demo/LR_vs_LSTM_vs_MLP_on_PIMA.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "LR_vs_LSTM_vs_MLP_on_PIMA.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "toc_visible": true,
      "authorship_tag": "ABX9TyOgPuu0jxe0vRGeXEezeyNo",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/lisatwyw/data-gym/blob/master/LR_vs_LSTM_vs_MLP_on_PIMA.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sYD4qX7ik4xw",
        "colab_type": "text"
      },
      "source": [
        "# Overview #\n",
        "\n",
        "Two diabetic datasets can be explored:\n",
        "\n",
        "1. UCI\n",
        "\n",
        "2. PIMA: \"This dataset is originally from the National Institute of Diabetes and Digestive and Kidney Diseases. The objective is to predict based on diagnostic measurements whether a patient has diabetes.\"\n",
        "\n",
        "Adopted from:\n",
        "\n",
        "- [Collab notebook](https://github.com/1UC1F3R616/myGoogleCollabNotebooks/blob/master/Pima_Indians_Diabetes.ipynb)\n",
        "\n",
        "- [MDPI 2019](https://www.mdpi.com/2076-3417/9/17/3532/pdf)\n",
        "\n",
        "\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BtG7k2Y0k80_",
        "colab_type": "text"
      },
      "source": [
        "# A) Mount and download datasets #"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_rup1_Ybj5jh",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 122
        },
        "outputId": "2b8ca93c-3906-4b33-87a0-c5b602087b01"
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')\n",
        "\n",
        "from pydrive.auth import GoogleAuth\n",
        "from pydrive.drive import GoogleDrive\n",
        "from google.colab import auth\n",
        "from oauth2client.client import GoogleCredentials"
      ],
      "execution_count": 1,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Go to this URL in a browser: https://accounts.google.com/o/oauth2/auth?client_id=947318989803-6bn6qk8qdgf4n4g3pfee6491hc0brc4i.apps.googleusercontent.com&redirect_uri=urn%3aietf%3awg%3aoauth%3a2.0%3aoob&response_type=code&scope=email%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdocs.test%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdrive%20https%3a%2f%2fwww.googleapis.com%2fauth%2fdrive.photos.readonly%20https%3a%2f%2fwww.googleapis.com%2fauth%2fpeopleapi.readonly\n",
            "\n",
            "Enter your authorization code:\n",
            "··········\n",
            "Mounted at /content/drive\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2d2ACaXRIjsD",
        "colab_type": "text"
      },
      "source": [
        "## Download UCI-Diabetes ##"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QwGqLC0I8nsQ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import os\n",
        "if os.path.isdir('/content/drive/My Drive/Colab Notebooks/opensource_datasets/' )==False:\n",
        "  try:\n",
        "    ! mkdir '/content/drive/My Drive/Colab Notebooks/opensource_datasets/'\n",
        "  except e as Exception:\n",
        "    pass  \n",
        "\n",
        "if os.path.isdir( '/content/drive/My Drive/Colab Notebooks/opensource_datasets/UCI-diabetes' )==False:\n",
        "  try:\n",
        "    ! mkdir '/content/drive/My Drive/Colab Notebooks/opensource_datasets/UCI-diabetes'\n",
        "  except e as Exception:\n",
        "    pass  \n",
        "    \n",
        "os.chdir('/content/drive/My Drive/Colab Notebooks/opensource_datasets/UCI-diabetes')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xqOOoKP38mrZ",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "! wget -O diabetes2.Z https://archive.ics.uci.edu/ml/machine-learning-databases/diabetes/diabetes-data.tar.Z\n",
        "! tar xvf diabetes2.Z"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4PLAuZJL9WcK",
        "colab_type": "text"
      },
      "source": [
        "## Download PIMA ##"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lADZ3VR7kENs",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import os\n",
        "if os.path.isdir('/content/drive/My Drive/Colab Notebooks/opensource_datasets/' )==False:\n",
        "  try:\n",
        "    ! mkdir '/content/drive/My Drive/Colab Notebooks/opensource_datasets/'\n",
        "  except e as Exception:\n",
        "    pass  \n",
        "\n",
        "if os.path.isdir( '/content/drive/My Drive/Colab Notebooks/opensource_datasets/PIMA' )==False:\n",
        "  try:\n",
        "    ! mkdir '/content/drive/My Drive/Colab Notebooks/opensource_datasets/PIMA'\n",
        "  except e as Exception:\n",
        "    pass  \n",
        "    \n",
        "os.chdir('/content/drive/My Drive/Colab Notebooks/opensource_datasets/PIMA')"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sP9ypL6QjsXL",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        },
        "outputId": "5814b33f-a508-409d-e0e3-bc4a55279589"
      },
      "source": [
        "! git clone https://github.com/lisatwyw/GlucoseLevels.git\n",
        "! ls"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "fatal: destination path 'GlucoseLevels' already exists and is not an empty directory.\n",
            "GlucoseLevels\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Q8ooVLDdlJLC",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        },
        "outputId": "64e92963-757b-4ae2-c5ae-097bde8ab731"
      },
      "source": [
        "os.chdir('/content/drive/My Drive/Colab Notebooks/opensource_datasets/PIMA/GlucoseLevels')\n",
        "! ls"
      ],
      "execution_count": 239,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "accuracy.png   diabetes2.csv  diabetes3.csv  diabetes.csv  loss.png\n",
            "ann_BGL.ipynb  diabetes2.Z    diabetes4.csv  glucose_RF.R  README.md\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2fhi-8rSIqY4",
        "colab_type": "text"
      },
      "source": [
        "# B) Load data  #"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DQO8W9hklSn_",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 238
        },
        "outputId": "50647cee-4759-4c06-c97a-5c5db10569d3"
      },
      "source": [
        "import pandas as pd\n",
        "col_names = ['pregnant', 'glucose', 'bp', 'skin', 'insulin', 'bmi', 'pedigree', 'age', 'label']\n",
        "feature_cols=['pregnant','insulin', 'bmi', 'skin', 'age','glucose','bp','pedigree']\n",
        "\n",
        "pima = pd.read_csv('diabetes.csv', header=None, names=col_names)\n",
        "print(pima.shape)\n",
        "pima.drop(pima.index[0], inplace=True)\n",
        "print(pima.shape)\n",
        "pima.head()"
      ],
      "execution_count": 272,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(769, 9)\n",
            "(768, 9)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "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>pregnant</th>\n",
              "      <th>glucose</th>\n",
              "      <th>bp</th>\n",
              "      <th>skin</th>\n",
              "      <th>insulin</th>\n",
              "      <th>bmi</th>\n",
              "      <th>pedigree</th>\n",
              "      <th>age</th>\n",
              "      <th>label</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>6</td>\n",
              "      <td>148</td>\n",
              "      <td>72</td>\n",
              "      <td>35</td>\n",
              "      <td>0</td>\n",
              "      <td>33.6</td>\n",
              "      <td>0.627</td>\n",
              "      <td>50</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>1</td>\n",
              "      <td>85</td>\n",
              "      <td>66</td>\n",
              "      <td>29</td>\n",
              "      <td>0</td>\n",
              "      <td>26.6</td>\n",
              "      <td>0.351</td>\n",
              "      <td>31</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>8</td>\n",
              "      <td>183</td>\n",
              "      <td>64</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>23.3</td>\n",
              "      <td>0.672</td>\n",
              "      <td>32</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>1</td>\n",
              "      <td>89</td>\n",
              "      <td>66</td>\n",
              "      <td>23</td>\n",
              "      <td>94</td>\n",
              "      <td>28.1</td>\n",
              "      <td>0.167</td>\n",
              "      <td>21</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>5</th>\n",
              "      <td>0</td>\n",
              "      <td>137</td>\n",
              "      <td>40</td>\n",
              "      <td>35</td>\n",
              "      <td>168</td>\n",
              "      <td>43.1</td>\n",
              "      <td>2.288</td>\n",
              "      <td>33</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "  pregnant glucose  bp skin insulin   bmi pedigree age label\n",
              "1        6     148  72   35       0  33.6    0.627  50     1\n",
              "2        1      85  66   29       0  26.6    0.351  31     0\n",
              "3        8     183  64    0       0  23.3    0.672  32     1\n",
              "4        1      89  66   23      94  28.1    0.167  21     0\n",
              "5        0     137  40   35     168  43.1    2.288  33     1"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 272
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0apfhWS-LFeS",
        "colab_type": "text"
      },
      "source": [
        "# C) Setup machine learning experiments #"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7lCD_2cTlJKL",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from keras.models import Sequential\n",
        "from keras.layers import LSTM\n",
        "from keras.layers import Dense\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "from sklearn import preprocessing\n",
        "from sklearn.metrics import roc_curve\n",
        "from sklearn.metrics import roc_auc_score\n",
        "from matplotlib import pyplot\n",
        "\n",
        "seed = 42\n",
        "np.random.seed(seed)\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "M1u6DM81mNjy",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "4ff9aed2-49cb-4f97-fe0e-a7afc6043d38"
      },
      "source": [
        "X = pima[feature_cols] # Features\n",
        "y = pima.label # Target variable\n",
        "\n",
        "X=X.to_numpy('float')\n",
        "y=y.to_numpy('int')\n",
        "\n",
        "X.shape"
      ],
      "execution_count": 274,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(768, 8)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 274
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bIZfLiTw3-UN",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from sklearn.model_selection import train_test_split\n",
        "X_train,X_test,y_train,y_test=train_test_split(X,y,test_size=0.5,random_state=0)\n",
        "\n",
        "X_train0=X_train.copy()\n",
        "X_test0=X_test.copy()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "N9gx7koH3_ky",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 323
        },
        "outputId": "c0593bb9-b3ef-4eb8-8c2b-7221873a4631"
      },
      "source": [
        "y_test"
      ],
      "execution_count": 258,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([1, 0, 0, 1, 0, 0, 1, 1, 0, 0, 1, 1, 0, 0, 0, 0, 1, 0, 0, 0, 1, 1,\n",
              "       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 0, 1,\n",
              "       1, 0, 0, 1, 1, 1, 0, 0, 1, 0, 0, 0, 0, 1, 1, 1, 1, 0, 0, 1, 1, 1,\n",
              "       1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0,\n",
              "       1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1,\n",
              "       0, 1, 1, 1, 1, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0,\n",
              "       0, 1, 0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 1, 0, 0,\n",
              "       1, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 1, 0, 0,\n",
              "       0, 0, 0, 0, 1, 1, 0, 1, 1, 0, 1, 1, 1, 0, 0, 0, 0, 0, 0, 1, 1, 0,\n",
              "       0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 1, 1, 0, 0,\n",
              "       1, 0, 1, 1, 0, 0, 1, 1, 0, 0, 0, 1, 0, 0, 1, 1, 0, 1, 0, 1, 0, 1,\n",
              "       0, 1, 0, 1, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0,\n",
              "       1, 0, 1, 1, 1, 1, 0, 0, 0, 0, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 0, 0,\n",
              "       0, 0, 0, 1, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, 0, 0, 1, 1, 0, 1,\n",
              "       1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1,\n",
              "       1, 1, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 1, 0, 0, 0, 1, 1, 1, 1,\n",
              "       0, 1, 0, 1, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 1, 1, 0, 1, 1, 0, 0, 0,\n",
              "       0, 1, 1, 0, 0, 1, 1, 0, 0, 0])"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 258
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yaa1EQE3KeBH",
        "colab_type": "text"
      },
      "source": [
        "## Try classical methods, like logistic regression ##"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "rnv-03bPl8CC",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 523
        },
        "outputId": "65c5377e-eb50-4e98-8c0b-7c4c611d3a12"
      },
      "source": [
        "from sklearn.linear_model import LogisticRegression\n",
        "\n",
        "# instantiate the model (using the default parameters)\n",
        "logreg = LogisticRegression(verbose=False ) \n",
        "\n",
        "# fit the model with data\n",
        "logreg.fit(X_train0,y_train)\n",
        "\n",
        "# Predict the response for test dataset \n",
        "y_pred=logreg.predict(X_test0)\n",
        "yp = logreg.predict_proba(X_test0)[:,1]\n",
        "\n",
        "# import the metrics class\n",
        "from sklearn import metrics\n",
        "cmat = metrics.confusion_matrix(y_test, y_pred)\n",
        "print( cmat )\n",
        "print(\"Accuracy:\",metrics.accuracy_score(y_test, y_pred))\n",
        "print(\"Precision:\",metrics.precision_score(y_test, y_pred, pos_label=1)) \n",
        "print(\"Recall:\",metrics.recall_score(y_test, y_pred, pos_label=1))\n",
        "\n",
        "auc = roc_auc_score(y_test, yp)\n",
        "print('AUC: %.3f' % auc)\n",
        "fpr, tpr, thresholds = roc_curve(y_test, yp)\n",
        "pyplot.plot([0, 1], [0, 1], linestyle = '-')\n",
        "pyplot.plot(fpr, tpr, marker = '.')\n",
        "pyplot.show()\n"
      ],
      "execution_count": 276,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[[227  26]\n",
            " [ 57  74]]\n",
            "Accuracy: 0.7838541666666666\n",
            "Precision: 0.74\n",
            "Recall: 0.5648854961832062\n",
            "AUC: 0.838\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/sklearn/linear_model/_logistic.py:940: ConvergenceWarning: lbfgs failed to converge (status=1):\n",
            "STOP: TOTAL NO. of ITERATIONS REACHED LIMIT.\n",
            "\n",
            "Increase the number of iterations (max_iter) or scale the data as shown in:\n",
            "    https://scikit-learn.org/stable/modules/preprocessing.html\n",
            "Please also refer to the documentation for alternative solver options:\n",
            "    https://scikit-learn.org/stable/modules/linear_model.html#logistic-regression\n",
            "  extra_warning_msg=_LOGISTIC_SOLVER_CONVERGENCE_MSG)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "rqiuJ3mnKWiH",
        "colab_type": "text"
      },
      "source": [
        "## Standardize input variables using training set only ##"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "PlJx3VxiqZ34",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 68
        },
        "outputId": "96ec536c-3fb3-4b14-9bbf-3d4cd1e2cdde"
      },
      "source": [
        "# raw data\n",
        "X_trn0 = np.expand_dims(X_train0,2)\n",
        "X_tst0 = np.expand_dims(X_test0,2)\n",
        "\n",
        "# after standardizing\n",
        "scaler = preprocessing.StandardScaler().fit(X_train0)\n",
        "X_train=scaler.transform(X_train0)\n",
        "X_test=scaler.transform(X_test0)\n",
        "X_trn = np.expand_dims(X_train,2)\n",
        "X_tst = np.expand_dims(X_test,2)\n"
      ],
      "execution_count": 297,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "384 [  3.77083333  88.0390625   32.14453125  20.71614583  33.390625\n",
            " 121.73958333  70.5546875    0.46521875] [1.14162326e+01 1.51816834e+04 6.53255951e+01 2.48547031e+02\n",
            " 1.43758870e+02 1.02317177e+03 3.45611593e+02 1.17655426e-01] (384, 8)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ca7fC6pEOP-p",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 374
        },
        "outputId": "ead6c716-abf2-48bb-af9b-fe16bb8f37c0"
      },
      "source": [
        "transformer = preprocessing.Normalizer().fit(X_train0) \n",
        "X_train2 = transformer.transform(X_train0)\n",
        "X_test2 = transformer.transform(X_test0)\n",
        "X_trn2 = np.expand_dims(X_train2,2)\n",
        "X_tst2 = np.expand_dims(X_test2,2)\n",
        "\n",
        "transformer = preprocessing.RobustScaler().fit(X_train0) \n",
        "X_train3 = transformer.transform(X_train0)\n",
        "X_test3 = transformer.transform(X_test0)\n",
        "X_trn3 = np.expand_dims(X_train3,2)\n",
        "X_tst3 = np.expand_dims(X_test3,2)\n",
        "\n",
        "\n",
        "df=pd.DataFrame( {'Raw':np.max(X_test0,0), '1':np.max(X_test,0), '2':np.max(X_test2,0), '3':np.max(X_test3,0) } ) \n",
        "print( df)\n",
        "\n",
        "df=pd.DataFrame( {'Raw':np.min(X_test0,0), '1':np.min(X_test,0), '2':np.min(X_test2,0),'3':np.min(X_test3,0) } ) \n",
        "print( df)\n",
        "\n",
        "print( scaler.n_samples_seen_,scaler.mean_, scaler.var_ , X_train0.shape,  X_train3.shape )"
      ],
      "execution_count": 299,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "       Raw         1         2         3\n",
            "0   17.000  3.915354  0.117208  2.800000\n",
            "1  680.000  4.804333  0.959673  4.505338\n",
            "2   57.300  3.112366  0.400734  2.706522\n",
            "3   99.000  4.965556  0.409096  2.356589\n",
            "4   72.000  3.220145  0.512539  2.687500\n",
            "5  199.000  2.415365  0.973682  1.940828\n",
            "6  114.000  2.336947  0.792729  2.625000\n",
            "7    1.893  4.162514  0.013417  4.310078\n",
            "      Raw         1        2         3\n",
            "0   0.000 -1.116030  0.00000 -0.600000\n",
            "1   0.000 -0.714522  0.00000 -0.334520\n",
            "2   0.000 -3.977090  0.00000 -3.521739\n",
            "3   0.000 -1.314028  0.00000 -0.713178\n",
