handwritten_digit_recognition.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "tMj-Z7EHU7aT"
   },
   "source": [
    "##Τεχνητή Νοημοσύνη: Εργαστηριακή Άσκηση 3\n",
    "###Μαυραειδή Λυδία ΑΜ:03117181\n",
    "###Τσακανίκα Χριστίνα ΑΜ:03117012"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "W2U8JPk5Vdof"
   },
   "source": [
    "##**1ο** Μέρος: Αξιολόγηση\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "OTIJvhZ_VBXu"
   },
   "outputs": [],
   "source": [
    "from sklearn.metrics import confusion_matrix, f1_score, precision_score, recall_score\n",
    "from matplotlib import pyplot as plt\n",
    "import numpy as np\n",
    "import random\n",
    "\n",
    "class Evaluate:\n",
    "    def __init__(self,y_true,y_pred,x):\n",
    "      self.y_true = y_true\n",
    "      self.y_pred = y_pred\n",
    "      self.x = x\n",
    "\n",
    "    def my_accuracy(self):\n",
    "      y_true = self.y_true\n",
    "      y_pred = self.y_pred\n",
    "      \n",
    "      right_predictions = 0 #number of right predictions\n",
    "      for i in range(len(y_true)):\n",
    "        if y_true[i] == y_pred[i]:\n",
    "          right_predictions += 1\n",
    "      \n",
    "      acc=right_predictions / len(y_true)\n",
    "      \n",
    "      return acc\n",
    "\n",
    "    def get_metrics(self):\n",
    "        precision = precision_score(self.y_true, self.y_pred, average = \"macro\")\n",
    "        recall = recall_score(self.y_true, self.y_pred, average = \"macro\")\n",
    "        f1 = f1_score(self.y_true, self.y_pred, average = \"macro\")\n",
    "        results = {\"precision\" : precision, \"recall\": recall, \"f1\": f1,\"accuracy\":self.my_accuracy()}\n",
    "        return results   \n",
    "    \n",
    "    def confusion_matrix(self):\n",
    "        cm = confusion_matrix(self.y_true, self.y_pred)\n",
    "        return cm \n",
    "\n",
    "    def get_sample_of_wrong(self, rows = 5, lines = 4):\n",
    "        wrong_indexes = [index for index in range(len(self.y_true)) if self.y_true[index] != self.y_pred[index]]\n",
    "        plt.figure(figsize=(10, 10))\n",
    "\n",
    "        #Shuffle a list (reorganize the order of the list items)\n",
    "        random.shuffle(wrong_indexes)\n",
    "        for i, index in enumerate(wrong_indexes[: rows*lines]):\n",
    "            plt.subplot(rows, lines, i + 1)\n",
    "            image = self.x[index]\n",
    "            plt.imshow(image.reshape((28, 28)))\n",
    "            plt.title(\"Real: {}, Pred: {}\".format(self.y_true[index], self.y_pred[index]))\n",
    "            plt.axis('off')\n",
    "\n",
    "    def get_evaluation_report(self):\n",
    "       metrics = self.get_metrics()\n",
    "       for m in metrics:\n",
    "         print(m+' : '+str(metrics[m]))\n",
    "       cm = self.confusion_matrix()\n",
    "       print(\"Confusion matrix: \")\n",
    "       print(cm)\n",
    "       print(\"Some images with wrong predictions\")\n",
    "       self.get_sample_of_wrong()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 292
    },
    "id": "mPE8SYNaVupL",
    "outputId": "2352ca41-9682-4670-ed71-c346b6955a90"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "precision : 0.7000000000000001\n",
      "recall : 0.6944444444444443\n",
      "f1 : 0.6888888888888888\n",
      "accuracy : 0.625\n",
      "Confusion matrix: \n",
      "[[3 1 0]\n",
      " [2 1 0]\n",
      " [0 0 1]]\n",
      "Some images with wrong predictions\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 3 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y_true = [1,0,2,0,0,1,1,0]\n",
    "y_pred = [0,0,2,0,1,1,0,0]\n",
    "x = np.random.rand(8,28*28)\n",
    "\n",
    "eval = Evaluate(y_true,y_pred,x)\n",
    "eval.get_evaluation_report()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "OxUY-ZijV20p"
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "from matplotlib import pyplot as plt\n",
    "\n",
    "train_dset = pd.read_csv('/content/sample_data/mnist_train_small.csv',header=None)\n",
    "test_dset = pd.read_csv('/content/sample_data/mnist_test.csv',header=None)\n",
    "## From https://colab.research.google.com/notebooks/mlcc/multi-class_classification_of_handwritten_digits.ipynb#scrollTo=JfFWWvMWDFrR\n",
    "def parse_labels_and_features(dataset):\n",
    "  \"\"\"Extracts labels and features.\n",
    "  \n",
    "  This is a good place to scale or transform the features if needed.\n",
    "  \n",
    "  Args:\n",
    "    dataset: A Pandas `Dataframe`, containing the label on the first column and\n",
    "      monochrome pixel values on the remaining columns, in row major order.\n",
    "  Returns:\n",
    "    A `tuple` `(labels, features)`:\n",
    "      labels: A Pandas `Series`.\n",
    "      features: A Pandas `DataFrame`.\n",
    "  \"\"\"\n",
    "  labels = dataset[0]\n",
    "  \n",
    "  # DataFrame.loc index ranges are inclusive at both ends.\n",
    "  features = dataset.loc[:,1:784]\n",
    "  # Scale the data to [0, 1] by dividing out the max value, 255.\n",
    "  features = features / 255\n",
    "\n",
    "  return labels, np.array(features)\n",
    "\n",
    "def show_image(features):\n",
    "  img = features.reshape((28,28))\n",
    "  plt.imshow(img)\n",
    "  plt.show()\n",
    "\n",
    "y_train,x_train = parse_labels_and_features(train_dset)\n",
    "y_test,x_test = parse_labels_and_features(test_dset)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "8fwLtQPPWPCR"
   },
   "source": [
    "Ποια είναι η μορφή των δεδομένων για μια εικόνα;"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "x1q_eM4dWFce",
    "outputId": "fe2e61f8-bf79-487c-cf33-3558c76b8d96"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape of an image: (784,)\n",
      "Max value of pixel: 1.0\n"
     ]
    }
   ],
   "source": [
    "print(\"Shape of an image: \"+str(x_train[0].shape))\n",
    "print(\"Max value of pixel: \"+str(np.max(x_train[0])))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "7kp1qp_yWYD1"
   },
   "source": [
    "Φορτώνουμε μια τυχαία εικόνα και την εμφανίζουμε, μαζί με την ετικέτα της"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 282
    },
    "id": "rjqhElUdWK6c",
    "outputId": "4113464d-68c4-4ef9-c4f0-e4a3a4049388"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Label = 8\n"
     ]
    }
   ],
   "source": [
    "i = np.random.randint(0,x_train.shape[0])\n",
    "show_image(x_train[i])\n",
    "\n",
    "print(\"Label = \" +str(y_train[i]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "F0Tapi3wW7uD"
   },
   "source": [
    "##2o Μέρος: Υλοποίηση KNN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "4U0YnuKyWmjt"
   },
   "outputs": [],
   "source": [
    "## Simple KNN\n",
    "from collections import Counter\n",
    "from numpy import linalg\n",
    "class KNN:\n",
    "  def __init__(self,x,y,k,distance=\"euclidian\"):\n",
    "    self.x = x\n",
    "    self.y = y\n",
    "    self.k = k\n",
    "    self.distance = distance\n",
    " \n",
    "  def get_distance(self,img1,img2):\n",
    "    \n",
    "    if self.distance=='euclidian':\n",
    "      #k_most_similar = self.get_knn(img)\n",
    "    \n",
    "      #prediction = self.y[k_most_similar[0]] \n",
    "      #uniq_label, counts = np.unique(nn_labels, return_counts = True)\n",
    "      #prediction = uniq_label[counts.argmax()]\n",
    "      vector = img1-img2\n",
    "      dist = linalg.norm(vector)\n",
    "       \n",
    "    elif self.distance=='cosine':\n",
    "      \n",
    "      product = np.inner(img1, img2)\n",
    "      img1 = linalg.norm(img1)\n",
    "      img2 = linalg.norm(img2)\n",
    "      dist = 1 - product/(img1*img2) \n",
    "      pass\n",
    "    return dist\n",
    "\n",
    "\n",
    "  ## Given an image as a vector, returns indexes of k nearest neighbors\n",
    "  def get_knn(self,img):\n",
    "    distances = list()\n",
    "    x = self.x\n",
    "    k = self.k\n",
    "    \n",
    "    for i in range(len(x)):\n",
    "      distances.append(self.get_distance(img, x[i]))\n",
    "    \n",
    "    # Sort distances, and return the indexes of k first elements\n",
    "    ans_indexes = np.argsort(distances)[:k]\n",
    "    return ans_indexes\n",
    "\n",
    "\n",
    "  ## Given an image as a vector, classify it according to KNN\n",
    "  def classify(self,img):\n",
    "    y = self.y\n",
    "    nn_labels = [y[i] for i in self.get_knn(img)]\n",
    "    uniq_label, counts = np.unique(nn_labels, return_counts = True)\n",
    "    prediction = uniq_label[counts.argmax()]\n",
    "    #prediction = Counter(nn_labels).most_common(1)[0][0]\n",
    "    return prediction\n",
    "\n",
    "\n",
    "knn = KNN(x_train,y_train,k=5,distance='euclidian')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "L8uYXTCXXq4J"
   },
   "source": [
    "Τώρα που είναι έτοιμος ο ταξινομητής ας δούμε τι προβλέπει σε μεμονομένες εικόνες"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 298
    },
    "id": "DNSt0FquWVQg",
    "outputId": "45557afb-79f5-40c7-a0f3-d7113d0120c1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Input image: \n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Prediction : 9\n"
     ]
    }
   ],
   "source": [
    "img = x_test[np.random.randint(0,x_test.shape[0])]\n",
    "print(\"Input image: \")\n",
    "show_image(img)\n",
    "print('Prediction : '+str(knn.classify(img)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "Zer0gaBQX0Eo"
   },
   "outputs": [],
   "source": [
    "## Παίρνουμε όλες τις προβλέψεις στο test set (πρώτες 100 εικόνες)\n",
    "## ..υπομονή μερικά λεπτά\n",
    "preds = [knn.classify(x_test[i]) for i in range(100)]\n",
    "labels = [y_test[i] for i in range(100)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 410
    },
    "id": "utwnhIXMY5YE",
    "outputId": "2f5bb45d-74d0-4840-d8df-5540281faba6"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "precision : 0.9749019607843138\n",
      "recall : 0.9568181818181818\n",
      "f1 : 0.962567733990148\n",
      "accuracy : 0.96\n",
      "Confusion matrix: \n",
      "[[ 8  0  0  0  0  0  0  0  0  0]\n",
      " [ 0 14  0  0  0  0  0  0  0  0]\n",
      " [ 0  1  6  0  0  0  0  1  0  0]\n",
      " [ 0  0  0 11  0  0  0  0  0  0]\n",
      " [ 0  0  0  0 14  0  0  0  0  0]\n",
      " [ 0  0  0  0  0  7  0  0  0  0]\n",
      " [ 0  0  0  0  0  0 10  0  0  0]\n",
      " [ 0  0  0  0  0  0  0 15  0  0]\n",
      " [ 0  0  0  0  0  0  0  0  2  0]\n",
      " [ 0  0  0  0  1  0  0  1  0  9]]\n",
      "Some images with wrong predictions\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "eval = Evaluate(labels,preds,x_test[:100])\n",
    "eval.get_evaluation_report()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "r7O2ZFP5ZNxG"
   },
   "source": [
    "##Έτοιμος ΚΝΝ classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 410
    },
    "id": "0G8861PIZGqC",
    "outputId": "0d710183-1c9d-429e-a099-ac14ef314866"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "precision : 0.9749019607843138\n",
      "recall : 0.9568181818181818\n",
      "f1 : 0.962567733990148\n",
      "accuracy : 0.96\n",
      "Confusion matrix: \n",
      "[[ 8  0  0  0  0  0  0  0  0  0]\n",
      " [ 0 14  0  0  0  0  0  0  0  0]\n",
      " [ 0  1  6  0  0  0  0  1  0  0]\n",
      " [ 0  0  0 11  0  0  0  0  0  0]\n",
      " [ 0  0  0  0 14  0  0  0  0  0]\n",
      " [ 0  0  0  0  0  7  0  0  0  0]\n",
      " [ 0  0  0  0  0  0 10  0  0  0]\n",
      " [ 0  0  0  0  0  0  0 15  0  0]\n",
      " [ 0  0  0  0  0  0  0  0  2  0]\n",
      " [ 0  0  0  0  1  0  0  1  0  9]]\n",
      "Some images with wrong predictions\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 4 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "\n",
    "k = 5\n",
    "knc = KNeighborsClassifier(n_neighbors=k)\n",
    "knc.fit(x_train, y_train)\n",
    "y_pred = knc.predict(x_test[:100])\n",
    "\n",
    "eval = Evaluate(y_test[:100],y_pred,x_test[:100])\n",
    "eval.get_evaluation_report()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "2RorYXI6ZcA0"
   },
   "source": [
    "##Σύκγριση υλοποιήσεων"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "FZ2KMINNZUet",
    "outputId": "223ea20d-66d6-4b96-9833-03f99408ab74"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 22.2 s, sys: 79 ms, total: 22.3 s\n",
      "Wall time: 22.3 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "knn = KNN(x_train,y_train,k=5,distance='euclidian')\n",
    "preds = [knn.classify(x_test[i]) for i in range(100)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "8U93tAi5ZhWL",
    "outputId": "87f517bb-825d-4270-c569-439f7ef97369"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 7.51 s, sys: 8.01 ms, total: 7.52 s\n",
      "Wall time: 7.52 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "knc = KNeighborsClassifier(n_neighbors=5)\n",
    "knc.fit(x_train, y_train)\n",
    "y_pred = knc.predict(x_test[:100])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "A8rG6UzBag2I"
   },
   "source": [
    "##Σχολιασμός"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "M-atIhsEak6_"
   },
   "source": [
    "Πράγματι, παρατηρούμε πως ο έτοιμος ταξινομητής \" KNeighborsClassifier\" παρουσιάζει τα ίδια ακριβώς αποτελέσματα \"precision : 0.9749019607843138, \n",
    "recall : 0.9568181818181818, \n",
    "f1 : 0.962567733990148,\n",
    "accuracy : 0.96\" με τον ταξινομητή που κατασκευάσαμε εμείς, για την ευκλίδεια απόσταση. Οι επιδόσεις του δικού μας ταξινομητή για τη συνάρτηση συνημιτόνου είναι καλέτερες του έτοιμου classifier.\n",
