoWJbsXjxYmpbQdEXWqVF\naT7YgxNrc8SQpp6aTBWZnZ2CggJ2R+jIkSMoKCjQ21TREC8hvtBxdXW1e6HDGB3y51JRUYFNmzYh\nLS0NH374Ib766iuVKjFuAru5EKINROvWrdmeXI6OjggLC0NhYaHZxkxpudjnN5/SoqFOrPYLE7Ly\n9/eHv78/xowZw/7OUFPFVq1asa9TV+LNCB6xWIxWrVo1G08gfr7Rw4cPERMTg1OnTmHevHmIiooy\nuOO8sVDbCoo9QQUPxe6gTqzNE7FYDD8/P/j5+SEyMpJ9XKlUorS0lBVCcXFxyM3NRWNjI9q2bati\nqtilSxfs3bsXFRUV+PTTT9mFX6FQ4NGjRxq9hGypcoxBV6l8dXU1tm7diiNHjmDWrFlYuXKlxXev\nqG0FxZ6w3aY2FIoGuJVjjY2NSExMxOjRo1Wew1SOAdCrcuy9995DYWEh8vPzkZaWhsDAQIPFzrFj\nxxAUFISAgAB88cUXTX7f2NiIyZMnQyKRoH///qxoq6iowODBg+Hu7t6kx5lMJsO7776LwMBABAcH\n48CBAwaNqTkgFovRuXNnDBkyBPPmzcO2bdtw5swZpKWlIS4uDiNHjsSjR4+waNEiBAUF4eDBg8jP\nz8eaNWvw448/4tq1a5DL5XB3d4eHhwe7Q8KEu2pra1FVVYXq6mrU1taivr4eMpmM9cCxNIzQefTo\nERobG+Hi4gI3NzfW6bm2thYbN27EyJEj4e3tjd9++w3Tp0+3SqjOkJsPANS2gmJV6A4Pxe6wNSdW\nQLieY6tXr8aTTz7JmrBVVFSYNM7mBJOT8ujRI0yePBn/+te/cP78eYSGhhplqsgNjVnDS4gQArlc\njvr6egCAi4uLSjuL+vp6xMXFISkpCTNmzEBaWhqcnZ3NOgZDsVXbCgpFHbRKi0IxA0JVjvn6+iIn\nJ4cmUmuB6QcVHBys83lcU0WmhP7BgwdwdnaGRCJhQ2NBQUHw8vJq4iXE/a+5vIQYodPQ0ABCSBOh\n09DQgD179mDPnj2YOnUq3nvvPbi4uBh9vMyNrdlWUFo8tEqL0jJhSpqFaszIIETy5sOHDwE8vts9\ne/Ys/P39ERMTAy8vL4FmYZ+IRCKdYod5niZTxdraWuTk5CAzMxOpqanYvn07ysrK4OTkxJoqMjtD\nmkwVDfUSYlpJ1NfXqxU6MpkMe/fuRVxcHCZOnIizZ8/Czc3NvAfPDAjRBoKpwKNQzAkVPJRmDb9a\nJSkpCenp6fj666/ZbXRzCCEhkjflcjkKCwsxaNAgrFu3DuvXr8fHH3+M+Ph4k8dL+QeRSAQ3NzeE\nhYUhLCyMfZxvqnjx4kXEx8ejuLgYYrFYo6ki8E8JvVwuV+slBDw+v4SQJl3l5XI5kpKS8O2332LM\nmDE4ffo03N3dLX9gKJRmBhU8lGZLUlISLl26hB49emDw4MHw9/dHeno6ZDIZgKaCRC6Xw9HREdev\nX4dYLEbPnj31/ltCVI61b98ebm5ueOWVVwAAEyZMQFxcnN5jopiGSCSCi4sLQkJCEBISwj7OJBXn\n5+cjMzMTmZmZ+PHHH9nEXF9fX5UcIa6p4rVr1+Du7o4OHTqwZoAXLlzAqlWrIJFI4OzsjNTUVERG\nRuL48eN6JfcaihBtWS5duoQ333wT9fX1GD58ODZs2GD2cVMopkKrtCjNFolEgqeffhr5+flYvXo1\nAOD27dvo27cvACA1NRU3b95EVVUVALBVLp9//jlb4cV4uuhCiMox5jVnzpwBAKSkpOgVuuEjRPVY\nQkICevXqhd69e2P48OEtKpmaMVUMCgrCuHHjsHTpUiQkJCA1NRXnzp3DqlWr0Lt3b+Tn52PdunUY\nMWIEBgwYgLCwMEyYMAFHjx5FcXExnJ2d4e7ujl69emH06NEoLy9Hbm4uvLy8kJCQgK5duyIiIgI7\nduww29iZ5Prjx48jMzMTCQkJuHnzpspzuMn18+fPx4IFCwCATa5ft25dk/d9//33sWPHDuTm5iI3\nNxfHjx8325gpFLPB9F3R8EOh2C0pKSlk69atJD09nXTr1o3cu3eP9O7dm6SnpxOZTEaefPJJMmLE\nCBIREUGGDh1KiouLCSGEDB06lMTFxZHGxkaD/t7Ro0dJQEAA8ff3J1FRUYQQQpYtW0Z+/vlnQggh\n9fX1ZMKECcTf35/069ePFBQUsK/t2rUrad++PXF3dyddunQh2dnZhBBC7t69S5577jkSGhpKhgwZ\nQu7du2fQmBQKBenRowe5c+cOaWxsJKGhoex7M8TGxpL333+fEEJIYmIimTRpEiGEkJqaGvLbb7+R\nbdu2kTlz5rDPl8vlpGPHjqSiooIQQsiCBQvIypUrDRpXSyErK4uMGDGC+Pj4kNWrV5NDhw6RL7/8\nksyYMYM899xzpF+/fqRLly7kk08+IWVlZSqv/fvvv0lqaiq5fPmy2cZz/vx5MmzYMPbfUVFRJDo6\nWuU5L7/8Mrlw4QIh5PG57tChg8rvv/vuO5XPQ0lJCZFKpey/ExISyHvvvWe2MVMoBqJR09CQFqVZ\ncurUKURHR6N///44c+YMlEolCgsL8fDhQ0gkEpSWlqKmpgZr165FcHAwpk+fjmPHjmHGjBkoKSnB\nn3/+iR9++AGFhYU4cOAApFIpKisrkZubC19fX3Tq1KnJ3xQiedPX1xfnzp0z+jgIYf1P/rcLVV1d\njTZt2qCqqgoSicToMTZn6urqMGzYMOzfv5+trBo1ahT7e6VSiZKSEnh7ezd5bfv27TFw4ECzjkeI\n5PqioiL4+PiovGdRUZFZx02hmAMa0qI0OxQKBX7//Xc8/fTTWLVqFb766isEBwejuLgY7u7uaNu2\nLW7duoWAgAAEBwdDJpMhMjISFy5cQGNjIyorKzF06FCkpKTgzTffRFJSEgCgpKQEW7duxYQJE9C/\nf3+76O+jboHjL0aaFjhNODo6IjY2FiEhIfDx8UF2djZmzpwpzATsnLCwMMyePVtjGblYLFYrdoSC\nCJBcr897Uii2ABU8lGaHg4MDevTogV9//RVbt27FnDlz0NDQgNLSUvj6+gJ43DuL6cUkk8lQVVUF\nT09PFBYWol27dhg9ejQUCgW6devG5jj4+fkhKioKKSkp+OSTTxATE2O1OeqLEAucXC7Hli1bcPXq\nVRQVFSEkJARr1qwxz4ApgiKEM7KPjw/7fE3vSaHYAlTwUJolI0aMwMKFC/HgwQN0794db7/9Ni5f\nvszeTRcUFKC+vh5FRUU4f/48Tp8+jdGjR+Pq1avo1q0bgMfhhr/++gt+fn4ghOCbb77Bq6++iv79\n+2PhwoXswmHLniFCLHBXrlyBSCRizeMmTpyI8+fPm3/wFLMjRHJ9p06d4OHhgYyMDBBCEB8fr9IU\nlkKxFWgOD6VZ4unpiYkTJ6o8NmHCBNTU1AAAHjx4gNDQUMyaNQv5+flYuHAhBg4ciBkzZrC7QDKZ\nDHfv3sVzzz2Hn3/+GRkZGTh69Cg8PT3xySefsC0AbBkhrP+9vb2RlZWF8vJytG/fHidPnoRUKhV8\nLhTTEaotS2xsrEpZ+rBhw6w4SwpFA9oymi2WU02hCIBCoSAKhYLI5XKVx2tra0n37t1Jfn6+yvMI\nIWTPnj0kNTWVEEJIeXk5GTp0KDlx4gS5dOkSmTZtGklLSyOnTp0iUqmUxMXFEUIIUSqVFpyV4QhR\nPbZt2zYilUpJaGgoGT16NFuxZei4AgMDiUQiaVIpRAghDQ0NZNKkScTf359ERESQu3fvEkIIOXny\nJHnmmWdIr169yLPPPktOnz7NvubPP/8kISEhRCKRkHnz5hk8JgqFYvdo1DRU8FBaHHV1dWTdunWk\ntrZW53PLyspIXV0dIYSQRYsWkbCwMLJ8+XLSoUMHVhhRDMeUcvkrV66QkpISQgghN27cIN7e3uxr\nwsPDSXp6OiGEkMjISHLs2DFLTIdCodgOGjUNbR5KoXBQKBRN2lGoo6ysDO3atWPNCimGYY5mqwxe\nXl4oLi5GeXk5Bg8ejKysLABAYmIizp07hy1btlhmUhQKxRbQWHFBk5YpFA7axA7THwkAOnbsSMWO\nCZirXH7//v3o06cPnJycmp0fjLEO2QAQFRUFiUQCqVSKEydOsI+vX78eTz/9NHr16oWpU6eisbHR\nInOhUGwBKngoFD1hOl9TTEfdzrKh5fKZmZn4v//7P2zfvl3v97QXTGkBkZWVhX379iE7OxtHjx7F\nBx98AEIIiouLsXnzZly6dAnXrl2DXC5XSUimUJo79OpNoVAsjqnl8oWFhRg7diz27NnDlsc3Jz8Y\nrkO2k5MT65DNJTk5GdOnTwfw2CH79OnTAIBDhw5h8uTJcHR0RNeuXSGRSFg3ZYVCgZqaGsjlctTW\n1trt8aFQjIEKHgqFYnFM8YOprKzEyJEjER0djYiICPb55vSDMTaclJKSgmeffRahoaHo27cv2/i1\nrq4OI0eOhFQqRUhICP773/9q/fvGhPw8PT1RUVHR5LXe3t4oKipC586d8fHHH8PX1xfe3t5o06YN\nhgwZYtTxoVDsESp4KBSKxeH6wfTs2ROTJ09m/WAOHz4MAJg5cyb+/vtvSCQSbNiwAdHR0QCAb775\nBrdv38Znn32GPn36ICwsDH///TcAIDY2FjNnzkRAQAAkEolRfjCmhJO8vLxw+PBhXL16Fd999x2m\nTZvGvubTTz9FdnY2Ll++jLS0NK0dxY0J+THP0fR4ZWUlkpOTcffuXRQXF+PRo0fYu3ev9oNBoTQj\naNYlhUKxCsY2W128eDEWL16s9j2feeYZXL9+3aRxmdJwNTQ0lH1Oz5490dDQAJlMhtatW+P5558H\n8LgXWVhYmNZebIaE/Dp37gyFQoGHDx+ibdu2GkN7KSkp6N69O9sEdOzYsfj9998xZcoUo44ThWJv\n