Evaluation.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "name": "Evaluation.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "machine_shape": "hm",
      "authorship_tag": "ABX9TyOn8O+2dGb7QNoqRKMOIInM"
    },
    "kernelspec": {
      "display_name": "Python 3",
      "name": "python3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "zNgw2aeIinHz",
        "outputId": "d168c740-a583-49b1-c817-8821d9d35b41"
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive', force_remount = True)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Mounted at /content/drive\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XmvhsNgxioBQ"
      },
      "source": [
        "from scipy.io import wavfile\n",
        "from scipy import signal\n",
        "from scipy.io import wavfile\n",
        "from scipy.signal import butter,filtfilt\n",
        "from scipy.signal import cwt\n",
        "from scipy.signal import hilbert\n",
        "from scipy.signal import resample\n",
        "from scipy.signal import decimate\n",
        "from scipy.signal import spectrogram\n",
        "from scipy.signal.windows import get_window\n",
        "from scipy.stats import kurtosis\n",
        "from scipy.integrate import simps\n",
        "\n",
        "import numpy as np\n",
        "\n",
        "import librosa\n",
        "import librosa.display\n",
        "import matplotlib.pyplot as plt\n",
        "import json\n",
        "import os\n",
        "import skimage.io\n",
        "import cv2\n",
        "import tensorflow as tf\n",
        "import csv\n",
        "import sklearn\n",
        "from sklearn import svm\n",
        "from sklearn.decomposition import PCA\n",
        "from sklearn.manifold import TSNE\n",
        "from sklearn.metrics import confusion_matrix, plot_confusion_matrix\n",
        "from sklearn.model_selection import KFold\n",
        "import pandas as pd\n",
        "import seaborn as sns\n",
        "from sklearn.model_selection import train_test_split\n",
        "import IPython.display as ipd\n",
        "from shutil import copyfile, copy\n",
        "from sklearn.utils.class_weight import compute_class_weight\n",
        "from keras import backend as K\n",
        "import keras\n",
        "\n",
        "import seaborn as sn\n",
        "\n",
        "from collections import Counter\n",
        "\n",
        "import pickle\n",
        "\n",
        "from google.colab.patches import cv2_imshow\n",
        "\n",
        "import warnings\n",
        "warnings.filterwarnings(\"ignore\")"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jzIuQEakbFY3"
      },
      "source": [
        "#Model Implementation"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fLEKgdLI-3mg"
      },
      "source": [
        "We preliminary implement the three AI models: A multi class convolutional neural network (DTLMC), a binary class convolutional neural network (DTLBC) and a support vector machine."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JyraXvQLmNS6"
      },
      "source": [
        "# implementation of diagnosis clasifier architecture\n",
        "\n",
        "class DTLMC(tf.keras.Model):\n",
        "\n",
        "  def __init__(self):\n",
        "    super(DTLMC, self).__init__()\n",
        "    self.MaxPool1 = tf.keras.layers.MaxPool2D(pool_size=(2,2))\n",
        "    self.Conv1 = tf.keras.layers.Conv2D(filters = 16,kernel_size=(5,5), activation=tf.nn.relu)\n",
        "    self.Conv2 = tf.keras.layers.Conv2D(filters = 16,kernel_size=(5,5), activation=tf.nn.relu)\n",
        "    self.MaxPool2 = tf.keras.layers.MaxPool2D(pool_size=(2,2))\n",
        "    self.Dropout1 = tf.keras.layers.Dropout(rate = 0.15)\n",
        "    self.Conv3 = tf.keras.layers.Conv2D(filters = 32,kernel_size=(5,5), activation=tf.nn.relu)\n",
        "    self.Conv4 = tf.keras.layers.Conv2D(filters = 32,kernel_size=(5,5), activation=tf.nn.relu)\n",
        "    self.MaxPool3 = tf.keras.layers.MaxPool2D(pool_size=(2,2))\n",
        "    self.Dropout2 = tf.keras.layers.Dropout(rate = 0.15)\n",
        "    self.flatten = tf.keras.layers.Flatten()\n",
        "    self.Dense1 = tf.keras.layers.Dense(256)\n",
        "    self.Dropout3 = tf.keras.layers.Dropout(rate = 0.30)\n",
        "    self.Dense2 = tf.keras.layers.Dense(3,activation=tf.nn.softmax)\n",
        "\n",
        "  def call(self, inputs):\n",
        "    x = self.MaxPool1(inputs)\n",
        "    x = self.Conv1(x)\n",
        "    x = self.Conv2(x)\n",
        "    x = self.MaxPool2(x)\n",
        "    x = self.Dropout1(x)\n",
        "    x = self.Conv3(x)\n",
        "    x = self.Conv4(x)\n",
        "    x = self.MaxPool3(x)\n",
        "    x = self.Dropout2(x)\n",
        "    x = self.flatten(x)\n",
        "    x = self.Dense1(x)\n",
        "    x = self.Dropout3(x)\n",
        "    x = self.Dense2(x)\n",
        "    return x\n",
        "\n",
        "class DTLBC(tf.keras.Model):\n",
        "\n",
        "  def __init__(self):\n",
        "    super(DTLBC, self).__init__()\n",
        "    self.MaxPool1 = tf.keras.layers.MaxPool2D(pool_size=(2,2))\n",
        "    self.Conv1 = tf.keras.layers.Conv2D(filters = 16,kernel_size=(5,5), activation=tf.nn.relu)\n",
        "    self.Conv2 = tf.keras.layers.Conv2D(filters = 16,kernel_size=(5,5), activation=tf.nn.relu)\n",
        "    self.MaxPool2 = tf.keras.layers.MaxPool2D(pool_size=(2,2))\n",
        "    self.Dropout1 = tf.keras.layers.Dropout(rate = 0.15)\n",
        "    self.Conv3 = tf.keras.layers.Conv2D(filters = 32,kernel_size=(5,5), activation=tf.nn.relu)\n",
        "    self.Conv4 = tf.keras.layers.Conv2D(filters = 32,kernel_size=(5,5), activation=tf.nn.relu)\n",
        "    self.MaxPool3 = tf.keras.layers.MaxPool2D(pool_size=(2,2))\n",
        "    self.Dropout2 = tf.keras.layers.Dropout(rate = 0.15)\n",
        "    self.flatten = tf.keras.layers.Flatten()\n",
        "    self.Dense1 = tf.keras.layers.Dense(256)\n",
        "    self.Dropout3 = tf.keras.layers.Dropout(rate = 0.30)\n",
        "    self.Dense2 = tf.keras.layers.Dense(2,activation=tf.nn.softmax)\n",
        "\n",
        "  def call(self, inputs):\n",
        "    x = self.MaxPool1(inputs)\n",
        "    x = self.Conv1(x)\n",
        "    x = self.Conv2(x)\n",
        "    x = self.MaxPool2(x)\n",
        "    x = self.Dropout1(x)\n",
        "    x = self.Conv3(x)\n",
        "    x = self.Conv4(x)\n",
        "    x = self.MaxPool3(x)\n",
        "    x = self.Dropout2(x)\n",
        "    x = self.flatten(x)\n",
        "    x = self.Dense1(x)\n",
        "    x = self.Dropout3(x)\n",
        "    x = self.Dense2(x)\n",
        "    return x\n",
        "\n",
        "class SVM():\n",
        "\n",
        "  def __init__(self, classWeights = None):\n",
        "    self.clf = sklearn.svm.SVC(kernel='linear', class_weight=classWeights)\n",
        "    # other possible kernels include: \"linear\", \"polynomial\", \"rbf\", \"sigmoid\"\n",
        "\n",
        "  def fit(self, X, y):\n",
        "    self.clf.fit(X,y)\n",
        "\n",
        "  def predict(self, X):\n",
        "    return self.clf.predict(X)\n",
        "\n",
        "  def evaluateScore(self, X_test, y_test):\n",
        "    return self.clf.score(X_test, y_test)\n",
        "\n",
        "  def getSupportVectors(self):\n",
        "    return self.clf.support_vectors_"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "z9ADeZ5B-2Bo"
      },
      "source": [
        "We furthermore implement a mediator class that evaluates and moderates the prediction results from all three classifiers combined."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "diZwYcpT-1Ch"
      },
      "source": [
        "class Mediator():\n",
        "  '''\n",
        "  The mediator class is designed to realize agreement inbetween the different classifiers.\n",
        "  It returns the predicted class if all of the three classifiers agree with their prediction.\n",
        "  If they do not agree, the mediator class returns a inconclusive result.\n",
        "  '''\n",
        "  def __init__(self, dtlmc, dtlbc, svm):\n",
        "    \"\"\"\n",
        "    Constructor. Takes the pretrained classifiers as arguments\n",
        "    dtlmc: the pretrained multiclass classifier\n",
        "    dtlbc: the pretrained binaryclass classifier\n",
        "    svm: the pretrained support vector machine\n",
        "    \"\"\"\n",
        "    self.multiClassClassifier = dtlmc\n",
        "    self.binaryClassClassifier = dtlbc\n",
        "    self.svm = svm\n",
        "  \n",
        "  def encodePrediction(self, prediction):\n",
        "      if prediction == 'healthy':\n",
        "        return 0\n",
        "      elif prediction == 'symptomatic':\n",
        "        return 2\n",
        "      elif prediction == 'COVID-19':\n",
        "        return 1\n",
        "\n",
        "  def predict(self, X_spectrogram, X_features):\n",
        "    \"\"\"\n",
        "    Collects predictions from all three classifiers and medaites them into single output\n",
        "    X: Features of test sample\n",
        "    \"\"\"\n",
        "    res_dtlmc = self.multiClassClassifier.predict(X_spectrogram)\n",
        "    res_dtlbc = self.binaryClassClassifier.predict(X_spectrogram)\n",
        "    res_svm = self.svm.predict(X_features)\n",
        "\n",
        "    argmax_dtlmc = np.argmax(res_dtlmc)\n",
        "    argmax_dtlbc = np.argmax(res_dtlbc)\n",
        "    argmax_svm = self.encodePrediction(res_svm)\n",
        "\n",
        "    res = 2\n",
        "\n",
        "    if argmax_dtlmc == 0 and argmax_dtlbc == 0 and argmax_svm == 0:\n",
        "      res = 0\n",
        "    if argmax_dtlmc == 2 and argmax_svm == 2 and argmax_dtlbc == 1:\n",
        "      res = 1\n",
        "    if argmax_dtlmc == 1 and argmax_svm == 1 and argmax_dtlbc == 0:\n",
        "      res = 0\n",
        "\n",
        "    return res\n",
        "\n",
        "\n",
        "    def analyzeResult(self, res):\n",
        "      \"\"\"\n",
        "      takes the result of the mediator class and translates it into human readable labels\n",
        "      res: result of the predict() function of the Mediator class\n",
        "      \"\"\"\n",
        "      if res == 0:\n",
        "        return 'Covid unlikely'\n",
        "      if res == 1:\n",
        "        return 'Covid likely'\n",
        "      if res == 2:\n",
        "        return 'Test inconclusive'"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8-eIQBgBppZP"
      },
      "source": [
        "# define two functions for one hot encoding because binary classifier needs different labels\n",
        "def oneHotEncoding(labels):\n",
        "  oneHotList = []\n",
        "  for label in labels:\n",
        "    if label == 'healthy':\n",
        "      oneHotList.append([1.0,0.0,0.0])\n",
        "    elif label == 'symptomatic':\n",
        "      oneHotList.append([0.0,0.0,1.0])\n",
        "    elif label == 'COVID-19':\n",
        "      oneHotList.append([0.0,1.0,0.0])\n",
        "\n",
        "  return np.array(oneHotList)\n",
        "\n",
        "def oneHotEncodingForBC(labels):\n",
        "  oneHotList = []\n",
        "  for label in labels:\n",
        "    if label == 'healthy':\n",
        "      oneHotList.append([1.0,0.0])\n",
        "    elif label == 'symptomatic':\n",
        "      oneHotList.append([1.0,0.0])\n",
        "    elif label == 'COVID-19':\n",
        "      oneHotList.append([0.0,1.0])\n",
        "  \n",
        "  return np.array(oneHotList)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qntx6-xhFuQp"
      },
      "source": [
        "Furthermore we implement some helper functions to visualize results and learning process:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RnYvZeuyFzoH"
      },
      "source": [
        "def plotConfusionMatrix(labellist, matrix, filename='', safeImage = False):\n",
        "\n",
        "  df_cm = pd.DataFrame(matrix, index = [i for i in labellist],\n",
        "                    columns = [i for i in labellist], dtype=float)\n",
        "  fig = plt.figure(figsize = (10,7))\n",
        "  sn.heatmap(df_cm, annot=True)\n",
        "  if safeImage == True:\n",
        "    fig.savefig('./drive/MyDrive/COUGHVID/Plots/' + filename)\n",
        "\n",
        "def plotTrainingLoss(train_loss, val_loss):\n",
        "  plt.plot(train_loss)\n",
        "  plt.plot(val_loss)\n",
        "  plt.ylabel('loss')\n",
        "  plt.xlabel('epoch')\n",
        "  plt.legend(['train loss','test loss'], loc='upper left')\n",
        "  plt.show()\n",
        "\n",
        "def plotTrainingAccuracy(train_acc, val_acc):\n",
        "  plt.plot(train_acc)\n",
        "  plt.plot(val_acc)\n",
        "  plt.ylabel('accuracy')\n",
        "  plt.xlabel('epoch')\n",
        "  plt.legend(['train accuracy', 'test accuracy'], loc='upper left')\n",
        "  plt.show()\n",
        "\n",
        "\n",
        "def evaluateAccuracy(X_test, y_test, metric, model):\n",
        "  predictions = []\n",
        "  for x in X_test:\n",
        "    predArray = [0,0]\n",
        "    x = np.expand_dims(x, axis=0)\n",
        "    prediction = model.predict(x)\n",
        "    predArray[np.argmax(prediction[0])] = 1\n",
        "    predictions.append(predArray)\n",
        "\n",
        "  metric.update_state(y_test, predictions)\n",
        "  return m.result().numpy(), predictions\n",
        "\n",
        "def getPredictions(X_test, model):\n",
        "  predictions = []\n",
        "  for x in X_test:\n",
        "    predArray = [0,0]\n",
        "    x = np.expand_dims(x, axis=0)\n",
        "    prediction = model.predict(x)\n",
        "    predArray[np.argmax(prediction[0])] = 1\n",
        "    predictions.append(predArray)\n",
        "  \n",
        "  return predictions\n",
        "\n",
        "def getAllMetrics(y_test, predictions, verbose = False):\n",
        "  # simplifies splitting y_test into classes\n",
        "  y_test_argmax = np.array([])\n",
        "  for y in y_test:\n",
        "    y_test_argmax = np.append(y_test_argmax, np.argmax(y))\n",
        "\n",
        "  covid_indizes = np.where(y_test_argmax==1)[0]\n",
        "  if verbose == True:\n",
        "    print('COVID Samples: ' + str(len(covid_indizes)))\n",
        "  symptom_indizes = np.where(y_test_argmax==2)[0]\n",
        "  if verbose == True:\n",
        "    print('Symptomatic Samples: ' + str(len(symptom_indizes)))\n",
        "  healthy_indizes = np.where(y_test_argmax==0)[0]\n",
        "  if verbose == True:\n",
        "    print('Healthy Samples: ' + str(len(healthy_indizes)))\n",
        "\n",
        "  tp_c = tf.keras.metrics.TruePositives()\n",
        "  tn_c = tf.keras.metrics.TrueNegatives()\n",
        "  fp_c = tf.keras.metrics.FalsePositives()\n",
        "  fn_c = tf.keras.metrics.FalseNegatives()\n",
        "  tp_c.update_state(y_test[covid_indizes], np.array(predictions)[covid_indizes])\n",
        "  tn_c.update_state(y_test[covid_indizes], np.array(predictions)[covid_indizes])\n",
        "  fn_c.update_state(y_test[covid_indizes], np.array(predictions)[covid_indizes])\n",
        "  fp_c.update_state(y_test[covid_indizes], np.array(predictions)[covid_indizes])\n",
        "  if verbose == True:\n",
        "    print('True Positives for COVID Samples: ' + str(tp_c.result().numpy()))\n",
        "    print('True Negatives for COVID Samples: ' + str(tn_c.result().numpy()))\n",
        "    print('False Negatives for COVID Samples: ' + str(fn_c.result().numpy()))\n",
        "    print('False Positives for COVID Samples: ' + str(fp_c.result().numpy()))\n",
        "\n",
        "  tp_s = tf.keras.metrics.TruePositives()\n",
        "  tn_s = tf.keras.metrics.TrueNegatives()\n",
        "  fp_s = tf.keras.metrics.FalsePositives()\n",
        "  fn_s = tf.keras.metrics.FalseNegatives()\n",
        "  tp_s.update_state(y_test[symptom_indizes], np.array(predictions)[symptom_indizes])\n",
        "  tn_s.update_state(y_test[symptom_indizes], np.array(predictions)[symptom_indizes])\n",
        "  fn_s.update_state(y_test[symptom_indizes], np.array(predictions)[symptom_indizes])\n",
        "  fp_s.update_state(y_test[symptom_indizes], np.array(predictions)[symptom_indizes])\n",
        "  if verbose == True:\n",
        "    print('True Positives for symptomatic Samples: ' + str(tp_s.result().numpy()))\n",
        "    print('True Negatives for symptomatic Samples: ' + str(tn_s.result().numpy()))\n",
        "    print('False Negatives for symptomatic Samples: ' + str(fn_s.result().numpy()))\n",
        "    print('False Positives for symptomatic Samples: ' + str(fp_s.result().numpy()))\n",
        "\n",
        "  tp_h = tf.keras.metrics.TruePositives()\n",
        "  tn_h = tf.keras.metrics.TrueNegatives()\n",
        "  fp_h = tf.keras.metrics.FalsePositives()\n",
        "  fn_h = tf.keras.metrics.FalseNegatives()\n",
        "  tp_h.update_state(y_test[healthy_indizes], np.array(predictions)[healthy_indizes])\n",
        "  tn_h.update_state(y_test[healthy_indizes], np.array(predictions)[healthy_indizes])\n",
        "  fn_h.update_state(y_test[healthy_indizes], np.array(predictions)[healthy_indizes])\n",
        "  fp_h.update_state(y_test[healthy_indizes], np.array(predictions)[healthy_indizes])\n",
        "  if verbose == True:\n",
        "    print('True Positives for healthy samples: ' + str(tp_h.result().numpy()))\n",
        "    print('True Negatives for healthy samples: ' + str(tn_h.result().numpy()))\n",
        "    print('False Negatives for healthy samples: ' + str(fn_h.result().numpy()))\n",
        "    print('False Positives for healthy samples: ' + str(fp_h.result().numpy()))\n",
        "\n",
        "  sensitivity_c = tp_c.result().numpy() / (tp_c.result().numpy() + fn_c.result().numpy())\n",
        "  specificity_c = tn_c.result().numpy() / (tn_c.result().numpy() + fp_c.result().numpy())\n",
        "  if verbose == True:\n",
        "    print('Sensitivity for the covid class: ' + str(sensitivity_c))\n",
        "    print('Specificity for the covid class: ' + str(specificity_c))\n",
        "\n",
        "  sensitivity_s = tp_s.result().numpy() / (tp_s.result().numpy() + fn_s.result().numpy())\n",
        "  specificity_s = tn_s.result().numpy() / (tn_s.result().numpy() + fp_s.result().numpy())\n",
        "  if verbose == True:\n",
        "    print('Sensitivity for the symptom class: ' + str(sensitivity_s))\n",
        "    print('Specificity for the symptom class: ' + str(specificity_s))\n",
        "\n",
        "  sensitivity_h = tp_h.result().numpy() / (tp_h.result().numpy() + fn_h.result().numpy())\n",
        "  specificity_h= tn_h.result().numpy() / (tn_h.result().numpy() + fp_h.result().numpy())\n",
        "  if verbose == True:\n",
        "    print('Sensitivity for the healthy class: ' + str(sensitivity_h))\n",
        "    print('Specificity for the healthy class: ' + str(specificity_h))\n",
        "\n",
        "  recall_h = tf.keras.metrics.Recall(class_id=0)\n",
        "  recall_c = tf.keras.metrics.Recall(class_id=1)\n",
        "  recall_s = tf.keras.metrics.Recall(class_id=2)\n",
        "  recall_h.update_state(y_test, predictions)\n",
        "  recall_c.update_state(y_test, predictions)\n",
        "  recall_s.update_state(y_test, predictions)\n",
        "  if verbose == True:\n",
        "    print('Recall Healthy Samples: ' + str(recall_h.result().numpy()))\n",
        "    print('Recall COVID Samples: ' + str(recall_c.result().numpy()))\n",
        "    print('Recall Symptomatic Samples: ' + str(recall_s.result().numpy()))\n",
        "\n",
        "  precision_h = tf.keras.metrics.Precision(class_id=0)\n",
        "  precision_c = tf.keras.metrics.Precision(class_id=1)\n",
        "  precision_s = tf.keras.metrics.Precision(class_id=2)\n",
        "  precision_h.update_state(y_test, predictions)\n",
        "  precision_c.update_state(y_test, predictions)\n",
        "  precision_s.update_state(y_test, predictions)\n",
        "  if verbose == True:\n",
        "    print('Precision Healthy Samples: ' + str(precision_h.result().numpy()))\n",
        "    print('Precision COVID Samples: ' + str(precision_c.result().numpy()))\n",
        "    print('Precision Symptomatic Samples: ' + str(precision_s.result().numpy()))\n",
        "\n",
        "  f1_score_h = 2*((precision_h.result().numpy()*recall_h.result().numpy())/(precision_h.result().numpy()+recall_h.result().numpy()))\n",
        "  f1_score_c = 2*((precision_c.result().numpy()*recall_c.result().numpy())/(precision_c.result().numpy()+recall_c.result().numpy()))\n",
        "  f1_score_s = 2*((precision_s.result().numpy()*recall_s.result().numpy())/(precision_s.result().numpy()+recall_s.result().numpy()))\n",
        "\n",
        "  m = tf.keras.metrics.Accuracy()\n",
        "  m.update_state(y_test, predictions)\n",
        "  if verbose == True:\n",