            "4  21.000 -1.033418  0.03246 -0.500000\n",
            "5   0.000 -3.805901  0.00000 -2.769231\n",
            "6   0.000 -3.795174  0.00000 -4.500000\n",
            "7   0.078 -1.128887  0.00025 -0.806202\n",
            "384 [  3.77083333  88.0390625   32.14453125  20.71614583  33.390625\n",
            " 121.73958333  70.5546875    0.46521875] [1.14162326e+01 1.51816834e+04 6.53255951e+01 2.48547031e+02\n",
            " 1.43758870e+02 1.02317177e+03 3.45611593e+02 1.17655426e-01] (384, 8) (384, 8)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "lVJMbkkGLTGM",
        "colab_type": "text"
      },
      "source": [
        "## Try logistic regression with the standardized input data ##"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "BdTBD2_gLLj8",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 724
        },
        "outputId": "ef49b065-8dea-42b9-8595-4ab66dc8d089"
      },
      "source": [
        "# instantiate the model (using the default parameters)\n",
        "logreg = LogisticRegression() \n",
        "\n",
        "# fit the model with data\n",
        "logreg.fit(X_train,y_train)\n",
        "\n",
        "# Predict the response for test dataset \n",
        "y_pred=logreg.predict(X_test)\n",
        "\n",
        "\n",
        "# import the metrics class\n",
        "from sklearn import metrics\n",
        "cmat = metrics.confusion_matrix(y_test, y_pred)\n",
        "\n",
        "print('\\nResults using Standard normalizer:')\n",
        "print( cmat )\n",
        "print(\"Accuracy:\",metrics.accuracy_score(y_test, y_pred))\n",
        "print(\"Precision:\",metrics.precision_score(y_test, y_pred, pos_label=1)) \n",
        "print(\"Recall:\",metrics.recall_score(y_test, y_pred, pos_label=1))\n",
        "\n",
        "\n",
        "# instantiate the model (using the default parameters)\n",
        "logreg = LogisticRegression() \n",
        "\n",
        "# fit the model with data\n",
        "logreg.fit(X_train2,y_train)\n",
        "\n",
        "# Predict the response for test dataset \n",
        "y_pred=logreg.predict(X_test2)\n",
        "\n",
        "# import the metrics class\n",
        "from sklearn import metrics\n",
        "cmat = metrics.confusion_matrix(y_test, y_pred)\n",
        "print('\\nResults using data Normalizer:')\n",
        "print( cmat )\n",
        "print(\"Accuracy:\",metrics.accuracy_score(y_test, y_pred))\n",
        "print(\"Precision:\",metrics.precision_score(y_test, y_pred, pos_label=1)) \n",
        "print(\"Recall:\",metrics.recall_score(y_test, y_pred, pos_label=1))\n",
        "\n",
        "\n",
        "# instantiate the model (using the default parameters)\n",
        "logreg = LogisticRegression() \n",
        "\n",
        "# fit the model with data\n",
        "logreg.fit(X_train3,y_train)\n",
        "\n",
        "# Predict the response for test dataset \n",
        "y_pred=logreg.predict(X_test3)\n",
        "\n",
        "cmat = metrics.confusion_matrix(y_test, y_pred)\n",
        "print('\\nResults using robust normalization:')\n",
        "print( cmat )\n",
        "print(\"Accuracy:\",metrics.accuracy_score(y_test, y_pred))\n",
        "print(\"Precision:\",metrics.precision_score(y_test, y_pred, pos_label=1)) \n",
        "print(\"Recall:\",metrics.recall_score(y_test, y_pred, pos_label=1))\n",
        "\n",
        "\n",
        "\n",
        "yp = logreg.predict_proba(X_test3)[:,1]\n",
        "cmat = metrics.confusion_matrix(y_test, y_pred)\n",
        "print( cmat )\n",
        "print(\"Accuracy:\",metrics.accuracy_score(y_test, y_pred))\n",
        "print(\"Precision:\",metrics.precision_score(y_test, y_pred, pos_label=1)) \n",
        "print(\"Recall:\",metrics.recall_score(y_test, y_pred, pos_label=1))\n",
        "\n",
        "auc = roc_auc_score(y_test, yp)\n",
        "print('AUC: %.3f' % auc)\n",
        "fpr, tpr, thresholds = roc_curve(y_test, yp)\n",
        "pyplot.plot([0, 1], [0, 1], linestyle = '-')\n",
        "pyplot.plot(fpr, tpr, marker = '.')\n",
        "pyplot.show()\n"
      ],
      "execution_count": 300,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\n",
            "Results using Standard normalizer:\n",
            "[[228  25]\n",
            " [ 56  75]]\n",
            "Accuracy: 0.7890625\n",
            "Precision: 0.75\n",
            "Recall: 0.5725190839694656\n",
            "\n",
            "Results using data Normalizer:\n",
            "[[245   8]\n",
            " [119  12]]\n",
            "Accuracy: 0.6692708333333334\n",
            "Precision: 0.6\n",
            "Recall: 0.0916030534351145\n",
            "\n",
            "Results using robust normalization:\n",
            "[[228  25]\n",
            " [ 56  75]]\n",
            "Accuracy: 0.7890625\n",
            "Precision: 0.75\n",
            "Recall: 0.5725190839694656\n",
            "[[228  25]\n",
            " [ 56  75]]\n",
            "Accuracy: 0.7890625\n",
            "Precision: 0.75\n",
            "Recall: 0.5725190839694656\n",
            "AUC: 0.848\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "l09uquYDKWfz",
        "colab_type": "text"
      },
      "source": [
        "## Train LSTM with raw input data ## "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1wHn35l7qZ20",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "e9e66433-3743-47d7-b689-b2d046feb052"
      },
      "source": [
        "from keras.callbacks import ModelCheckpoint\n",
        "C = [ModelCheckpoint(filepath='best.h5',monitor='val_accuracy',save_best_only=True)]\n"
      ],
      "execution_count": 302,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(384, 8, 1)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 302
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-Z5VjIh7q_FE",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "987c6627-ffc1-4290-eb38-cb2d219f1832"
      },
      "source": [
        "model = Sequential()\n",
        "model.add(LSTM(32, input_shape = (X.shape[1],1), return_sequences = True, kernel_initializer = 'uniform', activation ='relu'))\n",
        "model.add(LSTM(64, kernel_initializer = 'uniform', return_sequences = True, activation = 'relu'))\n",
        "model.add(LSTM(128, kernel_initializer = 'uniform', activation = 'relu'))\n",
        "model.add(Dense(256, activation = 'relu'))\n",
        "model.add(Dense(128, activation = 'relu'))\n",
        "model.add(Dense(64, activation = 'relu'))\n",
        "model.add(Dense(16, activation = 'relu'))\n",
        "model.add(Dense(1, activation = 'sigmoid'))\n",
        "\n",
        "from keras import optimizers \n",
        "        \n",
        "lr=0.002 \n",
        "b1=0.9; b2=0.999; ep=1e-08; dd=0.004\n",
        "opt = optimizers.Nadam()#(lr=lr, beta_1=b1, beta_2=b2, epsilon=ep, schedule_decay=dd)      \n",
        "\n",
        "model.compile(loss = 'binary_crossentropy', optimizer ='NADAM',  metrics = ['accuracy'])\n",
        "model.summary()\n",
        "history = model.fit(X_trn0, y_train, validation_split = 0.33, epochs = 500, batch_size = 64, verbose =1,callbacks=C)"
      ],
      "execution_count": 282,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_20\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "lstm_58 (LSTM)               (None, 8, 32)             4352      \n",
            "_________________________________________________________________\n",
            "lstm_59 (LSTM)               (None, 8, 64)             24832     \n",
            "_________________________________________________________________\n",
            "lstm_60 (LSTM)               (None, 128)               98816     \n",
            "_________________________________________________________________\n",
            "dense_96 (Dense)             (None, 256)               33024     \n",
            "_________________________________________________________________\n",
            "dense_97 (Dense)             (None, 128)               32896     \n",
            "_________________________________________________________________\n",
            "dense_98 (Dense)             (None, 64)                8256      \n",
            "_________________________________________________________________\n",
            "dense_99 (Dense)             (None, 16)                1040      \n",
            "_________________________________________________________________\n",
            "dense_100 (Dense)            (None, 1)                 17        \n",
            "=================================================================\n",
            "Total params: 203,233\n",
            "Trainable params: 203,233\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "Train on 257 samples, validate on 127 samples\n",
            "Epoch 1/500\n",
            "257/257 [==============================] - 2s 8ms/step - loss: 0.6926 - accuracy: 0.5914 - val_loss: 0.6857 - val_accuracy: 0.6929\n",
            "Epoch 2/500\n",
            "257/257 [==============================] - 0s 733us/step - loss: 0.6894 - accuracy: 0.6226 - val_loss: 0.6686 - val_accuracy: 0.6929\n",
            "Epoch 3/500\n",
            "257/257 [==============================] - 0s 768us/step - loss: 0.6827 - accuracy: 0.6148 - val_loss: 0.6507 - val_accuracy: 0.6929\n",
            "Epoch 4/500\n",
            "257/257 [==============================] - 0s 709us/step - loss: 0.6703 - accuracy: 0.6342 - val_loss: 1.2332 - val_accuracy: 0.6929\n",
            "Epoch 5/500\n",
            "257/257 [==============================] - 0s 670us/step - loss: 0.9028 - accuracy: 0.6265 - val_loss: 0.6376 - val_accuracy: 0.7087\n",
            "Epoch 6/500\n",
            "257/257 [==============================] - 0s 680us/step - loss: 0.6591 - accuracy: 0.6342 - val_loss: 0.6309 - val_accuracy: 0.6929\n",
            "Epoch 7/500\n",
            "257/257 [==============================] - 0s 663us/step - loss: 0.6759 - accuracy: 0.6187 - val_loss: 0.6588 - val_accuracy: 0.6929\n",
            "Epoch 8/500\n",
            "257/257 [==============================] - 0s 739us/step - loss: 0.6712 - accuracy: 0.6187 - val_loss: 0.6429 - val_accuracy: 0.6929\n",
            "Epoch 9/500\n",
            "257/257 [==============================] - 0s 676us/step - loss: 0.6621 - accuracy: 0.6148 - val_loss: 0.6233 - val_accuracy: 0.7008\n",
            "Epoch 10/500\n",
            "257/257 [==============================] - 0s 710us/step - loss: 0.6591 - accuracy: 0.6148 - val_loss: 0.6710 - val_accuracy: 0.6850\n",
            "Epoch 11/500\n",
            "257/257 [==============================] - 0s 687us/step - loss: 0.6784 - accuracy: 0.6381 - val_loss: 0.6341 - val_accuracy: 0.6929\n",
            "Epoch 12/500\n",
            "257/257 [==============================] - 0s 674us/step - loss: 0.6532 - accuracy: 0.6187 - val_loss: 0.6066 - val_accuracy: 0.6929\n",
            "Epoch 13/500\n",
            "257/257 [==============================] - 0s 663us/step - loss: 0.6504 - accuracy: 0.6187 - val_loss: 0.5982 - val_accuracy: 0.6929\n",
            "Epoch 14/500\n",
            "257/257 [==============================] - 0s 711us/step - loss: 0.6507 - accuracy: 0.6187 - val_loss: 0.6201 - val_accuracy: 0.6929\n",
            "Epoch 15/500\n",
            "257/257 [==============================] - 0s 688us/step - loss: 0.6764 - accuracy: 0.6226 - val_loss: 0.6039 - val_accuracy: 0.6929\n",
            "Epoch 16/500\n",
            "257/257 [==============================] - 0s 680us/step - loss: 0.6608 - accuracy: 0.6226 - val_loss: 0.5947 - val_accuracy: 0.7008\n",
            "Epoch 17/500\n",
            "257/257 [==============================] - 0s 728us/step - loss: 0.6517 - accuracy: 0.6342 - val_loss: 0.6468 - val_accuracy: 0.7087\n",
            "Epoch 18/500\n",
            "257/257 [==============================] - 0s 671us/step - loss: 0.6619 - accuracy: 0.6265 - val_loss: 0.6621 - val_accuracy: 0.6850\n",
            "Epoch 19/500\n",
            "257/257 [==============================] - 0s 666us/step - loss: 0.6699 - accuracy: 0.6381 - val_loss: 0.6467 - val_accuracy: 0.7087\n",
            "Epoch 20/500\n",
            "257/257 [==============================] - 0s 720us/step - loss: 0.6605 - accuracy: 0.6342 - val_loss: 0.6522 - val_accuracy: 0.6850\n",
            "Epoch 21/500\n",
            "257/257 [==============================] - 0s 710us/step - loss: 0.6649 - accuracy: 0.6342 - val_loss: 0.6393 - val_accuracy: 0.7087\n",
            "Epoch 22/500\n",
            "257/257 [==============================] - 0s 681us/step - loss: 0.6569 - accuracy: 0.6459 - val_loss: 0.6443 - val_accuracy: 0.7008\n",
            "Epoch 23/500\n",
            "257/257 [==============================] - 0s 693us/step - loss: 0.6603 - accuracy: 0.6498 - val_loss: 0.6316 - val_accuracy: 0.7087\n",
            "Epoch 24/500\n",
            "257/257 [==============================] - 0s 698us/step - loss: 0.6522 - accuracy: 0.6342 - val_loss: 0.6472 - val_accuracy: 0.6614\n",
            "Epoch 25/500\n",
            "257/257 [==============================] - 0s 695us/step - loss: 0.6566 - accuracy: 0.6459 - val_loss: 0.6149 - val_accuracy: 0.7165\n",
            "Epoch 26/500\n",
            "257/257 [==============================] - 0s 647us/step - loss: 0.6403 - accuracy: 0.6498 - val_loss: 0.5827 - val_accuracy: 0.7087\n",
            "Epoch 27/500\n",
            "257/257 [==============================] - 0s 711us/step - loss: 0.6337 - accuracy: 0.6381 - val_loss: 0.6298 - val_accuracy: 0.7165\n",
            "Epoch 28/500\n",
            "257/257 [==============================] - 0s 792us/step - loss: 0.6535 - accuracy: 0.6537 - val_loss: 0.6870 - val_accuracy: 0.5669\n",
            "Epoch 29/500\n",
            "257/257 [==============================] - 0s 747us/step - loss: 0.6811 - accuracy: 0.6187 - val_loss: 0.6581 - val_accuracy: 0.6850\n",
            "Epoch 30/500\n",
            "257/257 [==============================] - 0s 784us/step - loss: 0.6784 - accuracy: 0.6342 - val_loss: 0.6700 - val_accuracy: 0.7165\n",
            "Epoch 31/500\n",
            "257/257 [==============================] - 0s 682us/step - loss: 0.7113 - accuracy: 0.6109 - val_loss: 0.6682 - val_accuracy: 0.7165\n",
            "Epoch 32/500\n",
            "257/257 [==============================] - 0s 672us/step - loss: 0.6708 - accuracy: 0.6654 - val_loss: 0.6409 - val_accuracy: 0.6929\n",
            "Epoch 33/500\n",
            "257/257 [==============================] - 0s 673us/step - loss: 0.6590 - accuracy: 0.6187 - val_loss: 0.6224 - val_accuracy: 0.6929\n",
            "Epoch 34/500\n",
            "257/257 [==============================] - 0s 670us/step - loss: 0.6516 - accuracy: 0.6187 - val_loss: 0.5915 - val_accuracy: 0.6929\n",
            "Epoch 35/500\n",
            "257/257 [==============================] - 0s 669us/step - loss: 0.6440 - accuracy: 0.6187 - val_loss: 0.6061 - val_accuracy: 0.6929\n",
            "Epoch 36/500\n",
            "257/257 [==============================] - 0s 718us/step - loss: 0.6354 - accuracy: 0.6226 - val_loss: 0.6183 - val_accuracy: 0.6850\n",
            "Epoch 37/500\n",
            "257/257 [==============================] - 0s 718us/step - loss: 0.6496 - accuracy: 0.6304 - val_loss: 0.5968 - val_accuracy: 0.7087\n",
            "Epoch 38/500\n",
            "257/257 [==============================] - 0s 693us/step - loss: 0.6244 - accuracy: 0.6304 - val_loss: 0.6502 - val_accuracy: 0.6850\n",
            "Epoch 39/500\n",
            "257/257 [==============================] - 0s 671us/step - loss: 0.6610 - accuracy: 0.6459 - val_loss: 0.6321 - val_accuracy: 0.6929\n",
            "Epoch 40/500\n",
            "257/257 [==============================] - 0s 679us/step - loss: 0.6519 - accuracy: 0.6187 - val_loss: 0.6383 - val_accuracy: 0.7165\n",
            "Epoch 41/500\n",
            "257/257 [==============================] - 0s 792us/step - loss: 0.6504 - accuracy: 0.6342 - val_loss: 0.5991 - val_accuracy: 0.7008\n",
            "Epoch 42/500\n",
            "257/257 [==============================] - 0s 715us/step - loss: 0.6326 - accuracy: 0.6265 - val_loss: 0.6050 - val_accuracy: 0.7008\n",
            "Epoch 43/500\n",
            "257/257 [==============================] - 0s 692us/step - loss: 0.6488 - accuracy: 0.6304 - val_loss: 0.6426 - val_accuracy: 0.6929\n",
            "Epoch 44/500\n",
            "257/257 [==============================] - 0s 675us/step - loss: 0.6405 - accuracy: 0.6226 - val_loss: 0.5852 - val_accuracy: 0.6929\n",
            "Epoch 45/500\n",
            "257/257 [==============================] - 0s 664us/step - loss: 0.6304 - accuracy: 0.6187 - val_loss: 0.5856 - val_accuracy: 0.6929\n",
            "Epoch 46/500\n",
            "257/257 [==============================] - 0s 679us/step - loss: 0.6315 - accuracy: 0.6226 - val_loss: 0.6283 - val_accuracy: 0.6929\n",
            "Epoch 47/500\n",
            "257/257 [==============================] - 0s 726us/step - loss: 0.6526 - accuracy: 0.6265 - val_loss: 0.5808 - val_accuracy: 0.7008\n",
            "Epoch 48/500\n",
            "257/257 [==============================] - 0s 707us/step - loss: 0.6223 - accuracy: 0.6265 - val_loss: 0.9227 - val_accuracy: 0.6535\n",
            "Epoch 49/500\n",
            "257/257 [==============================] - 0s 683us/step - loss: 0.6757 - accuracy: 0.6654 - val_loss: 0.5772 - val_accuracy: 0.7165\n",
            "Epoch 50/500\n",
            "257/257 [==============================] - 0s 665us/step - loss: 0.5985 - accuracy: 0.6498 - val_loss: 0.6285 - val_accuracy: 0.6142\n",
            "Epoch 51/500\n",
            "257/257 [==============================] - 0s 680us/step - loss: 0.6016 - accuracy: 0.6926 - val_loss: 0.7059 - val_accuracy: 0.7323\n",
            "Epoch 52/500\n",
            "257/257 [==============================] - 0s 692us/step - loss: 0.6573 - accuracy: 0.6459 - val_loss: 0.5796 - val_accuracy: 0.7008\n",
            "Epoch 53/500\n",
            "257/257 [==============================] - 0s 696us/step - loss: 0.6223 - accuracy: 0.6304 - val_loss: 0.6137 - val_accuracy: 0.7008\n",
            "Epoch 54/500\n",
            "257/257 [==============================] - 0s 672us/step - loss: 0.6232 - accuracy: 0.6342 - val_loss: 0.5689 - val_accuracy: 0.7244\n",
            "Epoch 55/500\n",
            "257/257 [==============================] - 0s 790us/step - loss: 0.6028 - accuracy: 0.6537 - val_loss: 0.6795 - val_accuracy: 0.7165\n",
            "Epoch 56/500\n",
            "257/257 [==============================] - 0s 723us/step - loss: 0.6035 - accuracy: 0.6537 - val_loss: 0.5623 - val_accuracy: 0.7087\n",