    "Συγρίνοντας τους χρόνους εκτέλεσης, όπως φαίνεται παραπάνω, ο έτοιμος ταξινομητής καθυστερεί περισσότερο.\n",
    "Για μεγαλύτερες τιμές του k το μοντέλο γίνεται απλόυστερο, αφού η προσαρμογή είναι μειωμένη (underfit) και έχουμε ως αποτέλεσμα χαμηλότερο variance και υψηλότερο bias.\n",
    "Για υψηλότερες τιμές του k το μοντέλο γίνεται σύνθετο, αυξάνεται η προσαρμογή (overfit) και έχουμε ως αποτέλεσμα υψηλότερο variance και χαμηλότερο bias."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "TEMl8npjarEi"
   },
   "source": [
    "##3ο Μέρος: Naive Bayes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 265
    },
    "id": "VG1jorSjZnpU",
    "outputId": "212f74f7-9205-4bb3-c9b6-bc294f6e8454"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Στρογγυλοποίηση:\n",
    "x_train_r = np.round(x_train)\n",
    "x_test_r = np.round(x_test)\n",
    "show_image(x_test_r[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "gVrSXujXbKMw"
   },
   "outputs": [],
   "source": [
    "class NaiveBayes:\n",
    "  def __init__(self,x,y):\n",
    "    self.x = x\n",
    "    self.y = y\n",
    "    ## pC is a vector with the probability of each class\n",
    "    self.pC = np.zeros((10,))\n",
    "    ## pxC is an array with all probabilities p(xi|C)\n",
    "    self.pxC = np.zeros((x.shape[-1],10))\n",
    "    ## Compute the probabilities\n",
    "    self.compute_probabilities()\n",
    "\n",
    "  def compute_probabilities(self):\n",
    "\n",
    "    for i in range(len(self.x)):\n",
    "      for j in range(len(self.x[i])):\n",
    "        self.pxC[j][self.y[i]] += self.x[i][j]\n",
    "\n",
    "    for label in self.y:\n",
    "      self.pC[label]+=1\n",
    "    for j in range(10):\n",
    "      self.pxC[:, j] = self.pxC[:, j] / self.pC[j]\n",
    "    self.pC = self.pC/self.y.shape[0]\n",
    "\n",
    "  def predict(self,x):\n",
    "    \n",
    "    pcX = np.ones((10,))\n",
    "    \n",
    "    for i in range(10):\n",
    "      \n",
    "      pcX[i] = self.pC[i]\n",
    "      for j in range(28*28):\n",
    "        if x[j] != 0:   \n",
    "          pcX[i] *= self.pxC[j][i]\n",
    "        else:\n",
    "          pcX[i] *= (1-self.pxC[j][i])\n",
    "     \n",
    "    return np.argmax(pcX)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "t4Xh1s8ebiWV"
   },
   "source": [
    "##Αξιολόγηση του Naive Bayes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 747
    },
    "id": "37QOmeDIbe3_",
    "outputId": "4b9db9d3-5236-47c4-fb05-6fdc929f9c7b"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "precision : 0.8400974025974026\n",
      "recall : 0.850108225108225\n",
      "f1 : 0.8261529454950507\n",
      "accuracy : 0.85\n",
      "Confusion matrix: \n",
      "[[ 7  0  0  0  0  0  0  0  1  0]\n",
      " [ 0 13  0  1  0  0  0  0  0  0]\n",
      " [ 0  0  7  1  0  0  0  0  0  0]\n",
      " [ 0  0  1 10  0  0  0  0  0  0]\n",
      " [ 1  0  0  0 12  0  0  0  0  1]\n",
      " [ 0  0  0  2  1  4  0  0  0  0]\n",
      " [ 0  0  2  0  0  0  8  0  0  0]\n",
      " [ 0  0  1  0  0  0  0 13  0  1]\n",
      " [ 0  0  0  0  0  0  0  0  2  0]\n",
      " [ 0  0  0  0  1  0  0  0  1  9]]\n",
      "Some images with wrong predictions\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 15 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "nb = NaiveBayes(x_train_r,y_train)\n",
    "preds = [ nb.predict(i) for i in x_test_r[:100]]\n",
    "eval = Evaluate(y_test[:100], preds, x_test_r[:100])\n",
    "eval.get_evaluation_report()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Peo0bPOvcZyI"
   },
   "source": [
    "##Έτοιμος Naive Bayes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "id": "Bn5oa7c4bp6Y",
    "outputId": "b147292a-c031-4851-9238-33f840fb9669"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.6/dist-packages/sklearn/metrics/_classification.py:1272: UndefinedMetricWarning: Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "precision : 0.5992985558502799\n",
      "recall : 0.5716774891774892\n",
      "f1 : 0.5081873844349185\n",
      "accuracy : 0.6\n",
      "Confusion matrix: \n",
      "[[ 7  0  0  0  0  0  0  0  0  1]\n",
      " [ 0 14  0  0  0  0  0  0  0  0]\n",
      " [ 3  0  2  0  0  0  2  0  0  1]\n",
      " [ 0  0  1  3  0  0  1  0  5  1]\n",
      " [ 1  1  0  0  4  0  0  0  0  8]\n",
      " [ 2  0  0  0  1  0  0  0  3  1]\n",
      " [ 0  0  0  0  0  0 10  0  0  0]\n",
      " [ 0  0  1  0  0  0  0  8  0  6]\n",
      " [ 1  0  0  0  0  0  0  0  1  0]\n",
      " [ 0  0  0  0  0  0  0  0  0 11]]\n",
      "Some images with wrong predictions\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.6/dist-packages/sklearn/metrics/_classification.py:1272: UndefinedMetricWarning: Precision is ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "precision : 0.5771820010055304\n",
      "recall : 0.5655194805194805\n",
      "f1 : 0.4885160894473657\n",
      "accuracy : 0.59\n",
      "Confusion matrix: \n",
      "[[ 8  0  0  0  0  0  0  0  0  0]\n",
      " [ 0 14  0  0  0  0  0  0  0  0]\n",
      " [ 3  0  2  0  0  0  2  0  0  1]\n",
      " [ 0  1  1  2  0  0  1  0  5  1]\n",
      " [ 1  1  0  0  3  0  0  1  0  8]\n",
      " [ 2  0  0  0  1  0  0  0  3  1]\n",
      " [ 0  0  0  0  0  0 10  0  0  0]\n",
      " [ 0  0  1  0  0  0  0  9  0  5]\n",
      " [ 1  0  0  0  0  0  0  0  1  0]\n",
      " [ 0  1  0  0  0  0  0  0  0 10]]\n",
      "Some images with wrong predictions\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAiAAAAI+CAYAAACFVbH3AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOzdd3gVVfoH8O+bBEInUoWEJh0bWLAgiooFEOsiawMLdkXXsru2Bcva1rUiim0VUVTERRRZ20/EFVCaoggISglVWqSXJO/vj5mcmXM3N9x7czNzb/L9PE8e3sk5M3OGnEzOPWVGVBVEREREQcoIuwBERERU9bABQkRERIFjA4SIiIgCxwYIERERBY4NECIiIgocGyBEREQUuFAbICLSS0RWhlmGZBCR4SIyJuxyVFWVqB69KiIPhF2OqqoS1SPej0JSiepQIPeimBogIrJMRHaKyDYRWesWrk5FFy6iDE1EZKyIrBaR30XkaxE5Ko79p4jILvcaNojIeyLSrCLLXEZZaonISLccv4vI1DDKEbRUqEellGObiHyS4L7rQrwGEZG7RGSFiGwRkbdEpF7Q5QhDKtQjEWnpqz8lXyoit8a4fyrdj4aIyBK3LP8RkeZhlCNIqVCHIspzglt/Yv6jXxnuRfH0gPRX1ToAugLoBuCORApbDnUAzARwOIAGAF4DMCnO//Ab3GvoACAHwBORGUQkKwll3ZcX4FxDZ/ffPwVwzlQRdj2yyuF+nZrIvgAOA3AEgLsjMwRQjwYBuARADwDNAdQE8EwFnzOVhFqPVHWFr/7UAXAwgGIA4+M4TOj3IxHpBeBBAGfBuRctBTC2Is+ZQlLiXiQi1QA8BeCbBHZP63tR3EMwqroWwMdwfmgAABE5WkSmiUiBiHzvVuqStMtEZIGIbBWRX0Xk6njP6Z73V1V9XFXXqGqRqr4AoDqAjgkcaxOcG8VBbhmXichfRGQegO0ikrWPa2ojIl+61/QpgEaxnltEOgE4E8BVqrrevZbZ8V5DugurHiWTqq4CMBlePVIRuV5EFgNY7H7vDBH5zr2maSJySMn+ItJNROa41/Q2gBpxnL4/gJdVNV9VtwF4BMBAEamVrOtLBylUjwYBmKqqy+LdMcz7EYAzAIxT1fmqugfA/QCOF5G28V5HukqBOnQrgE8ALEz0AOl6L4q7ASIieQD6AFjibucCmATgATgt6NsAjBeRxu4uv8Gp5PUAXAbgCRE5LMqxR4rIyBjL0RVOA2RJAtfQCMB5AOb6vn0BgH5wPok03cc1vQlgNpxf9PsBDI44/jwRuTDK6bsDWA7gXnG6Xn8QkfPivYZ0lwL16A0RWS8in4jIoQleQwsAfWHXo7MBHAWgi4h0A/AKgKsBNAQwCsBEEckWkeoAJgB4Hc71joNTJ/3HLxCR48oqQkScDaB9IteSrlKgHkFEBE4D5LUEryHM+xHwv/UIcP+QVQVh1iERaQXgcgD3lfMa0vNepKr7/AKwDMA2AFsBKIDPAeS4aX8B8HpE/o8BDI5yrAkAbnLjXgBWxlKGiGPUA/ADgDvi2GcKgB0ACgCsAvAGgMa+67vclzfqNQFoCaAQQG1f2psAxsRYjjvd/8PhcBpQJ7j/t53j/X9It69UqUdwugprAqgFp9t1bUk54riGAjgNyZEAarppCuAkX97nANwfsf8i92d+PIDVAMSXNg3AAzGWYwiAnwG0BlAfwET3/MeE/XOuKvXId4yebnnqxLFPqtyPegPYAOAQ93diFJyhpAvC/jlXhToE4H0AA9341Vh//yOuIW3vRfH0gJytqnXd/+BO8Lr5WgEY4LaQCkSkAMBxAJoBgIj0EZEZIrLJTeuL+LoILSJSE8AHAGao6kNx7j5UVXNUNVdVL1LV9b60fF9c1jU1B7BZVbf78i+Poww7AeyF88Pdo6pfAvgCQLzzENJV6PVIVb9W1Z2qusOtQwVw/ojEcw05qtpKVa9T1Z2+tMh6dGvENbWAU4eaA1il7m+wK5569AqcsfopAObDqUMAkPYz8GMUej3yGQxgvDrdz/EI/X6kqp8BGAZnCGiZ+7UVVaMehVqHRKQ/gLqq+nY5ryFt70WJzAH5Ek5L7TH3W/lwWos5vq/aqvqwiGTDqdiPAWiqqjkAPoLdXRMz93gT4FxYsucA+P/zo14TgDUA9hOR2r78LeM4z7x9nLtKCLMelVacJB+rRD6Av0dcUy1VHQunHuW63fclYq5HqlqsqsNUtbWq5sH5xV/lflUZYdcj9wPRACQ4/FKGoO5HUNVnVbW9qjaF8/+TBeDH8l5AugixDp0M4AhxVuGsBTAQwM0i8n55r8mV8veiRJ8D8iSAU9yx8zEA+ovIaSKSKSI1xFkLnQdniCEbwHoAhSLSBwl+0hdnpvC7cHoQBqtqcUR6a3Em3rRO8Jr8ol6Tqi4HMAvOHI7q7rhY/ziOPRXACgB3iDO5rAeAE+F08VU1YdSjliLSw/3Z1RCR2+F8evnaTe8lIslqEL4I4BoROUoctUWkn4jUBTAdTtf5UBGpJiLnwpkfFOt1NBCRtu5xuwB4HMB9kb8XVUTg9cjnHACb4X3qA5A+9yP3WAe59aglnBV6T6nq5iSUO52EUYfugbMCqqv7NRHOPeMyoGrcixJqgLhdhaMB/E1V8+Es4boTzg8lH8DtADJUdSuAoQDegfNLeiGc/+RoF/K8iDwfJflYOBN/TgVQIN7a+5Ku8xZwuo3K/QmwrGtys1wIZ3LPJjjdl6MjrmO+iFwU5dh73WP3BfA7nIoxSFUTngGdrkKqR3XhjIduhlNXTgfQR1U3uukt4Ix/lpuqzgJwJYAR7vmWALjUTdsD4Fx3exOcTz/vRVyHv35HagTnk9d2OLPfX1FnZViVE1I9KjEYzqflyD8UaXE/grPa4U04cwm+hfPH6J7yljndhFGHVHWrqq4t+YLz4Xq7OquigCpwL5L//b1JTyJyN4D1qjoq7LJQ+hKRl+AsS6yKPVKUJLwfUXlVhXtRpWmAEBERUfrgy+iIiIgocGyAEBERUeDYACEiIqLAsQFCREREgSvzLXmnZAzgDNUQfFo8LlkPxUoJrEfhqEz1iHUoHJWpDgGsR2GJVo/YA0JERESBYwOEiIiIAscGCBEREQWODRAiIiIKHBsgREREFDg2QIiIiChwbIAQERFR4NgAISIiosCxAUJERESBYwOEiIiIAscGCBEREQWODRAiIiIKXJkvo0tr3Q824aoT60bNlvvItCBKQ0RERD7sASEiIqLAsQFCREREgWMDhIiIiAJXaeeAbDy4jomvHfyBlXZV/WUmPvyYi03c7OwFFV4uSj1rJnS2to/cP9/EX/7azsR1ptey8jWcvzvuc2Wv3Wriop9+jnt/Ikp/RSceZuIbXnjHxM+1b1da9qTZOvBoE+d8t8Erz6IlFXreaNgDQkRERIFjA4SIiIgCV2mHYBq+PN3Ek6YfY6V1/XC5iYd1mWTiV/J6WvkKV66qoNJR2LJa5JnYXwcA4Mzam01c3OILE2ecYLfXi1Hspfna8v7vR6Zdld/LxD+8aNdLf50lospr+WnZJm6QuS2w867tt8fEey/x7ksNzgisCBb2gBAREVHg2AAhIiKiwFXaIRi/yNUGk37vauLrGnpPQn2xWQN7Rw7BVFq72jc1sX/IBQC6PX2jif1PyvUP2wDA8gtbmrjBwkITb+oU/deq4U9evua3LbXS9szznt6rM3+Iegwqv41XesNfLS+xVwAs/M2rG3t2V7PScsd627VWel3nxd/9lOwiUiUi1apb2yed9F0o5ag7t4aJz7/iSxN/kWPf24oKfg+kPOwBISIiosCxAUJERESBYwOEiIiIAlcl5oD434wLANc0fM7EfWZfZeLmHHevkiKXzUZTmL/S2s59ZGWp+XLfj+28v6O7tV1j5rex7Ujl9ufb3zTxeRFzgNC2jB17eeGywh0mfmr9ickpWAy+/a2ViWv/s76VlvX57MDKQbHbes5h1vbTuc+YuPOEG0zcHt9UaDl276cmHrrfQhNPqWs/DRqcA0JERESVFRsgREREFLgqMQSzpW1ta7tZZk0TN38oM+jiUIrJCKkdXuMDDrmE5ek7/2jivx1i//z3W+B1U2/uLFZa9UMKTPzoQe+Z+Ilmdtf5pB3eyzD71YrtSZc71XtK5Te77XtWrxp7vQ3fudoNvNrK1+HzmE5FAdAe3uMenn3kKSttzBZvGK3T3d5jIooquEzHnPpjBZ8hPuwBISIiosCxAUJERESBYwOEiIiIAld554D4lt4+8fdnraRnC3zr7L7l0tuqaOOB3tsoY12GS5VH7Xe/8cXR89Ur4xjP7N/LxA/0aG3v96X3ePdHe7WLqUxZO716WHveGiut4dTxJj64uu9x8MvsR8VT6th8h7dMOy+r0Eq75cZ+Jq62ueKWTmc129/a/lfL/5h4r4bf/xB+CYiIiKjKYQOEiIiIAlephmAyc7ynAm65b7uJ1xbZTwv89LQDfVt8421VVKffWhOHtQyX0lvh2nUmrj1+nZXmX05Z+92NcR973ZBjrO0Dq3u36sc2dTRx63/9apcp7jNRsvjfsAwA4w7+h4lH/36IlVbts2CeWPvTfS2s7b3q1czBy3qbuOi39YGUJxLvvERERBQ4NkCIiIgocJVqCGbh/Z1MvODgESY+4tvBVr7mK38KrEyUmr44eJyJI1fB3HvFGBPP/mPrmI73yYgeJm748vTyFY6qpKxWXnf5iDtHWGnVxHti87invK7zhmtY11JFxtkbrO3mWd5Ku5ffPN1Ky8O0CitH5oHeEN2Yk0dZabvVe6Luisc7mLj27op9CV407AEhIiKiwLEBQkRERIFjA4SIiIgCl9ZzQLJa5Fnbi899zsRD8k8ycfNzOOeDgF8f9ZbJZWCOL8Vuh2dKbE9GPbz2MhPff993Ju65/TorX923ZsReSKqyFv4p18RHZttv4Z2/Z6eJG/y0A5QaMhs3NvHdHSZFzZf3YMXN+Yi08LocEx+Rbb9f99nNXUxce3w48z782ANCREREgWMDhIiIiAKXdkMw/mGXQyeusNJm7lYT59/R3sSZVnc7VRm+FxICwOd/9J5MWIyaJu747vVWvs7/yDdx4croT8r9Pq+nic/8ZqKJ//HgSCvffb/6loHz5Yfks7vfkSae84cnfCnZVr5rb7rJxDWnfVvRxaIYSa0aJj6t1u9WWveZg0y8PxYEVqZGrTdFTXtj6RFePvwcRHHKxB4QIiIiChwbIERERBS4tBuCWfBnb6b4hCbvW2ndnr7RxLlfeLOOM7t0sPLt3r9uTOdadYLXDSre6A7UnqCOwvberPTijXbXaae7vK63oi1bYjovJccvN2da280yvWGX85b0M3H7m+xVKrG+0Ms/PNPxPW/ly6Jz7SGYVSd69S2Xvefks6KP9xmwjnj3jguWnmLlq/Wf702soFRRvKnAxPevP8xKu7DtLBNPbdbWSitcsxbJ5H+K7tdd3/Kl2H0MO2c08m1xCIaIiIiqIDZAiIiIKHBsgBAREVHg0mIOyMYrvCdYLjrXe0tkrG8xPbP+WCtft2xvvwxfGyzyeNHSMiLabWWldV/kzUtpMiK4p+ER0Oolew5IZx1i4o5/+S2p52rzb2/mSPG5sT1JlaqejLr2/LNLev7XxFuKd5n4twcPsPJl755ZsQWjhBRv3WriT1Z1stK+6vqmidd8WN9OG3UM4lXQxZv9U6e1veT36ObLvDIh+v1HUmwCEXtAiIiIKHBsgBAREVHgQhuCycyxu6R2H9bOi/+82UqbfrA37JIB/xpYu/10dm1vSdTqvV4X1ZBRNyKa3EeSOyyyZkJna/uMK7wu1tkj2N4LUo3F66ztthetNHGsS21j5V+yHTkMR1Ri8fADre0PG3lLts9afJ6Jsz/ikEu62e/eGtb2CcMvMPG/D3rVSntk2PS4jz9rtzekXBRxjzmi+h7fVsRzInxaPuM9iTkVBop5pyQiIqLAsQFCREREgQt0CObXR72Zv0P7fWSlXZXzmYnLWmVyVf5JJp4x2X7ZWO5UbxZ59dlLvO9vCW71SbOz7ZcOzTnyUN8WX0RW0fwrppoPWmqlFZ5Qceftcdo8E5c1C52qnt8vPtrE8wY+baX9UrjXxNse8V60mY01FV8wSq6IF03W7+vFl/QaaqUVtLefmB2Lhi9GH7ZZ9Z43tDf7qFej5vOv2kkF7AEhIiKiwLEBQkRERIFjA4SIiIgCV+FzQLb/x3ui30++5bTVxH5K5V712kKTdthLdJ+/8BwT60xvnK0los/tKIq/qBXCX16qeKfe8LWJf9zSvELPldXCG7N/ocVEExdHtOtzv0itcVeqeFm5Xt27+Z63TZwt9i33j99fYuLGk7n0trLKnDLH2m44JbnH37nM94Tdo6Ln0x5dTSxff5fcQiSAPSBEREQUODZAiIiIKHAVPgTzxcHjTOxfnrg34qU4Hd+93sSd/5FvpelKDmNQ6fzDIABweO2pJp741nFWWi7WJvVch05cYeJieBXaX5cBoP23M8p1Xkp9kmXfSg/90Hvq7oA6G038xtYmVr6m9/hfeEmUIN/DT8t6EnMqDLv4sQeEiIiIAscGCBEREQWuwodguj1d+ovgIlcG+Lupk/2iMKq8CvNXWttzd7Qy8e6GGpk9ft29p+3+ft92K+neJnNNPOy3biaOHEJkfa4CDu1obd7f5PVSsz374ABrO+f7+F9KRvQ/fLe6dHoSM3tAiIiIKHBsgBAREVHg2AAhIiKiwFX4HJDcR4J7Ey3R+0u9ORsPn/2GlTZs48UmbvWmt4R2+YUtrXwZx2w28YeHPWfiZpk1rXzHzzvfxA2u9N7EXLhyVbzFpjSU2aWDia966/2o+bq84i3Lbv06l2RT8hXXKH3ex/qi3QGXJD7sASEiIqLAsQFCREREgavwIRiiIDV/yHvJYYu3N1ppc4c+Y+JqN3n59qr96kL/ixJ7zBtk4t3jmlr5Gr7sLaHkUtuqZ+F1+5m4f60tUfPlTdnjbWgSloYTRRhz+vMmXrDHG4654NU/W/nKeoFrGNgDQkRERIFjA4SIiIgCxwYIERERBY5zQKhy+dZ7c/LwvhdaSQuvbWDiE7r/ZOIZkw+28uVO9ZbU1p+9xMRFW35NWjEpPe3q393En/f/py+lVvCFIXLdt/RME28fmWviluNTa85HJPaAEBERUeDYACEiIqLAcQiGKq2iBYut7fZDvXi17/tlLU0rippCVdHqHt4S7ZZZ0Ydd3tjaxMTVtnjLcLkIlyrEyd5bwWtjZRkZUwt7QIiIiChwbIAQERFR4DgEQ0RUTg9t7GJtTz+ttYl1zQ8gov/FHhAiIiIKHBsgREREFDg2QIiIiChwnANCRBSjA/7qvQG5718PKyPn2oovDFGaYw8IERERBY4NECIiIgqcqPLZfERERBQs9oAQERFR4NgAISIiosCF2gARkV4ikj5vzolCRF4VkQfCLkdVVYnq0XARGRN2Oaoq1iMqL9ah+MTUABGRZSKyU0S2icha9w9unYouXBnlOUFENJ4/+hHXsC6saxDHXSKyQkS2iMhbIlIv6HKEIRXqkYi0dM/v/1IRuTXG/aeIyC53vw0i8p6INKvockcpyxARWeKW5T8i0jyMcgQtFeqRW44vRGS9+3v8vYicFce+qVSPzheRBSKyVUR+EpGzwyhHkFKhDolIExEZKyKrReR3EflaRI6KY/9UqkMni8hCEdnh/l60imW/eHpA+qtqHQBdAXQDcEciBS0vEakG4CkA3ySwe8k1HAbgCAB3l3L8in42yiAAlwDoAaA5gJoAnqngc6aSUOuRqq5Q1TolXwAOBlAMYHwch7nB3bcDgBwAT0RmqOh6JCK9ADwI4CwADQAsBTC2Is+ZYlLhfnQTgGaqWg/AVQDGxPkHIBXqUS6AMQBuAVAPwO0A3hSRJhV53hQRdh2qA2AmgMPh/A6/BmBSnA2hVKhDjQC8B+AeONcxC8Dbsewb9xCMqq4F8DGcH1pJAY4WkWkiUuB+EujlS7vM17r+VUSujvecEW4F8AmAhYkeQFVXAZgM4CC3jCoi14vIYgCL3e+dISLfudc0TUQOKdlfRLqJyBz3mt4GUCOO0/cH8LKq5qvqNgCPABgoIrUSvZ50lAL1qMQgAFNVdVm8O6rqJjgNl5J6tExE/iIi8wBsF5GsfVxTGxH50r2mTwE0iuP0ZwAYp6rzVXUPgPsBHC8ibeO9jnQWZj1S1XmqWliyCaAagBYJHCfMepQHoEBVJ6tjEoDtAKpMPQqrDqnqr6r6uKquUdUiVX0BQHUAHRM4Vph16FwA81V1nKruAjAcwKEi0mlfO8bdABGRPAB9ACxxt3MBTALwAJzWz20AxotIY3eX3+DcLOsBuAzAEyJS6iMERWSkiIws49ytAFwO4L54yx1xnBYA+gKY6/v22QCOAtBFRLoBeAXA1QAaAhgFYKKIZItIdQATALwO53rHATgv4vgFInJcWUWIiLMBtC/PNaWbMOuRL5/AaYC8luA1NILzs/fXowsA9IPzaaTpPq7pTQCz4fyy3w9gcMTx54nIhWUVoZT4oESuJV2FXY9E5EMR2QWnR3YKnE9/8V5DmPVoFoAFInKmiGSKM/yyG8C8eK8jXYVdh3x5u8JpgCxJ4BrCrEMHAvi+ZENVtwP4xf1+2VR1n18AlgHYBmArnJb+5wBy3LS/AHg9Iv/HAAZHOdYEADe5cS8AK2Mpg5v/fQAD3fhVAA/EsW/JNRQAWA5gJICabpoCOMmX9zkA90fsvwjACQCOB7Aa7jNU3LRpsZYFwBAAPwNoDaA+gInu+Y+J9VrS9StV6pHvGD3d8tSJY58pAHa49WgVgDcANPZd3+W+vFGvCUBLAIUAavvS3gQwJsZy9AawAcAhcIbxRsEZSrog7J9zFaxH1eD8Absl3eqRm/8K9/+z0C1Tv7B/xlWwDtUD8AOAO9KtDgF4GcDDEd/7GsCl+9o3nh6Qs1W1rvsf3AleF00rAAPcT/0FIlIA4DgAzQBARPqIyAwR2eSm9UV83Ttwj9MfQF1VjWlsqYxryFHVVqp6naru9KXl++JWAG6NuKYWcOZsNAewSt3/ZdfyOMrwCpyx+ikA5gP4wv1+2s+cjlGo9SjCYADj1RkKi8dQtx7lqupFqrrelxZZj6JdU3MAm9X5tFAi5nqkqp8BGAan23WZ+7UVrEeB1yNV3auqkwGcKiJnxrFr6PVIRHoDeBTO/2N1OB+yXnI/jVd2KVGHRKQmgA8AzFDVh+LcPfQ6BKchF7mQoh6c+1GZEpkD8iWc3ofH3G/lw2lZ5fi+aqvqwyKSDecG+RiApqqaA+Aj2F3HsToZwBHizFheC2AggJtF5P0EjlUaf4MiH8DfI66plqqOBbAGQK7bfV+iZcwnUS1W1WGq2lpV8+A0Qla5X1VGiPUIgPmlH4AEh1/KEFmPSr0mOPVoPxGp7csfcz0CAFV9VlXbq2pTOP8/WQB+LO8FpJOw61GELCRv7kRQ9agrnDlQs9x700w4w0m9y30FaSLMOuQebwKcDw7JmtdWIqg6NB/AoSUb7nHaut8vU6LPAXkSwCkiciicGdT9ReQ0dwyxhjhrofPgtKizAawHUCgifQCcmuA574Ez07er+zURwItwxuBK1l8n67nyLwK4RkSOEkdtEeknInUBTIfTXTVURKqJyLkAusd6YBFpICJt3eN2AfA4gPtUtThJZU8nYdSjEucA2AyvBwoAICKtxZmU3LqcxwfKuCZVXQ5n/P1eEakuzpyh/rEe2D3WQW49agngBQBPqermJJQ73QRej0Skk/tJuKZ7H7gYzvDsl256WtQjOKswepb0eIgz/60nqtAcEFcYdagagHcB7IQzvFMckZ4udejfAA4SkfNEpAaAvwGYp6r7XCiSUAPE7eYZDeBvqpoPZyngnXB+KPlwlnJlqOpWAEMBvAPnZn8hnIZDqUTkeRF5Pso5t6rq2pIvOD+07erM/gWcIZJpiVxPKeeaBeBKACPcci8BcKmbtgfOrN9LAWyC0xPzXsR1bBORnlEO3whOi3k7nJU4r6gz+7nKCaMe+QyG84kgstHaAk73Y7l7pMq6JjfLhXAmPm+CM5wy2r+/iMwXkYuiHL4GnHHabQC+hdMwvqe8ZU5HIdUjgTPb/zf3PDfBmZ82x01Pi3rkfvofDuBdEdkK59P9g6r6SXnLnU5CqkPHwpnMeiqAAvGeS1TytyNd6tB6OBNg/w7n/+QoAH+MpVyV5mV0IvISnGWJH4ddFkpfInI3gPWqOirsslD6Yj2i8qoKdajSNECIiIgoffBldERERBQ4NkCIiIgocGyAEBERUeDYACEiIqLAlfmWvFMyBnCGagg+LR6XrAcjpQTWo3BUpnrEOhSOylSHANajsESrR+wBISIiosCxAUJERESBYwOEiIiIAscGCBEREQWODRAiIiIKHBsgREREFDg2QIiIiChwbIAQERFR4NgAISIiosCxAUJERESBYwOEiIiIAlfmu2CIiIgoHEUnHmbibbdvMfHXh75j5Tv5qqtNnD1pZsUXLEnYA0JERESBYwOEiIiIAscGCBEREQUu7eaAdJxVzcTfD+9mpdX48Nugi0NElcyOc44y8erjJaZ9fhn4vLXd9u1rSk0btPx4K9/XM7qYuPlUNXGtf38TW2Ep7WV09epA4+dWWmmP5D5t4iaZtaIf5Kb1XjwpaUWrcOwBISIiosCxAUJERESBS4shGP9SpD838bqkrtjQOYziEFEl4h9yAYCvnh0VNa9/CGV0q6lR80UOyUTdx7890At74morG4dk0ptk2X9qt5x3hInfePQxE7fMsodZvttT3cTLC70huiOz7aHBOtV3m3hv+YoaKPaAEBERUeDYACEiIqLApeYQTEamtdng/uUm3lTsFTlr1SYrX2GMh/95ZHcTN5zjnavhS9PjKCSRRw4/0MQ3v/2ulbZsTyMTf3DKoSYuXLW64gtG+xTrSheg7GGXZIocBvIPyXA4Jj1ktW5p4mqv7bbSJrUbaeJVRd73uz5xg5WvxftrTbz0gv1N/OPVI5JVzFCxB4SIiIgCxwYIERERBY4NECIiIgpcSs4BWX91d2v7zAYfmvivvTbrwWAAACAASURBVC8wcVH+0oSOL3u9dle1837zEl5K6HBUATIbNbS2pV7d0jPu3mNtxjqvIjOnvnfsBvtZafq799ZJqV8v6jH2tPD2u/6lcSY+peZOK9/G7J9NPDHnOC9hVUxFpQrmfwIpALTFNVFyxn4M/zyNJU8cbeIeR/9k5Yt1Tol/nkq7f8ddPApAVl6utX3E+7+Y+MYG9lO6uz1+u4nz3l9j4uZLpln5iqyt/RHNgnwvrR3WRM2XatgDQkRERIFjA4SIiIgCl5JDML93tLszn5p4honbLCn/Utn6i71217knf2fizxClm58Ct/Cx1tb24lNeLDXf17vtNvRdt3nLFVeeYtejGo28oZG+B8w38aP7f2Hle3xzexPfst/i2ApchmO/ut7Ebed/V0ZOCkPkstZkD3G0+9MME3/tG44BYD8JldJaUbMG1va7b7Uy8azXW1tpzVZ6Qy32MEtiOjzi3duKk3C8oLAHhIiIiALHBggREREFjg0QIiIiClzKzAHJPLCjid8562kr7eLXbq6w8w6s972Jv+h8mZVWtKD84/+UmKzV1a3tkQVtTNy0WoGJF+y0l75d85D3GPQW1TZaaT2yYxsdvb2Bt3yuSMvIGMVZi/tZ2x3+5C0NTsZ4L6WeyDfqRnu8e+Qy3FhFLvOl1KMzf7C282Z6cayvCYnV7D32naT4x4VJPkMw2ANCREREgWMDhIiIiAKXMkMwv1zgLWFatreRlXbAyytMnIyurBr91pm4mu/7O9rkWPmyFyThZJSQNnfay60nP9raxLKf93MqXLrcyvdti2NMXNSkvpX2e4fyLbNudu0v1va4th+b+JdCbxnc7rubWvky1nHpbWXkf8LpLwOfT+qxBy0/3trmG3CrJv/UhBv++IGJL/jvVVa+dpgbWJmSiT0gREREFDg2QIiIiChwoQ3BZHZub21/dMk/THzaO7dbaW3zy/f0UznyYGv704O9p2oePOlPJu7w0UxQaioq+N3b8McRCvNXehv+GEC92fGfN6NWLS++rlbUfOc969XZ5v+dFjUfpTf/apeKHHZZ+mhnK60WOARTFS28zhtu7ryzsYk73ma/yTJdV9exB4SIiIgCxwYIERERBY4NECIiIgpcoHNAJMs73aKrGlpp/pZQx8eXWmnlXXqbsWyttT1nTw1vo6j0JxYSAcDiew818aIDno2ar05+Or2DkmK1JOLttcme9+G37pgtJuacj6opK89+svOrfUaZePCUISbusG5WYGWqSOwBISIiosCxAUJERESBC3QIJqOh97TTyK7M0xcONLGusZcYlVfR+vXW9s0/DoySk8hW3HhP1LQf9uw1cd0Vu4IoDgWsIodcIjWdXs/E/7MMl09CrRIWPLC/tX1A1jYT7/9pyjy4PGnYA0JERESBYwOEiIiIAhdon47u8F7YddmKnlbafW0mmHjg89dbaXV+9YrZ8k3v5WOFKxMbqtkx11uBc8f53nkndDkupv2LFtovJUNxuj6HrmpZe/OxJt5Tr4yMPid2mhc1rV01NfHiK7w6WuPkY0vLDgCo032DiWu+ZL/8sOaEb2MrFAUm8qVwo1tNrbBz+Y896M922rp/V9hpKWSZTZuY+LxD5lhpx79/q4nbj50RWJmCwh4QIiIiChwbIERERBQ4NkCIiIgocIHOASneutXEa0+ubaVdPfgmE2d0sJ8qeewf5pq4x+WLTbyhMPpAfoZ4xyhWu53VpcbLJj691m4TX/XZO1a+YesPNPHYyd5YcNv77LesFu/YEbUcFJ6NVxxjbc+6/RkTZyD6E3AzxasvRRr9Cac1pbqJl5z2QtzlO3J/e65TzbiPQBXN/3RSAGj7xDXlOl6sy3oj55r4z9vuT5VvLkBVtmqUNyfx+nr2z33+c95b4ytypmFmTn1re/EdXUx84el2mabd0N3EGV/NRXmwB4SIiIgCxwYIERERBS60R6sVb99ubTcZOc2LI/Iu88X59byuIWRn28fM8/Zcd6zdpeTX49LZJu5Y7XMTX9d/iJVPF3kvxWuze7p3nqhHplTS8OXp1nbXJjea+IUrR5i4RZY9hFbNFzfLqhP1+BuKvDp81dKzTfz9ijwrX4P/815+2OAVr0yNYZePUl95hz96Tr3a2v7q2VFRctp6HP2TideVqwSUCjZf6g0Pf9/9ORO3fdse4mv3U3KH2zIbNzbxb2e3M/FVt7xv5buw7scmPviTG6y0zj96j6Eo77AQe0CIiIgocGyAEBERUeDS7u02RVu2RE/0vXSuSRmTc6ed1tHEgxp5wziybpOVr3j3blDlkfeQN8x330OHRc2X6Xtp4gfzPjPxbi208vW78zYT57zuDae0g/3yQ6IS//NSuWdj28+/KuY0dE1iiSgIGbVqWdtn3vKFiT/Z4Q36th+91cqniF9WXq6Jf7or10r7sM9TJu5Uzfvb9/lOezpD3xuGmrhDxBOak7kahz0gREREFDg2QIiIiChwbIAQERFR4NJuDkiyDfziWhN3WDcrxJJQqntnm7281j/vg6qeHeccZeL/mdtB5LPwsYOs7Q8bek/EPWGo9zeo9tzY6tHe3odb283v85bGvtbqg6j73bS6l4k//cSbB9f6LvteVhPBvJmbPSBEREQUODZAiIiIKHBVcghGP/Fe/nPMBQtMvDGMwlDKWXVxJ9/WZ1HzUdWz5ImjTex/sdygPx9v5fO/xM4/VBPrk08j+Z+Q2Q58GV06yGrhDdm+drr9ssqzFvczce3x3rBL5oEdrXz5/by/VW37ecMs77QdYeVbXegt2O278HwTr/2wpZWv2QhvmkHrveEPIbMHhIiIiALHBggREREFrkoOwfjd3PwTEw875BIrrXjewqCLQylga4dkPuuP0pl/yAWwXwrn539SKQAMmn68Ly2xYRe/8r4Ej4K38BZvCKZHtv0K0+UftjHxjidamXj0WSOtfEf7HlDqfxLzxb+eaeXb9EhrE2dPmmni/bHSypfIk1UrEntAiIiIKHBsgBAREVHg2AAhIiKiwFXJOSDNJyw3cfc7vTcR7mhVz8pXY15gRaIUMvK0V0v9/nvrIt+gu7bCy0Lhipx7MXr1dzHtFzknJF7+ZbcAl96mA6lW3dr+e9+3o+b97k8joqb5Dck/wcSLnjjQxHXftutDdpq+gZs9IERERBQ4NkCIiIgocFVyCKZonddd9Y9NbU181WPjrXyjP2gRWJkodZxea7eJi3zr1pZPOMDKtz+HYCiJel5/tYnb/ZtDLulm90mHWNsD6kT/GV60rLeJf/jQe/JyvWX2ct39JntP6q5bUPnqBHtAiIiIKHBsgBAREVHg2AAhIiKiwFXJOSC6d4+J/zX2NBPXPHqDla8Rfg6sTESU+k5r3tXETad7y/ZjXXYbuby2+VRvklGtf38TmZ3SSPWPZ1nbfXMjl+37bTJRHqZFzVXZXwrBHhAiIiIKHBsgREREFLgqOQTj1+Lv0bu/iIiiWXfMFhOfhq5l5PTwiaZEHvaAEBERUeDYACEiIqLAVfkhGKJIR959rYnHD/+HiRt9v7u07ERElAD2gBAREVHg2AAhIiKiwLEBQkRERIHjHBCiCA1emW7iK185zsRZmB1GcYiIKiX2gBAREVHg2AAhIiKiwImq7jsXERERURKxB4SIiIgCxwYIERERBS7UBoiI9BKRlWGWIRlEZLiIjAm7HFVVJapHr4rIA2GXo6qqRPWI96OQsA7FJ6YGiIgsE5GdIrJNRNa6N8o6FV24iDK0dM/v/1IRuTXG/aeIyC53vw0i8p6INKvockcpyxARWeKW5T8i0jyMcgQtFepRRHlOcOtQzH/0I65hXVjXII67RGSFiGwRkbdEpF7Q5QhDqtQjEflCRNa7///fi8hZceybSvej80VkgYhsFZGfROTsMMoRJNah5BGRo0XkUxHZ5F7LuFjLEU8PSH9VrQOgK4BuAO5IpLCJUtUVqlqn5AvAwQCKAYyP4zA3uPt2AJAD4InIDCJSoc9GEZFeAB4EcBaABgCWAhhbkedMMaHWoxIiUg3AUwC+SWD3kms4DMARAO4u5fgV/YydQQAuAdADQHMANQE8U8HnTCWpUI9uAtBMVesBuArAmDj/AKTC/SgXwBgAtwCoB+B2AG+KSJOKPG+KYB1Kjv0AvACgNYBWALYC+FcsO8Y9BKOqawF8DOeHBsC0gKaJSIHbiuvlS7vM17r+VUSujvecUQwCMFVVl8W7o6pugtNwOcgt4zIR+YuIzAOwXUSy9nFNbUTkS/eaPgXQKI7TnwFgnKrOV9U9AO4HcLyItI33OtJZCtSjWwF8AmBhogdQ1VUAJsOrRyoi14vIYgCL3e+dISLfudc0TUQOKdlfRLqJyBz3mt4GUCOO0/cH8LKq5qvqNgCPABgoIrUSvZ50FGY9UtV5qlpYsgmgGoAWCRwnzPtRHoACVZ2sjkkAtgOoMvcj1qHy1SG37oxT1S2qugPACDgfjPYp7gaIiOQB6ANgibudC2ASgAfgfKK/DcB4EWns7vIbnD+69QBcBuAJETksyrFHisjIGMogcBogr8Vbfnf/RgDOAzDX9+0LAPSD04psuo9rehPAbDg/pPsBDI44/jwRubCsIpQSH5TItaSrMOuRiLQCcDmA+8p5DS0A9IVdj84GcBSALiLSDcArAK4G0BDAKAATRSRbRKoDmADgdTjXOw5OnfQfv0BEjkN0kfUoG0D78lxTugn7fiQiH4rILjg9aVMAzErgGsK8H80CsEBEzhSRTHGGX3YDmBfvdaQr1iEA5f+b5nc8gPkx5VTVfX4BWAZgG5yuFQXwOYAcN+0vAF6PyP8xgMFRjjUBwE1u3AvAyljKEHGMnm556sSxzxQAOwAUAFgF4A0AjX3Xd7kvb9RrAtASQCGA2r60NwGMibEcvQFsAHAInG7zUXCGki6I9/8h3b5SpR4BeB/AQDd+FcADCVxDAYDlAEYCqOmmKYCTfHmfA3B/xP6LAJwA55d0Ndxn8bhp02ItC4AhAH6G0+1ZH8BE9/zHhP1zrir1yHeManD+gN0Sxz4pcT9y81/h/n8WumXqF/bPmHUoveqQb79DAGwC0DOW/PH0gJytqnXd/+BO8LpoWgEY4H5aKxCRAgDHAWgGACLSR0RmiDNBpQDOJ8Z4ughLMxjAeHW6nuMxVFVzVDVXVS9S1fW+tHxfXNY1NQewWVW3+/Ivj7UAqvoZgGFwusuWuV9bAaT9zOkYhVqPRKQ/gLqq+nY5ryFHVVup6nWqutOXFlmPbo24phZw6lBzAKvU/a11xVyP4PSsjIVzE5oP4Av3+6xHAd+PVHWvqk4GcKqInBnHrqHfj0SkN4BH4fw/VofTOH5JRLqWtV8lwTqUhDpUQkTawRmSvklVv4pln0TmgHwJ51PjY+638uG0rHJ8X7VV9WERyYbzh/YxAE1VNQfAR7C7juMiIjUBDECCwy9l8P8hiHpNANYA2E9Eavvyt4zrRKrPqmp7VW0K5/8nC8CP5b2AdBJiPToZwBHizHxfC2AggJtF5P3yXpMrsh79PeKaaqnqWDj1KNcdTiwRcz1S1WJVHaaqrVU1D04jZJX7VWWEfT+KkIXkzZ0I6n7UFc5cullunZoJZyigd7mvIE2wDpX/b5o7rP0ZnB7f12PdL9HngDwJ4BQRORTODOr+InKaO4ZYQ5y10HlwWtTZANYDKBSRPgBOTfCcJc4BsBneJz4AgIi0FmcSYOtyHh8o45pUdTmcMbp7RaS6O0bfP9YDu8c6SBwt4cwefkpVNyeh3OkmjHp0D5wZ413dr4kAXoQzlluyjj9Z7yd4EcA1InKU+/OuLSL9RKQugOlwuj2Hikg1ETkXQPdYDywiDUSkrXvcLgAeB3CfqhYnqezpJPB6JCKd3E/CNd2f38VwhtW+dNPT4n4EYCaAniU9HuLMW+qJKjQHxMU6lPjftFwA/wdghKo+H0+hEmqAuN08owH8TVXz4SwpvRPODyUfzlKuDFXdCmAogHfgNBouhHPDL5WIPC8i+7qAwXBacpF/JFrA6TYq9yfAsq7JzXIhnImGm+AMp4z27y8i80XkoiiHrwFnfG0bgG/h/CG6p7xlTkdh1CNV3aqqa0u+AOwEsF2dWeSAU4+mJen6ZgG4Es6s8M1wJrld6qbtAXCuu70JTk/MexHXsU1EekY5fCM4n7y2w+n2fEVVX0hGudNNSPcjATAczoTE9XCWUw5U1Tluelrcj9xP/8MBvCsiW+F8un9QVT8pb7nTCetQuf6mDQFwAIDh4ntOVyzlqjQvoxORuwGsV9VRYZeF0peIvARnmfTHYZeF0hfvR1ReVaEOVZoGCBEREaUPvoyOiIiIAscGCBEREQWODRAiIiIKHBsgREREFLgy35J3SsYAzlANwafF45L1UJuUwHoUjspUj1iHwlGZ6hDAehSWaPWIPSBEREQUODZAiIiIKHBsgBAREVHg2AAhIiKiwLEBQkRERIFjA4SIiIgCxwYIERERBY4NECIiIgocGyBEREQUODZAiIiIKHBsgBAREVHg2AAhIiKiwJX5MjoiIiIqXeZ++5m4qH2elbb4uupR92v7r2ITZ3w5N/kFSxPsASEiIqLAsQFCREREgWMDhIiIiALHOSBEROWU1cIe/19yVQsTn3L6HBOPyP0mpuN1+u8l1nbbob+ZuHDtukSKSEnin/exaFhHEy8c8GzMx9jde6+Jj511mYlbXLvZyle4Zm0iRUwb7AEhIiKiwLEBQkRERIHjEAxVWlLNXga3/M4jTLy7cZGJG8y12+ENX5nhbajGdq4s71dp1Z+6W2k1Tthg4ksPmG7ij0452MpXuGp1TOei1LDj3KNMfMLfpllpExtPLHWf3Vpkbe/SQhPXkWwTz+/xmpXvkGe9IZm8P3jDMbHWT0qehfd3MPGic2IfdvHLlmomnn3kGBNP/a99z7rnzitNXPftGahs2ANCREREgWMDhIiIiAKXdkMwWa282eX557Ww0vL6LzPxBx0+NHGm2O2st7Z6s5jvf+0CE7d49FsrnxYWgtLX+ssOt7bnXfVM6RnPsTfPGn+SiYsKfo96fP8Qz+JHDjPxwoFRzhPhw8Y97W9wCCblZNSubW0v/OeBJv7hjKdMXFPsrvMVhTtMfPKXQ03c7INqVr46y718y86qY+KvBz1m5Zt39OsmPjO3v4kLV64q+wIo6WqvyCz1+8UotrY7fXKNt8/CbDuv7y/vW1c+buLja9jHnPjYP0181NG3mrjjsJ+sfEVbtpRd6BTFHhAiIiIKHBsgREREFDg2QIiIiChwKTMHJLNDWxMvuraxlXbzqZNN/Ie6/zVxk8xaUY+3sXinFxeJlTagjjdWN+D6ESY+bPsNVr79n7KX1lHq23P6kSZ+4q/PxbTPz3v32N8ojnFp40HtTbhwYPzL8ZadlWNtt/wu7kNQBVt+y6HW9pL+3v1iVZE3R+y0b6608rW527v/tF80B1Fl+OYT9PeWb9cSe67IluJd3kaRvZSXgtWiz7JSv3/snIus7Q6XzY7peDfO8+YIPf70CCvtkOrepJCF53v3mMPbXGzlaz5wt4l1926kC/aAEBERUeDYACEiIqLABToEk9n+ABP/PLy+lfZBD697qUO1iLVIPuO355p43o6WVtqv2xuZeNnT3kuCcub8ZuU7833vhVBX1s83sabMgBQlqvZfV5r4mOzYuqrP+L8bre0OW2YltUzR+J/GSqmpzeh8a7t9W2+opeOT3jBLy+9+sPLF+pPVY7yn4f50uX8Yz74ZHTnyZhPnreHQcJg+6viRiff6RmsbPlgzoePV+MB7/MOtRddbaW3/tsDEz7f40sT+p6cCwBFve0MyzQb8YmKNHF5OMewBISIiosCxAUJERESBYwOEiIiIAhforIfCRnVNPPN4e9livQxv3sfM3fYyyOEDLzVxxrK1Ji5avz7iDJtNVBfemwMjx2Ofe/EsE195m7fsaW8dUJrZcc5R1vY/Wo6M+xit3pV9Z6oAHV/aam0XR8lH4SlcHjEH5FJvu6yfV0YN73627C/eY/qLOm238k0+1r/s0nuswFvb7EcRtHzSW6PNehKu83892cRvtPnExFlbdln5Epnhlf3RTGv71yLvDd7zn//cxAdWt/90z/LNCTn6cu9xEo1GTUcqYw8IERERBY4NECIiIgpcoEMwMv17E3d/41YrrdH33rBLzg8FVpr+6C1xS8bCxRobSn/S5cGnLLK2f783CSejpMvM8ZZwd/rrj1ba4dmRufet9gJ7mbb/Hcj+N95Gvl2397Wp3b1J4Vl/STcT/3jViDJyesMuX+/2Pg++OuRMK1fGjrlJKxuVz6wlrb2NNrHts/r2Y63tQ8/x3mb70+jOMR3j3KnXmnhR7xej5tviPVQcjaLmSg3sASEiIqLAsQFCREREgQvt2Z8H/DV693VFz/Le1aD0VQ+zl9pPVm2HjRVcEkpEcZs8E4/MG13u460YkGttVy9obuLzb/zMxLc0KKsrPTYDlvT1Nn5eVu7jUXiWvO4Ns0w+3q4bTTP997fYxgWPqO49tXLtUfZTNZt/FX/5qGLUnev7eZ7ihRuOaGDla1DLe8rt1KGPWWl1MnzHuOdzJNPDZ79h4ocW2y/IazJhiYn/dxVp8NgDQkRERIFjA4SIiIgCxwYIERERBa5Kvv91e17py3DbP11obZeei8K26Lpa+84Uhzk3PRM1LQPefKFkzE3a+FRrE9fa8U30jJTy/PM+2mZFfxNqse9Ocs7iM6y099tPMnG2eLfj/978Tyvf8YXeYwv2f5Jvww1T7hve4xqOLPbepN184gIrn7bc38Tri+2/JnUq8KP/mbW9J4KfOdyem7TuHu8Nzv0f/rOJm42xH2dQtGVLBZXOxh4QIiIiChwbIERERBS4KjEE438xFACc3/trE5/443kmrjnL7oai1FRrabWwixCXi5aeauLak7yXinGIL70NbXeiiYuPtJ9mmTFzQWR2AIAWrrW2T+s1xMT17vVedDeu7cdWvlm3e8OEPTZ7Lxvb7zU+jTdoRRu8xzM0fcYbDvufp3Rv9oZCBt1lP/l727neiyhz6/9u4g87vZ+cQkbRNNMbKvz2Lq9OPX5tJyvf51d5T271P8E82dgDQkRERIFjA4SIiIgCVyWGYPJvOsza/rCJNzO40yc9THyALg2sTJS4FpO9LktcH145YrVljzcEqLv5dN3KQvd6Ty6VaXY3dazDa5lfzDHxrh8amnjKN/YwY68ae0282Tfas1+M56Fw1R8zI2LbiyXL+zN8VsM+UY9R3KKJt6F2DctYWfpTTRc83MLa/vLkp0zczDccc0uDhVa+bx5pbeKt93Sz0jK+TN6LEdkDQkRERIFjA4SIiIgCxwYIERERBa5SzQGRI723D175hrecad7OvaVlBwBk7PHFB9lLkYp/XAhKPVLsPZN0W/FuK63Al9Z76o1W2n3dJ5r42BrLYzpXNd+Lk5tlRn8Ca6Z4bfkitZ+ZuqvQG8+P7b2oVBX5l3dePeFKK23RH0eaePg575h49F/tMX5KTVl59hu3tx/ivXE7+6OZJi5a91v0g5SR9j9LgF0dLrP3GdT3Tybu88gUE0fOAXm77X9MfN59/ay03SdEL2K82ANCREREgWMDhIiIiAJXqYZg/E6q6T1x8OzaBVHz/XT5syZeMWiHlTZsdV8TL3mqi4nrvmUvqaJgFX/vPWXyD5fawyxZn882cTvYy8VGo0WpcVn8T9Ft/EV1K+3lll94ZdJonaCAPu1bPodlMZ2XqrbiatEX8hYUJfdljFQxCgYdY+I/3fWWlda71koT973nNhPv92rFPtnWP9zz5TfePbDZ15utfBfUXWfix1qPt9IGX+A91bXe2PL9LWQPCBEREQWODRAiIiIKXKUagtGZP5j4sMk3mXhJv1FR99lSvMvE72yxn/j2r5ZTTDzyHm/VxIdv8fmDqcI/5FIRind59eOrhQfaib4hmLLsapBp4hpl5KNw5N99rLVdfYsXN316GsLwZJ/Xo6Y9NfEME7cBX0aXqvbU9ZbQ+YdcAKB+hncn+OrvT5v4tHXXWfmyJ89ERSnauMnEr95wlpX2h1efM3GbLPuu1X7oTyZeN7Z8ZWAPCBEREQWODRAiIiIKHBsgREREFLhKNQfE/1bBwzt7b7Z9ZGNnK9+HD55o4lprvUehbsu1l1nOv6GZif87xztGe3xT/sJSWsioW9fEw475IKFj7HfpChPr6HIXiZIg88COJn5ryONW2hU/DvI2nkbFyvDmBy19sLuJe9f8NiKjd29rNDfWd+1SmJo8680f6tHsNivtx8u8N7Jn+PsBQuoSWN/VfkZzpkiUnMBXC9ubuAPKNwePPSBEREQUODZAiIiIKHCVagjm56cON/HiA7xlRJ1HX2/la/NW6UvX6kdsr3vDiznsUjUVH9jGxBfVnZLQMZas8Z6E2harylskSoKtHXNMfGA1e+i1Qc0dkdkrTHGPQ0y84JJnfSn2rfnI2ReYuOn733n7V1jJKJnaPvmztX3JiaeY+PXWn5r4msfftfLde6m35LrVw95PW2fPT6gcK4Z5S86vHOC9cO7i+v+w8mUE9MAA9oAQERFR4NgAISIiosCl9RBMZtMm1nbf7l7X