        "    print(\"Accuracy: \" + str(m.result().numpy()))\n",
        "\n",
        "  metrics_dict = {'Class':['Overall','Healthy', 'COVID', 'Symptomatic']}\n",
        "  metrics_dict['F1 Score'] = ['-',f1_score_h*100, f1_score_c*100, f1_score_s*100]\n",
        "  metrics_dict['Sensitivity'] = ['-',sensitivity_h*100.0,sensitivity_c*100.0,sensitivity_s*100.0]\n",
        "  metrics_dict['Specificity'] = ['-',specificity_h*100.0,specificity_c*100.0,specificity_s*100.0]\n",
        "  metrics_dict['Precision'] = ['-',precision_h.result().numpy()*100,precision_c.result().numpy()*100,precision_s.result().numpy()*100]\n",
        "  metrics_dict['Recall'] = ['-',recall_h.result().numpy()*100,recall_c.result().numpy()*100,recall_s.result().numpy()*100]\n",
        "  metrics_dict['Accuracy'] = [m.result().numpy(), '-','-','-']\n",
        "\n",
        "  metrics_df = pd.DataFrame(data=metrics_dict)\n",
        "\n",
        "  return metrics_df\n",
        "\n",
        "def getAllMetricsForBC(y_test, predictions, verbose = False):\n",
        "  # simplifies splitting y_test into classes\n",
        "  y_test_argmax = np.array([])\n",
        "  for y in y_test:\n",
        "    y_test_argmax = np.append(y_test_argmax, np.argmax(y))\n",
        "\n",
        "  covid_indizes = np.where(y_test_argmax==1)[0]\n",
        "  if verbose == True:\n",
        "    print('COVID Samples: ' + str(len(covid_indizes)))\n",
        "  healthy_indizes = np.where(y_test_argmax==0)[0]\n",
        "  if verbose == True:\n",
        "    print('No COVID Samples: ' + str(len(healthy_indizes)))\n",
        "\n",
        "  tp_c = tf.keras.metrics.TruePositives()\n",
        "  tn_c = tf.keras.metrics.TrueNegatives()\n",
        "  fp_c = tf.keras.metrics.FalsePositives()\n",
        "  fn_c = tf.keras.metrics.FalseNegatives()\n",
        "  tp_c.update_state(y_test[covid_indizes], np.array(predictions)[covid_indizes])\n",
        "  tn_c.update_state(y_test[covid_indizes], np.array(predictions)[covid_indizes])\n",
        "  fn_c.update_state(y_test[covid_indizes], np.array(predictions)[covid_indizes])\n",
        "  fp_c.update_state(y_test[covid_indizes], np.array(predictions)[covid_indizes])\n",
        "  if verbose == True:\n",
        "    print('True Positives for COVID Samples: ' + str(tp_c.result().numpy()))\n",
        "    print('True Negatives for COVID Samples: ' + str(tn_c.result().numpy()))\n",
        "    print('False Negatives for COVID Samples: ' + str(fn_c.result().numpy()))\n",
        "    print('False Positives for COVID Samples: ' + str(fp_c.result().numpy()))\n",
        "\n",
        "  tp_h = tf.keras.metrics.TruePositives()\n",
        "  tn_h = tf.keras.metrics.TrueNegatives()\n",
        "  fp_h = tf.keras.metrics.FalsePositives()\n",
        "  fn_h = tf.keras.metrics.FalseNegatives()\n",
        "  tp_h.update_state(y_test[healthy_indizes], np.array(predictions)[healthy_indizes])\n",
        "  tn_h.update_state(y_test[healthy_indizes], np.array(predictions)[healthy_indizes])\n",
        "  fn_h.update_state(y_test[healthy_indizes], np.array(predictions)[healthy_indizes])\n",
        "  fp_h.update_state(y_test[healthy_indizes], np.array(predictions)[healthy_indizes])\n",
        "  if verbose == True:\n",
        "    print('True Positives for healthy samples: ' + str(tp_h.result().numpy()))\n",
        "    print('True Negatives for healthy samples: ' + str(tn_h.result().numpy()))\n",
        "    print('False Negatives for healthy samples: ' + str(fn_h.result().numpy()))\n",
        "    print('False Positives for healthy samples: ' + str(fp_h.result().numpy()))\n",
        "\n",
        "  sensitivity_c = tp_c.result().numpy() / (tp_c.result().numpy() + fn_c.result().numpy())\n",
        "  specificity_c = tn_c.result().numpy() / (tn_c.result().numpy() + fp_c.result().numpy())\n",
        "  if verbose == True:\n",
        "    print('Sensitivity for the covid class: ' + str(sensitivity_c))\n",
        "    print('Specificity for the covid class: ' + str(specificity_c))\n",
        "\n",
        "  sensitivity_h = tp_h.result().numpy() / (tp_h.result().numpy() + fn_h.result().numpy())\n",
        "  specificity_h= tn_h.result().numpy() / (tn_h.result().numpy() + fp_h.result().numpy())\n",
        "  if verbose == True:\n",
        "    print('Sensitivity for the healthy class: ' + str(sensitivity_h))\n",
        "    print('Specificity for the healthy class: ' + str(specificity_h))\n",
        "\n",
        "  recall_h = tf.keras.metrics.Recall(class_id=0)\n",
        "  recall_c = tf.keras.metrics.Recall(class_id=1)\n",
        "  recall_h.update_state(y_test, predictions)\n",
        "  recall_c.update_state(y_test, predictions)\n",
        "  if verbose == True:\n",
        "    print('Recall Healthy Samples: ' + str(recall_h.result().numpy()))\n",
        "    print('Recall COVID Samples: ' + str(recall_c.result().numpy()))\n",
        "\n",
        "  precision_h = tf.keras.metrics.Precision(class_id=0)\n",
        "  precision_c = tf.keras.metrics.Precision(class_id=1)\n",
        "  precision_h.update_state(y_test, predictions)\n",
        "  precision_c.update_state(y_test, predictions)\n",
        "  if verbose == True:\n",
        "    print('Precision Healthy Samples: ' + str(precision_h.result().numpy()))\n",
        "    print('Precision COVID Samples: ' + str(precision_c.result().numpy()))\n",
        "\n",
        "  f1_score_h = 2*((precision_h.result().numpy()*recall_h.result().numpy())/(precision_h.result().numpy()+recall_h.result().numpy()))\n",
        "  f1_score_c = 2*((precision_c.result().numpy()*recall_c.result().numpy())/(precision_c.result().numpy()+recall_c.result().numpy()))\n",
        "\n",
        "  m = tf.keras.metrics.Accuracy()\n",
        "  m.update_state(y_test, predictions)\n",
        "  if verbose == True:\n",
        "    print(\"Accuracy: \" + str(m.result().numpy()))\n",
        "\n",
        "  metrics_dict = {'Class':['Overall','Healthy', 'COVID']}\n",
        "  metrics_dict['F1 Score'] = ['-',f1_score_h*100, f1_score_c*100]\n",
        "  metrics_dict['Sensitivity'] = ['-',sensitivity_h*100.0,sensitivity_c*100.0]\n",
        "  metrics_dict['Specificity'] = ['-',specificity_h*100.0,specificity_c*100.0]\n",
        "  metrics_dict['Precision'] = ['-',precision_h.result().numpy()*100,precision_c.result().numpy()*100]\n",
        "  metrics_dict['Recall'] = ['-',recall_h.result().numpy()*100,recall_c.result().numpy()*100]\n",
        "  metrics_dict['Accuracy'] = [m.result().numpy(), '-','-']\n",
        "\n",
        "  metrics_df = pd.DataFrame(data=metrics_dict)\n",
        "\n",
        "  return metrics_df\n",
        "  "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-aEcQFGkqRmB"
      },
      "source": [
        "#Evaluation on segmented data"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sM2JryCNPCD5"
      },
      "source": [
        "Load the melspecs of segmented data from drive"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "LHJthkBYMchL"
      },
      "source": [
        "dataSegmented = np.load('./drive/MyDrive/COUGHVID/melSpecsDataSegmented.npy', allow_pickle=True)\n",
        "labels = np.load('./drive/MyDrive/COUGHVID/melSpecsLabelsSegmented.npy', allow_pickle=True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "OKZMSgI3M_Zm",
        "outputId": "e13d03b7-a8a6-40af-9e86-f376188fdc73"
      },
      "source": [
        "print(dataSegmented.shape)\n",
        "print(labels.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "(24374, 240, 320)\n",
            "(24374,)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sIEWX8lMNQuO"
      },
      "source": [
        "##Evaluation with class weights"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8GQiU3-oZ82Y"
      },
      "source": [
        "###Evaluation of the multi class clasifier\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3lIkPl4rZ-Ch"
      },
      "source": [
        "labels_new_mc = oneHotEncoding(labels)\n",
        "data_new = np.expand_dims(dataSegmented, axis=3)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "114gK2QuuRt6",
        "outputId": "2afaeb4f-91b0-4c3c-85f9-25a80fb5a2f2"
      },
      "source": [
        "print(labels_new_mc.shape)\n",
        "print(data_new.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "(24374, 3)\n",
            "(24374, 240, 320, 1)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "gEv3FCswXYjc",
        "outputId": "97844eb5-5712-45ec-d5fc-2b6fdc879499"
      },
      "source": [
        "# split data for train and test validation\n",
        "X_train, X_test, y_train, y_test = train_test_split(data_new, labels_new_mc, test_size=0.2, random_state=0)\n",
        "\n",
        "\n",
        "## calculate class weights\n",
        "y_integers = np.argmax(y_train, axis=1)\n",
        "class_weights = compute_class_weight('balanced', np.unique(y_integers), y_integers)\n",
        "class_weights_dict = dict(enumerate(class_weights))\n",
        "\n",
        "class_weights_dict"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "text/plain": [
              "{0: 0.4392853924484095, 1: 4.075026123301986, 2: 2.0912698412698414}"
            ]
          },
          "execution_count": 12,
          "metadata": {},
          "output_type": "execute_result"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nRAl_G7Wuabv"
      },
      "source": [
        "lr_schedule = tf.keras.optimizers.schedules.ExponentialDecay(\n",
        "    initial_learning_rate=0.00001,\n",
        "    decay_steps=10000,\n",
        "    decay_rate=0.9)\n",
        "\n",
        "cb = tf.keras.callbacks.EarlyStopping(monitor='val_loss', min_delta=0.01, patience=25, verbose=1,\n",
        "    mode='auto', baseline=None, restore_best_weights=True)\n",
        "\n",
        "dtlmc = DTLMC()\n",
        "opt = tf.keras.optimizers.Adam(learning_rate=lr_schedule)\n",
        "dtlmc.compile(optimizer=opt,loss='categorical_crossentropy',metrics=['acc'])\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "JsYJEXl6toon",
        "outputId": "229b026b-202d-4738-de02-3c8943e0a374"
      },
      "source": [
        "history = dtlmc.fit(X_train,y_train, epochs=250, validation_data=(X_test,y_test), class_weight=class_weights_dict, callbacks=[cb])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/250\n",
            "610/610 [==============================] - 39s 33ms/step - loss: 5.3400 - acc: 0.3390 - val_loss: 1.6471 - val_acc: 0.1104\n",
            "Epoch 2/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 1.5544 - acc: 0.3440 - val_loss: 1.0631 - val_acc: 0.4123\n",
            "Epoch 3/250\n",
            "610/610 [==============================] - 17s 29ms/step - loss: 1.1431 - acc: 0.3335 - val_loss: 1.0831 - val_acc: 0.3342\n",
            "Epoch 4/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 1.1039 - acc: 0.3499 - val_loss: 1.0921 - val_acc: 0.3545\n",
            "Epoch 5/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 1.1004 - acc: 0.3324 - val_loss: 1.0841 - val_acc: 0.3844\n",
            "Epoch 6/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 1.0905 - acc: 0.3485 - val_loss: 1.0924 - val_acc: 0.4072\n",
            "Epoch 7/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 1.0875 - acc: 0.3605 - val_loss: 1.0888 - val_acc: 0.3497\n",
            "Epoch 8/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 1.0834 - acc: 0.3666 - val_loss: 1.0870 - val_acc: 0.3703\n",
            "Epoch 9/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 1.0758 - acc: 0.3797 - val_loss: 1.0878 - val_acc: 0.3614\n",
            "Epoch 10/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 1.0724 - acc: 0.3860 - val_loss: 1.0747 - val_acc: 0.4193\n",
            "Epoch 11/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 1.0638 - acc: 0.3928 - val_loss: 1.0749 - val_acc: 0.4033\n",
            "Epoch 12/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 1.0656 - acc: 0.3959 - val_loss: 1.0848 - val_acc: 0.3674\n",
            "Epoch 13/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 1.0555 - acc: 0.4035 - val_loss: 1.0943 - val_acc: 0.3268\n",
            "Epoch 14/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 1.0553 - acc: 0.4054 - val_loss: 1.0684 - val_acc: 0.3998\n",
            "Epoch 15/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 1.0487 - acc: 0.4099 - val_loss: 1.0718 - val_acc: 0.3797\n",
            "Epoch 16/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 1.0420 - acc: 0.4155 - val_loss: 1.0785 - val_acc: 0.3713\n",
            "Epoch 17/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 1.0354 - acc: 0.4126 - val_loss: 1.0721 - val_acc: 0.3360\n",
            "Epoch 18/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 1.0291 - acc: 0.4285 - val_loss: 1.0907 - val_acc: 0.3264\n",
            "Epoch 19/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 1.0225 - acc: 0.4309 - val_loss: 1.0854 - val_acc: 0.3317\n",
            "Epoch 20/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 1.0139 - acc: 0.4379 - val_loss: 1.0854 - val_acc: 0.3612\n",
            "Epoch 21/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 1.0056 - acc: 0.4483 - val_loss: 1.0872 - val_acc: 0.3629\n",
            "Epoch 22/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.9948 - acc: 0.4569 - val_loss: 1.0807 - val_acc: 0.3645\n",
            "Epoch 23/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.9820 - acc: 0.4620 - val_loss: 1.0759 - val_acc: 0.3848\n",
            "Epoch 24/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.9739 - acc: 0.4715 - val_loss: 1.0172 - val_acc: 0.4574\n",
            "Epoch 25/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.9651 - acc: 0.4763 - val_loss: 1.0772 - val_acc: 0.3754\n",
            "Epoch 26/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.9586 - acc: 0.4817 - val_loss: 1.0845 - val_acc: 0.3865\n",
            "Epoch 27/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.9435 - acc: 0.4908 - val_loss: 1.0487 - val_acc: 0.4244\n",
            "Epoch 28/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.9373 - acc: 0.4926 - val_loss: 1.0293 - val_acc: 0.4406\n",
            "Epoch 29/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.9235 - acc: 0.5032 - val_loss: 1.0761 - val_acc: 0.3865\n",
            "Epoch 30/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.9133 - acc: 0.5108 - val_loss: 1.0730 - val_acc: 0.4049\n",
            "Epoch 31/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.9092 - acc: 0.5107 - val_loss: 1.0210 - val_acc: 0.4505\n",
            "Epoch 32/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.8936 - acc: 0.5275 - val_loss: 1.0964 - val_acc: 0.3838\n",
            "Epoch 33/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.8868 - acc: 0.5283 - val_loss: 1.0670 - val_acc: 0.4146\n",
            "Epoch 34/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.8788 - acc: 0.5313 - val_loss: 1.0353 - val_acc: 0.4219\n",
            "Epoch 35/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.8669 - acc: 0.5380 - val_loss: 1.0253 - val_acc: 0.4615\n",
            "Epoch 36/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.8624 - acc: 0.5392 - val_loss: 1.0578 - val_acc: 0.4244\n",
            "Epoch 37/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.8492 - acc: 0.5485 - val_loss: 1.0832 - val_acc: 0.4043\n",
            "Epoch 38/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.8463 - acc: 0.5542 - val_loss: 1.0545 - val_acc: 0.4244\n",
            "Epoch 39/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.8327 - acc: 0.5533 - val_loss: 1.0740 - val_acc: 0.4254\n",
            "Epoch 40/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.8296 - acc: 0.5500 - val_loss: 1.0393 - val_acc: 0.4496\n",
            "Epoch 41/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.8184 - acc: 0.5616 - val_loss: 1.0418 - val_acc: 0.4501\n",
            "Epoch 42/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.8157 - acc: 0.5711 - val_loss: 1.0299 - val_acc: 0.4611\n",
            "Epoch 43/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.8062 - acc: 0.5726 - val_loss: 1.0435 - val_acc: 0.4552\n",
            "Epoch 44/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.7981 - acc: 0.5704 - val_loss: 1.0141 - val_acc: 0.4595\n",
            "Epoch 45/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.7982 - acc: 0.5778 - val_loss: 1.0416 - val_acc: 0.4480\n",
            "Epoch 46/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.7852 - acc: 0.5891 - val_loss: 1.0358 - val_acc: 0.4615\n",
            "Epoch 47/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.7805 - acc: 0.5880 - val_loss: 1.0634 - val_acc: 0.4336\n",
            "Epoch 48/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.7725 - acc: 0.5856 - val_loss: 1.0247 - val_acc: 0.4782\n",
            "Epoch 49/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.7690 - acc: 0.5892 - val_loss: 1.0198 - val_acc: 0.4773\n",
            "Restoring model weights from the end of the best epoch.\n",
            "Epoch 00049: early stopping\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "XhVUGMb7a5fD",
        "outputId": "7b1d1414-7118-494f-a4ae-3582c505a5c4"
      },
      "source": [
        "dtlmc.save('./drive/MyDrive/COUGHVID/Models/dtlmc', save_format=\"tf\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Assets written to: ./drive/MyDrive/COUGHVID/Models/dtlmc/assets\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0Yulq0QbxNaK"
      },
      "source": [
        "predictions = []\n",
        "for x in X_test:\n",
        "  predArray = [0,0,0]\n",
        "  x = np.expand_dims(x, axis=0)\n",
        "  prediction = dtlmc.predict(x)\n",
        "  predArray[np.argmax(prediction[0])] = 1\n",
        "  predictions.append(predArray)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 175
        },
        "id": "P2sq6OQlShUa",
        "outputId": "5569e95c-fe08-43aa-c04c-ed0f7c29cfc3"
      },
      "source": [
        "df = getAllMetrics(y_test, predictions)\n",
        "df"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Class</th>\n",
              "      <th>F1 Score</th>\n",
              "      <th>Sensitivity</th>\n",
              "      <th>Specificity</th>\n",
              "      <th>Precision</th>\n",
              "      <th>Recall</th>\n",
              "      <th>Accuracy</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>Overall</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>0.638291</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>Healthy</td>\n",
              "      <td>58.502</td>\n",
              "      <td>45.8266</td>\n",
              "      <td>72.9133</td>\n",
              "      <td>80.8704</td>\n",
              "      <td>45.8266</td>\n",
              "      <td>-</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>COVID</td>\n",
              "      <td>19.1436</td>\n",
              "      <td>29.3059</td>\n",
              "      <td>64.653</td>\n",
              "      <td>14.2145</td>\n",
              "      <td>29.3059</td>\n",
              "      <td>-</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>Symptomatic</td>\n",
              "      <td>30.5976</td>\n",
              "      <td>53.392</td>\n",
              "      <td>76.696</td>\n",
              "      <td>21.443</td>\n",
              "      <td>53.392</td>\n",
              "      <td>-</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "         Class F1 Score Sensitivity Specificity Precision   Recall  Accuracy\n",
              "0      Overall        -           -           -         -        -  0.638291\n",
              "1      Healthy   58.502     45.8266     72.9133   80.8704  45.8266         -\n",
              "2        COVID  19.1436     29.3059      64.653   14.2145  29.3059         -\n",
              "3  Symptomatic  30.5976      53.392      76.696    21.443   53.392         -"
            ]
          },
          "execution_count": 17,
          "metadata": {},
          "output_type": "execute_result"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "J1HZRrkuadCz"
      },
      "source": [
        "matrix = confusion_matrix(np.argmax(y_test, axis=1), np.argmax(predictions, axis=1))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7D6UjvAZz0fB"
      },
      "source": [
        "matrix = matrix/matrix.sum(axis=1)[:, np.newaxis]"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "CeDtcawQadr4",
        "outputId": "babf651d-7c03-496a-e699-807c712b29f6"
      },
      "source": [
        "matrix"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "text/plain": [
              "array([[0.45826558, 0.15338753, 0.38834688],\n",
              "       [0.38817481, 0.29305913, 0.31876607],\n",
              "       [0.31281407, 0.15326633, 0.5339196 ]])"
