LDA_&_QDA.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "LDA & QDA.ipynb",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "id": "FiD45DoKN6VY"
      },
      "source": [
        "import pandas as pd\n",
        "import numpy as np\n",
        "\n",
        "from sklearn.discriminant_analysis import LinearDiscriminantAnalysis, QuadraticDiscriminantAnalysis\n",
        "from sklearn.metrics import confusion_matrix, classification_report, precision_score, accuracy_score\n",
        "from sklearn.metrics import accuracy_score, confusion_matrix\n",
        "from sklearn import metrics\n",
        "\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "from scipy import linalg\n",
        "from sklearn.manifold import TSNE\n",
        "import matplotlib.pyplot as plt\n",
        "import matplotlib.patches as mpatches\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "import seaborn as sns\n",
        "from sklearn import preprocessing \n",
        "from sklearn.feature_selection import chi2, SelectKBest\n",
        "from sklearn.metrics import accuracy_score, confusion_matrix\n",
        "from matplotlib.pyplot import figure\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "\n",
        "%matplotlib inline"
      ],
      "execution_count": 20,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ecxFkC2jN8a9"
      },
      "source": [
        "\n",
        "def data_cleaner(df):\n",
        "    data = df.copy()\n",
        "    #remove the id, partlybad columns\n",
        "    data.drop([\"id\",\"partlybad\"],1,inplace=True)\n",
        "    #set new index using the date column\n",
        "    data.set_index([\"date\"],inplace=True)\n",
        "    #add binary 'class2' col for event/non-event\n",
        "    data[\"class2\"] = np.where(data[\"class4\"] == \"nonevent\", \"nonevent\",\"event\")\n",
        "    #arrange to put class2 col at the front\n",
        "    cols = ['class2'] + [col for col in data if col != 'class2']\n",
        "    data = data[cols]\n",
        "    return data\n",
        "\n",
        "##### New way to generate the data\n",
        "\n",
        "def data_generator(df):\n",
        "    '''\n",
        "    input:\n",
        "        df: dataframe to split stratified by variable class\n",
        "        class: string with variable to use as response and as variable to stratify by\n",
        "    output:\n",
        "        X_train, y_train: variables and response 60% data \n",
        "        X_val, y_val: variables and response 20% data \n",
        "        X_trainval, y_trainval: variables and response 80% data (for cross validation)\n",
        "        X_test, y_test: variables and response 20% data \n",
        "    '''\n",
        "    \n",
        "    X = df.drop([\"class2\",\"class4\"], axis=1)\n",
        "    y = df['class4']\n",
        "    \n",
        "    # 60%, 20%, 20%\n",
        "    # Generate TrainVal 80% and test 20% \n",
        "    X_trainval, X_test, y_trainval, y_test = train_test_split( X, y, test_size=0.20, random_state=42, stratify=y)\n",
        "    \n",
        "    # Furthermore split train into Train and Val\n",
        "    X_train, X_val, y_train, y_val= train_test_split( X_trainval, y_trainval, test_size=0.25, random_state=42, stratify=y_trainval)\n",
        "    \n",
        "    return X_train, X_val, X_trainval, X_test, y_train, y_val, y_trainval, y_test\n",
        "\n",
        "def convert_binary(y):\n",
        "    '''\n",
        "    Parameters\n",
        "    ----------\n",
        "    y : repsonse variable with 4 classes\n",
        "    Returns\n",
        "    -------\n",
        "    y : reponse variable 2 classes\n",
        "    '''\n",
        "    y_bin = np.where(y == \"nonevent\", \"nonevent\",\"event\")\n",
        "    return y_bin\n",
        "\n",
        "url = \"https://raw.githubusercontent.com/williwilliams3/TermProjectIML/master/data/train.csv\"\n",
        "df = pd.read_csv(url)\n",
        "df = data_cleaner(df)\n",
        "\n",
        "X_train, X_val, X_trainval, X_test, y_train, y_val, y_trainval, y_test = data_generator(df)\n",
        "\n",
        "# If needed conversion from 4 classes to binary classes\n",
        "y_train_bin = convert_binary(y_train)\n",
        "y_val_bin = convert_binary(y_val)\n",
        "y_trainval_bin = convert_binary(y_trainval)\n",
        "y_test_bin = convert_binary(y_test)\n"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "bGmRe4M1OJXH",
        "outputId": "2451e602-4a76-49b0-91f2-898bfcda34d2"
      },
      "source": [
        "\n",
        "lda = LinearDiscriminantAnalysis()\n",
        "model = lda.fit(X_train, y_train)\n",
        "\n",
        "print(model.priors_)"
      ],
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[0.22077922 0.07142857 0.21428571 0.49350649]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "HMDP9UyiOkge",
        "outputId": "ddf62007-5f95-4d1d-b81c-8d657b2b1fdc"
      },
      "source": [
        "y_pred_test =model.predict(X_test)\n",
        "print(np.unique(pred, return_counts=True))\n"
      ],
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(array(['II', 'Ia', 'Ib', 'nonevent'], dtype='<U8'), array([10,  4, 20, 18]))\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 490
        },
        "id": "z7EqcEaqOpZ8",
        "outputId": "6d0c79e1-0c84-4aeb-d965-b9ecc708a421"
      },
      "source": [
        "print(classification_report(y_test, pred, digits=3))\n",