            "Epoch 57/500\n",
            "257/257 [==============================] - 0s 738us/step - loss: 0.5908 - accuracy: 0.6615 - val_loss: 0.6379 - val_accuracy: 0.7244\n",
            "Epoch 58/500\n",
            "257/257 [==============================] - 0s 713us/step - loss: 0.6532 - accuracy: 0.6693 - val_loss: 0.6379 - val_accuracy: 0.7087\n",
            "Epoch 59/500\n",
            "257/257 [==============================] - 0s 683us/step - loss: 0.6506 - accuracy: 0.6848 - val_loss: 0.6578 - val_accuracy: 0.6614\n",
            "Epoch 60/500\n",
            "257/257 [==============================] - 0s 686us/step - loss: 0.6819 - accuracy: 0.6459 - val_loss: 0.6442 - val_accuracy: 0.7244\n",
            "Epoch 61/500\n",
            "257/257 [==============================] - 0s 684us/step - loss: 0.6538 - accuracy: 0.6887 - val_loss: 0.6354 - val_accuracy: 0.7323\n",
            "Epoch 62/500\n",
            "257/257 [==============================] - 0s 686us/step - loss: 0.6467 - accuracy: 0.6926 - val_loss: 0.6201 - val_accuracy: 0.7323\n",
            "Epoch 63/500\n",
            "257/257 [==============================] - 0s 676us/step - loss: 0.6334 - accuracy: 0.6848 - val_loss: 0.5840 - val_accuracy: 0.7402\n",
            "Epoch 64/500\n",
            "257/257 [==============================] - 0s 661us/step - loss: 0.5963 - accuracy: 0.6926 - val_loss: 0.8889 - val_accuracy: 0.3071\n",
            "Epoch 65/500\n",
            "257/257 [==============================] - 0s 730us/step - loss: 0.7350 - accuracy: 0.4747 - val_loss: 0.6620 - val_accuracy: 0.6929\n",
            "Epoch 66/500\n",
            "257/257 [==============================] - 0s 777us/step - loss: 0.6641 - accuracy: 0.6187 - val_loss: 0.6236 - val_accuracy: 0.6929\n",
            "Epoch 67/500\n",
            "257/257 [==============================] - 0s 723us/step - loss: 0.6642 - accuracy: 0.6187 - val_loss: 0.6176 - val_accuracy: 0.6929\n",
            "Epoch 68/500\n",
            "257/257 [==============================] - 0s 715us/step - loss: 0.6621 - accuracy: 0.6187 - val_loss: 0.6278 - val_accuracy: 0.6929\n",
            "Epoch 69/500\n",
            "257/257 [==============================] - 0s 711us/step - loss: 0.6499 - accuracy: 0.6420 - val_loss: 0.8121 - val_accuracy: 0.3150\n",
            "Epoch 70/500\n",
            "257/257 [==============================] - 0s 660us/step - loss: 0.7509 - accuracy: 0.5525 - val_loss: 0.6154 - val_accuracy: 0.6929\n",
            "Epoch 71/500\n",
            "257/257 [==============================] - 0s 711us/step - loss: 0.6474 - accuracy: 0.6187 - val_loss: 0.6001 - val_accuracy: 0.6929\n",
            "Epoch 72/500\n",
            "257/257 [==============================] - 0s 673us/step - loss: 0.6289 - accuracy: 0.6226 - val_loss: 0.6598 - val_accuracy: 0.6929\n",
            "Epoch 73/500\n",
            "257/257 [==============================] - 0s 674us/step - loss: 0.6836 - accuracy: 0.6304 - val_loss: 0.6019 - val_accuracy: 0.6929\n",
            "Epoch 74/500\n",
            "257/257 [==============================] - 0s 700us/step - loss: 0.6443 - accuracy: 0.6187 - val_loss: 0.5917 - val_accuracy: 0.6929\n",
            "Epoch 75/500\n",
            "257/257 [==============================] - 0s 683us/step - loss: 0.6350 - accuracy: 0.6187 - val_loss: 0.6168 - val_accuracy: 0.7008\n",
            "Epoch 76/500\n",
            "257/257 [==============================] - 0s 705us/step - loss: 0.6359 - accuracy: 0.6187 - val_loss: 0.6068 - val_accuracy: 0.7008\n",
            "Epoch 77/500\n",
            "257/257 [==============================] - 0s 736us/step - loss: 0.6155 - accuracy: 0.6887 - val_loss: 0.5953 - val_accuracy: 0.6772\n",
            "Epoch 78/500\n",
            "257/257 [==============================] - 0s 766us/step - loss: 0.5956 - accuracy: 0.6926 - val_loss: 0.5738 - val_accuracy: 0.6929\n",
            "Epoch 79/500\n",
            "257/257 [==============================] - 0s 704us/step - loss: 0.5951 - accuracy: 0.6615 - val_loss: 0.5867 - val_accuracy: 0.7717\n",
            "Epoch 80/500\n",
            "257/257 [==============================] - 0s 732us/step - loss: 0.5969 - accuracy: 0.7121 - val_loss: 0.5666 - val_accuracy: 0.7402\n",
            "Epoch 81/500\n",
            "257/257 [==============================] - 0s 782us/step - loss: 0.5853 - accuracy: 0.7432 - val_loss: 0.6655 - val_accuracy: 0.4252\n",
            "Epoch 82/500\n",
            "257/257 [==============================] - 0s 800us/step - loss: 0.6413 - accuracy: 0.5564 - val_loss: 0.5953 - val_accuracy: 0.6850\n",
            "Epoch 83/500\n",
            "257/257 [==============================] - 0s 713us/step - loss: 0.6433 - accuracy: 0.6654 - val_loss: 0.7325 - val_accuracy: 0.3071\n",
            "Epoch 84/500\n",
            "257/257 [==============================] - 0s 692us/step - loss: 0.7007 - accuracy: 0.5097 - val_loss: 0.6785 - val_accuracy: 0.6929\n",
            "Epoch 85/500\n",
            "257/257 [==============================] - 0s 731us/step - loss: 0.6835 - accuracy: 0.6187 - val_loss: 0.6751 - val_accuracy: 0.6929\n",
            "Epoch 86/500\n",
            "257/257 [==============================] - 0s 722us/step - loss: 0.6825 - accuracy: 0.6187 - val_loss: 0.6751 - val_accuracy: 0.6929\n",
            "Epoch 87/500\n",
            "257/257 [==============================] - 0s 716us/step - loss: 0.6825 - accuracy: 0.6187 - val_loss: 0.6744 - val_accuracy: 0.6929\n",
            "Epoch 88/500\n",
            "257/257 [==============================] - 0s 763us/step - loss: 0.6820 - accuracy: 0.6187 - val_loss: 0.6735 - val_accuracy: 0.6929\n",
            "Epoch 89/500\n",
            "257/257 [==============================] - 0s 712us/step - loss: 0.6804 - accuracy: 0.6226 - val_loss: 0.6704 - val_accuracy: 0.7323\n",
            "Epoch 90/500\n",
            "257/257 [==============================] - 0s 696us/step - loss: 0.6810 - accuracy: 0.6187 - val_loss: 0.7045 - val_accuracy: 0.4016\n",
            "Epoch 91/500\n",
            "257/257 [==============================] - 0s 659us/step - loss: 0.6922 - accuracy: 0.5837 - val_loss: 0.6590 - val_accuracy: 0.6929\n",
            "Epoch 92/500\n",
            "257/257 [==============================] - 0s 676us/step - loss: 0.6777 - accuracy: 0.6187 - val_loss: 0.6524 - val_accuracy: 0.6929\n",
            "Epoch 93/500\n",
            "257/257 [==============================] - 0s 687us/step - loss: 0.6751 - accuracy: 0.6187 - val_loss: 0.6635 - val_accuracy: 0.7165\n",
            "Epoch 94/500\n",
            "257/257 [==============================] - 0s 701us/step - loss: 0.6689 - accuracy: 0.6342 - val_loss: 0.6402 - val_accuracy: 0.6929\n",
            "Epoch 95/500\n",
            "257/257 [==============================] - 0s 713us/step - loss: 0.6618 - accuracy: 0.6187 - val_loss: 0.6401 - val_accuracy: 0.6929\n",
            "Epoch 96/500\n",
            "257/257 [==============================] - 0s 879us/step - loss: 0.6809 - accuracy: 0.6187 - val_loss: 0.6311 - val_accuracy: 0.6929\n",
            "Epoch 97/500\n",
            "257/257 [==============================] - 0s 676us/step - loss: 0.6731 - accuracy: 0.6187 - val_loss: 0.6259 - val_accuracy: 0.6929\n",
            "Epoch 98/500\n",
            "257/257 [==============================] - 0s 680us/step - loss: 0.6734 - accuracy: 0.6187 - val_loss: 0.6233 - val_accuracy: 0.6929\n",
            "Epoch 99/500\n",
            "257/257 [==============================] - 0s 660us/step - loss: 0.6706 - accuracy: 0.6187 - val_loss: 0.6141 - val_accuracy: 0.6929\n",
            "Epoch 100/500\n",
            "257/257 [==============================] - 0s 715us/step - loss: 0.6634 - accuracy: 0.6187 - val_loss: 0.6077 - val_accuracy: 0.6929\n",
            "Epoch 101/500\n",
            "257/257 [==============================] - 0s 688us/step - loss: 0.6365 - accuracy: 0.6187 - val_loss: 0.6124 - val_accuracy: 0.6929\n",
            "Epoch 102/500\n",
            "257/257 [==============================] - 0s 680us/step - loss: 0.6511 - accuracy: 0.6187 - val_loss: 0.6658 - val_accuracy: 0.6929\n",
            "Epoch 103/500\n",
            "257/257 [==============================] - 0s 683us/step - loss: 0.7020 - accuracy: 0.6187 - val_loss: 0.6677 - val_accuracy: 0.6929\n",
            "Epoch 104/500\n",
            "257/257 [==============================] - 0s 681us/step - loss: 0.6792 - accuracy: 0.6187 - val_loss: 0.6678 - val_accuracy: 0.6929\n",
            "Epoch 105/500\n",
            "257/257 [==============================] - 0s 671us/step - loss: 0.6787 - accuracy: 0.6187 - val_loss: 0.6681 - val_accuracy: 0.6929\n",
            "Epoch 106/500\n",
            "257/257 [==============================] - 0s 717us/step - loss: 0.6787 - accuracy: 0.6187 - val_loss: 0.6675 - val_accuracy: 0.6929\n",
            "Epoch 107/500\n",
            "257/257 [==============================] - 0s 709us/step - loss: 0.6783 - accuracy: 0.6187 - val_loss: 0.6667 - val_accuracy: 0.6929\n",
            "Epoch 108/500\n",
            "257/257 [==============================] - 0s 665us/step - loss: 0.6779 - accuracy: 0.6187 - val_loss: 0.6665 - val_accuracy: 0.6929\n",
            "Epoch 109/500\n",
            "257/257 [==============================] - 0s 735us/step - loss: 0.6778 - accuracy: 0.6187 - val_loss: 0.6658 - val_accuracy: 0.6929\n",
            "Epoch 110/500\n",
            "257/257 [==============================] - 0s 688us/step - loss: 0.6775 - accuracy: 0.6187 - val_loss: 0.6650 - val_accuracy: 0.6929\n",
            "Epoch 111/500\n",
            "257/257 [==============================] - 0s 659us/step - loss: 0.6770 - accuracy: 0.6187 - val_loss: 0.6641 - val_accuracy: 0.6929\n",
            "Epoch 112/500\n",
            "257/257 [==============================] - 0s 705us/step - loss: 0.6765 - accuracy: 0.6187 - val_loss: 0.6639 - val_accuracy: 0.6929\n",
            "Epoch 113/500\n",
            "257/257 [==============================] - 0s 697us/step - loss: 0.6765 - accuracy: 0.6187 - val_loss: 0.6632 - val_accuracy: 0.6929\n",
            "Epoch 114/500\n",
            "257/257 [==============================] - 0s 685us/step - loss: 0.6761 - accuracy: 0.6187 - val_loss: 0.6630 - val_accuracy: 0.6929\n",
            "Epoch 115/500\n",
            "257/257 [==============================] - 0s 699us/step - loss: 0.6761 - accuracy: 0.6187 - val_loss: 0.6624 - val_accuracy: 0.6929\n",
            "Epoch 116/500\n",
            "257/257 [==============================] - 0s 671us/step - loss: 0.6757 - accuracy: 0.6187 - val_loss: 0.6623 - val_accuracy: 0.6929\n",
            "Epoch 117/500\n",
            "257/257 [==============================] - 0s 717us/step - loss: 0.6757 - accuracy: 0.6187 - val_loss: 0.6617 - val_accuracy: 0.6929\n",
            "Epoch 118/500\n",
            "257/257 [==============================] - 0s 781us/step - loss: 0.6754 - accuracy: 0.6187 - val_loss: 0.6617 - val_accuracy: 0.6929\n",
            "Epoch 119/500\n",
            "257/257 [==============================] - 0s 684us/step - loss: 0.6754 - accuracy: 0.6187 - val_loss: 0.6611 - val_accuracy: 0.6929\n",
            "Epoch 120/500\n",
            "257/257 [==============================] - 0s 688us/step - loss: 0.6751 - accuracy: 0.6187 - val_loss: 0.6604 - val_accuracy: 0.6929\n",
            "Epoch 121/500\n",
            "257/257 [==============================] - 0s 678us/step - loss: 0.6747 - accuracy: 0.6187 - val_loss: 0.6603 - val_accuracy: 0.6929\n",
            "Epoch 122/500\n",
            "257/257 [==============================] - 0s 692us/step - loss: 0.6747 - accuracy: 0.6187 - val_loss: 0.6597 - val_accuracy: 0.6929\n",
            "Epoch 123/500\n",
            "257/257 [==============================] - 0s 692us/step - loss: 0.6744 - accuracy: 0.6187 - val_loss: 0.6596 - val_accuracy: 0.6929\n",
            "Epoch 124/500\n",
            "257/257 [==============================] - 0s 693us/step - loss: 0.6744 - accuracy: 0.6187 - val_loss: 0.6590 - val_accuracy: 0.6929\n",
            "Epoch 125/500\n",
            "257/257 [==============================] - 0s 712us/step - loss: 0.6741 - accuracy: 0.6187 - val_loss: 0.6590 - val_accuracy: 0.6929\n",
            "Epoch 126/500\n",
            "257/257 [==============================] - 0s 677us/step - loss: 0.6741 - accuracy: 0.6187 - val_loss: 0.6585 - val_accuracy: 0.6929\n",
            "Epoch 127/500\n",
            "257/257 [==============================] - 0s 675us/step - loss: 0.6738 - accuracy: 0.6187 - val_loss: 0.6578 - val_accuracy: 0.6929\n",
            "Epoch 128/500\n",
            "257/257 [==============================] - 0s 699us/step - loss: 0.6735 - accuracy: 0.6187 - val_loss: 0.6571 - val_accuracy: 0.6929\n",
            "Epoch 129/500\n",
            "257/257 [==============================] - 0s 723us/step - loss: 0.6731 - accuracy: 0.6187 - val_loss: 0.6569 - val_accuracy: 0.6929\n",
            "Epoch 130/500\n",
            "257/257 [==============================] - 0s 719us/step - loss: 0.6731 - accuracy: 0.6187 - val_loss: 0.6564 - val_accuracy: 0.6929\n",
            "Epoch 131/500\n",
            "257/257 [==============================] - 0s 676us/step - loss: 0.6728 - accuracy: 0.6187 - val_loss: 0.6563 - val_accuracy: 0.6929\n",
            "Epoch 132/500\n",
            "257/257 [==============================] - 0s 679us/step - loss: 0.6729 - accuracy: 0.6187 - val_loss: 0.6558 - val_accuracy: 0.6929\n",
            "Epoch 133/500\n",
            "257/257 [==============================] - 0s 690us/step - loss: 0.6726 - accuracy: 0.6187 - val_loss: 0.6552 - val_accuracy: 0.6929\n",
            "Epoch 134/500\n",
            "257/257 [==============================] - 0s 683us/step - loss: 0.6723 - accuracy: 0.6187 - val_loss: 0.6545 - val_accuracy: 0.6929\n",
            "Epoch 135/500\n",
            "257/257 [==============================] - 0s 770us/step - loss: 0.6720 - accuracy: 0.6187 - val_loss: 0.6544 - val_accuracy: 0.6929\n",
            "Epoch 136/500\n",
            "257/257 [==============================] - 0s 763us/step - loss: 0.6720 - accuracy: 0.6187 - val_loss: 0.6538 - val_accuracy: 0.6929\n",
            "Epoch 137/500\n",
            "257/257 [==============================] - 0s 744us/step - loss: 0.6717 - accuracy: 0.6187 - val_loss: 0.6532 - val_accuracy: 0.6929\n",
            "Epoch 138/500\n",
            "257/257 [==============================] - 0s 710us/step - loss: 0.6715 - accuracy: 0.6187 - val_loss: 0.6525 - val_accuracy: 0.6929\n",
            "Epoch 139/500\n",
            "257/257 [==============================] - 0s 737us/step - loss: 0.6711 - accuracy: 0.6187 - val_loss: 0.6519 - val_accuracy: 0.6929\n",
            "Epoch 140/500\n",
            "257/257 [==============================] - 0s 709us/step - loss: 0.6708 - accuracy: 0.6187 - val_loss: 0.6518 - val_accuracy: 0.6929\n",
            "Epoch 141/500\n",
            "257/257 [==============================] - 0s 772us/step - loss: 0.6708 - accuracy: 0.6187 - val_loss: 0.6513 - val_accuracy: 0.6929\n",
            "Epoch 142/500\n",
            "257/257 [==============================] - 0s 686us/step - loss: 0.6706 - accuracy: 0.6187 - val_loss: 0.6507 - val_accuracy: 0.6929\n",
            "Epoch 143/500\n",
            "257/257 [==============================] - 0s 681us/step - loss: 0.6704 - accuracy: 0.6187 - val_loss: 0.6500 - val_accuracy: 0.6929\n",
            "Epoch 144/500\n",
            "257/257 [==============================] - 0s 677us/step - loss: 0.6701 - accuracy: 0.6187 - val_loss: 0.6494 - val_accuracy: 0.6929\n",
            "Epoch 145/500\n",
            "257/257 [==============================] - 0s 695us/step - loss: 0.6698 - accuracy: 0.6187 - val_loss: 0.6487 - val_accuracy: 0.6929\n",
            "Epoch 146/500\n",
            "257/257 [==============================] - 0s 702us/step - loss: 0.6696 - accuracy: 0.6187 - val_loss: 0.6481 - val_accuracy: 0.6929\n",
            "Epoch 147/500\n",
            "257/257 [==============================] - 0s 670us/step - loss: 0.6693 - accuracy: 0.6187 - val_loss: 0.6480 - val_accuracy: 0.6929\n",
            "Epoch 148/500\n",
            "257/257 [==============================] - 0s 704us/step - loss: 0.6693 - accuracy: 0.6187 - val_loss: 0.6476 - val_accuracy: 0.6929\n",
            "Epoch 149/500\n",
            "257/257 [==============================] - 0s 686us/step - loss: 0.6691 - accuracy: 0.6187 - val_loss: 0.6476 - val_accuracy: 0.6929\n",
            "Epoch 150/500\n",
            "257/257 [==============================] - 0s 685us/step - loss: 0.6692 - accuracy: 0.6187 - val_loss: 0.6472 - val_accuracy: 0.6929\n",
            "Epoch 151/500\n",
            "257/257 [==============================] - 0s 719us/step - loss: 0.6690 - accuracy: 0.6187 - val_loss: 0.6473 - val_accuracy: 0.6929\n",
            "Epoch 152/500\n",
            "257/257 [==============================] - 0s 772us/step - loss: 0.6691 - accuracy: 0.6187 - val_loss: 0.6475 - val_accuracy: 0.6929\n",
            "Epoch 153/500\n",
            "257/257 [==============================] - 0s 702us/step - loss: 0.6691 - accuracy: 0.6187 - val_loss: 0.6478 - val_accuracy: 0.6929\n",
            "Epoch 154/500\n",
            "257/257 [==============================] - 0s 672us/step - loss: 0.6693 - accuracy: 0.6187 - val_loss: 0.6476 - val_accuracy: 0.6929\n",
            "Epoch 155/500\n",
            "257/257 [==============================] - 0s 683us/step - loss: 0.6692 - accuracy: 0.6187 - val_loss: 0.6477 - val_accuracy: 0.6929\n",
            "Epoch 156/500\n",
            "257/257 [==============================] - 0s 678us/step - loss: 0.6692 - accuracy: 0.6187 - val_loss: 0.6474 - val_accuracy: 0.6929\n",
            "Epoch 157/500\n",
            "257/257 [==============================] - 0s 677us/step - loss: 0.6691 - accuracy: 0.6187 - val_loss: 0.6476 - val_accuracy: 0.6929\n",
            "Epoch 158/500\n",
            "257/257 [==============================] - 0s 683us/step - loss: 0.6692 - accuracy: 0.6187 - val_loss: 0.6472 - val_accuracy: 0.6929\n",
            "Epoch 159/500\n",
            "257/257 [==============================] - 0s 710us/step - loss: 0.6690 - accuracy: 0.6187 - val_loss: 0.6473 - val_accuracy: 0.6929\n",
            "Epoch 160/500\n",
            "257/257 [==============================] - 0s 678us/step - loss: 0.6691 - accuracy: 0.6187 - val_loss: 0.6476 - val_accuracy: 0.6929\n",
            "Epoch 161/500\n",