            ]
          },
          "execution_count": 20,
          "metadata": {},
          "output_type": "execute_result"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 428
        },
        "id": "3wz-uy9P6b_R",
        "outputId": "f050b6c3-7fa9-4804-8ee2-35f32d4fa2b8"
      },
      "source": [
        "labellist = ['healthy','covid', 'symptomatic']\n",
        "plotConfusionMatrix(labellist, matrix)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x504 with 2 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "id": "yQqBvnKptysO",
        "outputId": "806eceba-0e23-4ea4-d834-1c2d1061c012"
      },
      "source": [
        "plotTrainingLoss(history.history['loss'], history.history['val_loss'])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "id": "rFA3NfcKt4Zm",
        "outputId": "af514932-6b90-416e-8e4e-a61064b17d31"
      },
      "source": [
        "plotTrainingAccuracy(history.history['acc'], history.history['val_acc'])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EiicnAL8ZxWx"
      },
      "source": [
        "###Evaluation of the binary class classifier"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "N9gHG4UTZ0pN",
        "outputId": "f95704b4-1b64-4c15-dfb4-7521ad733f76"
      },
      "source": [
        "labels_new_bc = oneHotEncodingForBC(labels)\n",
        "data_new_bc = np.expand_dims(dataSegmented, axis=3)\n",
        "print(data_new_bc.shape)\n",
        "print(labels_new_bc.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "(24374, 240, 320, 1)\n",
            "(24374, 2)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zwjKm2NhaaB5"
      },
      "source": [
        "X_train, X_test, y_train, y_test = train_test_split(data_new_bc, labels_new_bc, test_size=0.2, random_state=0)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ArJ_g-UfBjKE",
        "outputId": "4fd7ca27-7820-40b5-d330-8f6f479905e0"
      },
      "source": [
        "## calculate class weights\n",
        "y_integers = np.argmax(y_train, axis=1)\n",
        "class_weights_bc = compute_class_weight('balanced', np.unique(y_integers), y_integers)\n",
        "class_weights_dict_bc = dict(enumerate(class_weights_bc))\n",
        "\n",
        "class_weights_dict_bc"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "text/plain": [
              "{0: 0.5445431188561215, 1: 6.112539184952978}"
            ]
          },
          "execution_count": 11,
          "metadata": {},
          "output_type": "execute_result"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "c8GBJbGwfF5g"
      },
      "source": [
        "lr_schedule = tf.keras.optimizers.schedules.ExponentialDecay(\n",
        "    initial_learning_rate=0.00001,\n",
        "    decay_steps=10000,\n",
        "    decay_rate=0.9)\n",
        "\n",
        "cb = tf.keras.callbacks.EarlyStopping(monitor='val_loss', min_delta=0.01, patience=25, verbose=1,\n",
        "    mode='auto', baseline=None, restore_best_weights=True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "stcWFfhmanE1"
      },
      "source": [
        "dtlbc = DTLBC()\n",
        "opt = tf.keras.optimizers.Adam(learning_rate=lr_schedule)\n",
        "dtlbc.compile(optimizer=opt,loss='categorical_crossentropy',metrics=['acc'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "kPSClLjuaq9C",
        "outputId": "8ed9af82-e683-44d2-ab83-065175a4a1ad"
      },
      "source": [
        "history = dtlbc.fit(X_train,y_train, epochs=250, validation_data=(X_test,y_test), class_weight=class_weights_dict_bc, callbacks=[cb])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/250\n",
            "610/610 [==============================] - 38s 33ms/step - loss: 2.0809 - acc: 0.5060 - val_loss: 1.1129 - val_acc: 0.1327\n",
            "Epoch 2/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.9978 - acc: 0.5106 - val_loss: 0.6678 - val_acc: 0.6111\n",
            "Epoch 3/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.8083 - acc: 0.5135 - val_loss: 0.6152 - val_acc: 0.7319\n",
            "Epoch 4/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.7475 - acc: 0.5337 - val_loss: 0.7794 - val_acc: 0.3536\n",
            "Epoch 5/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.7164 - acc: 0.5366 - val_loss: 0.7367 - val_acc: 0.4238\n",
            "Epoch 6/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.7033 - acc: 0.5458 - val_loss: 0.5642 - val_acc: 0.8328\n",
            "Epoch 7/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.6899 - acc: 0.5599 - val_loss: 0.7438 - val_acc: 0.4059\n",
            "Epoch 8/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.6823 - acc: 0.5718 - val_loss: 0.6927 - val_acc: 0.5387\n",
            "Epoch 9/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.6653 - acc: 0.5911 - val_loss: 0.6180 - val_acc: 0.7028\n",
            "Epoch 10/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.6601 - acc: 0.6018 - val_loss: 0.6286 - val_acc: 0.6755\n",
            "Epoch 11/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.6502 - acc: 0.6179 - val_loss: 0.7010 - val_acc: 0.5157\n",
            "Epoch 12/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.6404 - acc: 0.6260 - val_loss: 0.6928 - val_acc: 0.5333\n",
            "Epoch 13/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.6308 - acc: 0.6394 - val_loss: 0.6082 - val_acc: 0.6905\n",
            "Epoch 14/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.6286 - acc: 0.6405 - val_loss: 0.7554 - val_acc: 0.4379\n",
            "Epoch 15/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.6205 - acc: 0.6461 - val_loss: 0.6442 - val_acc: 0.6261\n",
            "Epoch 16/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.6124 - acc: 0.6592 - val_loss: 0.6708 - val_acc: 0.5739\n",
            "Epoch 17/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.5978 - acc: 0.6731 - val_loss: 0.6195 - val_acc: 0.6556\n",
            "Epoch 18/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.5977 - acc: 0.6680 - val_loss: 0.5684 - val_acc: 0.7298\n",
            "Epoch 19/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.5879 - acc: 0.6869 - val_loss: 0.7069 - val_acc: 0.5374\n",
            "Epoch 20/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.5808 - acc: 0.6851 - val_loss: 0.6068 - val_acc: 0.6763\n",
            "Epoch 21/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.5698 - acc: 0.6965 - val_loss: 0.6176 - val_acc: 0.6544\n",
            "Epoch 22/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.5639 - acc: 0.7005 - val_loss: 0.5904 - val_acc: 0.6896\n",
            "Epoch 23/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.5580 - acc: 0.7065 - val_loss: 0.5945 - val_acc: 0.6872\n",
            "Epoch 24/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.5567 - acc: 0.7101 - val_loss: 0.5421 - val_acc: 0.7467\n",
            "Epoch 25/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.5493 - acc: 0.7104 - val_loss: 0.6601 - val_acc: 0.6068\n",
            "Epoch 26/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.5427 - acc: 0.7210 - val_loss: 0.5985 - val_acc: 0.6761\n",
            "Epoch 27/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.5368 - acc: 0.7278 - val_loss: 0.6704 - val_acc: 0.5973\n",
            "Epoch 28/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.5239 - acc: 0.7333 - val_loss: 0.5428 - val_acc: 0.7274\n",
            "Epoch 29/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.5276 - acc: 0.7296 - val_loss: 0.5980 - val_acc: 0.6804\n",
            "Epoch 30/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.5158 - acc: 0.7364 - val_loss: 0.5699 - val_acc: 0.7030\n",
            "Epoch 31/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.5134 - acc: 0.7432 - val_loss: 0.5644 - val_acc: 0.7087\n",
            "Epoch 32/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.5070 - acc: 0.7452 - val_loss: 0.4893 - val_acc: 0.7809\n",
            "Epoch 33/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4987 - acc: 0.7518 - val_loss: 0.5517 - val_acc: 0.7194\n",
            "Epoch 34/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4826 - acc: 0.7657 - val_loss: 0.5530 - val_acc: 0.7145\n",
            "Epoch 35/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4868 - acc: 0.7581 - val_loss: 0.6149 - val_acc: 0.6599\n",
            "Epoch 36/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4774 - acc: 0.7635 - val_loss: 0.5594 - val_acc: 0.7108\n",
            "Epoch 37/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4778 - acc: 0.7622 - val_loss: 0.5242 - val_acc: 0.7440\n",
            "Epoch 38/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4765 - acc: 0.7660 - val_loss: 0.5686 - val_acc: 0.6993\n",
            "Epoch 39/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4671 - acc: 0.7708 - val_loss: 0.5647 - val_acc: 0.7034\n",
            "Epoch 40/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4617 - acc: 0.7718 - val_loss: 0.5756 - val_acc: 0.6919\n",
            "Epoch 41/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4547 - acc: 0.7798 - val_loss: 0.5666 - val_acc: 0.7019\n",
            "Epoch 42/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4555 - acc: 0.7746 - val_loss: 0.5357 - val_acc: 0.7321\n",
            "Epoch 43/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4437 - acc: 0.7857 - val_loss: 0.5499 - val_acc: 0.7204\n",
            "Epoch 44/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4508 - acc: 0.7790 - val_loss: 0.6024 - val_acc: 0.6738\n",
            "Epoch 45/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4426 - acc: 0.7796 - val_loss: 0.5840 - val_acc: 0.6872\n",
            "Epoch 46/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4410 - acc: 0.7857 - val_loss: 0.4997 - val_acc: 0.7621\n",
            "Epoch 47/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4425 - acc: 0.7853 - val_loss: 0.4950 - val_acc: 0.7664\n",
            "Epoch 48/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4311 - acc: 0.7939 - val_loss: 0.4723 - val_acc: 0.7869\n",
            "Epoch 49/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4283 - acc: 0.7946 - val_loss: 0.5083 - val_acc: 0.7491\n",
            "Epoch 50/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4221 - acc: 0.7955 - val_loss: 0.5140 - val_acc: 0.7471\n",
            "Epoch 51/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4152 - acc: 0.8025 - val_loss: 0.4900 - val_acc: 0.7737\n",
            "Epoch 52/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4108 - acc: 0.7997 - val_loss: 0.4736 - val_acc: 0.7811\n",
            "Epoch 53/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4100 - acc: 0.8011 - val_loss: 0.5478 - val_acc: 0.7270\n",
            "Epoch 54/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4118 - acc: 0.8008 - val_loss: 0.5357 - val_acc: 0.7333\n",
            "Epoch 55/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4046 - acc: 0.8079 - val_loss: 0.5563 - val_acc: 0.7108\n",
            "Epoch 56/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4003 - acc: 0.8074 - val_loss: 0.4825 - val_acc: 0.7774\n",
            "Epoch 57/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4005 - acc: 0.8098 - val_loss: 0.4985 - val_acc: 0.7606\n",
            "Epoch 58/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.4011 - acc: 0.8098 - val_loss: 0.5144 - val_acc: 0.7508\n",
            "Epoch 59/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3882 - acc: 0.8135 - val_loss: 0.4606 - val_acc: 0.7922\n",
            "Epoch 60/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3891 - acc: 0.8129 - val_loss: 0.5011 - val_acc: 0.7594\n",
            "Epoch 61/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3830 - acc: 0.8185 - val_loss: 0.5867 - val_acc: 0.6976\n",
            "Epoch 62/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3826 - acc: 0.8183 - val_loss: 0.4947 - val_acc: 0.7655\n",
            "Epoch 63/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3825 - acc: 0.8197 - val_loss: 0.4660 - val_acc: 0.7879\n",
            "Epoch 64/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3756 - acc: 0.8206 - val_loss: 0.5597 - val_acc: 0.7175\n",
            "Epoch 65/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3687 - acc: 0.8312 - val_loss: 0.5335 - val_acc: 0.7358\n",
            "Epoch 66/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3639 - acc: 0.8273 - val_loss: 0.4663 - val_acc: 0.7883\n",
            "Epoch 67/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3674 - acc: 0.8247 - val_loss: 0.5210 - val_acc: 0.7487\n",
            "Epoch 68/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3660 - acc: 0.8282 - val_loss: 0.4372 - val_acc: 0.8139\n",
            "Epoch 69/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3638 - acc: 0.8294 - val_loss: 0.5200 - val_acc: 0.7495\n",
            "Epoch 70/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3568 - acc: 0.8349 - val_loss: 0.5536 - val_acc: 0.7276\n",
            "Epoch 71/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3589 - acc: 0.8294 - val_loss: 0.4381 - val_acc: 0.8121\n",
            "Epoch 72/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3515 - acc: 0.8362 - val_loss: 0.5351 - val_acc: 0.7446\n",
            "Epoch 73/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3458 - acc: 0.8375 - val_loss: 0.5172 - val_acc: 0.7545\n",
            "Epoch 74/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3480 - acc: 0.8380 - val_loss: 0.5048 - val_acc: 0.7655\n",
            "Epoch 75/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3477 - acc: 0.8382 - val_loss: 0.5726 - val_acc: 0.7163\n",
            "Epoch 76/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3343 - acc: 0.8452 - val_loss: 0.4869 - val_acc: 0.7774\n",
            "Epoch 77/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3347 - acc: 0.8455 - val_loss: 0.4407 - val_acc: 0.8121\n",
            "Epoch 78/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3394 - acc: 0.8376 - val_loss: 0.4423 - val_acc: 0.8119\n",
            "Epoch 79/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3385 - acc: 0.8466 - val_loss: 0.4883 - val_acc: 0.7766\n",
            "Epoch 80/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3337 - acc: 0.8445 - val_loss: 0.4344 - val_acc: 0.8191\n",
            "Epoch 81/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3289 - acc: 0.8475 - val_loss: 0.5470 - val_acc: 0.7387\n",
            "Epoch 82/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3301 - acc: 0.8484 - val_loss: 0.5047 - val_acc: 0.7684\n",
            "Epoch 83/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3287 - acc: 0.8476 - val_loss: 0.5254 - val_acc: 0.7516\n",
            "Epoch 84/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3257 - acc: 0.8527 - val_loss: 0.4443 - val_acc: 0.8133\n",
            "Epoch 85/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3236 - acc: 0.8482 - val_loss: 0.4302 - val_acc: 0.8275\n",
            "Epoch 86/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3202 - acc: 0.8522 - val_loss: 0.5034 - val_acc: 0.7684\n",
            "Epoch 87/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3137 - acc: 0.8551 - val_loss: 0.5728 - val_acc: 0.7184\n",
            "Epoch 88/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3166 - acc: 0.8553 - val_loss: 0.4859 - val_acc: 0.7822\n",
            "Epoch 89/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3124 - acc: 0.8555 - val_loss: 0.4700 - val_acc: 0.7953\n",
            "Epoch 90/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3158 - acc: 0.8539 - val_loss: 0.4365 - val_acc: 0.8209\n",
            "Epoch 91/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3085 - acc: 0.8601 - val_loss: 0.5025 - val_acc: 0.7717\n",
            "Epoch 92/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.2969 - acc: 0.8633 - val_loss: 0.4756 - val_acc: 0.7912\n",
            "Epoch 93/250\n",
            "610/610 [==============================] - 18s 29ms/step - loss: 0.3007 - acc: 0.8618 - val_loss: 0.4700 - val_acc: 0.7971\n",
            "Restoring model weights from the end of the best epoch.\n",
            "Epoch 00093: early stopping\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "glbNpB1-eaAL",
        "outputId": "7f279efd-d6e6-4216-cbb5-de04045830bd"
      },
      "source": [
        "dtlbc.save('./drive/MyDrive/COUGHVID/Models/dtlbc', save_format=\"tf\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Assets written to: ./drive/MyDrive/COUGHVID/Models/dtlbc/assets\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jEswg5HFpokM"
      },
      "source": [
        "predictions = getPredictions(X_test, dtlbc)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 143
        },
        "id": "ecavkV0fb70m",
        "outputId": "c01e5eaa-2bd1-4e9e-e751-c8ff00f672d3"
      },
      "source": [
        "results = getAllMetricsForBC(y_test, predictions)\n",
        "results"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Class</th>\n",
              "      <th>F1 Score</th>\n",
              "      <th>Sensitivity</th>\n",
              "      <th>Specificity</th>\n",
              "      <th>Precision</th>\n",
              "      <th>Recall</th>\n",
              "      <th>Accuracy</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>Overall</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>0.813949</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>Healthy</td>\n",
              "      <td>89.5447</td>\n",
              "      <td>86.5805</td>\n",
              "      <td>86.5805</td>\n",
              "      <td>92.719</td>\n",
              "      <td>86.5805</td>\n",
              "      <td>-</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>COVID</td>\n",
              "      <td>15.6279</td>\n",
              "      <td>21.5938</td>\n",
              "      <td>21.5938</td>\n",
              "      <td>12.2449</td>\n",
              "      <td>21.5938</td>\n",
              "      <td>-</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "     Class F1 Score Sensitivity Specificity Precision   Recall  Accuracy\n",
              "0  Overall        -           -           -         -        -  0.813949\n",
              "1  Healthy  89.5447     86.5805     86.5805    92.719  86.5805         -\n",
              "2    COVID  15.6279     21.5938     21.5938   12.2449  21.5938         -"
            ]
          },
          "execution_count": 16,
          "metadata": {},
          "output_type": "execute_result"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 428
        },
        "id": "zdTC4Kxvsu6W",
        "outputId": "dce4fb31-0b86-47d4-f2ee-726132ff6646"
      },
      "source": [
        "matrix = confusion_matrix(np.argmax(y_test, axis=1), np.argmax(predictions, axis=1), normalize='true')\n",
        "labellist = ['No-COVID', 'COVID']\n",
        "plotConfusionMatrix(labellist, matrix)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x504 with 2 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "id": "KS4WIWAQt7hh",
        "outputId": "9d379006-764f-4aa4-b87b-d9b32661e886"
      },
      "source": [
        "plotTrainingLoss(history.history['loss'], history.history['val_loss'])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "id": "s90jj9-Cr4xc",
        "outputId": "2bce0843-190c-4bcf-859b-6ab7f0b6b336"
      },
      "source": [