0B0eCoVC4SBEBRmXyspK/Pzzz3jxxRc1jsGUkN/o0aORmJiIxsZGFBQU4NatWwgPD4evry8uXLiA\n+vp6EEJw6tQp6pBNaVHQHR4KhULhYM4KspMnT6o8T6FQYMqUKZg/fz6bbK0OU1pABAcHY+LEiQgO\nDoaTkxNiY2MhEokQHh6O8ePHsyKsT58+eOeddww9PBSK3UKNBykUCoXDhQsXsGLFChw7dgyAelPE\nyMhIrFixgjVFfOqpp1BWVgbgcQjpxRdfxO7du1WSqoHHeUkeHh5Yv3695SZEobQsNHpR6BI8FAqF\n0qIQiUQOAHIAvAigBEAGgNcIIdmc53wA4GlCyAcikWgygFcIIZNFIlEbAGcBrCSEHOC97+cAAgkh\nEyw0FQqFwoEKHgqFQuEhEomGAdiIx3mOOwkh0SKRaCWAi4SQwyKRyBnAHgB9AJQDmEwIuSMSiRYD\nWAQgD4/vNAmAoQCcAdwDkA2g8X+PxxBC4iw8NQqlxUIFD4VCoVAolGYPrdKiUCgUCoXS7KGCh0Kh\nUCgUSrOHCh4KhUKhUCjNHip4KBQKhUKhNHv+H7GvwywHxSTxAAAAAElFTkSuQmCC\n"
                    }, 
                    "output_type": "display_data"
                }
            ], 
            "execution_count": 10, 
            "source": "# Evaluate the different parameters for the model & plot\n\nset_of_learning_rates = np.arange(1e-2, 3e-2, step=1e-2)\nset_of_regs = np.arange(1e-2, 3e-2, step=1e-2)\n\nbest_params = cross_validation(set_of_learning_rates, set_of_regs, X_train, y_train)\n\nbest_learning_rate = best_params[1]\nbest_reg = best_params[0]"
        }, 
        {
            "metadata": {
                "collapsed": false
            }, 
            "cell_type": "code", 
            "outputs": [
                {
                    "output_type": "stream", 
                    "text": "Score model\nBEST LEARNING RATE (CROSS-ENTROPY) = 0.01\nBEST REG (CROSS-ENTROPY) = 0.01\nTrain the model with the best hyperparameters\niteration 49999: loss 0.575395\nPredict test data\n > Prediction accuracy: 0.70\nCROSS-ENTROPY = 0.701030927835\n[2 0 0 3 0 0 0 3 3 0 3 2 0 0 2 2 0 0 0 3 0 0 3 3 2 2 0 2 3 3 4 3 2 3 2 2 3\n 3 0 3 2 0 0 0 2 4 3 0 2 3 3 0 0 2 2 3 0 3 3 0 3 3 2 0 2 4 3 2 2 0 3 2 2 0\n 2 4 3 3 0 0 3 0 4 2 0 2 2 2 3 4 2 0 3 0 0 2 3]\n[2 0 0 3 0 0 3 3 3 0 3 1 0 0 1 2 3 0 0 4 0 0 3 3 1 2 3 1 4 3 4 4 2 3 1 2 3\n 3 3 3 2 0 0 0 1 4 3 0 1 3 3 3 0 2 2 0 3 3 3 3 3 3 1 0 1 4 3 2 1 0 3 2 2 0\n 1 4 4 3 0 3 4 0 4 2 0 2 1 2 3 3 2 0 3 3 3 2 3]\n", 
                    "name": "stdout"
                }, 
                {
                    "metadata": {}, 
                    "data": {
                        "text/plain": "<matplotlib.figure.Figure at 0x7f25a1a47cd0>", 