5NFz/2jiNndFzignis3QMe+U+xi5Y6sloSSUTHUXei/f+mHP3go9V9b+TU28dERjK+2T7s/4tryXzA1bf6iVr+n53pOY/U/npfRQtGGjtb3lDO9p2uNnNDLxSRFPTD3v2FdNvPd974WXexH95ZdlqSXRVq1EH3L569ojre0ud68xcWFCpfCwB4SIiIgCxwYIERERBY4NECIiIgpcWs8BWXB/K2v7mhxv3ezPQzt5CcWJjZdR1bT6Nm+p2uk155iYSx4rj6KfvGWR8/c0t9IeOWC8ic8ZdYOVljMvynyeiAdHbjl6p4nvPHyyiS+tt9rKt029J6EeNvMiE+dd/7uV+hGjWgAAIABJREFUr3iXvR+lt6LN3hykf3X0/o49dPMFVr5W5/xq4htyPzfxiTWTOw/ohlXHWduff9nVxB2ftR8dULhqBZKFPSBEREQUODZAiIiIKHBpNwRTeJL3tNOxvZ+30q546UYT502fBqJE7Gpcvpd9fb3L7qavuWq7ifkasdTzyIJTre25R3pDuUvOGGVnPgOlyhT7s1yRegN2/mW+7T+71srXebi3PHP/pQtMXN7ljZSe9n/S/ru1+0kvfqpNXxM/nlPHyrfo6tombvqVVxc3HGYfv94Sb6ywyYwtJpZFS618bXd4TwuvyLrIHhAiIiIKHBsgREREFDg2QIiIiChwaTcHZN0N3vKjguJaVlrrV34xMcdQKVHtH17obVwc//63PXS1td1wbulvX6bUkHebvaSx01Xe27Prdtpkpc08fKyJb13b3cQf/XyglW/v5mwTd37WW1Lbfv4cKx/vUxSrwqXLo6Z1uKb079d/o/TvA/Z8tLDmprEHhIiIiALHBggREREFLu2GYB4/xHs76XX/HmKltV3Lrm4qP/9TCsvSdcYgE7e6wnuLZaPt9hsnufQ2tRUtsZcgHvDnpVFyAn3hX9foDaC0wffRj59wyYgqN/aAEBERUeDYACEiIqLApd0QzF0LzzFx+3vmWml8WRglW9/cw6Km5WG+idnNTkQUH/aAEBERUeDYACEiIqLAsQFCREREgUu7OSANzvjZxJzzQURElJ7YA0JERESBYwOEiIiIAieqfE4jERERBYs9IERERBQ4NkCIiIgocKE2QESkl4is3HfO1CYiw0VkTNjlqKpYjygZKlE9elVEHgi7HFUR61B8YmqAiMgyEdkpIttEZK1buDoVXbiIMjQRkbEislpEfheRr0XkqDj2nyIiu9xr2CAi74lIs4oscxllGSIiS9yy/EdEmodRjqClSD1q6Z7f/6UicmuM+6dEPRKRo0XkUxHZJCLrRWRcWPU5aKlQj9xytBaRL0Rkh4gsFJHecezrv4Z1IV6DiMhdIrJCRLaIyFsiUi/ocgQtheqQvxzbROSTBPcNrQ65ZblQRJaLyHYRmSAiDWLZL54ekP6qWgdAVwDdANyRSEHLoQ6AmQAOB9AAwGsAJsX5H36Dew0dAOQAeCIyg4hU6LNRRKQXgAcBnAXnOpYCGFuR50wxodYjVV2hqnVKvgAcDOeRMuPjOEzo9QjAfgBeANAaQCsAWwH8q4LPmUrCvh8Bzu/tXAANAdwF4F0RaRzH/iXXcBiAIwDcHZkhgHo0CMAlAHoAaA6gJoBnKvicqSIV6pAph/t1aiL7IsQ6JCIHAhgFpx41BbADwMhY9o17CEZV1wL4GM4PraQAR4vINBEpEJHv3T+yJWmXicgCEdkqIr+KyNXxntM976+q+riqrlHVIlV9AUB1AB0TONYmOH9wDnLLuExE/iIi8wBsF5GsfVxTGxH50r2mTwE0iuP0ZwAYp6rzVXUPgPsBHC8ibeO9jnQWVj0qxSAAU1V1Wbw7hlmPVHWyqo5T1S2qugPACDh/RKqUsOqRiHSAc9Mfpqo7VXU8gB8AnJfANawCMBlePVIRuV5EFgNY7H7vDBH5zr2maSJyiK8s3URkjntNbwOoEcfp+wN4WVXzVXUbgEcADBSRWvFeR7pKoXtRwkKuQxcB+EBVp7p16B4A54pI3X3tGHcDRETyAPQBsMTdzgUwCcADcD7R3wZgvO+TwG9w/ujWA3AZgCdEpNRXjIrISBGJqeUkIl3hNECWJHANjeDcKPyv070AQD84n2ib7uOa3gQwG84fjPsBDI44/jwRubCsIpQSHxTvdaSzVKhHIiJwGiCvJXgNYdcjv+MB3+t5q4gQ69GBAH5V1a2+733vfj/ea2gBoC/senQ2gKMAdBGRbgBeAXA1nN6WUQAmiki2iFQHMAHA6+71jkNEI8j9g3NcWUWIiLMBtI/3OtJVCtyL3hBnGPUTETk0wWsIsw4dCKfuAwBU9RcAe+D0EJdNVff5BWAZgG1wunkVwOcActy0vwB4PSL/xwAGRznWBAA3uXEvACtjKUPEMerB+bRxRxz7TIHTNVQAYBWANwA09l3f5b68Ua8JQEsAhQBq+9LeBDAmxnL0BrABwCFwujtHwRkCuCDe/4d0+0rBetTTLU+ddKtHEcc8BMAmAD3D/hlXlXoEp7t5RsT3/g7g1TivoQDAcjhd1jXdNAVwki/vcwDuj9h/EYAT4DQ8V8N9ppObNg3AAzGWYwiAn+EM5dUHMNE9/zFh/5wrex1y8/eA83egFpwhoLUl5UijOvQ5gGsivrcKQK997RtPD8jZqlrX/Q/uBK+7uBWAAW4LqUBECgAcB6AZAIhIHxGZIc5kuQI4rbR4hiwsIlITwAdwfvkfinP3oaqao6q5qnqRqq73peX74rKuqTmAzaq63Zd/eawFUNXPAAyD03W/zP3aCiDtZ07HKCXqkWswgPHqdBvGI/R6VEJE2sHper1JVb+Kd/80FnY92gbng5BfPTi/y/FcQ46qtlLV61R1py8tsh7dGnFNLeDUoeYAVql713fFU49egTOXZQqcHrQv3O9XhftR2HUIqvq1OkN4O9y/ZwVwPhjFcw1h16GEfxcSmQPyJYBXATzmfisfTmsxx/dVW1UfFpFsOH9oHwPQVFVzAHwEu8svZu7xJsD55Uj2uJv/Pz/qNQFYA2A/Eanty98yrhOpPquq7VW1KZz/nywAP5b3AtJJmPUIMA3ZAUhw+KUMgdUjEWkF4DM4n2xeL3fJ01CI9Wg+gAMixrkPRfKGwSLr0d8jrqmWqo6FU49y3eHEEjHXI1UtVtVhqtpaVfPglH+V+1UlhH0viixOko9VosLqEJw6Y4aOROQAOMN4P0fdw5Xoc0CeBHCKO141BkB/ETlNRDJFpIY4a6Hz4MzRyAawHkChiPQBEO8sXwCAiFQD8C6AnXC6wooj0luLM/GmdYLX5Bf1mlR1OYBZAO4VkeruuFj/OK6jhogcJI6WcFYyPKWqm5NQ7nQTeD3yOQfAZnif+ACkVT3KBfB/AEao6vNJKGs6C7weqerPAL4DMMw9xzlwhsLGA+Z5EMl6z8WLAK4RkaPc+0ZtEennNn6mwxnKGyoi1UTkXADdYz2wiDQQkbbucbsAeBzAfZH31yogjL9pLUWkh/v7X0NEbofTk/K1m54WdQjOMHR/EenpfqC6D8B7as+PKlVCDRC3y3k0gL+paj6cJaV3wvmh5AO4HUCGW4ChAN6Bc7O/EM4YY6lE5HkRiXYzPRbOxJ9TARSIt266pLuqBZxuo3K33Mu6JjfLhXAm92yCM5wyOuI65ovIRVEOXwPOWP82AN/C+eHfU94yp6OQ6lGJwXA+5UT+gqdLPRoC4AAAw32/C/EOJVUKIdajP8JZ+rgZwMMA/uAbjmsBZxy93FR1FoAr4ax02gxnsuSlbtoeAOe625sADATwXsR1+O+TkRrB+QS/Hc5Q3ivqrDCsUkKqQ3XhzM3YDOd+czqAPqq60U1PizqkqvMBXAOnIfKbe13XxVKuSvMyOhG5G8B6VR0VdlkofbEeUTKIyEtwltt/HHZZKD1VhTpUaRogRERElD74MjoiIiIKHBsgREREFDg2QIiIiChwbIAQERFR4Mp8S94pGQM4QzUEnxaPS9aDaFIC61E4KlM9Yh0KR2WqQwDrUVii1SP2gBAREVHg2AAhIiKiwLEBQkRERIFjA4SIiIgCxwYIERH9f3v3HR9lkT9w/DvZhBYIBJBOCC2AoICcXdFTrGBFrOchYj/LqXjceZazeyenhxQrduU8FCzcCSqHeiIookiR3gQp0ntJsvP7Y5eZZ/aXDQlJZneTz/v14vX6PpnZ55klw+4wFfCOBggAAPCOBggAAPCOBggAAPCOBggAAPCOBggAAPCOBggAAPCu2LNgAABA4qkenU28dJD71X1H109NfFrmAhNffu9dTr56r02toNIdHHpAAACAdzRAAACAdzRAAACAd8wBAYCDEMrKshdNGzlp4aU/mXjloF+Z+OlrnnPy9ayxz8Qdx/zOxB3u+9HJV7htW5nKitS06epjTTzsnuEm/vPSC5x8rz1wjok//LqHiestT645H7HoAQEAAN7RAAEAAN4xBIOUFmrfxrmed0dDE7f6QDtpmTNXlulZhc0aONcLrsk0cfs3bFd62tdznHy6oKBMz0VyWnepXRb51f1PO2nnzrdd5N91HBr3HuFAPL/fCBMfuvdmJ1/7v9mllYUbN5W2qEiAglPtUMiyK+PnUyFbC3KbbnTS/tzqdRP3f9PWidx73aGVOmKH/FLp04YeEAAA4B0NEAAA4F1SDsGojGruD7q0N+Gi39YxcTizsET3a/Wee11rxgoTF677pfQFRNL48LN3nOuwBIZdzvVYkMCz+lxwlZv2zWyPBYEv29rFTxvf8X0TB4dZOk263snX6q2QiT8e9ayJ5/zGHdLp2PBGE+cNZAgmFVw7cqyJF+9t7KT9c7EdnmnXYIOJF01s6+QbObmviXOnJveKloNBDwgAAPCOBggAAPCOBggAAPAuYXNAwid2d65XnFXDxCedOstJG9nitbI9rLd7OWprjo2fsIP39V+ufGNslV1IuW3osC7ZvCDgYBScYsfup1w2JJBS7f9njurx9G0m7jD0OyctvGePiQ/9fKCJ55z0gpOvYWN2Qk12oXp1neszav1s4pF/6OekNX//GxPvDvy8hayrkLIlK3pAAACAdzRAAACAd16HYH4efJyJT+k33Un7T9NpJv5ij9ud2fl/A0zc9uHAjpNbtsd9VkELu2vlkr6ZTtqsy+zOhBc9aLtRf33utU6+Zhe4B0Ih+Rw76AbnusF1don14l8aOmmH/LOWiTf/ZoeJr+kwxcn3u3pLSl2Ob/YqE4e273HSGBSqPHYO2mri7LQacfNlKLu8tsYmuzQ8OOQSq/0tdjfLJyYf5qR91X20iY+6+RYTNxr+1QFKDF/2/Mpdl10rLSNBJUkd9IAAAADvaIAAAADvaIAAAADvKnwOyO7zjzLxGzc8ZeL6oXwnX7fhfzBxzlPuUrXWe+yy3OC2xmEpxiq7BKrtNDfpyE2/N/HMm4eZ+KMezzv5+p9il8+l/3dGcU9DgmS95f5y89+ycStZE/d1tewuyfLeWac5af2ft3N/aqdVj3uPGXttfPetdi5KjXnfFJEblUFY27k+4WI+gTYU2rkeNTcW+0llBE+5/WClOwfkrgZ2O//rb7LbvI8bfkiJ7o2Ktzfb/TpN4//3B8TfEAAA8I4GCAAA8K7Ch2BeHfqkiVuk1zRxp3/e5uRr+5hdTlayDsuD1/JJO5xy10VHm/iJJl87+XY1tsuosiq4TEicWtMWO9c7ta2BtYt5XbsM282evpvFtrB+/fRdJm42tvRLZcNj3SXk0s2GA+ouN/E4YQgmWazpXeBcp4mKkzO+2JPg09rn2ot19tRcaeL+3ne3sKfE11xpd80tnLfIfYDWkkzoAQEAAN7RAAEAAN5V+BBM6wzbiZ0fOCgsfWfpu6fKi95rly9sza9TTE5UBYWbNzvXd/x0nonfbP1x3Nf9afXpJs74/AcTJ1cnJ8oi1K61c/3OYS8HruKvkGr2RNl2KK29uuDAmZBUaiyKqQ+nFZ1PRCTU0O7U/cv5eSYecMd4J98Nde0qvwm77U7OZ9bcVaIynbOwj3O9fVhLE9ca+3Vsdu/oAQEAAN7RAAEAAN7RAAEAAN5V+ByQLk/fZOLsBXYOSNtvVjj5fI54pnXtZOKHmr1g4km7Gzj5sicsMDGLLFNDcBlbqEVTJ23JVc1NvLf5PolnZu6wwFW1uPmmv9nVxM1b/hw3X0XS29wToYO7aaLsCuu7C7GbhmrGyVmx2FUz+dVbHH8DibWX7nWuhwyZauJeNT8y8Ynf/8bJ9/IHdg5HtW12dtnIr1fHfdbqPi1M3OiCn5y0j562J8Ef3eEOE7d4LDGnKlOrAQCAdzRAAACAdxU+BNP88aK7dhK5yGxh/7ombhzoUj3x0/5OvrzNHCqWatLatTLxUxNecdJap9co4V3iD7sETR8cGKoZXMJbl4N5+fYgx4Fzfuuk1e/DEEx5WnitW2eKO4CuIiXquSi57GnuMOz96+0Q7f1HuMtrB33bz8RtH7HDMw3nzC/Rs4r7/mw0PDDsMtxNO/I+O+wy75aRJj5l+kAnX8anfg5fpQcEAAB4RwMEAAB4V+FDMMlg2ePHOtfzL7H9UnkfXW/iDjd/5+RjR8vUEzx86carb3XSVh9vdyq87dL3TTywrjtTPFFGbc0x8YztuU7a58vamrjVcPv/hvpfzqzwclVlL57yUty0j3Zlm/iRBWc7afVlYYWVCcmpYMVK53p6t5CNpaWT1lrszsk+B9dyn5pt4icutp8p9z33spPvsbaHeykPPSAAAMA7GiAAAMA7GiAAAMC7SjsHZMfFx5h47pXuWqRnt7Qxcae/2mWLhQWcQFmZpE9yl5LlTLLxcxvsibcD/zxMSuKngt3O9cqCLBO/ucHOM1q8raGTb98LdkfWjJ3x99TNnLXGxAUrVzlprWVWicqI8tWzhrtjbnC8Pjjvo34f5nwg+YW3252T33jxDBPfOOhJJ5868jAT6+mzpaLQAwIAALyjAQIAALyrVEMwweW2z/d7zsR9F/d28hVcm2niwkVLKr5gqBQunnW1c914UKBDfs0vJqy2zT1osZq41/EwAJh80kT9v5/4sPok96M5eBjdgBWnBlK2eCkPKp+6y+wnTk3l7v68vbX9jqw9veLKQA8IAADwjgYIAADwLuWGYNJb2R3llj+Z5aR9d/RTJr5oYV+bcPo6J58uWFsxhUOlVvCpu7qlcEHRBy2i8gjH7IccPBQu/+NgfSjfVTD1DtsQ97krhnQwcS35ulyfi6pjX23b/7AtvMdJq/fFMhNX5NAwPSAAAMA7GiAAAMA7GiAAAMC7lJsD0uxfdufS91q856T9dWM3E296zZ4s2rBBvnuThvYUy8K5C8q1fOlNmzjX+9rZXTDT/vd9uT4LFW/KngwTNxnGeDus7IX5B85UCqEsO6ftqEbxT2iuM8fOD4m/ry5QvPrX2DqWHzPXqWDtutjsFYIeEAAA4B0NEAAA4F1SDsGEGtR3rhcPt0tv323+XCAlw8k3uMFcGz9i4xn3uffPULbjcnm+XUpXJ809bCyeT7Z1ca7HfnG0iUef5x5s1iS018TX5pxQovsjebTJ2Gbifad1d9KqTfzWd3GQRFYPsP+2cyeU/X6F4+wQzN+bjXPSzp5ntxXIWFqynXWRYMruohvq1N5JKvzR0+GFyt3JN71JYxPfkTPRxIN/PjPmhdvFB3pAAACAdzRAAACAd0k5BLP11Dznem7PESbu8M7vTVx/duxBUSVTfZud8bv6tMA88jR3JnBGLTvLPXOKPZznmps+dPKF69i94raHazhp3avRxktlTUM1Tbypo3tgU5OJsblR2fSZf55zPb7j+ybOabjZxKHGjZx8hevs4YTpzZuZeOsxLZ18bQbNM/HLOR+YeMimQ518Na6znyMFBRxbmAqCUwke/ffrTtp1f7HfY9mvf2MTwmVf1xRcibliRAMn7cVur5n4b6vssMvu3u5OqL7w7QgAALyjAQIAALyjAQIAALxLyjkgdcZ951yfu/BKE3dab3dvK/h5dZmflfev0r/m3293dO+x3i7HfDK3t5M2pF7twNWPpX8YgIRJv9qdZ/b9Z/ZU2vEdx5r45cm5Tr6Xlh1n4nvy/m3iM2ptjfus4LyPz/se7qQVLl1SsgIjaRRu2GjiS1//vZM291E7r/FXtW82cdPPNjr5ZLWdS7T7qLYm3tbK3YIi7Vz7uhGd34xbpsvev8XEeXfPMnF41664r6lI9IAAAADvaIAAAADvknIIRufvc69n2qGLcGzmBChcvz5uWsHy+IdIIfV8v8/WuBbvujtQshiy8itYsdK5vvRD210+r+9wEw+ou9zJN7Cb/RwIF/Op9eLWNib+4rzOJmbIpXLJfWiGc92+7k0mHnb7SybuOniDk29DoR1qOayajTeH3V27+82/3MRXvXSbiXMmuDuatps+zcTJ8F1KDwgAAPCOBggAAPCOBggAAPAuKeeAAMliV7i6iQtW/ZzAkiAZdPjjbBP3+szOBxn11JNOvtbp9kiG4Em26z5t4eTLGbXIxIXrl5dXMZFkYuc1tr/NzsV4+raOsdlLrZrY+WktA7EuKnMSoQcEAAB4RwMEAAB4xxAMqqQmL9nddg9tdbOTVlDHnkjZbLLdCbO2fF3xBUNSC+4YWWusrQ+3jD0+7mvSxS7JbS7uMv2yn30KpC56QAAAgHc0QAAAgHcMwaBKCu/ZY+I2g6cmsCQAUDXRAwIAALyjAQIAALyjAQIAALyjAQIAALyjAQIAALyjAQIAALxTWif7cTUAAKCyoQcEAAB4RwMEAAB4l9AGiFLqZKXUqkSWoTwopV5RSj2c6HJUVZWoHv1FKfVGostRVVGPUFaVqA55+U4rUQNEKbVcKbVbKbVDKbU2WrjaFV24A5Rjh1Lq44N87bpEvYdoWW5RSi1TSm1TSn2rlDohEeXwLRnqkVIqJ1B/9v/RSqk7S/j6z5RSe6Kv26CUGquUalrR5Y5TlmuUUoujZZmglGqWiHL4liT1qJFSarRSarVSaqtSaopS6uhSvD6Z6tGpSqn5SqldSqnJSqlWiSiHT8lQh4ooR5X7TitND8g5WuvaItJNRLqLyJ8OpqDl4Bytde3on9MP5rUicoSI/EpE7onNoJSq0PNxoh9Sj4vIRSJSV0RGicg4pVSoIp+bRBJaj7TWPwXqT20ROUxEwiLybiluc3P0tXkiUk9EnorN4KEenSwij4rIeSJSX0SWicjoinxmkkn051FtEZkuIj0k8vf/qoj8u5RfAMlQjxqKyFgRuVci7+NbEXm7Ip+ZRBJdh5xyVMXvtFIPwWit14rIRIn80vYX4Bil1FdKqS1KqR+iH4770wYopeYppbYrpZYqpa4v7TPLm9b6ZxH5SES6iIhE/wf8O6XUIhFZFP1ZH6XUzOh7+kopdfj+1yuluiulvou+p7dFpEYpHp8rInO11jN0ZAnSayLSUEQalc+7Sw1JVI9+KyJfaK2Xl/aFWutNEmm47K9Hy5VSg5VSs0Rkp1Iq/QDvqbVS6vPoe/pEIvWgpPqIyBit9Vyt9T4ReUhEeiql2pb2faSyRNUjrfVSrfWTWus1WutCrfXzIlJNRDocxL0SWY8ulMjn0Rit9R4R+YuIdFVKdSzt+0hVSfRZdNBS9jtNa33APyKyXER6ReMWIjJbRIZGr5uLyEYROVsiDZrToteHRNN7i0hbEVEicpKI7BKRI6JpJ4vIqsBzRorIyAOUY52IrBeRj0Wka0nKX8R7aCkic0Xkoei1FpFPJPI/gJoSaQ3/IiJHi0hIRPpHX19dIh8yK0TkdhHJkEirL19EHg48a4uInBCnHFkiMiNw71tE5HuJLomuzH+SpR4F8ikRWSIiV5XiPXwmItdE44Yi8l8ReT3w/mZG61fNErynqSLyZLRe9RSR7SLyRuBZs0Tk8jjlGBJ8j9FnaRE5L9G/56pWj6J5u4nIHhGpm2L1aKiIPBPzszki0jfRv+eqUIekin+nleaN7ohWbC0ik0SkXjRt8P5/OIH8E0Wkf5x7vScitxX1yypBOY6P/mXWkkh32dr95SjFe9gS/cseKSI1A7+sUwJ5n9n/iwz8bEG0svUUkdXBv1wR+Sr4yzpAOZSI3B39BReIyAYROTLR/yB9/EmWehS4x4nR8tQuxWs+k8gHzhYR+VlE3hT7wbRcRK4O5I37nkQkJ/r7zwykvSWBL44DlKNXtO4cHv038ZxEhpIuS/TvuQrWoyyJfIH9KQXr0SgReTzmZ1OkFI3yVPyTLHVIqvh3WmmGYM7XWteJ/gV3FNvN10pE+kW7dbYopbaIyAki0lRERCl1llJqmlJqUzTtbCldF6GhtZ6itd6ttd6ltX5MIn/xJ5byPdTTWrfSWt+ktd4dSFsZiFuJyJ0x76mliDSL/vlZR//mo1aUogwDRWSAiHSWSMvzNyIyXlWRCYSSBPUooL+IvKu13lHK190arUfNtdZXaK3XB9Ji61G899RMRDZrrXcG8pe4HmmtPxWR+yXSdb88+me7iKT8DPwSSop6pJSqKSIfisi06GdSaSS8HknkCywr5mdZEqlLlV3C61BV/047mDkgn4vIKxLpAhaJvMnXo38J+/9kaq0fV0pVl8gH5BARaay1rici/5FIi6k86HK+134rReSRmPdUS2s9WkTWiEhzpVTwuTmleE43ERmvtV6otQ5rrSdE73lcmd9BCkl0PYp+cfSTyOTB8hRbj4p8TxL5nWcrpTID+UtTj0RrPUJr3V5r3Vgifz/pEuk+rzISWY+i93tPIo2+8p4H4KsezRWRrvsvovdpG/15lZDoz6LY4pTzvfZLyu+0g90H5B8icppSqquIvCEi5yilzlBKhZRSNVRkLXQLibSGqktkfKtAKXWWiJR2lq+ImOWTxyulqkWfcZdEWp1TouknK6V08XcpsRdE5Aal1NEqIlMp1VspVUci460FInKrUipDKXWhiBxVintPF5HeSqk20XufJpFZ8FXqiyPKez0KuEBENovI5OAPlVK50QlcuWW8v0gx70lrvUIiKw4eiNbpE0TknJLeOHqvLtE6lCMiz0tkDHtzOZQ71STi8yhDRN4Rkd0S6ZoPx6SnRD0SkXEi0kUp1VcpVUNE7hORWVrr+eVQ7lTCd1oCvtMOqgES7Sp8TUTu01qvlMhSwLsl8ktZKSJ3iUia1nq7iNwqIv+SyIf95SLyQbz7KqWeVUo9Gye5jkTGsTZLZMz0TBE5S2u9MZreUiLjVmWmtf5WRK4VkeHR5y0WkauiafskMnP8KhHZJCKXSGQZW/B97FBKxetGe01E/imRMeBtIvK0iFxfBf/hKNOyAAAP9ElEQVTBJ6oe7ddfIv/Lif0H3lIi3Y8/l/4duYp7T9Esl0tk4tYmiQynvBZ8vVJqrlLqiji3ryGRsf4dIvKNRD5E7i1rmVNRgurRcRJZiXS6iGxRdh+H/f/uU6IeRf/u+orIIxL5OzlaRC4ta5lTDd9piflOqzSH0SmlXpTIssSJiS4LUpdS6h4RWa+1fi7RZUHqoh6hrKrCd1qlaYAAAIDUwWF0AADAOxogAADAOxogAADAOxogAADAu2JPyTstrR8zVBPgk/CY8tqIJilQjxKjMtUj6lBiVKY6JEI9SpR49YgeEAAA4B0NEAAA4B0NEAAA4B0NEAAA4B0NEAAA4B0NEAAA4B0NEAAA4B0NEAAA4B0NEAAA4B0NEAAA4B0NEAAA4B0NEAAA4F2xh9EBQGUXPrG7c7348gwTf3r2k05abnqtMj0rpNz/8w1e183EY744xsQd7p7tlnHnzjI9F0hG9IAAAADvaIAAAADvaIAAAADvmANSztLq1HGuN13QxcR1Vu41cWjyd97KBMCV1qWjie98+Q0n7dc19wSuajppYdFF3m/q3pBz3X/CdSY+sft8E5+SPd/J17fedBM/fNEME9987AlOvkX3/MrEGR9/W2QZUHXtuNjOH1p3vq2/Y4971snXOaNaka+PnZtUqMMmLpBCJ+3s395g4vRJM6Qs6AEBAADe0QABAADeMQQTlOZ2o+4+t4eJ1x5t0xp0+8XJ99ahr5o4pNxbjti43cQfvXGciZtOLlNJAZRBYZ3qJs7N2BKTWiPu68Jiu6YHLD/dxFuuPcTJl/fjNyZeF/j5aGnm5FPdTzPxglvtcxee/ryT7+zBDU0c+t4+q3D9+rhlRepbf8OxJi4809bT17u97OTrlGGHQtLEfgm9vj3HyXfehPNN3OBb+53WcOZ2J9/SO23avJ7us5b2tWl5k4ov/4HQAwIAALyjAQIAALyrVEMwhScfYeLqS+wwScHKVU6+9NatTLzoOtslevFZXzr5HjjEziAOzhKesKu6k6/X2EH2uRvcNl2rp2aauOmur4p/Ayg11b2zc734CrsKqeGhG5y0DVtq23wnv2LinrMvcPJ1qGfrzrR3u8Z9dssX5pl4a688E29r7Q7ltXx5kYnpMk8OauoPJr78wUFOWpMrl8d93eYRtku79pivgykHVQ79/VwTd7jWfhznPXODk2/h2faz6PCRV5k4px/1KdUtfPYoE3/d+yknLTut6KGVUdvaOPkGzLarpuoOyzJxtc/dHXXz9n4jRYld21X9OztdYEDuyU5ah1vtCs6i14SVHD0gAADAOxogAADAOxogAADAu5SbAxLq1N7EtV/Y5KSNbv2iiafZTUdl9p6WTr4jatrdB3tUs+P1ywt2OfkOm2rHYbPG2fkDDf673MnXbs20uOUNx01BeVh7Yl3net5lw0r0uvzA4OWkLu/Ez3jbF3GTHr3iMBOflzXBxJ2ruf+sTjz5EhM3uNq2+QvWrhMkXoNRU53r/FHx89aWNRVWDl1QYOKOt7hj9xe072Pi8Uc9Y+KLr7vLydfwefe9IPmNPt3+PhukuTvvHjvTfnbUfsrO7aj+5VwnX8M9C4u898HO0WjxsZ3TtH6kuxNqsJ6WFT0gAADAOxogAADAu5QYgklv2sTEF479n4mvylod9zXHBFbKdq+2zEn7YGdjEw94/jIT5/zbHdJpOWtOkfcuvw4opLK7Gwa7yeP/U/pf17dNfGS/W0zceBhDMChaeM8e53rplFwT57S33fQZ58csw3U3UEUK+MOii0w8ucu7Ttrej+2ut9mf2m0cKnpoP/zDvANnKgf0gAAAAO9ogAAAAO9ogAAAAO+Scg5IcM6HiEj78XZL7eLmfezQdu3tCdMHmrjuG3WcfJnv2i2UW4i/cTWUvyYj3a2Fz/vkUhP3GjPDSXt9id3yuFb1fSb+7LAxTr7bVh9v4qHNppRLOYGyyPk4MCfkqoQVA+VE9bBHSLzd6TkTj9nhbhnR/I35JnYXw1YO9IAAAADvaIAAAADvknIIJnNMvnP99yZFn+B3y+rjnOvl/e1Jlc1+/LH8C4akE7srX+E8e/LsJ5cf5aQ1nrfUxKpahol7nXKjk6/2HLs8tlcXm7bm8r1Ovrk9XypRGR/dYHdMbf7BShOznBuomhbcaJdSNwrVMvGfvznfyddu4/feypQI9IAAAADvaIAAAADvkmYIZu9ZR5r44Zb/iEm13VXdp19h4qYXxhzAEy76QB5UTeFZ8+Om6Xy7CqbGh+4QX3BopMayFfZ+PY89qHIs2dXQ3nvFymJyAkVb3rtGoouAchSqWfSalmpLahb588qKHhAAAOAdDRAAAOAdDRAAAOBd0swBuWnov0zcNt0dB+v69ZUmbnHZYhPrcGXcGw7JJL2V3Znwq0uHxKQWPS7//NZc53r99c0CV9vKp2CoUup03FTkz/d9eEjMTxYXmQ/JZejRo038U8EuE7ces9HJV9m/4egBAQAA3tEAAQAA3iVsCCbt8I7O9ZHVvzTxtrDbLsocl2Vi1aG1iX86p76Tb08z22F136njTPz6qmOcfBs/aGHiZv9ZY+LCxctKVHZUHT/+samJs9NKthRyyBdnOdd5s4reyReIJ5SV5Vx3bWQP4bz/l+4mbvJPd6l5Ze+yT1Wh7Gznun2GHWp5YZPd0bugbsxnzHFdS3b/H+zQW3jnzoMoYWLQAwIAALyjAQIAALxL2BDM0kvcLqmcdHsgz7bwHiftrYefMHGGsj9vHjjER0Rk5j67h+XyfLv75DPtRzv52g62q2xm324PvvvNs7c7+Zr/9au45UcldpQ9PO5PJ48v0Us2FO42cfNPVTE5gaKp6tVNvHRQZyft/ZbDTXzl8tNMXLi56NUxSC5bTu/gXLdNn2TihxrNtAljZsrBeDBw4OX4ET1N3OjNWU6+ZBueoQcEAAB4RwMEAAB4RwMEAAB4l7A5IGec8W3ctKyY5Y6Pru9h4ndm2rjVO+5Ye+bctSYOnjoaOvRMJ9+8O+wSt9ln2rHVF68f5uQbmHaLiVs8xnyQquKy1yaa+Io6a4rJaZ3xjz+YuOk71BWUXlotO6dt9sDhcfMtGWXnE9SXqRVaJpSPrIXbneu+i+1S/dU77PfRhmXu1hJBmc3tPQbmuZ8x9zWcbeP7bXxkn8ucfI0vW2XiZJgPQg8IAADwjgYIAADwLmFDMEvOaeBcdxgy0MR1P3eHYBqPs7u85a2PP3RTEOfnhT8udK7zrrHx4cNuM/GiC59x8h3ya7v7oDweGO7ROm4ZkBpUuq36K+86ykk7O/OJwFX83U8fWN/NxC3fs3UlXj1EclIZ1Zzrfb8+3MTLe7sfkZ0es7slh7dstfEed+uAkkqrYetXvWJWfHf45DoTt3+FnXVTjf5+rnO9+yQbZ8u6QFwyE+u0dK4/PPJUe3H3ehNO7+FuQdHhgZtM3HbQtBI+reLQAwIAALyjAQIAALxL2BBMwZq1znXbK9bGyVmxByzVWBeKmzap81gT96lt+8zC27cXlR0p5KfBdtjlh5uGxaSW7NC5se+eaOKWS1n5kkpCne1Kkr1DdztpEzs9F/+FF9nwplV2x8lJC45w77/a7mra7sEfTBzetcvJt/BRO4w3P3eEiefucwfy8obaIR4d5si5qi72O6j6ys0mPqnxvLivq74xufockqs0AACgSqABAgAAvKMBAgAAvEvYHJBESuvS0cT9L/kkbr5eP15g4mo7fqrQMsGvjmcsKlG+rYGTmXs+f5eTlvO4XRLOwuzkl966lYkzRm4x8fvt/uPkG7U1x8TDXz3PSTvrErvzaF6mnbc28tQv4j73suPPMPHSze7yyand/27ifJ1h4r7jbnPytfs+8UsmkTx2n+9uHdD7ockmHlR/gYlvWHWik6/V8/NNnAwziegBAQAA3tEAAQAA3lWJIZhdFxztXN/7xEsmPrXmXhNP3+t2pGf8JbAvnV5RMYVDUpuzr46JWz7kLrVl2CW1hOtmmnhMO7vE/p5fejj5ZveyB4I13+j+zmf91cZptRqb+K0BZzj5nhs01MSj20yU+OyS758K7HLgdncw5FJVBHfDjd1RN5RlD6qb9zc7deB/Zz/p5GsasgcZ3r7Gft8tvbujky9944yyFbac0QMCAAC8owECAAC8S+khmFBD90C7Vf3t7oY559hDoz5qN9TJV1PZw6em7LVtsEEP3ujky54yVQBUbldlu8MsF103yMStRsxx0gq3bTNxQQ/7ebP7JHdnykwV3MnUPewunqYhm2/RMHfYuOPwjbYMCxYLUkuoQX3nem0/W3e2t7Y/L2yy18n3xLHvmPj8zM9MvDmsnHx5b9tD5jo8Yg9fTbYhl1j0gAAAAO9ogAAAAO9ogAAAAO+8zgEJ5bW1Fzr+Isaf+jZxrvPr2LyFuXaZ0qNHjnXy9c0selfTJQXunm9nfHKdiTv9zZ4imL2QOR+V2bbLjzHxC7lDAik1477m1fXHB644BTmVpW22v79T59hjbSd1ecfJN/NmezryBwOynbSdYTtPo0eNwK6oGbHzPOz1BzvtPUbcfLGTa0tbu/vpl3+2c9UWXDjSydchzY7xt/8dc0CSkUp3v06XPnikib/77VNO2jNb7LYO19SdbeKsNPck7oLAfqV3rj3WxHPuPNzJ1+4zu2w7GXY4LSl6QAAAgHc0QAAAgHdeh2BGfPKqieuHQk5aSOyyouoqw0kLHgiWHxi62RUzitN3cV8TLxlvh3tajl/v5MubZw8RS6XuKpTNnmzb3m4aij/skq9trZg3rLOJ6wq7U6ayghUrTVyrjx0iOeuEa518D4x60cTnZm6W+Ow9vt8XdlKuGmUPk8sdMc/EGZu/dfIdEoh75tvXTH1guJNv1nl2eKb79ttN3OaPDBsniwXDj3CuF58zInDlDtHdkW0Pw9yt7efSH9e5u/J++YRdjp012n7+hOS7shQ1adADAgAAvKMBAgAAvKMBAgAAvPM6B+SGVieYOHxCNydNp9u20N767hyQ2ou3mjhtk11KV7Dq55gnrDNRs0DMPA+IiGSes7ZE+XrP62fium8y76My0vn7TBya7I6nP9jmiNjspdZS7PbuJf38afCinc/R58UecfO1EeZ9JKNDprnzGo9qfqmJj23inqb+1Shbxxq/9oOJw7t2OfmyKvm8M3pAAACAdzRAAACAdwk7DTfty5lx02rFXIfjxECpvGgXPS4cYrvgY3exDD0SPGXZ7ToFgKJkvxIzNPaKDRe5KXJIYBitKn+n0QMCAAC8owECAAC8S9gQDOBb5jtfm/jeW8818dttJzj5qi+zO+cWVHyxAKBKogcEAAB4RwMEAAB4RwMEAAB4xxwQVEk7e9p5Hn0kdtfJlQIAqFj0gAAAAO9ogAAAAO+U1jrRZQAAAFUMPSAAAMA7GiAAAMA7GiAAAMA7GiAAAMA7GiAAAMA7GiAAAMC7/wNW2TuEYhBj2QAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 720x720 with 20 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 20 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.naive_bayes import GaussianNB\n",
    "\n",
    "gnb = GaussianNB()\n",
    "y_pred = gnb.fit(x_train, y_train).predict(x_test[:100])\n",
    "\n",
    "eval = Evaluate(y_test[:100], y_pred, x_test[:100])\n",
    "eval.get_evaluation_report()\n",
    "\n",
    "y_pred = gnb.fit(x_train_r, y_train).predict(x_test_r[:100])\n",
    "\n",
    "eval = Evaluate(y_test[:100], y_pred, x_test_r[:100])\n",
    "eval.get_evaluation_report()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "78n0_v6_cmJ6"
   },
   "source": [
    "##Σύγκριση υλοποιήσεων"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "Jtd-uj_UclMh",
    "outputId": "b213509d-8d68-4fb8-a79f-fceb72ed9a25"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 1min 13s, sys: 12 ms, total: 1min 13s\n",
      "Wall time: 1min 14s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "nb = NaiveBayes(x_train_r,y_train)\n",
    "preds = [ nb.predict(i) for i in x_test_r[:100]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "hfcp3c7ecfq9",
    "outputId": "d8b6bf2f-b9bb-4016-d8ba-5aee715656fb"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 323 ms, sys: 4.99 ms, total: 328 ms\n",
      "Wall time: 335 ms\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "gnb = GaussianNB()\n",
    "y_pred = gnb.fit(x_train, y_train).predict(x_test[:100])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "FBPnCVCzd4WN"
   },
   "source": [
    "##Σχολιασμός"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "jSVljcb5dz22"
   },
   "source": [
    "Κατά τον υπολογισμό της πιθανότητας \"pxC\", χάρη στη στρογγυλοποιήση των \"x_train & x_test\" sets, γνωρίζαμε για την κάθε εικόνα της κλάσης πόσες φορές ήταν \"1\" το συγκεκριμένο χαρακτηριστικό και διαιρώντας με το πλήθος των στοιχείων της κλάσης λαμβάναμε τη ζητούμενη πιθανότητα.\n",
    "Ο δικός μας NaiveBayes ταξινομητής, εμφανίζει υψολότερες επιδόσεις και εγκυρότητα από τον έτοιμο. \n",
    "Ο τελευταίος, για τα στρογγυλοποιημένα σύνολα παρουσιάζει χαμηλότερες επιδόσεις.\n",
    "Ωστόσο, ο έτοιμος ταξινομητής είναι αισθητά ταχύτερος από αυτόν που υλοποιήσαμε εμείς."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "-aWBz2IzeIGx"
   },
   "source": [
    "##4ο Μέρος: Multi-Layer Perceptron"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "VQvd13RddogR"
   },
   "outputs": [],
   "source": [
    "class Layer:\n",
    "    def __init__(self):\n",
    "        \"\"\"Here we can initialize layer parameters (if any) and auxiliary stuff.\"\"\"\n",
    "        # A dummy layer does nothing\n",
    "        pass\n",
    "    \n",
    "    def forward(self, input):\n",
    "        \"\"\"\n",
    "        Takes input data of shape [batch, input_units], returns output data [batch, output_units]\n",
    "        \"\"\"\n",
    "        # A dummy layer just returns whatever it gets as input.\n",
    "        return input\n",
    "\n",
    "    def backward(self, input, grad_output):\n",
    "        # The gradient of a dummy layer is precisely grad_output, but we'll write it more explicitly\n",
    "        #changes that should be done at weights\n",
    "        #helps refresh of weights correct train\n",
    "        num_units = input.shape[1]\n",
    "        \n",
    "        d_layer_d_input = np.eye(num_units)\n",
    "        \n",
    "        return np.dot(grad_output, d_layer_d_input) # chain rule"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "zUk3YNCGeSIc"
   },
   "outputs": [],
   "source": [
    "class ReLU(Layer):\n",
    "    def __init__(self):\n",
    "        \"\"\"ReLU layer simply applies elementwise rectified linear unit to all inputs\"\"\"\n",
    "        pass\n",
    "    \n",
    "    def forward(self, input):\n",
    "        \"\"\"Apply elementwise ReLU to [batch, input_units] matrix\"\"\"\n",
    "        relu_forward = np.maximum(0,input) #returns an array with elements >=0\n",
    "        return relu_forward\n",
    "    \n",
    "    def backward(self, input, grad_output):\n",
    "        \"\"\"Compute gradient of loss w.r.t. ReLU input\"\"\"\n",
    "        relu_grad = input > 0 #true if input>0\n",
    "        return grad_output*relu_grad #int*True = int\n",
    "                                     #int*False = 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "4zcoJ1h5eWkA"
   },
   "outputs": [],
   "source": [
    "class Dense(Layer):\n",
    "    def __init__(self, input_units, output_units, learning_rate=0.1):\n",
    "        \"\"\"\n",
    "        Weight Initialization\n",
    "        \"\"\"\n",
    "        self.input_units = input_units\n",
    "        self.output_units = output_units\n",
    "        \n",
    "        self.learning_rate = learning_rate\n",
    "        self.weights = np.random.normal(loc=0.0, \n",
    "                                        scale = np.sqrt(2/(input_units+output_units)), \n",
    "                                        size = (input_units,output_units))\n",
    "        self.biases = np.zeros(output_units)\n",
    "        \n",
    "    def forward(self,input):\n",
    "        \"\"\"\n",
    "        Perform an affine transformation:\n",
    "        f(x) = <W*x> + b\n",
    "        \n",
    "        input shape: [number of inputs, input units]\n",
    "        output shape: [number of inputs, output units]\n",
    "        \"\"\"\n",
    "        \n",
    "        output = np.dot(input,self.weights) + self.biases\n",
    "\n",
    "        return output\n",
    "    \n",
    "    def backward(self,input,grad_output):\n",
    "        # compute d f / d x = d f / d dense * d dense / d x\n",
    "        # where d dense/ d x = weights transposed\n",
    "        grad_input = np.dot(grad_output, self.weights.T)\n",
    "        \n",
    "        #The T attribute is the transpose of the array,\n",
    "        # compute gradient w.r.t. weights and biases\n",
    "        grad_weights = np.dot(input.T, grad_output)\n",
    "        grad_biases = grad_output.mean(axis=0)*input.shape[0]\n",
    "        \n",
    "        assert grad_weights.shape == self.weights.shape and grad_biases.shape == self.biases.shape\n",
    "        \n",
    "        # Here we perform a stochastic gradient descent step. \n",
    "        self.weights = self.weights - self.learning_rate * grad_weights\n",
    "        self.biases = self.biases - self.learning_rate * grad_biases\n",
    "        \n",
    "        return grad_input"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "hJKJC7rCelYO"
   },
   "outputs": [],
   "source": [
    "def softmax_crossentropy_with_logits(logits,reference_answers):\n",
    "    \"\"\"Compute crossentropy from logits[batch,n_classes] and ids of correct answers\"\"\"\n",
    "    logits_for_answers = logits[np.arange(len(logits)),reference_answers]\n",
    "    xentropy = - logits_for_answers + np.log(np.sum(np.exp(logits),axis=-1))\n",
    "    return xentropy\n",
    "\n",
    "def grad_softmax_crossentropy_with_logits(logits,reference_answers):\n",
    "    \"\"\"Compute crossentropy gradient from logits[batch,n_classes] and ids of correct answers\"\"\"\n",
    "    ones_for_answers = np.zeros_like(logits)\n",
    "    ones_for_answers[np.arange(len(logits)),reference_answers] = 1\n",
    "    \n",
    "    softmax = np.exp(logits) / np.exp(logits).sum(axis=-1,keepdims=True)\n",
    "    \n",
    "    return (- ones_for_answers + softmax) / logits.shape[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "6pvwkzPAewN6"
   },
   "outputs": [],
   "source": [
    "class MLP:\n",
    "    def __init__(self, shapes, input_dim):\n",
    "        '''\n",
    "        Κατασκευή του MLP. Τα επίπεδα και το μέγεθός τους ορίζεται από την μεταβλήτη shape, ενώ\n",
    "        το μέγεθος της εισόδου ορίζεται από την μεταβλήτη input_dim\n",
    "        '''\n",
    "        self.shapes = shapes\n",
    "        self.network = [Dense(input_dim, shapes[0])]\n",
    "        self.network.append(ReLU())\n",
    "        for i in range(1, len(self.shapes) - 1):\n",
    "            self.network.append(Dense(shapes[i-1],shapes[i]))\n",
    "            self.network.append(ReLU())\n",
    "        self.network.append(Dense(shapes[i],shapes[-1]))\n",
    "\n",
    "    def forward(self, X):\n",
    "        \"\"\"\n",
    "        Αγόριθμος διφάνειας 33\n",
    "        \"\"\"\n",
    "        activations = []\n",
    "        input = X\n",
    "        \n",
    "        for l in self.network:\n",
    "            activations.append(l.forward(input))\n",
    "            \n",
    "            input = activations[-1]\n",
    "            \n",
    "            \n",
    "            pass\n",
    "\n",
    "                   \n",
    "        assert len(activations) == len(self.network)\n",
    "        return activations\n",
    "\n",
    "    def predict(self,X):\n",
    "        \"\"\"\n",
    "        Προβλέπει την έξοδο του δικτύου για ένα ή περισσότερα στιγμιότυπα εισόδου\n",
    "        \"\"\"\n",
    "        logits = self.forward(X)[-1]\n",
    "        return logits.argmax(axis=-1)\n",
    "\n",
    "    def fit(self, X, y):\n",
    "        \n",
    "        layer_activations = self.forward(X)\n",
    "        layer_inputs = [X]+layer_activations  #layer_input[i] is an input for network[i]\n",
    "        logits = layer_activations[-1]\n",
    "        \n",
    "        # Compute the loss and the initial gradient\n",
    "        loss = softmax_crossentropy_with_logits(logits,y)\n",
    "        loss_grad = grad_softmax_crossentropy_with_logits(logits,y)\n",
    "        \n",
    "        # Propagate gradients through the network\n",
    "        # Reverse propogation as this is backprop\n",
    "        for layer_index in range(len(self.network))[::-1]:\n",
    "            layer = self.network[layer_index]\n",
    "            loss_grad = layer.backward(layer_inputs[layer_index],loss_grad) #grad w.r.t. input, also weight updates\n",
    "        return np.mean(loss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "id": "eEr2XDmAfOEj",
    "outputId": "a94cff8d-ae9c-404e-c3aa-7d7a611f6989"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 24\n",
      "Train accuracy: 0.75\n",
      "Val accuracy: 0.7517\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "precision : 0.7703072400261106\n",