        "plotTrainingAccuracy(history.history['acc'], history.history['val_acc'])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEGCAYAAABo25JHAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOydd3zU5f3A30/23iSMBALI3lMUUBzUgaKiuG3VKrbOVuuv1lql1lar1lqtWkedVXGLAxwg4ECQKSB7kxAySchO7u75/fG5b+5yubtcxoWR5/165XV333u+3+9zl+T5PJ+ttNYYDAaDofMScrgnYDAYDIbDixEEBoPB0MkxgsBgMBg6OUYQGAwGQyfHCAKDwWDo5IQd7gm0lLS0NJ2dnX24p2EwGAxHFatWrSrSWnfx9t5RJwiys7NZuXLl4Z6GwWAwHFUopfb4es+YhgwGg6GTYwSBwWAwdHKMIDAYDIZOzlHnI/BGfX09OTk51NTUHO6pGJohKiqKzMxMwsPDD/dUDAaDk2NCEOTk5BAfH092djZKqcM9HYMPtNYUFxeTk5ND7969D/d0DAaDk2PCNFRTU0NqaqoRAkc4SilSU1ON5mYwHGEcE4IAMELgKMH8ngyGI49jwjRkMBgMxxJl1fUs21nM9oIKMhKiyEyOJislhoz4SMJC23//bgRBO1BaWsobb7zBjTfe2OJzzz77bN544w2SkpKCMDODwXC0UFlr47Vle5i3Po8NuWU4vLSK+dM5g/nlpPb3rxlB0A6Ulpby9NNPexUENpuNsDDfX/O8efOCObVWo7VGa01IyDFjPTQYOoxam52N+w9xsKqOervGZtfYHI6GR4Wie1I0WSnRJMdGMOeHvfxnyU5KKusY0yuZW0/rx4l90xjSPYHC8lpyDlaz72AVY3slB2W+RhC0A3fddRc7duxg5MiRTJ06lWnTpvGnP/2J5ORkNm/ezNatWzn//PPZt28fNTU13HbbbcyaNQtwlcyoqKjgrLPOYtKkSSxdupQePXowd+5coqOjG93r448/5oEHHqCuro7U1FRef/11MjIyqKio4JZbbmHlypUopbjvvvu48MIL+eyzz7j77rux2+2kpaWxcOFCZs+eTVxcHL/73e8AGDp0KJ988gkAZ5xxBscffzyrVq1i3rx5PPTQQ6xYsYLq6mouuugi/vznPwOwYsUKbrvtNiorK4mMjGThwoVMmzaNJ554gpEjRwIwadIknnrqKUaMGNFRvwqDIajsKa7kvVU5xEaGcd7IHnRNjAJk47R2XykLNxXww+4SftxXSq3N0aJrT+6Xxm+n9md0z8aLfWxkGNlpse32GbxxzAmCP3/8Exv3H2rXaw7unsB95w7x+f5DDz3Ehg0bWLt2LQCLFy9m9erVbNiwoSFM8sUXXyQlJYXq6mrGjRvHhRdeSGpqaqPrbNu2jTfffJPnn3+eiy++mPfee48rr7yy0ZhJkyaxbNkylFK88MILPPzww/zjH//gL3/5C4mJiaxfvx6AgwcPUlhYyPXXX8/XX39N7969KSkpafazbtu2jVdeeYUJEyYA8Ne//pWUlBTsdjunnXYa69atY+DAgVxyySW89dZbjBs3jkOHDhEdHc0vf/lLXn75ZR5//HG2bt1KTU2NEQKGowKtNXuKq9hbUsXBqjrKquupqbcTHxVOQlQ4dq15f3UOS7YWogCHhoc+28zEvmn0TotlwaZ88spqCA1RDO2ewJUTejG2VzLdkqIJC1FEhIUQGqIIDwkhNFRht2tyS2WXn1daw4nHpTIuO+Wwff5jThAcKYwfP75RrPwTTzzBBx98AMC+ffvYtm1bE0HQu3fvht30mDFj2L17d5Pr5uTkcMkll5CXl0ddXV3DPRYsWMCcOXMaxiUnJ/Pxxx9z0kknNYxJSWn+D61Xr14NQgDg7bff5rnnnsNms5GXl8fGjRtRStGtWzfGjRsHQEJCAgAzZ87kL3/5C4888ggvvvgiV199dbP3Mxjag7X7Snl75T6qam3U2zV2h+bcEd05e1jXhki1OpuDf3y5hfdW5ZCZHEP/jDh6pcayLb+cZTtLOHDIf1hzenwkt57aj8vG96S63s4Hq3P4YG0uK3aXcHL/Ltx5xgBOG5RBYnRgyZI9U2M4gdTmB3YAx5wg8Ldz70hiY12q3OLFi1mwYAHff/89MTExTJkyxWssfWRkZMPz0NBQqqurm4y55ZZbuP3225k+fTqLFy9m9uzZLZ5bWFgYDodLbXWfi/u8d+3axaOPPsqKFStITk7m6quv9psDEBMTw9SpU5k7dy5vv/02q1atavHcDIaWcKCshoc/28z7a3KJiwwjJTaCiLAQqmptfPbTAU4flMED5w+lzubgljdX82NOGacPSqeqzs7CTQUUV9aRFhfJhD4pTOiTysCu8STFRJAUE05UeCjlNfUcqrZRXW9nSPcEwt0idm7/2QB+O7U/docOSiRPR3LMCYLDQXx8POXl5T7fLysrIzk5mZiYGDZv3syyZctafa+ysjJ69OgBwCuvvNJwfOrUqTz11FM8/vjjgJiGJkyYwI033siuXbsaTEMpKSlkZ2c3+ARWr17Nrl27vN7r0KFDxMbGkpiYSH5+PvPnz2fKlCkMGDCAvLw8VqxYwbhx4ygvLyc6OpqwsDCuu+46zj33XCZPnkxycnAcW4Zjm1qbnc155eSV1XCgrJr88lqq6+zU1NuprrdTZ3NQb9fU2x38sKsEu0Pz6yl9uXFKX+KjZDduszt46bvd/OPLLUx9bAkaCFHwzBWjOWtYt4Z7lVXXkxAV5jO/JS4yjG6JvueqlCIs9OjPjTGCoCXY60GFQEhoo8OpqalMnDiRoUOHctZZZzFt2rRG75955pn85z//YdCgQQwYMKCR6aWlzJ49m5kzZ5KcnMypp57asIjfc8893HTTTQwdOpTQ0FDuu+8+ZsyYwXPPPceMGTNwOBykp6fz5ZdfcuGFF/Lqq68yZMgQjj/+ePr37+/1XiNGjGDUqFEMHDiQrKwsJk6cCEBERARvvfUWt9xyC9XV1URHR7NgwQLi4uIYM2YMCQkJXHPNNa3+jIZjm+o6O++u2sf8DQeYeFwaVxzfk6SYCBwOzcfr9vPwZ1vILXVpw2EhipiIUKLC5ScyLISw0BAiQhVnDe3Kb6f2JyslptE9wkJDuP6kPvxsSAZ/mvsTNruDhy8aTmZy43GBmnGOdZTWXoJVj2DGjh2rPRvTbNq0iUGDBgX/5vkbIToJEroH/15HKfv372fKlCls3rzZZ+hph/2+DEGh3u7g0c+3sKOwkqmD05k6uCspsRFex2qtKauup7C8lsKKWpbtKOa1ZXs4WFVPZnI0OQeriQ4PZcboHmzILePHnDIGdUvg11P60ictlq6JUaTERBAScvTvug83SqlVWuux3t4zGkFLcNTLj8Err776Kn/84x957LHHTP7BMUp5TT03vr6ab7YVkZEQyYJN+dz9wQbG9kpmTK9kRmYlMbBrAutzy/hqcwFLthZQVFHXcL5ScPqgDGad1IexvZLZkl/OC9/s4u2V+0iNjeTRmSO4YFQPQo+WhX/N/2Dfcpj+5OGeSZswGkFL2L9WNILk7ODf6xjGaARHJwfKarj6pR/YVlDBgxcMY+bYTH7af4jPNhxg8dYCNueVY3NLh02MDufk/l0YnplIekIUXeIi6ZUaQ/ek6CbXrqi1EREaQkTYUbSB2DIf5lwO2gF37oTYIyMCyBdGI2gPtAa0/NINhmMEh0Pz7fYiiipqCVGKkBBFdZ3N6aitIf9QDQer6imrrudAWQ0hCl68ehwn95ce6EN7JDK0RyK/O2MANfV2ftpfxqa8cgZ0jWdUVlLA0TRxkYdhKdIaVr8CA89t+SK+fw28ey3EpEFlgbzud3rL51BTBs9Mgil3wagrWn5+O2EEQaBYmtNRpkEZDBYOh26wtdfU23l3VQ7//XYXu4oqm4xVCtLiIslIiCQ5JoKslBhO6pfGZcf3ZGDXBK/XjwoPZUyvFMb0cstXqa+RAIsw7z6Ew0ruKvj4NqgthxNvCfy80n3wxiUiBK76AP49Fvavbp0gWPEClO2FTR8ZQXB04NQEjCAwHMHU2Rw4tEYp+VNdu6+UBRvzWbi5gF1FlUSFhxAbEUatzUFFrY3hmYk8cdkohvdIxKE1Dg1R4SGkx0e1j5nmjZkQ3x1mPOt/3Jr/QXJvyJ7Y9ntqDQvvh0O5EBImP8edBoPPazxu2xfyWLStZdd//3oRcD+fC2nHQVo/yF3d8nnWVcH3T8vzPUvBYW8SkdhRBFUQKKXOBP4FhAIvaK0f8ni/J/AKkOQcc5fW+kitwuZ8YkxDhsOHFWK5q6iSmIhQYiLCnCaZQ6zLKWVnUWWTvUpEaAgn9E3l3OHdqLU5qKyz4dBw3ojujO+dErweEfZ62LscYrv4H6c1zL8Lep3QPoIgdxV8+xjEdRUhUFMqO+7+ZzXWTCxBULw98GvXVcLeZXDy/0G608/VfTTsXCSfoyXf5epXoaoIxl0nmsGB9dB9ZODntyNBEwRKqVDgKWAqkAOsUEp9pLXe6DbsHuBtrfUzSqnBwDwgO1hzahPat0bQljLUAI8//jizZs0iJiam+cGGYwqtdcAL8ao9Jcz+aCPrc8uavJeREMmwHklMG9aNqIjQhj/TPmmxTO7f5fDY4Iu2gb0WDuVAVQnE+ChxUlkIdeVwYEP73Hf9OxAaCTf/AFGJsOUzePMS2LkY+v9MxlQ47foqtGUawYH1gIbuo1zHuo+CdXPg0H5I7BHYdWx1sPQJ6HkiTL5DBMHub489QQCMB7ZrrXcCKKXmAOcB7oJAA5bBMRHYH8T5tA0/PgJ/ZagD4fHHH+fKK688rIKguXLZhvalpt7OS9/t5pnF24mPCmd0r2TG9EwiOTaC/aU15JVVc7CqnlAlyVGlVXUs2FRARkIk/7p0JOcM7051vZ2qWhshIYq0uMjmb9rRHFjv9nwd9JnifVzxDnks3+9fYASC3QYb3oP+Z4gQAOh7KkQlwYZ3XYJg+0J5HHK+jK8pc433x34pLEk3t2KKPUY731sduCBY95aYrs59QvKSUvrAnu/gxJsDO7+dCWasVg9gn9vrHOcxd2YDVyqlchBtwKvHRik1Sym1Uim1srCwMBhzbZ4GjaCpaci9DPWdd94JwCOPPMK4ceMYPnw49913HwCVlZVMmzaNESNGMHToUN566y2eeOIJ9u/fzymnnMIpp5zS5Nr3338/48aNY+jQocyaNQsr3Hf79u2cfvrpjBgxgtGjR7Njh/wz/f3vf2fYsGGMGDGCu+66C4ApU6ZghdwWFRWRnZ0NwMsvv8z06dM59dRTOe2006ioqOC0005j9OjRDBs2jLlz5zbM49VXX2X48OGMGDGCq666ivLycnr37k19veRVHDp0qNFrQ1NqbXbyD9Xw0Y/7Of2xJfz9s82MzU5hZM8kVuwqYfbHG7ltzlr+/tlmPliTy4bcMlbvLWXp9iI27j/ETaf05as7pnDeSImzj4sMIz0h6sgUAiCLf4hzc5G3zve4kh2u5/k/te2eu5aIhjFsputYWAQMng6bPxW7PIhZKC4Dhlwgr4sCNA/l/Qix6RDvKlNB12HyOQP1Ezjs8O0/oetw8V0AZE8SQeCwez+nZCd8dEv7aU0eHO4t4GXAy1rrfyilTgBeU0oN1brxaqu1fg54DiSPwO8V59/VeCfSHnQdBqf9yZpNk7c9y1B/8cUXbNu2jR9++AGtNdOnT+frr7+msLCQ7t278+mnnwJSNygxMZHHHnuMRYsWkZaW1uTaN998M/feey8AV111FZ988gnnnnsuV1xxBXfddRcXXHABNTU1OBwO5s+fz9y5c1m+fDkxMTEBlZ1evXo169atIyUlBZvNxgcffEBCQgJFRUVMmDCB6dOns3HjRh544AGWLl1KWloaJSUlxMfHM2XKFD799FPOP/985syZw4wZMwgPNyn7xRW1PPL5FrYXVFBeY6O8RsIvK+tc/+QDu8bzxnXHc+Jxrt/5/tJqKmttdEuKbn9TTm0FVB+EpKz2va4/DqyX/52KAhEKvijeAShAQ/4G6D259fdc/y5EJkK/nzU+PvQisclv+wIGngM7FkrYaJqzvErxNsgc0/z189aK+cbdnBceLf6C/WsCm+N3j4vwm/my6zq9Jsn88n+CbsNdYwu3wjf/EHNXSBhkTYCuQwO7TwsIpiDIBdz/6jKdx9z5JXAmgNb6e6VUFJAGFARxXq3Dj0bgyRdffMEXX3zBqFFiR6yoqGDbtm1MnjyZO+64g9///vecc845TJ7c/B/8okWLePjhh6mqqqKkpIQhQ4YwZcoUcnNzueAC2c1ERUlzjAULFnDNNdc0mJgCKTs9derUhnFaa+6++26+/vprQkJCyM3NJT8/n6+++oqZM2c2CCpr/HXXXcfDDz/M+eefz0svvcTzzz/f7P2OdRZtKeDOd9ZxqKaesb2SyU6LaahpnxwTTnJsBN2Toji5f3qT7FlviVbtxpd/gg3vw+2bIKIFJkhbXfOhnw4HVORDgtsuWWsRBIPOcQoCP5uzkh1iGqkpE0HQWuqrYdPHMOQ8CI9q/F72JNEANrwL8V3lXv1Ol0ilQP0EdVVQuBkGTmv6XvfRsHFu8w7jVS9LRNPQC2GQWxST5STf/a1LEKx/F967TgTNhF9LiGt81+bn2QqCKQhWAP2UUr0RAXApcLnHmL3AacDLSqlBQBTQNtvPWQ81P6Y11Dib3QQQPqq15g9/+AM33HBDk/dWr17NvHnzuOeeezjttNMadvteb1lTw4033sjKlSvJyspi9uzZfstA+8K97LTn+e5lp19//XUKCwtZtWoV4eHhZGdn+73fxIkT2b17N4sXL8ZutzN0aPvvVA4HOwsrWL6rhAtG9SAq3Hs4n9aaZTtLWLG7hMiwECLDQtiSX8GbP+xlUHoMC7NfJXHspTDw7A6evRccDjGL1JTC5k9g+MWBnbfhPfjwRrilGdv3mldh3v/BrWtc4w7th+oSMX9UFMhOvK7KuxAq3gmpfcFe1zbTx9bPxOnsbhayCAkVM9DKlySaSIVCn1NEyCX3Eo2gOfJ/ko1gNy8O3R6jJTmtxPlZvLHhffj4N6KtXPAsuJdhScyUigW7v4UTboTKYph3J2SOhcvmQGxTa0F7EjQfgdbaBtwMfA5sQqKDflJK3a+Umu4cdgdwvVLqR+BN4Gp9pNa8aJhW0+l5lqE+44wzePHFF6moqAAgNzeXgoIC9u/fT0xMDFdeeSV33nknq1ev9nq+hbUIp6WlUVFRwbvvvtswPjMzkw8//BCA2tpaqqqqmDp1Ki+99BJVVWIHtUxD2dnZDb0BrGt4o6ysjPT0dMLDw1m0aBF79uwB4NRTT+Wdd96huLi40XUBfv7zn3P55Ze3rdpoVYnEZR8BrNpTwgVPL+UP76/nlEcX887KfdjdyiZorVm6vYhLnl3GZc8v47Evt/Lg/M3M/ngjb/6wl2sn9uajCZtI3D4Xvv93y25eUQiVRe38iRBzRkW+PF/7RmDn1FfDF/eCrUacoP7YsUiig7a4RX5bGkDXYfKjHVCwqem5WsvimdIXMobKjttuC2yOnqx/Vxb5bB+a9tCLZJ4rX4SeE6RcDEBqv8B8BHlOR7G3yJ7ulsPYh3lo5xJ4f5bcd+YrEOrFhNrgJ3DAF3+URLfpTwZdCECQfQTOnIB5HsfudXu+EWiHwOGOwLdpyLMM9SOPPMKmTZs44YQTAIiLi+N///sf27dv58477yQkJITw8HCeeeYZAGbNmsWZZ55J9+7dWbRoUcN1k5KSuP766xk6dChdu3Zt6AgG8Nprr3HDDTdw7733Eh4ezjvvvMOZZ57J2rVrGTt2LBEREZx99tn87W9/43e/+x0XX3wxzz33XJMS2e5cccUVnHvuuQwbNoyxY8cycOBAAIYMGcIf//hHTj75ZEJDQxk1ahQvv/xywzn33HMPl112Weu/2udPlSiPs/7e+mu0kM82HODpxdsZ0j2BX5yYzcCuCSzYmM/Nb66mW2I0f71gKM9/vZM7313Hf5bsID0+iqo6G6XV9ewprqJrQhR/nj6EmWMz0VqigEJDFEmOUnjyQQlf3LNUdsNx6f4nY6uVUMKv/yGLzLWfBf5B6quhPE9MK77Y+jmgYNwvYcV/oSy3+eiW5c9K2CdI1d1B53ofp7XE1YMIgvHXy3NLEGQMcZkzDvzY1A5ffgDqK2UXHREngqdkJ3TxXhrdJ9UHResYd73vpKzMsZDUE0r3wnFuWcBp/cTJ7HA03qV7sn8txKRCgpfvLn0QhEWJw3jYRU3f//7fYpq6bI5v01yvSZJYt/wZ+PFNmPw7V65CkDFF5wKlqlj+gEBUw2Al4RxlvPvuu8ydO5fXXnst4HMa/b4qi+GRPmJC+NU3QZqli6KKWu6b+xOfrs+jV2oMB8pqqLU5GJmVxPrcMoZ0T+Clq8eRGheJ1ppP1+fx6tI9aDQxEWHERIRyQt9ULh6b5d1s9MGvxbF30X/h7Z/DOf+Esdf6ntDOxfDJ7WInj+8mES937Qvcjr/47/D1w3DtF76dnc9NgZBwye59YhScdq/ErvuiqgT+NVJ2r0Vb5O/94le8jz24G/41QiJpqg/C/+2QMMy3rhJ7/61rRFg81AuGzoBzH298/u5v4eVpcOX7knj27GS46EWxobeEde/A+9fBdQtlwffFgtkSsfOr71xO15UvwSe/gdvWiZnIF89MEqF+1fve339hqgghb4L80QESPusvw7p0Lzw+DFCQ0ht+vVT8A+2EKTrXHrgLTO0QG2Mn55ZbbmH+/PnMm9eGZPBCp7mgYJPsjMPaLxSyzubgq80FbNxfRnFlHcUVdSzfVUxlrZ07zxjArJP6UFFj451V+3h9+V5O6pfGk5ePbojYUUpxzvDunDM8wP4Te5fDj2/ApN/CoOli7tj4kW9BULwDXpshtuEr3xOTyJuXiHkh0Azb3FXgsMF718IN30CURx2g8gNyvVP/JFpDzxNg7Zsw6Xbfm5mvHxVb++mz4asHvJt0Gj6zUxuY8nv49A7YvkAW8QPrXU5PpcQ85M1hbOUQpPYVQRgSJrb4lgqC/ashLNq7/d6dSbeLGcc98iatn3Mu23wLgvoa+Vvt/zPv74PTT/Cq/B5D3ZbW8nyoONA4GsgbST0hqReU7pENRDsKgeYwgiBQGgmCo0uLChZPPtkONditRcZRLwuAlZzjpM7maFHNG601W/MreGflPj5Yk0txZR1KQUpMBCmxEYzNTuH/zhhAv4x4AJJjI5h1Ul9mneTDwRco9dUw7w4xG5x0pyx+g6fD0id9J0l9/5TsIK+ZJ+aTSvHBsG954IIgfwOkDxHb+qe3w4znGy/w276Ux/5nyuPIyyUePXeV953zwd3ww3Mw8grIGCymia2f+RbSe5dJuOboX8CiB2HzPDhuKhzc1biIWrfhsvP2rKdTsgNCIyAxS46n9fefS2D973kKsf1rZXEPbWZJi0qQ34s7qU5BULTdZTLa9bVE+Ex/EiJiZU4Om39B0300LP+P/C7cBY0VOtu1GUEAMPE20Qr7TGl+bDtyzAiClqTqtw6Hj+eGltDEFFmwSbQrbRdnnJsgmPPDXu75cAOXje/J784Y0NBW0O7QLNiUz87CSjISIumaEIVSisVbCvhiYz67iioJD1WcPiiDi8dmMblfWnCaizvsYlte/66ELdYektjwCGck1qDpYobYMr9pZcmKQlj7Ooy41GVDj02F1OMgZ0Vg968qkezU8bOkrs+iB6DvaTDSzV+z9TMRThlD5PXg8yXCZ+3r3gXBd/+SXfkpd8vrjMHyuynaKrt6T/Yug6xx4vwccCZs/NjlVHVf+LoOA1u11PXpMsB1vHiHaESWcMgYAnu+9/55836Et66ECTfBhF+5jjscstiOaKWfKi4dIhMaRw4tmC3CMjwGzvu36zO5ZxR7kuX04e1b1lgQ5P0oj96+P0/G/bJFU28vjglBEBUVRXFxMampqcETBkYjaDNaa4qLixvyHgDZPfUYIwuNlb4PfLU5nz9+uIFeqTG8vnwP8zcc4A9nDeRgVR0vL91NzsHqJtcPD1VM6JPKtROzOXtYN1KDnXH77WNiOolMkEV/xKWNk6G6j4LEnt5LDK94XhyjJ3gk02cdL4t3IAXMrJj7rkMlFHLnYjHPdOkv36mtVo4Nv9h1ragEie3f8B6c8WDTePucFaKNWO1Y0wc777Wx6UJWVSLmkmFOM86AaeLsXPYf57zcxltCIW9dY0FgRQxZZAwVH0v1QYhOdh3fuQTmXCEmq5/ebywIirdDXUXj+j8tQSkRwFYuwf41IgS6DIQ1r0HfU0QQRCeL+cYXyb3FvLX7OykkZ3FgnQg7K0rpCOSYEASZmZnk5OQQ1PITNaWuXIKSUO/hX4ZmiYqKIjMzU15oDQUbZRENj27Ydf24r5SbXl/DoG7xvDXrBHYVVfLHD9ZzxzuysxqXncw90wYx8bg0CstrOXCohuo6O+N6p5AQ1YG/l13fQMYwuG5B0wUVZIEZdK4s+jWHXPb7ukoxvww4u2l0TNZ42a37i0e3sGLuM4bJjnrGc/DimfDiWXDuvyA+QxbIfmc0Pm/YxbLY7vnOVeIARKso3CK1eSxSjxNHc8FGmmBpLj0lOo4+U8ROv+VTia5xL8PQZYCYgA78CMOdcf4Oh3xO9/tlOHfS+Rtd5rEN78MHN8hcuo2E9W/Ld2hpXv7COgMlrZ8s4CCRVeExcPU88dl8/BtZxJsLElEKek0UB7i7IM9bF5hZ6DByTAiC8PBwevfuHdybfP5HV1z4Dd9AN9Nqsc1UFMjOL32Q/KN9/zQ7DhRz7curSY2L4MWrxxEbGcbQHom8f+NEFm7Kp2tiFMMzXTur+Khw+nSJOzzzL9wiC6k3IWAxeDose0pCG62wwjWvy+eeeFvT8VnHy+O+5c0LgvwNEmkTnyGvE3vArMXwzi/gw1+JNhIWBb1P8riH04SRt7axICjeLkldGW5mjdBwsdt7EwR7vxczkhVDHxEji/qWT0UbcF80Q8Pl9+zuMD6UK1qR++e0TFj5G6DXiWJaW3i/RDBd9ibkrBKH/L4fZFiVE2YAACAASURBVKcOsoMPi4Y0N02jpaT2cxaCyxNT37CLxFR34X/hP5MkomfIjOavkz1RspctQV5T1tRfcgRyFDUIPczY67w/N7QeK2IofRCVqUPBUc/vnpyDXWtevmY86fGuBTY0RPGzIV0bCYHDSvVBiQRJaybePXO8JDktfVIiSvb9AN8/Kcd7Tmg6Pm2AOF/3LW9+DgfWN160QRavqz6A438tna96n9w0FDU6WUwVbqY4wOWktRZji4zB3iOH9i6TXbL79a1Mam/28K7DZXdsFVazis25m4biu4o2sW85vHM1LPyzhJ1e9YHMu+fx4lPa/a3rnEAdxf6whNGiB8SXYZl2knuJdgXef1+e9HJqMXuc2oWltXX141s4AjCCIFBsbpmvttrDN4+OImelxE3XNs14bg9qbXbyt0sW5lMbwpn5kdj8L+95kHm3Tua49Bbs8rWW9oEdSeFWeewy0P+4kBCxZxdukWid/06V3aU3bcAanzkW9jXjMLbXN41OsQgNl1IrP/8Ipj3q/fxuI11OTIv8DWIGsqJoLNIHQdk+2d1a2GolecpzcRxwtggZ94Qti35TpezEj2/Ka/fQUQulRBBteE98K1P/IrtyK5QyMl5MQNZCazmKmwsbbQ4rhHTN6+JfcTczDZ0hOQZW5JXf6/SXFpaWmcmKGGoudPQwYwRBoLgv/u5C4Vhl8yeQv75l3ZsCoKy6ngfnbWLY7C9Y+PUSDuo4Hv3+IOk9B2KPSGRm96KWF19b+zr8a3jgpYTbg8LN8tglAHPEpN/C3bmSXHXpm1JnZoCfGkRZx4spxn3h9aRom9OM4ycSpc/Jvp2b3UZIvHqVW4Xa/J/k83gWmUt3aggFm13H9q+Vcg2egiAmBW770Xv446Dp0GOsONjrKsV8EhYlrSwbzXuK7P6veAcm3trULp89STYqdVVtdxRbNGglGsZ6idxJ7hVYEqlSYtLas1Re561zlq0OTrG49sIIgkBxFwSdwTRk1VZvp9o3dTYHL367i5MfWcRz3+zknOHdOCO9lIjuQ9j457N4+drjCe0xwuX4awmrX5Mkv+1ftstcA6Joq9il/UWRuBMSKgldA8+W6CJ/pQyyxgNaFjtfuEcMtQZrx+uuFeT/1NQsBK4yB+5+gn3ORLKsAMwlFkrBGX+VkhhL/y0aQUqfpt/FxN/CnTu8axUgpRgc9eKstmr7tLWzV0SM5DJEJYkG0BayJ4lZrnSvU1s5srUBMIIgcGy1riYbx7ppSGuX/bgy8EiseruDpduLyCtzhXZqrZm/Po+p/1zC/Z9sZFiPRD69ZTKPzRxBauUOYnsMJTrCGUPebYQsRrYWCNriHa5FacdXgZ/XVgo3izkhGM3Ge4wBFSL+BBCb+u5vG38vB9ZLFE5zPgpfdPMQBFZOgjdBkNRT6gBZgkBraf+Y0hfimulH7EnPCaIZfPcvEfreaiSFhPj/XntOkO9n97dyjbY6ii0m3w5nPdz2jN5eJ8rjjkVO892RLwiOiaihNjP3ZrH1+msTZ6uRePHqkiNTEGgtDbR7T/G/2wyEkp1Q6zRLVDTfGqKsup45P+zl5aW7ySsTs9nQHgmcMiCdZTuLWbH7IP0z4nj5mnFMGeAswFaWKwlY7kW1uo0Ubatwk//EHXfWvQUoqRG/46vWl6mY44zqmPF8YHV+Cre4wibbm6gEid/ftwx++hAW/U1q/kz8DUz9s4zJ3+AMyWxluGxMikQVWRqYtch7EwRKye/JchiveQ32LoWzHmndvU+fLUl25XnNR0Z5IypB/j72fCd/9211FFv4qwnVEtKHSL2lFS84s5GPfEFgNAKQqol7fWQzWtjrXHHg9iNQEOz9Hl67AHYtbvu13Fvu+dEIKmttPPblVk58cCEPzt9M35QIlgz8kL+cHEdEaAj/XrSdXUVVPDhjGPNunewSAtAoYqgBy87rGc3iC4dDHI99pkhJhPqqwKJtPMlbJz6RzZ/A/y505Yv4orZCnKctrZDZErLGSzLYO78AtJhDlv8HypwVQQ9s8O8fCITuI1zfdUPEkA9TU/pgGVO6T0Kpsyc3TppqCal9XVVKU1ohCMDpJ1ghGk1bHcXtTUiINKVvcBQf2RFDYASBqNtVxc07gC2NAI5MjcD6R24Pm/7+NS4nnhdBYHdo3l6xj1MeXcwTC7cxZUA6n9wyif+dl0Sv3W9zVcTXvH/jRNb+6Wd8+/tTuGx8z6YlHqzdZRc3QZDcW77jQP0E+5aJHXbEZbIwhIS1zjy06iX5vNMeg5wf4LXzGztRPSkKMGKoLQybKY7V85+BG5fB+U+LH2Txg6KlVRa0vWVht5ES415dKqammFQpleyN9MGiDb/zCzFVTX+ybZrnyf8Ho65q2lIyULIny+asvrLtjuJgYCXDRSZAUvZhnUogGNOQ1bCjucXdVutKeT8SBYEVxeIv0iRQ9q8Wu6a2NxEExRW1/Op/q1ix+yCjeibxzJVjGNPL+b1sd36Xu5YAfyIxxo/ZomCzRFPEprqOhYTI7mn3d5LY49760Bs/vgnhsVIyISJWYvN3fCWmh0CpLYd1b0uy0LhfSjbsO7+A/82A6xd5jxQp3CKPwRQEvU6E6xe6Xif3kppCy552lX3wtXsPFMvBemCd01E81HdkTIbznrmr4OxHpUxyW4hOlho+rcXyE2hH2x3FwcDKJ+g6rO2m2g7gyJ9hsCk/II/NagS1Lo3gSDQNWYtTdWnbruOwi7rdfZRkrboJgh2FFcx4Zinrcsp4dOYI3v/1iS4hAC5/Qu7q5s0rhZsg3ctCOnSG2MP/OUTs9juXeD+/vlrs54PPc5UaOO5UmXtLtKL170j44Vhnh7WBZ8MZfxOtyFf55cLNEm+fHORsdk8m3wER8VIQDQIrYuYPy6SSu1o+qz/BYgmf7Mnewys7mqhE2ay0l6O4vek6XP5/AklCOwIwgqA8Tx6ba5Voq5VkFmhZVEtH0aARtFEQFG4RW3uP0dii03CUF7C/tJpFmwuY8fRSKmpsvDlrAheNyWxa4M/SrrTdFUftDYdDNIIuXsp0jL1WeuSecJP4PV6d7r393+ZPxdnsXmnTqlmzc3Fgn1VrKY2cMRQyXd3f6H+G/+sUbpGIofZwULaEmBSY9BsxicR3917auiXEpkFCphRxs1V7dxS7j738bWmzeKTscCfeKsKxo38PgRAaJia9k39/uGcSEEH9jSqlzlRKbVFKbVdK3eXl/X8qpdY6f7Yqpdq4irWCgDWCGqkpExpx5CWUVRa7du5tMA3ZHZpd66RL2HUL7Dy76hD2iiImPrSAa15eQVpcBB/eNJHRPZO9X6CiQHZoYVFO85AXtBbnfH2l7zZ8qX3hZ3+BXzsd+Lu8dC776QMpr9xrkutYt5FicgjUT5C7WswiY69pbBJJ6ilhjb4+Q+HmwBLJgsGEX8vi7dG3odV0d8sw9icIQASkuynvcDP0Qjj5zsM9C9/EprVro6VgEjRRqpQKBZ4CpgI5wAql1EfOPsUAaK1/6zb+FqDjvT6WRtCc3d9eK31ow6KaTyhrrvdpe2NpA9AqQVBrs/PeqlyeWbKd6w99SZfQaKriejOsy3GE77Tzj3OyCY1L4ZSB6f6re1bki10/MUsae7jjcEjJgG//Kc7g+O5ScsAf8RlSrsAzEsjhEI1jwNmNv+eQUIkg2vGVq/pj+QExHVnanDsrXxQfw7CLm77X+2QpPubZbaq+WjJyh1/if+7BIjxafAehEc2PDYRuIyRaSoUE1+dhOKIJ5mo1Htiutd6pta4D5gDn+Rl/GfBmEOfjnZb4CMIinRqBH6GhNTwxUoqMdRSWIEjIbJFpKP9QDS98s5Mpjyzm7g/WkxITwbldDhDVazRv3HAiJ42SHeKMARGcN7JH8yWeKwqkwFrvkyTOvcLN0bxwtjhha8vh3CfgtrWQmNn8JLOOlzBB9x4QRVslgsVK3HGn76ki3BfeD//9GfxjgPTptYp/WeStk3o2w2c2be8IUp6hrlwc5+4UbxcH5eHSCMBZmK2NZiELy0+Q2s9/FVXDMU0wBUEPwL0SWI7zWBOUUr2A3oBXnV4pNUsptVIptbLdew40aAR+BIHW8n5YlAgDf4Kg9pDsGL/9p+weO4LCLZL5mTG4WY2gpt7Oi9/u4qJnlnL83xbywKebyEyO5tVrx/Phr8aRfGgLYVYT9Ng0eQw0u7giX7o99Zkir3c7TToH98CyZ2TnffMKGPOLwFXmzHFy3dI9rmNWwbFeXhK6+p4KKGkaU1cFJ98ljt2Xz3YVcts4F148QxZTX8Xfsp2lmz2d1R0RMdSRWBE3zZmFDMc0R4qX5VLgXa213dubWuvngOcAxo4d277twQLRCOz18hgWIQuYv6ghK/68qliyXsdc3S7T9Itls45KcsW4e6Gy1sb1r65k6Y5iBnVL4I6p/TlrWFeOS3eaTfavEbOXZX+OdSaABZBdLOPypVNWt5ESYbVriUQBLfqrmB5On93ykgxZ4+Vx3woxE4GYheK7eY/aScyEqz8VgWRVlBx5ueQGvHqedNNa/aqEml7yP1ctf09iUyUqZ9eSxnbows1SBrk1GbFHInHp0m944DmHeyaGw0gwBUEukOX2OtN5zBuXAjcFcS6+sTQCe51v274lJMKixE/gTyOodgqCkHBpTj7q54H5C9a+IQvMiFbYnq0GKRGxPsNHS6vquPqlFazPLeMfM0dw4RinWWbdO5BTIztDa8dsJejEOuvIBBKOWV8j2khcutjUe02U3XTeOonTn3ibNE5pKelDxI6f84OYcbQWQdDzBN8x756N35N7wTWfSeb16lclC/mcfzavlfSZAsufFc3CKjtRuEUcyUeJEzAgpj9xuGdgOMwEUxCsAPoppXojAuBS4HLPQUqpgUAy0EyNhyBQXy029Yh4sQfbarzXmbGcw2FRohX4EwRVB+VxzC+k1sj2L13hiP5Y/qxcv6WCwGqQ0mWA2N9rD6Eddv700UbyD9XSp0ssfdJieem73ewsrOTpK0ZzxhBnSdyDu+F9jzIB0SmQ1Euex6TITj4Q05AVOmplpvY5GbbOh7k3Scz3pN/6PtcfoWGioVgO49I9UL7fu3/AH/EZcO180Xp6nxxYSeHeU8TXs/d7EbT11VKS4SioHWMwtISgCQKttU0pdTPwORAKvKi1/kkpdT+wUmv9kXPopcAcrQ9DR3jLLJSSLSn2vgSBpRGERjijhgLQCMZeC5vnyUISiCCoKvHutGwO9wYpRVtBO3jv+y38b9leslKiWbKlkDq7g+jwUF68ehyT+qW5zt3wvjz+4mMx/+xfI4lD1iIZEiplByoDMA1Z5iNLEFjtEQ+sg5890LbG3VnHi8+lrhL2OPcLLRUEIAKpz5TAx/c6QTS7XUtEeLxzjdQYOvvhlt/bYDiCCaqPQGs9D5jncexej9ezgzkHv1iCIDnbKQh8LPDW8TArj8BP+KjlI4jrCsffAAvucxbGaqbwVFVx60JOreJtXQY07Mqf/WINJ/TpyxvXH49DQ+7BamIiQ0mL8zBnbHhfnLHWom311HUntktgpqEGjcDpV0gfLOeGRcG461v+udzJGi9JavvXiKM4Ksl7Mlp7ExEr38/OxRIBtXU+TPsHDDgr+Pc2GDqQIyRF8DBR4SYIwLfDuEEQRDqjhvw4lqtLACU74DFXi337+6f9z6O+WhKs6ipbMHlBF2yWJK7EnuioRABiHOU8OGMYSilCQxQ997xL2sfXuHrFgti689fDUC+LvzuxaU1NQ988Jjt0dzxNQ0pJFuqlb7Q9LNHK+t23XMw0PU/ouDyNPlNEkP/4Bky5u/UVNw2GI5jOLQjKAxUElrM4gISyqhIxQYSEijDod7o4Ov1haRG1FQFPXWvNfXM3sHTZUvaH92RrYSXL9stCf93YFLLTYl2Dty+ELZ/Cj3Ncx9a/K/b/IRf4v1FsetOooVUvSVcwdyoKAOUKOQVx2raHPT0mReLcN38qcfytMQu1ln5TASUF307+v467r8HQgXRyQZAnpp54Z5XLQDSC5hLKqksaJ/vEpvsvaQxiFgKp9+LwGkHbhKcX7+CV7/cwIDSXFZVd+Nk/v+bvi0Swnd3Pw89h7ei/+otoHVpLMlX2JN/hkw3z9zAN1ZZL6eeDuxubyCryxZ/Q2kYpzZE1XipfQscKgh6j4TfrpHNVIA5mg+EopJMLggOSpWm1pvO1wNvdfATNJZRVlUjkjUVMioRV2m1+zil2Pa9rXiv4cE0uj3y+hYuHJZLmKOL0k0/mrrMGktldBFporUdSWWWhaD3leRLSmrcWSnY0bxYCaUVYV+5KjrMqcmq71LK3qCjwXcu+PbDyCcJjOr7RR1JPIwQMxzRHSkLZ4aE8T7SBMKcNuzmNIDSy+YSy6pLGC2JMKqAlTNXdbOJOI0FQKaYlHyzdXsSd7/7ICX1SeWCSgm0Q22MIvxrYF8anwN9pml1cUSCO4IoC+PZxKNom0TCDp/v+HBYNuQSFsiBaDXBAopSsUgtWVnGwyHQKgsxxwdM6DIZOitEI4ru6koN8laJ29xE0l1BWddBDI3BWa3Rf7Juc42Y68uMwXp9TxqzXVtE7LZb/XDWGiGKPTllWvwR3QWCrcwqhdMnstdfB+rfhuNNdjXb84S4IQHrbhjk1KPcs5mBrBF0GSqTQYH/lqgwGQ2swgiAgjaAlCWXFjX0E1mLrz0/gLiRqy70O2V5Qzi9e+oHE6HBeuXY8idHh0sAlNMLl7A4JFWHgLgiqnPb92LTGvWKHXuh7Pu40lJlwCoL8jVJ6Ib67aBYgPoeK/Ob9DW0hJARuWiZdxAwGQ7vSeU1DtRVSIC6+q5sg8JVHYGkEEaIR+DIN2WolDLTFGoGHaciDfSVVXPnCD4QoxevXHU+3ROeOvHiHNP92r98TldS4Aqm1k7fMNqfcLZnDge6s3QvPaS1VRYecLyGhliCoKZPvJJgagcFgCBqdVyOw4t4baQQ+qoW61xoKi5IyxN6cv9auP8bN5GJpB9UBagRuzmKHQ/P5Twe44oXlVNXZeO2X4xuHhRZta1r8LCqxsUZg7eStnX1kPEz4lQi1QGgwDRWIT6WmVOr/pPWX+2vdNKvYYDAcVXRejcAqNheX0bxGYOUNhEa4FlBbDYTGNR5nLfYt1QiqSxp28lUVZewvKOeHXQd54Zud7CyqJCslmpevHc+gbm4lKOw2idoZ5FE10lMQWBqBL0d1c0TESInryiIxC4GzkbmG2jIRAlZiXjCdxQaDIWh0YkHgXLziu7kyX5tNKHNWHwXvSWUNGoGbIAiPkfP8CAJdWUyOTieLUu5/fwVz7FIWenhmIv++fBRnDulKWKiH8la6Bxw2SD2u8fGoRInxt7DqBLVlkbayiwucEUPpg12fv3ib0QgMhqOcTiwInBpBfFfX4t5sraFIV4SRt7GWRhDj1tdVKdEQrKqkHjgcmrLiA2yozSYrFM4fnMAJQ0aSnRrL8MzEpg3iLYq3y2Nqv8bHo5Mgz0MjCIuSXX1rie0ii33+RlfTdOu+RVtdOQZGIzAYjko6sSA4IGGQUYmyWIeE+e4oZqsVs5BSboLAi/ZQ5cU0BCIYfGgED83fxO31pWT07A+5K5jQIwpGBlC3v0EQeNEIPH0EseltS4iKTRcto7rEaRZCGseHx4ifIDRcvp+oNlQYNRgMh43O6yy2cgisBTIsyr9GYPkRrKbh3kxD1V5MQyDOYy/O4v8s2cFr32wiStUzalB/EUzeMovz1knTHHeKtsnC63mvqETJBLac2ZWFrfcPWMSmiQZVuFXMQiDhnKnHiUZg5RCY7FuD4aikkwuCbq7XYVH+fQSWJuDPsVxVIou5VbLCwkMjKKqo5abXV/PQ/M1cNEjGqphUKXvsKQhK98Gzk2HdnMbHi7dLK0bPxdfKSq49JI+VBW032cSliyCz1zbubWtFDgU7q9hgMASVTiwI8kQjsPCnEdjrXH4Evz6Cg0136OAUBCVorfnox/387J9f8+XGfO48YwD3ntbVNSYyrmkegRXmuuubxseLdzQ1C4HLPGPlElQWtYNG0MX13NIIQARR6V5pTm8cxQbDUUvnFARae9EIIv3nEVgCoME05EMj8PQPgByrPshzS7Zx65tryEqJ4ZNbJ3HTKccRXuNmToqIa1qK2upBvNetk2dthbRr9NZA3dIIasrEnFRZ6MohaC2WIFGhrtpC4GwOr6WAndEIDIajls7pLK4tlwzgQDUCdx+BP9NQdUnjZDILZ+G5579YzdnD+vHkZaMJDXGadKrcIo0i4pqahqyd/cFdLr9GyQ455hkxBC5BUF0q5zpsjXf0rcESJGn9GjdtT+vvem40AoPhqCWoGoFS6kyl1Bal1Hal1F0+xlyslNqolPpJKfVGMOfTgGfZBZBcAn/VR61EsoaEMm8aQbFXjeBQiOQFDE6y8fcLh7uEgHUOOAVBbFPTkHu5CEsr8BUxBK7ewDVl3j9na7AEibtZCKS8hYXRCAyGo5agCQKlVCjwFHAWMBi4TCk12GNMP+APwESt9RDgN8GaTyOsMFF3p65fjaDGLWrISijzYRry8BHYHZqnf5AcgvundiM+yqOEclWxdAqLSvTuLLZMQ6GRsHeZPC9yCoKUPk3n4G4aamtWsUV8BqCk2Jw7ETGQ2FOex3VtcprBYDg6CKZGMB7YrrXeqbWuA+YAnpXOrgee0lofBNBae/REDBLuJSMswiL95xGEeTqLPcJHHQ7ZvbtpBFprHpy3iW9zNQDZ0b60iGRn5dB4LxpBmQiBrPGNNYLELFmIPXEXBFbGb1t9BNHJ8IuPpV2jJ2lO85QxDRkMRy3BFAQ9gH1ur3Ocx9zpD/RXSn2nlFqmlDrT24WUUrOUUiuVUisLCwu9DWkZXgVBtP8OZU2ihjzMSDWlUozOqRHU2Rzc8faPvPDtLk4Z6ewX4C2prKrYlYkcEdu0DHVNqZh7ep4AB9bL+8XbvTuKQfwMKlTOs1pMtofZpvdkiWryxPITGNOQwXDUcrijhsKAfsAU4DLgeaVUk/RUrfVzWuuxWuuxXbq00fEJvjUCvz4CK2rIR62hamcJiegUDtXUc83LP/D+mlxun9qf28+f4BzjpQJpVUljQeCpEVSXSkhorxNE0Oz7wSkIvPgHQPIKrOziygIxOwXSgKa19Dsduo9uHIFlMBiOKoIZNZQLZLm9znQecycHWK61rgd2KaW2IoJhRRDn5dZoxi0CptmEsqjG53hqD87on7rIZC59dhlb88t5dOYILhqTKeGqoRE+NIISSOktzyPiwFEv87Oc0jVlsrBnjpNFfdPHkizmLWLIwhIEDhvEpDXuV9DeHHe6/BgMhqOWYGoEK4B+SqneSqkI4FLgI48xHyLaAEqpNMRUtDOIcxIaNAI3x61fjcBtYQ7z4Sx27vbf/qmSjXmHeObKMSIEQHbpvuoNuXc0swrDuTuMLdNQZLw4aze8J8d9aQQggqC61FlnqB00KIPBcEwTNEGgtbYBNwOfA5uAt7XWPyml7ldKWV3TPweKlVIbgUXAnVprP4X72wlrEQ910wjC/fgIGkUN+QgfdWoEL60uY8boHkwd7OE8jUltWoFU68Y+gkgvgsAyDYH4CazSEWnNCAIraijOCAKDweCfoCaUaa3nAfM8jt3r9lwDtzt/Og57vTy2yEfgFARKeW1g76gqJgSoj0zinmmDm14jOrmpRlBbLqYgdx8BNM4utkxDAD0nwPL/yLwTs/BJdJIkn9lrIXmc73EGg8HA4XcWHx4a+gu4C4IoMRk57E3H22ubCg0PZ/GG7buxa8VvzhlLSqyXNpAxqU2dxe7JZOBmGnI6jB0OEQRWkliW0+mc0se/3b9BIyhqe+iowWA45umcgqDBNOQhCKCpycduE6er9b51npv2sLe4ik07dlMZmsAFo33s1GNSmmoEVR6NbDx9BLWHAO0yDSV0k3DNjKH+P19UopiF6irankxmMBiOeQISBEqp95VS05RSx4bg8Goa8tGu0hIaTSKMRCPYXVTJJc99T5KqIDohzXdHsZhUCTF17ytgCQYrCc0yDVmCwGowY5mGAH4+F85+xP/ni0oC7dRsTHy/wWBohkAX9qeBy4FtSqmHlFIDmjvhiMbmbXH3ERbqdWwE2GvZXlDBxc9+T63NwaTuIYTH+9l9x6RKHoB77aAG05BTEER6mIassdFuqRUJ3b2XunbHXXCYqCGDwdAMAQkCrfUCrfUVwGhgN7BAKbVUKXWNUirc/9lHIH41Ao8yE94EQWgkFRUVXPrcMhxa8+b1E4i1l3kvQW1hvVfl5ifw5SOwsou9aQSB4N4y0ggCg8HQDAGbepRSqcDVwHXAGuBfiGD4MigzCyb2WsDZp9gi3IePwDIVufkI6lUEa3cXoBTMmXUCA7rGey041whrsa/2EAQq1LXQN5iGnBqBVXCupb2AjUZgMBhaQEDho0qpD4ABwGvAuVrrPOdbbymlVgZrckHDXudqRm/h00fQuBxFRa2NPcV1hDgcvHLNeI5Ld+7iq0v8l3Kw+hS4O4ytHAJrHuExgPJvGgqEaKMRGAyGwAk0j+AJrfUib29orce243w6BltdY1MPuF7XewgCN42g3u7gptdX86t6xZCMSBK7J8h7dVUyLhCNwNM0ZB0HEQjuzWlabRpyjo9MdGk6BoPB4INATUOD3YvBKaWSlVI3BmlOwcde17i8BPjWCJzRQbaQCP7w/nqWbC2kV0YKieFu0T/VHmGg3mgQBO4aQUnTc9x7ElSXSn2hiPgAPpQbliAwoaMGgyEAAhUE12utG8JdnP0Drg/OlDoA97LSFr7yCJyC4f7523l3VQ63ntaP7qlJjRPKrF2+P2dxRByEhDf2EVR78StEuvUtrimVRT2khVG7liAwoaMGgyEAAl1hQpVbgLyz+5iX9NmjBHt9QBqB1pq1uw8AsOOgnScvG8XtU/s3SShzaQR+BIG3wnOeZdbOBQAAFaFJREFUpiFoXIravbxESwiPFkFnNAKDwRAAgfoIPkMcw886X9/gPHZ04t5fwML5urq6kne+382K3QdZsauEYRUbeT4C/n7JWDIHd3eOdSWUAa4GMP5MQ9b7lvbgcPgwDcU3jhpqacSQRWpf6DKwdecaDIZORaCC4PfI4v9r5+svgReCMqOOwIoacsfZv3jO0m38eX8SXROiGNc7hZlRXeFHyExziwhyJpQ1YLWEbK5dY0yKSxDsWy7Zv/EevX4jYqFCtJBGdYZaynULmn5Gg8Fg8EJAgkBr7QCecf4c/XgTBE7T0O78Eu4+eyDXT+4j5SLWboIfaZJQ1sg0VFkg9v/mdu8xKVCwWXwAH/4KknrC8Esaj2lkGiqV+kKtwcpJMBgMhmYINI+gH/AgMBhoiEfUWvcJ0ryCixfTUJUjlBggKz6Eqyf2dtUM8lWOwt00VFEg8frNOXWjU8Sf8PndcHAPXDMPohIaj4mMax/TkMFgMARIoM7ilxBtwAacArwK/C9Ykwo6XpzFT3+bA8CZA5MIC3X7WhoEgVs8flikmIa0ltcVBYFF6MSkSlXQ1a/AxFuh14lNx0S4Rw210llsMBgMLSBQQRCttV4IKK31Hq31bGBa8KYVZDzCR3cWVvDsN3uwEUZmnEf10IaEMg/TkHZIeWqAivzm/QPgcgxnDIVT/uh9jJVHUFcl82ytj8BgMBgCJFBnca2zBPU2pdTNSBP6uOBNK8jY66kjjCUb81m5u4TPfzpAVHgoIRFRXvoRWCUmfFQqDQ0XjaDb8ObvmzFYylBc8GzTqCWLiDhAQ7mziocxDRkMhiATqEZwGxAD3AqMAa4EftHcSUqpM5VSW5RS25VSd3l5/2qlVKFSaq3z57qWTL6l1NrsfLouj7ziMr7YcpDrX13JS9/tJi0ukscvGUlIWJSXzOIaKQwX6iYzGxrY10kYaGVhYBpBnylw507o6qexjOXkPZQrj8Y0ZDAYgkyzGoEzeewSrfXvgArgmkAu7DzvKWAqkAOsUEp9pLXe6DH0La31zS2bdst5e8U+/jpvE2XV9XwXVUfvjCTennYCwzMTiQp3tn2c700Q1Db2D4BbA/sacf5qe+AtIZtzKEc6y0mUOQWBMQ0ZDIYg06wg0FrblVKTWnHt8cB2rfVOAKXUHOA8wFMQdAgZiVGc1L8LM8dk0v3jEHpkdYHeHpnA4b4EgfdQU2y1rlLR7VXOoUEjEOc1UX4qmhoMBkM7EKiPYI1S6iPgHaDSOqi1ft/POT2AfW6vc4DjvYy7UCl1ErAV+K3Wep+XMW3m5P5dOLm/sySztzwCcGYMe6k15KkRWILBXieOYgjMNBQIliAoM6Yhg8HQMQQqCKKAYuBUt2Ma8CcIAuFj4E2tda1S6gbgFY97AKCUmgXMAujZs2cbb4n3MtQgx+q9dCjzHGs5jm01blnF7aUROE1Dh4xpyGAwdAyBZhYH5BfwIBfIcnud6Tzmfl23Cmy8ADzs4/7PAc8BjB07VrdiLo2x1zYtOgcQFu0lashfpdI6ySqG9jcNGY3AYDB0EIFmFr+EaACN0Fpf6+e0FUA/pVRvRABcClzucd1ubt3OpgObAplPm9DaaRryoRFYzWAsvBaos0xDtWIaCo2ESI8M4dZiNbA/lAPhsd4FlsFgMLQjgZqGPnF7HgVcAOz3d4LW2ubMOfgcCAVe1Fr/pJS6H1iptf4IuFUpNR3JWC5BeiIHF6txvacDGJw+goLGx7z5CBqZhpyho8ojEa21WA3sa8ogoUf7XNNgMBj8EKhp6D3310qpN4FvAzhvHjDP49i9bs//APwhoJm2Fx49iBsRFgk2Tx+Bn7aWNqezOK4d+wK7F4szZiGDwdABtLD1VQP9gKOz/ZW3TGGLcC8+AluNb0Fgr3XWGWqniCHr2iFOc5DJKjYYDB1AoD6Cchr7CA4gPQqOPhoEgTdncWQLE8pqxVmcNa595xgRKyWoTcSQwWDoAAI1DbWwe/oRjLey0hbe8gjstT57F1BfJd3JAs0qDpSIOFe/YoPBYAgyAZmGlFIXKKUS3V4nKaXOD960gojlLPaVUOY1j8AzocwpRA7tB3T7N4m3IoeMachgMHQAgfoI7tNaN8RVaq1LgfuCM6UgY7WY9CUIHPXgsLuOefMRWOeW7pXH9hYElsPYmIYMBkMHEKgg8DYu0NDTI4vmooagsXnIa9SQU0Moc9YDak9nMbhCSI1pyGAwdACBCoKVSqnHlFJ9nT+PAauCObGgYbWY9JVHAI0dxl41AqejudRZFqndNQJjGjIYDB1HoILgFqAOeAuYA9QANwVrUkHFn2ko3EMQaC3jPX0ESkn4qVUPqN2dxcY0ZDAYOo5Ao4YqgSaNZY5K/OUReGoEfs1IUVBbJmUgItu5WVukMQ0ZDIaOI9CooS+VUklur5OVUp8Hb1pBxK9pyMNH0NCvOMrLWOf57ZlVbGFpBMY0ZDAYOoBATUNpzkghALTWBznqM4v9+AisEFJ/OQeWRtHejmJwlaI2GoHBYOgAAhUEDqVUQyMApVQ2XqqRHhUEZBqqbfzoq3cBQGwQNIIu/UXAxKa1/7UNBoPBg0BDQP8IfKuUWgIoYDLORjFHHX5LTHj4CBoEgTfTUBA1giEXyI/BYDB0AIE6iz9TSo1FFv81wIdAtf+zjlAC2eV7+gi8mZGsY8EQBAaDwdCBBFp07jrgNqTL2FpgAvA9XtpKHvH4KzERHi2PVilquz+NwHksGM5ig8Fg6EAC9RHcBowD9mitTwFGAaX+TzlC8VtiwkMjqCiUx+hkL2ONRmAwGI4NAhUENVrrGgClVKTWejMwIHjTCiKBRA1ZJqGCn+QxfWDTsZazub2TyQwGg6GDCdRZnOPMI/gQ+FIpdRDYE7xpBRFbIM5ip0aQvxESs7yHcTY4i40gMBgMRzcBaQRa6wu01qVa69nAn4D/As2WoVZKnamU2qKU2q6U8pmZrJS6UCmlnQ7p4GKvE23AW49hzzyCgk2QPsj7dYwgMBgMxwgtriCqtV4SyDilVCjwFDAVyAFWKKU+0lpv9BgXj/gglrd0Lq3CXuc9hwAa+wjs9VC0FfpN9T42NBIiE1wOZoPBYDhKaW3P4kAYD2zXWu/UWtchxerO8zLuL8DfkUJ2wcdW6728BEBIqPQLttVA8XbpTZAxxPvYYRfB5NuDN0+DwWDoIIIpCHoA+9xe5ziPNaCUGg1kaa0/DeI8GmOZhnwRHi2CIN9yFPswDR13Gkz6bfvPz2AwGDqYYAoCvyilQoDHgDsCGDtLKbVSKbWysLCwbTduThBYDewLNoIKhbT+bbufwWAwHOEEUxDkAllurzOdxyzigaHAYqXUbiRJ7SNvDmOt9XNa67Fa67FdurQxgatZQeBsYF+wCVKP856BbDAYDMcQwRQEK4B+SqneSqkI4FLgI+tNrXWZ1jpNa52ttc4GlgHTtdYrgzgn760n3bE0gvyfIGNwUKdiMBgMRwJBEwRaaxtwM/A5sAl4W2v9k1LqfqXU9GDdt1nsdd5zCCzCoqGyCEr3QLoPR7HBYDAcQwS1Ab3Weh4wz+PYvT7GTgnmXBqw1/oOHwXRCPLWynNfjmKDwWA4hjhszuLDhr2+GY0gCmrK5LkxDRkMhk5A5xMEttrmfQQA4TGQlN0hUzIYDIbDSecTBPb65vMIALoMhJDO9/UYDIbOR+db6ey1zecRgDELGQyGTkPnEwTNmoachefSjSAwGAydg84nCAJxFoMRBAaDodPQCQVBc+GjTkHgq9icwWAwHGMENY/giKQ5Z3H2RCjbB7GmF7HBYOgcdD5B4K8MNcDAafJjMBgMnYTOZRrSuvmicwaDwdDJ6FyCwGEDtH8fgcFgMHQyOpcgsPtpXG8wGAydlM4lCGy18mh6DBgMBkMDnUsQ2Ovl0fgIDAaDoYFOJgicGoERBAaDwdBA5xIENqePwJiGDAaDoYHOJQiMs9hgMBia0MkEgWUaMhqBwWAwWARVECilzlRKbVFKbVdK3eXl/V8ppdYrpdYqpb5VSgW30ptxFhsMBkMTgiYIlFKhwFPAWcBg4DIvC/0bWuthWuuRwMPAY8GaD+AWPmoEgcFgMFgEUyMYD2zXWu/UWtcBc4Dz3AdorQ+5vYwFdBDn4+YjMILAYDAYLIJZdK4HsM/tdQ5wvOcgpdRNwO1ABHBqEOdjBIHBYDB44bA7i7XWT2mt+wK/B+7xNkYpNUsptVIptbKwsLD1NzOCwGAwGJoQTEGQC2S5vc50HvPFHOB8b29orf+/vbuLkauuwzj+fWytvNQI6EqwRSjSiNXwoitBUSTQCxBSuAAFwaDBcEMj+BIFXzA28UIxohdE2wCmRpSXCrExVdRKiFwAXSiibSU0CFJS7BoBReNud3m8OGfas7NTqG3PTNv/80mamfOyM/85/c8+e/7nnN9ZZnvY9vDQ0G7cJyDXEURETNNmEKwB5kuaJ2kWcBGwsrmCpPmNyXOAJ1psT/YIIiJ6aO0Yge0JSYuBe4AZwC2210laAozYXgkslrQQ2Ao8D1zWVnuAlJiIiOih1TuU2V4FrOqad13j+VVtvv80GRqKiJhm4AeL+yolJiIipiksCFJiIiKiW2FB0CkxkT2CiIiOsoJgYqw6UCwNuiUREXuNsoJgcmvOGIqI6FJYEIwlCCIiuhQWBOMJgoiILmUFwcR4SlBHRHQpKwgmx3PqaERElwKDIHsEERFN5QVBhoYiIqYoKwgmctZQRES3soIgQ0MREdMkCCIiCldWEEyMpwR1RESXsoJgcjwF5yIiuhQWBGO5jiAiokthQZCicxER3coKgomxXEcQEdGl1SCQdJakxyVtlHRNj+WflbRe0mOSVks6qs32VHsEGRqKiGhqLQgkzQBuBM4GFgAXS1rQtdpaYNj28cAK4FtttQeojxHkYHFERFObewQnAxttP2l7HLgNOK+5gu17bf+nnnwAmNtie+oSE9kjiIhoajMI5gDPNKY31fN25HLgl70WSLpC0oikkdHR0V1rzeQE+OUcLI6I6LJXHCyWdCkwDFzfa7ntZbaHbQ8PDQ3t2ptMjlWPCYKIiClmtvjazwJHNqbn1vOmkLQQ+DLwIdtjrbVmcrx6TBBEREzR5h7BGmC+pHmSZgEXASubK0g6CVgKLLK9pcW2VOUlIKePRkR0aS0IbE8Ai4F7gA3AHbbXSVoiaVG92vXAbOBOSY9KWrmDl9t92SOIiOipzaEhbK8CVnXNu67xfGGb7z/FtiDIWUMREU17xcHivtgWBLmOICKiqZwgmKiPQ+c6goiIKcoJgsmt1WOGhiIipigoCDrXEWRoKCKiqaAg6Jw+mj2CiIimcoJgIgeLIyJ6KScItg0NZY8gIqKpoCDoHCzOBWUREU3lBMG200cTBBERTeUEQUpMRET0VGAQ5BhBRERTgUGQs4YiIprKCYLDjoEF58HMAwbdkoiIvUqr1Uf3KsedU/2LiIgpytkjiIiInhIEERGFSxBERBQuQRARUbhWg0DSWZIel7RR0jU9lp8m6RFJE5IuaLMtERHRW2tBIGkGcCNwNrAAuFjSgq7V/gp8AvhJW+2IiIhX1ubpoycDG20/CSDpNuA8YH1nBdtP1ctebrEdERHxCtocGpoDPNOY3lTP+79JukLSiKSR0dHRPdK4iIio7BMXlNleBiwDkDQq6eldfKk3AX/fYw3bd2U7bJdtUcl2qOzP2+GoHS1oMwieBY5sTM+t5+0W20O7+rOSRmwP724b9nXZDttlW1SyHSqlboc2h4bWAPMlzZM0C7gIWNni+0VExC5oLQhsTwCLgXuADcAdttdJWiJpEYCk90raBFwILJW0rq32REREb60eI7C9CljVNe+6xvM1VENG/bKsj++1N8t22C7bopLtUClyO8j2oNsQEREDlBITERGFSxBERBSumCB4tbpH+ytJR0q6V9J6SeskXVXPP0zSbyQ9UT8eOui29oOkGZLWSvpFPT1P0oN1v7i9PsNtvybpEEkrJP1Z0gZJ7yuxP0j6TP2d+JOkn0o6oMT+AIUEwU7WPdpfTQCfs70AOAW4sv7s1wCrbc8HVtfTJbiK6iy2jm8CN9g+FngeuHwgreqv7wG/sn0ccALV9iiqP0iaA3waGLb9LmAG1SnuJfaHMoKARt0j2+NAp+7Rfs/2ZtuP1M//RfWln0P1+ZfXqy0Hzh9MC/tH0lzgHOCmelrAGcCKepX9fjtIegNwGnAzgO1x2y9QYH+gOmvyQEkzgYOAzRTWHzpKCYI9VvdoXybpaOAk4EHgcNub60XPAYcPqFn99F3gC0CnyOEbgRfqa16gjH4xDxgFflgPkd0k6WAK6w+2nwW+TVUBeTPwIvAw5fUHoJwgKJ6k2cDPgKtt/7O5zNU5xPv1ecSSzgW22H540G0ZsJnAu4Hv2z4J+Dddw0CF9IdDqfaC5gFvAQ4GzhpoowaolCBope7RvkLSa6lC4Fbbd9Wz/ybpiHr5EcCWQbWvT04FFkl6impo8AyqsfJD6qEBKKNfbAI22X6wnl5BFQyl9YeFwF9sj9reCtxF1UdK6w9AOUFQbN2jehz8ZmCD7e80Fq0ELqufXwb8vN9t6yfb19qea/toqv//39m+BLgX6Nwdr4Tt8BzwjKS317POpLpHSFH9gWpI6BRJB9Xfkc52KKo/dBRzZbGkD1ONEc8AbrH9jQE3qS8kfQD4PfBHto+Nf4nqOMEdwFuBp4GP2P7HQBrZZ5JOBz5v+1xJx1DtIRwGrAUutT02yPa1TdKJVAfMZwFPAp+k+qOwqP4g6evAR6nOrFsLfIrqmEBR/QEKCoKIiOitlKGhiIjYgQRBREThEgQREYVLEEREFC5BEBFRuARBRB9JOr1T+TRib5EgiIgoXIIgogdJl0p6SNKjkpbW9zF4SdINdQ371ZKG6nVPlPSApMck3d2p5S/pWEm/lfQHSY9Ielv98rMb9wO4tb6yNWJgEgQRXSS9g+qK01NtnwhMApdQFSYbsf1O4D7ga/WP/Aj4ou3jqa7g7sy/FbjR9gnA+6mqXEJVAfZqqntjHENV4yZiYGa++ioRxTkTeA+wpv5j/UCqImwvA7fX6/wYuKuu73+I7fvq+cuBOyW9Hphj+24A2/8FqF/vIdub6ulHgaOB+9v/WBG9JQgiphOw3Pa1U2ZKX+1ab1frszRr10yS72EMWIaGIqZbDVwg6c2w7f7OR1F9XzqVKT8G3G/7ReB5SR+s538cuK++G9wmSefXr/E6SQf19VNE7KT8JRLRxfZ6SV8Bfi3pNcBW4Eqqm7icXC/bQnUcAapyxT+of9F3qnlCFQpLJS2pX+PCPn6MiJ2W6qMRO0nSS7ZnD7odEXtahoYiIgqXPYKIiMJljyAionAJgoiIwiUIIiIKlyCIiChcgiAionD/A6LUWOswoGrFAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "YaFg1Shuub_L"
      },
      "source": [
        "As we can see, the model clearly overfits since the train accuracy increases while the test accuracy stays more or less the same."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XY4ZsqBS6O-0"
      },
      "source": [
        "###Evaluation of the SVM"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "fFf7ZJSp6RBQ",
        "outputId": "045ce633-471d-4927-c835-45fefa6e2362"
      },
      "source": [
        "features_svm = np.load('./drive/MyDrive/COUGHVID/mfccfeatures.npy', allow_pickle=True)\n",
        "labels_svm = np.load('./drive/MyDrive/COUGHVID/mfccfeatureslabels.npy', allow_pickle=True)\n",
        "print(features_svm.shape)\n",
        "print(labels_svm.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "(23490, 256)\n",
            "(23490,)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ZlCQnYY97nz3",
        "outputId": "06ad6480-504a-43f8-eb1e-19a416a23c1f"
      },
      "source": [
        "np.unique(labels_svm)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "text/plain": [
              "array(['COVID-19', 'healthy', 'symptomatic'], dtype=object)"
            ]
          },
          "execution_count": 17,
          "metadata": {},
          "output_type": "execute_result"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "6k_rE0JN6cAh",
        "outputId": "66cf9074-c7dc-40ba-9158-418fd7c4c698"
      },
      "source": [
        "## calculate class weights\n",
        "class_weights_svm = compute_class_weight('balanced', np.unique(labels_svm), labels_svm)\n",
        "class_weights_dict_svm = {'COVID-19':class_weights_svm[0], 'healthy':class_weights_svm[1], 'symptomatic':class_weights_svm[2]}\n",
        "\n",
        "class_weights_dict_svm"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "text/plain": [
              "{'COVID-19': 4.042333505420753,\n",
              " 'healthy': 0.4403824521934758,\n",
              " 'symptomatic': 2.0752716671084017}"
            ]
          },
          "execution_count": 19,
          "metadata": {},
          "output_type": "execute_result"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fxcTHwJq61rd"
      },
      "source": [
        "X_train_svm, X_test_svm, y_train_svm, y_test_svm = train_test_split(features_svm, labels_svm, test_size=0.2, random_state=0)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4n_ANQF46-e4"
      },
      "source": [
        "svm = SVM(classWeights=class_weights_dict_svm)\n",
        "svm.fit(X_train_svm, y_train_svm)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DsZUhbH77Jjy"
      },
      "source": [
        "predictions = []\n",
        "for x in X_test_svm:\n",
        "  X = np.expand_dims(x, axis=0)\n",
        "  predictions.append(svm.predict(X)[0])\n",
        "predictions_oneHot = oneHotEncoding(predictions)\n",
        "y_test_svm_oneHot = oneHotEncoding(y_test_svm)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 175
        },
        "id": "x1s7dbRh7PhT",
        "outputId": "8e1d5903-a10c-4eac-f847-d0d8a274a654"
      },
      "source": [
        "results = getAllMetrics(y_test_svm_oneHot, predictions_oneHot)\n",
        "results"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Class</th>\n",
              "      <th>F1 Score</th>\n",
              "      <th>Sensitivity</th>\n",
              "      <th>Specificity</th>\n",
              "      <th>Precision</th>\n",
              "      <th>Recall</th>\n",
              "      <th>Accuracy</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>Overall</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>0.629333</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>Healthy</td>\n",
              "      <td>57.5944</td>\n",
              "      <td>43.8503</td>\n",
              "      <td>71.9251</td>\n",
              "      <td>83.8875</td>\n",
              "      <td>43.8503</td>\n",
              "      <td>-</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>COVID</td>\n",
              "      <td>19.5251</td>\n",
              "      <td>41.573</td>\n",
              "      <td>70.7865</td>\n",
              "      <td>12.7586</td>\n",
              "      <td>41.573</td>\n",
              "      <td>-</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>Symptomatic</td>\n",
              "      <td>32.7801</td>\n",
              "      <td>48.4663</td>\n",
              "      <td>74.2331</td>\n",
              "      <td>24.7649</td>\n",
              "      <td>48.4663</td>\n",
              "      <td>-</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "         Class F1 Score Sensitivity Specificity Precision   Recall  Accuracy\n",
              "0      Overall        -           -           -         -        -  0.629333\n",
              "1      Healthy  57.5944     43.8503     71.9251   83.8875  43.8503         -\n",
              "2        COVID  19.5251      41.573     70.7865   12.7586   41.573         -\n",
              "3  Symptomatic  32.7801     48.4663     74.2331   24.7649  48.4663         -"
            ]
          },
          "execution_count": 26,
          "metadata": {},
          "output_type": "execute_result"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 428
        },
        "id": "kaLA3ReT9ZE6",
        "outputId": "888e1ff7-a762-4c57-93d2-6a12bab752e1"
      },
      "source": [
        "matrix = confusion_matrix(np.argmax(oneHotEncoding(y_test_svm), axis=1), np.argmax(oneHotEncoding(predictions), axis=1), normalize=\"true\")\n",
        "labellist = ['healthy','covid', 'symptomatic']\n",
        "plotConfusionMatrix(labellist, matrix)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x504 with 2 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Qb7XEpK2NUAt"
      },
      "source": [
        "##Evaluation with balanced datasets"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "VeeEMZS7N4of",
        "outputId": "896058f7-854a-4c96-88bf-4bb0940a8197"
      },
      "source": [
        "covid_indizes = np.where(labels=='COVID-19')\n",
        "print('COVID Samples: ' + str(len(covid_indizes[0])))\n",
        "symptom_indizes = np.where(labels=='symptomatic')\n",
        "print('Symptomatic Samples: ' + str(len(symptom_indizes[0])))\n",
        "healthy_indizes = np.where(labels=='healthy')\n",
        "print('Healthy Samples: ' + str(len(healthy_indizes[0])))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "COVID Samples: 1984\n",
            "Symptomatic Samples: 3904\n",
            "Healthy Samples: 18486\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "oZTmRJR9Oazv"
      },
      "source": [
        "As we can see the dataset is highly unbalanced. Therefor we take from each class only the amount of samples that we have available of the most underrepresented class."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Z_F9A0gpxvJ0"
      },
      "source": [
        "##Evaluation of DTLMC"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "v90l-snQNha6",
        "outputId": "83c602c3-7632-48b9-cbc5-f70470f7b26d"
      },
      "source": [
        "data_new = dataSegmented[covid_indizes]\n",
        "data_new = np.append(data_new,dataSegmented[symptom_indizes[0][:len(covid_indizes[0])]], axis=0)\n",
        "data_new = np.append(data_new,dataSegmented[healthy_indizes[0][:len(covid_indizes[0])]], axis=0)\n",
        "\n",
        "data_new = np.expand_dims(data_new, axis=3)\n",
        "\n",
        "labels_new = labels[covid_indizes]\n",
        "labels_new = np.append(labels_new,labels[symptom_indizes[0][:len(covid_indizes[0])]])\n",
        "labels_new = np.append(labels_new,labels[healthy_indizes[0][:len(covid_indizes[0])]])\n",
        "\n",
        "labels_new = oneHotEncoding(labels_new)\n",
        "\n",
        "print(labels_new.shape)\n",
        "print(data_new.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "(5952, 3)\n",
            "(5952, 240, 320, 1)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "X4yDrOnlOqta"
      },
      "source": [
        "# split data for train and test validation\n",
        "X_train, X_test, y_train, y_test = train_test_split(data_new, labels_new, test_size=0.2, random_state=0)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gTwbkVB4Outx"
      },
      "source": [
        "lr_schedule = tf.keras.optimizers.schedules.ExponentialDecay(\n",
        "    initial_learning_rate=0.00001,\n",
        "    decay_steps=10000,\n",
        "    decay_rate=0.9)\n",
        "\n",
        "cb = tf.keras.callbacks.EarlyStopping(monitor='val_loss', min_delta=0.01, patience=25, verbose=1,\n",
        "    mode='auto', baseline=None, restore_best_weights=True)\n",
        "\n",
        "dtlmc = DTLMC()\n",
        "opt = tf.keras.optimizers.Adam(learning_rate=lr_schedule)\n",
        "dtlmc.compile(optimizer=opt,loss='categorical_crossentropy',metrics=['acc'])\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "jn5XRSeuOx22",
        "outputId": "396be704-072b-4548-8bf0-6f24fcd42070"
      },
      "source": [
        "history = dtlmc.fit(X_train,y_train, epochs=250, validation_data=(X_test,y_test), callbacks=[cb])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/250\n",
            "149/149 [==============================] - 6s 36ms/step - loss: 3.8623 - acc: 0.3333 - val_loss: 1.2285 - val_acc: 0.3703\n",
            "Epoch 2/250\n",
            "149/149 [==============================] - 4s 28ms/step - loss: 2.2516 - acc: 0.3422 - val_loss: 1.1260 - val_acc: 0.3686\n",
            "Epoch 3/250\n",
            "149/149 [==============================] - 4s 28ms/step - loss: 1.7151 - acc: 0.3394 - val_loss: 1.1063 - val_acc: 0.3594\n",
            "Epoch 4/250\n",
            "149/149 [==============================] - 4s 28ms/step - loss: 1.4115 - acc: 0.3516 - val_loss: 1.0934 - val_acc: 0.3787\n",
            "Epoch 5/250\n",
            "149/149 [==============================] - 4s 28ms/step - loss: 1.2550 - acc: 0.3693 - val_loss: 1.0904 - val_acc: 0.3661\n",
            "Epoch 6/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 1.1921 - acc: 0.3579 - val_loss: 1.0890 - val_acc: 0.3921\n",
            "Epoch 7/250\n",
            "149/149 [==============================] - 4s 28ms/step - loss: 1.1441 - acc: 0.3703 - val_loss: 1.0899 - val_acc: 0.3728\n",
            "Epoch 8/250\n",
            "149/149 [==============================] - 4s 28ms/step - loss: 1.1156 - acc: 0.3875 - val_loss: 1.0880 - val_acc: 0.3787\n",
            "Epoch 9/250\n",
            "149/149 [==============================] - 4s 28ms/step - loss: 1.1062 - acc: 0.3879 - val_loss: 1.0880 - val_acc: 0.3812\n",
            "Epoch 10/250\n",
            "149/149 [==============================] - 4s 28ms/step - loss: 1.0997 - acc: 0.3968 - val_loss: 1.0856 - val_acc: 0.3887\n",
            "Epoch 11/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 1.0871 - acc: 0.4098 - val_loss: 1.0842 - val_acc: 0.3871\n",
            "Epoch 12/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 1.0784 - acc: 0.4161 - val_loss: 1.0835 - val_acc: 0.3971\n",
            "Epoch 13/250\n",
            "149/149 [==============================] - 4s 28ms/step - loss: 1.0616 - acc: 0.4333 - val_loss: 1.0806 - val_acc: 0.4022\n",
            "Epoch 14/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 1.0608 - acc: 0.4302 - val_loss: 1.0781 - val_acc: 0.3887\n",
            "Epoch 15/250\n",
            "149/149 [==============================] - 4s 28ms/step - loss: 1.0491 - acc: 0.4507 - val_loss: 1.0770 - val_acc: 0.4030\n",
            "Epoch 16/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 1.0508 - acc: 0.4480 - val_loss: 1.0772 - val_acc: 0.3980\n",
            "Epoch 17/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 1.0422 - acc: 0.4541 - val_loss: 1.0750 - val_acc: 0.4005\n",
            "Epoch 18/250\n",
            "149/149 [==============================] - 4s 28ms/step - loss: 1.0332 - acc: 0.4608 - val_loss: 1.0731 - val_acc: 0.4047\n",
            "Epoch 19/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 1.0257 - acc: 0.4764 - val_loss: 1.0709 - val_acc: 0.4198\n",
            "Epoch 20/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 1.0219 - acc: 0.4862 - val_loss: 1.0723 - val_acc: 0.4131\n",
            "Epoch 21/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 1.0233 - acc: 0.4825 - val_loss: 1.0726 - val_acc: 0.4081\n",
            "Epoch 22/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 1.0044 - acc: 0.4974 - val_loss: 1.0693 - val_acc: 0.4106\n",
            "Epoch 23/250\n",
            "149/149 [==============================] - 4s 28ms/step - loss: 1.0014 - acc: 0.4854 - val_loss: 1.0718 - val_acc: 0.4005\n",
            "Epoch 24/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.9969 - acc: 0.5026 - val_loss: 1.0734 - val_acc: 0.4139\n",
            "Epoch 25/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.9840 - acc: 0.5203 - val_loss: 1.0707 - val_acc: 0.4139\n",
            "Epoch 26/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.9814 - acc: 0.5222 - val_loss: 1.0746 - val_acc: 0.4198\n",
            "Epoch 27/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.9723 - acc: 0.5224 - val_loss: 1.0750 - val_acc: 0.4114\n",
            "Epoch 28/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.9687 - acc: 0.5295 - val_loss: 1.0696 - val_acc: 0.4030\n",
            "Epoch 29/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.9586 - acc: 0.5360 - val_loss: 1.0699 - val_acc: 0.4139\n",
            "Epoch 30/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.9482 - acc: 0.5472 - val_loss: 1.0711 - val_acc: 0.4064\n",
            "Epoch 31/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.9417 - acc: 0.5524 - val_loss: 1.0674 - val_acc: 0.4131\n",
            "Epoch 32/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.9407 - acc: 0.5581 - val_loss: 1.0697 - val_acc: 0.4139\n",