            "257/257 [==============================] - 0s 674us/step - loss: 0.6692 - accuracy: 0.6187 - val_loss: 0.6473 - val_accuracy: 0.6929\n",
            "Epoch 162/500\n",
            "257/257 [==============================] - 0s 687us/step - loss: 0.6690 - accuracy: 0.6187 - val_loss: 0.6469 - val_accuracy: 0.6929\n",
            "Epoch 163/500\n",
            "257/257 [==============================] - 0s 711us/step - loss: 0.6689 - accuracy: 0.6187 - val_loss: 0.6464 - val_accuracy: 0.6929\n",
            "Epoch 164/500\n",
            "257/257 [==============================] - 0s 670us/step - loss: 0.6687 - accuracy: 0.6187 - val_loss: 0.6459 - val_accuracy: 0.6929\n",
            "Epoch 165/500\n",
            "257/257 [==============================] - 0s 687us/step - loss: 0.6685 - accuracy: 0.6187 - val_loss: 0.6458 - val_accuracy: 0.6929\n",
            "Epoch 166/500\n",
            "257/257 [==============================] - 0s 689us/step - loss: 0.6685 - accuracy: 0.6187 - val_loss: 0.6455 - val_accuracy: 0.6929\n",
            "Epoch 167/500\n",
            "257/257 [==============================] - 0s 684us/step - loss: 0.6683 - accuracy: 0.6187 - val_loss: 0.6450 - val_accuracy: 0.6929\n",
            "Epoch 168/500\n",
            "257/257 [==============================] - 0s 688us/step - loss: 0.6682 - accuracy: 0.6187 - val_loss: 0.6450 - val_accuracy: 0.6929\n",
            "Epoch 169/500\n",
            "257/257 [==============================] - 0s 701us/step - loss: 0.6682 - accuracy: 0.6187 - val_loss: 0.6452 - val_accuracy: 0.6929\n",
            "Epoch 170/500\n",
            "257/257 [==============================] - 0s 685us/step - loss: 0.6683 - accuracy: 0.6187 - val_loss: 0.6449 - val_accuracy: 0.6929\n",
            "Epoch 171/500\n",
            "257/257 [==============================] - 0s 682us/step - loss: 0.6682 - accuracy: 0.6187 - val_loss: 0.6451 - val_accuracy: 0.6929\n",
            "Epoch 172/500\n",
            "257/257 [==============================] - 0s 693us/step - loss: 0.6682 - accuracy: 0.6187 - val_loss: 0.6453 - val_accuracy: 0.6929\n",
            "Epoch 173/500\n",
            "257/257 [==============================] - 0s 683us/step - loss: 0.6683 - accuracy: 0.6187 - val_loss: 0.6451 - val_accuracy: 0.6929\n",
            "Epoch 174/500\n",
            "257/257 [==============================] - 0s 689us/step - loss: 0.6682 - accuracy: 0.6187 - val_loss: 0.6447 - val_accuracy: 0.6929\n",
            "Epoch 175/500\n",
            "257/257 [==============================] - 0s 676us/step - loss: 0.6681 - accuracy: 0.6187 - val_loss: 0.6443 - val_accuracy: 0.6929\n",
            "Epoch 176/500\n",
            "257/257 [==============================] - 0s 682us/step - loss: 0.6680 - accuracy: 0.6187 - val_loss: 0.6443 - val_accuracy: 0.6929\n",
            "Epoch 177/500\n",
            "257/257 [==============================] - 0s 673us/step - loss: 0.6679 - accuracy: 0.6187 - val_loss: 0.6445 - val_accuracy: 0.6929\n",
            "Epoch 178/500\n",
            "257/257 [==============================] - 0s 672us/step - loss: 0.6680 - accuracy: 0.6187 - val_loss: 0.6448 - val_accuracy: 0.6929\n",
            "Epoch 179/500\n",
            "257/257 [==============================] - 0s 714us/step - loss: 0.6681 - accuracy: 0.6187 - val_loss: 0.6451 - val_accuracy: 0.6929\n",
            "Epoch 180/500\n",
            "257/257 [==============================] - 0s 702us/step - loss: 0.6682 - accuracy: 0.6187 - val_loss: 0.6454 - val_accuracy: 0.6929\n",
            "Epoch 181/500\n",
            "257/257 [==============================] - 0s 719us/step - loss: 0.6684 - accuracy: 0.6187 - val_loss: 0.6453 - val_accuracy: 0.6929\n",
            "Epoch 182/500\n",
            "257/257 [==============================] - 0s 731us/step - loss: 0.6683 - accuracy: 0.6187 - val_loss: 0.6455 - val_accuracy: 0.6929\n",
            "Epoch 183/500\n",
            "257/257 [==============================] - 0s 741us/step - loss: 0.6684 - accuracy: 0.6187 - val_loss: 0.6452 - val_accuracy: 0.6929\n",
            "Epoch 184/500\n",
            "257/257 [==============================] - 0s 680us/step - loss: 0.6683 - accuracy: 0.6187 - val_loss: 0.6454 - val_accuracy: 0.6929\n",
            "Epoch 185/500\n",
            "257/257 [==============================] - 0s 724us/step - loss: 0.6683 - accuracy: 0.6187 - val_loss: 0.6456 - val_accuracy: 0.6929\n",
            "Epoch 186/500\n",
            "257/257 [==============================] - 0s 686us/step - loss: 0.6684 - accuracy: 0.6187 - val_loss: 0.6454 - val_accuracy: 0.6929\n",
            "Epoch 187/500\n",
            "257/257 [==============================] - 0s 676us/step - loss: 0.6683 - accuracy: 0.6187 - val_loss: 0.6456 - val_accuracy: 0.6929\n",
            "Epoch 188/500\n",
            "257/257 [==============================] - 0s 688us/step - loss: 0.6684 - accuracy: 0.6187 - val_loss: 0.6453 - val_accuracy: 0.6929\n",
            "Epoch 189/500\n",
            "257/257 [==============================] - 0s 723us/step - loss: 0.6683 - accuracy: 0.6187 - val_loss: 0.6449 - val_accuracy: 0.6929\n",
            "Epoch 190/500\n",
            "257/257 [==============================] - 0s 678us/step - loss: 0.6681 - accuracy: 0.6187 - val_loss: 0.6444 - val_accuracy: 0.6929\n",
            "Epoch 191/500\n",
            "257/257 [==============================] - 0s 724us/step - loss: 0.6680 - accuracy: 0.6187 - val_loss: 0.6444 - val_accuracy: 0.6929\n",
            "Epoch 192/500\n",
            "257/257 [==============================] - 0s 676us/step - loss: 0.6680 - accuracy: 0.6187 - val_loss: 0.6446 - val_accuracy: 0.6929\n",
            "Epoch 193/500\n",
            "257/257 [==============================] - 0s 694us/step - loss: 0.6681 - accuracy: 0.6187 - val_loss: 0.6449 - val_accuracy: 0.6929\n",
            "Epoch 194/500\n",
            "257/257 [==============================] - 0s 682us/step - loss: 0.6682 - accuracy: 0.6187 - val_loss: 0.6452 - val_accuracy: 0.6929\n",
            "Epoch 195/500\n",
            "257/257 [==============================] - 0s 688us/step - loss: 0.6683 - accuracy: 0.6187 - val_loss: 0.6450 - val_accuracy: 0.6929\n",
            "Epoch 196/500\n",
            "257/257 [==============================] - 0s 668us/step - loss: 0.6682 - accuracy: 0.6187 - val_loss: 0.6447 - val_accuracy: 0.6929\n",
            "Epoch 197/500\n",
            "257/257 [==============================] - 0s 726us/step - loss: 0.6681 - accuracy: 0.6187 - val_loss: 0.6448 - val_accuracy: 0.6929\n",
            "Epoch 198/500\n",
            "257/257 [==============================] - 0s 690us/step - loss: 0.6681 - accuracy: 0.6187 - val_loss: 0.6450 - val_accuracy: 0.6929\n",
            "Epoch 199/500\n",
            "257/257 [==============================] - 0s 684us/step - loss: 0.6682 - accuracy: 0.6187 - val_loss: 0.6453 - val_accuracy: 0.6929\n",
            "Epoch 200/500\n",
            "257/257 [==============================] - 0s 694us/step - loss: 0.6683 - accuracy: 0.6187 - val_loss: 0.6456 - val_accuracy: 0.6929\n",
            "Epoch 201/500\n",
            "257/257 [==============================] - 0s 696us/step - loss: 0.6684 - accuracy: 0.6187 - val_loss: 0.6455 - val_accuracy: 0.6929\n",
            "Epoch 202/500\n",
            "257/257 [==============================] - 0s 686us/step - loss: 0.6684 - accuracy: 0.6187 - val_loss: 0.6457 - val_accuracy: 0.6929\n",
            "Epoch 203/500\n",
            "257/257 [==============================] - 0s 710us/step - loss: 0.6684 - accuracy: 0.6187 - val_loss: 0.6459 - val_accuracy: 0.6929\n",
            "Epoch 204/500\n",
            "257/257 [==============================] - 0s 675us/step - loss: 0.6686 - accuracy: 0.6187 - val_loss: 0.6457 - val_accuracy: 0.6929\n",
            "Epoch 205/500\n",
            "257/257 [==============================] - 0s 694us/step - loss: 0.6685 - accuracy: 0.6187 - val_loss: 0.6454 - val_accuracy: 0.6929\n",
            "Epoch 206/500\n",
            "257/257 [==============================] - 0s 715us/step - loss: 0.6683 - accuracy: 0.6187 - val_loss: 0.6449 - val_accuracy: 0.6929\n",
            "Epoch 207/500\n",
            "257/257 [==============================] - 0s 690us/step - loss: 0.6681 - accuracy: 0.6187 - val_loss: 0.6444 - val_accuracy: 0.6929\n",
            "Epoch 208/500\n",
            "257/257 [==============================] - 0s 698us/step - loss: 0.6680 - accuracy: 0.6187 - val_loss: 0.6439 - val_accuracy: 0.6929\n",
            "Epoch 209/500\n",
            "257/257 [==============================] - 0s 768us/step - loss: 0.6678 - accuracy: 0.6187 - val_loss: 0.6434 - val_accuracy: 0.6929\n",
            "Epoch 210/500\n",
            "257/257 [==============================] - 0s 701us/step - loss: 0.6676 - accuracy: 0.6187 - val_loss: 0.6428 - val_accuracy: 0.6929\n",
            "Epoch 211/500\n",
            "257/257 [==============================] - 0s 665us/step - loss: 0.6675 - accuracy: 0.6187 - val_loss: 0.6428 - val_accuracy: 0.6929\n",
            "Epoch 212/500\n",
            "257/257 [==============================] - 0s 677us/step - loss: 0.6675 - accuracy: 0.6187 - val_loss: 0.6425 - val_accuracy: 0.6929\n",
            "Epoch 213/500\n",
            "257/257 [==============================] - 0s 693us/step - loss: 0.6673 - accuracy: 0.6187 - val_loss: 0.6426 - val_accuracy: 0.6929\n",
            "Epoch 214/500\n",
            "257/257 [==============================] - 0s 717us/step - loss: 0.6674 - accuracy: 0.6187 - val_loss: 0.6423 - val_accuracy: 0.6929\n",
            "Epoch 215/500\n",
            "257/257 [==============================] - 0s 687us/step - loss: 0.6673 - accuracy: 0.6187 - val_loss: 0.6424 - val_accuracy: 0.6929\n",
            "Epoch 216/500\n",
            "257/257 [==============================] - 0s 686us/step - loss: 0.6673 - accuracy: 0.6187 - val_loss: 0.6421 - val_accuracy: 0.6929\n",
            "Epoch 217/500\n",
            "257/257 [==============================] - 0s 703us/step - loss: 0.6672 - accuracy: 0.6187 - val_loss: 0.6418 - val_accuracy: 0.6929\n",
            "Epoch 218/500\n",
            "257/257 [==============================] - 0s 742us/step - loss: 0.6671 - accuracy: 0.6187 - val_loss: 0.6413 - val_accuracy: 0.6929\n",
            "Epoch 219/500\n",
            "257/257 [==============================] - 0s 919us/step - loss: 0.6670 - accuracy: 0.6187 - val_loss: 0.6414 - val_accuracy: 0.6929\n",
            "Epoch 220/500\n",
            "257/257 [==============================] - 0s 749us/step - loss: 0.6670 - accuracy: 0.6187 - val_loss: 0.6411 - val_accuracy: 0.6929\n",
            "Epoch 221/500\n",
            "257/257 [==============================] - 0s 739us/step - loss: 0.6669 - accuracy: 0.6187 - val_loss: 0.6407 - val_accuracy: 0.6929\n",
            "Epoch 222/500\n",
            "257/257 [==============================] - 0s 662us/step - loss: 0.6668 - accuracy: 0.6187 - val_loss: 0.6403 - val_accuracy: 0.6929\n",
            "Epoch 223/500\n",
            "257/257 [==============================] - 0s 654us/step - loss: 0.6667 - accuracy: 0.6187 - val_loss: 0.6398 - val_accuracy: 0.6929\n",
            "Epoch 224/500\n",
            "257/257 [==============================] - 0s 704us/step - loss: 0.6665 - accuracy: 0.6187 - val_loss: 0.6394 - val_accuracy: 0.6929\n",
            "Epoch 225/500\n",
            "257/257 [==============================] - 0s 658us/step - loss: 0.6664 - accuracy: 0.6187 - val_loss: 0.6389 - val_accuracy: 0.6929\n",
            "Epoch 226/500\n",
            "257/257 [==============================] - 0s 677us/step - loss: 0.6663 - accuracy: 0.6187 - val_loss: 0.6385 - val_accuracy: 0.6929\n",
            "Epoch 227/500\n",
            "257/257 [==============================] - 0s 709us/step - loss: 0.6661 - accuracy: 0.6187 - val_loss: 0.6381 - val_accuracy: 0.6929\n",
            "Epoch 228/500\n",
            "257/257 [==============================] - 0s 712us/step - loss: 0.6660 - accuracy: 0.6187 - val_loss: 0.6376 - val_accuracy: 0.6929\n",
            "Epoch 229/500\n",
            "257/257 [==============================] - 0s 707us/step - loss: 0.6659 - accuracy: 0.6187 - val_loss: 0.6372 - val_accuracy: 0.6929\n",
            "Epoch 230/500\n",
            "257/257 [==============================] - 0s 761us/step - loss: 0.6658 - accuracy: 0.6187 - val_loss: 0.6372 - val_accuracy: 0.6929\n",
            "Epoch 231/500\n",
            "257/257 [==============================] - 0s 757us/step - loss: 0.6658 - accuracy: 0.6187 - val_loss: 0.6374 - val_accuracy: 0.6929\n",
            "Epoch 232/500\n",
            "257/257 [==============================] - 0s 708us/step - loss: 0.6659 - accuracy: 0.6187 - val_loss: 0.6372 - val_accuracy: 0.6929\n",
            "Epoch 233/500\n",
            "257/257 [==============================] - 0s 679us/step - loss: 0.6659 - accuracy: 0.6187 - val_loss: 0.6370 - val_accuracy: 0.6929\n",
            "Epoch 234/500\n",
            "257/257 [==============================] - 0s 678us/step - loss: 0.6658 - accuracy: 0.6187 - val_loss: 0.6371 - val_accuracy: 0.6929\n",
            "Epoch 235/500\n",
            "257/257 [==============================] - 0s 751us/step - loss: 0.6658 - accuracy: 0.6187 - val_loss: 0.6373 - val_accuracy: 0.6929\n",
            "Epoch 236/500\n",
            "257/257 [==============================] - 0s 743us/step - loss: 0.6659 - accuracy: 0.6187 - val_loss: 0.6376 - val_accuracy: 0.6929\n",
            "Epoch 237/500\n",
            "257/257 [==============================] - 0s 729us/step - loss: 0.6660 - accuracy: 0.6187 - val_loss: 0.6380 - val_accuracy: 0.6929\n",
            "Epoch 238/500\n",
            "257/257 [==============================] - 0s 686us/step - loss: 0.6661 - accuracy: 0.6187 - val_loss: 0.6379 - val_accuracy: 0.6929\n",
            "Epoch 239/500\n",
            "257/257 [==============================] - 0s 706us/step - loss: 0.6660 - accuracy: 0.6187 - val_loss: 0.6377 - val_accuracy: 0.6929\n",
            "Epoch 240/500\n",
            "257/257 [==============================] - 0s 678us/step - loss: 0.6660 - accuracy: 0.6187 - val_loss: 0.6374 - val_accuracy: 0.6929\n",
            "Epoch 241/500\n",
            "257/257 [==============================] - 0s 705us/step - loss: 0.6659 - accuracy: 0.6187 - val_loss: 0.6371 - val_accuracy: 0.6929\n",
            "Epoch 242/500\n",
            "257/257 [==============================] - 0s 700us/step - loss: 0.6658 - accuracy: 0.6187 - val_loss: 0.6371 - val_accuracy: 0.6929\n",
            "Epoch 243/500\n",
            "257/257 [==============================] - 0s 726us/step - loss: 0.6658 - accuracy: 0.6187 - val_loss: 0.6369 - val_accuracy: 0.6929\n",
            "Epoch 244/500\n",
            "257/257 [==============================] - 0s 738us/step - loss: 0.6658 - accuracy: 0.6187 - val_loss: 0.6370 - val_accuracy: 0.6929\n",
            "Epoch 245/500\n",
            "257/257 [==============================] - 0s 719us/step - loss: 0.6658 - accuracy: 0.6187 - val_loss: 0.6373 - val_accuracy: 0.6929\n",
            "Epoch 246/500\n",
            "257/257 [==============================] - 0s 776us/step - loss: 0.6659 - accuracy: 0.6187 - val_loss: 0.6372 - val_accuracy: 0.6929\n",
            "Epoch 247/500\n",
            "257/257 [==============================] - 0s 729us/step - loss: 0.6658 - accuracy: 0.6187 - val_loss: 0.6374 - val_accuracy: 0.6929\n",
            "Epoch 248/500\n",
            "257/257 [==============================] - 0s 746us/step - loss: 0.6659 - accuracy: 0.6187 - val_loss: 0.6376 - val_accuracy: 0.6929\n",
            "Epoch 249/500\n",
            "257/257 [==============================] - 0s 778us/step - loss: 0.6660 - accuracy: 0.6187 - val_loss: 0.6380 - val_accuracy: 0.6929\n",
            "Epoch 250/500\n",
            "257/257 [==============================] - 0s 708us/step - loss: 0.6661 - accuracy: 0.6187 - val_loss: 0.6379 - val_accuracy: 0.6929\n",
            "Epoch 251/500\n",
            "257/257 [==============================] - 0s 726us/step - loss: 0.6660 - accuracy: 0.6187 - val_loss: 0.6377 - val_accuracy: 0.6929\n",
            "Epoch 252/500\n",
            "257/257 [==============================] - 0s 702us/step - loss: 0.6660 - accuracy: 0.6187 - val_loss: 0.6374 - val_accuracy: 0.6929\n",
            "Epoch 253/500\n",
            "257/257 [==============================] - 0s 680us/step - loss: 0.6659 - accuracy: 0.6187 - val_loss: 0.6371 - val_accuracy: 0.6929\n",
            "Epoch 254/500\n",
            "257/257 [==============================] - 0s 719us/step - loss: 0.6658 - accuracy: 0.6187 - val_loss: 0.6367 - val_accuracy: 0.6929\n",
            "Epoch 255/500\n",
            "257/257 [==============================] - 0s 707us/step - loss: 0.6657 - accuracy: 0.6187 - val_loss: 0.6363 - val_accuracy: 0.6929\n",
            "Epoch 256/500\n",
            "257/257 [==============================] - 0s 686us/step - loss: 0.6656 - accuracy: 0.6187 - val_loss: 0.6359 - val_accuracy: 0.6929\n",
            "Epoch 257/500\n",
            "257/257 [==============================] - 0s 701us/step - loss: 0.6655 - accuracy: 0.6187 - val_loss: 0.6359 - val_accuracy: 0.6929\n",
            "Epoch 258/500\n",
            "257/257 [==============================] - 0s 674us/step - loss: 0.6656 - accuracy: 0.6187 - val_loss: 0.6357 - val_accuracy: 0.6929\n",
            "Epoch 259/500\n",
            "257/257 [==============================] - 0s 715us/step - loss: 0.6655 - accuracy: 0.6187 - val_loss: 0.6358 - val_accuracy: 0.6929\n",
            "Epoch 260/500\n",
            "257/257 [==============================] - 0s 668us/step - loss: 0.6655 - accuracy: 0.6187 - val_loss: 0.6361 - val_accuracy: 0.6929\n",
            "Epoch 261/500\n",
            "257/257 [==============================] - 0s 681us/step - loss: 0.6656 - accuracy: 0.6187 - val_loss: 0.6364 - val_accuracy: 0.6929\n",
            "Epoch 262/500\n",
            "257/257 [==============================] - 0s 686us/step - loss: 0.6657 - accuracy: 0.6187 - val_loss: 0.6363 - val_accuracy: 0.6929\n",
            "Epoch 263/500\n",
            "257/257 [==============================] - 0s 708us/step - loss: 0.6657 - accuracy: 0.6187 - val_loss: 0.6366 - val_accuracy: 0.6929\n",
            "Epoch 264/500\n",
            "257/257 [==============================] - 0s 798us/step - loss: 0.6657 - accuracy: 0.6187 - val_loss: 0.6364 - val_accuracy: 0.6929\n",
            "Epoch 265/500\n",