        "confusion_matrix = pd.crosstab(y_test, y_pred_test, rownames=['Actual'], colnames=['Predicted'])\n",
        "sns.heatmap(confusion_matrix, annot=True)\n",
        "\n",
        "print('Accuracy on test set: ', metrics.accuracy_score(y_test, y_pred_test))"
      ],
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "          II      0.200     0.182     0.190        11\n",
            "          Ia      0.000     0.000     0.000         4\n",
            "          Ib      0.300     0.545     0.387        11\n",
            "    nonevent      0.722     0.500     0.591        26\n",
            "\n",
            "    accuracy                          0.404        52\n",
            "   macro avg      0.306     0.307     0.292        52\n",
            "weighted avg      0.467     0.404     0.418        52\n",
            "\n",
            "Accuracy on test set:  0.40384615384615385\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "TDcQ0AdAOr_d",
        "outputId": "d8b57050-3d1b-4aa1-b0fe-0a81c3bd9f0c"
      },
      "source": [
        "qda = QuadraticDiscriminantAnalysis()\n",
        "model2 = qda.fit(X_train, y_train)\n",
        "print(model2.priors_)"
      ],
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[0.22077922 0.07142857 0.21428571 0.49350649]\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/sklearn/discriminant_analysis.py:691: UserWarning: Variables are collinear\n",
            "  warnings.warn(\"Variables are collinear\")\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 494
        },
        "id": "a7kxHlYtOx_w",
        "outputId": "3f1684fe-f656-4ebc-eb1f-fd5ff09de843"
      },
      "source": [
        "y_pred_test=model2.predict(X_test)\n",
        "\n",
        "print(classification_report(y_test, pred, digits=3))\n",
        "confusion_matrix = pd.crosstab(y_test, y_pred_test, rownames=['Actual'], colnames=['Predicted'])\n",
        "sns.heatmap(confusion_matrix, annot=True)\n",
        "\n",
        "print('Accuracy on test set: ', metrics.accuracy_score(y_test, y_pred_test))\n"
      ],
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "              precision    recall  f1-score   support\n",
            "\n",
            "          II      0.200     0.182     0.190        11\n",
            "          Ia      0.000     0.000     0.000         4\n",
            "          Ib      0.300     0.545     0.387        11\n",
            "    nonevent      0.722     0.500     0.591        26\n",
            "\n",
            "    accuracy                          0.404        52\n",
            "   macro avg      0.306     0.307     0.292        52\n",
            "weighted avg      0.467     0.404     0.418        52\n",
            "\n",
            "Accuracy on test set:  0.4423076923076923\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AO9_RSUxO5Zp"
      },
      "source": [
        "def select_n_best_feature(X, y, n):\n",
        "    # Normalize X values to be between 0 and 1\n",
        "    x = X.values\n",
        "    min_max_scaler = preprocessing.MinMaxScaler()\n",
        "    x_scaled = min_max_scaler.fit_transform(x)\n",
        "    X = pd.DataFrame(x_scaled)\n",
        "\n",
        "    # Use SelectKBest class to extract top n best features\n",
        "    bestfeatures = SelectKBest(score_func=chi2, k=n)\n",
        "    fit = bestfeatures.fit(X,y)\n",
        "    dfscores = pd.DataFrame(fit.scores_)\n",
        "    dfcolumns = pd.DataFrame(X.columns)\n",
        "\n",
        "    # Concat two dataframes for better visualization and print best features\n",
        "    featureScores = pd.concat([dfcolumns,dfscores],axis=1)\n",
        "    featureScores.columns = ['Feature_index','Score']\n",
        "    #print(featureScores.nlargest(n,'Score'))\n",
        "\n",
        "    # Filter X. Leave only top n features\n",
        "    #X = bestfeatures.fit_transform(X, y)\n",
        "    \n",
        "    # Return X\n",
        "    #return X\n",
        "    return bestfeatures"
      ],
      "execution_count": 26,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "fIJqNyzAP6eI",
        "outputId": "1aea49a6-288c-43d7-d365-f08440280a2a"
      },
      "source": [
        "#normalise data\n",
        "\n",
        "def normalize_0to1(X):\n",
        "    # Normalize X values to be between 0 and 1\n",
        "    x = X.values\n",
        "    min_max_scaler = preprocessing.MinMaxScaler()\n",
        "    x_scaled = min_max_scaler.fit_transform(x)\n",
        "    return pd.DataFrame(x_scaled)\n",
        "\n",
        "X_train_norm = normalize_0to1(X_train)\n",
        "X_val_norm = normalize_0to1(X_val)\n",
        "X_test_norm = normalize_0to1(X_test)\n",
        "\n",
        "acc_rate = np.zeros(100)\n",
        "\n",
        "for i in range(1,100): \n",
        "    bestFeaturesFilter = select_n_best_feature(X_train, y_train, i)\n",
        "    X_train_filter = bestFeaturesFilter.fit_transform(X_train_norm, y_train)\n",
        "    X_val_filter = bestFeaturesFilter.fit_transform(X_val_norm, y_val)\n",
        "    X_test_filter = bestFeaturesFilter.fit_transform(X_test_norm, y_test) \n",
        "    \n",
        "    lda = LinearDiscriminantAnalysis()\n",
        "    model = lda.fit(X_train_filter, y_train)\n",
        "\n",
        "    y_pred_val=model.predict(X_val_filter)\n",
        "\n",
        "#    print('Accuracy on test set: ',metrics.accuracy_score(y_test_bin, y_pred_test))\n",
        "    acc_rate[i] = metrics.accuracy_score(y_val, y_pred_val)\n",