                        "image/png": "iVBORw0KGgoAAAANSUhEUgAAAkYAAAEzCAYAAADU2syDAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzt3X+MbPd51/HP4ywXtslNbvdCi4pJblt+pKmItzaElgh1\nQ7PXTiw5oFKJoKh3K5D4o6IRIaihf3C30poffwScKkilarCTCreQVE5dqdpF6O6oBKnlrp3rmLq5\nDcTbxI0wIlOn7pWD1vWXP+aZ3dnZMzNnzpxznu/seb+k0d2Ze853nvOZ7zn7ZL7nOpZSEgAAAKS7\nogsAAADIBY0RAACAozECAABwNEYAAACOxggAAMDRGAEAALiVugYyM/7dPwAAWBopJRt/rdZvjFJK\nS/u4fv16eA1dfpA/+Xf5Qf7k3+VHVP6TsJTmDg8Po0voNPKPRf6xyD8W+cfKLX8aIwAAAEdj5La2\ntqJL6DTyj0X+scg/FvnHyi1/m7bONtdAZqmusQAAAJpkZkpN33y9zHq9XnQJnUb+scg/FvnHIv9Y\nueVPYwQAAOBYSgMAAJ3DUhoAAMAMNEYutzXOriH/WOQfi/xjkX+s3PKnMQIAAHDcYwQAADqHe4wA\nAABmoDFyua1xdg35xyL/WOQfi/xj5ZY/jREAAIDjHiMAANA53GMEAAAwA42Ry22Ns2vIPxb5xyL/\nWOQfK7f8aYwAAAAc9xgBAIDOWfgeIzO7y8yeNrMn6y0NAAAgD/MspX1Q0nNNFRIttzXOriH/WOQf\ni/xjkX+s3PIv1RiZ2d2S3ivp55stB+fe3p509ergsbcXXc10JWo9eHhPT12+qqcuX9XBw3ul9+uK\nwnzQmmXPv/HzazjWhz88dawccuSy0qKU0syHpE9LWpf0g5KenLBNAqba3U1pdTUlafBYXR28lqMS\ntd7c2U13dLLNHa2m29d2lucYG1aUz82dbmYRYdnzb/z8Knk9yiHHZbp0LhPvW872M0UvntpAelDS\nx/3nDUm/OmG7do8Iy2dz8+TMHj42N6OrKlai1oO1s9t8Y2VteY6xYUX5HKx1M4sIy55/4+dXyetR\nDjku06VzmUxqjFZKfKn0TkkPmdl7Ja1Kumhmn0op/ej4hltbW7py5Yok6dKlS1pfX9fGxoakkzXE\nXJ8/8sgjS1XvUj7v9zV4Jm3dI209o+Pn2eXf7w+eS9rekDZ6OlV/r9fT7aO+7hs+9z/vHXn+2D3S\nY8+cbB96PAHzvyifi8PnmR1/9HPyL1d/refXyPn8wHdLH/lfJ9ejM9v768O/Pzjq6+Ver7U8+v2R\nCja2pd6GBpeodt5/Ged/0fPhz4eHh5qqqFua9NA5Xkrb39+PLuH8G/k+WNunvw/OLv8ptQ7N+qp/\n0n45aiL/HJYglgX5n9X4+TV6jl9boqW0bS3LZaW0qOu/qi6lndr4HDdGaMnubkqbm4OLWu5ndola\nb+7spoO1zXSwtnlysVymY2xYYT5ozbLn3/j5VXKsHHL0UpO21fXLSm0mNUb8Bx4Rwn7alK4vx3yp\nWusyHSOwbOo8v5bpXF2mWnPH/4nsDKNrkGgf+cci/1jkH+z56AK6Lbf5T2MEAADgaIzc8O51xCD/\nWOQfi/yDfWd0Ad2W2/ynMQIAAHA0Ri63Nc6uIf9Y5B+L/INxj1Go3OY/jREAAICjMXK5rXF2DfnH\nIv9Y5B+Me4xC5Tb/aYwAAAAcjZHLbY2za8g/FvnHIv9g3GMUKrf5T2MEAADgaIxcbmucXUP+scg/\nFvkH4x6jULnNfxojAAAAR2Pkclvj7Bryj0X+scg/GPcYhcpt/tMYAQAAOBojl9saZ9eQfyzyj0X+\nwbjHKFRu85/GCAAAwNEYudzWOLuG/GORfyzyD8Y9RqFym/80RgAAAI7GyOW2xtk15B+L/GORfzDu\nMQqV2/ynMQIAAHA0Ri63Nc6uIf9Y5B+L/INxj1Go3OY/jREAAICjMXK5rXF2DfnHIv9Y5B+Me4xC\n5Tb/aYwAAAAcjZHLbY2za8g/FvnHIv9g3GMUKrf5T2MEAADgaIxcbmucXUP+scg/FvkH4x6jULnN\nfxojAAAAR2Pkclvj7Bryj0X+scg/GPcYhcpt/tMYAQAAOBojl9saZ9eQfyzyj0X+wbjHKFRu85/G\nCAAAwNEYudzWOLuG/GORfyzyD8Y9RqFym/80RgAAAI7GyOW2xtk15B+L/GORfzDuMQqV2/ynMQIA\nAHA0Ri63Nc6uIf9Y5B+L/INxj1Go3OY/jREAAICjMXK5rXF2DfnHIv9Y5B+Me4xC5Tb/aYwAAAAc\njZHLbY2za8g/FvnHIv9g3GMUKrf5T2MEAADgaIxcbmucXUP+scg/FvkH4x6jULnNfxojAAAAR2Pk\nclvj7Bryj0X+scg/GPcYhcpt/tMYAQAAOBojl9saZ9eQfyzyj0X+wbjHKFRu85/GCAAAwM1sjMzs\nj5vZb5rZ583sWTO73kZhbcttjbNryD8W+cci/2DcYxQqt/k/szFKKf0/Se9KKX2fpHVJ7zGzdzRe\n2RI4eHhPT12+qqcuX9XBw3vld9zbk65eHTz29iaOVWb8yjUsoqD+eXcd/lxW1XymFjGj/qJay7xn\n2WOsWn/pLMaOM2SulKy/6n5ljnsRhUNVPX+n7Hf7oQ8vPq9LqnP8iDlV9bwsyr/q9ahpU+f18Oc6\nxhqd2CXPm1yuI41KKZV