            "257/257 [==============================] - 0s 690us/step - loss: 0.6657 - accuracy: 0.6187 - val_loss: 0.6366 - val_accuracy: 0.6929\n",
            "Epoch 266/500\n",
            "257/257 [==============================] - 0s 688us/step - loss: 0.6657 - accuracy: 0.6187 - val_loss: 0.6365 - val_accuracy: 0.6929\n",
            "Epoch 267/500\n",
            "257/257 [==============================] - 0s 681us/step - loss: 0.6657 - accuracy: 0.6187 - val_loss: 0.6362 - val_accuracy: 0.6929\n",
            "Epoch 268/500\n",
            "257/257 [==============================] - 0s 754us/step - loss: 0.6656 - accuracy: 0.6187 - val_loss: 0.6359 - val_accuracy: 0.6929\n",
            "Epoch 269/500\n",
            "257/257 [==============================] - 0s 663us/step - loss: 0.6655 - accuracy: 0.6187 - val_loss: 0.6356 - val_accuracy: 0.6929\n",
            "Epoch 270/500\n",
            "257/257 [==============================] - 0s 692us/step - loss: 0.6655 - accuracy: 0.6187 - val_loss: 0.6356 - val_accuracy: 0.6929\n",
            "Epoch 271/500\n",
            "257/257 [==============================] - 0s 694us/step - loss: 0.6655 - accuracy: 0.6187 - val_loss: 0.6354 - val_accuracy: 0.6929\n",
            "Epoch 272/500\n",
            "257/257 [==============================] - 0s 673us/step - loss: 0.6655 - accuracy: 0.6187 - val_loss: 0.6355 - val_accuracy: 0.6929\n",
            "Epoch 273/500\n",
            "257/257 [==============================] - 0s 696us/step - loss: 0.6655 - accuracy: 0.6187 - val_loss: 0.6358 - val_accuracy: 0.6929\n",
            "Epoch 274/500\n",
            "257/257 [==============================] - 0s 721us/step - loss: 0.6655 - accuracy: 0.6187 - val_loss: 0.6361 - val_accuracy: 0.6929\n",
            "Epoch 275/500\n",
            "257/257 [==============================] - 0s 671us/step - loss: 0.6656 - accuracy: 0.6187 - val_loss: 0.6361 - val_accuracy: 0.6929\n",
            "Epoch 276/500\n",
            "257/257 [==============================] - 0s 686us/step - loss: 0.6656 - accuracy: 0.6187 - val_loss: 0.6359 - val_accuracy: 0.6929\n",
            "Epoch 277/500\n",
            "257/257 [==============================] - 0s 675us/step - loss: 0.6655 - accuracy: 0.6187 - val_loss: 0.6356 - val_accuracy: 0.6929\n",
            "Epoch 278/500\n",
            "257/257 [==============================] - 0s 673us/step - loss: 0.6655 - accuracy: 0.6187 - val_loss: 0.6353 - val_accuracy: 0.6929\n",
            "Epoch 279/500\n",
            "257/257 [==============================] - 0s 676us/step - loss: 0.6654 - accuracy: 0.6187 - val_loss: 0.6350 - val_accuracy: 0.6929\n",
            "Epoch 280/500\n",
            "257/257 [==============================] - 0s 725us/step - loss: 0.6654 - accuracy: 0.6187 - val_loss: 0.6350 - val_accuracy: 0.6929\n",
            "Epoch 281/500\n",
            "257/257 [==============================] - 0s 658us/step - loss: 0.6654 - accuracy: 0.6187 - val_loss: 0.6348 - val_accuracy: 0.6929\n",
            "Epoch 282/500\n",
            "257/257 [==============================] - 0s 701us/step - loss: 0.6653 - accuracy: 0.6187 - val_loss: 0.6349 - val_accuracy: 0.6929\n",
            "Epoch 283/500\n",
            "257/257 [==============================] - 0s 677us/step - loss: 0.6654 - accuracy: 0.6187 - val_loss: 0.6352 - val_accuracy: 0.6929\n",
            "Epoch 284/500\n",
            "257/257 [==============================] - 0s 664us/step - loss: 0.6654 - accuracy: 0.6187 - val_loss: 0.6355 - val_accuracy: 0.6929\n",
            "Epoch 285/500\n",
            "257/257 [==============================] - 0s 718us/step - loss: 0.6655 - accuracy: 0.6187 - val_loss: 0.6355 - val_accuracy: 0.6929\n",
            "Epoch 286/500\n",
            "257/257 [==============================] - 0s 695us/step - loss: 0.6655 - accuracy: 0.6187 - val_loss: 0.6353 - val_accuracy: 0.6929\n",
            "Epoch 287/500\n",
            "257/257 [==============================] - 0s 672us/step - loss: 0.6654 - accuracy: 0.6187 - val_loss: 0.6350 - val_accuracy: 0.6929\n",
            "Epoch 288/500\n",
            "257/257 [==============================] - 0s 744us/step - loss: 0.6654 - accuracy: 0.6187 - val_loss: 0.6347 - val_accuracy: 0.6929\n",
            "Epoch 289/500\n",
            "257/257 [==============================] - 0s 739us/step - loss: 0.6653 - accuracy: 0.6187 - val_loss: 0.6348 - val_accuracy: 0.6929\n",
            "Epoch 290/500\n",
            "257/257 [==============================] - 0s 729us/step - loss: 0.6653 - accuracy: 0.6187 - val_loss: 0.6346 - val_accuracy: 0.6929\n",
            "Epoch 291/500\n",
            "257/257 [==============================] - 0s 719us/step - loss: 0.6653 - accuracy: 0.6187 - val_loss: 0.6347 - val_accuracy: 0.6929\n",
            "Epoch 292/500\n",
            "257/257 [==============================] - 0s 685us/step - loss: 0.6653 - accuracy: 0.6187 - val_loss: 0.6346 - val_accuracy: 0.6929\n",
            "Epoch 293/500\n",
            "257/257 [==============================] - 0s 669us/step - loss: 0.6653 - accuracy: 0.6187 - val_loss: 0.6343 - val_accuracy: 0.6929\n",
            "Epoch 294/500\n",
            "257/257 [==============================] - 0s 707us/step - loss: 0.6652 - accuracy: 0.6187 - val_loss: 0.6341 - val_accuracy: 0.6929\n",
            "Epoch 295/500\n",
            "257/257 [==============================] - 0s 663us/step - loss: 0.6652 - accuracy: 0.6187 - val_loss: 0.6337 - val_accuracy: 0.6929\n",
            "Epoch 296/500\n",
            "257/257 [==============================] - 0s 744us/step - loss: 0.6651 - accuracy: 0.6187 - val_loss: 0.6334 - val_accuracy: 0.6929\n",
            "Epoch 297/500\n",
            "257/257 [==============================] - 0s 727us/step - loss: 0.6651 - accuracy: 0.6187 - val_loss: 0.6331 - val_accuracy: 0.6929\n",
            "Epoch 298/500\n",
            "257/257 [==============================] - 0s 715us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6328 - val_accuracy: 0.6929\n",
            "Epoch 299/500\n",
            "257/257 [==============================] - 0s 682us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6324 - val_accuracy: 0.6929\n",
            "Epoch 300/500\n",
            "257/257 [==============================] - 0s 710us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6325 - val_accuracy: 0.6929\n",
            "Epoch 301/500\n",
            "257/257 [==============================] - 0s 738us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6323 - val_accuracy: 0.6929\n",
            "Epoch 302/500\n",
            "257/257 [==============================] - 0s 754us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6325 - val_accuracy: 0.6929\n",
            "Epoch 303/500\n",
            "257/257 [==============================] - 0s 698us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6323 - val_accuracy: 0.6929\n",
            "Epoch 304/500\n",
            "257/257 [==============================] - 0s 675us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6321 - val_accuracy: 0.6929\n",
            "Epoch 305/500\n",
            "257/257 [==============================] - 0s 655us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6323 - val_accuracy: 0.6929\n",
            "Epoch 306/500\n",
            "257/257 [==============================] - 0s 650us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6325 - val_accuracy: 0.6929\n",
            "Epoch 307/500\n",
            "257/257 [==============================] - 0s 677us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6328 - val_accuracy: 0.6929\n",
            "Epoch 308/500\n",
            "257/257 [==============================] - 0s 682us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6332 - val_accuracy: 0.6929\n",
            "Epoch 309/500\n",
            "257/257 [==============================] - 0s 644us/step - loss: 0.6651 - accuracy: 0.6187 - val_loss: 0.6332 - val_accuracy: 0.6929\n",
            "Epoch 310/500\n",
            "257/257 [==============================] - 0s 708us/step - loss: 0.6651 - accuracy: 0.6187 - val_loss: 0.6330 - val_accuracy: 0.6929\n",
            "Epoch 311/500\n",
            "257/257 [==============================] - 0s 666us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6332 - val_accuracy: 0.6929\n",
            "Epoch 312/500\n",
            "257/257 [==============================] - 0s 662us/step - loss: 0.6651 - accuracy: 0.6187 - val_loss: 0.6331 - val_accuracy: 0.6929\n",
            "Epoch 313/500\n",
            "257/257 [==============================] - 0s 725us/step - loss: 0.6651 - accuracy: 0.6187 - val_loss: 0.6333 - val_accuracy: 0.6929\n",
            "Epoch 314/500\n",
            "257/257 [==============================] - 0s 672us/step - loss: 0.6651 - accuracy: 0.6187 - val_loss: 0.6332 - val_accuracy: 0.6929\n",
            "Epoch 315/500\n",
            "257/257 [==============================] - 0s 675us/step - loss: 0.6651 - accuracy: 0.6187 - val_loss: 0.6330 - val_accuracy: 0.6929\n",
            "Epoch 316/500\n",
            "257/257 [==============================] - 0s 698us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6327 - val_accuracy: 0.6929\n",
            "Epoch 317/500\n",
            "257/257 [==============================] - 0s 669us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6324 - val_accuracy: 0.6929\n",
            "Epoch 318/500\n",
            "257/257 [==============================] - 0s 692us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6321 - val_accuracy: 0.6929\n",
            "Epoch 319/500\n",
            "257/257 [==============================] - 0s 702us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6318 - val_accuracy: 0.6929\n",
            "Epoch 320/500\n",
            "257/257 [==============================] - 0s 693us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6319 - val_accuracy: 0.6929\n",
            "Epoch 321/500\n",
            "257/257 [==============================] - 0s 777us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6321 - val_accuracy: 0.6929\n",
            "Epoch 322/500\n",
            "257/257 [==============================] - 0s 683us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6324 - val_accuracy: 0.6929\n",
            "Epoch 323/500\n",
            "257/257 [==============================] - 0s 683us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6324 - val_accuracy: 0.6929\n",
            "Epoch 324/500\n",
            "257/257 [==============================] - 0s 674us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6322 - val_accuracy: 0.6929\n",
            "Epoch 325/500\n",
            "257/257 [==============================] - 0s 759us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6324 - val_accuracy: 0.6929\n",
            "Epoch 326/500\n",
            "257/257 [==============================] - 0s 709us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6326 - val_accuracy: 0.6929\n",
            "Epoch 327/500\n",
            "257/257 [==============================] - 0s 660us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6326 - val_accuracy: 0.6929\n",
            "Epoch 328/500\n",
            "257/257 [==============================] - 0s 692us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6328 - val_accuracy: 0.6929\n",
            "Epoch 329/500\n",
            "257/257 [==============================] - 0s 675us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6327 - val_accuracy: 0.6929\n",
            "Epoch 330/500\n",
            "257/257 [==============================] - 0s 689us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6325 - val_accuracy: 0.6929\n",
            "Epoch 331/500\n",
            "257/257 [==============================] - 0s 684us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6323 - val_accuracy: 0.6929\n",
            "Epoch 332/500\n",
            "257/257 [==============================] - 0s 732us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6320 - val_accuracy: 0.6929\n",
            "Epoch 333/500\n",
            "257/257 [==============================] - 0s 681us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6317 - val_accuracy: 0.6929\n",
            "Epoch 334/500\n",
            "257/257 [==============================] - 0s 674us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6315 - val_accuracy: 0.6929\n",
            "Epoch 335/500\n",
            "257/257 [==============================] - 0s 693us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6312 - val_accuracy: 0.6929\n",
            "Epoch 336/500\n",
            "257/257 [==============================] - 0s 697us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6309 - val_accuracy: 0.6929\n",
            "Epoch 337/500\n",
            "257/257 [==============================] - 0s 682us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6309 - val_accuracy: 0.6929\n",
            "Epoch 338/500\n",
            "257/257 [==============================] - 0s 666us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6311 - val_accuracy: 0.6929\n",
            "Epoch 339/500\n",
            "257/257 [==============================] - 0s 718us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6311 - val_accuracy: 0.6929\n",
            "Epoch 340/500\n",
            "257/257 [==============================] - 0s 699us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6313 - val_accuracy: 0.6929\n",
            "Epoch 341/500\n",
            "257/257 [==============================] - 0s 684us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6312 - val_accuracy: 0.6929\n",
            "Epoch 342/500\n",
            "257/257 [==============================] - 0s 704us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6314 - val_accuracy: 0.6929\n",
            "Epoch 343/500\n",
            "257/257 [==============================] - 0s 688us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6317 - val_accuracy: 0.6929\n",
            "Epoch 344/500\n",
            "257/257 [==============================] - 0s 800us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6320 - val_accuracy: 0.6929\n",
            "Epoch 345/500\n",
            "257/257 [==============================] - 0s 745us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6324 - val_accuracy: 0.6929\n",
            "Epoch 346/500\n",
            "257/257 [==============================] - 0s 742us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6324 - val_accuracy: 0.6929\n",
            "Epoch 347/500\n",
            "257/257 [==============================] - 0s 773us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6326 - val_accuracy: 0.6929\n",
            "Epoch 348/500\n",
            "257/257 [==============================] - 0s 694us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6326 - val_accuracy: 0.6929\n",
            "Epoch 349/500\n",
            "257/257 [==============================] - 0s 672us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6324 - val_accuracy: 0.6929\n",
            "Epoch 350/500\n",
            "257/257 [==============================] - 0s 722us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6322 - val_accuracy: 0.6929\n",
            "Epoch 351/500\n",
            "257/257 [==============================] - 0s 672us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6323 - val_accuracy: 0.6929\n",
            "Epoch 352/500\n",
            "257/257 [==============================] - 0s 680us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6322 - val_accuracy: 0.6929\n",
            "Epoch 353/500\n",
            "257/257 [==============================] - 0s 701us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6323 - val_accuracy: 0.6929\n",
            "Epoch 354/500\n",
            "257/257 [==============================] - 0s 696us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6326 - val_accuracy: 0.6929\n",
            "Epoch 355/500\n",
            "257/257 [==============================] - 0s 683us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6325 - val_accuracy: 0.6929\n",
            "Epoch 356/500\n",
            "257/257 [==============================] - 0s 700us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6324 - val_accuracy: 0.6929\n",
            "Epoch 357/500\n",
            "257/257 [==============================] - 0s 702us/step - loss: 0.6650 - accuracy: 0.6187 - val_loss: 0.6322 - val_accuracy: 0.6929\n",
            "Epoch 358/500\n",
            "257/257 [==============================] - 0s 710us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6319 - val_accuracy: 0.6929\n",
            "Epoch 359/500\n",
            "257/257 [==============================] - 0s 695us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6320 - val_accuracy: 0.6929\n",
            "Epoch 360/500\n",
            "257/257 [==============================] - 0s 666us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6318 - val_accuracy: 0.6929\n",
            "Epoch 361/500\n",
            "257/257 [==============================] - 0s 748us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6320 - val_accuracy: 0.6929\n",
            "Epoch 362/500\n",
            "257/257 [==============================] - 0s 681us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6319 - val_accuracy: 0.6929\n",
            "Epoch 363/500\n",
            "257/257 [==============================] - 0s 676us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6317 - val_accuracy: 0.6929\n",
            "Epoch 364/500\n",
            "257/257 [==============================] - 0s 709us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6315 - val_accuracy: 0.6929\n",
            "Epoch 365/500\n",
            "257/257 [==============================] - 0s 681us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6312 - val_accuracy: 0.6929\n",
            "Epoch 366/500\n",
            "257/257 [==============================] - 0s 674us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6309 - val_accuracy: 0.6929\n",
            "Epoch 367/500\n",
            "257/257 [==============================] - 0s 702us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6307 - val_accuracy: 0.6929\n",
            "Epoch 368/500\n",
            "257/257 [==============================] - 0s 687us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6307 - val_accuracy: 0.6929\n",
            "Epoch 369/500\n",