        "#    print('Accuracy on train set: ',metrics.accuracy_score(y_train_bin, y_pred_train))\n",
        "    print('for  k = %i' %i, acc_rate[i])\n",
        "\n",
        "f, ax = plt.subplots(1)\n",
        "plt.plot(acc_rate,color=\"green\")\n",
        "ax.set_xlim(1, 100)"
      ],
      "execution_count": 31,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "for  k = 1 0.6346153846153846\n",
            "for  k = 2 0.6153846153846154\n",
            "for  k = 3 0.6538461538461539\n",
            "for  k = 4 0.5384615384615384\n",
            "for  k = 5 0.5\n",
            "for  k = 6 0.46153846153846156\n",
            "for  k = 7 0.5576923076923077\n",
            "for  k = 8 0.5192307692307693\n",
            "for  k = 9 0.23076923076923078\n",
            "for  k = 10 0.07692307692307693\n",
            "for  k = 11 0.09615384615384616\n",
            "for  k = 12 0.19230769230769232\n",
            "for  k = 13 0.21153846153846154\n",
            "for  k = 14 0.4423076923076923\n",
            "for  k = 15 0.17307692307692307\n",
            "for  k = 16 0.057692307692307696\n",
            "for  k = 17 0.15384615384615385\n",
            "for  k = 18 0.038461538461538464\n",
            "for  k = 19 0.36538461538461536\n",
            "for  k = 20 0.17307692307692307\n",
            "for  k = 21 0.19230769230769232\n",
            "for  k = 22 0.25\n",
            "for  k = 23 0.2692307692307692\n",
            "for  k = 24 0.23076923076923078\n",
            "for  k = 25 0.2692307692307692\n",
            "for  k = 26 0.25\n",
            "for  k = 27 0.36538461538461536\n",
            "for  k = 28 0.23076923076923078\n",
            "for  k = 29 0.38461538461538464\n",
            "for  k = 30 0.3269230769230769\n",
            "for  k = 31 0.057692307692307696\n",
            "for  k = 32 0.1346153846153846\n",
            "for  k = 33 0.36538461538461536\n",
            "for  k = 34 0.3076923076923077\n",
            "for  k = 35 0.057692307692307696\n",
            "for  k = 36 0.1346153846153846\n",
            "for  k = 37 0.15384615384615385\n",
            "for  k = 38 0.1346153846153846\n",
            "for  k = 39 0.28846153846153844\n",
            "for  k = 40 0.2692307692307692\n",
            "for  k = 41 0.23076923076923078\n",
            "for  k = 42 0.28846153846153844\n",
            "for  k = 43 0.3076923076923077\n",
            "for  k = 44 0.28846153846153844\n",
            "for  k = 45 0.28846153846153844\n",
            "for  k = 46 0.1346153846153846\n",
            "for  k = 47 0.057692307692307696\n",
            "for  k = 48 0.23076923076923078\n",
            "for  k = 49 0.17307692307692307\n",
            "for  k = 50 0.23076923076923078\n",
            "for  k = 51 0.17307692307692307\n",
            "for  k = 52 0.23076923076923078\n",
            "for  k = 53 0.4230769230769231\n",
            "for  k = 54 0.3076923076923077\n",
            "for  k = 55 0.4423076923076923\n",
            "for  k = 56 0.40384615384615385\n",
            "for  k = 57 0.34615384615384615\n",
            "for  k = 58 0.17307692307692307\n",
            "for  k = 59 0.34615384615384615\n",
            "for  k = 60 0.11538461538461539\n",
            "for  k = 61 0.15384615384615385\n",
            "for  k = 62 0.3269230769230769\n",
            "for  k = 63 0.3076923076923077\n",
            "for  k = 64 0.4230769230769231\n",
            "for  k = 65 0.36538461538461536\n",
            "for  k = 66 0.3269230769230769\n",
            "for  k = 67 0.4423076923076923\n",
            "for  k = 68 0.25\n",
            "for  k = 69 0.23076923076923078\n",
            "for  k = 70 0.4230769230769231\n",
            "for  k = 71 0.4423076923076923\n",
            "for  k = 72 0.4807692307692308\n",
            "for  k = 73 0.5\n",
            "for  k = 74 0.40384615384615385\n",
            "for  k = 75 0.25\n",
            "for  k = 76 0.11538461538461539\n",
            "for  k = 77 0.11538461538461539\n",
            "for  k = 78 0.07692307692307693\n",
            "for  k = 79 0.25\n",
            "for  k = 80 0.19230769230769232\n",
            "for  k = 81 0.1346153846153846\n",
            "for  k = 82 0.15384615384615385\n",
            "for  k = 83 0.21153846153846154\n",
            "for  k = 84 0.21153846153846154\n",
            "for  k = 85 0.15384615384615385\n",
            "for  k = 86 0.17307692307692307\n",
            "for  k = 87 0.3076923076923077\n",
            "for  k = 88 0.3076923076923077\n",
            "for  k = 89 0.3269230769230769\n",
            "for  k = 90 0.19230769230769232\n",
            "for  k = 91 0.19230769230769232\n",
            "for  k = 92 0.19230769230769232\n",
            "for  k = 93 0.28846153846153844\n",
            "for  k = 94 0.2692307692307692\n",
            "for  k = 95 0.25\n",
            "for  k = 96 0.28846153846153844\n",
            "for  k = 97 0.40384615384615385\n",
            "for  k = 98 0.17307692307692307\n",
            "for  k = 99 0.21153846153846154\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(1.0, 100.0)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 31