+SPoWSQeS/krB36Uuubmzm+5oNSUpJSnd0Wq6ubM7e8fd3ZRWT/ZLq6vp\n9rWdM2PdePfZ18bHr1zDIgrqT7vl3vPUrtsqvWvRcZbJZ5H6i2p9/NrsvMseY9XPrmwWt6/tnDrO\nV1cupG/qQrtzpWT9TR73PPNzXNFUublT7fx9RRfSqysX5t6v7s+pzmtGxPWn6nlZ9GHe3NmtdD1q\nWlGuo/Na2+Xn9ayxjif2TrnzJuR3ToO8bznb6xS9eGajwTdLn5f0B5L+xYRt2j2iYAdrmyeTyB8H\na5uzd9w8u983VtbOvPZ1nX1tfPzKNSyioP60We49T+26rdK7Fh1nmXwWqb+o1hsrs/Mue4xVP7uy\nWRTNqdbnSsn6Gz/ukvNzXNFUKaqj7Plb13m/iDqvGRHXn6rnZdGHebC2Wel61LRZc0zb5ed12fma\n1sqdNyG/cxo0qTFaKfmt0muSvs/M3ijps2b2tpTSc+PbbW1t6cqVK5KkS5cuaX19/fhu8+EaYq7P\nH3nkkbnqPTjq62VJG37sPUm3j/q6b/h80v4j2z92j/TYMyfPR//+czrSG2eMf+q5/3lx1vsv+nxS\n/SX27/dHRnjimo8yeD4zf3/f4fuXyWeR+vv9kXfsXZfU063X+nrXyP7S2bxPvcMT16Sxv190/lwc\n2X78eEaf/9fXjvR6f769IW30zm4/+n7zzv+qzyfWX+F8+wMd6aGx8e4deT7v/Bx/fmq+3rMlPbOl\noTPjjx1P0fwc/v3w87jz2pEe9Ncf0eA+hbcXbF+1/qLnVfMveh5x/Zl4fo397cFRXy/3eif7Dz7M\nU/nfPuof769fuF+j16Po30fjxzN6Pl/v+d/3+zM/v1nXi96GtN2TekdHx39/fL0oGL/O+TP6vK3r\nz/Dnw8NDTVXULU17SPpnkj5U8Hp7bV4D9vf359q+jqW04Veiy7qUNs9XumO7Jo3tOi3/ppbSptVf\nVOvcS2lT4mlzSUnbs5fS5p3/VbV93LUtpW3rzFLaPOfv6FJa0X77i87rks7VUtoc52XReX9qKU37\nS7OUNvXCMudYx+fITrnzpqnPvK3rzzhN+MaoTCP0JyW9yX9elfTrkt5bsF3LhxTv5s5uOljbTAdr\nm/NNjt3dlDY3BxPQJ1/RWGXGr1zDIgrqn3PXtLk5365V81mk/qJay7xn2WOsWn/pLMaOM2SulKy/\n6n5ljnsRw89S2zoZqur5W+N5v4g6x4+YU1XPy6L8q16PmjZtXs9bbOlzpOR5k8t1pA6TGiMb/N1k\nZvaXJH1Sg/uM7pL0H1NKDxdsl2aNhdPsp03p+vJmRv3LoSvHOa7O4y4aq+r4Xf08ckH+J+qc18vI\nzJRSsvHXy/xz/WdTSvemlNZTSm8vaorOg9E1SLSP/GORfyzyj0X+sXLLn//yNQAAgKMxcsO71xGD\n/GORfyzyj0X+sXLLn8YIAADA0Ri53NY4u4b8Y5F/LPKPRf6xcsufxggAAMDRGLnc1ji7hvxjkX8s\n8o9F/rFyy5/GCAAAwNEYudzWOLuG/GORfyzyj0X+sXLLn8YIAADA0Ri53NY4u4b8Y5F/LPKPRf6x\ncsufxggAAMDRGLnc1ji7hvxjkX8s8o9F/rFyy5/GCAAAwNEYudzWOLuG/GORfyzyj0X+sXLLn8YI\nAADA0Ri53NY4u4b8Y5F/LPKPRf6xcsufxggAAMDRGLnc1ji7hvxjkX8s8o9F/rFyy5/GCAAAwNEY\nudzWOLuG/GORfyzyj0X+sXLLn8YIAADA0Ri53NY4u4b8Y5F/LPKPRf6xcsufxggAAMDRGLnc1ji7\nhvxjkX8s8o9F/rFyy5/GCAAAwNEYudzWOLuG/GORfyzyj0X+sXLLn8YIAADA0Ri53NY4u4b8Y5F/\nLPKPRf6xcsufxggAAMDRGLnc1ji7hvxjkX8s8o9F/rFyy5/GCAAAwNEYudzWOLuG/GORfyzyj0X+\nsXLLn8YIAADA0Ri53NY4u4b8Y5F/LPKPRf6xcsufxggAAMDRGLnc1ji7hvxjkX8s8o9F/rFyy5/G\nCAAAwNEYudzWOLuG/GORfyzyj0X+sXLLn8YIAADA0Ri53NY4u4b8Y5F/LPKPRf6xcsufxggAAMDR\nGLnc1ji7hvxjkX8s8o9F/rFyy5/GCAAAwNEYudzWOLuG/GORfyzyj0X+sXLLn8YIAADA0Ri53NY4\nu4b8Y5F/LPKPRf6xcsufxggAAMDRGLnc1ji7hvxjkX8s8o9F/rFyy39mY2Rmd5vZDTN7zsyeNbOf\naKMwAACAtpX5xuhVSR9KKb1N0g9I+nEze2uzZbXn4OE9PXX5qi7+8D/XwcN70eXUYm9Punp18Nib\n85DG9x3m89Tlq43mU9cac9X6F8msjDJ1FNXQdP51zv+6M2z7M2lrrheZNf/LZLFIXuPHXudYk16r\ns/5FzXv9KTtXirar8zxv6/qwLNf/2qSU5npI+qykHyp4PS2bmzu76Y5WU5JSktIdraabO7utvb+2\n689sdzel1ZNDSqurg9eq7PvgyvR8mqh/EVXrXySzMsrMs6IaHr9Wz/yc9DnVOf/rzrCO8abNz/Fj\nf0UX0jd1Ya65XnX+z7tfmSwWyatoHjy4slvbWDfevVNp/td5Do5a9LpV9rwp2u7xa7u1ned1nL/T\n5nX078c2eN9yts8penHSQ9IVSYeS3lDwd60eUB0O1jaPP/R9//NgbbO192+isdg8OaTjx2bJQxrf\nd09nBxvNp8769/f3Fx6jav2LZFbG6DwrqmNSDTdWZu9XxqTPqc75X3eGdYw3bX4WfSbzzvU6G6Np\n879MFovkVZTFnjZrG+vrWqs0/+s8B0fNm/+4MufzpO1urGyeOc6q53nZOqaZNq/rGL+sOq7/VUxq\njFbKfrNkZm+Q9BlJH0wp/WHRNltbW7py5Yok6dKlS1pfXz/+imx4c1VOz28f9XWf137L/7zof7ZS\nz/M6Vtf40sZwROmex6RnHiu9f78/sv89W3rmmb6unowmaSyfGuu/devWQvsvWv/xFhs9qbetfr+n\nXq/Gz9vfxavTwVFfL/d6x3/f749ssbEt9TZ067W+3jWt/kXqqXn+F9U/+nmEjDdlfh4c9fWyTj6P\nkXc7fj6aj54fjNHU+Ttr/s+anwvnNXb8z6g/eHXO8+F4/oyNN/68zPxfZP40fv2ZcTzj59dw+xO9\n09fnEuON11M0/rznr6ZsX8f4bV7/yx5vr9fT4eGhpirqlsYfklYk7WrQFE3aps1GrxbRXxU2vpS2\nrepLadta7qW0OepfJLMy5l5K8xqWdimthgzrGO9cLqVNyGKRvKYupdUw1txLaQ2cg6OyWUpb8Dxn\nKW1xmvCNUdnG6FOS/vWMbdo8ntrc3NlNB2ub6WBts/UPvanGYnfXv5re1twXl/F9p+WTW2OUUvX6\nF8msjDLzrKiGOubnrAahrvlfd4aLjjdrfo4f+7xzva3GKKVyWSyS1/ix1znWpNfqrH8edVy3yp43\nRdvVeZ4vev7OmteRvx/bULkxkvROSX+kwbftn5f0tKQHCrZr+5hqFbHG2XRjscj4ZfbN7R6jUVXr\nz6HZq7uGMuPVmX9E/U3XEXmP0bzv2fRxNy3iunger/9Va4ioa+nuMUop/TdJr5u1HQAAwLLjv3zt\nTm5yRATyj0X+scg/FvnHyi1/GiMAAABHY+TG//ki2kX+scg/FvnHIv9YueVPYwQAAOBojFxua5xd\nQ/6xyD8W+cci/1i55U9jBAAA4GiMXG5rnF1D/rHIPxb5xyL/WLnlT2MEAADgaIxcbmucXUP+scg/\nFvnHIv9YueVPYwQAAOBojFxua5xdQ/6xyD8W+cci/1i55U9jBAAA4GiMXG5rnF1D/rHIPxb5xyL/\nWLnlT2MEAADgaIxcbmucXUP+scg/FvnHIv9YueVPYwQAAOBojFxua5xdQ/6xyD8W+cci/1i55U9j\nBAAA4GiMXG5rnF1D/rHIPxb5xyL/WLnlT2MEAADgaIxcbmucXUP+scg/FvnHIv9YueVPYwQAAOBo\njFxua5xdQ/6xyD8W+cci/1i55U9jBAAA4GiMXG5rnF1D/rHIPxb5xyL/WLnlT2MEAADgaIxcbmuc\nXUP+scg/FvnHIv9YueVPYwQAAOBojFxua5xdQ/6xyD8W+cci/1i55U9jBAAA4GiMXG5rnF1D/rHI\nPxb5xyL/WLnlT2MEAADgaIxcbmucXUP+scg/FvnHIv9YueVPYwQAAOBojFxua5xdQ/6xyD8W+cci\n/1i55U9jBAAA4GiMXG5rnF1D/rHIPxb5xyL/WLnlT2MEAADgaIxcbmucXUP+scg/FvnHIv9YueVP\nYwQAAOBojFxua5xdQ/6xyD8W+cci/1i55U9jBAAA4GiMXG5rnF1D/rHIPxb5xyL/WLnlT2MEAADg\naIxcbmucXUP+scg/FvnHIv9YueVPYwQAAOBojFxua5xdQ/6xyD8W+cci/1i55U9jBAAA4GY2Rmb2\nCTN70cy+0EZBUXJb4+wa8o9F/rHIPxb5x8ot/zLfGD0q6f6mC0GN9vakq1dPfm5r3xxUrT+H486h\nhkXUXf+y51GnMlnUmVdE9l35vLtynMsspTTzIektkr4wY5u0zPb391t/T203kNnubkqrqylJg/FX\nVwevNbBvnfXXkn/V+hfJrC4N1lDmc1o4/7rrr2G8Oudn0VhVxy/ab2r+ZbKoM/+I86HF95w7/zrl\ncK0ZUee8XkTE79+UUvK+5Uw/wz1G581HPyq98srJ81deGbzW9L45qFp/DsedQw2LqLv+Zc+jTmWy\nqDOviOy78nl35TiX3Eqdg21tbenKlSuSpEuXLml9ff147XB413muz4evtfr+zx+/dX3jD8eTdO2J\nOcfv94/3v/bEYIyNafvXWP/wtYWOf5H6/c/r/kOv