            "257/257 [==============================] - 0s 705us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6306 - val_accuracy: 0.6929\n",
            "Epoch 370/500\n",
            "257/257 [==============================] - 0s 690us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6304 - val_accuracy: 0.6929\n",
            "Epoch 371/500\n",
            "257/257 [==============================] - 0s 675us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6305 - val_accuracy: 0.6929\n",
            "Epoch 372/500\n",
            "257/257 [==============================] - 0s 720us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6304 - val_accuracy: 0.6929\n",
            "Epoch 373/500\n",
            "257/257 [==============================] - 0s 716us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6305 - val_accuracy: 0.6929\n",
            "Epoch 374/500\n",
            "257/257 [==============================] - 0s 683us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6308 - val_accuracy: 0.6929\n",
            "Epoch 375/500\n",
            "257/257 [==============================] - 0s 700us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6308 - val_accuracy: 0.6929\n",
            "Epoch 376/500\n",
            "257/257 [==============================] - 0s 659us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6307 - val_accuracy: 0.6929\n",
            "Epoch 377/500\n",
            "257/257 [==============================] - 0s 801us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6305 - val_accuracy: 0.6929\n",
            "Epoch 378/500\n",
            "257/257 [==============================] - 0s 684us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6302 - val_accuracy: 0.6929\n",
            "Epoch 379/500\n",
            "257/257 [==============================] - 0s 675us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6300 - val_accuracy: 0.6929\n",
            "Epoch 380/500\n",
            "257/257 [==============================] - 0s 704us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6301 - val_accuracy: 0.6929\n",
            "Epoch 381/500\n",
            "257/257 [==============================] - 0s 766us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6300 - val_accuracy: 0.6929\n",
            "Epoch 382/500\n",
            "257/257 [==============================] - 0s 664us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6298 - val_accuracy: 0.6929\n",
            "Epoch 383/500\n",
            "257/257 [==============================] - 0s 663us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6299 - val_accuracy: 0.6929\n",
            "Epoch 384/500\n",
            "257/257 [==============================] - 0s 652us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6298 - val_accuracy: 0.6929\n",
            "Epoch 385/500\n",
            "257/257 [==============================] - 0s 664us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6296 - val_accuracy: 0.6929\n",
            "Epoch 386/500\n",
            "257/257 [==============================] - 0s 679us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6294 - val_accuracy: 0.6929\n",
            "Epoch 387/500\n",
            "257/257 [==============================] - 0s 664us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6292 - val_accuracy: 0.6929\n",
            "Epoch 388/500\n",
            "257/257 [==============================] - 0s 701us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6290 - val_accuracy: 0.6929\n",
            "Epoch 389/500\n",
            "257/257 [==============================] - 0s 722us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6287 - val_accuracy: 0.6929\n",
            "Epoch 390/500\n",
            "257/257 [==============================] - 0s 734us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6285 - val_accuracy: 0.6929\n",
            "Epoch 391/500\n",
            "257/257 [==============================] - 0s 674us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6285 - val_accuracy: 0.6929\n",
            "Epoch 392/500\n",
            "257/257 [==============================] - 0s 744us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6288 - val_accuracy: 0.6929\n",
            "Epoch 393/500\n",
            "257/257 [==============================] - 0s 716us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6287 - val_accuracy: 0.6929\n",
            "Epoch 394/500\n",
            "257/257 [==============================] - 0s 686us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6289 - val_accuracy: 0.6929\n",
            "Epoch 395/500\n",
            "257/257 [==============================] - 0s 692us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6292 - val_accuracy: 0.6929\n",
            "Epoch 396/500\n",
            "257/257 [==============================] - 0s 668us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6295 - val_accuracy: 0.6929\n",
            "Epoch 397/500\n",
            "257/257 [==============================] - 0s 689us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6299 - val_accuracy: 0.6929\n",
            "Epoch 398/500\n",
            "257/257 [==============================] - 0s 691us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6300 - val_accuracy: 0.6929\n",
            "Epoch 399/500\n",
            "257/257 [==============================] - 0s 709us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6299 - val_accuracy: 0.6929\n",
            "Epoch 400/500\n",
            "257/257 [==============================] - 0s 750us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6301 - val_accuracy: 0.6929\n",
            "Epoch 401/500\n",
            "257/257 [==============================] - 0s 760us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6303 - val_accuracy: 0.6929\n",
            "Epoch 402/500\n",
            "257/257 [==============================] - 0s 725us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6307 - val_accuracy: 0.6929\n",
            "Epoch 403/500\n",
            "257/257 [==============================] - 0s 730us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6307 - val_accuracy: 0.6929\n",
            "Epoch 404/500\n",
            "257/257 [==============================] - 0s 676us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6309 - val_accuracy: 0.6929\n",
            "Epoch 405/500\n",
            "257/257 [==============================] - 0s 734us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6309 - val_accuracy: 0.6929\n",
            "Epoch 406/500\n",
            "257/257 [==============================] - 0s 704us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6311 - val_accuracy: 0.6929\n",
            "Epoch 407/500\n",
            "257/257 [==============================] - 0s 672us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6310 - val_accuracy: 0.6929\n",
            "Epoch 408/500\n",
            "257/257 [==============================] - 0s 740us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6312 - val_accuracy: 0.6929\n",
            "Epoch 409/500\n",
            "257/257 [==============================] - 0s 669us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6312 - val_accuracy: 0.6929\n",
            "Epoch 410/500\n",
            "257/257 [==============================] - 0s 768us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6310 - val_accuracy: 0.6929\n",
            "Epoch 411/500\n",
            "257/257 [==============================] - 0s 840us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6311 - val_accuracy: 0.6929\n",
            "Epoch 412/500\n",
            "257/257 [==============================] - 0s 764us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6310 - val_accuracy: 0.6929\n",
            "Epoch 413/500\n",
            "257/257 [==============================] - 0s 721us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6312 - val_accuracy: 0.6929\n",
            "Epoch 414/500\n",
            "257/257 [==============================] - 0s 708us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6311 - val_accuracy: 0.6929\n",
            "Epoch 415/500\n",
            "257/257 [==============================] - 0s 710us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6313 - val_accuracy: 0.6929\n",
            "Epoch 416/500\n",
            "257/257 [==============================] - 0s 677us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6312 - val_accuracy: 0.6929\n",
            "Epoch 417/500\n",
            "257/257 [==============================] - 0s 670us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6311 - val_accuracy: 0.6929\n",
            "Epoch 418/500\n",
            "257/257 [==============================] - 0s 682us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6309 - val_accuracy: 0.6929\n",
            "Epoch 419/500\n",
            "257/257 [==============================] - 0s 717us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6306 - val_accuracy: 0.6929\n",
            "Epoch 420/500\n",
            "257/257 [==============================] - 0s 708us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6304 - val_accuracy: 0.6929\n",
            "Epoch 421/500\n",
            "257/257 [==============================] - 0s 715us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6301 - val_accuracy: 0.6929\n",
            "Epoch 422/500\n",
            "257/257 [==============================] - 0s 675us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6301 - val_accuracy: 0.6929\n",
            "Epoch 423/500\n",
            "257/257 [==============================] - 0s 709us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6303 - val_accuracy: 0.6929\n",
            "Epoch 424/500\n",
            "257/257 [==============================] - 0s 680us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6303 - val_accuracy: 0.6929\n",
            "Epoch 425/500\n",
            "257/257 [==============================] - 0s 720us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6305 - val_accuracy: 0.6929\n",
            "Epoch 426/500\n",
            "257/257 [==============================] - 0s 696us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6304 - val_accuracy: 0.6929\n",
            "Epoch 427/500\n",
            "257/257 [==============================] - 0s 690us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6303 - val_accuracy: 0.6929\n",
            "Epoch 428/500\n",
            "257/257 [==============================] - 0s 669us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6304 - val_accuracy: 0.6929\n",
            "Epoch 429/500\n",
            "257/257 [==============================] - 0s 697us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6307 - val_accuracy: 0.6929\n",
            "Epoch 430/500\n",
            "257/257 [==============================] - 0s 684us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6307 - val_accuracy: 0.6929\n",
            "Epoch 431/500\n",
            "257/257 [==============================] - 0s 737us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6305 - val_accuracy: 0.6929\n",
            "Epoch 432/500\n",
            "257/257 [==============================] - 0s 787us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6303 - val_accuracy: 0.6929\n",
            "Epoch 433/500\n",
            "257/257 [==============================] - 0s 820us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6301 - val_accuracy: 0.6929\n",
            "Epoch 434/500\n",
            "257/257 [==============================] - 0s 743us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6299 - val_accuracy: 0.6929\n",
            "Epoch 435/500\n",
            "257/257 [==============================] - 0s 756us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6299 - val_accuracy: 0.6929\n",
            "Epoch 436/500\n",
            "257/257 [==============================] - 0s 727us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6301 - val_accuracy: 0.6929\n",
            "Epoch 437/500\n",
            "257/257 [==============================] - 0s 711us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6301 - val_accuracy: 0.6929\n",
            "Epoch 438/500\n",
            "257/257 [==============================] - 0s 679us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6303 - val_accuracy: 0.6929\n",
            "Epoch 439/500\n",
            "257/257 [==============================] - 0s 701us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6303 - val_accuracy: 0.6929\n",
            "Epoch 440/500\n",
            "257/257 [==============================] - 0s 681us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6301 - val_accuracy: 0.6929\n",
            "Epoch 441/500\n",
            "257/257 [==============================] - 0s 759us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6299 - val_accuracy: 0.6929\n",
            "Epoch 442/500\n",
            "257/257 [==============================] - 0s 692us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6300 - val_accuracy: 0.6929\n",
            "Epoch 443/500\n",
            "257/257 [==============================] - 0s 703us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6302 - val_accuracy: 0.6929\n",
            "Epoch 444/500\n",
            "257/257 [==============================] - 0s 707us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6302 - val_accuracy: 0.6929\n",
            "Epoch 445/500\n",
            "257/257 [==============================] - 0s 696us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6301 - val_accuracy: 0.6929\n",
            "Epoch 446/500\n",
            "257/257 [==============================] - 0s 751us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6302 - val_accuracy: 0.6929\n",
            "Epoch 447/500\n",
            "257/257 [==============================] - 0s 725us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6305 - val_accuracy: 0.6929\n",
            "Epoch 448/500\n",
            "257/257 [==============================] - 0s 750us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6308 - val_accuracy: 0.6929\n",
            "Epoch 449/500\n",
            "257/257 [==============================] - 0s 757us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6308 - val_accuracy: 0.6929\n",
            "Epoch 450/500\n",
            "257/257 [==============================] - 0s 781us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6307 - val_accuracy: 0.6929\n",
            "Epoch 451/500\n",
            "257/257 [==============================] - 0s 716us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6305 - val_accuracy: 0.6929\n",
            "Epoch 452/500\n",
            "257/257 [==============================] - 0s 752us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6303 - val_accuracy: 0.6929\n",
            "Epoch 453/500\n",
            "257/257 [==============================] - 0s 776us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6304 - val_accuracy: 0.6929\n",
            "Epoch 454/500\n",
            "257/257 [==============================] - 0s 762us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6303 - val_accuracy: 0.6929\n",
            "Epoch 455/500\n",
            "257/257 [==============================] - 0s 763us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6304 - val_accuracy: 0.6929\n",
            "Epoch 456/500\n",
            "257/257 [==============================] - 0s 775us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6307 - val_accuracy: 0.6929\n",
            "Epoch 457/500\n",
            "257/257 [==============================] - 0s 777us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6307 - val_accuracy: 0.6929\n",
            "Epoch 458/500\n",
            "257/257 [==============================] - 0s 747us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6306 - val_accuracy: 0.6929\n",
            "Epoch 459/500\n",
            "257/257 [==============================] - 0s 784us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6304 - val_accuracy: 0.6929\n",
            "Epoch 460/500\n",
            "257/257 [==============================] - 0s 760us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6301 - val_accuracy: 0.6929\n",
            "Epoch 461/500\n",
            "257/257 [==============================] - 0s 751us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6299 - val_accuracy: 0.6929\n",
            "Epoch 462/500\n",
            "257/257 [==============================] - 0s 713us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6297 - val_accuracy: 0.6929\n",
            "Epoch 463/500\n",
            "257/257 [==============================] - 0s 689us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6294 - val_accuracy: 0.6929\n",
            "Epoch 464/500\n",
            "257/257 [==============================] - 0s 711us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6291 - val_accuracy: 0.6929\n",
            "Epoch 465/500\n",
            "257/257 [==============================] - 0s 692us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6289 - val_accuracy: 0.6929\n",
            "Epoch 466/500\n",
            "257/257 [==============================] - 0s 715us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6286 - val_accuracy: 0.6929\n",
            "Epoch 467/500\n",
            "257/257 [==============================] - 0s 700us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6287 - val_accuracy: 0.6929\n",
            "Epoch 468/500\n",
            "257/257 [==============================] - 0s 723us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6289 - val_accuracy: 0.6929\n",
            "Epoch 469/500\n",
            "257/257 [==============================] - 0s 707us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6292 - val_accuracy: 0.6929\n",
            "Epoch 470/500\n",
            "257/257 [==============================] - 0s 790us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6292 - val_accuracy: 0.6929\n",
            "Epoch 471/500\n",
            "257/257 [==============================] - 0s 691us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6294 - val_accuracy: 0.6929\n",
            "Epoch 472/500\n",
            "257/257 [==============================] - 0s 700us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6294 - val_accuracy: 0.6929\n",
            "Epoch 473/500\n",
            "257/257 [==============================] - 0s 771us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6293 - val_accuracy: 0.6929\n",