      "recall : 0.7446128565208234\n",
      "f1 : 0.7343594729722056\n",
      "accuracy : 0.7517\n",
      "Confusion matrix: \n",
      "[[ 933    0    7   10    0    4   19    1    6    0]\n",
      " [   0 1051    1    5    0    1    2    0   75    0]\n",
      " [  30   31  676  124   21    0   46   28   64   12]\n",
      " [  18   10   22  848    0   18    8   23   55    8]\n",
      " [   4   29    6    0  565    0   60   15   16  287]\n",
      " [  79   50    6  247   19  254   47   35   90   65]\n",
      " [  53   25   11    7    9    1  843    1    8    0]\n",
      " [   9   31   32    2   10    1    1  903   18   21]\n",
      " [  22   26   21  104    4    1   32   52  670   42]\n",
      " [  20   26   15   13   46    4    3   90   18  774]]\n",
      "Some images with wrong predictions\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 20 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from IPython.display import clear_output\n",
    "import numpy as np\n",
    "\n",
    "network = MLP([100, 200, 100, 10], 28*28)\n",
    "train_log = []\n",
    "val_log = []\n",
    "\n",
    "#Αυτά πρέπει να τρέχουν αφού έχετε φτιάξει σωστά τον κώδικα της MLP\n",
    "for epoch in range(25):\n",
    "     network.fit(x_train,y_train)\n",
    "    \n",
    "     train_log.append(np.mean(network.predict(x_train)==y_train))\n",
    "     val_log.append(np.mean(network.predict(x_test)==y_test))\n",
    "    \n",
    "     clear_output()\n",
    "     print(\"Epoch\",epoch)\n",
    "     print(\"Train accuracy:\",train_log[-1])\n",
    "     print(\"Val accuracy:\",val_log[-1])\n",
    "     plt.plot(train_log,label='train accuracy')\n",
    "     plt.plot(val_log,label='val accuracy')\n",
    "     plt.legend(loc='best')\n",
    "     plt.grid()\n",
    "     plt.show()\n",
    "\n",
    "y_pred = network.predict(x_test)\n",
    "\n",
    "eval = Evaluate(y_test, y_pred, x_test)\n",
    "eval.get_evaluation_report()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Dnwo5tyJfuA-"
   },
   "source": [
    "##Έτοιμο Multi-Layer Perceptron"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 914
    },
    "id": "Eh90-KHofWUX",
    "outputId": "35054469-966f-4ed4-9579-2aeaa2bb347f"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.6/dist-packages/sklearn/neural_network/_multilayer_perceptron.py:571: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (25) reached and the optimization hasn't converged yet.\n",
      "  % self.max_iter, ConvergenceWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "precision : 0.9674121749328176\n",
      "recall : 0.9671022741195809\n",
      "f1 : 0.9671582122271513\n",
      "accuracy : 0.9675\n",
      "Confusion matrix: \n",
      "[[ 965    1    1    1    2    4    1    2    2    1]\n",
      " [   0 1122    3    1    0    1    2    4    2    0]\n",
      " [   4    6  994   10    5    1    2    8    2    0]\n",
      " [   0    1    1  983    0   11    0    8    5    1]\n",
      " [   3    0    1    0  952    0    4    2    3   17]\n",
      " [   4    0    2   12    2  858    6    1    4    3]\n",
      " [   7    3    2    1    7    7  930    1    0    0]\n",
      " [   0    4   10    2    1    0    0 1004    1    6]\n",
      " [   6    2    7   21    5    9    4   12  905    3]\n",
      " [   4    4    0    8    8    7    1    9    6  962]]\n",
      "Some images with wrong predictions\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAiAAAAI+CAYAAACFVbH3AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOzdd5gURfoH8O+7gSW7ZIlLXKIHiAoqKHcKCIjxFAUF8cyinvn0p2C+OxMmEAyoCOoJnAkOMYIJVBBEEQkSJIiAgIQl7W79/uimumvcWWZmZ6snfD/Psw9vb1f3VO/WzhQVRSkFIiIiIpsygs4AERERpR9WQIiIiMg6VkCIiIjIOlZAiIiIyDpWQIiIiMg6VkCIiIjIukArICLSU0TWBZmHeBCRu0RkYtD5SFcsRxQPLEdUVixD0YmoAiIiq0Vkj4jsEpGNIvKiiFQt78yVkI+PRWSziOwQkW9F5PQorp0lInvdZ9giIv8Vkfrlmd8w+Rjs5uHgV4GIKBHpYjsvtiVCORKRJiE//13uz//GCK9PlHJUQUSmuD9TJSI9bechKAlSjuqKyKsiskFEfheRz0WkaxTXJ0o5Ssv3owQpQynxXuTm5RIRWeHm5V0RaRDJddG0gAxQSlUF0AlAZwC3xZLRMroOQH2lVHUAlwGYGOUPfLj7DPkAcgGMCk0gIllxyWkYSqlJSqmqB78AXAVgJYBvyvN1E0ig5Ugp9XPIz/8IAMUApkZxm8DLkeszABcA2GjhtRJN0O9HVQF8DaALgJoAXgIwPcoPscDLUZq/H/G9KA7c//w8AOB0OH8LqwC8Gsm1UXfBKKU2ApgJ55d2MAPdROQLEdnutkz09J0bJiJLRGSniKwUkcujfU3fay9SShUePASQDaBxDPfZCueX3MHN42oRuVVEFgHYLSJZh3imZiIy232m9wHUjvWZAAwFMEGl2ZK0QZajEEMAfKKUWh3thUGWI6XUfqXUY0qpzwAURZv3VBFUOVJKrVRKPaqU+kUpVaSUegZABQCtY7gX348CxPeiMpehUwFMVkotVkrtB3AvgBNEpEUkGT/kF4DVAE5240YAvgPwuHvcEMBvAPrBqdD0co/ruOf7A2gBQACcCKAAwJHuuZ4A1vleZwyAMYfIyzQAe+FUQN4FkBHhM8wCcIkb1wbwEYCXfc+3EE5lplIEzzQHwKMAcgCcAGAngIm+11oEYFAEecqD8+HRLJJnSPavRCpHbjoB8BOAi6J4hkQsR+sA9Az695uu5chN2wnO+9JhSVyO0ub9KNHKEJL4vQjAw/5ndF9LATj9kM8QxS9rl5spBeBDALnuuVsPPrQv/UwAQ8Pc600A15X0y4riB58NoC+AG6L8ZRUA2A5gPYBJvh/+agAX+9KGfSYATQAUAqjiO/eK/5cVRZ7uBDCrvP7IEu0rActRDzc/VZO8HKVjBSSRylF1OB9gtyV5OUqb96MELENJ+14E4GQAWwD8CU5lZxycrqTzD3VtNF0wZyilqrk/4DbwmmjyAJzjNutsF5HtALoDqA8AItJXROaKyFb3XD+UrYkQSqkDSqkZAHqLyGlRXHqtUipXKdVQKTVYKbXZd26tLy7tmRoA2KaU2u1LvybGRxkCp+84nSRMOYLzxzdVKbUryusSrRylo4QoRyJSCcA7AOYqpf4Z5eWJVo7S7f0oIcqQK2nfi5RSHwAYCacLaLX7tRPOf4xKFfXgFKXUbBF5EU6zyxlwHvJlpdSloWlFJMfN1BAAbymlDojIm3Cam+IhC05TWDwoX1zaM+UBqCEiVXy/sCYh1x+SiBwP5xc/Jcb8JrWgy5H7wXEOgDNjvUcYVstRuguyHLn3exPOG228xgEcxPcjS/heVPYypJQaDWC0e798AHcA+P5Q18W6DshjAHqJSEcAEwEMEJE+IpIpIhXFmQvdCM6grBwAmwEUikhfAL1jeUERaePWPCuJSLaIXACnr2q2e76pOFOYmsb4TH5hn0kptQbAPAB3izMVsjuAATG8xsEa78445DdZWS9HPmcC2AbgY/83k6kciUiOiFR0Dyu4949X5T6ZBPF+lA3nw3oPnKb54pDzSVOOXOn+fsT3ohjLkHuvDuJoAuAZOONpth3q2pgqIG4zzwQAI5RSa+FMv7kdzi9lLYCb4QwO3QngWgCvw/kBDwLwdikPMlZExoY7DeAuAJvc17kOwECl1MHpYo3hNButj+WZ/Ep7JjfJIABdAWyF0/Q0IeQ5FovI4HD3dz80zkV6NXf+QUDl6KChcP5HEFrLT5pyBGApnA/AhnD6c/fAaWpNKwGVo+PgjP7vDWC7eOs49HDPJ0054vsR34tQtjJUEc6YkV0AvoIzoPXOSPIlf3zm5CQidwDYrJQaF3ReKHmxHFE8sBxRWaVDGUqZCggRERElD25GR0RERNaxAkJERETWsQJCRERE1rECQkRERNaVuhBZr4xzOEI1AO8XT06ptRxYjoKRSuWIZSgYqVSGAJajoIQrR2wBISIiIutYASEiIiLrWAEhIiIi61gBISIiIutYASEiIiLrWAEhIiIi61gBISIiIutYASEiIiLrWAEhIiIi61gBISIiIutYASEiIiLrWAEhIiIi61gBISIiIutYASEiIiLrWAEhIiIi61gBISIiIuuygs4AUXnJatTQOD5m+iodX13zax3fsK6vkW7OJ+113HLkAh0X790b7ywSEaUttoAQERGRdayAEBERkXUJ2QUjndsbxxtGFut4wdGTIrpHpnh1qyJVbJx7c3eujm9fcIaO1coqRrpmt82J6LUoMRU2qGkc3177bd9RRR093+RjI13GBbN03DL3ch3nX/41KL1l1qur471HNNbxqvMl7DUrTnlGx8VQEb3OJ3srGMePDDhbx0U/LIvoHpS4Mtvl6/jHK2voOKPWPiNd9rLKOm72+hYd7zu8mpHurKfe0/FTrw/QcZO7vih7ZssRW0CIiIjIOlZAiIiIyLrE7IJZ8bNxnJXRQMdf7wvfhPnClu46/mRNSx3/75injXSnVfHucVr3F7wT3Y1k6NFxoI5zT1utY1VYGDYPlDjkW7OpesAZF+n451vDl6Pvjp2g4y7tvJkzBdWrG+mKduwoYw4p0W0afpxxfPGV03V8We700OQlKvb9P++BLUcY5w7LKtDxlbnLddy9ojnj6r7GXtmr8ENEL0sJbMS0V3XcOccbIpAR0iZQfKJv+MGQjBKvCb1uVKs9cctneWMLCBEREVnHCggRERFZxwoIERERWZeQY0CKd+40juuctlTHI9GllCu9vq8m+E7HV3W6zEj1W8fDdLy5xwEdL+s7zkj3ecfXdXzErcN13Pj+xJ7aRA61z5zShq+9MtH4r6VcuN4LX20+U8c9T7naSFb19bllyR4lqE1Xe+M+3rrlQeNcvcwcHX+/3xtHdN4c8z2mwvfe9MmGs3brOGvpWiOdZGfr+K9ffa/jOr7XAYDV53mvlT8TlOSOzvGmbbee5ZWdiosqhb2m4b+9z51lzx1lnFvR15vq3fS58FPCEw1bQIiIiMg6VkCIiIjIuoTsgom34oXmvLUaC7245iTvR9D6ObMZdXmvZ3X80LDxOn7i/jZxziElklYfXqLjpSd5ZWDbubuMdFVfB6WAzJDp1W0HLdHxyA3mRoXLHmmn4+rveu8rzXcuRCSKQo4zqnkrWhYjvLwp/L9iKgm3Iq6/myUa/uUpKmz0hjCElrdEw1JNRERE1rECQkRERNaxAkJERETWpcUYkNL4l1Wv+p059Q29vLBWxm5Qemjzb+93/WtPb2r3k51fNdI90vpMHRctXVH+GaNyEbqk/m/Hh09bFV/quLQxG5FaPqKDjutlfqTjl3bkGemqLPTmhnMjiOR3xdoTdVyv1u9RX39+l6+M4/l7m3oH63+NNVvWsQWEiIiIrGMFhIiIiKxL+y4YvwPdzBVYM8Wrn10wZ5iOW2CBtTyRfUWLvZV3T5x9jY5//MtzRrqHqlW0lidKDQdONldy/mjgQ74jrwt41GtnGOmarOfqy6lkyWNe19vHjzyp49OOudhM+NV3iESXiqt1PL1hN+9Egu/YzRYQIiIiso4VECIiIrIu7btgNl9xrI7f7/qQce6OTd65Vn/zVkeMx+h3Sg45S32bQ/0luHxQEsvI1OHa3hWMU/7N7ZYe8NatbPRhQfnniwJTaZO3CWqGrx1g/Z+rGekampNdwvJvbrf/cO8emUtKSp042AJCRERE1rECQkRERNaxAkJERETWpcUYkIzKlY3jpWPa6virkx7W8Ud7Ghrp3n65h47r7+U0OCKK3oabuur4+8GPh0036OkbdNzwM77fpIviCEcVZjVupOMuVT4JuUfJu+smOraAEBERkXWsgBAREZF1KdUFk9kuX8dLrs3V8T//PNlId3bVT3U8Z18VHd8/erCRrv7jbAYlorKp23td2HNv7a6t48ZPLtQxp/qntqyP5uv4+BHX6rjhdztLSg4A2Nuqno5Pq7LNOLdgv1diKmz07lGExMYWECIiIrKOFRAiIiKyLqm7YDLr1TWOz57qjQy+qPoGHZc2Qvime67U8eEvssuFTNMvfVDHGTBnU0EERCUp6nmkjt9t+6yOQ7tW7n7R6/ZtVMD3n3RU6/k5UV8TOnOmcwWvLYEroRIRERGVghUQIiIiso4VECIiIrIuqceAFP26yTh+5oEzdfxYPa9//skrxxrpelQs1PHvrbzv14hz/ij5FRtxyFgilZyrD1L8ZeYeZhwX3rFFx9ni7YZ77sreRrpGD3DcB0UvI6TtIAPJOR6NLSBERERkHSsgREREZF1Sd8GEyn3Zm85Uq1ZNHf/0N3O67vEV1+s4c29yNl1R+VHHd9JxtYzPdTxpZ30jXebWXTouBKWz1Ve3N44XtPU2nfulcJ+Ol7/a2khXF+yCoej9cQO75GxLSM5cExERUVJjBYSIiIisS6kuGL/Kb3p1qyHV1xvnWk+52osf+kbH3ACKAOC+id7KlTUyKur4we/NGQyNV35vLU+UeKSz1+3yzmUPhpzN0VH3/96k45aj2eVCZVfaLJj1J3rvWU0+tpalmLAFhIiIiKxjBYSIiIisYwWEiIiIrEvqMSAZlc3dSVfe1lHHPzQfreP395jpWr+wQ8fFe/eWU+4oWXWp4K1cWdpOypSGfDsgr7nd+/9bo6ycklIDAPKmHSjXLFH6KW0a7oFWe+xmpgzYAkJERETWsQJCRERE1iV1F8zy5/KN4yUnPuU78ppK/3n9UCNdxYVflWe2KEXVeanyoRNRSvvtkm46XnjcE2HT/emzv+m46QfzyzVPlH5Km4Z7XaePdDwtwbdYZQsIERERWccKCBEREVmXFF0wkuONMF81wdvMaWn38aEpddT+eW+107x35oAoUpni1cuLVZGOt7Uy/1zMrekoHfzeMrJ0ze/dr+NIV1jefMWxOq4zlu9ZFF5ps2CKkDwbrLIFhIiIiKxjBYSIiIisYwWEiIiIrAtsDMi2i441jqtd4O1Ym3Xyz8a5Nf/oouPvuz+p49AVTv3TbTnug2J12vJTdDy15XQdN3r7FyNdESjd1O+0scTvt3v9GuO41Y/zdOxfsXnjRZ2MdBdf7ZWv/4zkftwUmdBpuNnird48bfhfdJyJb5DI2AJCRERE1rECQkRERNYF1gVTeNZW4/jdtm/oOH/slca5D/o+qOOP93gruz164XlGuopzuMIpld345lN9RxUDywclnvtavVHi91UNc8O54q4ddNxr3Kc6Pqvag0a6vpNu1nHzt7wVU7kFIoXKmbdcxz2/O8c4N+uIybazExdsASEiIiLrWAEhIiIi61gBISIiIusCGwMy9oiJId/x6kIrBow1zpy+3OvvKj57n47lt2/LJW+U3o6ddJOOf7jwqVJSUroZ9vkwHf/wl2d0vKSX+Z6FXl7onzLZdvZwI1mL273lAjjug0pTtGOHjn+f1cE4t6C1N4V7Swdv3Fq9j8s/X2XBFhAiIiKyjhUQIiIisi6wLpjBr19rHP9wgdfU3eoNcxpumzHbdFz0m7kaJVG8Nf+H1yx+qm8VXmCV/cxQQmlz+2YdP/a/djr+e80fjHT3bj5SxzOf6K7j/CmLjXRcTZdiUWtxoXHcuYLXlrCvpu3cxI4tIERERGQdKyBERERknSgVfux1r4xzODA7AO8XT5ag8xBPLEfBSKVyxDIUjFQqQwDLUVDClSO2gBAREZF1rIAQERGRdayAEBERkXWsgBAREZF1rIAQERGRdayAEBERkXWlTsMlIiIiKg9sASEiIiLrWAEhIiIi6wKtgIhITxFZF2Qe4kFE7hKRiUHnI12xHFE8sBxRWbEMRSeiCoiIrBaRPSKyS0Q2isiLIlK1vDMXkocm7uv7v5SI3Bjh9bNEZK973RYR+a+I1C/vfJeQjwoiMsX9mSoR6Wk7D0FJhHLk5uM4EflKRHaKyCIR6X7oq/S1iVKOBof8LRS45anLoa9ObixH8SMi7URknohsc78+EJF2h74yuSVQGfpYRDaLyA4R+VZETo/i2kQpQzF/pkXTAjJAKVUVQCcAnQHcFmU+y0Qp9bNSqurBLwBHACgGMDWK2wx3r80HkAtgVGgCEcmKS4ZL9xmACwBstPBaiSbQciQiNQG8A+AhOGXgQQDviEiNKG4TeDlSSk0K+Xu4CsBKAN+U5+smEJaj+NgA4K8AagKoDeBtAK+V82smikDLkOs6APWVUtUBXAZgYpSViEQoQ0CMn2lRd8EopTYCmAnnlwYAEJFuIvKFiGx3a3E9feeGicgS938JK0Xk8mhfM4whAD5RSq2O9kKl1FY4FZcObh5Xi8itIrIIwG4RyTrEMzUTkdnuM70P5w830tfer5R6TCn1GYCiaPOeKgIsR8cB2KiUmqyUKlJKTQSwGcBZMTxDYOWoBEMBTFBpNq2N5ajM70fblVKr3XIjcN6TWkb7DMksyM80pdQipVThwUMA2QAax3CfpPxMi7oCIiKNAPQFsMI9bghgOoD74NSibwIwVUTquJdsAnAqgOoAhgEYJSJHhrn3GBEZE0EeBE4F5KVo8+9eXxvA2QAW+L59PoD+cGqR9Q7xTK8AmA/nl3QvnDd///0XicigWPKWLgIuR6E7MwrcP9wonyEhypGI5AE4AcCEaJ8h2bEcAYhDORKR7QD2AngSwAPRPkMyC/ozTUSmicheAF8CmAVgXgzPEHgZiolS6pBfAFYD2AVgJ5xa2ocAct1ztwJ4OST9TABDw9zrTQDXuXFPAOsiyUPIPXq4+akaxTWzABQA2A5gPYBJAOr4nu9iX9qwzwSgCYBCAFV8514BMDGG51gHoGe01yXrVyKUIwC13DJwPpz/bQyF05U3LonL0Z0AZgX9+2U5SvpyVAVOV17/oH/H6VCGQu6RDacSdEMU1yRiGYrqMy2aFpAzlFLV3B9wG3hNNHkAznGbdba7NenuAOoDgIj0FZG5IrLVPdcPZWtqhvtDm6qU2hXlddcqpXKVUg2VUoOVUpt959b64tKeqQGAbUqp3b70a2J5iDQVaDlSSv0G4HQANwD4FcApAD6A84cTqUQrRzG3BiYxlqM4lyP3HmMBTBCRurHcI8kkzGeaUuqAUmoGgN4icloUlyZUGYpW1INTlFKzReRFAA8DOAPOQ76slLo0NK2I5MDplxoC4C2l1AEReRN/bLqMmIhUAnAOgDNjvUcY/r7z0p4pD0ANEani+4U1CbmeDiHIcqSUmg3gaPfeWXAGbz4Sy71Kur0vLvdyJCLHw3kDmRJjfpMay1Hc348yAFQG0BBOV0PKC/ozLUQWgBZxulfCf6bFug7IYwB6iUhHABMBDBCRPiKSKSIVxZkL3QhABQA5cAZnFYpIXwC9y5jnMwFsA/Cx/5si0lScKUBNy3h/oJRnUkqtgdNHd7c404+6AxgQzc1FJEdEKrqHFdz7x6sAJ5NAypGIdBaRbBGpDudNZ61SaqZ7LmnKketga+DOOOQ3WbEcxViORKSX+xyZ7nM8Cuf9dUkc8p1MrJchEWnjtqZUcsvRBXDGcs12zydFGXLzGtNnWkwVELeZZwKAEUqptXCaIm+H80tZC+BmABnum+K1AF6HU6gHwZnmFe4hxorI2EO8/FA4NbnQ2lljOM1G66N/IlNpz+QmGQSgK4CtAEYiZPCfiCwWkcGlvMRSAHvg/C9jphvnlTXfySbAcnQLgC3ua9SH2ZqWNOXI/YM/F+nX/WJgOSpTOcoF8CqA3wH8BOd/36copfaWNd/JJKAyJADugtPStBnOlNyBSqmDU+mTpQwBMX6mpcxmdCJyB4DNSqlxQeeFkhfLEcUDyxGVVTqUoZSpgBAREVHy4GZ0REREZB0rIERERGQdKyBERERkHSsgREREZF2pC5H1yjiHI1QD8H7x5JRaE4TlKBipVI5YhoKRSmUIYDkKSrhyxBYQIiIiso4VECIiIrKOFRAiIiKyjhUQIiIiso4VECIiIrKOFRAiIiKyjhUQIiIiso4VECIiIrKOFRAiIiKyjhUQIiIiso4VECIiIrKOFRAiIiKyjhUQIiIiso4VECIiIrKOFRAiIiKyLivoDBAlHBEdrp/aTsc571Y3ktV7Y4WOizZvLv98kXWZrVsax7+eWEfHW48u1PGyfmONdNmSqeMDqqjE74ee6/HtQB0f1m8FiFIdW0CIiIjIOlZAiIiIyLrAumAku4JxvOb2o3RcaZMyztV76VsdFxcU6DizVk0jXacPtug4v9JGHb/apkHZMktpRR3XUceTjxyj47O/uclIt2rM4Tpucq5X9qDM8kvJq/nEtcbxlPqv6DhcN0uoSM/d3up/Oh51ymAjXYV3vz50ZilpSE6OjjOqVtHx8ltaG+m69lii44lNZ+k4tEx1nDNUx43/+n28slnu2AJCRERE1rECQkRERNYF1gWTkd/MOD7vrFk6vrHWN8a5x67spOPn5x+v4/M7mc2Sd9ddoOMJOxrGI5uUBjLr1DGOC0Zu13H/WdfouNV9XxjpMv7URsf7enXRcfZ78+KdRbJo2/RWOn6q4WTj3AHldbtkwJstFTq7Jdw5//dDz/WvvEvHfZ83Z9Ucv/A8Hdfov7z0B6CEkFGxoo73nHSEca7yTet1/Eb+W74zM8Pe74Dy2guKUWyce+2o53R844lXenmYvQCJjC0gREREZB0rIERERGQdKyBERERkXWBjQIoWLzWOv+joTct955LrjXNbT9in4zu6TdfxzuKKRjp/P+nEDi/qeNLJ/Y102R/Mjz7DlLKKm9Q1jmd1mKTjPiOGhL9u0Y863vdgcx1nf5AZkjD8NExKPEp54zRCpzuGW9U0NN1Nv5yg46/Gddbxb77VUwFgcf/REd3PnydKDr8N9H7vnz7wRETX/GdnfeN4r/I+F4dVXxuaXGud7ZWdbfne52Kt2RG9bGDYAkJERETWsQJCRERE1iXkZnS1npsTcuzFr+NwhFOn+QEdN/vUa4ba2C3HSNf4gzJmkFJKxnKzafOBLd5qhL918FYprDU3/D3ubOl1DT7ZtK9xrnDl6rJlkKyqeeoyHd84r5tx7pH6XiEobRpulnhdKHU/2aTjWs/9ZKTrOd1b8XROp/+Evd/czq/puHi9t9LuqQ27gBJDVrM84/jiW98Om3bJfm8a7YVPe0MOGo3+1kj387XeqszDhj8Z9n6z91TWcd3ZXnlL9M5ftoAQERGRdayAEBERkXWsgBAREZF1CTkGJN7OOcecizT3vuyAckKJqGjHDuP48zPa6njg2+/p+MU6fYx0DWfv1nFelrdM+/ajzHFKVTkGJGl9/sxRxvGBEZ/ruLRps/6xItdN8tL9dLR5/ypP5uq4+HlvbEfo/cK91pp7jjXS5Y0wx8+RPYWr1hjHzz98mo6/vMzcoXbpqPY6bvAf773DXGAdePuKB31HlXT0e/F+I93fX7hOx42XmVtGJDK2gBAREZF1rIAQERGRdWnRBUMUDf+02Y+7N9LxnrvNBtJLXnxTx20reNPg/nz750a6+dOq6bi4oCBe2SQLaj9jdmkc2chr6u7dz9v12N/lAphTdB9v4JWHjPXmiqYtpx1d4jWR7q5LiavmeK/sbBhvnquGkuf0r3rA7FKrl1lyutvWm93Bje9Pnm4XP7aAEBERkXWsgBAREZF17IIhKkXR9t913Oo6szn06f+do+Nez3mbTc27pKORThWYI+ApeflnmXy+zmsu98+OAcLPWgntPvFvRleM8LNqwt2Ps16S37aLvHL05YWPGOeyxZuxOXKTt7ndxnNyYdpVLnkrb2wBISIiIutYASEiIiLrWAEhIiIi6zgGhCgO7tz4Zx2reRzzkQ78U3TPfOYY41zreV7ffbgddAFzbEcs03D39TWXVs2Z8XVEeafEsbm7t4t75Yzwq3RPnnm8jpuvTY2xP2wBISIiIutYASEiIiLr0qIL5pgqPxnHX7fzNgkq+mGZ7exQith/mPfn83NBDd+ZjfYzQwllxaAmOvZvRudfFRUIP0U30mm4okBJ7o7jp4U9d/vGrjpu9a8fdFxUUuIkxBYQIiIiso4VECIiIrIupbpgClev1XHnL4foeEHXCUa6+zrW1HH1H0AUkxpXrdHxljFNdVyNXTBpr2iZ1+37k2+iSrfLhhvp5o58SsexzILZfc12I12Fd2PLLyWmNz73Ck+r7V8GmJPywRYQIiIiso4VECIiIrKOFRAiIiKyLqXGgKDYm5zU6F7v29nTM0tITBSddbcdZxx/3fIxHf914SAdp8oUOYqPNfd4u52e3Pcb41xZp+EqZa6sSsnnyafP0vEFtz5unFt69hgdtym6Wset715ipPPv2p1M2AJCRERE1rECQkRERNalVBeMZHmPs+rs6jp+e3dlI12Nb7fqmM3lVJrM6l45evBv441zj209QsfFK3+2lidKDJmtW+r41xPrGOe+Gjlax8WYr+PQzeiKEdlmdO/tqaLjh4dfqOOa73LzuWRX74kvdNzu6MuMc9/9eZyOfzzXK1PrztpjpLvwxht1XGVK8kzXZQsIERERWccKCBEREVmXUl0wmU0a6bhn74U6nl/QzEjHDegoUgc6ttBx/8qfGOfuuv8vOq55YI61PJE9+/oebRzvvtqbbdC1nrcS7pT6rxjp/HrgrOoAACAASURBVF0r4Wa6lHYu/50rjXRN3/R2navAbpeU1fLCBcbxcdf8Xccv3jBKx20r5Bjpxj3szci7uPINOq41d5ORzr9CbyJgCwgRERFZxwoIERERWccKCBEREVmXUmNA/J5q+JmO79vypwBzQslm6zBv5cpnR3h9q8csuMBIV2uiN71SgZLVsmfNcR7L+o3VcQbMlUuLfb9p/7RZ/5iP0HP+sR2h03Bv+uUEHX80o7OO80dwTBEB9Z70pujeMt+bovvL8VWMdPP+7q2g+sk/n9Dx3Zu6GOkW/qWWjou2bYtbPmPFFhAiIiKyjhUQIiIisi6lumBWDmkQdBYoBZx743s6fnRjLx3XPH2VkU4VFlrLE5Wdf0qtfzrtss5jjXSxTJuNdPO4E+6+zkhXd7Y3TTJvGbtdKDz54lsdN1poru7dbbdXrp666Skdj6w730jX5tErdJw/zDwXBLaAEBERkXWsgBAREZF1KdUFE24qwkW55uY8n/TxVperMHNeeeaIksDPI44zjs+q9qCOPxp8jI5V4RJreaKyC13F9MPnvI29ws1mAUqfteI/598g7q4fTzPS7fzGm22Q55vRUgtmNws3w6RYFBcUGMd1x3izZa7vP1DHn3YyV+id0vNpHf9f52E6VgsWxzuLEWELCBEREVnHCggRERFZxwoIERERWZdSY0Bqf1dyj+pOZT5m1l72vKajjGrVdLz8rvY6fu4McxrmpX/zprRlfxv8VDWKjX+qLWCO+4h0qq1/pVIA+Gqct1pptXXeNOyaITvU1owhv0SR8r+XAcDaq47Q8SvtH/WdMct2hwremKai6hW8+8U3exFjCwgRERFZxwoIERERWZdSXTDrBpS8MuXAry81jpvMXmAjOxSwjCrmhk295qzX8Yyan+q42duXGenyP/iqfDNGVoiY8/LDbRDnn04LAA8Pv1DHFUK6VkKn0RKVp1+v9ZYIKM72vn/A7IHBwksf9x15ZfuzvRWNdFdM8d7rWn7traxaXLZsxowtIERERGQdKyBERERkHSsgREREZF1KjQEJp8p7VYPOAlmSWdtbAvvIDzYZ5w4or2/05EEX6zh/ttnPT6mhypO5xnHLMy/X8eL+o3XsH/MB/HHcB1FQpt7obQvRJKuSjotLGbVx2o9n6rjwgXrGueYfemOYghr34ccWECIiIrKOFRAiIiKyLmW7YBbv961SuLiglJSUSoq2/Kbjrztlhpz1pltm4htLOaKghHal5L/rxWfC2+W4AtjlQonpmo79ddz4vf06vqXe+0a6fhNu1nGze7zVm7MOrCvH3JUdW0CIiIjIOlZAiIiIyLqU6oLJH+Y1Pd2KrjoWfFtSciIiooRVtN3bUHG112uIq9DdSNfUt0Kvuf5vYmMLCBEREVnHCggRERFZxwoIERERWccKCBEREVnHCggRERFZxwoIERERWSdKJdOkHSIiIkoFbAEhIiIi61gBISIiIusCrYCISE8RSezdciIgIneJyMSg85GuUqgcvSgi9wWdj3TFckRllUJlyMpnWkQVEBFZLSJ7RGSXiGx0C3jV8s5cmLxcJyKrRGS3iCwRkfwIr5slInvdZ9giIv8Vkfrlnd8weTlJRH4UkQIR+VhE8oLIh22JUo5EpKn7cy9wfw8nR3Gt/xl+DfAZRET+T0R+FpEdIvKaiFS3nY8gJEo58uXnRBFR0XzosxwFKxHKkIjUFZFXRWSDiPwuIp+LSNdDX6mvT4jPNBEZ7Obh4FeB+/fQ5VDXRtMCMkApVRVAJwCdAdwWa4ZjJSKXAPgbgP4AqgI4FcCWKG4x3H2GfAC5AEaV8Brluj+OiNQG8F8AdwKoCWAegP+U52smmMDLEYBXASwAUAvA/wGYIiJ1orj+4DMcCeAoAHeEJijvcgRgCIALARwPoAGASgCeLOfXTCSJUI4gItkAHgfwZQyXsxwFK+gyVBXA1wC6wPkseAnA9CgrQoF/pimlJimlqh78AnAVgJUAvjnUtVF3wSilNgKYCeeXBgAQkW4i8oWIbBeRb0Wkp+/cMLelYqeIrBSRy6N9Tfc+GQBGArheKfWDcvyklNoawzNsBTAVQAf33qtF5FYRWQRgt4hkHeKZmonIbPeZ3gdQO4qXPwvAYqXUZKXUXgB3AegoIm2ifY5kFmA5yofzhj9SKbVHKTUVwHcAzo7hGdYDmAGvHCkRuVpElgNY7n7vVBFZ6D7TFyLyJ19eOovIN+4z/QdAxShefgCA55VSa5VSuwD8G8BAEakc7XMks6DKkc+NAN4D8GOsN2A5ClZQZUgptVIp9ahS6helVJFS6hkAFQC0juFeQX6mhRoKYIKKYIpt1BUQEWkEoC+AFe5xQwDTAdwHpxZ3E4Cpvv9RboLTUlEdwDAAo0TkyDD3HiMiY8K8dCP3q4OIrBWnG+Zut2IS7TPUhvOBs8D37fPhtKzkAqh3iGd6BcB8OL+ke+H8wP33XyQig8K8fHvA255XKbUbwE/u99NGgOWoPYCVSqmdvu99ixh+/iLSGEA/mOXoDABdAbQTkc4AxgO4HE5ryzgAb4tIjohUAPAmgJfd552MkEqQ+0ZhbnsZkoWQOAdAq2ifI5kFWI4gTtfpxQDuKeMzsBwFKMgyFJK2E5wKyIoYniHIzzR/ujwAJwCYEFHGlVKH/AKwGsAuADvh7Pb7IYBc99ytAF4OST8TwNAw93oTwHVu3BPAugjzcJz72tPdH2hTAMsAXBrh9bMAFADYDmA9gEkA6vie72Jf2rDPBKAJgEIAVXznXgEwMcJ8PA/gXyHf+xzARZFcn8xfCVKOLgQwN+R79wN4Mcpn2A5gDYAxACq55xSAv/jSPg3g3pDrlwI4Ec4f6Qa4a/G4574AcF+E+bjELf9NARwG4G339Y8N+vecDuXITf8WgIFu/GKkvzuWo+C/EqUM+e5RHU5L7G1RXDMLCfCZFnLPOwHMijR9NK0HZyilqrk/4DbwmmjyAJzj1rK3i8h2AN0B1AcAEekrInNFZKt7rh9ia97Z4/77oFJqu1JqNZz/CfSL4h7XKqVylVINlVKDlVKbfefW+uLSnqkBgG3Kabk4aE0UedgFp7D5VYfzh5AOgi5H8fj5n+GWozyl1FVKqT2+c6Hl6MaQZ2oMpww1ALBeuX+1rmjK0Xg4Y1lmAVgM4GP3+0k/Aj9CgZYjERkAoJpSqizjt1iOghX0exHc+1UC8A6c/xj9M8rLE+EzzW8InLEsEYllDMhsOLX9h91vrYVTs8r1fVVRSv1LRHLg9Es9DKCeUioXwP9gNvlFaimA/XBqqzo7MdwnHP+9wj4TgF8A1BCRKr70TaJ4ncUAOh48cO/Twv1+2giwHC0G0FxEqvm+1xHx+/mHlqP7Q56pslLqVTjlqKGI+J8h4nKklCpWSo1USjVVSjWCk//17lfaCLAcnQTgKHFmUGwEMBDA30XkrbI+k4vlyJIAyxDc+70Jp8JX1vFIoWx9pgEAROTgQOYpkV4T6zogjwHoJSIdAUwEMEBE+ohIpohUFGcudCM4/Vk5ADYDKBSRvgB6x/KCSqkCOLNFbhGRau79LwMwDdBTK5WINI3xmfzCPpNSag2cmSt3i0gFt291QBT3fgPOOJazRaQigBEAFimlYh7ElsSCKEfLACwEMNJ9jTMB/AnOm8rBefzxqtg+C+AKEekqjioi0t+t/MyB0+x5rYhki8hZAI6J9MYiUlNEWrj3bQfgUQD3KKWK45T3ZGK9HMFpas6HM3CxE5yui2fhjAlgOUo+1suQODOopsBp3R8a+jNPos+0g4YCmKrM8XWliqkC4jbzTAAwQim1FsDpAG6H80tZC+BmABluRq4F8DqAbQAGwflDLZGIjBWRsaW89HA4Tegb4PzhvQKnCRFwmiTXIA4199KeyU0yCM4Asa1wZuYYA25EZLGIDA5z781wBgvdD+dn0hXAeWXNczIKsBydB2fa4zYA/wLwV1/TZWM4fehlppSaB+BSAE+5r7UCwEXuuf1wZkRdBKccDYQzPdv/HLtEpEeY29eG8z+v3XBmUIxXzij6tBNEOVJK7VRKbTz4BedDZLfyZuWxHCWRgN6LjoMzmLU3gO3iraNx8HeVFJ9p7vmKAM5FFN0vQAptRicidwDYrJQaF3ReKHmJyHMAJiulZgadF0peLEdUVunwmZYyFRAiIiJKHtyMjoiIiKxjBYSIiIisYwWEiIiIrGMFhIiIiKwrdZe8XhnncIRqAN4vnhzTojaJiuUoGKlUjliGgpFKZQhgOQpKuHLEFhAiIiKyjhUQIiIiso4VECIiIrKOFRAiIiKyjhUQIiIiso4VECIiIrKOFRAiIiKyjhUQIiIiso4VECIiIrKOFRAiIiKyjhUQIiIiso4VECIiIrKOFRAiIiKyjhUQIiIiso4VECIiIrKOFRAiIiKyLivoDETr5xHH6bjJPV8EmBNKB1kNG+j4QNO6xjkpKvYO5i7S4a/XHGeka/DKjzou+m1rnHNIRJSc2AJCRERE1rECQkRERNYlZBdMZu5hxvGPj7fQcX7jNTpW91jLEiUhyckxjrede2SJ6X7rqIzj6/tN13GdrFU6bpi1zUhX5Ku/f7e3sY4vOexxI127rpfpuOUF7IJJN8uf6Krj3CXm//levvVRHd902jAdFy/6EUSpji0gREREZB0rIERERGRdQnbBFLZrahwvPflZHZ8yxGvOzsJ6W1miJJFRpYqOf3ysnXFuRb/RcX41bxbM8TlrfN/PNFK9evwzOh429SIdN716i5GucOOvcc0dBWfnwG46nn+m182y8TQzXdsKlW1liVLIvr5HG8c7m3gf5dmnbw573dxOU3TcbPqlOs6/9Os45i5ybAEhIiIi61gBISIiIutYASEiIiLrEnIMSM2H1hrHj25to+PMgkLb2aEE5h/zAQA/jvLGfazoNy6ie7y5O9c43quydfzRtrY6/vbZI8LeY1vPvTq+55i3jXOnV/HGKi3q9rKO//zS2Ua6qoNq6pgrpiaXzPatjeNb7puo45vW99bx8t/rGOmGNpmjY/XDinLKHSWyzNYtdbxysFk+co/yxnP4x28AC8v8uqv6e2Mrmz17qXHO1pgQtoAQERGRdayAEBERkXUJ0wWz61xv2lqDzMXGuY+O8JrZBd+W+bUyOnrN6vtredPgsj6ab6QrOMtbwfCXYyOrq9X6zjzOnTCn5IQUF7t7tTeOV/QfGzbttuI9Or5g+bk6zjhrl5GuaPvvviPvXC2E/13Wet6LJ6Cxce6OcWd5+TvV6xb6uMNUI90pr5/u5ekkdsEkk3V9ahnHvSt5v7/rl3jdMzcc84GRbtU+r8ldFbJ7OZVsufxYHYdOjT2mrjdt/4kGUxDO9IKKOr52gzf19p0FncJek/dfL86ZYXaltJ7ndS8/0cA75++OAYB+rf+q46Kl5dc1yBYQIiIiso4VECIiIrKOFRAiIiKyLmHGgGw4uUjHzYozS0kZXmbLZjo+YvKqsOlaVJyt41qZXh//h7+bS3d3qvqOjodVN6cGhzN/n3n8xvVddLywc0S3oCi0+scPEae9YV1f76C3t+x5UTn3vdf62vdndmr4dC+0elXHg0+9wThXcdpX8c4WlVFGB295gBeuecw4137acB3nX+H97iZM62qk27LOmwKeD/6OU9Wd+dOM4/6VvWn7/nEeIx4aZqSrPc4/7uyAjvIR2TRZ/zgUAJjZ4OkS0/nzAJTvuA8/toAQERGRdayAEBERkXWBdcH4u0sA4Oh2K3X82xWHG+cKT2qh44orNnnfX2N2i1R4vkDHLSp66TJ9u5YCwJDqJe+ie8t0c2rTu5kddTzs7DElXhOqS455PHJrI9/RuojuQaVbNuYYHU9rHNqkKGGve6HJLB13G3K1jmuOL9+p0rXGe03rt199pI4fqPuNka5+pjclPOO6kJ1xzRZcCkqG1z285i7v7XP1gdpGsnb/9n5/nFybnvzdJ0+Ma2OceyLcNaVM9Y+Uf2XVe25+IWw6f7fLja+YXT95cchHJNgCQkRERNaxAkJERETWBdYFk/vSduP4pabeCoGDx/Q2zvWpNUvHrwzvr+OskC6YzXu8FVOnXOLdQ4W0yo9vYI74PSh/2qKQbzTV4fMnN9HxwGrLjWRVM7x+lzOXm9Mcsi7yYjbFxm7Z0163y3cDvAbMDFQIe03r164yjquu8erb9V6wuEJtsTfDq7A4sjr/B+3eMI774cgwKckmday3IeEPx72o4yMeM8tag1VflHj9/7X+n3F8/bpB8cscEYAlN9XQsX+2Tah7l3mfVXkjglmxmy0gREREZB0rIERERGQdKyBERERkndUxIPv7HKXj+xuZKwfevdlbse23EU2Nc09d5+0Yefjm3To2J9cCVU9ZiZKETsysGiZ/offDQm+Vzbd7eDtajn7uRCPZN8e8rOPlv9YxzjVdGzKuhCKS2b61cTyz3ygdV5JKYa+7cPVJOm796M/GucL1G+KUu+hkNfXGD7WqNL+UlJRwMsxVmbff7r3/3LHJGw/S8FFzFVMV5nZdczYax9nbYlv1mdLbvr5HG8d7r9mm41Wdng1NXqIDb/k/q+ysfBqKLSBERERkHSsgREREZJ3VLpiNl3g7tTXKMpvRp7zVQ8d5H5lT2Op+5MV/6CaxRDX2Vmf1d7lQ+Thv6ofGcYuskrtdxv6eZxz/PsjrYCtcH9kGguXt15Mb6vjSw96M6JoP9+QcOhGVu4wOrYzjOZ1e0XGPG72pt9UK58Z0//pfFJX4fck2p5erA/tjuj8lF/8qpoA5pXZA54U6fqJBZN0spak3a7OOSy6F5Y8tIERERGQdKyBERERkXbl3wfz0SDcdLzzWm/lyxdqTjXRN75un43AjyG1bMcrL+5d/fUTH+TOuM9K1uW6JjpsVLjXOJcqzJJsh1bcYx0VhfpDdKv1kHE9t3UfH2WuC6YL57ZJjjeMuly0Mk9K0utDbTPGRQUNDzn5X1mxRDNYMqGkcv7vH2zDwsLe9GW6xdg0X5Xhz9PzvlXW/NtNVey22Lh5KfP4ZLTc/aXbvl7aSaVn97+MpOvZvTAcAD11zoY5zZoQUxjhiCwgRERFZxwoIERERWccKCBEREVlX7mNAlp43RsfFyNbxAWXWfRJhmtny0V2N4/OO86YDd51yo47z3jMnLRXv3g2Kr33qgHGchZJXjOxUwSzCFf7xi44LKh1jnKv0lrlaZVn5p0pKtpePyXc+ZKRrklUZkVhbWN07+IpjPhJB4w93Gcf9r/b65OfN8Xb0fuVtc3XkXN+G2ZtO8MryYRlmGXz4Qe/9ceSq03Us/9hkpONYstTin277p3u9MWKxjvnwj+G48ZVhOq6y3ky321sRAI8MeiHs6/Z/3pvm2+a5K41z8dw5ly0gREREZB0rIERERGRduXfBZIpXxylWXtdFk0rbjHS/NWuu48JVa8o1T4UnddHxusu85tFFxz9upPtojzcFb8EL7XRcvOjHcswdAcARL11rHC8ZOjqi66a3fkfH257aY5zr1uPG0ORl0q+nt7HcqPqf+c5E1uVy9PzzjeOaj1bRcSa+KVPeKE6+NLvC/jzsEh1Xuc1r335nyMNhb1FZvA6UyhnmVphDJ3nN201Het0zqjiotSkpXvzdLFtHmeeOqet9xj3RIPpprqHTZkf3P1XHeUvDd5HU9sVPjGij4xvvMZcOyD1qc4lxvLEFhIiIiKxjBYSIiIisK/cumCJV8hqBI+uYq0MOfOlw38HhxrnCXzZG/brF3Tvp+KdzzY29Fpzlrch6+g9eM3jnSdcb6Rp+UqjjnEXltxoc/VGLexYYxy1rX67jab2f0HGb7PCbttXIMDewW3r+mDAp7Wk5zXuO1sO/Nc4lwkwwCqHM+ScVZnorNh+Y6X3/Ghwf9hYFZ3mz6z59apxx7vCvfF0t7HZJaqEbya0cXEfH+9YXGuf65E+L+v7XbvBWTF161IGQsyuivp9fPGe2RIMtIERERGQdKyBERERkHSsgREREZF25jwGZtLOujrtV8qYeNcsypxH9p8W7Oj7/9T7Gud3XeFNg9zzo7Rh6WoNFCKdlzus6vvk1c2fRc865QsdVVv+q4+a/BNMPRn9UvNdcmS//Mm8Mzs35F+h4Z/vaRro7Hxqv45Mq7Sun3JXutV11jOOR087RcZsR3rTOYo75SAvr/+zFqw6YK6tW/mixjmPdUZcS0+FzCsOe698/shVPm02/VMf5l6beOES2gBAREZF1rIAQERGRdeXfBdOmkY7/ObmvjvftrmCkO6ejt6rkq81nGucwPbLXKlBek3anadfpuOUHZnOXzPGmP4ZvJKNEVbTsJx1X9sUA8Ojqc3V8X5MqxrkWty3R8TONZ+k4AxJTPvzdix9vb6vjX4bnma87b66O2cyeflQlb3rtgZCyxo0sk5t/6m3RUnMq7M6e3uqi99z8AsLxT6/9apP53tH2YW/F8FScpM0WECIiIrKOFRAiIiKyrty7YPzy/uVrgF5kbvK0qK23Mc7oV7cb53pUXhbR/Yc86202lv/AFzHkkJKdWuDNKqhkLqaKDW958QmDrvauKaUavvkUbyZNnXfNVVdrLvA1jy5e6jvzfWSZpbSQvTk76CxQGYSucPprzzolptt6Uw3jeFX/p8Pes81z3iaE/lVIDwtZ0TQVu1382AJCRERE1rECQkRERNaxAkJERETWWR0DouYvDn9u0Y86ntE+1zg3A8dEdP9G4LgPikz1V+YeOhGAwyaGP5fq/bMUHw1ne5P9sweZu+tm1fd2/o5l128qf1tHmce7vQ2R0bufd9An1xzXGG6cBwDkgatuA2wBISIiogCwAkJERETWWe2CISJKNxXe9TYRK1CZxrldRzXRccV32AWTiOZ2mmJ+o1PJ6Xr+7VLjOG8Gu1kOhS0gREREZB0rIERERGQdKyBERERkHceAEBFZctH3Q43jjFxvTEhF25mhiHRb+FfjePs8byn25pM26zhn6deg6LAFhIiIiKxjBYSIiIisYxcMEZElWZNqGscbTzqg49yXbeeGInFYP3OHWv+OtVwNuWzYAkJERETWsQJCRERE1rELhojIkuqvzg05DigjRAmALSBERERkHSsgREREZB0rIERERGQdKyBERERkHSsgREREZB0rIERERGSdKKWCzgMRERGlGbaAEBERkXWsgBAREZF1gVZARKSniKwLMg/xICIXichnQecjXbEcUTykUDm6S0QmBp2PdMQyFJ2IKiAislpE9ojILhHZKCIvikjV8s5cCfnoJCKfisjvIrJORO6M4toXRWS/+wxbReR9EWlTnvkNk48ebh78X0pEzradF9tYjuJLRJ4RkaUiUiwiFwWRhyAkUDk6TkS+EpGdIrJIRLpHce0sEdnrPsMWEfmviNQvz/yGyUdT9/3H/34U8d9DskqgMnSviHwnIoUicleU1yZKGermvg9uFZHNIjI50nxE0wIyQClVFUAnAJ0B3BZLZsvoFQCfAKgJ4EQAV4nIaVFc/6D7DI0AbALwYmgCcZRby5BS6lOlVNWDXwBOBbALwLvl9ZoJhuUofr4FcBWAb8r5dRJRoOVIRGoCeAfAQwByATwI4B0RqRHFbYa7z5Dv3mNUCa9ja7+uXN/70r2WXjNoifBetALALQCmx3h9IpShGgCeAdAUQB6AnQBeiOTCqN8glVIbAcyE80sDoGtAX4jIdhH5VkR6+s4NE5El7v8SVorI5dG+pk9TAJOUUkVKqZ8AfAagfQzPUADnQ6iDm8dZInK/iHwOoABAcxFp46vVLRWRc33PVEtE3haRHSLyFYAWZXimoQCmKKV2l+EeSYflqOzlSCk1Win1IYC90eY9VQRYjo4DsFEpNdktRxMBbAZwVgzPsBXAVHjlaLWI3CoiiwDsFpGsQzxTMxGZ7T7T+wBqx/hMaSnI9yKl1EtKqRlwPrTL8gyBlSGl1Az372CH+574FIDjI7k26gqIiDQC0BdOzQ0i0hBO7e0+OP+jvAnAVBGp416yCc7/8qsDGAZglIgcGebeY0RkTCkv/xiAISKSLSKtARwL4IMYnqEqgMEAFvi+fSGAywBUg/NG8j6cD5e6AM4DMEZE2rlpR8N5068P4GL3y3//aSLyjwjyUQXAXwG8FO0zJDuWIwBxKkfpLOByJCUcd4jhGWoDOBtmOTofQH84/6uth9Kf6RUA8+F8aNwL5z81/vsvEpFBh8jGGnG6I19w85M2Ai5DcZEgZeigEwAsjiilUuqQXwBWw+km2AlAAfgQTpMdANwK4OWQ9DMBDA1zrzcBXOfGPQGsiyQPbvrj4BSSQjcfd0dx7Ytw3uy3A9gI4G0ALdxzswDc40s7EMCnIdePAzASQCaAAwDa+M49AOCzSPPiu+5CAKvgrseS6l8sR+VWjj4DcFHQv990KkcAarll4HwA2XDesIsBjIvw+llwWsm2A1gPYBKAOr7nu9iXNuwzAWjiluMqvnOvAJgYYT6qAjgKQBacD6kpAGYG/TtOhzIUco+JAO6K8pqEKEMh9/wTgK0AekSSPpoWkDOUUtXcH3AbeE00eQDOcZt1tovIdgDd4fyvDiLSV0Tmuk3Q2wH0QwxNhOL0ub4L4B4AFQE0BtBHRK6K4jYPK6VylVKHK6VOU07z+0FrfXEegK4hzzQYwOEA6sD5Y/WnXxPt87iGApig3N9cmmA5in85SkeBliOl1G8ATgdwA4BfAZwCpxUtmhkQ17rlqKFSarBSarPvXGg5CvdMDQBsU2YXbsTlSCm1Syk1TylVqJT6FcBwAL1FpFoUz5GsAi1DcRJ4GTpIRFoCmAGnMvZpJNdEPThFKTVbRF4E8DCAM+A85MtKqUtLyFAOnH6pIQDeUkodEJE38cemy0g0B1CklJrgHq8Tkdfg/PLj0cTlrwSsBTBbKdUrNJGIZMKpLTYG8KP77SbRvpiINIZT8MsyliFpsRzFpxyluwDLEZRSswEc7d47LPnqSwAAIABJREFUC8BKAI/Ecq+Sbu+LS3umPAA1RKSK7wOkScj1sbxu2qwRFWQZKmfWypB7jw8A3KuUejnS62ItZI8B6CUiHeE0HQ0QkT4ikikiFcWZC90IQAUAOXD6wgtFpC+A3jG+5jI4kwsGiUiGiBwOp4l70cEE4kwn6xnj/f2mAcgXkQvdcQLZInK0iLRVShUB+C+Au0SkstufP7T025XoQgBfhPzvOd2wHJWhHIlIBRGpCOfNL9v9maXNB4dPEOUIItLZ/Z1Wh/PhtVYpNdM9d3B6a9OyPRqAUp5JKbUGwDwAd7vloTuAAVE8Q1cRae3+LdQC8ASAWUqp3+OQ72QSVBnKdv+GMwBkua+V6Z5LljLUEMBHAJ5SSo2NJlMxvVm5zTwTAIxQSq2F0xR5O5xfyloANwPIUErtBHAtgNcBbAMwCE6feYlEZKyIlPgASqkdcEaYX+/eayGA7+EMqjnYorATwHexPFPIa+2EU6jOA7ABTl//v+EUPMBppqzqfv9FhEw5EpEZInL7IV5mCNJw8Kkfy1GZy9F7APbAGdPyjBufUNZ8J5sgypHrFgBb3NeoD+BM37nGcJqx18f4WFppz+QmGQSgK5y+95Fwfhb+51gsIoPD3L45nC7JnXD+DvbBGdeSVgIsQ8/C+bs9H8D/ufGF7rlkKUOXwClHd4lvPZlI8pUym9GJyAUA2iulgpjLTSmC5YjiQUTuALBZKTUu6LxQckqHMpQyFRAiIiJKHunYX0xEREQBYwWEiIiIrGMFhIiIiKxjBYSIiIisK3Uhsl4Z53CEagDeL56ciIvaxIzlKBipVI5YhoKRSmUIYDkKSrhyxBYQIiIiso4VECIiIrKOFRAiIiKyjhUQIiIiso4VECIiIrKOFRAiIiKyjhUQIiIiso4VECIiIrKOFRAiIiKyjhUQIiIiso4VECIiIrKu1L1giCgy9eZU1/GEvE+Mcy3+c4WOW14/11qeiIgSGVtAiIiIyDpWQIiIiMg6VkCIiIjIOo4BIYqD0HEffj8NHOsdDPTCHldfbqSr/MaX8c4WEaWgwz6rZRy/3vxDHRcU79dx3yuGG+kqTvuqfDMWJbaAEBERkXWsgBAREZF1CdMFk9m2lY53ta5hnNvWysvmrjZe81Klw/Ya6ern7tDxW21e1/Excy8J+7pV/ldNxzXHz4kix5TuCs7s6jtaqKMha04w0n0+t52O/d0xG04QI13LN+KbPyJKTcXKfO84oIq8cyjWcUHdTCNdxfLNVtTYAkJERETWsQJCRERE1iVMF8yRr/6o47vrfBuHO1bQ0XfHTgibqvhYpeOrLjebzjcM9EYaF67+OQ55olTS7JYlJX5/1YNtjeOWb/hWP/XNgjFmxwDoc32nuOWNyp+/C+7T0eNiuod/JhRnQVE8FPi6YxJ9WAFbQIiIiMg6VkCIiIjIOlZAiIiIyLqEGQOycHsj7yDGMSC3bzpSxzPWtA2bbnaX53VcPcObmDS20adGurZXXK3jZv/gGJB0t2JUN+N4Zp43hsPY8fYN7nibKsyp1qFjPRairPz36wGOB6HwDpzcRccjGo8OOZutozd3tUKyYAsIERERWccKCBEREVmXMF0wB/6yScdnf9zXODe15YwSr+n0uLnRToOHvWbL+sUlT5EEgPOze5b4/e1vNTGO845eF/YelH6O7/ZD2HMtrw/f7RLpiqnADlDw6s2pruMJebFNr/V3yfnLTWmbFvpXxuWquBRqX02vm6VtdrZxrkB5K4SPHneGjg/HFxHdWx3b0Thefql3//yL50WVz2iwBYSIiIisYwWEiIiIrGMFhIiIiKxLmDEgKPaWj91/aVXj1Pnje+l4fNPpOn776geNdAMybtFxw3+F7/tSB7z+MsnJ0fGZjc3pvzWydut4KuqGvR+lLv/4jVjHA4TuenuQf5dcAGgJTt+1JR7LqBtTr0PGABm/S9+YktIYY4xCrilt7IhfnwZczj+VHOh9lI6rXh5+TOJne70d5A9/LLJxH3576+UYx9d1fU/HM5Ab9f0ixRYQIiIiso4VECIiIrIucbpgfIqW/WQc/97di7vedoOOR/3tWSPdW1d6XTJDV9yo4ypTwq8qmOHrgrmp5lLj3PM7GoUmpzRTWvO8OY02/BRaf9O6/5rSpu5S/MV799oG8HbSnrmh7KuiRtrNUhp/Ptgdk/y2tfJ2df+qzVvl9jpVl203jp97uZ+OG0Y4lTcWbAEhIiIi61gBISIiIusSsgumNI3+6TUH3f3T34xzH416UsfjH3lUx9d/dpaRrnDjrzpePd5c/dTvwWmn67gF5kSfWUoKpW84Ft6vx3rdLv57NLvFXIXX37Tub8KvDG44ZlPo7yUcfzfZqgfNTS39m8SFbk5YVv7X/cMMKV93nblSa/hum9ByzQ3uEl/xiZ2N4/1/+T2i6+78wfusqoOlpaQsWdEPy4zjhiHH5YUtIERERGQdKyBERERkHSsgREREZF3SjQHxq/q6OY2xY4drdfzNxY/rOO8dsx/t81eO0/Gsrg/peJcy62O1v1Wg1GROm4xsCqV/5UsAaHCmVz4ina7Lfnh7QsdARLqSrX/cR+jvyz/uo7TdkcMJLUPmVGxvTFFpq+Ia40Mi3F0X4A67yeDeF54zjrvklJyu85dDjOMmF63VcVFo4gTGFhAiIiKyjhUQIiIisi6pu2BC5Y3wpsq2r3e1jlecGtL0epN/ZbdKOmo3/mojWdOJnHqbSsxpk163i7miqSnS1SlDm9b9uOJpMGKZdht63YTRod1z0XfXNfjE66pr+Ya9suB/XUpcGZ28LrW2FUI/c7yVUBfsL9bx4U+afTNFO8KvxJzI2AJCRERE1rECQkRERNalVBeMX7t7N3gHp4ZPt0vt03GzN81mLDZgJrfQlSp/GjhWx6Wtdumf0RLp5nGcYZC8Yt0ErrQyZLOrhZKPZHkfvT/dmq3jylKhpOQAgKse8GZ51vo4NYYHsAWEiIiIrGMFhIiIiKxjBYSIiIisS9kxID+flxdRun7fD9Zx1fmLyys7ZIl/3Id/zAdgTo30j+fYMMpcMdLPv+pkaatTUuIJHZeB0ZGN9Yh0N1z/yqU2dzYOLdfhcNXdxJXRoqmOF58wPmy64eu767jeOyt1XFguubKPLSBERERkHSsgREREZF1KdcFkdGij49eHP+w7UzHsNe93eE3H/fqbK6HmTP86bnmj8hG64Zi/ebr0jb8o1YV2QfR5o5OO/V11fywXwXStlMYs5+FXYzVXdU3O1TFTkpjdvCvuqhLRZXM3NNXx4RsjW9k3mbAFhIiIiKxjBYSIiIisS+oumMzatYzj5RfU0HF+ttftsrqwwEg39IYbdfzfxx7V8ZrTzPvnT49HLine/M3R/lVLQ0Xa5VLarAJ226SmZPu9Rrqxnn/WTqJ0HxGQ8ac2xnG4mS83bzS7lLOn55ZbnhIBW0CIiIjIOlZAiIiIyDpWQIiIiMi6pB4DsumMfOP4xwtH63hH8V4dD7npRiNd1ale3+gjdxyv45YtN8Y7i1QONpwQfuXSHldfruPS+sDNnXLNaY2cykiJJtIde7n6aeLIatpExy2eXxHRNct6VzeOa/2WGrvehsMWECIiIrKOFRAiIiKyLqm7YHY3CN8UP3GHNx2t6uTImiVXbqhtHLfE2tgyRuWqtGmzpTVB+6fv+u9hdrkAvx7LbheyL9Lp5X5/WO2XGyYmjII29XT8SP03wqZ7bJs3lKB4x65yzVOiYQsIERERWccKCBEREVmXdF0wkuVlucspPxjnZu3N1vGMc/2zHJaa9+jSXsd9q7+q4w9yzVk1lDgi3YzLz5zpEr7rxr96JMAVJCkYkXa7+LsMk21FV3LsUwd0POmZPjqud+CLILKDzHp1jeN9HRrrOOvD+eX2umwBISIiIutYASEiIiLrWAEhIiIi65JuDEjm4d7Uppfy3jHO3belg46Ll/6k453nmWMB/n6PN+6jR8VC78Q75u66lDiM6bWjw6ebucE/PiT8WJE+DTp59+aYDwpIvTnVD50oBHe8TQ6764X/eD3i3eE6zn8imHEffruObWocn3X/ezqe0b78duRlCwgRERFZxwoIERERWZd0XTC/9mkS9tyZh32j431fH6Pje+uOCXtN+8+H6jjvmdTe+CdV+DecC92YjiucUiIzp5MDE/Iim3prbLLIDecS0vKnzN/t+L7e7/arfeb7VOux3mapqnyzldDYAkJERETWsQJCRP/f3p2HR1HkfQD//hIgIAECJBxyY0hAlENQAQ/wRBEFL1RQQXYVRBdXUFddFQ/0dT1XQTx2VVbwYBHl8sDjFUVAuQWRQ7kE5AoQjnBmUu8f3anqmjcTZiZJz0z4fp4nz/PrdHV3NSl6aupqIiLfsQJCREREvku4MSB1Zm3X8ZbAAWtfm0qpOm5dJ/QUzE4P36HjZu8t1nFBaWSQypy3Dzwz6CWT3e9u59nimA+KL8FjlkIJHr/EcR/xT1WwR3OcVdkst541Y5C1L2vBAl/yFO/YAkJERES+YwWEiIiIfJdwXTCB1WaF0+6v3WftSz5jt44PLzert9WdF7DS1f7cTNctOHy4tLNIRFSkUG9kDsYp44kna/A8a7vn4A5mH+K7y6XKZDvvn00uu9VPvdgCQkRERL5jBYSIiIh8l3BdMF4Nn4ruJT7H88pzRBQ73hVNAWDWK2a1THvmC7tgqPxjCwgRERH5jhUQIiIi8h0rIEREROS7hB4DQkSUSIJXNO3+MVfupeMXW0CIiIjId6yAEBERke9EKU5KJSIiIn+xBYSIiIh8xwoIERER+S6mFRAR6SYim2KZh9IgIgNE5PtY5+N4VY7K0VgRGRnrfByvylE5elRExsc6H8ejclSGfPlMC6sCIiLrReSgiOwXka3ugzK1rDNXRD7aicgsEdkjIptE5OEIjh0rIkfce9glIl+KSMuyzG+IfGSJyBQR2eHmY4aIZPudj1iIl3LkyU9XEVGRfOgH3cO2WN6DiGSIyHvu/4fdIvJuLPLht3gpRyLSRUTmicg+EVkqImdHcOxMETnk3kOOiHwkIvXLMr/F5OUCEVkpIgdE5BsRaRKLfPgpjspQwn+muXlRIpLn5mW/iPw7nOMiaQG5XCmVCqAdgPYAHogmoyX0HoDvANQC0BXAEBG5IoLjn3HvoSGA7QDGBicQR1m2DKUBmAogG0BdAPMATCnD68WbeChHEJGKAF4C8OOx0hah8B5OA9ARwENFnN+PNXY+ArAVQGMAdQA858M140VMy5GI1AIwDcCzcP5PPwNgmojUjOA0d7r3kOWe48UirlOm5UhE0uGUo4fhPFcXAJhQlteMI/HwLCoPn2mF2iqlUt2fP4dzQMSZUkptBTADzh8NACAinURkjojkishPItLNs+8WEVnhfktYKyKDijhtuJoCeFcpFVBKrQHwPYDWUdzDATh/+FPcPM4UkSdFZDaAAwCai0hLt0a5S0RWiUgfzz3VFpGpIrJXROYBOCmCa89TSr2plNqllDoK56GTLSK1I72PRBbjcgQAwwF8AWBltCdQSm0G8BlMOVIicoeI/ArgV/d3PUVkiXtPc0SkTeHxItJeRBa59zQBQOVwry0iFwNoBOBepdQepdRRpdTiaO8lUcWwHHUBsFUpNdF9Ho0HsAPAVVHcwy4Ak2DK0XoR+ZuILAWQJyIVjnFPzUTkW/eevgSQHsHlrwKw3L2PQwAeBdA2Vt+kY4GfaSX7TCuJiCsgItIQwKUAfnO3GwD4BMBIOLW4ewBMEpEM95DtAHoCqA7gFgAvishpIc49RkTGFHP5fwK4WUQqitNt0RnAV1HcQyqAfgC8D+ybANwGoBqcB8mXcP6gdQBcD2CMiJzspn0FwCEA9QEMdH+8558uIveHmZ1z4TzIdkZ6H4ksluVInCbmgQAeL+E9NALQA3Y56g3gTAAni0h7AG8BGASgNoDXAUwVkRQRqQRgMoBxcO53IoCrg86fK6Gb9TsBWAXgPyKyU0Tmi0jXktxPIorx80iK2D4lintIh/O395ajGwBcBqdlpC6Kv6f3ACyEU/F4AkD/oPMvFZG+IS7fGsBPhRtKqTwAaxDFh2Ci4mcagNL5TPtOnO6sj0SkaVgZV0od8wfAegD7AeyD8zb7rwGkufv+BmBcUPoZAPqHONdkAHe5cTcAm8LJg5u+C5xCku/m47EIjh3r/gPnwmm2ngrgJHffTACPe9JeB2BW0PGvAxgBIBnAUQAtPfueAvB9uHnxHNcQwGYAN0R6bCL+xFE5mgLgOk+5GBnFPeQC2ABgDIAq7j4F4HxP2lcBPBF0/Co4Ta3nAvgD7lo87r454eYFwBvu9f4EoCKcB0ougPRY/52Ph3IEp0KZC6eiUBHOh34BgNfDPH4mnG+mue4z4F0AGZ77G+hJG/Ke4HS/5QOo6tn3HoDxYebjTQBPB/1uNoABsf47l/cy5KYvF59p7vOsEpwK82gAPwOocKzjImkB6a2Uqub+A7eEaeZrAuBa99tarojkAjgbTk0KInKpiPzgNvvkwvnGGEkTIdzz1ALwOZxvrZXhND93F5EhEZzmOaVUmlKqnlLqCuU0eRXa6ImbADgz6J76AagHIAPOO3S86TdEcT8ZcLoAxiil3o/0+AQW63J0OYBqSqmS9HP3dstRE6XUEKXUQc++4HI0POieGgE40f3ZrNz/va5IytFBAOuV0513VCn1gXvts6K7pYQT03KknBbLXgCGAdgG4BI431wjmQEx1C1HDZRS/ZRSOzz7gstRqHs6EcBu5bRcFIqkHO2H803eqzqcD+byLtbPonLzmaaU+k4pdUQplQvgLgDNALQ61nHRjAH5Fk7Nq3DA20Y4tcU0z09VpdTTIpICp2/zOQB1lVJpAD7F/2+6DEdzAAGl1DtKqXyl1CYAH8D545cG7wfBRgDfBt1TqlLqdjhNWflwCkuhxpFcSJyBal8AmKqUerKkGU9EMSxHFwDo6DYVboXzzeCvIlJaA4GDy9GTQfd0glvh3AKggYh47yGScrQ06FrB1z4uxLAcQSn1rVLqdKVULTjN3S3hDCovDcHlqMh7glOOaopIVU/6SMrRcgBtCzfc85zk/v64wM+0kn+mhbj2Mf9Noh0Z+08AF4lIWwDjAVwuIt1FJFlEKoszF7ohnCaZFLg3KCKXArg4ymuuhjOgt6+IJIlIPTgfHksLE4gzCLBblOf3mg4gS0RucvvmKorI6SLSSikVgDNq/FEROcHtQ+tf/OkMEakOpzlvtlIq3HEi5VUsytHDcGYdtHN/pgL4F5y+3MJ5/KX1Qf4vAINF5ExxVBWRy0SkGoC5cP7TD3XL11UAzojg3B/D+eDp7/57XQOnS292KeU9kcSiHBUOIq7o/p9+DsBGpdQMd19T93nUtGS3BqCYe1JKbYAzc+UxEakkzpihyyM498cAThGRq0WkMoBHACxVSkU9ODtB8TMt+s+01uJMJ04WZyzK83C6FVcc69ioKiBuU+E7AB5RSm2E0xT5IJw/ykYA9wJIUkrtAzAUwH8B7AbQF84DP9SNvCYir4W45l44I7bvds+1BE4/00j32EZwmg2XRXNPQdfaB6dQXQ+nn34rgH/AKXgAcCeAVPf3YwG8HXQfn4nIgyFOfyWA0wHcImbO9H4RKWmNM+HEqBztU0ptLfyB05WRp5yZCIDzLWBOKd3fAgC3wukT3Q2nr3eAu+8InPI8AMAuOA+ej4LuY7+InBPi3LsAXAFngNweAPcD6KWUyimNvCeSWJQj130Actxr1Ifzf7tQIzjN2JujvC2tuHtyk/SFM/B5F5w+/XeC7mO5iPQLce4dcAbAPgnn3+RMOM+94wo/00r0mVYXztTtvQDWwpnZ01M5szyLVW5eRiciNwJorZSKyboSVD6Is4DOxMJvskTREJGHAOxQSr0e67xQYjoePtPKTQWEiIiIEgdfRkdERES+YwWEiIiIfMcKCBEREfmOFRAiIiLyXbFvWrwo6VqOUI2BLwsmRrUwUrxiOYqN8lSOWIZiozyVIYDlKFZClSO2gBAREZHvWAEhIiIi37ECQkRERL5jBYSIiIh8xwoIERER+Y4VECIiIvIdKyBERETkO1ZAiIiIyHesgBAREZHvWAEhIiIi37ECQkRERL5jBYSIiIh8V+zL6IiIiKh0bRnWRccLh48Kma7lN3/WceaNi3VcoVmTkMcUbNth4gMHos2iL9gCQkRERL5jBYSIiIh8xwoIERER+Y5jQIiCJNeto+Phc77W8bTc9la61T1q6TiwbXvZZ4yIEtKefp2sbe+4jwIU6PimdZdY6Vre84eO8z2/v2DaMivdHTVXmWM+HaLjrFvnR5Vfv7AFhIiIiHzHCggRERH5LiG6YDY+bKYs1T13s4lP2GelG1xvpo7PqWwarL4+mGKle2DFlTre/1NtHTf/cI+VrmDJL9FlmBLayoeb6bhb5aMmrjfPSnfWhaaps8a77IIpj5Izm1nbl05ZpOMhaetCHnflbz10vOIHc47MEYutdAWHDpU0i5QAOg1bEFa6zS+3sLZTt/5QZLqvuja1f/GtCSdcOEbHI065yUpW8PPKsPLhF7aAEBERke9YASEiIiLfxWUXTMWZ9a3txS1e0vGUvHQdj9/S2Ur3p7n9zcZ20+1yage7qbR7Q9MM9cRpS3S8f8BhK92Zbw7XcZMRc8LJOiWgwz1Ot7Zv7/ZVWMdt76x0XOPdUs0SxVDOIPNc6XOnXRYGp63VcQFCm5T5iY6TMkXHv19vr0zpPcdLO87T8Q+jO1rpao6dW1yWKc71TltkbecEDur4sqfu1XG9r1dZ6QIhzrfhDfsz8o6aM4pMF6hR2dqWIlPFDltAiIiIyHesgBAREZHvWAEhIiIi38XlGJA+9ezV2yogWcd/+6aPjrMG2ekysbXI8x0M2l6UUkXHl516s47X9KlmpftiwDM6vjLnPh3XHcXxIOVJ44ftftdhNX/VcbKYOnpA2b3+tRex/l4erH3GHks274bndZyalBKcXBuy6Vwdz97YLGS6ZZ3f0XHDClVCpnu+vplyufuJmda+zm3NeLTsN3bqOLDiV1B8yrnNlKuOKfZ02gWHq+s44zUzvifUmA8AOHphBx0v6fSGtc/7ZGo59Q4dZ89bYqVTiC98ghIREZHvWAEhIiIi38VlF8zji3pa2/26vqXjxtNLPpFIHfZMt13wsw6bBy1WN+bCc0weBplpTl+NsrtqKPFsesCsrju98aigvaaMBXe7eKX9ylUsy4O/9pxubRfX7XLxL1fpuEpfM6W20Y6fi0oOAMj81yAdr+7xWsh0160xLyJ7tPE0a9/KPq/oeOT5bXQ8//qTrXTskolPSUHf9btVMc+Vp/+3gY5XrW5gpXv/4ld13CFloY4rSrKVbuxeMy231Wizonfg6JEoc+wPtoAQERGR71gBISIiIt+xAkJERES+i8sxIJnPHrW2d55jJtLmV/ZvMdlJ35+h40cu/tizh2NAElFyRoaOL7rGvNk2KWiB4pE5p+h4/HJTBlZ6xiJRYvt9hBkDdFXqM0F7zVTZi5Zfbe2perPnWbRjR1jXqrAz9GO258peOk7qbfru+48fYKWb19Gs9f9Q+lIdt7ztHCtd5t0cAxIv0t8w02sH9r/E2jeu6Zc6npI92ezIDn0+72i0o0HzaZ+ecI2OmyxPnGUi2AJCREREvmMFhIiIiHwXl10wavFya3vktm46/uNC0xCVNbFs89Ewe3vZXoB8tW5MPR1Pq2emVW8O2G8o/eox06wd6Bpv74+kaO26xaxMOevPz+q4RpK9OumI7e11XLW/PdU6f+u2iK9b0CD0dO2Nu9N03GjfJh2fODjXSvfYZ+1M/jLM6pYvXz7WSvfy3S0jzh+Vvd1n7bK2T3nyTh0vHfByic8/dYApz1fnmbfr1v8+z0onc34q8bVKE1tAiIiIyHesgBAREZHv4rILJtin33TU8f9c8YGOx7bvYaUL7rqJVIVmTaztF7Mm6PjaWYN13AKLSnQd8kdSO3uVyP8981XP1gk6Ou/De6x0mR95XhzVtVPI8x/KqFTE2She5TUw3Wk1kiqHTLd6fx0d528p+gWXkVh9/ps6Dl5Xt+aEqkUeE3zdaf8x3YIj7jFdMKdVyrHS7e1rymv19+wXoFHsqM5tre1mnX8P67hWk01XTepas/ppzxu/t9JNWXuqjrN7mplQj9xur/J74+t367jB07GfLcMWECIiIvIdKyBERETkO1ZAiIiIyHcJMQak8ZdmZdSeN2zR8YN322+tbHFzya5z8KR0a7tVRRM3nJQQ/1TksXKI3b9eJ9mM1PjkQKqOs8fYUysDYZ5/V7bpk+UYEPLKGdTZs2XGjHnfeAsANWas0HFx5a728qLfapqebE8h3tPMfKesfuxskk82XGY/IX7Onqpj77igLo8NtdJlvW3egKs8b7Zd8mZNK13DXDP+8UCH1jrOm1jRSrfwLy/puF2HgTpufO2y4rJfZtgCQkRERL5jBYSIiIh8lxD9ChW/WKDjU6ebJqrWLTda6fIrmNtR+flhnds79Tbj8TXWvhd2tdFxlSnzQImlZr29IfcNm2L66076LbrpigcbhNtZQ+Wd90WHANB98Owi020bfZK1nbo3vLIXSOF3xUSTVM28tPQ/fUdb+wo8HS8tp9yh4yzPC+wAIOidc1ogd0+IPYBaaLpjHm9+mrXv72vNFO4VZ40z132/v5Wu2Q3+rJjKUk1ERES+YwWEiIiIfJcQXTBeLcaZkcAPjbNXeev7vGnKanFXeE2bq28/Uccrm3xs7TtviFn9tArYBZNoxrUZa21vD5gukxbjTfdM8OqUROHwdru0+8KeSfVYncWeLc8KrF+vttKF24lX929rivz9b0cPW9snzgn94jvy17Z3zWdL+xT7KTP3kJnBmTXEv8+W/p/fpuPVvczK0P/t9IaV7r4uJl1ZvsCOLSBERESibLWqAAAHrElEQVTkO1ZAiIiIyHesgBAREZHvEm4MiMw204gGvv0Xa98j/T7U8cTmXXScv3a9la7gnPY6XtD3BR2f93NfKx2n3iaepDYtdZyRbE+F7DLLvFnypCVLEEqFBqbvFmlFr0AJAGd3MKtYLvrIrD7YN3NBUckBAB+ub6fjejdusfYF9oaeNkwlV2ehWVF5U/5BHTeuYK9SOaH5F2Zjc7RXM+M+zr5niI6r74xuyneSmAmZFcWswLsxv4aVLvkbvqk7VqznBoDh2V+FTDs51zs9NrwlI0qDNd6klwlbVbLbIvIamhV2U1F22AJCREREvmMFhIiIiHyXcF0wXo1G2qvGPV7nSh2nvmCasxvd1chKd9Fr3+i4+9KbdJzW63crXahV6Ch+7WmVpuPaSfaLuhqm5+q46Tyzr3bFPCvdxdU/1/E5lUM3j77deKbZaBxe/npUW6rj+1sMtHd6VjCk0pfy2XwdD7hzmI5HjXrZSteqov0Cr3DsKbCnv1724D06Tnt/bnDyY1r7dGdre3GTF3V8VJn8PbL6GitdDfwW8bWodOw+234IXJM6JWTaH17oqOMaiK5bLhqqc1vPlnnR3b/3NLfSeaeLl+V6z2wBISIiIt+xAkJERES+S+guGCi7kyR1nRkd/m3vN3W8b5a9Ct1Dm3vouNZNu3QcOBp6xgMlhrSZa3U8/7BdPr5u/VGRxySLXQ8PqPDWRvUe5z1mc+CAlW7EH5fqeP2IbB1XWhh6tgyVrcrTzGyAXpcOtfZVqWP+fqPav2/te3PruTqeu9o0W5/8xE4rXdrayLtdvHpfbDfLp4jpdlnteU5VfqVmia5DsVHjXf+6Xbz6j51e5O9zjlaztgM7dxWZrrSxBYSIiIh8xwoIERER+Y4VECIiIvJdXI4BSa5e3dred2ErHW+6JHT//KnZpv+/elJlHY/c2tFKl9PDjBUJ7Panr4v8Edi2XcdDnr3T2vfE8Ld1XDvJTL2tlWxPw90VMGWnTSUzCW1dvj0hrc8bw3WcvtRM1z3hd3tF04KlK3VcCRz3EW+Kexvp89lX2r/IMc+LrJ1mGmO0a1kmt2qh49UD03U8Nv3ZoJRm2viffjFLB9T4ZD4oPiV5vt9nTbvd2pdVhm9XTzrBrOz7+7hm1r4+qabMelfU3XbE/swF/HmrMltAiIiIyHesgBAREZHv4rILZss4+6U+Czq+quOFR0wz+Ld5La10n245RcdnLLreHNPhv1a6S4+cVSr5pPiW8ao9FfLlV1sWma5CE3ulXARMN9+D303T8bJDdnNmw6fmFHm+8CbxUiIIrCrdlUWTszOt7RVDzTTa1VeM9uyxV/Edtdt01dQeuE/H/r3GjI6l5vyt1vbUPPO3/cvZ9ovpvm5iXl6Zv2FjxNeSlBRre98V5iWX9Yeu0fGi5mOtdN5nU/v5/XRc7+/BV1gZ/IsywRYQIiIi8h0rIEREROS7uOyCqVXVXknyNE9TUcO7zb78dRusdJVgtutmZOg4sMRuFM+5ro251lslW7GQEl9wE2hB1/Y67uRp6Vzmz8BwKmeSM03XXdcPf7L2TalVdFO3t8sFAMaP7q7jjK18ZsWj/LXrre0Vh8xQgntrL7P27ZtqZtrNeOpcRCrzr79Y21Majy4y3dDN9rkXj/F01cz3rAK+fFXEeSgNbAEhIiIi37ECQkRERL5jBYSIiIh8F5djQHZPamBtT77frArY57x7dVwraAyIV2DHDh1fsPwqa19+r91m461oc0nlVcXlv+v4y4NViklJ5JAK9qN0w0Nn6HjItZ/oeHDaWoSyLt8MMvp8kN13nzGb4z4SzVszu+p4QG97BeQH082YkPtfsMcFhSMpqO1g8MbzdDxzrlmOIuuR5Va6mvtMObLXdY4NtoAQERGR71gBISIiIt/FZRdM8AqW57cdpuOH7v1Yx5N+7GqlCzWVKO/9+tZ2owHrdHw46lxSeRXI2anj9UfSi0lJxzNvt8uGB8+w9i29dZSOvS/9Oqrsc2wJHNRx35Gme7k2u1wSXouhP+p40LPXW/sKaqTqeOWd5kVwnU/91Uo3d5k9HbtQ3e/ttoOaH5suncy8H8x1IshvLLAFhIiIiHzHCggRERH5jhUQIiIi8l1cjgEJ1uo+s1zxPx64Usd3TZxupfvg7z10nPqVWap2Z9cjVro9c83SyM1gv8GQyOuVsb10fKTDfmtfUyz1OzsUR5Kamrcoz7n1uaC9KSjKtb/1sLZzRjXVce0POe6jvMrfuMn+heftD1mDTbzTToUszAvr/PE+1iMUtoAQERGR71gBISIiIt8lRBdMYO9eHTd7wDRTTptkT8Pt8PpCHT8/2jRd3fWH3QWz5jnTBZOoTVfkjxOfmRPrLFCcCvxmpvNfds8wa1/zu0y38Y/ftdJx5qOLrXRVD/0IouMVW0CIiIjId6yAEBERke8SogsmFLXgZ2t7RQcT98Bpnj3B652uBBFRaak24Qdre8cEEzeH6TZmly+RwRYQIiIi8h0rIEREROQ7VkCIiIjId6yAEBERke9YASEiIiLfsQJCREREvhOlVKzzQERERMcZtoAQERGR71gBISIiIt+xAkJERES+YwWEiIiIfMcKCBEREfmOFRAiIiLy3f8BtXAhUABbmCUAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 720x720 with 20 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light",
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "from sklearn.neural_network import MLPClassifier\n",
    "\n",
    "epochs = 25\n",
    "mlp = MLPClassifier(hidden_layer_sizes=(100, 200, 100), max_iter = epochs)\n",
    "\n",
    "mlp.fit(x_train,y_train)\n",
    "\n",
    "y_pred = mlp.predict(x_test)\n",
    "eval = Evaluate(y_test, y_pred, x_test)\n",
    "eval.get_evaluation_report()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "JeKmO4y1gIsp"
   },
   "source": [
    "##Σύγκριση υλοποιήσεων"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "lKCtJIcfgTD1"
   },
   "source": [
    "Όπως φάνηκε παραπάνω, ο έτοιμος Multi-Layer Perceptron εμφανίζει υψηλότερες επιδόσεις σε precision, recall, f1 από τον δικό μας. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "2eiuHVwLgQib"
   },
   "source": [
    "###Αξιολόγηση- Συμπεράσματα"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "d8MRX1zA_MDj"
   },
   "source": [
    "Συμπερασματικά, έχοντας υλοποιήσει και τους τρεις ταξινομητές (KNN, Naive Bayes, MLP) καταλήγουμε στο ότι οι έτοιμοι ταξινομητές knn & Multi-Layer Perceptron εμφάνιζαν υψηλότερα precision, recall\n",
    "και f1. Στην περίπτωση του knn, ο δικός μας ταξινομητής, για ευκλίδεια απόσταση, ταυτίστηκε με τον έτοιμο. Τέλος, αξιοσημείωτο γεγονός ήταν τα υψηλότερα  precision, recall\n",
    "και f1, για cosine απόσταση και το ότι τα τρία αυτά χαρακτηριστικά ήταν υψηλότερα για το δικό μας Naive Bayes συγκριτικά με τον έτοιμο."
   ]
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "name": "Untitled10.ipynb",
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}