35fanA/9/uC519CTTh1P9vO/7vrrGK/O+fl8\nQT5tzn/fZuL8rDv/We/H/K/2fPheqnB9buB59Pu3nf/w58PDQ01V9DXS+EMdWEqL0PRSWlL1pbQy\n+0Z85TpV1foXyawuDdbQyudUd/01jLdMS2lTlcmizvwjzocW3zP0upXDtWZELktpUTRhKa1sY3RF\n0rMztmnzeGp3bu4xSmlwom1uDh7znnRz7JvdPUYpVa9/kczq0lANrdxjlFL99S843jI1RjPzL5NF\nnflHnA8tvWfoPUYp5XGtcbk0RrndYzRzKc3MHtfgG8DLZvYVSddTSo/O2g+B7r9/8Gh73xxUrT+H\n486hhkXUXf+y51GnMlnUmVdE9l35vLtynEtsZmOUUvq7bRQSbXStE+0j/1jkH4v8Y5F/rNzyvyu6\nAAAAgFzQGLnxu/TRLvKPRf6xyD8W+cfKLX8aIwAAAEdj5HJb4+wa8o9F/rHIPxb5x8otfxojAAAA\nR2Pkclvj7Bryj0X+scg/FvnHyi1/GiMAAABHY+RyW+PsGvKPRf6xyD8W+cfKLX8aIwAAAEdj5HJb\n4+wa8o9F/rHIPxb5x8otfxojAAAAR2Pkclvj7Bryj0X+scg/FvnHyi1/GiMAAABHY+RyW+PsGvKP\nRf6xyD8W+cfKLX8aIwAAAEdj5HJb4+wa8o9F/rHIPxb5x8otfxojAAAAR2Pkclvj7Bryj0X+scg/\nFvnHyi1/GiMAAABHY+RyW+PsGvKPRf6xyD8W+cfKLX8aIwAAAEdj5HJb4+wa8o9F/rHIPxb5x8ot\nfxojAAAAR2Pkclvj7Bryj0X+scg/FvnHyi1/GiMAAABHY+RyW+PsGvKPRf6xyD8W+cfKLX8aIwAA\nAEdj5HJb4+wa8o9F/rHIPxb5x8otfxojAAAAR2Pkclvj7Bryj0X+scg/FvnHyi1/GiMAAABHY+Ry\nW+PsGvKPRf6xyD8W+cfKLX8aIwAAAEdj5HJb4+wa8o9F/rHIPxb5x8otfxojAAAAR2Pkclvj7Bry\nj0X+scg/FvnHyi1/GiMAAABHY+RyW+PsGvKPRf6xyD8W+cfKLX8aIwAAAEdj5HJb4+wa8o9F/rHI\nPxb5x8otfxojAAAAR2Pkclvj7Bryj0X+scg/FvnHyi1/GiMAAABHY+RyW+PsGvKPRf6xyD8W+cfK\nLX8aIwAAAEdj5HJb4+wa8o9F/rHIPxb5x8otfxojAAAAV6oxMrMHzOyLZvY7ZvaTTRcVIbc1zq4h\n/1jkH4v8Y5F/rNzyn9kYmdldkj4u6X5J3yvp/Wb21qYLAwAAaFuZb4zeIelLKaXfTSkdSfolSe9r\ntqz2ja5xHjy8p6cuX9VTl6/q4OG9ia8VGd+uaL+9PenqVR3/PI+ydZTZr+pYRfVXHX+4zb97419e\nuK5F6j+P5jnOKvO/zs+pzrHKzs+qg1WdP9P2m/cei6bPkabfs+n6573uzpr/Vd8zQpk6irKIvC7m\ndo+RUkpTH5J+WNLPjTz/gKSfKdguLbP9/f2UUko3d3bTHa2mJKUkpTtaTTfevXPmtZs7u2fGGN/3\nFV1I39SFU/s9fm03rQ432VZaXU1p9+xQhYpqK6qjzH5lj2nc7m46U//j16qNP1rX/oJ1lVVUf9n8\nl8m8xznv/K/zc6o6r4uUnZ+lxh8ZTNtKaXWwX5X5M+vzGOZfRp15Rbxn0/UXjT/rujtt/le9xjb9\nmVSto2gu7uzEXhfnmf918r7lbN9T9OKpDaS/XdAYfaxgu3aPqCEHa5vHk2r4+LrWzrx2sLZZat/x\nx42VzZOn20pSSptnhypdW1EddR7TuM3Robz+GyvVxq+zrrKK6i+b/zKpepxlP5M6P6eq87pI2flZ\navyRwbR9sl+VXOucd3XmFfGeTddfNH7Z626d19imP5OqdRTNxbW1s6+dx+viuEmN0UqJL5VekPTm\nked3S/pa0YZbW1u6cuWKJOnSpUtaX18/vqlq+FVZ7s8v+rH0/M+NCc8Pjvp6udc7tf/to77um7H/\niZ70xLWTZyXqKxr/Ytn9x+r5nI70xrH6To0/YbxTe4zWX2H8ouN5+4Tx6vp8J9afyfyr63m/P3K8\nveuSeur3T45/0flfdf6UnZ9F51eZ8crOz1Lj9/vH2197YjDGMJ9550/Vz6PpvMo+Pzjq62Xl93mX\nHf/Wa31/dWNq/hPnf9v51P37a2z74y161yVJR0cje9QxPzN9Pvz58PBQUxV1S6MPSa+T9D8lvUXS\nBUm3JH1PwXZtN3u1CllKU1rupTSvf2mX0ubMf5nMe5znciltyvycdyltONippbQ55s+sz4OltBaX\n0qbk37mlNM/i1FJawHVx6ZbSBvvqAUm3JX1J0kcmbNPuEdVs9IO5ubObDtY208Ha5vGkKnqtyPh2\nRfvt7g6+ptzcnH/yla2jzH5Vxyqqv+r4w21+9uJ9C9e1SP3n0TzHWWX+1/k51TlW2flZdbCq82fa\nfvP+Ymj6HGn6PZuuf97r7qz5X/U9I5SpoyiLyOtibo2RDf5ucWaW6hoLAACgSWamlJKNv85/+RoA\nAMDRGLnRm7PQPvKPRf6xyD8W+cfKLX8aIwAAAMc9RgAAoHO4xwgAAGAGGiOX2xpn15B/LPKPRf6x\nyD9WbvnTGAEAADjuMQIAAJ3DPUYAAAAz0Bi53NY4u4b8Y5F/LPKPRf6xcsufxggAAMBxjxEAAOgc\n7jECAACYgcbI5bbG2TXkH4v8Y5F/LPKPlVv+NEYAAACOe4wAAEDncI8RAADADDRGLrc1zq4h/1jk\nH4v8Y5F/rNzypzFyt27dii6h08g/FvnHIv9Y5B8rt/xpjNxLL70UXUKnkX8s8o9F/rHIP1Zu+dMY\nAQAAOBojd3h4GF1Cp5F/LPKPRf6xyD9WbvnX+s/1axkIAACgBUX/XL+2xggAAGDZsZQGAADgaIwA\nAAAcjZEkM3vAzL5oZr9jZj8ZXc95Z2Z3m9kNM3vOzJ41s5/w17/VzP6zmd02sz0ze1N0reeVmd1l\nZk+b2ZP+/IqZ/YZn/4tmthJd43lmZm8ys0+b2W+b2W+Z2V9l/rfDzP6Rmf0PM/uCmf0HM7vA/G+W\nmX3CzF40sy+MvDZxvpvZz5jZl8zslpmtt11v5xsjM7tL0scl3S/peyW938zeGlvVufeqpA+llN4m\n6Qck/bhn/hFJ/yWl9Bcl3ZD0TwNrPO8+KOm5kef/StJHPfuXJP29kKq642OSfi2l9D2S7pH0RTH/\nG2dm3yHpH0q6N6X0dkkrkt4v5n/THtXgd+yowvluZu+R9N0ppT8v6R9I+tk2C5VojCTpHZK+lFL6\n3ZTSkaRfkvS+4JrOtZTS/04p3fKf/1DSb0u6W4PcP+mbfVLS34yp8Hwzs7slvVfSz4+8/Dck/bL/\n/ElJf6vturrCzC5K+usppUclKaX0akrpG2L+t+V1kl7v3wqtSvqapHeJ+d+YlNLnJP3+2Mvj8/19\nI69/yvfLdf86AAACnUlEQVT7TUlvMrNvb6POIRoj6c9I+urI8xf8NbTAzK5IWpf0G5K+PaX0ojRo\nniT9qbjKzrV/I+mfSEqSZGaXJf1+Suk1//sXJH1HUG1d8F2S/q+ZPerLmT9nZt8i5n/jUkpfk/RR\nSV+R9HuSviHpaUkvMf9b921j8/3b/PXx38m/p5Z/J9MYSWf+GwbyXxholpm9QdJnJH3Qvzki94aZ\n2YOSXvRv7IZz33T2POCzaM6KpHsl/duU0r2S7miwrEDmDTOzSxp8I/EWDZqf10t6T8GmfBZxwn8n\n0xgN/tfBm0ee363BV6tokH+N/RlJv5BS+hV/+cXhV6Zm9qcl/Z+o+s6xd0p6yMy+LOkXNVhCe0SD\nr6uH1wPOgWa9IOmrKaUDf/7LGjRKzP/mvVvSl1NK/ZTSH0l6QtJfk3SJ+d+6SfP9BUl/dmS71j8P\nGiPppqQ/Z2ZvMbMLkv6OpCeDa+qCfy/puZTSx0Zee1LSlv98TdKvjO+ExaSUfiql9OaU0ndpMNdv\npJQ+IGlf0o/4ZmTfIF8++KqZ/QV/6Yck/ZaY/234iqTvN7M/YWamk+yZ/80b/2Z6dL5v6STzJyX9\nqCSZ2fdrsMz5YjslDvBfvtbgn+tr8K9E7pL0iZTSvwwu6Vwzs3dK+nVJz2rwFWmS9FOS/ruk/6TB\n/1r4iqQfSSnl9X+7fI6Y2Q9K+scppYfM7Ds1+IcH3yrp85I+4P8YAQ0ws3s0uPn9j0n6sqQf0+Cm\nYOZ/w8zsugb/o+BIg7n+9zX4VoL53xAze1zShqTLkl6UdF3SZyV9WgXz3cw+LukBDZaZfyyl9HSr\n9dIYAQAADLCUBgAA4GiMAAAAHI0RAACAozECAABwNEYAAACOxggAAMDRGAEAADgaIwAAAPf/Ac3I\nzlk+B1d9AAAAAElFTkSuQmCC\n"
                    }, 
                    "output_type": "display_data"
                }
            ], 
            "execution_count": 14, 
            "source": "# Calculate the final score for the model\nscore_model(X_train, y_train, X_test, y_test, best_learning_rate, best_reg)"
        }, 
        {
            "metadata": {}, 
            "cell_type": "markdown", 
            "source": " # 70% of samples correctly classified"
        }
    ], 
    "nbformat": 4, 
    "nbformat_minor": 0
}