            "Epoch 474/500\n",
            "257/257 [==============================] - 0s 768us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6294 - val_accuracy: 0.6929\n",
            "Epoch 475/500\n",
            "257/257 [==============================] - 0s 739us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6294 - val_accuracy: 0.6929\n",
            "Epoch 476/500\n",
            "257/257 [==============================] - 0s 810us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6295 - val_accuracy: 0.6929\n",
            "Epoch 477/500\n",
            "257/257 [==============================] - 0s 749us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6295 - val_accuracy: 0.6929\n",
            "Epoch 478/500\n",
            "257/257 [==============================] - 0s 745us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6293 - val_accuracy: 0.6929\n",
            "Epoch 479/500\n",
            "257/257 [==============================] - 0s 714us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6292 - val_accuracy: 0.6929\n",
            "Epoch 480/500\n",
            "257/257 [==============================] - 0s 696us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6293 - val_accuracy: 0.6929\n",
            "Epoch 481/500\n",
            "257/257 [==============================] - 0s 687us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6292 - val_accuracy: 0.6929\n",
            "Epoch 482/500\n",
            "257/257 [==============================] - 0s 733us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6290 - val_accuracy: 0.6929\n",
            "Epoch 483/500\n",
            "257/257 [==============================] - 0s 702us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6288 - val_accuracy: 0.6929\n",
            "Epoch 484/500\n",
            "257/257 [==============================] - 0s 781us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6289 - val_accuracy: 0.6929\n",
            "Epoch 485/500\n",
            "257/257 [==============================] - 0s 754us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6291 - val_accuracy: 0.6929\n",
            "Epoch 486/500\n",
            "257/257 [==============================] - 0s 846us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6294 - val_accuracy: 0.6929\n",
            "Epoch 487/500\n",
            "257/257 [==============================] - 0s 814us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6298 - val_accuracy: 0.6929\n",
            "Epoch 488/500\n",
            "257/257 [==============================] - 0s 696us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6301 - val_accuracy: 0.6929\n",
            "Epoch 489/500\n",
            "257/257 [==============================] - 0s 728us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6305 - val_accuracy: 0.6929\n",
            "Epoch 490/500\n",
            "257/257 [==============================] - 0s 693us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6306 - val_accuracy: 0.6929\n",
            "Epoch 491/500\n",
            "257/257 [==============================] - 0s 693us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6305 - val_accuracy: 0.6929\n",
            "Epoch 492/500\n",
            "257/257 [==============================] - 0s 705us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6304 - val_accuracy: 0.6929\n",
            "Epoch 493/500\n",
            "257/257 [==============================] - 0s 681us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6302 - val_accuracy: 0.6929\n",
            "Epoch 494/500\n",
            "257/257 [==============================] - 0s 696us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6299 - val_accuracy: 0.6929\n",
            "Epoch 495/500\n",
            "257/257 [==============================] - 0s 691us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6297 - val_accuracy: 0.6929\n",
            "Epoch 496/500\n",
            "257/257 [==============================] - 0s 745us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6297 - val_accuracy: 0.6929\n",
            "Epoch 497/500\n",
            "257/257 [==============================] - 0s 705us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6296 - val_accuracy: 0.6929\n",
            "Epoch 498/500\n",
            "257/257 [==============================] - 0s 699us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6298 - val_accuracy: 0.6929\n",
            "Epoch 499/500\n",
            "257/257 [==============================] - 0s 693us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6300 - val_accuracy: 0.6929\n",
            "Epoch 500/500\n",
            "257/257 [==============================] - 0s 707us/step - loss: 0.6648 - accuracy: 0.6187 - val_loss: 0.6300 - val_accuracy: 0.6929\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sZDxYwOdIb0j",
        "colab_type": "text"
      },
      "source": [
        "## Examine convergence ##"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4DH7Sb2djkWV",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 573
        },
        "outputId": "96209e9f-35a5-4db7-9863-c9227fd968a5"
      },
      "source": [
        "plt.plot(history.history['accuracy'], label = 'train')\n",
        "plt.plot(history.history['val_accuracy'], label='validation')\n",
        "plt.title('model accuracy')\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'validation'], loc = 'upper left')\n",
        "plt.show()\n",
        "\n",
        "plt.plot(history.history['loss'])\n",
        "plt.plot(history.history['val_loss'])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'validation'],loc = 'upper left')\n",
        "plt.show()"
      ],
      "execution_count": 286,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "lOwAsqvzK4JN",
        "colab_type": "text"
      },
      "source": [
        "## Evaluate the derived model (obtained from final epoch) ##"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3p5WO16i3EHE",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 282
        },
        "outputId": "87a810ca-0e55-4dad-e783-209f83cf7a96"
      },
      "source": [
        "y_pred2=model.predict(X_tst0)\n",
        "y_pred2=(y_pred2>.5).astype(int)\n",
        "\n",
        "# todo: use boot-strap estimation\n",
        "probs = model.predict_proba(X_tst0)\n",
        "probs = probs[:,0]\n",
        "auc = roc_auc_score(y_test, probs)\n",
        "print('AUC: %.3f' % auc)\n",
        "fpr, tpr, thresholds = roc_curve(y_test, probs)\n",
        "pyplot.plot([0, 1], [0, 1], linestyle = '-')\n",
        "pyplot.plot(fpr, tpr, marker = '.')\n",
        "pyplot.show()"
      ],
      "execution_count": 287,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "AUC: 0.500\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2LjOotjg3WAZ",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 156
        },
        "outputId": "709db01e-5b1a-4cdf-8f83-e478a95d4b4a"
      },
      "source": [
        "cmat = metrics.confusion_matrix(y_test, y_pred2)\n",
        "print( cmat )\n",
        "print(\"Accuracy:\",metrics.accuracy_score(y_test, y_pred2))\n",
        "print(\"Precision:\",metrics.precision_score(y_test, y_pred2, pos_label=1)) \n",
        "print(\"Recall:\",metrics.recall_score(y_test, y_pred2, pos_label=1))"
      ],
      "execution_count": 288,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[[253   0]\n",
            " [131   0]]\n",
            "Accuracy: 0.6588541666666666\n",
            "Precision: 0.0\n",
            "Recall: 0.0\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/sklearn/metrics/_classification.py:1272: UndefinedMetricWarning: Precision is ill-defined and being set to 0.0 due to no predicted samples. Use `zero_division` parameter to control this behavior.\n",
            "  _warn_prf(average, modifier, msg_start, len(result))\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AgzmnBuNKwpD",
        "colab_type": "text"
      },
      "source": [
        "## Evaluate using the best model  ##"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6Ck4i8YY3DCv",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 333
        },
        "outputId": "6862f95c-6fac-47d8-addb-a18db9a74e6d"
      },
      "source": [
        "! ls\n",
        "model.load_weights('best.h5')\n",
        "\n",
        "y_pred2=model.predict(X_tst0)\n",
        "y_pred2=(y_pred2>.5).astype(int)\n",
        "\n",
        "# todo: use boot-strap estimation\n",
        "probs = model.predict_proba(X_tst0)\n",
        "probs = probs[:,0]\n",
        "auc = roc_auc_score(y_test, probs)\n",
        "print('AUC: %.3f' % auc)\n",
        "fpr, tpr, thresholds = roc_curve(y_test, probs)\n",
        "pyplot.plot([0, 1], [0, 1], linestyle = '-')\n",
        "pyplot.plot(fpr, tpr, marker = '.')\n",
        "pyplot.show()\n"
      ],
      "execution_count": 289,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "accuracy.png   diabetes2.csv  diabetes4.csv  loss.png\n",
            "ann_BGL.ipynb  diabetes2.Z    diabetes.csv   README.md\n",
            "best.h5        diabetes3.csv  glucose_RF.R\n",
            "AUC: 0.787\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yx1l183PIGy5",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 102
        },
        "outputId": "cfc1462a-d18c-4ed6-f94d-56d8f00b971a"
      },
      "source": [
        "cmat = metrics.confusion_matrix(y_test, y_pred2)\n",
        "print( cmat )\n",
        "print(\"Accuracy:\",metrics.accuracy_score(y_test, y_pred2))\n",
        "print(\"Precision:\",metrics.precision_score(y_test, y_pred2, pos_label=1)) \n",
        "print(\"Recall:\",metrics.recall_score(y_test, y_pred2, pos_label=1))\n"
      ],
      "execution_count": 290,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[[239  14]\n",
            " [ 98  33]]\n",
            "Accuracy: 0.7083333333333334\n",
            "Precision: 0.7021276595744681\n",
            "Recall: 0.25190839694656486\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "4WlgjQ2WL5Z8",
        "colab_type": "text"
      },
      "source": [
        "## Train LSTM with standardized input data ##"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DrPCFs7XL3UP",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "8c57056b-4b97-48a1-f35c-0edae0e765d4"
      },
      "source": [
        "history = model.fit(X_trn3, y_train, validation_split = 0.33, initial_epoch=0, epochs = 500, batch_size = 64, verbose = 1,callbacks=C)\n"
      ],
      "execution_count": 303,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Train on 257 samples, validate on 127 samples\n",
            "Epoch 1/500\n",
            "257/257 [==============================] - 0s 718us/step - loss: 0.7589 - accuracy: 0.6187 - val_loss: 0.6306 - val_accuracy: 0.6929\n",
            "Epoch 2/500\n",
            "257/257 [==============================] - 0s 683us/step - loss: 0.6652 - accuracy: 0.6187 - val_loss: 0.6446 - val_accuracy: 0.6929\n",
            "Epoch 3/500\n",
            "257/257 [==============================] - 0s 723us/step - loss: 0.6685 - accuracy: 0.6187 - val_loss: 0.6437 - val_accuracy: 0.6929\n",
            "Epoch 4/500\n",
            "257/257 [==============================] - 0s 721us/step - loss: 0.6677 - accuracy: 0.6187 - val_loss: 0.6382 - val_accuracy: 0.6929\n",
            "Epoch 5/500\n",
            "257/257 [==============================] - 0s 681us/step - loss: 0.6662 - accuracy: 0.6187 - val_loss: 0.6305 - val_accuracy: 0.6929\n",
            "Epoch 6/500\n",
            "257/257 [==============================] - 0s 681us/step - loss: 0.6649 - accuracy: 0.6187 - val_loss: 0.6238 - val_accuracy: 0.6929\n",
            "Epoch 7/500\n",
            "257/257 [==============================] - 0s 706us/step - loss: 0.6659 - accuracy: 0.6187 - val_loss: 0.6301 - val_accuracy: 0.6929\n",
            "Epoch 8/500\n",
            "257/257 [==============================] - 0s 735us/step - loss: 0.6658 - accuracy: 0.6187 - val_loss: 0.6287 - val_accuracy: 0.6929\n",
            "Epoch 9/500\n",
            "257/257 [==============================] - 0s 703us/step - loss: 0.6647 - accuracy: 0.6187 - val_loss: 0.6242 - val_accuracy: 0.6929\n",
            "Epoch 10/500\n",
            "257/257 [==============================] - 0s 749us/step - loss: 0.6654 - accuracy: 0.6187 - val_loss: 0.6305 - val_accuracy: 0.6929\n",
            "Epoch 11/500\n",
            "257/257 [==============================] - 0s 686us/step - loss: 0.6651 - accuracy: 0.6187 - val_loss: 0.6286 - val_accuracy: 0.6929\n",
            "Epoch 12/500\n",
            "257/257 [==============================] - 0s 696us/step - loss: 0.6643 - accuracy: 0.6187 - val_loss: 0.6152 - val_accuracy: 0.6929\n",
            "Epoch 13/500\n",
            "257/257 [==============================] - 0s 692us/step - loss: 0.6489 - accuracy: 0.6187 - val_loss: 0.6247 - val_accuracy: 0.6929\n",
            "Epoch 14/500\n",
            "257/257 [==============================] - 0s 707us/step - loss: 0.6389 - accuracy: 0.6187 - val_loss: 0.5843 - val_accuracy: 0.6929\n",
            "Epoch 15/500\n",
            "257/257 [==============================] - 0s 775us/step - loss: 0.6217 - accuracy: 0.6187 - val_loss: 1.0626 - val_accuracy: 0.6929\n",
            "Epoch 16/500\n",
            "257/257 [==============================] - 0s 748us/step - loss: 0.8052 - accuracy: 0.6187 - val_loss: 0.6143 - val_accuracy: 0.6929\n",
            "Epoch 17/500\n",
            "257/257 [==============================] - 0s 679us/step - loss: 0.6301 - accuracy: 0.6187 - val_loss: 0.5841 - val_accuracy: 0.6929\n",
            "Epoch 18/500\n",
            "257/257 [==============================] - 0s 699us/step - loss: 0.6148 - accuracy: 0.6187 - val_loss: 0.6135 - val_accuracy: 0.6929\n",
            "Epoch 19/500\n",
            "257/257 [==============================] - 0s 691us/step - loss: 0.6228 - accuracy: 0.6187 - val_loss: 0.5687 - val_accuracy: 0.6929\n",
            "Epoch 20/500\n",
            "257/257 [==============================] - 0s 772us/step - loss: 0.5990 - accuracy: 0.6187 - val_loss: 0.5649 - val_accuracy: 0.6929\n",
            "Epoch 21/500\n",
            "257/257 [==============================] - 0s 702us/step - loss: 0.5984 - accuracy: 0.6187 - val_loss: 0.5650 - val_accuracy: 0.6929\n",
            "Epoch 22/500\n",
            "257/257 [==============================] - 0s 684us/step - loss: 0.5900 - accuracy: 0.6187 - val_loss: 0.5887 - val_accuracy: 0.6614\n",
            "Epoch 23/500\n",
            "257/257 [==============================] - 0s 704us/step - loss: 0.5950 - accuracy: 0.6654 - val_loss: 0.5636 - val_accuracy: 0.7244\n",
            "Epoch 24/500\n",
            "257/257 [==============================] - 0s 653us/step - loss: 0.5875 - accuracy: 0.7004 - val_loss: 0.5742 - val_accuracy: 0.7244\n",
            "Epoch 25/500\n",
            "257/257 [==============================] - 0s 747us/step - loss: 0.5918 - accuracy: 0.6809 - val_loss: 0.5576 - val_accuracy: 0.7244\n",
            "Epoch 26/500\n",
            "257/257 [==============================] - 0s 683us/step - loss: 0.5851 - accuracy: 0.6887 - val_loss: 0.5576 - val_accuracy: 0.6929\n",
            "Epoch 27/500\n",
            "257/257 [==============================] - 0s 695us/step - loss: 0.6006 - accuracy: 0.6187 - val_loss: 0.5502 - val_accuracy: 0.6929\n",
            "Epoch 28/500\n",
            "257/257 [==============================] - 0s 721us/step - loss: 0.5904 - accuracy: 0.6187 - val_loss: 0.5418 - val_accuracy: 0.6929\n",
            "Epoch 29/500\n",
            "257/257 [==============================] - 0s 690us/step - loss: 0.5886 - accuracy: 0.6226 - val_loss: 0.5593 - val_accuracy: 0.7559\n",
            "Epoch 30/500\n",
            "257/257 [==============================] - 0s 689us/step - loss: 0.5981 - accuracy: 0.6770 - val_loss: 0.5608 - val_accuracy: 0.7480\n",
            "Epoch 31/500\n",
            "257/257 [==============================] - 0s 759us/step - loss: 0.5851 - accuracy: 0.7004 - val_loss: 0.5491 - val_accuracy: 0.7402\n",
            "Epoch 32/500\n",
            "257/257 [==============================] - 0s 697us/step - loss: 0.5770 - accuracy: 0.7004 - val_loss: 0.5469 - val_accuracy: 0.7402\n",
            "Epoch 33/500\n",
            "257/257 [==============================] - 0s 678us/step - loss: 0.5857 - accuracy: 0.6926 - val_loss: 0.6905 - val_accuracy: 0.5276\n",
            "Epoch 34/500\n",
            "257/257 [==============================] - 0s 750us/step - loss: 0.6870 - accuracy: 0.5331 - val_loss: 0.6884 - val_accuracy: 0.4724\n",
            "Epoch 35/500\n",
            "257/257 [==============================] - 0s 710us/step - loss: 0.6859 - accuracy: 0.6187 - val_loss: 0.6865 - val_accuracy: 0.5748\n",
            "Epoch 36/500\n",
            "257/257 [==============================] - 0s 704us/step - loss: 0.6847 - accuracy: 0.5992 - val_loss: 0.6821 - val_accuracy: 0.6614\n",
            "Epoch 37/500\n",
            "257/257 [==============================] - 0s 679us/step - loss: 0.6738 - accuracy: 0.6265 - val_loss: 0.6718 - val_accuracy: 0.6693\n",
            "Epoch 38/500\n",
            "257/257 [==============================] - 0s 689us/step - loss: 0.6654 - accuracy: 0.6381 - val_loss: 0.6016 - val_accuracy: 0.6929\n",
            "Epoch 39/500\n",
            "257/257 [==============================] - 0s 685us/step - loss: 0.6058 - accuracy: 0.6187 - val_loss: 0.7450 - val_accuracy: 0.6929\n",
            "Epoch 40/500\n",
            "257/257 [==============================] - 0s 696us/step - loss: 0.6468 - accuracy: 0.6187 - val_loss: 0.5256 - val_accuracy: 0.6929\n",
            "Epoch 41/500\n",
            "257/257 [==============================] - 0s 701us/step - loss: 0.6004 - accuracy: 0.6187 - val_loss: 0.5298 - val_accuracy: 0.6929\n",
            "Epoch 42/500\n",
            "257/257 [==============================] - 0s 703us/step - loss: 0.5943 - accuracy: 0.6187 - val_loss: 0.5257 - val_accuracy: 0.6929\n",
            "Epoch 43/500\n",
            "257/257 [==============================] - 0s 686us/step - loss: 0.6074 - accuracy: 0.6187 - val_loss: 0.5227 - val_accuracy: 0.6929\n",
            "Epoch 44/500\n",
            "257/257 [==============================] - 0s 671us/step - loss: 0.5878 - accuracy: 0.6187 - val_loss: 0.5360 - val_accuracy: 0.6929\n",
            "Epoch 45/500\n",
            "257/257 [==============================] - 0s 739us/step - loss: 0.5805 - accuracy: 0.6187 - val_loss: 0.5209 - val_accuracy: 0.6929\n",
            "Epoch 46/500\n",