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAD4CAYAAAAAczaOAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO2deXxb5Znvv69ky5YtebdjZd8cCmGLSQOUtkDZsrTQO10Geju0vdMyc++lQAst9LYDLR2YUoYyZUo703Z6O7fL0GmnS9qEfW3ZmpBAQgiJAyFxgh073hfZkqX3/iG/ypF0JB1JR7Zkvd/PJ59YR0dHr+Wj5/zO73ne5xVSSjQajUZTGjhmewAajUajmTl00NdoNJoSQgd9jUajKSF00NdoNJoSQgd9jUajKSHKZuuNm5qa5NKlS2fr7TUajaYoeemll45LKZuzff2sBf2lS5eyffv22Xp7jUajKUqEEIdyeb22dzQajaaE0EFfo9FoSggd9DUajaaE0EFfo9FoSggd9DUajaaE0EFfo9FoSggd9DUajaaEKLig/2DHg3T0dcz2MDQajWZOUlBBfzQwygd/8UHuevau2R6KRqPRzEkKKug/9uZjBEIBjo4cne2haDQazZykoIL+1o6tAHSNdM3ySDQajWZuUjBBX0p5IuiP6qCv0Wg0+aBggv6uY7s4OnKUxbWL6R3rZSo8NdtD0mg0mjnHrAX9nrGemMdK5X/yjE8ikQnPazQajSZ3Zi3odw51su3otujjLR1bOMt3Fmt8awDt62s0Gk0+mFV757qHriMsw/T7+3n+yPNsbNuIz+MDtK+v0Wg0+WDWFlFBwAtHXuBnu35GmaOMsAyzqW0TPu900NdKX6PRaGxn1oJ+dXk1p8w/hZsfu5l1C9bRVNXE2vlrownc7tHu2RqaRqPRzFlmzd4RQnDfhvvoGu3id/t+x/qV63E6nFSUVdDgbtD2jkaj0eSB2Qv6CM5ZeA5/dfpfAbCpbVP0OZ/Hp4O+RqPR5IFZs3eEEADcc+k9LKpZxBUnXRF9zuf1aU9fo9Fo8oAlpS+EWC+E2CeEOCCEuCXJPh8VQrwmhNgjhPh52mMSCfrN1c3ccdEduMvd0ee00tdoNJr8kFbpCyGcwP3AJcARYJsQYrOU8jXDPm3Al4DzpJQDQogWC8dN+pzP46N7tBspZcr9NBqNRpMZVpT+OuCAlPJNKWUAeAC4Im6fzwD3SykHAKSUaafTKqVvRqunlUAowMDEgIXhaTQajcYqVoL+AqDT8PjI9DYjq4BVQohnhRAvCCHWpztoSqWva/U1Go0mL9hVvVMGtAEXAFcBPxBC1MXvJIS4RgixXQixfXJiMunB9KxcjUajyQ9Wgv5RYJHh8cLpbUaOAJullEEp5UFgP5GLQAxSyu9LKddKKde6K93xT0fRSl+j0Wjyg5Wgvw1oE0IsE0K4gCuBzXH7/JaIykcI0UTE7nkz1UHTJXJBK32NRqOxm7RBX0o5BVwLPAzsBf5TSrlHCHG7EOLy6d0eBvqEEK8BTwJfkFL2pTpuqkSux+WhqrxKt2LQaDQam7E0OUtKuRXYGrftVsPPEvj89D9LpFL6Qghdq6/RaDR5oGBWzopHz8rVaDQa+5nVhmup0Epfo9Fo7GdWG66lwufRSl+j0WjspmCVfqunlZHACGOBsRkakUaj0cx9ClfpT9fq6woejUajsY+CVfq6Vl+j0Wjsp+CVvhVffywwxu/3/d6WcWk0Gs1cpnCDfgZK/6e7fsrlD1zOS2+/ZMvYNBqNZq5SsPZOY1UjZY4yS0r/4OBBALZ0bLFlbBqNRjNXKVil7xAO5lXPo3ssfSK3czjS+Xlrx9Y0e2o0Gk1pU7BKH6zPyj08dBiAPx/9M71jvTmPTaPRaOYqBav0wfqs3MNDhzml+RQkkocOPGTH8DQajWZOMnu9dywsfWtlVm4oHOLo8FGuOOkK5lXP076+RqPRpKCglf6qxlX0jvfSOdSZdJ+u0S5CMsSS2iVsbNvIw288zFR4ys6hajQazZyhoD39DW0bgNQJWuXnL65dzMa2jQxODPJ85/P2DFKj0WjmGAWt9E9uOpkltUvYesBa0L9k+SWUOcp0FY9Go9EkoaCDvhCCTW2beOzNx5iYmjDdRwX9RbWLqK2s5d2L3619fY1Go0lCQds7ABvbNjIeHOeZQ8+YPn946DC1FbXUVNRE9l+5kd09u1PmATQajaZUKWilD3DhsgupLKtMatl0DneyuHZx9PGmVZsAePDAg7kPUqPRaOYYBa/0q8qruHDphUktm8NDh2OC/slNJ+MUzqjto9FoNJoTFLzSB9jUtokD/Qfo6OtIeC4+6AshqCqvYjw4bss4NRqNZi5R8EofIr4+JDZUGw2M0u/vjwn6gA76Go1GkwRLQV8IsV4IsU8IcUAIcYvJ858UQvQKIV6e/vdpOwe5rH4ZJzednODrq2TtoppFMdt10NdoNBpz0gZ9IYQTuB/YAJwCXCWEOMVk119IKc+c/vdDC8fNaKAb2zby9KGnGQ2MRrep7ppa6Ws0Go01rCj9dcABKeWbUsoA8ABwRa5vnImnDxFfPxAK8Nibj0W3GSdmGXGXu3XQ12hmgD09e1h872J2dO2Y7aFoLGIl6C8AjEXvR6a3xfMhIcQuIcSvhBCLTJ5HCHGNEGK7EGJ7f39/RgM9b/F5eF3eGIvn8NBhHMLBfO/8mH210tdo8o+Ukuseuo7O4U7+fPTPsz0cjUXsSuT+HlgqpTwdeBT4d7OdpJTfl1KulVKubWpsyugNXE4Xl664lK0dW5FSApGg7/P4KHeWx+yrg75Gk39+8/pveOLgEwC6RLqIsBL0jwJG5b5welsUKWWflHJy+uEPgbPSHTRTTx8ivv7RkaPs7tkNJE7MUuigr9HkF3/Qz42P3MipLaeyqGaRDvpFhJWgvw1oE0IsE0K4gCuBzcYdhBA+w8PLgb3pDpqppw+wYWWk6+aW/ZHSzfgafYUO+hpNfrnn+Xt4a/At7lt/H8vql+mgX0SkDfpSyingWuBhIsH8P6WUe4QQtwshLp/e7TohxB4hxCvAdcAn0x03G6Xv8/po97Wz9cBWwjJM51ASpV+mg75Gky+ODB/hH/70D3zo5A9x4bILWVSzKFpJpyl8LHn6UsqtUspVUsoVUso7prfdKqXcPP3zl6SUq6WUZ0gpL5RSvp7umNkofYg0VHuu8zn2Hd/HZGhSK32NZoa5+9m7CYVD3H3J3UCkeu7I8BFC4dAsj0xjhdlbLjFLNq3aRFiG+eGOyFSA+IlZoIO+RpMvpJRs3r+ZS1dcyrL6ZUAk6E+Fp+ge7Z7l0WmsUHRB/53z30lTVRM/fuXHQGKNPkSCfkiGCIaCMzw6jWZu8/rx13lr8C02tW2KblPfQe3rFwdFF/SdDifrV66n3x+p808W9AGt9jUam1H9r9RSpnDiblsH/eKg6II+RHx9iAT3BndDwvM66Gs0+WFrx1ZOazktRmypn3UytzgoyqB/2crLcAgHi2sXm1YB6aCv0djP8OQwfzz8x2jXW0VtZWTlOq30i4Oy2R5ANjS4G7h0xaXUVdaZPq+DvkZjP4++8ShT4akYP1+xuHaxDvpFQlEGfYDfXfk7HML8RkUHfY3GfrZ2bKW2opZzF52b8JwO+sVDUdo7EOnFU+Ywv2a5y92ADvoajV2EZZitB7Zy2crLTL93eoJW8VC0QT8VWulrNPbycvfLdI92m1o7EFH6x8eP6+9cEaCDvkajScuW/VsQCNavXG/6fLSCZ0ir/UJHB30LvHDkBZ48+KQtx9LMbX7w0g94e+Tt2R5GUo4MH+Enr/wk49dt6djCOxe8k5bqFtPn9QSt4kEHfQvc9tRtfOHRL9hyLM3cpWesh2v+cA3ffPabsz2UpHz/pe9z9W+vZiwwZvk1T731FC8efZEPn/zhpPvoCVrFgw76FhieHI5Zm1ejMaNnrAcgZnW3QkPdhQxNDlnaPxQOcf1D17OkdgnXrrs26X4LahYgEDqZWwTMyaDvLrO3emdkcoSxoHVlpClNesd6Aejo76Cjr2OWR2NO12gXAEMT1oL+D3b8gF3HdvGPl/5jtCrODJfThc/r00q/CJiTQb/cWU65o9y2oD8aGNVJYU1aesd7oz8XqtrvGpkO+haU/oB/gK888RXOX3I+Hzr5Q2n317X6xcGcDPpgb3vlkcBIRh6opjRR9k5LdUu0MVmhodofW1H6tz11GwMTA9y34T5Lix7pZROLgzkd9P1TfluONRoYZTI0qReJ0ADw3W3f5b9e+6+E7b1jvQgEV516FU8fenpG80Bb9m/hG3/6Rsp9QuEQx8aOAemV/v6+/Xx323f5m7P+htPnnW5pDItrF9M53ImU0tqgNbPCnA76dij9QChAIBQAdN2/JsK9L9zLv7z0Lwnbe8d7aXA3cPlJlxMIBXj8zcdnbEw/evlH3PbUbSnXkOgd7yUsw0CkOCEVLx55kZAM8dl1n7U8hsW1i5mYmuD4+HHLr9HMPDrop8Go1nQyVwMRa0R540Z6xnporm7m3YvfjdflnVFfv2uki0AowGu9r6XcR5HO3lF3Ao1VjZbHoGv1iwMd9NNgDPpa6WsgEhBVFYyR3vFeWqpbcDldXLLiErYe2DpjVofy6nd270y6j3HM6ewddVGorai1PAZdq18czNmg7y532xKkRyZHoj/rZK5mYmqCQChAv7+fyanJmOd6x3pprmoGIgv9HBk+wu6e3Xkfk5QyGtB3dO1Iup9xDVsrSr/CWUFFWYXlcWilXxxYCvpCiPVCiH1CiANCiFtS7PchIYQUQqy1b4jZoe0dTT4weuEqKaroHT8R9NVyglv257+KZ2hyiImpCSB10Ff2Tkt1S1qlPzw5TG2ldZUP0FTVRGVZpZ6gVeCkDfpCCCdwP7ABOAW4Sghxisl+XuB64EW7B5kN2t7R5AOjQjZ65KFwiL7xPpqrI0F/vnc+a1rX8OCBB/M+JjWO5qpmXu5+OZqsTdhvtIu6yjrmVc9Lb+9MDlFTUZPROIQQumwzC8IyzCU/uYQHO/J/roA1pb8OOCClfFNKGQAeAK4w2e/rwF3AhI3jyxq7gv5IQNs7mhMYg6XRI+/z9yGRMQ3J3rvkvWx/e3veS33VODa0bWAsOJZ0NnDXaBc+j4/aytr09s7EUEZ+vmKeZ150voLGGr1jvTz25mM82/nsjLyflaC/ADDerx2Z3hZFCNEOLJJSpryXFUJcI4TYLoTY3tvbm2rXnKkq0/aOxn6SKX3VgkHZOwDtvnb8U3729e3L65jUODaujKxdmyyZ2zXShc/ro7ai1pLSz9TeAWh0N9Lv78/4daWMumjP1LyOnBO5QggH8C3gxnT7Sim/L6VcK6Vc29zcnG73nLBN6RsSudre0RiDpTExqlowKHsHYE3rGiC1z24HahzvW/Y+XE5X0vfrGu2i1dNKTUVN3pR+o7uRPn9fxq8rZdRFu5CC/lFgkeHxwultCi9wKvCUEOIt4Bxg82wnc1XQz7VkLkbpa3un5IlR+gZ7x9iCQXFy88lUllWysyt5GaUddI12UVlWSVNVE6e1nGYa9KWUdI92R+yditq0k7OyVfoN7gb6xvv0rNwMUOeR0UrOJ1aC/jagTQixTAjhAq4ENqsnpZRDUsomKeVSKeVS4AXgcinl9ryM2CJV5VVIJJOhyfQ7p2CuJXKnwlNJE32a9Cilv6R2SUzQN7N3yhxlnD7vdHZ051fpK69eCEG7r52d3TsTgq6q8Il6+hbq9LNS+lWNTIYmbWuBUgoUnNKXUk4B1wIPA3uB/5RS7hFC3C6EuDzfA8wWu3rqjwRGqC6vBorf05dSsvzby/nOn78z20MpWpTSX9W4KtbTn7Z34mewrmldw86uxCBsJ8qrV+/X7+9PqKBRY1WefiAUiJZ5xhMKhxgJjGRt7wD0jWuLxypRpT9ZOEofKeVWKeUqKeUKKeUd09tulVJuNtn3gtlW+WBf0B8NjOKt8FJVXlX09s5oYJTO4U4ee/Ox2R5K0TI0OUR1eTULaxYm2DsN7gbKHGUx+7f72hmaHOLg4MG8jUkpffV+kJhHUGNVSh+ST9BSNkO29g6gk7kZoHIyBaP0ixU7lb7X5aW6vLro7R2VYMt3YnEuMzw5TE1FDT6Pj2Ojx6LlmKoFQzwqCOfT1+8e7abV0wrAafNOwyEcCRU8Sum3elqjCj6ZxZNNCwaFutPRyVzrFF31TqGigr4/mJu3OBoYxePyUO2qLnp7R6mvoyNHdS11lqgEp8/rIyRD0eBmbMFg5NSWU3EKZ94utP6gn8GJwajSryqv4uSmkxPeT6lJnze90ldJ3kwnZ8EJpa/tHeuoC3IhJXKLkrzYO0Ue9I1fxHxXlMxVVIJTBVn1he0d740p11RUllWyumV13pK5xmCuUMlcI6rCp7aiNr3Sn96ebZ0+aHvHKsa+SVrp54ht9s7kSETp52jvvD3y9oz1GT82eoxjo8cSthu/iJkoz4mpCfb37bdlbMWOUenDiVvznrEeU6UPkSC8o2tHXpK5Rq9esaZ1DW+PvB0zj8BY4aMUfDKln4u9E1X62t6xhKqqqi6vZjQwOiOVdXM26KtFnO1Q+lF7J4dE7lX/dRWffdD6ghS58InffoJP/u6TCdvVF7GmoiZlC954/m3Hv3Ha907T6o0TSl956F0jXdG+O2aePkSCcM9Yj2k75lwxVuUo1s6PTJF54cgLMfupfaL2Th6UvrvcjbvMrc8Vi6gLc1tjGzAzZeFzNujbncjN1d7pGevh6PDR9DvawIH+A7w1+FbCdmXvXLD0goyUfvdoN4FQQFtCTCt9o70z2kW/vx+JTKn0IT+Wmgoa6iIEcPbCs/G6vDENvIwVPkrBJ5uglYvSh0gyVyt9a6iL9qrGVcDMWDw66KchqvRztHf8QX/aCTF2oDxCs0Rtv78fr8vL2QvO5o2BNywtjg0nEkyZ3B3MVYYmIvaOu9xNbUUt3aPdpi0YjJwx7wwEIi/J3K7RLhzCEXPBMVvExVjhk9beyUHpw3QrBp3ItYS6+2triCj9majV10E/BVLKSCLX5c3Z3vFP+S0H2VwYDYwyHhyn39/PVHgq5rk+fx8N7oao8ny5+2XLxwRd6hkMBfFP+aMK2Of1xVxgk9k73govbY1teUnmdo10Ma96Hk6HM2b7prZN0UVc4it8nA4nHpcnZclmmaMMd5k7qzE1uBu0vWMRpfRV0NdKPwfsCPr+KT9hGcbj8uTctXOmlH5Mu984tdXv76exqjHjRmA66EeIV8A+j4+ukS7TFgzxtPva82LvdI12xfj5ivUr1wOwtWOraYVPbUXy9srKwhJCZDUmbe9YR1VVLaiJNC6eibLNORv0lUrJJVCrYOet8OZcpz8xNcHw5HDeG1EZWwPEWzx9/j4a3Y3M88xjvne+ZbtGnYj7+/bPWFlZIaI88Hiln87egUgy99DQIVPb4+jwUR4+8HBWYzJ69UbUIi5bOraYVvik6r+TbbM1RUNlg7Z3LKL+fl6XF9BKPyecDicVzoqcgr7y1zwuD1XlVUxMTWS1IEYoHCIYDhKW4bz/UWOagI3HrlnQN94XLalTZYRWGA2MIhBIJK90v2LfYIsMpYyVJ95a3UrXyAl7R9Wom7G6eTWAaenrfS/exxUPXJGVIOgaMQ/6ELF4nut8jtd6XwNMlH6SoK9mHWdLY1Wkp77utJme7tFufF4fHpcH0EE/Z3Ltqa/+ACqRC2TVPdD4mnxbPGYLeyj6/f3RwNTe2s7e43stfT4jkyOc0XoGUNrJ3AR7x+vDP+XnjYE3aHA3UO4sT/pa1Z5gYGIg4bnj48eZDE0SCAUyGk8oHKJ3vDemcsfIxraNhGWY//fK/4uM13BxSNVTP9sOm4pGdyMhGUrbvllz4qLtrYgofZ3IzRG7gr5K5EJ2PfWNrSDy/UUwW9gDIutwDkwMRJX+Gt8awjLM7mO70x5zNDDKSY0n0VLdUtK+fnwpowqiu47tSunnA9RX1gMw4E8M+upCkKmf2zPWQ1iGTT19gHUL1tHobuSPh/+IQzhoqmqKPpdXe0c3XbOMsne00reJqvIqxqdysHcCsfYOZNde2djCNt8VPF2jXSzwLkAgYjz9oYkhwjIcVZzJujGaoSqY1rSuKe2gb6L0Afb27k3p5wPUVdYBMDgxmPCc2pbpF97MqzfidDijCd34Cp+Uidxclb5uumaJaFWVwd7RidwcsU3pV3ij9k42x5tRe2e0i4U1C2msaoyxd9QXUNk7i2oW0ehutGTXjAQirSjafe3s6d3D5FRuC9MUK8mUfjAcTKv0VdA3s3eiSj/DW3uz2bjxbGzbaLpPqtWzVPVOtuie+tYwTqxzOV24nC6t9HOlqrwqpy6bxkSuXfZO3pX+9HT75qrmGHtHfQHVrbcQgjW+NbzU9VLK44VlmLHAGN6KiNKfCk/xas+r+fsFCphkSh+S1+grKsoqcJe5c1L6hwYP8YVHvhC9c0yn9AEuW3EZDuFI2Ke2shb/lJ9gKBizXUrJ8ORwyds7wVCQWx67hY6+jry9R7SUdvpv43F5dNDPFTsTubnU/c+00vd5fDRXN8fYO+oLaFzZaWX9yoQVluIZD44jkVGlD6WbzB2aGKKyrBKX0wVE1HKFswJIXaOvqHfXm3v609vSfeH/sP8P/OPz/8g9z90DnAga8zzzkr6msaqRz5/zeT66+qMx25N12lRNv0rd3vnOn7/DXc/exW9e/03e3iN60Z4WD16XV9s7uWJrIjeHJRNnytOfnJqk399Pq6eVluqWWKXvj1X6ANWu9K0ljJ/B8vrl1FbUlqyvH297CCGiX9h0nj5ELJ7ByVilH5bhqM2S7guv7gju/NOdHBk+QtdIF/WV9VSWVaZ83d2X3s3VZ1wdsy1ZT/1cWzDAiaR1sSr9nrEevvb014DECjg7idpzWunbh7vcnVudfmCEMkcZLqfLPnsnj0r/2FiknbLPM23vjCXaO8ZactVPKFU7V+PdjhCCM1vPLO2gHxcM1Rc2nb0DkWAYr/SHJoaQROrZ033hByYGKHOUEQqHuOWxW5LOxrVCMqWfa7M1gHJnOTUVNUXr6X/lia8wFhzD6/LSM56/xYa6RrtwCmdUMHgrvLpkM1fsUPpelxchhH32Th6VvjGx11zVHNN/p9/fj0BEE4pA9EKWbIFsOJHXUHXE7b52dh3bldDXpxQwm7QUVfoW7J26yroET9/4ON0XfnBikOaqZm561038bPfP+OPhP6b081ORrOmaHUofppuuFaG9s6NrBz/c8UOuW3cdqxpX5V3pz/PMwyEiYVgrfRvItV+OqloBcrJ3ZkrpGxN7LdUtSGRUbfX5+6irrIsp24uWoaa4ezEqfYgEff+Un33H9+XldyhkzEoZVdC1Yu/Uu+sTqneMj60o/Xp3PV9695dY4F3A8fHj2Sv9JD31c1kq0UgxNl2TUnLdg9fRVNXEreffSnN1c8KsdjvpGu2KmVhXUEFfCLFeCLFPCHFACHGLyfN/K4TYLYR4WQjxJyHEKfYPNXPsUPoq2FkJkMlQSr/B3ZDfoG9U+tNBSJ20qtmaESsXMuNcBSDarG02krnf3fZd/vYPfztr0/tT2TuWlH5FGqVvwdOvq6yj2lXNNy/5Zsz7Z0rU3olX+jbYO1A8TddueOgG2v65jbZ/bmP5fct5tvNZ7rzoTmorayN5sTwq/e7R7pi/30wlcsvS7SCEcAL3A5cAR4BtQojNUsrXDLv9XEr5L9P7Xw58C1ifh/FmRFV5VbRTprqFygS1Pi7kthKXsk9aPa15nZHbNdqFQNBS3RINQuqkVc3WjCh7J9XvZEzkAiyuXQzEzvydCQ70H+BzD3+OQCjARcsu4iOrPzKj7w/mSv9jp30Mp8OZtBWCkXp3fXSSnDofjR5/WqXvH2C+dz4AV516FQcHDnLFO67I9NcAkit9O+2dN/rfyOkYM8EvX/slVeVVrFuwDoBPnPEJPnXmp4DIhdxsXQq76Brt4izfWdHHM6X00wZ9YB1wQEr5JoAQ4gHgCiAa9KWUxkhWDRREpyWlziemJqI/Z4JaHxfAIRxZr56l7J151fNM67Ttonu0m+bqZsocZVGlr05as+X8srF31GtmutvmjY/ciMvpYkX9Cm569CY2rdqU1d80F8wmLS2rX8Yt7064+TWlvrIeSaQOPn6GrkCk/UwHJwZZ3RJp3CaE4Mvv/XKmv0KUZKtn2aX0i8HekVLSO9bL58/9PN+4+BsJzzdXNeOf8jMWGIsKJLsIhUP0jPXE2HNeV+EkchcAnYbHR6a3xSCE+N9CiDeAbwLXmR1ICHGNEGK7EGJ7b2/+bpsUufbUV4lc4/FySeTO88zLu6cfX02Ss70Tl8h1Opy4y9w5LSiTKY+88Qib923mK+/5Ct/b9D0ODx3m7mfvnrH3h8iXdDQwmpMCjs7KNah75en7vL60t/YDEwPUVdSl3Mcq5c5y3GVu00SuQziiF/lsaXQ3MjgxmFVX2pliaHKIYDiYtPIq3iK1k2jfJIO943F5mAxNJkyYsxvbErlSyvullCuAm4GvJNnn+1LKtVLKtc3N6T3QXMk16BsTuRAJktkq/XJHOfWV9Xmv3lHKodHdiEDE2DsNlQ0x+1v5fOKVPpDz2gKZEAwFueGhG1jZsJIbzrmB85eez0dO+Qh3PXtX2olldhLfSz8b6t3TTdcMydsB/wBOEbGHUin9sAwzNDEUPYYdmDVdG5oYoqaiJusFVBQN7gYkMq93trmSbvEbdTHIh8Wjii7iE7mQ/7toK0H/KLDI8Hjh9LZkPAB8MJdB2YUdSj8h2GWhcCemJqJrqg5NDuUtEWlU+k6HkwZ3A73jvQRDQYYnhxOVvoW5ByOBEcod5dFZqGCP92j1M7h/2/3sPb6Xb136LSrKIrNf777kbiSSmx65ibHAWPRfPrHD6zZruqaSszUVNSlv7Ycnh5HImJLbXDHrqZ9r3x1FMczKTbf4TXxezCpWzm2z1czU3XQhBP1tQJsQYpkQwgVcCWw27iCEaDM83ATkr2FFBhSSveMuc1NbWctUeCqrnvzpCIVDHBs9FnO7qFoxKGWZkMi1YO8Yk9nG1+Wi9I+NHqPp7mC2sbMAACAASURBVCY279uccr9AKMDXnv4al624jPeven90+5K6Jdx83s388rVf4vkHT/TfV5/6atZjSocdXrdZe2VVhpnuQqouFOoYdmDWUz/XtsqKYmi6lk7pZ2Pv/POL/8xJ3zkprTev7lLj7R3If9BPm8iVUk4JIa4FHgacwI+klHuEELcD26WUm4FrhRAXA0FgAPhEPgdtlVyCfigcYjw4bo+9M+WnsqwypkzO7iRkn7+PkAzF3C6qVgzxzdYUVu2deH83V6V/oP8A/f5+rnvwOi5Zfkm0MiqePT17GJwY5FNnfirBbvg/7/k/LKxZGA2g//TiP/HKsfyt6mVH/Xoqpe91eVN+pur3tFXpJ7F37FD6xdB0Ld2C9tko/eePPE9Hfwd3/PEO0+Sw4omDTzDfOz9aDQcnKuTyXbZppXoHKeVWYGvctlsNP19v87hsIZegr4K7UeVWu6o5Nnos42P5g/6IvWMok8t2Uk0yzNrsNlc181rva6bN1sC6vWO821Gvy8VOUeM5NHSIu5+7m1vPv9V0P9XuQTV6M+Jyuvh0+6ejj7d0bMlrgLGlJ42Zpz8xQH1lROmn+rKr19jq6VfUcnQ41qkdnhyOloXmwlywdzwuD5VllRl5+krB3/vCvXy6/dOsbFiZsE8wFOThNx7mo6d8NEbMFJKnX7SooJ9Ne2VjW2Xj8bKt03eXuZNOiLEDsza7qs7YrNkazJ7SV+NZt2Ad3/jTN5ImZHd278Tr8rKiYUXaYza487sYtx32jsflwSEcpkrfqr0zI56+jfZOISv93rHeaGA3QwiR0KI8HZ3DnVyy/BJcThc3PnKj6T7PdT7H8ORwdK0DxUwtmVgSQT+bQG1atZKDvWNU+vmYoGWm9FuqW+j39ydduNshHFSWVaYt2YwP+rl6+ioQ/OADP0Ai+eKjXzTdb0fXDs5sPdPSxLpGd2PBK32HcFBXWRfr6fsjSl/ZO8ma36nX2Onp11Ymrp5ll71TW1mLQzgK29Mf7007kzqTVgxT4SmODh9l3YJ1/N17/47N+zbzyBuPJOy3pWML5Y5yLl5+ccx2rfRtIJdZtPEzUWE62GXZZTPG08+gVv/gwEHT7ZNTkzG35mYlYM3VzUhkdCGIeHsH0v9OZoncnJX+eB9O4eS0ltO4+byb+cWeX/DMoWdi9gmFQ7xy7BVTa8cMNe0/X5VRdk1aqq+sj2mvbFT6kPxczZfSHwuORWvppZS2Ve84hIP6yvqCtnd6xnrSdkfNpBVD10gXIRlice1irj/7+kiZ8UM3JNTdb+3YynuXvNf0ewU66OdELko/vueMOl6u1Ttg3d758cs/jvQDOfxswnNff+brtP1zW/Si0D3aTU1FTUyCWKmY1/tep8xRluDNw3RP/RTrCJvZO9le/BR9/j4a3A0IIfjieV9kYc1Cbn/69ph99vftZzw4Hu31k44GdwOBUCBv8weGJocod5Sn7V2fDqPS9wf9TIYmqXfXpy3XG5gYwCEcCYEiF+LvPP1TfqbCU7bYO1D4s3J7x3vTNsrLpBWDsikX1y6moqyCey+7l73H93L/tvuj+xwaPMSe3j1satuU8PqZSuTqoJ8E4/q4impXdbSXTyYY6/TBmtIfmhji5sduBuB3+36X8PxvX/8t/ik/Nz16ExBbo69QJ/Trx1+PBtl4qsqrMk7kKqWfrao2zg6uKq/iY6d+jKcPPR1je6VK4pqRbw95aCLidec6aaneXR9V7Ub1Hl0YO4mfq+4IsukhlYz489GuuxlFoTdd6x2zYO9k4Ol3DkcaFyyqiUxr2tS2ifUr1/PVp74avVvY2hGph4n38+FEYYVW+jmgVFlWSj9JIhcyTwz7gxGlry4gVpT+15/5Or1jvayoXxE9URRKLbQ1tPHrvb/miYNPxMzGVahb1zf630jw8xXp/HlTpe+qRiJT9uFPhVL6io1tG5kKT/HoG49Gt+3s3kllWSUnN59s6ZjqePnykO2yPeoq66KVONGKnGlPH1IrfTutHUjsqW9XszVFvvMsuSCltGTvNFc3Mx4ct3Rnq5T+otpI0BdCcO9l9zIWHOPLT0T6JG3p2MLy+uWsalyV8HqHcFBdXq0TubngEA7cZdmtnpUskQuZ99RXdfoO4cDr8qZV+vuO7+PbL36b/7Hmf/C/3vm/2NO7h0ODh6LPq4vALz/yS5bWLeX6h67nyPCRRKU/rWJCMpRQuaNIZVlNhaeYmJpILNnMYW0BiARm40XoXYveRW1FbczFbUfXDk5rOY0yh6Wq4ryXCNpV1WJcPSuanJ2enAXJb+0HJwZtTeJCYqdNu5V+viuqcmF4cphgOJhW6cf3sErF4aHD0dnVinc0vYPPrvssP9zxQ57rfI4nDj7BprZNSe8YZ6LT5pwO+pC9D2+ayM1yyUSl9MF8Qkw8n3v4c1SVV3HnRXdGvT9jQNzSsYUV9Ss4fd7p3HPpPbza8yqHhg4lBH1j4tYsiat+p2S/j9mFz/g425Oz398fcxEqd5Zz6YpL2XpgK2EZRkrJzu6dlq0dyL+9Mzw5bJvST2XvJFX6fvuVfnwJcT6UfqHaO+lq9BWZTNA6PHQ4ZrKV4tbzb6WpqokPPvBB/FN+U2tH4XF5GA3qoJ8TVeVVKROVyVCKy9hSNdscgSrZhOna6BT2zpb9W3jwwIPcdv5ttFS3sKpxFcvrl7OlY0vkWEE/Txx8go1tGxFC8N/e8d+4aNlFAAn2TpmjLBpckyn9VPaOmcUF2V/8FGa9/Te1baJ7tJuXu1/mrcG3GJwYtJzEhRmwd6YbkeVKfWU9k6FJ/EF/rL2TJpE7ODFo68QsSK707fg9ISI0RgOjBEKBjF63/e3tCZPG7CZdCwZFJq0YkgX9uso67rzoTnrHe3GXuTl/yflJjzET6+SWRtDPUum7y9wx9kI2tkZYhgmEApaV/ve2f48ltUu4dt21QMQX3NS2iScOPoE/6Oept57CP+WP3gEIIfj2+m/jdXk5reW0hOOp29Nknn6qz8csmQ25Kf2JqQnGg+MJdx4b2jYAkTuaTJO4UDz2jrEVQyaJXDvbKitaPa2UO8p5/ODjgD2dRI1k23/nL37xF3zxMfO5G3aRrgWDIhOl3zncGU3ixvOpMz/Fuxa9iw++44NJ246AtndsIdugbzopKQuFq5KdKqmcTum/1PUS5y89P6ar5ca2jfinIgF/a8fWiFpYekItrG5ZTf/N/Vy28rKE46mTNmUiN0N7JxdPX9kv8XceLdUtvHP+O9nSsYWd3TsjdfzzEi9iyXA5XXhcnvxW79gQDI2tGIz9dNIlcvOh9GsqavjcOZ/jxy//mG1Ht9lu72TTsExKybGxY2x/e7stY0iGVXvHanvl0cAo/f5+U6UPka63T3/yaX72Fz9LeZyZWDJRB/0kjAYTq1aysXdUpU/U3qmsTTojt2uki+7R7gRb44KlF+Auc7OlYwtbOrZw0fKLEurFkyU81Umd1N5xVSf9fdTJZ1ayCdkpfaX6zC5Cm9o28eKRF3nkjUc4pfmUjGviG9wNeVH6YRm21dOHE0q/uryacmd5VFCYfeEnpiaYmJqw3dMH+PJ7v8y86nlc99B10TsPs/kc2ZBNw7KJqQkCoQAdfR15tTms2jsel4cKZ0XaC1fnUKRcM1nQh8h3NF3Jr1b6NpCLvWPWUhgyU7iqjXLU3jHpd6JQi43H2xqVZZVctPwifrrrpxwcPGg6sSMZUaWfJJGrPh+zmvukSj8HTz9Z8zeI3NFIJNve3paRtaPIV4ngaGAUibStegciiVnVVhkidyoup8v0C5+PFgyKmooavnHxN3jhyAv8+yv/jtflxelw2nLsbJS+ynNIZF67pvaO91JdXp3SaoHp/jsWWjEYJ2blgg76NpAPeyeT4yl7x0oid2dXJOif2XpmwnOb2jZFLxYbVm6w/P7q9jRVIlciTXv8J0vk5qT0kzR/Azhr/lnR8WaSxFXkq0TQzlJGFeSV0jeq92TtlfPRgsHI1WdczboF6zg8dNg2awdOCI5MulQam9Gp3E4+sFKjr2iuak57txKt0U/i6VtlJtbJLYmgn02XzfgFVNSxIDOFq9476ulX1jIZmmRyajJh3x3dO1jZsNK0ekKVea1uXs2SuiWW3z+dp5/KskqWyM3F009l7ziEI3pBy0rpW5gB+vibj/Ngx4Omz/3gpR+wp2dPwnY7ve7oOrkTA9G2yopk7ZXz0VbZiEM4uG/9fYB9SVyIXIQdwpGRvWNsRqfufPOBlRYMipbqlrQXrs7hThzCkXNb6lxnu1thzgd9j8uT1WLk8evjgn32Dpi3YtjRtSNpsFtcu5iPnfYxPrvus5bfG+D8pefznsXvMe3rDamtmnzaO8nuPD7T/hnevfjdnDX/rIyPbcXe+fozX+e6h65L2D40McQ1f7iGe56/J+E5peKMzeyyJd7TNwZyb8XsKH2Asxeezc3n3ZyyhjxTnA4nje7GrOydpqqmvCp9Ky0YFFbtnfne+ZQ7y3Mal7fCS0iGsp7tbgVr0x2LGBUIpJQZ9U0xaz+QTddOs0QuRIKM8fZywD/AW4Nv8Tdn/U3SY6XL/Jtx+rzTeeZTzyR9PtWFLDpXobw6ZrvL6aLcUZ61vVPhrEi6cth5i8/jj5/6Y8bHhRMNvsIynLRHTb+/nwP9BxieHI65o3q5+2XAXF2mst0yxeV0UVVeFfH0/QOcPu/06HMel8f01j6fnr6RVCs9ZYtastMq6gJ34dIL+fXeXzMxNZFzkzszesZ6LP89rdo7ufr5EGudpss3ZMucV/oN7gamwlMZBygze0e1dcimZFMp/Wi/kzilnyyJm2/S2TtV5VWmib1qV3Y99VWztVwbl5nR6G4kLMMpS2JVUHmlOzZJqD7/PT17Eqy3VLZbNtRX1ie1d1Ip/XzZO/lELdlpFXWBe9+y9xGSIV7tedX2MUkpLfXSV7RUtzAWHEsp9uwK+ulKd+1gzgf9bCftmCVyIfNgp+wdY50+JDZdU2oymwRmLqSyakYmEztsKrKtMohvtmYn6m+dyuJR9kG8daAeB8NB9vTuSXjOzotxXWUdff4+hieHLSVy1Zjzae/kCysq2Yi6wL1v2fuA/CRzRwIjBEIBy55+utLTsAxzZPhIzklcOKH081mrP+eDfjYLNAdCAYLhoGnv8kyrgZLaO3FKf0f3DhbWLLR8ItqFsm5Mlb7JXAXj67JN5CZLKudKtBVDkgt8MBSMBtUd3YlB/x1N74j+rFC2m50X43p3fbSBnpVE7uDEIFXlVTET9oqFTJcbHJgYoLq8mraGNmoravMS9K3W6CvSlZ72jvUyGZq03d7JF3M+6GczFVzdYpoq/QyDXdJEbpzSt1tNWiVakWTyO5nlNRTZKn1jL327Sdd0zXihVXdWELng7T2+lw+d/CFqKmpinsuH7VZXWcebA29Gf1YkVfp5aLY2UzRXN9Pv709YPSoZKrkthGCNb43lCp49PXt4z/99D2f/8GzO/uHZnPtv58a06jZitQVD9HdIo/TtqtGHxHVy9x3fxxn/cgYvHnkx52MrLAV9IcR6IcQ+IcQBIcQtJs9/XgjxmhBilxDicSGE9ZrCPJONvfPtF78NRFr+xpNqBqsZCXX6JuvkjgXG2Hd834xbO2DB3kmyUlOq7pyp6PP30VCZH3snXdM1dTFfUruE13pfi96F7T62m7AMc5bvLNa0rom5C1BK01alX1kfVfRGnz5ZIndw0v62yjOFCqxWv3/GPEd7azu7ju1iKjyV9nW/ef03/Onwn6ivrKfB3cDuY7v5ya6fmO5rtQWDIl0rBjuDfrzS39KxhV3HdiU0U8yFtEFfCOEE7gc2AKcAVwkhTonbbSewVkp5OvAr4Ju2jTBHMrV3DvQf4J7n7+HqM65m7fy1Cc+nW2kqnvg6fbNE7ivHXkEiZ0Xpp7R3bFb6UsqIvZMvpZ/mAm/0i41JQqOab/e180r3K9F1Y3d277TddjOqduPPHpeHydBkgiouaqWfYSsG44S1dl87E1MTvH789bSv29G1g7aGNh76+EM8+N8f5IKlFyS1huy2d+yamAWJidytHVs5teVUWy4oCitKfx1wQEr5ppQyADwAXGHcQUr5pJRSRY0XgIW2jTBHMm25e+MjN+JyuvjGRebla7naO2WOMqrLq2PsHWUnFJq9Y7ZUoiIbT38sOEYwHMxbIlcpxGQXeJUQjU8S7ujaQX1lPYtrF7OmdQ3+KT/7+vZFn7P772JU7cafk7VXzkeztZki01YMA/4TrSnW+CJ3V1Z8/fi/U7uvnb3H95qKmUyVvtflxeV0Jb1wdQ53UlVeZct5bUzkjkyO8MyhZ9i40r65E2At6C8AOg2Pj0xvS8ZfA6ZTHoUQ1wghtgshtvf2Wk/u5EIm3RcfeeMRNu/bzFfe85Wkt1OZ2jv+oB+ncMZM2ohvr7yjawdNVU0s8Kb6WPNDqlnGdiv9VLNx7cDpcEYqY5Jc4JXSX9O6hvrK+pig3+5rRwgRDRw7unYwGhjNi+1mDODxSh8Sg34+lkqcKax2qVQYlf5JjSfhLnOnDfr9/n4ODR2KCfprWtcQlmF2H9udsH/PWA/V5dVJ54rEI4RImZBW5Zp2lCEbz4HH3nyMYDho64Q5sDmRK4T4OLAWuNvseSnl96WUa6WUa5ubZ65KxcoKPsFQkBseuoGVDSu54Zwbku6Xsb1jWEBFEd90bUf3iaAz0zgdTiqcFRnbO6laMicjVbM1u0j1tzYuT6iShMFQkN09u6MB46Smk6gsq2Rn1052HduVF9vNGMBjZuQmqdEe8A8Uraefqb1j9PSdDidntp6ZNplrVu5svHjHk0kLBkWqVgx21ehDxAZ2CicjkyNs6dhCbUWtaW4xF6zMyD0KGM2qhdPbYhBCXAx8GThfSpnYWGYWsdKT5ae7fsre43vZfOVmKsoqku6Xqa1hNqOwtvJE07V+fz97evZw2bmJvfBnCrO5B1LKtHX6mdo7qZqt2UVjVfJWDMZ2Bmta1/CdP3+HV469QiAUiAaJMkcZZ8w7gx3dO1hWvwyw33ZTQU1ZfQqzGm3V1rlYlX6DuwGBsGTvhMKhhN91TesafrLrJylnWUeT7b4TQX9x7WIa3A2mF4xMWjAo5nnmsaVjCxV/nxgbAqEAn17z6YyOlwwhRLR0d2vHVi5dcWnOrR3isRL0twFtQohlRIL9lcDH4ga6BvhXYL2U0vqc6xlCTc9PxcvdL+N1eXn/qven3C/jOv2pE+vjKmoqaqIB6NYnbyUkQ3z89I9bPqbdmF3IJkOThGQoudJ3VUfmM4SClk/KfNs7EPlbHx8/bvrcwMQALqcLd5mbdl87k6FJfr7750CiSvz57p+zrG5ZXmw3FdTqKuti7u7M7J2hiSEksmiVvtPhpLGq0ZK9o+5+jb9ru6+d727/Lm/0v0FbY5vp63Z272Rx7WKaqpqi24QQkUosE6XfM9aTcWO0r57/Vc6cZ962QQjB1WdcndHxUuGt8PKnw3+ia7QrozbqVkkb9KWUU0KIa4GHASfwIynlHiHE7cB2KeVmInaOB/jl9El8WEp5ue2jzZJGd2N0MkwyOoc7Lfly1eURTz+V8jDiD5rbO4cGD7H72G6+t/17/M+1/5NTW05N/4vkCbMLmSodTFqyaejZU+e0pkLTNVuzg0Z3I/uO7zN9TlXBqIAA8JNdP8Hj8sQElDWta/je9u/x+/2/z4vtpiyd+EAeX6MNxd2CQWG1FYNxJTGFusva2b0zadDf0bXDNO/S7mvn2y9+O0GY9I73ckbrGRn9DmcvPJuzF56d0WuyxePyRO9Q1q9cb/vxLXn6UsqtUspVUsoVUso7prfdOh3wkVJeLKWcJ6U8c/pfwQR8sObpW/XlVF271XbNZkpfefrXP3Q9dZV13H7h7ZaOlS/Mau6TddhUZDNzUP0N8unpp7qrM9a7r2pcRVV5FcfHj3Nm65kxF3AVaI6PH6e91f6KKqPSN2L2mRZzCwaF1VYMZhe41S2rKXeUJ03mjgZG2d+339SCa/e1EwgFeK33teg2KWVW9s5Mos6Dd85/J/M882w//pyfkQuRQDDgHyAsw0n3sRr0M10yMZmn3z3azZNvPcnXL/x6XpWvFdTdi5FoL/1kJZtZtFfu9/fjcXny2k6g0d3I0OSQ6YQeY727ShJC4sSrU1tOjS4/afSJ7UJdeOLVu1kiNxoIi9TeAWuticGwboDhd3U5XZzacmrSoP9Kd/I5LurvanztaGCUydBkQQd9dR7YXbWjKImg31jViETGrMpjxB/00zvea03pZ9hTP5m9A3Bay2lcc9Y1lo6TT6rKqxJ+H5VMtFvp5/sCp+4ijItxKOLr3VVQiA8YFWUVrG5ebfqcHXhcHpzCmVTpGxO5ZpZHsdFSlX4REki+bsCa1killdnCIqlmTLc1tuFxeWKCfqYtGGYDdR7ooJ8D6Wbldg6nX9RYkWmwM7N31C3bfRvuS7qg+UySjb2TzYIy+Wy2pkjVdC2+3v3sBRGPdt2CdQn7nr3gbBrcDSyvX277GIUQ+Lw+5ntik4nq7slU6Rexp6/676Rrp2AsqTXS7mvn+PhxjgwfSXjNju4dtFS3mCZmHcLBGfPOiKngUevu2rEgTr5oqW6h1dNq2hHADkoi6KdruqZWsrcyjVrV9x4bPWbpvc2U/l+d/lds+8w2Llh6gaVj5BszeyddIjcbpZ/PZmuKVH/rwYnYHjYfO+1jPP/Xz3NKc3xXEbjzojv506f+ZClZnw2P/dVj/N35fxezzSEcVJdXxyRy54qnD+lnxSdT+sZkbjw7u3amTLa3+9p5uftlQuEQgVCALz3+JVY1ruLCZRdm/HvMFHdedCfPfPKZvJ17pRH00/RkyaRhks8TmanbNdpl6b3NPH13uTtvV/FsMLN30ir9LDz9mbR34u/qpJQJk5ycDifnLDwn6XFObj45b+M8qekk088ifqbz4MQgDuFImlspBqy2YhiYGEiYuwCR1d8EIsHXn5iaYE/vnpQzptt97YwFxzjQf4D7XryP/X37ufeyewu6TXVLdUvSSiU7KImgn87eOTx0GIFgQU36emzVnqF7tNvSe5vZO4WG2ezadIlcdTEoFntnLDhGSIYKXjF7K7yMBg3VO4Yy02LFaisG1YIh/netdlXzjqZ3JCj9V3teZSo8lTLvoi4IWzu2cvvTt7OxbWPevPJioSSCfjp75/DQYXxen6Wrv9flpaq8iq4Ra0rfHyz8oK/q9I2JsnSJXKXGrNo7YRlmYGIg/0o/yd86mV9caMS3Vy7mtsoKq60Y4pePNLLGlzjRykqjwlOaT8HldPGlx7/ExNQE9152byZDn5OURNCvq6xDIFImcq22RRVC4PP4TO2dAf9ATE0wmPfeKTSqXdVIZLT3P0SCuUM4ki5KHVX6Keyd14+/Hg1gQxNDhGU470q/pqIGp3Am/K2LxRuPt3eKua2ywqq9Y2y2Fk97aztHho/EXDh2dO2gtqKWZXXLkh6z3FnOaS2nMRma5Pqzr2dV46osfoO5RUkE/Wj3xRSefiYNk1o9raZB//anb+eCH18QfSylNPX0Cw2zShzVdyeZrVBZVolAJFX6b/S/wRn/cgZXPHAFUsoZabYGkYtyg7sh4W9dLPXuXpc3pmTzrcG38jJBZyZpdDciEGntHWNb5Xjik7ljgTF+v//3rFuwLq31df6S81ngXZCQOC9VSiLoQ/Kma1LKjIO+z+sztXc6+jvoHe+Nqt/JUKTvXDHYOxA74SxVh02IBNdUi8Tf+MiNBEIBnnzrSX6999cz0mxNYdZ0rVjq3Y1K/9DgIfYe38tFyy6a5VHlhuq/k4u9oybSKYvnrmfv4ujIUf7uvekD+V2X3MXr174eXcCo1CmZoJ9sen6fvw//lD+zoJ/E3lH1/uq5+EXRCxWzSpzR4GjSck1Fsp76j77xKL/b9zv+/sK/57SW07jxkRs5OhxpzJpvewdIrfQL3NM3rpO7tWMrQF6abs00VhZIT2Xv1LvrWVa3jJ3dO3lr8C3ufu5urjz1St6z5D1p37vMUZZSwJQaJRP0G92NponcbNa39Hl8DE8OJ9S2q2Opu4D4VbMKFbMlE0cmR9J+Ucy6cwZDQa5/6HpW1K/gpnfdxH0b7uPQ0CFue+o2IP/2Dpj/rYvJ01d5kK0HtrK8fvmc8KHTtWIwK6mNRyVzv/DoFxAIvnlxwazKWlSUTtBP0mc9k4lZCrOyzeHJ4aiaVNvj18ctVMyWTBwNjKatDTdT+t/d9l32Ht/Lty77FhVlFVyw9AI+fMqH2d0TWcFotuwd9bdRLTAKFfWZ+oN+Hn/zcTau3FjU5ZqKVIuQQEQgBcPBlBfl9tZ2DvQf4Fev/YovvftLLKrNfU3aUqRkgn5DZeItP2Sn9NUUbqOvry4ecMLeUdUwxWjvjAQsKP249g3Hx49z21O3cemKS/nAqg9Et999yd3RC99MJFLN/tYD/gFqK2pxOpx5f/9c8FZ4CckQDx14CP+Un02rit/agfSdNq2U1Krmd0vrlnLTu26yd4AlRMkE/caqRoYnhwmGgjHbDw8dprKsMmYBhnSYzcpVfj7MDXund6w3rSqPV/oPdjzI0OQQd7zvjhh1urRuKXe87w4uWHrBjATdVk8r48HxmMXnByeT+8WFhLrQ/mLPL3CXuTl/yfmzPCJ7aK6K9N8JhUOmzydrwWDk3IXn0tbQxnc2fKfghVQhUzJBXwUw5e0qDg9nvqixsneMSl/dMbicrqJL5MbbO+PBcY6OHKWtIfVU8HhPf3/ffhzCwenzTk/Y9/Pnfp4nP/GkjaNOzsqGlUCkmkqRqhywkFCW2u/3/56Lll9U8OeOVZqrm5HI5OsXm7RVjqfeXc/+z+6fM3c/s0XJBP1kMzU7h6xPzFI0VTVR5iiLUfqHhw7jFE5ObTn1RNCfKg5PP97eOdB/ACBt/w+PyxNj73T0d7C0bums9zVRw1rZwAAADadJREFU4+7oMwT9ieKY5KSU/nhwnI0r5067gHStGKwofY09lE7QT9KIK5uV7B3CwbzqeTGJ3MNDh1lQs4CFNQuj26OefoHbO/F1+vv79gNYUvpGe6ejvyPta2aCFfUrgFilH99hs1Ax5lHmUo+YdK0YiqVNxlygZIK+WSOuYCjI2yNvZxz0YXqCVpynv6hmUaSGf6S47J34GblKISubJBkelyf6GiklHX2FEfTd5W4W1SxKsHeKQUWquRGrm1ezpG7JLI/GPtK1YiiWGdNzgZIJ+mb2ztGRo0hkVkG/1dOa4Okvrl2Mz+Ojd7yXYChYNIlcp8NJhbMiqvQ7+jto9bSmnZxV7TqxSHzPWA8jgZG8toTNhLbGthh7p9iU/lxS+XDC3vna019j/U/Xs/6n6/nmsyfq7JWnX1tZ2CW1c4GSCfpm7ZWzKddUGGflhmWYzqHOSNCfTvIeGztWNHX6MN1Tf9qf7+jvsDQhyFj1oyyhQplItKphVXRMwVCQseBYUSj9VY2r+MvVf8ln2j8z20OxlaaqJq489Uq8Li+DE4PsPb6XW5+8NXrODU4M4nV5C2IlublOyXzCNRU1lDnKYuydbCZmKXweH71jvUyFp+gd6yUYDkaVPkQqe4qlTh+I6aOzv29/TJ19MoyrZykrpRDsHYgo/YGJAfrG+wjLMFAcfnFlWSUPfPiB2R6G7TiEg//40H9EHz/+5uNc/JOLeeLgE3zgpA8UTaJ9LmBJ6Qsh1gsh9gkhDgghbjF5/r1CiB1CiCkhxIftH2buqO6LZko/m5l9Pq8PieTY6LGYO4boxK3RrqKxd+DEkonDk8P0jPVYCt7Gqp+Ovg7KHGUF40Or8Xf0d1gqB9TMLO9Z8h48Lk+0v1CxlNTOBdIGfSGEE7gf2ACcAlwlhIhfVPQw8Eng53YP0E7iG3EdHjpMU1VTtHolE5Si7x7tjk7MWlSzKKZFgz/oRyBmvYTRCmrJROWDW/Hm45X+8vrlBXN7bizb1OWAhYfL6eKS5ZewpWMLUsqUzdY09mJF6a8DDkgp35RSBoAHgCuMO0gp35JS7gLCeRijbcQ34lITs7LBqOiNSn9e9TwEgq6RiNKvLKssit4pqqVCJjaNseqnUMo1Fcvrl+MQjojS1+WABcnGto10Dneyp3dPyrbKGnuxEvQXAJ2Gx0emt2WMEOIaIcR2IcT23t7UbVbzgbERl5SS/X37s/LzIXZW7uGhw3hcHuoq6yh3ltNU1UTXaMTTLwY/H07YO1bLNeGE0h+ZHKGjz1ryd6ZwOV0srVvK/r79RdNhs9TYsHIDAFv2b9FKfwaZ0eodKeX3pZRrpZRrm5ubZ/KtgVh75+e7f86B/gN88B0fzOpY8Urf2MpB1fAXw/q4CmXv7O+PXAitXKyUp9/R34F/yl9QSh8idysd/R26BrxAWVCzgDNbz2RLx5a0bZU19mEl6B8FjHJ44fS2oqPRHVH6o4FRvvjYF1k7fy1Xn3F1VsdyOV00uhvpGumic7gzxiZSNfzFsD6uImrv9HVYrrVXSl8tUF0oNfqKtoZIrX6xrJpVimxcuZHnOp9jJDCi7bcZwkrQ3wa0CSGWCSFcwJXA5vwOKz80uBsYD45z25O38fbI29y3/j4cIvubHaXoDw8djrGJfB5fJJE77ekXA1VlVRF7JwNvXnn6Lx97GSicck1FW2MbI4ER9vXto8JZUTQX4FJi06pNhGSk86a+KM8MaSOelHIKuBZ4GNgL/KeUco8Q4nYhxOUAQoh3CiGOAB8B/lUIsSefg84WNSv33hfu5eOnf5xzF52b0/FaPa0cHDxIz1hPjNJXQX88OF409k61q5rj48fp9/dbDt5K6b/a8yoVzoqCW9RC/R5/PvpnHVAKlLMXnB2dOKntnZnBksyVUm6VUq6SUq6QUt4xve1WKeXm6Z+3SSkXSimrpZSNUsrV+Rx0tqima1XlVdx18V05H8/n8bGnJ3J9iwn6Xh/BcJCjw0eLRl1Wl1dHFZfVhKwqdQ2EAqxsWJnTXVM+UL/H68df19ZBgeJ0OLlsxWWAVvozRWF9S/PMvOp5AHz5PV9mvnd+zsfzeXzRQBmv9AEODh4sGqVvnKtg1Zt3OpxR+6rQ/HyAJXVLKHOUIZFaRRYw71/1fuBEfx5NfimMmTQzxHmLz2PzlZvZ0LbBluOpsk1IVPoQ6UlTLJ6+qsRxCAfL65dbfp3H5WFiaqLg/HyAMkcZy+uXs79vv1aRBcxfrv5LaipqWLdg3WwPpSQoKaXvEA4+cNIHbJs1qhQ9wALviakLqpwTiqPvDpxIyi6pXZLRDGL1ukIM+nBiXNreKVycDifvX/X+opjEOBcoqaBvNyq4t3paqSiriG43XgyKzd7J1KZRydxCtHfgRNCvq9BKX6MBHfRzQtk48a0cql3V0bVOiyXoK3tnVUNms2qjryug2bhG1Li00tdoIuignwNK0Zv171EXhKLx9JVNk4XSry6vjrm7KSTU76M9fY0mgg76OeCt8DLfO5/VzYkVqioIFounr6qZ2n3tGb1uYc1C1vjWFKwfu7p5NeWOcpbVLZvtoWg0BUFJVe/kg5f/5mVqKmoStiu/v1jsnZObT6bzc50srFmY0evu33g/U+GpPI0qd3xeH2/d8FZMcl2jKWV00M8RteBzPMWm9IGMAz6cSOQWMnbMydBo5gra3skTytMvFqWv0WhKAx3084RS+sWSyNVoNKWBDvp5Iqr0i8je0Wg0cx8d9PPE6ubVtFS3cHLTybM9FI1Go4miE7l5wuf1ceymY7M9DI1Go4lBK32NRqMpIXTQ12g0mhJCB32NRqMpIXTQ12g0mhJCB32NRqMpIXTQ12g0mhJCB32NRqMpISwFfSHEeiHEPiHEASHELSbPVwghfjH9/ItCiKV2D1Sj0Wg0uZM26AshnMD9wAbgFOAqIcQpcbv9NTAgpVwJ3AvcZfdANRqNRpM7VpT+OuCAlPJNKWUAeAC4Im6fK4B/n/75V8BFolBX1dBoNJoSxkrQXwB0Gh4fmd5muo+UcgoYAhrjDySEuEYIsV0Isb23tze7EWs0Go0ma2Y0kSul/L6Ucq2Ucm1zs/niIxqNRqPJH1aC/lFgkeHxwultpvsIIcqAWqDPjgFqNBqNxj6sBP1tQJsQYpkQwgVcCWyO22cz8Inpnz8MPCGllPYNU6PRaDR2kLa1spRySghxLfAw4AR+JKXcI4S4HdgupdwM/BvwEyHEAaCfyIVBo9FoNAWGpX76UsqtwNa4bbcafp4APmLv0DQajUZjN3pGrkaj0ZQQOuhrNBpNCaGDvkaj0ZQQOuhrNBpNCSFmq7JSCNELHJqVNy8MmoDjsz2IAkJ/HrHoz+ME+rOI5SQppTfbF1uq3skHUsqSnpIrhNgupVw72+MoFPTnEYv+PE6gP4tYhBDbc3m9tnc0Go2mhNBBX6PRaEoIHfRnj+/P9gAKDP15xKI/jxPozyKWnD6PWUvkajQajWbm0Upfo9FoSggd9DUajaaE0EF/BhBCLBJCPCmEeE0IsUcIcf309gYhxKNCiI7p/+tne6wzhRDCKYTYKYT4w/TjZUKIF4UQB4QQv5hu410SCCHqhBC/EkK8LoTYK4Q4t8TPjc9Nf09eFUL8hxCispTODyHEj4QQPUKIVw3bTM8HEeG+6c9llxCiPd3xddCfGaaAG6WUpwDnAP97enH5W4DHpZRtwOPTj0uF64G9hsd3AfdKKVcCA8Bfz8qoZodvAw9JKd8BnEHkcynJc0MIsQC4DlgrpTyVSDv3Kymt8+PHwPq4bcnOhw1A2/S/a4DvpT26lFL/m+F/wO+AS4B9gG96mw/YN9tjm6Hff+H0ifs+4A+AIDLjsmz6+XOBh2d7nDP0WdQCB5kuqjBsL9VzQ6233UBk8ugfgMtK7fwAlgKvpjsfgH8FrjLbL9k/rfRnGCHEUmAN8CIwT0rZNf1UNzBvloY10/wT8EUgPP24ERiUUk5NPz5C5MtfCiwDeoH/O213/VAIUU2JnhtSyqPAPwKHgS5gCHiJ0j0/FMnOB3WRVKT9bHTQn0GEEB7gv4AbpJTDxudk5DI95+tnhRDvB3qklC/N9lgKhDKgHfielHINMEaclVMq5wbAtFd9BZGL4XygmkSro6TJ9XzQQX+GEEKUEwn4P5NS/np68zEhhG/6eR/QM1vjm0HOAy4XQrwFPEDE4vk2UCeEUL2gFgJHZ2d4M84R4IiU8sXpx78ichEoxXMD4GLgoJSyV0oZBH5N5Jwp1fNDkex8OAosMuyX9rPRQX8GEEIIIusI75VSfsvwlHFB+U8Q8frnNFLKL0kpF0oplxJJ0D0hpfzvwJPAh6d3K4nPAkBK2Q10CiFOmt50EfAaJXhuTHMYOEcIUTX9vVGfR0meHwaSnQ+bgaunq3jOAYYMNpApekbuDCCEeDfwR2A3J3zs/0PE1/9PYDGRNtMflVL2z8ogZwEhxAXATVLK9wshlhNR/g3ATuDjUsrJ2RzfTCGEOBP4IeAC3gQ+RUSQleS5IYT4GvCXRKredgKfJuJTl8T5IYT4D+ACIi2ljwG3Ab/F5HyYvjB+h4gFNg58SkqZsgunDvoajUZTQmh7R6PRaEoIHfQ1Go2mhNBBX6PRaEoIHfQ1Go2mhNBBX6PRaEoIHfQ1Go2mhNBBX6PRaEqI/w9gIENoJE4SWAAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "C3YBvlaxP6GX",
        "outputId": "735671ed-c2e2-4d2b-9760-9b5db3b542e7"
      },
      "source": [
        "bestFeaturesFilter = select_n_best_feature(X_train, y_train, 3)\n",
        "X_train_filter = bestFeaturesFilter.fit_transform(X_train_norm, y_train)\n",
        "X_val_filter = bestFeaturesFilter.fit_transform(X_val_norm, y_val)\n",
        "X_test_filter = bestFeaturesFilter.fit_transform(X_test_norm, y_test) \n",
        "\n",
        "lda = LinearDiscriminantAnalysis()\n",
        "model = lda.fit(X_train_filter, y_train)\n",
        "\n",
        "y_pred_test=model.predict(X_test_filter)\n",
        "\n",
        "#    print('Accuracy on test set: ',metrics.accuracy_score(y_test_bin, y_pred_test))\n",
        "acc_rate = metrics.accuracy_score(y_test, y_pred_test)\n",
        "#    print('Accuracy on train set: ',metrics.accuracy_score(y_train_bin, y_pred_train))\n",
        "print('Accuracy on test with 3 features', acc_rate)"
      ],
      "execution_count": 33,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Accuracy on test with 3 features 0.40384615384615385\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CW_wvtneRN-d"
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}