            "Epoch 33/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.9274 - acc: 0.5610 - val_loss: 1.0692 - val_acc: 0.4039\n",
            "Epoch 34/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.9240 - acc: 0.5591 - val_loss: 1.0686 - val_acc: 0.4165\n",
            "Epoch 35/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.9184 - acc: 0.5644 - val_loss: 1.0745 - val_acc: 0.4190\n",
            "Epoch 36/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.9104 - acc: 0.5677 - val_loss: 1.0816 - val_acc: 0.4047\n",
            "Epoch 37/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.8987 - acc: 0.5789 - val_loss: 1.0860 - val_acc: 0.4307\n",
            "Epoch 38/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.8972 - acc: 0.5839 - val_loss: 1.0780 - val_acc: 0.4282\n",
            "Epoch 39/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.8873 - acc: 0.5864 - val_loss: 1.0813 - val_acc: 0.4173\n",
            "Epoch 40/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.8773 - acc: 0.6016 - val_loss: 1.0743 - val_acc: 0.4274\n",
            "Epoch 41/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.8687 - acc: 0.6003 - val_loss: 1.0855 - val_acc: 0.4291\n",
            "Epoch 42/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.8600 - acc: 0.6070 - val_loss: 1.0783 - val_acc: 0.4257\n",
            "Epoch 43/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.8530 - acc: 0.6144 - val_loss: 1.0842 - val_acc: 0.4198\n",
            "Epoch 44/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.8452 - acc: 0.6175 - val_loss: 1.1040 - val_acc: 0.4207\n",
            "Epoch 45/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.8519 - acc: 0.6116 - val_loss: 1.0840 - val_acc: 0.4139\n",
            "Epoch 46/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.8361 - acc: 0.6268 - val_loss: 1.0894 - val_acc: 0.4215\n",
            "Epoch 47/250\n",
            "149/149 [==============================] - 4s 29ms/step - loss: 0.8149 - acc: 0.6415 - val_loss: 1.1085 - val_acc: 0.4223\n",
            "Restoring model weights from the end of the best epoch.\n",
            "Epoch 00047: early stopping\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RlV4BJy1PqHd"
      },
      "source": [
        "predictions = []\n",
        "for x in X_test:\n",
        "  predArray = [0,0,0]\n",
        "  x = np.expand_dims(x, axis=0)\n",
        "  prediction = dtlmc.predict(x)\n",
        "  predArray[np.argmax(prediction[0])] = 1\n",
        "  predictions.append(predArray)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 175
        },
        "id": "bfXBCAtSPxOk",
        "outputId": "d72830d7-a6de-4845-d7c8-9a797c157232"
      },
      "source": [
        "df = getAllMetrics(y_test, predictions)\n",
        "df"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Class</th>\n",
              "      <th>F1 Score</th>\n",
              "      <th>Sensitivity</th>\n",
              "      <th>Specificity</th>\n",
              "      <th>Precision</th>\n",
              "      <th>Recall</th>\n",
              "      <th>Accuracy</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>Overall</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>0.607053</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>Healthy</td>\n",
              "      <td>42.9907</td>\n",
              "      <td>46.3476</td>\n",
              "      <td>73.1738</td>\n",
              "      <td>40.0871</td>\n",
              "      <td>46.3476</td>\n",
              "      <td>-</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>COVID</td>\n",
              "      <td>35.9264</td>\n",
              "      <td>33.5092</td>\n",
              "      <td>66.7546</td>\n",
              "      <td>38.7195</td>\n",
              "      <td>33.5092</td>\n",
              "      <td>-</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>Symptomatic</td>\n",
              "      <td>43.4676</td>\n",
              "      <td>42.8916</td>\n",
              "      <td>71.4458</td>\n",
              "      <td>44.0594</td>\n",
              "      <td>42.8916</td>\n",
              "      <td>-</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "         Class F1 Score Sensitivity Specificity Precision   Recall  Accuracy\n",
              "0      Overall        -           -           -         -        -  0.607053\n",
              "1      Healthy  42.9907     46.3476     73.1738   40.0871  46.3476         -\n",
              "2        COVID  35.9264     33.5092     66.7546   38.7195  33.5092         -\n",
              "3  Symptomatic  43.4676     42.8916     71.4458   44.0594  42.8916         -"
            ]
          },
          "execution_count": 26,
          "metadata": {},
          "output_type": "execute_result"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 428
        },
        "id": "hI9QEgy8qzy7",
        "outputId": "e061a258-7828-465d-aa38-bae3b295a5b4"
      },
      "source": [
        "matrix = confusion_matrix(np.argmax(y_test, axis=1), np.argmax(predictions, axis=1), normalize=\"true\")\n",
        "labellist = ['healthy','covid', 'symptomatic']\n",
        "plotConfusionMatrix(labellist, matrix)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAjAAAAGbCAYAAADawqrfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO3dd5weVdXA8d/ZTUIJnVCTYKgivYsIgtQgVVB6ExEREBTxFRVBERRQsSAqQQVUmopIgFCFRCyU0IuUEJA0ICHUkIRk97x/PJPlSdjyJG6bPL8vn/lk5s6dec4my+bknntnIjORJEkqk4aeDkCSJGl+mcBIkqTSMYGRJEmlYwIjSZJKxwRGkiSVTp+u/oBZU8a6zEmdasCQXXo6BC1Etl52nZ4OQQuhW8fdHN35eZ35d23fAWt0a+wLyhEYSZJUOl0+AiNJkrpYc1NPR9DtHIGRJEml4wiMJElll809HUG3M4GRJKnsmusvgbGEJEmSSscRGEmSSi4tIUmSpNKxhCRJktT7OQIjSVLZWUKSJEml44PsJEmSej9HYCRJKjtLSJIkqXRchSRJktT7OQIjSVLJ+SA7SZJUPpaQJEmSej9HYCRJKrs6LCE5AiNJUtk1N3Xe1oGIGBoRT0fEmIg4rZ1++0dERsQWxfGQiJgeEQ8X26+q+m4eEY8V9/xZRERHcTgCI0mSahIRjcBFwC7AeOD+iBiemU/O029J4GTg3nlu8VxmbtLKrX8JfK7oPwIYCtzcXiyOwEiSVHbZ3Hlb+7YCxmTm2Mx8F7ga2KeVft8FzgNmdHTDiFgFWCoz78nMBH4H7NvRdSYwkiSVXXNz523tGwiMqzoeX7S1iIjNgMGZeVMr168eEQ9FxKiI2K7qnuPbu2drLCFJkqQWEXEscGxV07DMHFbjtQ3ABcBRrZyeBKyWma9GxObAXyNi/QWN0wRGkqSy68RVSEWy0lbCMgEYXHU8qGibY0lgA2BkMQ93ZWB4ROydmaOBmcVnPBARzwHrFNcPaueerbKEJElS2XVfCel+YO2IWD0i+gEHAcPnnMzMNzJzQGYOycwhwD3A3pk5OiJWKCYBExFrAGsDYzNzEvBmRGxdrD46Ari+o0AcgZEkSTXJzNkRcSJwK9AI/DYzn4iIs4DRmTm8ncs/BpwVEbOAZuC4zJxanDseuAxYjMrqo3ZXIIEJjCRJpZfZ8fNbOu+zcgSVpc7VbWe00XeHqv1rgWvb6DeaSumpZiYwkiSVnU/ilSRJ6v0cgZEkqezq8G3UJjCSJJVdHZaQTGAkSSq7Gl7CuLBxDowkSSodR2AkSSo7S0iSJKl06nASryUkSZJUOo7ASJJUdpaQJElS6VhCkiRJ6v0cgZEkqezqcATGBEaSpJLrzrdR9xaWkCRJUuk4AiNJUtlZQpIkSaVTh8uoLSFJkqTSqWkEJiI2zMzHujoYSZK0ACwhtekXEbEIcBlwRWa+0XUhSZKk+WIJqXWZuR1wKDAYeCAiroyIXbo0MkmSpDbUPIk3M5+NiNOB0cDPgE0jIoBvZOZfuipASZLUAUtIrYuIjYDPAHsAtwN7ZeaDEbEq8G/ABEaSpJ5ShyWkWkdgLgR+Q2W0ZfqcxsycWIzKSJIkdZuaEpjM3L6dc7/vvHAkSdJ8s4TUuoj4KPBt4APFNQFkZq7RdaFJkqSamMC06TfAl4EHgPp7Y5QkSepVak1g3sjMm7s0EkmStGCcxDu3iNis2L0rIn5AZbXRzDnnM/PBLoxNkiTVwhLS+/xonuMtqvYT2LFzw5EkSepYuwlMZn4cICLWyMyx1eciwgm8kiT1BnVYQqr1bdR/bqXtT50ZyMLsH/eMZs+DjmH3A47m17//Y5v9br/rH2zw0d15/D/PtLQ9PeZ5Dj32y+xz6Of55OFfYObMd7sjZPVyO+38MUY/eDsPPXInXz7l8+87f8KJR3Pv6Fv45z03MfzG3zN48Kot56a+8Qx3/+sG7v7XDVx1zcXdGbZ6sS122Jxfj7yES+/+DQcc/+n3nd/vc59k2N8u5pe3/YJzr/o+Kw5cseXcZ79xNMPu+BWX3HkxX/jOcd0ZtuZobu68rSQ6mgOzLrA+sHRE7Fd1ailg0a4MbGHR1NTE2T+6iEt+8j1WXnEABx5zMh/f9sOsufoH5uo3bdo7/OFP17PReh9saZs9u4nTzjqf73/rq6y79hq8/sab9OnT2N1fgnqZhoYGfnTBt9l37yOZMOEl7vr7dYwY8TeefmpMS59HH32SHbbbl+nTZ/DZYw7hrLNP4zNHngTA9Okz2G6bvXoqfPVCDQ0NnHD2CXz9kG8wZdIULrzxp9xz+728+OyLLX2ee/w5vrjHScycMZM9D9+DY755NN87/lzW2/xDrL/Fehy36/EA/OgvP2SjrTfk0Xse66kvR3WioxGYDwJ7AssAe1VtmwGf69rQFg6P/ecZVhu0KoMHrkLfvn3ZfaftufPue97X78JLfsfRh32afov0a2n7130PsM6aq7Pu2pVq3TJLL0VjowlMvdt8i40ZO/a/vPDCOGbNmsVf/nwje+yx81x97v77PUyfPgOA++97mFVXXbknQlVJfHCTdZj4wkReevElZs+azcjho/jIrlvP1eeRfz/KzBmVNRz/efApBqw8AIDMpN8i/ejTrw99+/WlT99GXpvyerd/DXUvmztvK4mO5sBcD1wfER/JzH93U0wLlVcmT2HlFVdoOV5pxQE89sTTc/V58ukxvPTKFLbfZisuvfK9at1/x00gIjj2y9/ktdffYPedt+foQ98/tKv6suqqKzFh/KSW4wkTXmKLLTdus//hR36a228f1XK86KKLMPLvf2X27Nn8+IKLuenG27s0XvV+y688gMkTJ7ccT5k0hXU3/WCb/YcetCv3jxwNVJKZR/79KFeNvoKIYPjlNzBuzLguj1nzKFHpp7N0VEK6kMpqIyLi4HnPZ+ZJbVx3LHAswC9+dDbHHPG+S1Vobm7m/AuHcc43v/K+c7Obmnjo0Se4+tc/ZdFFF+GYk77Oeh9ci6232LQHIlUZHXDgPmy66YZ8YughLW0bfOhjTJr0MkOGDGb4TX/gySee5vnnX2znLtJ7dvzkx1l7o3X46qf/D4BVh6zC4LUGc+hWhwPw/Su/xwZbrc/j9z3Rk2GqC0XEUOCnQCPw68w8t41++1OZQ7tlZo6OiF2Ac4F+wLvAVzPzzqLvSGAVYM77FnfNzFfai6OjZdSja/ty5paZw4BhALOmjM0FucfCYsUVBvDSK+/9y+blV6aw4grLtxxPe2c6Y8b+l8+cWPlhMGXqa3zxa9/hwvPOZKUVB7D5xhuw7DJLA7DdR7bkyaefM4GpcxMnvszAQau0HA8cuDKTJr78vn477LANp/7f8Xxi6CG8++57k78nTar0feGFcfzj7nvZaOP1TGDq3KsvTWGFVd8bKR6wygCmvPTq+/ptuu0mHPzFgzj10//HrHdnAbDNbtvw1ENPMeOdSsly9F2j+dBmHzKB6W7dNAITEY3ARcAuwHjg/ogYnplPztNvSeBk4N6q5inAXsWLoDcAbgUGVp0/NDNrzjvanQOTmZe3t9X6IfVsg3XX4cXxExk/8SVmzZrFzX8bxce3fa+2vOQS/fnHiGu47drLue3ay9lo/XW58Lwz2eBD6/DRrTbn2bEvMH3GDGbPbmL0w4+x5uqr9eBXo97gwQceZc01h/CBDwyib9++7PepPRkx4m9z9dloo/X4yc/O5qADPs+Uye/9RbTMMkvRr19lntVyyy/L1ltvzlNVk39Vn55+5BkGDlmVlQavRJ++fdhh7+255/a55+qtuf6anHTuSZx59Hd449U3WtonT5zMRh/ekIbGBhr7NLLh1hvyoiWk7pfZeVv7tgLGZObYzHwXuBrYp5V+3wXOA2a8F2I+lJkTi8MngMUiYpEF/ZJrfZnjCsDXgPWoWn2UmT7IrgN9+jTyjS9/gc+fcjpNTU18cs9dWWuND/DzS37H+uuuw8e327rNa5deakmOOGg/DvrsyUQE231kS7bfZqtujF69UVNTE6d+5Tv85a+X0djYwB9+/2ee+s+zfOP0L/HQg49x84i/8d1zTqP/Ev25/PcXAjB+3EQOPvDzrPPBtfjJz86mubmZhoYGfnzBr+ZavaT61NzUzEXf+iXf+8PZNDQ2cts1t/HfZ17kiK8czjOPPsM9t9/L5775WRZbfFFO/9U3AHhl4mS+ffR3uPumf7DxNhtz8e2/JBNGjxrNvXfc28EnqjerngZSGFZUVqAyYlKdoY4HPjzP9ZsBgzPzpoj4ahsfsz/wYGbOrGq7NCKagGuBszPbz6aig/NzgrkNuAY4FTgOOBKYnJlf6+jaei8hqfMNGLJLT4eghcjWy67T0yFoIXTruJujOz9v+lVndtrftYsd/J02Y4+ITwFDM/OY4vhw4MOZeWJx3ADcCRyVmS8Uc1tOrS4NRcT6wHAq81yeK9oGZuaEovR0LfCHzPxde3HW+iC75TPzN8CszByVmUfjawQkSeoduu9BdhOAwVXHg4q2OZYENgBGRsQLwNbA8IjYAiAiBgHXAUfMSV4AMnNC8etbwJVUSlXtqjWBmVX8Oiki9oiITYHlarxWkiQtHO4H1o6I1SOiH3AQldEUADLzjcwckJlDMnMIcA+wd7EKaRngJuC0zPznnGsiok9EDCj2+1J5/tzjHQVS0xwY4OyIWBr4CnAhlSfxfrnGayVJUlfqpgfQZebsiDiRygqiRuC3mflERJwFjM7M4e1cfiKwFnBGRJxRtO0KTANuLZKXRuAO4JKOYqkpgcnMG4vdN4CP13KNJEnqJt34ILvMHAGMmKftjDb67lC1fzZwdhu33Xx+46iphBQR60TE3yLi8eJ4o4g4fX4/TJIkqTPUOgfmEuDrFHNhMvNRKnUvSZLU07rvOTC9Rq1zYBbPzPsi5lpZNbsL4pEkSfOrDt+FVOsIzJSIWJP33ov0KWBS+5dIkiR1jVpHYE6g8m6jdSNiAvA8cGiXRSVJkmpXhyMwtSYwE4BLgbuoPP/lTSpP4z2ri+KSJEm16qZl1L1JrQnM9cDrwIPAxA76SpIkdalaE5hBmTm0SyORJEkLJJvLs3qos9SawPwrIjbMzMe6NBpJkjT/nAMzt4h4jMrKoz7AZyJiLDATCCAzc6OuD1GSJGluHY3A7NktUUiSpAXnJN65ZeZ/uysQSZK0gOpwDkytD7KTJEnqNWqdxCtJknorJ/FKkqTSMYGRJEmlU6K3SHcW58BIkqTScQRGkqSys4QkSZJKx2XUkiRJvZ8jMJIklZ1P4pUkSaVjCUmSJKn3cwRGkqSSS1chSZKk0rGEJEmS1Ps5AiNJUtm5CkmSJJWOJSRJkqTezxEYSZLKzlVIkiSpdCwhSZIk9X6OwEiSVHauQpIkSaVjCUmSJKltETE0Ip6OiDERcVo7/faPiIyILaravl5c93RE7Da/96zmCIwkSSXXXe9CiohG4CJgF2A8cH9EDM/MJ+fptyRwMnBvVdt6wEHA+sCqwB0RsU5xusN7zssRGEmSyq45O29r31bAmMwcm5nvAlcD+7TS77vAecCMqrZ9gKszc2ZmPg+MKe5X6z3nYgIjSZJaRMSxETG6aju26vRAYFzV8fiirfr6zYDBmXnTPLdu69oO79kaS0iSJJVdJ07izcxhwLAFuTYiGoALgKM6LaA2mMBIklR23beMegIwuOp4UNE2x5LABsDIiABYGRgeEXt3cG1792yVJSRJklSr+4G1I2L1iOhHZVLu8DknM/ONzByQmUMycwhwD7B3Zo4u+h0UEYtExOrA2sB9Hd2zLY7ASJJUdt30HJjMnB0RJwK3Ao3AbzPziYg4CxidmW0mHkW/PwJPArOBEzKzCaC1e3YUiwmMJEkll934ILvMHAGMmKftjDb67jDP8TnAObXcsyOWkCRJUuk4AiNJUtnV4asETGAkSSq7bnoSb29iCUmSJJWOIzCSJJWdJSRJklQ6dZjAWEKSJEml4wiMJEkll1l/IzAmMJIklZ0lJEmSpN7PERhJksquDkdgujyBGb/T57v6I1Rnhi25dU+HoIXILY0zejoE6X/Wne9C6i0sIUmSpNKxhCRJUtnV4QiMCYwkSWVXf69CsoQkSZLKxxEYSZJKrh4n8ZrASJJUdnWYwFhCkiRJpeMIjCRJZVeHk3hNYCRJKrl6nANjCUmSJJWOIzCSJJWdJSRJklQ2lpAkSZJKwBEYSZLKzhKSJEkqmzSBkSRJpVOHCYxzYCRJUuk4AiNJUslZQpIkSeVThwmMJSRJklQ6jsBIklRylpAkSVLp1GMCYwlJkiTVLCKGRsTTETEmIk5r5fxxEfFYRDwcEf+IiPWK9kOLtjlbc0RsUpwbWdxzzrkVO4rDERhJkkquu0ZgIqIRuAjYBRgP3B8RwzPzyapuV2bmr4r+ewMXAEMz8wrgiqJ9Q+Cvmflw1XWHZuboWmNxBEaSpLLL6LytfVsBYzJzbGa+C1wN7DNXKJlvVh32B1p70+TBxbULzARGkiS1iIhjI2J01XZs1emBwLiq4/FF27z3OCEingPOB05q5WMOBK6ap+3Sonz0rYjoMJOyhCRJUsl1ZgkpM4cBw/7He1wEXBQRhwCnA0fOORcRHwbeyczHqy45NDMnRMSSwLXA4cDv2vsMR2AkSSq5bI5O2zowARhcdTyoaGvL1cC+87QdxDyjL5k5ofj1LeBKKqWqdpnASJKkWt0PrB0Rq0dEPyrJyPDqDhGxdtXhHsCzVecagAOomv8SEX0iYkCx3xfYE6genWmVJSRJkkquu1YhZebsiDgRuBVoBH6bmU9ExFnA6MwcDpwYETsDs4DXqCofAR8DxmXm2Kq2RYBbi+SlEbgDuKSjWExgJEkquex49VAnflaOAEbM03ZG1f7J7Vw7Eth6nrZpwObzG4clJEmSVDqOwEiSVHL1+CoBExhJkkquhtVDCx1LSJIkqXQcgZEkqeSytYf1L+RMYCRJKjlLSJIkSSXgCIwkSSVXjyMwJjCSJJVcPc6BsYQkSZJKxxEYSZJKzhKSJEkqne58F1JvYQlJkiSVjiMwkiSVnO9CkiRJpdNsCUmSJKn3cwRGkqSSq8dJvCYwkiSVXD0uo7aEJEmSSscRGEmSSq4eXyVgAiNJUslZQpIkSSqBdkdgIuIGoM2Bqczcu9MjkiRJ86UenwPTUQnph8Wv+wErA38ojg8GXu6qoCRJUu1cRj2PzBwFEBE/yswtqk7dEBGjuzQySZKkNtQ6B6Z/RKwx5yAiVgf6d01IkiRpfmR23lYWta5C+jIwMiLGAgF8APh8l0UlSZJq5hyYNmTmLRGxNrBu0fRUZs7surAWLottswXLfe14oqGBt667mTd+e81c55f89J4sdeDeZFMzOX06U876MbPGvkj/T+zI0kce0NKv3zqrM/Gg43n36ee6+0tQL7Pyxzdi07MOJxobGHvlSJ76+Q1znV/ziJ1Y66hdyKZmZr8zg9Ff/Q1vPjOh5fziA5dn6KjzeeKH1/L0r0Z0c/Tq7TbYfhMOOeNoGhob+Ps1f2PEL6+b6/wOh+7KTocPpbm5mRnTZnD513/FxDHjeyha1avIdsaLImLHzLwzIvZr7Xxm/qWjD3h+411KNCDVBRoaGDT8Ul76/NeY/fIUVr3y50w+7XvMGvtiS5fovzg57R0AFt/+Iyx54F68fPw35rpN37WGsNJPvsP4PY/s1vB7o/teWbGnQ+hR0RB84p8/YuSB32f6pKnscvN3+ffxF82VoPRZYjFmvz0dgFV33Yy1jtqZvx9yfsv5bS45GTJ59cExdZ/A3LLIjJ4OoVeJhgbOvetCfnjYWUx96VXOGH4eF3/xx3MlKIsusRgziu+vTXbegh0PH8oFR57dUyH3Spe+cG23Dok8tNo+nfZ37aYvXl+K4ZyORmC2B+4E9mrlXAIdJjD1bpENPsiscROZPeElAKbdMpLFd9iGN6oSmDnJC0AstmirRcgldt+RabeM7PJ41fstt+mavPXCy0x7cTIAL15/DwN323yuBGZO8gLQZ/FF5noYwsChmzPtxVeY/Y6DqHq/NTZZi1f++xKTx1UWmt53wz/YdNct50pgZlR9fy2y+KK09w9hdY96/CPoaBXSmcXuMZnZ1A3xLHQaVxxA00uTW46bXpnCIhuu+75+Sx64N0sfvj/Rtw+TPvd/7zvff7fteflLZ76vXfVnsZWXY/qEV1uO35k0leU3XfN9/dY6ahc++Pndaejbh7s+fQ5QSWbWPWEvRh34fT74hT26LWaVx7IrLcfUiVNajqdOmsqam6z9vn47Hj6U3Y7Ziz59+3D+Id/uxgililpXIT0fEcMiYqeIKMXQUtm8dc1wxu95JFN/8muW+dwhc51bZMN1yRkzmTXmhZ4JTqU05rLbuekjp/DIOVez3pf2BWD9U/fnmWE3O/qi/9mdv7+Fr21/An869/fs9cX9ezqcutec0WlbWdSawKwL3AGcQCWZ+XlEbNtW54g4NiJGR8Toq16t74ldTa9MoXHlFVqOG1ccwOyXp7TZf9otI+n/8Y/O1dZ/tx14++a7uixGlcv0l6ay2MDlW44XX2U5pr/0Wpv9X/zrvxk4tPIYp+U3W5ONv3Uwe973E9b53FA+dNI+rPWZXbo8ZpXHay9PZblVB7QcL7fKcrz28qtt9r/3hn+y6S5bdUdoakdmdNpWFjUlMJn5Tmb+MTP3AzYFlgJGtdN/WGZukZlbHLz8oE4KtZxmPvE0fVcbSJ+BK0OfPvQfugPvjPr3XH36rDawZX+xj32YWS++N5eBCPrvtj3TbjGBUcXUh8ey5Oor03/wCjT0bWS1fbZmwq0PzNVnidVXatlfdedNePv5yhysO/f9Ljdu9SVu3OpLPHPJLfznZ9cz5tLbuzV+9W7PPzKGFYeswoBBK9LYtw9b7bUtD90+93NLVxqySsv+RjtuzssvTOruMKXa30YdEdsDBwJDgdHAAe1fIQCamnn1+z9n5V9+HxoaeOuvtzLruf+yzPFH8u4Tz/DOqH+z1EH7sNjWm5Kzmmh+6y0mf+u91SKLbr4hs1+a3DIJWMqmZh78xmVsf9XXKsuorx7Fm89MYIOv7s/UR55n4m0PsvbRu7LSdhvQPKuJd9+Yxr0n/aqnw1ZJNDc1c8UZv+Yrv/sWDY0N3P3HO5n47Dj2/fJBvPDYGB6+YzQ7Hbk76310I5pmz2baG9P49Vd+3tNh173uLP1ExFDgp0Aj8OvMPHee88dRqdg0AW8Dx2bmkxExBPgP8HTR9Z7MPK64ZnPgMmAxYARwcnYwO7zdZdRVwbwAPAT8ERiemdNq+SLBZdTqfPW+jFqdy2XU6grdvYz6nlX367S/a7ee+Jc2Y4+IRuAZYBdgPHA/cHBmPlnVZ6nMfLPY3xs4PjOHFgnMjZm5QSv3vQ84CbiXSgLzs8y8ub04ax2B2WhOMJIkqXfpxhGYrYAxmTkWICKuBvYBWhKYefKF/sz1IIf3i4hVgKUy857i+HfAvkC7CUytk3iXiojrIuKVYrs2Iup7coskSQuh6oU4xXZs1emBwLiq4/FF27z3OCEingPOpzKyMsfqEfFQRIyKiO2q7lm94qfVe86r1hGYS4ErgU8Xx4cVbS5fkCSph3Xm6qHMHAYM+x/vcRFwUUQcApwOHAlMAlbLzFeLOS9/jYj1F/Qzah2BWSEzL83M2cV2GbBCRxdJkqSu19yJWwcmAIOrjgcVbW25mko5iMycmZmvFvsPAM8B6xTXV1d1OronUHsC82pEHBYRjcV2GND2gwEkSdLC6H5g7YhYPSL6AQcBw6s7FC9/nmMP4NmifYViEjARsQawNjA2MycBb0bE1sXDco8Aru8okFpLSEcDFwI/pjIZ51/AUTVeK0mSulDSPZN4M3N2RJwI3EplGfVvM/OJiDgLGJ2Zw4ETI2JnYBbwGpXyEcDHgLMiYhaVwZ7jMnNqce543ltGfTMdTOCF2hOYs4AjM/M1gIhYDvghlcRGkiT1oOZufGBJZo6gstS5uu2Mqv2T27juWuDaNs6NBt63vLo9tZaQNpqTvBQfNJXKE3klSZK6Xa0jMA0Rsew8IzA1P8VXkiR1neZuKiH1JrUmIT8C/h0RfyqOPw2c0zUhSZKk+dFdc2B6k5oSmMz8XUSMBnYsmvarfmywJElSd6q5DFQkLCYtkiT1MjU8v2Wh4zwWSZJKrh5LSLWuQpIkSeo1HIGRJKnkLCFJkqTSqccExhKSJEkqHUdgJEkquXqcxGsCI0lSyTXXX/5iCUmSJJWPIzCSJJWc70KSJEmlkz0dQA+whCRJkkrHERhJkkquHp8DYwIjSVLJNUf9zYGxhCRJkkrHERhJkkquHifxmsBIklRy9TgHxhKSJEkqHUdgJEkquXp8lYAJjCRJJVePT+K1hCRJkkrHERhJkkrOVUiSJKl06nEOjCUkSZJUOo7ASJJUcvX4HBgTGEmSSq4e58BYQpIkSaXjCIwkSSVXj5N4TWAkSSq5epwDYwlJkiSVjiMwkiSVnCMwkiSpdDI6b+tIRAyNiKcjYkxEnNbK+eMi4rGIeDgi/hER6xXtu0TEA8W5ByJix6prRhb3fLjYVuwoDkdgJElSTSKiEbgI2AUYD9wfEcMz88mqbldm5q+K/nsDFwBDgSnAXpk5MSI2AG4FBlZdd2hmjq41FhMYSZJKrhtLSFsBYzJzLEBEXA3sA7QkMJn5ZlX//hSPqcnMh6ranwAWi4hFMnPmggRiAiNJUsl1ZgITEccCx1Y1DcvMYcX+QGBc1bnxwIdbuccJwClAP2DHec8D+wMPzpO8XBoRTcC1wNmZ2e7z+UxgJElSiyJZGdZhx/bvcRFwUUQcApwOHDnnXESsD5wH7Fp1yaGZOSEilqSSwBwO/K69z3ASryRJJZeduHVgAjC46nhQ0daWq4F95xxExCDgOuCIzHyuJf7MCcWvbwFXUilVtcsERpKkkmuOzts6cD+wdkSsHhH9gIOA4dUdImLtqsM9gGeL9mWAm4DTMvOfVf37RMSAYr8vsCfweEeBWEKSJEk1yczZEXEilRVEjcBvM/OJiDgLGJ2Zw4ETI2JnYBbwGu+Vj04E1gLOiIgzirZdgWnArUXy0iP2Yz0AABglSURBVAjcAVzSUSwmMJIklVx3PsguM0cAI+ZpO6Nq/+Q2rjsbOLuN224+v3GYwEiSVHI+iVeSJKkEHIGRJKnkalg9tNAxgZEkqeRqWD200DGBkSSp5JwDI0mSVAKOwEiSVHLOgekCd7y6Uld/hOrMF169q6dD0ELkrWu+2NMhSP+z5jpMYSwhSZKk0rGEJElSydXjJF4TGEmSSq7+CkiWkCRJUgk5AiNJUslZQpIkSaVTj0/itYQkSZJKxxEYSZJKrh6fA2MCI0lSydVf+mIJSZIklZAjMJIklZyrkCRJUunU4xwYS0iSJKl0HIGRJKnk6m/8xQRGkqTSq8c5MJaQJElS6TgCI0lSydXjJF4TGEmSSq7+0hdLSJIkqYQcgZEkqeTqcRKvCYwkSSWXdVhEsoQkSZJKxxEYSZJKzhKSJEkqnXpcRm0JSZIklY4jMJIklVz9jb84AiNJUuk1k522dSQihkbE0xExJiJOa+X8cRHxWEQ8HBH/iIj1qs59vbju6YjYrdZ7tsYERpIk1SQiGoGLgN2B9YCDqxOUwpWZuWFmbgKcD1xQXLsecBCwPjAU+EVENNZ4z/exhCRJUsl14yqkrYAxmTkWICKuBvYBnpzTITPfrOrfn/cqXPsAV2fmTOD5iBhT3I+O7tkaExhJkkquMx9kFxHHAsdWNQ3LzGHF/kBgXNW58cCHW7nHCcApQD9gx6pr75nn2oHFfof3nJcJjCRJalEkK8M67Nj+PS4CLoqIQ4DTgSM7I7ZqJjCSJJVcN5aQJgCDq44HFW1tuRr4ZQ3Xzs89ASfxSpJUetmJ/3XgfmDtiFg9IvpRmZQ7vLpDRKxddbgH8GyxPxw4KCIWiYjVgbWB+2q5Z2scgZEkSTXJzNkRcSJwK9AI/DYzn4iIs4DRmTkcODEidgZmAa9RlI+Kfn+kMjl3NnBCZjYBtHbPjmKpKYEpMqVJmTmjOF4MWCkzX5iPr1uSJHWB7nwXUmaOAEbM03ZG1f7J7Vx7DnBOLffsSK0lpD8x9+9PU9EmSZJ6WHNmp21lUWsC0ycz351zUOz365qQJEmS2ldrAjM5IvaecxAR+wBTuiYkSZI0P7ITt7KodRLvccAVEfFzIKg8cOaILotKkiTVrJZ3GC1sakpgMvM5YOuIWKI4frtLo5IkSWpHuwlMRByWmX+IiFPmaQcgMy/owtgkSVINOvNVAmXR0QhM/+LXJVs5V3+/W5Ik9ULduYy6t2g3gcnMi4vdOzLzn9XnIuKjXRaVJElSO2pdhXRhjW2SJKmbNZOdtpVFR3NgPgJsA6wwzzyYpag87leSJPUw58C8Xz9giaJf9TyYN4FPdVVQkiRJ7eloDswoYFREXJaZ/+2mmCRJ0nxwEm/b3omIHwDrA4vOaczMHbskKkmSVLMs0TuMOkutk3ivAJ4CVge+A7wA3N9FMUmSJLWr1gRm+cz8DTArM0dl5tGAoy+SJPUCrkJq26zi10kRsQcwEViua0KSJEnzwzkwbTs7IpYGvkLl+S9LAV/usqgkSVLNXEbdhsy8sdh9A/h414UjSZLUsZoSmIhYHfgiMKT6mszcu2vCkiRJtSrT3JXOUmsJ6a/Ab4AbqM9SmyRJvVY9LqOuNYGZkZk/69JIJEmSalRrAvPTiDgTuA2YOacxMx/skqgkSVLN6rE0UmsCsyFwOJVnv8z5fUp8FowkST3OVUht+zSwRma+25XBSJIk1aLWBOZxYBnglS6MZaE1aIeN+Mh3DicaG3j6qpE8ctENc53/0GE7st5Ru5BNzcyaNoO7v/YbXn92IitssgbbnffZSqeABy+4jhduGd0DX4F6m9123YELLjiLxoYGfnvpVZz/g4vmOv+lk4/l6KMPZvbs2UyZPJVjjj2FF1+c0HJ+ySWX4LFHRnL98Fs4+Uund3f46oX++fR4zr/hXpoz+eSW63D0Dhu12u+Ox17g1Cvu4ooT92L9QQN4bNxkvvuXf1VOZnLczpuy4wYf6MbIBa5Cas8ywFMRcT9zz4FxGXUHoiH46NlHMuKQc5k2aSr73nQW/73tAV5/dmJLnzF//Tf/+cOdAKy2y2ZsfeZh3HLY+Ux9ajzXfeJbZFMzi624DPvfdg7/vf1Bsqkeq52ao6GhgZ/99ByGfuJgxo+fxD3/HsENN97Gf/7zbEufhx9+nA9vvTvTp8/g88cewbnfP51DDv1Cy/nvfPur3P2Pe3oifPVCTc3NfP/6e/jVZ3djpaUX59Cf38D2H1qNNVdaZq5+02bO4sp/PsmGg1doaVtrpWW58sS96NPYwOQ33+GAn17Pxz40mD6Ntb6pRp2hHlch1foddibwSeB7wI+qNnVghU3W5M0XXuatFyfTPKuJ566/hw/suvlcfWa9Pb1lv+/ii0Dxjdg0492WZKXPIn2pw+9PtWKrLTfluede4PnnX2TWrFn88Y/Xs/deu83VZ+SofzF9+gwA7r3vAQYNXKXl3GabbshKK63A7bf/vVvjVu/1+LgpDF5+SQYtvyR9+zSy28ZrMPLJF9/X76LbHuSoHTakX5/GlrbF+vVpSVbend1ERLeFrTpX65N4R0XESsCWRdN9mWk5qQb9V1mWtydNbTme9tJUVtx0zff1W+/Indnwc7vT0K8PNx34vZb2FTZdk+1/+DmWGDSAkSf/ytEXserAlRk3/r0RvPETJrHVlpu22f8zRx3MLbfeBUBE8IPzz+CIo05ipx236/JYVQ6vvPkOKy/dv+V4paUX57Fxk+fq858JU3j59Wl8bN3BXD7q8bnOPfbiZM788z+Y9PrbnHPAxxx96QH1WEKq6bssIg4A7qMymfcA4N6I+FQ7/Y+NiNERMfrv055tq5uqPHn5HVyz7Ve473tXs+lJ+7a0T37oOf6802n8dY8z2PjEvWhcpG8PRqmyOeSQ/dhi84354Y9+CcAXjjuSm2+5kwkTJvVwZCqT5ubkhzfezyl7bNnq+Q1XW4G/nPJJrjhxL34z8lFmzprdzREqO/G/sqh1Dsw3gS3njLpExArAHcCfW+ucmcOAYQCXDDqsPL8bXWDapNdYYpX3Xtzdf+XlmDbptTb7P3f9PWz7vc8wap7218dMZPa0GSz7wUFMefT5LopWZTBxwksMHrRqy/GggaswceJL7+u3047b8fXTTmLHnfbn3XcrCwi33npztv3ohznu80eyxBL96devL9OmTeMb3/x+t8Wv3mfFpRbnpTemtRy//MY7rLjUeyMy096dxXMvv8Yxw24B4NW3p/Oly+/gJ0fuzPqDBrT0W2PFZVi8Xx/GvPz6XO1SV6g1gWmYp2T0KrXPn6lrkx8Zy1Krr8ySg1dg2ktTWXOfrbnrxF/M1Wep1VfizedfBmC1nTbhjecrfxktOXgF3p74KtnUzBIDl2fpNVflrXmGdVV/7h/9MGuttTpDhgxmwoSXOOCAfTj8iBPm6rPJJuvzi4vOZY+9DmPy5Fdb2o848ovv7R9+AJtvvpHJi1h/0ABefPVNJkx9ixWXWpxbHxnL9w7evuX8kov2Y+QZh7Qcf/bimzlljy1Zf9AAJkx9i5WW7k+fxgYmvvY2L0x+g1WXXaInvoy61lyHkyRrTWBuiYhbgauK4wOBm7smpIVLNjXzr29dzu5X/B/R0MDT14zitWcmsPmp+zP5ked58fYHWf+oXRm47fo0z25i5hvTGPXliwFYaat12O34vWie3UQ2J//85mXMfO3tHv6K1NOampo4+UunM+KmK2lsaOCyy6/hySef4dtnnsroBx7hxhtv57zvf4sllujP1VdVvpfGjZvAJ/f7TA9Hrt6qT2MDp+29NV/47W00Nyf7bLE2a620LL+47UHWGzSAHdZbrc1rH3rhZX478jH6NDbQEPD1fT/Csv0X7cboBZSo8NN5otalVxGxH7BtcXh3Zl5Xy3X1XkJS5/vCK3f1dAhaiLx1zRc77iTNp8U+eVq3rsfabuBOnfZ37d0T/laKtWQ1jcBExHmZ+TXgL620SZKkHuQqpLbt0krb7p0ZiCRJWjDNZKdtHYmIoRHxdESMiYjTWjl/SkQ8GRGPRsTfIuIDRfvHI+Lhqm1GROxbnLssIp6vOrdJR3G0OwITEV8AjgfWiIhH5zQDSwD/7PCrlCRJC42IaAQuojKwMR64PyKGZ+aTVd0eArbIzHeKPOJ84MDMvAvYpLjPcsAY4Laq676ama2ubm5NRyWkK6lM1v0+UJ1lvZWZU1u/RJIkdadufJXAVsCYzBwLEBFXA/sALQlMkajMcQ9wWCv3+RRwc2a+s6CBtFtCysw3MvOFzDwYWL4Icm9gyIJ+oCRJ6lydWUKqfhhtsR1b9VEDgXFVx+OLtrZ8ltZXLR/Eeyub5zinKDv9OCIW6ehrrvVJvN8CLqeSxAwALo0IX2ErSdJCJjOHZeYWVduwBblPRBwGbAH8YJ72VYANgVurmr8OrEvllUXLAR0uEqr1OTCHARtn5oziw88FHgbOrvF6SZLURbrxFQATgMFVx4OKtrlExM5UnuK/fWbOnOf0AcB1mTlrTkNmznm/ycyIuBQ4taNAal2FNBGofjLRIq0FLEmSul9mdtrWgfuBtSNi9YjoR6UUNLy6Q0RsClwM7N3Gi58PZp7yUTEqQ0QEsC/weCvXzaXWEZg3gCci4nYqD/zbBbgvIn4GkJkn1XgfSZJUUpk5OyJOpFL+aQR+m5lPRMRZwOjMHE6lZLQE8KdKPsKLmbk3QEQMoTKCM+8r/64o3rMYVCo8x3UUS60JzHXFNsfIGq+TJEldrDsfZJeZI4AR87SdUbW/czvXvkArk34zc8f5jaOmBCYzL5/fG0uSpO7Rjcuoe41aVyHtGREPRcTUiHgzIt6KiDe7OjhJkqTW1FpC+gmwH/BY1mOaJ0lSL1aP70KqNYEZBzxu8iJJUu/Tjcuoe41aE5j/A0ZExCigZT13Zl7QJVFJkiS1o9YE5hzgbSrPgunXdeFIkqT51VyHBZJaE5hVM3ODLo1EkiQtkHosIdX6JN4REbFrl0YiSZJUo1pHYL4AnBoRM4FZVJ6Ul5m5VJdFJkmSamIJqQ2ZuWRXByJJkhaMJaQ2RMS1EfGJiKi15CRJktRlak1IfgkcCjwbEedGxAe7MCZJkjQfmjM7bSuLmhKYzLwjMw8FNgNeAO6IiH9FxGciom9XBihJktqXnfhfWdRcEoqI5YGjgGOAh4CfUklobu+SyCRJktpQ0yTeiLgO+CDwe2CvzJxUnLomIkZ3VXCSJKljZSr9dJZal1FfBdySmW9GxOkRsRlwdmY+mJlbdGF8kiSpA2Uq/XSWWktIpxfJy7bAzsBvqEzslSRJ6na1JjBNxa97AMMy8yZ8J5IkSb1CZnOnbWVRawlpQkRcDOwCnBcRizAfE4AlSVLXabaE1KYDgFuB3TLzdWA54KtdFpUkSVI7an2VwDvAX6qOJwGT2r5CkiR1l3QVkiRJKhtLSJIkSSXgCIwkSSVnCUmSJJVOPT6J1xKSJEkqHUdgJEkquXp8lYAJjCRJJeccGEmSVDouo5YkSSoBR2AkSSo5S0iSJKl0XEYtSZJUAo7ASJJUcvVYQnIERpKkkmsmO23rSEQMjYinI2JMRJzWyvlTIuLJiHg0Iv4WER+oOtcUEQ8X2/Cq9tUj4t7intdERL+O4jCBkSRJNYmIRuAiYHdgPeDgiFhvnm4PAVtk5kbAn4Hzq85Nz8xNim3vqvbzgB9n5lrAa8BnO4rFBEaSpJLLzE7bOrAVMCYzx2bmu8DVwD7zxHJXZr5THN4DDGrvhhERwI5Ukh2Ay4F9OwrEBEaSpJJrzuy0LSKOjYjRVduxVR81EBhXdTy+aGvLZ4Gbq44XLe55T0TMSVKWB17PzNk13hNwEq8kSaqSmcOAYf/rfSLiMGALYPuq5g9k5oSIWAO4MyIeA95YkPs7AiNJUsllJ/7XgQnA4KrjQUXbXCJiZ+CbwN6ZObMlzswJxa9jgZHApsCrwDIRMWdQpdV7zssERpKkkuvMElIH7gfWLlYN9QMOAoZXd4iITYGLqSQvr1S1LxsRixT7A4CPAk9mZeLNXcCniq5HAtd3FIgJjCRJqkkxT+VE4FbgP8AfM/OJiDgrIuasKvoBsATwp3mWS38IGB0Rj1BJWM7NzCeLc18DTomIMVTmxPymo1icAyNJUsl154PsMnMEMGKetjOq9ndu47p/ARu2cW4slRVONTOBkSSp5GqYu7LQsYQkSZJKxxEYSZJKrh7fhWQCI0lSydVjAmMJSZIklY4jMJIklVz9jb9A1OOwU28VEccWj3CW/md+P6mz+T2l3sQSUu9ybMddpJr5/aTO5veUeg0TGEmSVDomMJIkqXRMYHoXa8vqTH4/qbP5PaVew0m8kiSpdByBkSRJpWMCI0mSSscEphNExJCIeLwT7nNURPy82N83ItarOjcyIrb4Xz9D9S0izoqI973qPiJ2iIgbeyIm9T4RsUxEHN8Nn7NDRGxTdXxcRBzR1Z+rhYMJTO+1L7Beh72k+ZCZZ2TmHT0dh3q9ZYAuT2CAHYCWBCYzf5WZv+uGz9VCwASm8zRGxCUR8URE3BYRi0XEmhFxS0Q8EBF3R8S6ABGxV0TcGxEPRcQdEbFS9Y2Kf5HsDfwgIh6OiDWLU5+OiPsi4pmI2K7o+/eI2KTq2n9ExMbd9DWrm0XEERHxaEQ8EhG/L0b/7iza/hYRq0XE0hHx34hoKK7pHxHjIqJvRFwWEZ8q2odGxFMR8SCwX49+YapJ8Wd5U/Hn/3hEHBgRf606v0tEXFfsvx0RPyh+Jt0REVsVI7ljI2Lvos9REXF90f5sRJxZ3OpcYM3i588PouIHxWc+FhEHFtfvEBGjinuMjYhzI+LQ4ufUY3N+drX2My8ihgDHAV8uPme7iPh2RJxaXLNW0feRiHiw6uegVJGZbv/jBgwBZgObFMd/BA4D/gasXbR9GLiz2F+W91aAHQP8qNg/Cvh5sX8Z8KmqzxhZ1e8TwB3F/pHAT4r9dYDRPf374dZl32frA88AA4rj5YAbgCOL46OBvxb71wMfL/YPBH6dVd9XwKLAOGBtIIrv2Rt7+mt06/B7YH/gkqrjpYGngBWK4yuBvYr9BHYv9q8DbgP6AhsDDxftRwGTgOWBxYDHgS2Kn2mPz/O5twONwErAi8AqVEZQXi/2FwEmAN8prjm56mdTWz/zvg2cWvU5LcfAvcAni/1FgcV7+vffrXdtjsB0nucz8+Fi/wEqPwC2Af4UEQ8DF1P5nxxgEHBrRDwGfJXKX0y1+Ms89wf4E7BnRPSl8hfYZQv+JaiX2xH4U2ZOAcjMqcBHqPylBfB7YNti/xoqiQvAQcVxtXWpfM8+m5kJ/KErA1eneQzYJSLOi4jtMvMNKn/uh0XEMlS+H24u+r4L3FJ13ajMnFXsD6m65+2Z+WpmTqfyM2Zb3m9b4KrMbMrMl4FRwJbFufszc1JmzgSeo5IozfnMOZ8zXz/zImJJYGBmXgeQmTMy8532rlH9MYHpPDOr9puo/Ov49czcpGr7UHH+QiojLRsCn6fyr4v5+YwmijeJF/9T3w7sAxwAXPG/fRlaSAwHhkbEcsDmwJ09HI86QWY+A2xGJTk4OyLOAC6lMuJ7MJUEd3bRfVaRnAI0U/z8yMxmip8fc24778fMZ1jVP/uaq46rP2dBf+ZJbTKB6TpvAs9HxKcBihrynLkpS1MZaoVKCag1bwFL1vhZvwZ+RuVfQq8tYLzq/e6kMg9qeYAiOfkXlREWgEOBuwEy823gfuCnVEpDTfPc6ylgSNW8goO7OHZ1gohYFXgnM/8A/ADYLDMnAhOB06kkM/Nrl4hYLiIWo7J44J+8/+fP3cCBEdEYESsAHwPum4/PaOtnXqs/5zLzLWB8ROwLEBGLRMTi8/F5qgMmMF3rUOCzEfEI8ASVURKo1Hn/FBEPAFPauPZq4KvFpLd2J69l5gNUEqYF+eGlksjMJ4BzgFHF99QFwBeBz0TEo8DhVOYdzHENlX+Zz1s+IjNnUHmz8E3FJN5Xujh8dY4NgfuKsvSZwNlF+xXAuMz8zwLc8z7gWuBR4NrMHJ2ZrwL/LCbt/oDKHJpHgUeoJNL/l5kvzcdnfJvWf+bdAHxyziTeea45HDip+N7+F7DyAnxtWoj5KoGFQPGvspHAusXwsKQ6EpXnRz2Umb+Zz+uOArbIzBO7JDCpCzkCU3JReejTvcA3TV6k+lOMamyEE7FVZxyBkSRJpeMIjCRJKh0TGEmSVDomMJIkqXRMYCRJUumYwEiSpNL5f8L2nV+JA4l7AAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 720x504 with 2 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 282
        },
        "id": "QasROACecGjG",
        "outputId": "885b35d7-82f0-49ce-ea47-2dfcd2cd5ddc"
      },
      "source": [
        "plotTrainingLoss(history.history['loss'], history.history['val_loss'])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "id": "klv0LSvycNHk",
        "outputId": "02979c4c-1297-41e7-81df-59852e9f2db5"
      },
      "source": [
        "plotTrainingAccuracy(history.history['acc'], history.history['val_acc'])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "C74uxBZ989rG"
      },
      "source": [
        "##Evaluation of DTLBC"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "oWK0mNnw9DaF",
        "outputId": "ccad9e09-ecb3-4303-98ff-c0b415109ed1"
      },
      "source": [
        "data_new = dataSegmented[covid_indizes]\n",
        "data_new = np.append(data_new,dataSegmented[symptom_indizes[0][:int((len(covid_indizes[0])/2))]], axis=0)\n",
        "data_new = np.append(data_new,dataSegmented[healthy_indizes[0][:int((len(covid_indizes[0])/2))]], axis=0)\n",
        "\n",
        "data_new = np.expand_dims(data_new, axis=3)\n",
        "\n",
        "labels_new = labels[covid_indizes]\n",
        "labels_new = np.append(labels_new,labels[symptom_indizes[0][:int((len(covid_indizes[0])/2))]])\n",
        "labels_new = np.append(labels_new,labels[healthy_indizes[0][:int((len(covid_indizes[0])/2))]])\n",
        "\n",
        "labels_new = oneHotEncodingForBC(labels_new)\n",
        "\n",
        "print(labels_new.shape)\n",
        "print(data_new.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "(3968, 2)\n",
            "(3968, 240, 320, 1)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "h0lW8DkkB-hi"
      },
      "source": [
        "assert labels_new.shape[0]/2 == len(covid_indizes[0])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "RIv0sZ_H9OC9"
      },
      "source": [
        "X_train, X_test, y_train, y_test = train_test_split(data_new, labels_new, test_size=0.2, random_state=0)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Lhiw4fAH9OVo"
      },
      "source": [
        "lr_schedule = tf.keras.optimizers.schedules.ExponentialDecay(\n",
        "    initial_learning_rate=0.00001,\n",
        "    decay_steps=10000,\n",
        "    decay_rate=0.9)\n",
        "\n",
        "cb = tf.keras.callbacks.EarlyStopping(monitor='val_loss', min_delta=0.01, patience=25, verbose=1,\n",
        "    mode='auto', baseline=None, restore_best_weights=True)\n",
        "\n",
        "dtlbc = DTLBC()\n",
        "opt = tf.keras.optimizers.Adam(learning_rate=lr_schedule)\n",
        "dtlbc.compile(optimizer=opt,loss='categorical_crossentropy',metrics=['acc'])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Qxph7Vap9STv",
        "outputId": "4317bd8a-6c4b-4a02-b2df-5393c4c8347b"
      },
      "source": [
        "history = dtlbc.fit(X_train,y_train, epochs=250, validation_data=(X_test,y_test), callbacks=[cb])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "Epoch 1/250\n",
            "100/100 [==============================] - 4s 37ms/step - loss: 9.4064 - acc: 0.5101 - val_loss: 1.8421 - val_acc: 0.5126\n",
            "Epoch 2/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 5.4685 - acc: 0.5072 - val_loss: 0.9406 - val_acc: 0.5139\n",
            "Epoch 3/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 4.0460 - acc: 0.4978 - val_loss: 1.6999 - val_acc: 0.5176\n",
            "Epoch 4/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 3.0113 - acc: 0.5104 - val_loss: 0.7791 - val_acc: 0.4798\n",
            "Epoch 5/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 2.4126 - acc: 0.4943 - val_loss: 0.7805 - val_acc: 0.4975\n",
            "Epoch 6/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 1.6300 - acc: 0.5255 - val_loss: 0.7839 - val_acc: 0.4975\n",
            "Epoch 7/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 1.3142 - acc: 0.5066 - val_loss: 0.7378 - val_acc: 0.5126\n",
            "Epoch 8/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 1.0185 - acc: 0.5221 - val_loss: 0.7447 - val_acc: 0.4975\n",
            "Epoch 9/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.8926 - acc: 0.5107 - val_loss: 0.7341 - val_acc: 0.4924\n",
            "Epoch 10/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.8076 - acc: 0.5243 - val_loss: 0.7130 - val_acc: 0.4950\n",
            "Epoch 11/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.7580 - acc: 0.5299 - val_loss: 0.7054 - val_acc: 0.5239\n",
            "Epoch 12/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.7420 - acc: 0.5391 - val_loss: 0.7083 - val_acc: 0.5013\n",
            "Epoch 13/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.7226 - acc: 0.5454 - val_loss: 0.7032 - val_acc: 0.5214\n",
            "Epoch 14/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.7099 - acc: 0.5520 - val_loss: 0.6935 - val_acc: 0.5365\n",
            "Epoch 15/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.6959 - acc: 0.5728 - val_loss: 0.6919 - val_acc: 0.5428\n",
            "Epoch 16/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.6866 - acc: 0.5882 - val_loss: 0.6934 - val_acc: 0.5252\n",
            "Epoch 17/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.6855 - acc: 0.5725 - val_loss: 0.6906 - val_acc: 0.5693\n",
            "Epoch 18/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.6845 - acc: 0.5747 - val_loss: 0.6880 - val_acc: 0.5416\n",
            "Epoch 19/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.6725 - acc: 0.5907 - val_loss: 0.6863 - val_acc: 0.5466\n",
            "Epoch 20/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.6595 - acc: 0.6118 - val_loss: 0.6979 - val_acc: 0.5302\n",
            "Epoch 21/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.6550 - acc: 0.6087 - val_loss: 0.6843 - val_acc: 0.5630\n",
            "Epoch 22/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.6546 - acc: 0.6153 - val_loss: 0.6853 - val_acc: 0.5579\n",
            "Epoch 23/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.6473 - acc: 0.6185 - val_loss: 0.6902 - val_acc: 0.5466\n",
            "Epoch 24/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.6372 - acc: 0.6421 - val_loss: 0.6846 - val_acc: 0.5617\n",
            "Epoch 25/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.6400 - acc: 0.6314 - val_loss: 0.6841 - val_acc: 0.5668\n",
            "Epoch 26/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.6279 - acc: 0.6427 - val_loss: 0.6890 - val_acc: 0.5882\n",
            "Epoch 27/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.6347 - acc: 0.6437 - val_loss: 0.6835 - val_acc: 0.5718\n",
            "Epoch 28/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.6259 - acc: 0.6569 - val_loss: 0.6871 - val_acc: 0.5630\n",
            "Epoch 29/250\n",
            "100/100 [==============================] - 3s 29ms/step - loss: 0.6145 - acc: 0.6613 - val_loss: 0.6879 - val_acc: 0.5617\n",
            "Epoch 30/250\n",
            "100/100 [==============================] - 3s 29ms/step - loss: 0.6190 - acc: 0.6534 - val_loss: 0.6888 - val_acc: 0.5768\n",
            "Epoch 31/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.6142 - acc: 0.6613 - val_loss: 0.6862 - val_acc: 0.5781\n",
            "Epoch 32/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.6046 - acc: 0.6651 - val_loss: 0.6932 - val_acc: 0.5504\n",
            "Epoch 33/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.6115 - acc: 0.6604 - val_loss: 0.6908 - val_acc: 0.5844\n",
            "Epoch 34/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.5960 - acc: 0.6821 - val_loss: 0.7016 - val_acc: 0.5504\n",
            "Epoch 35/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.5936 - acc: 0.6837 - val_loss: 0.6921 - val_acc: 0.5617\n",
            "Epoch 36/250\n",
            "100/100 [==============================] - 3s 29ms/step - loss: 0.5948 - acc: 0.6742 - val_loss: 0.6912 - val_acc: 0.5642\n",
            "Epoch 37/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.5816 - acc: 0.6934 - val_loss: 0.6912 - val_acc: 0.5919\n",
            "Epoch 38/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.5821 - acc: 0.6922 - val_loss: 0.6958 - val_acc: 0.5806\n",
            "Epoch 39/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.5741 - acc: 0.6972 - val_loss: 0.6944 - val_acc: 0.5869\n",
            "Epoch 40/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.5771 - acc: 0.6903 - val_loss: 0.6901 - val_acc: 0.5768\n",
            "Epoch 41/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.5776 - acc: 0.6912 - val_loss: 0.6940 - val_acc: 0.5844\n",
            "Epoch 42/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.5647 - acc: 0.7114 - val_loss: 0.6943 - val_acc: 0.5693\n",
            "Epoch 43/250\n",
            "100/100 [==============================] - 3s 29ms/step - loss: 0.5651 - acc: 0.7086 - val_loss: 0.6960 - val_acc: 0.5793\n",
            "Epoch 44/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.5626 - acc: 0.7001 - val_loss: 0.6946 - val_acc: 0.5831\n",
            "Epoch 45/250\n",
            "100/100 [==============================] - 3s 29ms/step - loss: 0.5526 - acc: 0.7171 - val_loss: 0.7062 - val_acc: 0.5932\n",
            "Epoch 46/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.5493 - acc: 0.7171 - val_loss: 0.7049 - val_acc: 0.5730\n",
            "Epoch 47/250\n",
            "100/100 [==============================] - 3s 29ms/step - loss: 0.5429 - acc: 0.7388 - val_loss: 0.7036 - val_acc: 0.5680\n",
            "Epoch 48/250\n",
            "100/100 [==============================] - 3s 28ms/step - loss: 0.5377 - acc: 0.7284 - val_loss: 0.7138 - val_acc: 0.5693\n",
            "Epoch 49/250\n",
            "100/100 [==============================] - 3s 29ms/step - loss: 0.5429 - acc: 0.7212 - val_loss: 0.7053 - val_acc: 0.5806\n",
            "Epoch 50/250\n",
            "100/100 [==============================] - 3s 29ms/step - loss: 0.5316 - acc: 0.7350 - val_loss: 0.7093 - val_acc: 0.5856\n",
            "Epoch 51/250\n",
            "100/100 [==============================] - 3s 29ms/step - loss: 0.5345 - acc: 0.7331 - val_loss: 0.7126 - val_acc: 0.5819\n",
            "Epoch 52/250\n",
            "100/100 [==============================] - 3s 29ms/step - loss: 0.5238 - acc: 0.7489 - val_loss: 0.7117 - val_acc: 0.5831\n",
            "Restoring model weights from the end of the best epoch.\n",
            "Epoch 00052: early stopping\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DJ4kZFLk_zet"
      },
      "source": [
        "predictions = getPredictions(X_test, dtlbc)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 143
        },
        "id": "J0sIaL1qArh6",
        "outputId": "2374aebe-d618-457e-b2eb-7273ce9b9858"
      },
      "source": [
        "results = getAllMetricsForBC(y_test, predictions)\n",
        "results"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Class</th>\n",
              "      <th>F1 Score</th>\n",
              "      <th>Sensitivity</th>\n",
              "      <th>Specificity</th>\n",
              "      <th>Precision</th>\n",
              "      <th>Recall</th>\n",
              "      <th>Accuracy</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>Overall</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>0.571788</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>Healthy</td>\n",
              "      <td>55.0265</td>\n",
              "      <td>54.3081</td>\n",
              "      <td>54.3081</td>\n",
              "      <td>55.7641</td>\n",
              "      <td>54.3081</td>\n",
              "      <td>-</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>COVID</td>\n",
              "      <td>59.1346</td>\n",
              "      <td>59.854</td>\n",
              "      <td>59.854</td>\n",
              "      <td>58.4323</td>\n",
              "      <td>59.854</td>\n",
              "      <td>-</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "     Class F1 Score Sensitivity Specificity Precision   Recall  Accuracy\n",
              "0  Overall        -           -           -         -        -  0.571788\n",
              "1  Healthy  55.0265     54.3081     54.3081   55.7641  54.3081         -\n",
              "2    COVID  59.1346      59.854      59.854   58.4323   59.854         -"
            ]
          },
          "execution_count": 36,
          "metadata": {},
          "output_type": "execute_result"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 428
        },
        "id": "WoN4s4m6A-qc",
        "outputId": "a3695b59-4c84-4d3f-bdb3-08e30cbcf3fb"
      },
      "source": [
        "matrix = confusion_matrix(np.argmax(y_test, axis=1), np.argmax(predictions, axis=1), normalize='true')\n",
        "labellist = ['No-COVID', 'COVID']\n",
        "plotConfusionMatrix(labellist, matrix)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x504 with 2 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "id": "PevQWxQFDBIA",
        "outputId": "2d3de548-fc73-44bf-c1f8-ee372cd743df"
      },
      "source": [
        "plotTrainingLoss(history.history['loss'], history.history['val_loss'])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXgAAAEGCAYAAABvtY4XAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO3deZRV5Z3u8e9vn6EGoKqYVEo0DBqlmBW8KLERbXHAIVkmtvdK2tjdMVmrb4abXBu918SkV2fFXO3ENm20TWLaRGM0mqRjNIoxKhpHQBQUFEFUQGQuKGo6w+/+sXcNlAUUUKcOtc/zWeusfYZd5333qXOe/e73vOfd5u6IiEj8BMWugIiIFIYCXkQkphTwIiIxpYAXEYkpBbyISEwli12BzoYNG+ajRo0qdjVERPqNxYsXb3H34d09dlgF/KhRo1i0aFGxqyEi0m+Y2bt7e0xdNCIiMaWAFxGJKQW8iEhMHVZ98N3JZDKsW7eO5ubmYlelXyovL2fkyJGkUqliV0VE+thhH/Dr1q1j0KBBjBo1CjMrdnX6FXdn69atrFu3jtGjRxe7OiLSxw77Lprm5maGDh2qcD8IZsbQoUN19CNSog77gAcU7odAr51I6eoXAb8v7s6HO5vZ1ZwpdlVERA4r/T7gzYwtu1rY1ZwtyPPv2LGDH/3oRwf1t+effz47duzo8frf+ta3uOmmmw6qLBGRrvp9wAMkAiOXL8yJS/YV8NnsvncqjzzyCDU1NYWolojIfing9+Oaa65h9erVTJkyhauvvpqnnnqK008/nYsuuoi6ujoAPvnJT3LyySczfvx47rjjjva/HTVqFFu2bGHt2rWMGzeOz3/+84wfP545c+bQ1NS0z3KXLl3KjBkzmDRpEp/61KfYvn07ALfccgt1dXVMmjSJyy67DICnn36aKVOmMGXKFKZOncquXbsK8lqISP9y2A+T7OzbD73OGxt2fuT+5kwOgPJU4oCfs662iusvHL/Xx2+44QaWL1/O0qVLAXjqqadYsmQJy5cvbx96eOeddzJkyBCampqYPn06l1xyCUOHDt3jeVatWsW9997Lj3/8Yy699FIefPBB5s2bt9dy//Zv/5Yf/vCHzJo1i29+85t8+9vf5uabb+aGG27gnXfeoaysrL3756abbuLWW29l5syZNDQ0UF5efsCvg4jETyxa8GbQl6eWPeWUU/YYV37LLbcwefJkZsyYwfvvv8+qVas+8jejR49mypQpAJx88smsXbt2r89fX1/Pjh07mDVrFgBXXHEFCxcuBGDSpElcfvnl3H333SST4f555syZfO1rX+OWW25hx44d7feLSGnrV0mwt5b2uu2N7GzOUjeiqk/qMWDAgPbrTz31FH/60594/vnnqays5Iwzzuh23HlZWVn79UQisd8umr15+OGHWbhwIQ899BDf+c53WLZsGddccw1z587lkUceYebMmTz22GOceOKJB/X8IhIfsWjBt/XBewGa8YMGDdpnn3Z9fT2DBw+msrKSlStX8sILLxxymdXV1QwePJhnnnkGgF/84hfMmjWLfD7P+++/z+zZs/ne975HfX09DQ0NrF69mokTJzJ//nymT5/OypUrD7kOItL/9asW/N4kAsPdcQ+7a3rT0KFDmTlzJhMmTOC8885j7ty5ezx+7rnncvvttzNu3DhOOOEEZsyY0Svl3nXXXXzxi1+ksbGRMWPG8LOf/YxcLse8efOor6/H3fnyl79MTU0N3/jGN3jyyScJgoDx48dz3nnn9UodRKR/s0K0eg/WtGnTvOsJP1asWMG4ceP2+XdbG1pYv6OJE4+qIp2MxUFJr+rJaygi/ZOZLXb3ad09Fos0TARhsz13GO2sRESKLV4BX6Cx8CIi/VEsAj6pgBcR+YhYBLxa8CIiH6WAFxGJqVgEfGCGoYAXEeksFgFvZgSBkcvne/25D2W6YICbb76ZxsbGbh8744wz6DosVESkt8Qi4CH8ojXX+/le0IAXESmk2AR8IjCyBWjBd50uGODGG29k+vTpTJo0ieuvvx6A3bt3M3fuXCZPnsyECRO47777uOWWW9iwYQOzZ89m9uzZ+yzn3nvvZeLEiUyYMIH58+cDkMvl+NznPseECROYOHEiP/jBD4DupwwWEemqf01V8MdrYOOybh86OpMDHFIHuElHTYTzbtjrw12nC16wYAGrVq3ipZdewt256KKLWLhwIZs3b6a2tpaHH34YCOeoqa6u5vvf/z5PPvkkw4YN22sZGzZsYP78+SxevJjBgwczZ84cfve733HMMcewfv16li9fDtA+PXB3UwaLiHQVmxZ8X00ZvGDBAhYsWMDUqVM56aSTWLlyJatWrWLixIk8/vjjzJ8/n2eeeYbq6uoeP+fLL7/MGWecwfDhw0kmk1x++eUsXLiQMWPGsGbNGr70pS/x6KOPUlUVzpbZ3ZTBIiJd9a902EdLe+v2JnY0tTK+tufBejDcnWuvvZYvfOELH3lsyZIlPPLII1x33XWcddZZfPOb3zyksgYPHsyrr77KY489xu23387999/PnXfe2e2UwQp6EekqNi34RGDkCzBlcNfpgs855xzuvPNOGhoaAFi/fj2bNm1iw4YNVFZWMm/ePK6++mqWLFnS7d9355RTTuHpp59my5Yt5HI57r33XmbNmsWWLVvI5/Nccskl/Mu//AtLlizZ65TBIiJdxabZlwgMJ5xwLNmLcwZ3nS74xhtvZMWKFZx66qkADBw4kLvvvpu3336bq6++miAISKVS3HbbbQBcddVVnHvuudTW1vLkk092W8aIESO44YYbmD17Nu7O3Llzufjii3n11Ve58soryUdfHn/3u9/d65TBIiJdxWK6YIBtu1tZt72RE48aRDp54OdmjTNNFywSX7GfLhg6pivI6tesIiJADANe0xWIiIT6RcD3pBtJAd+9w6kLTkT61mEf8OXl5WzdunW/QdX2xaoCvoO7s3XrVsrLy4tdFREpgoKOojGz/wX8A+DAMuBKd28+kOcYOXIk69atY/Pmzftcz935cEczzZuTbCpPHXSd46a8vJyRI0cWuxoiUgQFC3gzOxr4MlDn7k1mdj9wGfCfB/I8qVSK0aNH73c9d+eT1z3KlTNHce35GjEiIlLoLpokUGFmSaAS2FCogsyM6soU9U2ZQhUhItKvFCzg3X09cBPwHvABUO/uCwpVHkB1RYodjQp4EREoYMCb2WDgYmA0UAsMMLN53ax3lZktMrNF++tn35+aCrXgRUTaFLKL5q+Bd9x9s7tngN8Ap3Vdyd3vcPdp7j5t+PDhh1RgtQJeRKRdIQP+PWCGmVWamQFnASsKWJ764EVEOilkH/yLwAPAEsIhkgFwR6HKA7XgRUQ6K+g4eHe/Hri+kGV0VlORpqElSyaXJ5U47H/DJSJSULFKweqKcH+1U614EZGYBXxl+AvWHQp4EZF4BXxNRRpA/fAiIsQs4KsqwhZ8vX7sJCISr4Cvibpo1IIXEYlZwFdXKOBFRNrEMuA1H42ISMwCPpUIGJBOqAUvIkLMAh6iGSWbWotdDRGRootfwFem9UMnERHiGPAVSfXBi4gQw4CvqUirD15EhBgGvGaUFBEJxS7gaypTmotGRIQYBnxVRYrWbJ7mTK7YVRERKarYBXzbdAX6olVESl3sAl7TFYiIhGIb8Dsa9WMnESltsQt4zQkvIhKKXcCri0ZEJBS/gNec8CIiQAwDflBZEjMFvIhI7AI+CCycUVLDJEWkxMUu4EHTFYiIQIwDXtMViEipi23AqwUvIqUutgGvk36ISKmLZcDXVKb0S1YRKXmxDPi2Lpp83otdFRGRoollwNdUpMk7NLRmi10VEZGiiWXAt09XoLHwIlLC4hnwmq5ARCSmAa8Jx0RE4h3wmq5AREpZLAO+Rl00IiLxDHh10YiIFDjgzazGzB4ws5VmtsLMTi1keW0qUgnSiYAdTfqxk4iUrmSBn//fgEfd/dNmlgYqC1weAGZGlaYrEJESV7CAN7Nq4K+AzwG4eyvQZ03qcLoCBbyIlK5CdtGMBjYDPzOzV8zsJ2Y2oOtKZnaVmS0ys0WbN2/utcI1o6SIlLpCBnwSOAm4zd2nAruBa7qu5O53uPs0d582fPjwXitcZ3USkVJXyIBfB6xz9xej2w8QBn6fqFELXkRKXMEC3t03Au+b2QnRXWcBbxSqvK70JauIlLpCj6L5EnBPNIJmDXBlgctrV1OZYldLlmwuTzIRy+H+IiL7VNCAd/elwLRClrE3bT922tmcZciAdDGqICJSVLFt2rZNV6AzO4lIqYptwGu6AhEpdbEP+B0KeBEpUTEO+LDfXSNpRKRUxTjg1UUjIqUt9gG/fbcCXkRKU2wDPp0MGFFdzjtbGopdFRGRoohtwAPUjaji9Q07i10NEZGiiHfA11axenMDzZlcsasiItLnYh3w42uryDu8uXFXsasiItLnYh3wdSOqAXjjA3XTiEjp6VHAm9lXzKzKQj81syVmNqfQlTtUIwdXMKgsyesb6otdFRGRPtfTFvzfuftOYA4wGPgscEPBatVLgsAYN6KKN/RFq4iUoJ4GvEXL84FfuPvrne47rNXVVrFy4y5yeS92VURE+lRPA36xmS0gDPjHzGwQkC9ctXpPXW0Vja053t26u9hVERHpUz2dD/7vgSnAGndvNLMh9OHJOw5F3YgqIPyidczwgUWujYhI3+lpC/5U4E1332Fm84DrgH7xzeXxRw4kGZh+8CQiJaenAX8b0Ghmk4GvA6uBnxesVr2oLJng+CMH6YtWESk5PQ34rLs7cDHw7+5+KzCocNXqXXUjqjQWXkRKTk8DfpeZXUs4PPJhMwuAVOGq1bvqaqvYvKuFTbuai10VEZE+09OA/xughXA8/EZgJHBjwWrVy9q/aFU3jYiUkB4FfBTq9wDVZnYB0Ozu/aIPHsIWPGjKAhEpLT2dquBS4CXgM8ClwItm9ulCVqw3VVekGDm4Qi14ESkpPR0H/3+B6e6+CcDMhgN/Ah4oVMV6m75oFZFS09M++KAt3CNbD+BvDwvja6t5Z8tudrdki10VEZE+0dOQftTMHjOzz5nZ54CHgUcKV63eV1dbhTus1NzwIlIievol69XAHcCk6HKHu88vZMV6m75oFZFS09M+eNz9QeDBAtaloGqry6muSOmLVhEpGfsMeDPbBXQ3z64B7u5VBalVAZgZ42ureEMn/xCRErHPgHf3fjMdQU/UjajiFy+8SzaXJ5noV98Ri4gcsJJKubraKlqyed7ZornhRST+Si7gQV+0ikhpKKmAHzt8IOlkoLnhRaQklFTApxIBJ2hueBEpESUV8NAxZUE4vb2ISHwVPODNLGFmr5jZHwpdVk/U1VaxbXcrH+5sKXZVREQKqi9a8F8BVvRBOT0yPvqi9XWNhxeRmCtowJvZSGAu8JNClnMgThxRhRm8tk4BLyLxVugW/M3APwH5ApfTYwPLkkyoreb5NVuLXRURkYIqWMBHZ37a5O6L97PeVWa2yMwWbd68uVDV2cNpY4fyynvbaWrN9Ul5IiLFUMgW/EzgIjNbC/wKONPM7u66krvf4e7T3H3a8OHDC1idDqeOHUom5yx6d1uflCciUgwFC3h3v9bdR7r7KOAy4M/uPq9Q5R2IU0YPIRkYf3lb3TQiEl8lNw4eoDKdZOqxNTy/ekuxqyIiUjB9EvDu/pS7X9AXZfXUaWOHsWx9PfVNmWJXRUSkIEqyBQ/hF615hxc1mkZEYqpkA37KsTWUpwKeW62AF5F4KtmAL0smmD5qCM8r4EUkpko24CHsh3/zw11s3qV5aUQkfko84IcC6FetIhJLJR3wE46uZlB5UsMlRSSWSjrgE4ExY8xQ/eBJRGKppAMeYObYoby3rZH3tzUWuyoiIr2q5AP+tOOGAWg0jYjETskH/PFHDGTYwDKeUz+8iMRMyQe8mXHa2KE8t3qrztMqIrFS8gEP4XDJTbtaWL25odhVERHpNQp4wh88AZq2QERiRQEPHDu0kpGDK3hOwyVFJEYU8JHTxg7l+TVbyeXVDy8i8aCAj8w8bhj1TRlWfLCz2FUREekVCvjIqWPCeWkWvPFhkWsiItI7FPCRI6rKmX3CcH754rs0Z3LFro6IyCFTwHfyD6ePYUtDK/+1dH2xqyIicsgU8J2cNnYo40ZU8ZNn3tGPnkSk31PAd2JmfP700aza1MDTb20udnVERA6JAr6LCybVcmRVGT955p1iV0VE5JAo4LtIJwOuOG0Uz769hTc2aMikiPRfCvhuXH7Kx6hIJfjps2rFi0j/pYDvRnVlikunjeT3r67nw53Nxa6OiMhBUcDvxd99YjTZvHPXc2uLXRURkYOigN+Ljw0dwDl1R3HPi+/R2JotdnVERA6YAn4fPv9Xo6lvyvDA4nXFroqIyAFTwO/DSccOZsoxNfz02Xc0y6SI9DsK+H0If/g0hne3NvLECk1CJiL9iwJ+P84ZfyTDBpbxO81PIyL9jAJ+P5KJgPMnHsUTKzbR0KIvW0Wk/1DA98CFk2tpyebVTSMi/YoCvgdOPnYwI6rLeejVDcWuiohIjyngeyAIjLkTR/D0W5upb8wUuzoiIj2igO+hCyfXksk5j72+sdhVERHpkYIFvJkdY2ZPmtkbZva6mX2lUGX1hUkjqzl2SCUPvaZuGhHpHwrZgs8CX3f3OmAG8I9mVlfA8grKzLhg0gieW72VLQ0txa6OiMh+FSzg3f0Dd18SXd8FrACOLlR5feHCybXk8s4fl6ubRkQOf33SB29mo4CpwIvdPHaVmS0ys0WbNx/ep8k78ahBHHfEQI2mEZF+oeABb2YDgQeBr7r7R06R5O53uPs0d582fPjwQlfnkJgZF06q5eW129hYr3niReTwVtCAN7MUYbjf4+6/KVhBO96D+r6Z8fGCySNwh4eXfdAn5YmIHKxCjqIx4KfACnf/fqHKobURfjgNXritYEV0Nnb4QOpGVKmbRkQOe4Vswc8EPgucaWZLo8v5vV5KuhJGfQLeerTXn3pvLpxcy9L3d/D+tsY+K1NE5EAVchTNs+5u7j7J3adEl0cKUtjHz4Wtb8OWtwvy9F1dMGkEAH94Td00InL4iscvWT9+Trhc9dj+1800hV06r9530MUdM6SSKcfUqJtGRA5r8Qj4wR+DI+p61k3z5h9h6yp49ZeHVOSFk2t544OdvL2p4ZCeR0SkUOIR8BC24t99Dprr973esgfC5dq/QMvBh/MFk0ZQlgy49jev0ZLNHfTziIgUSowC/lzIZ2H1n/e+TuM2WLUAak+CfAbWPHXQxR1ZVc5Nn5nMy2u3c82Dy3DXOVtF5PASn4AfOR0qBsOb++imWfH7MNjPvxHSg8KwPwQXTq7l62d/nN++sp5//3PffMErItJTyWJXoNcECTh+Thja+Vx4u6vXfg1Dj4ejT4axs2HV4+AOZgdd7P888zje2bKbf338LUYNG8CFk2sPYSNERHpPfFrwEPbDN22DdYs++lj9Onj3LzDxM2Ggf/wc2LUBPlx+SEWaGd+9ZCLTRw3m679+lSXvbT+k5xMR6S3xCvixZ0GQ7H40zfIHAYeJnw5vH3d2uHyrB0Mr96MsmeA/PjuNEdXlXPXzRfoBlIgcFuIV8BU1cOyp3Yf2sl+HXTNDx4a3Bx0JI6Yccj98myED0vz0ium0ZvP8/V0vs7NZp/YTkeKKV8BDOJpm0+vhBGRtNq2Ejctg4qVd1j0H1r0cjq7pBccdMZDb5p3Mms27+cLPF9Oc0fBJESmeeAY87NmKX/ZrsADGf2rPdY+fA56Ht5/oteJnHjeMGz8ziefXbOVr9y8ll9fwSREpjvgF/LDjYMjYjn549zDgR88Ku2U6qz0JKof1WjdNm09NHcl1c8fxyLKNfOv3r2uMvIgURfwCHsJW/DsLw1+qrnsZdrwLky796HpBAMf9Nbz9p3BoZS/6h9PH8IVZY/jFC+/yQ42RF5EiiGfAn3Au5FrhnafhtfshWQ4nXtD9uh+fEw6tXL+416txzbkncslJI/n+42/xyxff2/8fiIj0ongG/LGnQlkVrPgDvP7bsEVfXtX9umPPBEv0ynDJrsyMGy6ZyOwThnPd75bxqE7WLSJ9KJ4Bn0jBcWfBa7+Cxi3hj5v2pmIwHPPfer0fvk0qEXDr5Scx+ZgavvyrV/jhE6uob9QQShEpvHgGPIStds9DeTUcf/Z+1p0DG1+DnYU5gUdlOsmdV0znr44fxr8+/han3fAE33n4DZ24W0QKKr4Bf9zZYddL3cWQLNv3usfPCZcFasUDDB6Q5idXTOePXzmds+uO5M6/rOX0//dn5j/wGms2a055Eel9djgN4Zs2bZovWtTNPDIH670XYNjHoXLIvtdzhx9MgNopcNk9vVf+Pry/rZEfP7OG+15+n5ZsnqnH1nDO+KM4Z/xRjB42oE/qICL9n5ktdvdp3T4W64A/EA99NRwv/09r9t/i70VbGlr41Uvv8ejrG1m+ficAJxw5iHPGH8nZdUcxbsQgkon4HmiJyKFRwPfEm3+Eey+DqfNgwPCw/94d8HCZz4Hn9lziUPMxOGJceMrAwaO6n6a4h9Ztb2TB6x/y6OsbWbR2G3mHdDLghCMHMW7EIOpGVDFuRBUnHDWI6ooUdgjTHItIPCjge6J1N/xoBuzcEE5rgIXTCrddDxLh9SAR9u0HyXAn0NBp6GOyHIafAMNOCEfyZFsg1wK5THg9nw2PDlIVkKrstKwMu5Eqh0aXIWz3QTy7Ic/yD1tZvnE3KzY2sG13a3tRgcGg8hSDypNUtS0rUtRUpKipTFFTmaamMsXgyjTVFSkGlCUZkE5Q2bZMJ0kndWQgsk/ukG0OP8PJMkikD/z8Ee7hZ7+tUQi050vnZeLgTs+hgC+klgbY8iZ8+AZsWgGb3oCt0S9XE+nwkoyWQSoM/EwTZBrDZWsjZHaHO4u9MjxZjifSZCxFxpO45/F8DvM8eNvSyRLQ6gmyBGQ9QYYkWRI0kw4vHi6bSJOxNIkgQTIRkEgkSEXLZCJBMoBkAAmj03XDzcASuAU4AViAWwLMomWAW9C+M0wlApLJJOlkkmQiIJ1Kkg4gnW8i7c0kMo1YpjF8PfI5KBsI6YGQHgBlg8LrqfLoaIo9j6o8F35wcpnoAxRdh3AHm0iHO+JEOrxtQZcPZ3Td83v+fdvSLPyfJaJL23XoOIrLR3Xw3J5HfbQtuqn3Hh/yIPxFddtrZomo/NaogdDacR06NTI6/U2uNXofNXa8rzKN4bYny6NGRDkkK8Jl9P/qqEPnoOmi7XXu/Bq3Xdrq85HGUDKsY5DsdAk6BV220+uW7/jf4nuWi0ePR+t4Plqnc52j/6l7tN27O16L1t1hOH9kfT5a57bHccg0d7yO2aaPfBZJlodhnywP3w9t29J5u9reE3uE+j4MOAKuXrX/9bqxr4CPzxmdiqVsYDgN8dEnH/xzuIcnC2/cGs5s2bg1vDRtC9+g2RYs24LlWinLNlPWFj7W5ajCrD2cstkMmdZWWjMt5DIt5DPNkGnCss1Ytp4g20SQawEc8nk855Bt+yA5eQwH3I1cdB0gwDGcBHkC8gSdrnfc7wS27zd13o0m0jRRRrOV02wVYAEV3kQlzVR4E+W0HNDLmCcgb2EXWcKznWrdTwVJSJR17Kws6LQjyUM+WibTnY4IKyA1IBwenM+GIdW0LQytbPgewKNpOdpDFPYZQu1hndozvIn+bo/uzKhee+wMotALEnseAQfJjp1Umz32v0GnAA72DPO2sG+7btZxNJyuDLtZ05VhCO+xrV3+tmv9sY4dYucj7SAZ7mSzLe2fyfaW/Ud2Zp23M3rMEh078q47eyesawEo4A8HZuFc9hU1HfPVH6JkdKnolWeDbC5Pay4ffj6AvHvYIHPI4e33tx0Ruju5XJbGliyNLRmaWrM0trTS3Jpld2ueXbkkDa05GltyNLRkaWzN0tiaI+9ONudk804+lyWV2x3tiAzHyDvh0QJGJm805QOas0Zj1mjOQWsuT3MmR0smTzaXJUXHJRGFWEfwd7SkMyTIdrkApMiRJBstc6QIW645EuSwaBmQi3Z34bMa5akElakE5ekkZkY2lyeTD/ehmbyTyztmkA6cVADpwEgnnHQCEokUyVSadDpFeSpBeTKgPBXWvjWbI5NzWrPh/yOTy1OWDBhYlqKqPMmg8iSDylMMLE+SSgQEBonAMLPwurXtuMHb/m/R/6wslWBgWbK9O29AWZLKdIKgU8u+8wF/KmmUJROUJQPSyYBkVE5n7k7eO94vju/xPGaQTgT6PqlAFPDSI8lE0O9G8+TyTnMmR1MmR3MmRzbnJAIjCDrCLgiMbM7bdzCNrTl2t2ZpbMnhOKlEGF7paJlKBLg7zZk8Ldlc+7Ilk6c5m2t/jqbo+ZpaczhhyKYSRiIwkkFAIrCwezefJ5t3srk82ZyTyTut0fM2Z3LUN2XYFNU/MGuvTyphpJMBA8uStGTyrNveSENLll3NWXY1ZyjGLNWBhYMC8tFOI5f3HtUjERiV6QQD0kkGlIU7lopUor2xDnseXwQGgVn7jisR7cTKUgkqokt5KqAilaAslSARWPv/um3dRNDxGqYSbRcjHb3PkwkjGf2vUomO90km1/H/yuTCWlWkwzIr0wkq0uEynQjad2zh6+Dk82EDviIVdl32BQW8xFYisLA1WlZab3N3p7E13KHlPbzkohZ025GDYeEyug7QnAl3brtbsuxuyYXL1lx7C7+tlR31VJPJ5WnNduzgWnN5WrJ5rG3n2R6q4Q41CPZspbcFeFNrx1Hc7taw3MbWXEcrv+0SrZ/38Igy3+noIJvz9h1u2069KZPb44jjcJIMLNwRRTuHI6vK+PUXT+v9cnr9GUWkqMys5HZq3XF3MrmwBZ2LWtKdL5lcRzdXJuvt18MjqTy5nJPNhy31vDvJIGzZpxJhyz4Z7bD2PHILdyyt2Xx0hBHu7NqONsKjv2gH1JqPllnKUwc/vHpf9C4QkVgyM9LJ0u7b71+dqiIi0mMKeBGRmFLAi4jElAJeRCSmFPAiIjGlgBcRiSkFvIhITCngRURi6rCaLtjMNgPvHuSfDwO29GJ1DmeltK2g7ZA+tBIAAAUtSURBVI27UtreQmzrx9x9eHcPHFYBfyjMbNHe5kSOm1LaVtD2xl0pbW9fb6u6aEREYkoBLyISU3EK+DuKXYE+VErbCtreuCul7e3TbY1NH7yIiOwpTi14ERHpRAEvIhJT/T7gzexcM3vTzN42s2uKXZ/eZmZ3mtkmM1ve6b4hZva4ma2KloOLWcfeZGbHmNmTZvaGmb1uZl+J7o/dNptZuZm9ZGavRtv67ej+0Wb2YvSevs/M0sWua28ys4SZvWJmf4hux3Z7zWytmS0zs6Vmtii6r8/ey/064M0sAdwKnAfUAf/dzOqKW6te95/AuV3uuwZ4wt2PB56IbsdFFvi6u9cBM4B/jP6ncdzmFuBMd58MTAHONbMZwPeAH7j7ccB24O+LWMdC+AqwotPtuG/vbHef0mn8e5+9l/t1wAOnAG+7+xp3bwV+BVxc5Dr1KndfCGzrcvfFwF3R9buAT/ZppQrI3T9w9yXR9V2EQXA0MdxmDzVEN1PRxYEzgQei+2OxrW3MbCQwF/hJdNuI8fbuRZ+9l/t7wB8NvN/p9rrovrg70t0/iK5vBI4sZmUKxcxGAVOBF4npNkfdFUuBTcDjwGpgh7tno1Xi9p6+GfgnIB/dHkq8t9eBBWa22Myuiu7rs/eyTrrdz7m7m1nsxrqa2UDgQeCr7r4zbOiF4rTN7p4DpphZDfBb4MQiV6lgzOwCYJO7LzazM4pdnz7yCXdfb2ZHAI+b2crODxb6vdzfW/DrgWM63R4Z3Rd3H5rZCIBouanI9elVZpYiDPd73P030d2x3mZ33wE8CZwK1JhZW+MrTu/pmcBFZraWsDv1TODfiO/24u7ro+Umwh34KfThe7m/B/zLwPHRt/Bp4DLg90WuU1/4PXBFdP0K4L+KWJdeFfXJ/hRY4e7f7/RQ7LbZzIZHLXfMrAI4m/A7hyeBT0erxWJbAdz9Wncf6e6jCD+rf3b3y4np9prZADMb1HYdmAMspw/fy/3+l6xmdj5hv14CuNPdv1PkKvUqM7sXOINwmtEPgeuB3wH3A8cSTq98qbt3/SK2XzKzTwDPAMvo6Kf9P4T98LHaZjObRPglW4KwsXW/u/+zmY0hbOEOAV4B5rl7S/Fq2vuiLpr/7e4XxHV7o+36bXQzCfzS3b9jZkPpo/dyvw94ERHpXn/vohERkb1QwIuIxJQCXkQkphTwIiIxpYAXEYkpBbxILzCzM9pmRxQ5XCjgRURiSgEvJcXM5kVzsC81s/+IJvtqMLMfRHOyP2Fmw6N1p5jZC2b2mpn9tm3ebjM7zsz+FM3jvsTMxkZPP9DMHjCzlWZ2j3WeQEekCBTwUjLMbBzwN8BMd58C5IDLgQHAIncfDzxN+GthgJ8D8919EuEva9vuvwe4NZrH/TSgbWbAqcBXCc9NMIZw7hWRotFsklJKzgJOBl6OGtcVhBM95YH7onXuBn5jZtVAjbs/Hd1/F/DraG6Ro939twDu3gwQPd9L7r4uur0UGAU8W/jNEumeAl5KiQF3ufu1e9xp9o0u6x3s/B2d50/Joc+XFJm6aKSUPAF8Opqbu+3cmB8j/By0zWb4P4Bn3b0e2G5mp0f3fxZ4OjrL1Doz+2T0HGVmVtmnWyHSQ2phSMlw9zfM7DrCM+wEQAb4R2A3cEr02CbCfnoIp3K9PQrwNcCV0f2fBf7DzP45eo7P9OFmiPSYZpOUkmdmDe4+sNj1EOlt6qIREYkpteBFRGJKLXgRkZhSwIuIxJQCXkQkphTwIiIxpYAXEYmp/w/QyDedxES+sgAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "id": "7aOdzew1DkdK",
        "outputId": "53c11a9f-f21d-43cc-8711-d2ffbd1a2c71"
      },
      "source": [
        "plotTrainingAccuracy(history.history['acc'], history.history['val_acc'])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "YxosYkIQBHHI"
      },
      "source": [
        "This shows that the class balancing approach does not perform better for the DTLBC. The results are not remarkably better than guessing. This is most likely due to the fact that the dataset is too small for the NN to learn features."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FBHElZVGUqmp"
      },
      "source": [
        "##Evaluation of the Support Vector Machine classifier"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8eWLkkSrU6ba"
      },
      "source": [
        "features_svm = np.load('./drive/MyDrive/COUGHVID/mfccfeatures.npy', allow_pickle=True)\n",
        "labels_svm = np.load('./drive/MyDrive/COUGHVID/mfccfeatureslabels.npy', allow_pickle=True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "ukjeHBjZVqeh",
        "outputId": "edb05f0a-828e-41aa-b9d8-c09009bdc0a4"
      },
      "source": [
        "print(features_svm.shape)\n",
        "print(labels_svm.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "(23490, 256)\n",
            "(23490,)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FStmAEWbtLlr"
      },
      "source": [
        "Perform class balancing by hand. This is not too huge of a problem because SVMs should perform well an small datasets aswell."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "54BMGlcjt3zO",
        "outputId": "3ca3b7c5-9160-4817-bfe5-695541e778d1"
      },
      "source": [
        "covid_indizes = np.where(labels_svm=='COVID-19')\n",
        "print(len(covid_indizes[0]))\n",
        "symptom_indizes = np.where(labels_svm=='symptomatic')\n",
        "print(len(symptom_indizes[0]))\n",
        "healthy_indizes = np.where(labels_svm=='healthy')\n",
        "print(len(healthy_indizes[0]))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "1937\n",
            "3773\n",
            "17780\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nZEFJSOsuBoB"
      },
      "source": [
        "data_new_svm = features_svm[covid_indizes]\n",
        "data_new_svm = np.append(data_new_svm,features_svm[symptom_indizes[0][:len(covid_indizes[0])]], axis=0)\n",
        "data_new_svm = np.append(data_new_svm,features_svm[healthy_indizes[0][:len(covid_indizes[0])]], axis=0)\n",
        "\n",
        "\n",
        "labels_new_svm = labels_svm[covid_indizes]\n",
        "labels_new_svm = np.append(labels_new_svm,labels_svm[symptom_indizes[0][:len(covid_indizes[0])]])\n",
        "labels_new_svm = np.append(labels_new_svm,labels_svm[healthy_indizes[0][:len(covid_indizes[0])]])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "x4QdCBogV-w7"
      },
      "source": [
        "X_train_svm, X_test_svm, y_train_svm, y_test_svm = train_test_split(data_new_svm, labels_new_svm, test_size=0.2, random_state=0)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ipBMQukmUpW_"
      },
      "source": [
        "svm = SVM()\n",
        "svm.fit(X_train_svm, y_train_svm)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "96_8hiW7ixUI"
      },
      "source": [
        "pickle.dump(svm, open('./drive/MyDrive/COUGHVID/Models/svm', 'wb'))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "B14LUoKKUy0p"
      },
      "source": [
        "predictions = []\n",
        "for x in X_test_svm:\n",
        "  X = np.expand_dims(x, axis=0)\n",
        "  predictions.append(svm.predict(X)[0])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4fo_gfAynhby"
      },
      "source": [
        "predictions_oneHot = oneHotEncoding(predictions)\n",
        "y_test_svm_oneHot = oneHotEncoding(y_test_svm)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 699
        },
        "id": "ZjXj7915th_j",
        "outputId": "63366227-68b3-4e32-cdb6-e80cd3b4c7a1"
      },
      "source": [
        "results = getAllMetrics(y_test_svm_oneHot, predictions_oneHot, verbose=True)\n",
        "results"
      ],
      "execution_count": null,
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "COVID Samples: 377\n",
            "Symptomatic Samples: 384\n",
            "Healthy Samples: 402\n",
            "True Positives for COVID Samples: 182.0\n",
            "True Negatives for COVID Samples: 559.0\n",
            "False Negatives for COVID Samples: 195.0\n",
            "False Positives for COVID Samples: 195.0\n",
            "True Positives for symptomatic Samples: 165.0\n",
            "True Negatives for symptomatic Samples: 549.0\n",
            "False Negatives for symptomatic Samples: 219.0\n",
            "False Positives for symptomatic Samples: 219.0\n",
            "True Positives for healthy samples: 155.0\n",
            "True Negatives for healthy samples: 557.0\n",
            "False Negatives for healthy samples: 247.0\n",
            "False Positives for healthy samples: 247.0\n",
            "Sensitivity for the covid class: 0.4827586\n",
            "Specificity for the covid class: 0.7413793\n",
            "Sensitivity for the symptom class: 0.4296875\n",
            "Specificity for the symptom class: 0.71484375\n",
            "Sensitivity for the healthy class: 0.38557214\n",
            "Specificity for the healthy class: 0.6927861\n",
            "Recall Healthy Samples: 0.38557214\n",
            "Recall COVID Samples: 0.4827586\n",
            "Recall Symptomatic Samples: 0.4296875\n",
            "Precision Healthy Samples: 0.44159544\n",
            "Precision COVID Samples: 0.40625\n",
            "Precision Symptomatic Samples: 0.4532967\n",
            "Accuracy: 0.6210949\n"
          ]
        },
        {
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>Class</th>\n",
              "      <th>F1 Score</th>\n",
              "      <th>Sensitivity</th>\n",
              "      <th>Specificity</th>\n",
              "      <th>Precision</th>\n",
              "      <th>Recall</th>\n",
              "      <th>Accuracy</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>Overall</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>-</td>\n",
              "      <td>0.621095</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>Healthy</td>\n",
              "      <td>41.1687</td>\n",
              "      <td>38.5572</td>\n",
              "      <td>69.2786</td>\n",
              "      <td>44.1595</td>\n",
              "      <td>38.5572</td>\n",
              "      <td>-</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>COVID</td>\n",
              "      <td>44.1212</td>\n",
              "      <td>48.2759</td>\n",
              "      <td>74.1379</td>\n",
              "      <td>40.625</td>\n",
              "      <td>48.2759</td>\n",
              "      <td>-</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>Symptomatic</td>\n",
              "      <td>44.1176</td>\n",
              "      <td>42.9688</td>\n",
              "      <td>71.4844</td>\n",
              "      <td>45.3297</td>\n",
              "      <td>42.9688</td>\n",
              "      <td>-</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "         Class F1 Score Sensitivity Specificity Precision   Recall  Accuracy\n",
              "0      Overall        -           -           -         -        -  0.621095\n",
              "1      Healthy  41.1687     38.5572     69.2786   44.1595  38.5572         -\n",
              "2        COVID  44.1212     48.2759     74.1379    40.625  48.2759         -\n",
              "3  Symptomatic  44.1176     42.9688     71.4844   45.3297  42.9688         -"
            ]
          },
          "execution_count": 154,
          "metadata": {},
          "output_type": "execute_result"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "z1t0Togh3tSU"
      },
      "source": [
        "As we can see, the SVM classifier performs slightly better than pure guessing. We can further analyse the results bei taking a look at the confusion matrix. "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7VK3sXUpbW6e"
      },
      "source": [
        "predictions = oneHotEncoding(predictions)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "WYv4-c5pcx2u"
      },
      "source": [
        "y_test_oneHot = oneHotEncoding(y_test_svm)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 428
        },
        "id": "ngiQQ0NKZ6A3",
        "outputId": "4ca33836-c9ad-4b7c-bbcf-5f7eb1f9c7fa"
      },
      "source": [
        "matrix = confusion_matrix(np.argmax(y_test_oneHot, axis=1), np.argmax(predictions, axis=1), normalize=\"true\")\n",
        "labellist = ['healthy','covid', 'symptomatic']\n",
        "plotConfusionMatrix(labellist, matrix)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 720x504 with 2 Axes>"
            ]
          },
          "metadata": {},
          "output_type": "display_data"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mCPrMGTXeHNY"
      },
      "source": [
        "## Evaluation of the entire mediator"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "j66wU2EcfwrX"
      },
      "source": [
        "Load trained classifiers from drive:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ONV71NWXHbx-"
      },
      "source": [
        "dtlmc = new_model = tf.keras.models.load_model('./drive/MyDrive/COUGHVID/Models/dtlmc')\n",
        "dtlbc = tf.keras.models.load_model('./drive/MyDrive/COUGHVID/Models/dtlbc')\n",
        "svm = pickle.load(open('./drive/MyDrive/COUGHVID/Models/svm','rb'))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OHwz2LPbeJRs"
      },
      "source": [
        "mediator = Mediator(dtlmc, dtlbc, svm)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JYIXC4oVVSXV"
      },
      "source": [
        "X_spectrograms = np.load('./drive/MyDrive/COUGHVID/EvaluationSet/evaluationMelspecs.npy', allow_pickle=True)\n",
        "X_features = np.load('./drive/MyDrive/COUGHVID/EvaluationSet/evaluationFeatures.npy', allow_pickle=True)\n",
        "labels = np.load('./drive/MyDrive/COUGHVID/EvaluationSet/evaluationLabels.npy', allow_pickle=True)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "N5F05tLbIU3z",
        "outputId": "28a31735-20fe-42d9-c746-e9575da5400f"
      },
      "source": [
        "print(X_spectrograms.shape)\n",
        "print(X_features.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "name": "stdout",
          "output_type": "stream",
          "text": [
            "(600, 240, 320, 1)\n",
            "(600, 256)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "n4Sy6xS0WnIu"
      },
      "source": [
        "results = [] \n",
        "for index in range(0,600):\n",
        "  spectrogram = np.expand_dims(X_spectrograms[index], axis=0)\n",
        "  features=np.expand_dims(X_features[index], axis=0)\n",
        "  result = mediator.predict(spectrogram, features)\n",
        "  results.append(result)\n",
        "\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "0AUKOFb-Wq54",
        "outputId": "0e57f405-e9d7-403b-c43e-3b30e7065af0"
      },
      "source": [
        "Counter(results)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "text/plain": [
              "Counter({0: 223, 2: 377})"
            ]
          },
          "execution_count": 65,
          "metadata": {},
          "output_type": "execute_result"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "eE6ZOI1bmMRk",
        "outputId": "f29a4f32-7001-42ef-ec34-e12b5a57fd87"
      },
      "source": [
        "Counter(labels)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "data": {
            "text/plain": [
              "Counter({'COVID-19': 52, 'healthy': 455, 'symptomatic': 93})"
            ]
          },
          "execution_count": 66,
          "metadata": {},
          "output_type": "execute_result"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "x3QCriWnY1dB"
      },
      "source": [
        "We can see that the mediator class never predicts a COVID infection because inter classifier agreement is too never given. This is not surprising considering the low precission for COVID infections the classifiers score."
      ]
    }
  ]
}