            "257/257 [==============================] - 0s 744us/step - loss: 0.5842 - accuracy: 0.6187 - val_loss: 0.5179 - val_accuracy: 0.6929\n",
            "Epoch 47/500\n",
            "257/257 [==============================] - 0s 823us/step - loss: 0.5750 - accuracy: 0.6187 - val_loss: 0.5294 - val_accuracy: 0.6929\n",
            "Epoch 48/500\n",
            "257/257 [==============================] - 0s 782us/step - loss: 0.5777 - accuracy: 0.6187 - val_loss: 0.5197 - val_accuracy: 0.6929\n",
            "Epoch 49/500\n",
            "257/257 [==============================] - 0s 745us/step - loss: 0.5637 - accuracy: 0.6342 - val_loss: 0.5143 - val_accuracy: 0.6929\n",
            "Epoch 50/500\n",
            "257/257 [==============================] - 0s 675us/step - loss: 0.5641 - accuracy: 0.6148 - val_loss: 0.5693 - val_accuracy: 0.6850\n",
            "Epoch 51/500\n",
            "257/257 [==============================] - 0s 737us/step - loss: 0.5847 - accuracy: 0.6654 - val_loss: 0.5322 - val_accuracy: 0.6929\n",
            "Epoch 52/500\n",
            "257/257 [==============================] - 0s 670us/step - loss: 0.5665 - accuracy: 0.6187 - val_loss: 0.5244 - val_accuracy: 0.6929\n",
            "Epoch 53/500\n",
            "257/257 [==============================] - 0s 713us/step - loss: 0.5696 - accuracy: 0.6187 - val_loss: 0.5127 - val_accuracy: 0.6929\n",
            "Epoch 54/500\n",
            "257/257 [==============================] - 0s 718us/step - loss: 0.5584 - accuracy: 0.6187 - val_loss: 0.5438 - val_accuracy: 0.6929\n",
            "Epoch 55/500\n",
            "257/257 [==============================] - 0s 718us/step - loss: 0.5739 - accuracy: 0.6187 - val_loss: 0.5336 - val_accuracy: 0.6929\n",
            "Epoch 56/500\n",
            "257/257 [==============================] - 0s 685us/step - loss: 0.5584 - accuracy: 0.6187 - val_loss: 0.6207 - val_accuracy: 0.6457\n",
            "Epoch 57/500\n",
            "257/257 [==============================] - 0s 694us/step - loss: 0.6292 - accuracy: 0.6537 - val_loss: 0.5302 - val_accuracy: 0.6929\n",
            "Epoch 58/500\n",
            "257/257 [==============================] - 0s 679us/step - loss: 0.5585 - accuracy: 0.6965 - val_loss: 0.5219 - val_accuracy: 0.6929\n",
            "Epoch 59/500\n",
            "257/257 [==============================] - 0s 688us/step - loss: 0.5653 - accuracy: 0.6187 - val_loss: 0.5204 - val_accuracy: 0.6929\n",
            "Epoch 60/500\n",
            "257/257 [==============================] - 0s 721us/step - loss: 0.5626 - accuracy: 0.6187 - val_loss: 0.5186 - val_accuracy: 0.6929\n",
            "Epoch 61/500\n",
            "257/257 [==============================] - 0s 680us/step - loss: 0.5593 - accuracy: 0.6187 - val_loss: 0.5184 - val_accuracy: 0.7323\n",
            "Epoch 62/500\n",
            "257/257 [==============================] - 0s 690us/step - loss: 0.5581 - accuracy: 0.7160 - val_loss: 0.6549 - val_accuracy: 0.5433\n",
            "Epoch 63/500\n",
            "257/257 [==============================] - 0s 704us/step - loss: 0.6645 - accuracy: 0.5409 - val_loss: 0.6731 - val_accuracy: 0.4567\n",
            "Epoch 64/500\n",
            "257/257 [==============================] - 0s 743us/step - loss: 0.6712 - accuracy: 0.5447 - val_loss: 0.6665 - val_accuracy: 0.5748\n",
            "Epoch 65/500\n",
            "257/257 [==============================] - 0s 738us/step - loss: 0.6633 - accuracy: 0.6304 - val_loss: 0.6483 - val_accuracy: 0.7480\n",
            "Epoch 66/500\n",
            "257/257 [==============================] - 0s 764us/step - loss: 0.6452 - accuracy: 0.6342 - val_loss: 0.6089 - val_accuracy: 0.7165\n",
            "Epoch 67/500\n",
            "257/257 [==============================] - 0s 705us/step - loss: 0.6094 - accuracy: 0.6732 - val_loss: 0.5277 - val_accuracy: 0.6929\n",
            "Epoch 68/500\n",
            "257/257 [==============================] - 0s 672us/step - loss: 0.5686 - accuracy: 0.6187 - val_loss: 0.5159 - val_accuracy: 0.6929\n",
            "Epoch 69/500\n",
            "257/257 [==============================] - 0s 683us/step - loss: 0.5629 - accuracy: 0.6187 - val_loss: 0.5875 - val_accuracy: 0.6929\n",
            "Epoch 70/500\n",
            "257/257 [==============================] - 0s 692us/step - loss: 0.6037 - accuracy: 0.6187 - val_loss: 0.5391 - val_accuracy: 0.6929\n",
            "Epoch 71/500\n",
            "257/257 [==============================] - 0s 689us/step - loss: 0.5769 - accuracy: 0.6770 - val_loss: 0.5103 - val_accuracy: 0.6929\n",
            "Epoch 72/500\n",
            "257/257 [==============================] - 0s 709us/step - loss: 0.5539 - accuracy: 0.6187 - val_loss: 0.5555 - val_accuracy: 0.6772\n",
            "Epoch 73/500\n",
            "257/257 [==============================] - 0s 664us/step - loss: 0.5837 - accuracy: 0.6654 - val_loss: 0.5491 - val_accuracy: 0.6929\n",
            "Epoch 74/500\n",
            "257/257 [==============================] - 0s 661us/step - loss: 0.5726 - accuracy: 0.6187 - val_loss: 0.5323 - val_accuracy: 0.6850\n",
            "Epoch 75/500\n",
            "257/257 [==============================] - 0s 689us/step - loss: 0.5591 - accuracy: 0.6887 - val_loss: 0.5201 - val_accuracy: 0.6929\n",
            "Epoch 76/500\n",
            "257/257 [==============================] - 0s 684us/step - loss: 0.5524 - accuracy: 0.6304 - val_loss: 0.5184 - val_accuracy: 0.6929\n",
            "Epoch 77/500\n",
            "257/257 [==============================] - 0s 663us/step - loss: 0.5515 - accuracy: 0.6187 - val_loss: 0.5739 - val_accuracy: 0.6929\n",
            "Epoch 78/500\n",
            "257/257 [==============================] - 0s 718us/step - loss: 0.5925 - accuracy: 0.6187 - val_loss: 0.5798 - val_accuracy: 0.6929\n",
            "Epoch 79/500\n",
            "257/257 [==============================] - 0s 690us/step - loss: 0.5919 - accuracy: 0.6187 - val_loss: 0.5349 - val_accuracy: 0.6929\n",
            "Epoch 80/500\n",
            "257/257 [==============================] - 0s 679us/step - loss: 0.5667 - accuracy: 0.6187 - val_loss: 0.5415 - val_accuracy: 0.6929\n",
            "Epoch 81/500\n",
            "257/257 [==============================] - 0s 716us/step - loss: 0.5742 - accuracy: 0.6187 - val_loss: 0.5281 - val_accuracy: 0.6929\n",
            "Epoch 82/500\n",
            "257/257 [==============================] - 0s 748us/step - loss: 0.5584 - accuracy: 0.6187 - val_loss: 0.5310 - val_accuracy: 0.6929\n",
            "Epoch 83/500\n",
            "257/257 [==============================] - 0s 706us/step - loss: 0.5588 - accuracy: 0.6187 - val_loss: 0.5236 - val_accuracy: 0.6929\n",
            "Epoch 84/500\n",
            "257/257 [==============================] - 0s 683us/step - loss: 0.5555 - accuracy: 0.6187 - val_loss: 0.6196 - val_accuracy: 0.6929\n",
            "Epoch 85/500\n",
            "257/257 [==============================] - 0s 671us/step - loss: 0.6461 - accuracy: 0.6187 - val_loss: 0.6564 - val_accuracy: 0.6929\n",
            "Epoch 86/500\n",
            "257/257 [==============================] - 0s 717us/step - loss: 0.6628 - accuracy: 0.6187 - val_loss: 0.6656 - val_accuracy: 0.6929\n",
            "Epoch 87/500\n",
            "257/257 [==============================] - 0s 694us/step - loss: 0.6672 - accuracy: 0.6187 - val_loss: 0.6681 - val_accuracy: 0.6929\n",
            "Epoch 88/500\n",
            "257/257 [==============================] - 0s 685us/step - loss: 0.6677 - accuracy: 0.6187 - val_loss: 0.6682 - val_accuracy: 0.7480\n",
            "Epoch 89/500\n",
            "257/257 [==============================] - 0s 696us/step - loss: 0.6658 - accuracy: 0.7004 - val_loss: 0.6583 - val_accuracy: 0.6929\n",
            "Epoch 90/500\n",
            "257/257 [==============================] - 0s 690us/step - loss: 0.6576 - accuracy: 0.6187 - val_loss: 0.6445 - val_accuracy: 0.6929\n",
            "Epoch 91/500\n",
            "257/257 [==============================] - 0s 738us/step - loss: 0.6492 - accuracy: 0.6187 - val_loss: 0.6205 - val_accuracy: 0.6929\n",
            "Epoch 92/500\n",
            "257/257 [==============================] - 0s 737us/step - loss: 0.6295 - accuracy: 0.6187 - val_loss: 0.6040 - val_accuracy: 0.6929\n",
            "Epoch 93/500\n",
            "257/257 [==============================] - 0s 701us/step - loss: 0.6118 - accuracy: 0.6187 - val_loss: 0.5381 - val_accuracy: 0.6929\n",
            "Epoch 94/500\n",
            "257/257 [==============================] - 0s 695us/step - loss: 0.5657 - accuracy: 0.6187 - val_loss: 0.5130 - val_accuracy: 0.7323\n",
            "Epoch 95/500\n",
            "257/257 [==============================] - 0s 685us/step - loss: 0.5531 - accuracy: 0.7160 - val_loss: 0.5171 - val_accuracy: 0.7165\n",
            "Epoch 96/500\n",
            "257/257 [==============================] - 0s 667us/step - loss: 0.5459 - accuracy: 0.6965 - val_loss: 0.5128 - val_accuracy: 0.7087\n",
            "Epoch 97/500\n",
            "257/257 [==============================] - 0s 747us/step - loss: 0.5434 - accuracy: 0.7121 - val_loss: 0.5890 - val_accuracy: 0.5591\n",
            "Epoch 98/500\n",
            "257/257 [==============================] - 0s 686us/step - loss: 0.5925 - accuracy: 0.5953 - val_loss: 0.5512 - val_accuracy: 0.5827\n",
            "Epoch 99/500\n",
            "257/257 [==============================] - 0s 714us/step - loss: 0.5657 - accuracy: 0.6187 - val_loss: 0.5640 - val_accuracy: 0.5827\n",
            "Epoch 100/500\n",
            "257/257 [==============================] - 0s 735us/step - loss: 0.5706 - accuracy: 0.6226 - val_loss: 0.5541 - val_accuracy: 0.6142\n",
            "Epoch 101/500\n",
            "257/257 [==============================] - 0s 676us/step - loss: 0.5591 - accuracy: 0.6420 - val_loss: 0.6982 - val_accuracy: 0.7638\n",
            "Epoch 102/500\n",
            "257/257 [==============================] - 0s 694us/step - loss: 0.6225 - accuracy: 0.6848 - val_loss: 0.5179 - val_accuracy: 0.7244\n",
            "Epoch 103/500\n",
            "257/257 [==============================] - 0s 691us/step - loss: 0.5428 - accuracy: 0.7315 - val_loss: 0.5756 - val_accuracy: 0.6693\n",
            "Epoch 104/500\n",
            "257/257 [==============================] - 0s 679us/step - loss: 0.6009 - accuracy: 0.6848 - val_loss: 0.5890 - val_accuracy: 0.7008\n",
            "Epoch 105/500\n",
            "257/257 [==============================] - 0s 707us/step - loss: 0.6030 - accuracy: 0.7121 - val_loss: 0.5689 - val_accuracy: 0.6929\n",
            "Epoch 106/500\n",
            "257/257 [==============================] - 0s 756us/step - loss: 0.5836 - accuracy: 0.6265 - val_loss: 0.5227 - val_accuracy: 0.6929\n",
            "Epoch 107/500\n",
            "257/257 [==============================] - 0s 810us/step - loss: 0.5493 - accuracy: 0.6265 - val_loss: 0.5534 - val_accuracy: 0.6929\n",
            "Epoch 108/500\n",
            "257/257 [==============================] - 0s 763us/step - loss: 0.5894 - accuracy: 0.6265 - val_loss: 0.5569 - val_accuracy: 0.6929\n",
            "Epoch 109/500\n",
            "257/257 [==============================] - 0s 691us/step - loss: 0.5777 - accuracy: 0.6265 - val_loss: 0.5393 - val_accuracy: 0.6850\n",
            "Epoch 110/500\n",
            "257/257 [==============================] - 0s 735us/step - loss: 0.5573 - accuracy: 0.6420 - val_loss: 0.5157 - val_accuracy: 0.7323\n",
            "Epoch 111/500\n",
            "257/257 [==============================] - 0s 762us/step - loss: 0.5346 - accuracy: 0.7160 - val_loss: 0.5284 - val_accuracy: 0.7402\n",
            "Epoch 112/500\n",
            "257/257 [==============================] - 0s 761us/step - loss: 0.5383 - accuracy: 0.7237 - val_loss: 0.5207 - val_accuracy: 0.7087\n",
            "Epoch 113/500\n",
            "257/257 [==============================] - 0s 743us/step - loss: 0.5304 - accuracy: 0.7082 - val_loss: 0.5301 - val_accuracy: 0.6850\n",
            "Epoch 114/500\n",
            "257/257 [==============================] - 0s 710us/step - loss: 0.5321 - accuracy: 0.7004 - val_loss: 0.5288 - val_accuracy: 0.7087\n",
            "Epoch 115/500\n",
            "257/257 [==============================] - 0s 675us/step - loss: 0.5284 - accuracy: 0.7160 - val_loss: 0.5196 - val_accuracy: 0.6929\n",
            "Epoch 116/500\n",
            "257/257 [==============================] - 0s 686us/step - loss: 0.5227 - accuracy: 0.7198 - val_loss: 0.5196 - val_accuracy: 0.6929\n",
            "Epoch 117/500\n",
            "257/257 [==============================] - 0s 680us/step - loss: 0.5225 - accuracy: 0.7315 - val_loss: 0.5346 - val_accuracy: 0.7402\n",
            "Epoch 118/500\n",
            "257/257 [==============================] - 0s 705us/step - loss: 0.5308 - accuracy: 0.7198 - val_loss: 0.5419 - val_accuracy: 0.7087\n",
            "Epoch 119/500\n",
            "257/257 [==============================] - 0s 713us/step - loss: 0.5273 - accuracy: 0.7237 - val_loss: 0.5403 - val_accuracy: 0.7008\n",
            "Epoch 120/500\n",
            "257/257 [==============================] - 0s 683us/step - loss: 0.5296 - accuracy: 0.7237 - val_loss: 0.5570 - val_accuracy: 0.6772\n",
            "Epoch 121/500\n",
            "257/257 [==============================] - 0s 695us/step - loss: 0.5322 - accuracy: 0.7043 - val_loss: 0.5277 - val_accuracy: 0.7402\n",
            "Epoch 122/500\n",
            "257/257 [==============================] - 0s 733us/step - loss: 0.5402 - accuracy: 0.7237 - val_loss: 0.5194 - val_accuracy: 0.7165\n",
            "Epoch 123/500\n",
            "257/257 [==============================] - 0s 787us/step - loss: 0.5167 - accuracy: 0.7276 - val_loss: 0.4979 - val_accuracy: 0.7323\n",
            "Epoch 124/500\n",
            "257/257 [==============================] - 0s 747us/step - loss: 0.5278 - accuracy: 0.7198 - val_loss: 0.5019 - val_accuracy: 0.7795\n",
            "Epoch 125/500\n",
            "257/257 [==============================] - 0s 666us/step - loss: 0.5305 - accuracy: 0.7121 - val_loss: 0.5326 - val_accuracy: 0.7087\n",
            "Epoch 126/500\n",
            "257/257 [==============================] - 0s 687us/step - loss: 0.5520 - accuracy: 0.7354 - val_loss: 0.5620 - val_accuracy: 0.6693\n",
            "Epoch 127/500\n",
            "257/257 [==============================] - 0s 691us/step - loss: 0.5733 - accuracy: 0.6732 - val_loss: 0.5345 - val_accuracy: 0.6850\n",
            "Epoch 128/500\n",
            "257/257 [==============================] - 0s 689us/step - loss: 0.5382 - accuracy: 0.7004 - val_loss: 0.5195 - val_accuracy: 0.7165\n",
            "Epoch 129/500\n",
            "257/257 [==============================] - 0s 717us/step - loss: 0.5267 - accuracy: 0.7276 - val_loss: 0.5302 - val_accuracy: 0.7087\n",
            "Epoch 130/500\n",
            "257/257 [==============================] - 0s 675us/step - loss: 0.5342 - accuracy: 0.7198 - val_loss: 0.5419 - val_accuracy: 0.6299\n",
            "Epoch 131/500\n",
            "257/257 [==============================] - 0s 788us/step - loss: 0.5446 - accuracy: 0.6537 - val_loss: 0.5354 - val_accuracy: 0.6614\n",
            "Epoch 132/500\n",
            "257/257 [==============================] - 0s 675us/step - loss: 0.5302 - accuracy: 0.6887 - val_loss: 0.5380 - val_accuracy: 0.6850\n",
            "Epoch 133/500\n",
            "257/257 [==============================] - 0s 677us/step - loss: 0.5254 - accuracy: 0.6926 - val_loss: 0.5356 - val_accuracy: 0.6772\n",
            "Epoch 134/500\n",
            "257/257 [==============================] - 0s 695us/step - loss: 0.5199 - accuracy: 0.7004 - val_loss: 0.5493 - val_accuracy: 0.7087\n",
            "Epoch 135/500\n",
            "257/257 [==============================] - 0s 680us/step - loss: 0.5170 - accuracy: 0.7121 - val_loss: 0.5684 - val_accuracy: 0.6378\n",
            "Epoch 136/500\n",
            "257/257 [==============================] - 0s 670us/step - loss: 0.5740 - accuracy: 0.6459 - val_loss: 0.5587 - val_accuracy: 0.6378\n",
            "Epoch 137/500\n",
            "257/257 [==============================] - 0s 667us/step - loss: 0.5473 - accuracy: 0.6848 - val_loss: 0.5362 - val_accuracy: 0.6772\n",
            "Epoch 138/500\n",
            "257/257 [==============================] - 0s 673us/step - loss: 0.5238 - accuracy: 0.7004 - val_loss: 0.5610 - val_accuracy: 0.6772\n",
            "Epoch 139/500\n",
            "257/257 [==============================] - 0s 699us/step - loss: 0.5277 - accuracy: 0.6926 - val_loss: 0.5801 - val_accuracy: 0.6850\n",
            "Epoch 140/500\n",
            "257/257 [==============================] - 0s 747us/step - loss: 0.5300 - accuracy: 0.6965 - val_loss: 0.5308 - val_accuracy: 0.7087\n",
            "Epoch 141/500\n",
            "257/257 [==============================] - 0s 679us/step - loss: 0.5120 - accuracy: 0.7315 - val_loss: 0.5679 - val_accuracy: 0.7874\n",
            "Epoch 142/500\n",
            "257/257 [==============================] - 0s 654us/step - loss: 0.5390 - accuracy: 0.7354 - val_loss: 0.5171 - val_accuracy: 0.7244\n",
            "Epoch 143/500\n",
            "257/257 [==============================] - 0s 680us/step - loss: 0.5163 - accuracy: 0.7160 - val_loss: 0.5181 - val_accuracy: 0.7323\n",
            "Epoch 144/500\n",
            "257/257 [==============================] - 0s 696us/step - loss: 0.5182 - accuracy: 0.7237 - val_loss: 0.5159 - val_accuracy: 0.7638\n",
            "Epoch 145/500\n",
            "257/257 [==============================] - 0s 725us/step - loss: 0.5217 - accuracy: 0.7160 - val_loss: 0.5294 - val_accuracy: 0.6850\n",
            "Epoch 146/500\n",
            "257/257 [==============================] - 0s 715us/step - loss: 0.5318 - accuracy: 0.6965 - val_loss: 0.5781 - val_accuracy: 0.6063\n",
            "Epoch 147/500\n",
            "257/257 [==============================] - 0s 676us/step - loss: 0.5788 - accuracy: 0.6304 - val_loss: 0.5542 - val_accuracy: 0.6378\n",
            "Epoch 148/500\n",
            "257/257 [==============================] - 0s 741us/step - loss: 0.5447 - accuracy: 0.6693 - val_loss: 0.5504 - val_accuracy: 0.6772\n",
            "Epoch 149/500\n",
            "257/257 [==============================] - 0s 739us/step - loss: 0.5366 - accuracy: 0.6809 - val_loss: 0.5609 - val_accuracy: 0.6772\n",
            "Epoch 150/500\n",
            "257/257 [==============================] - 0s 759us/step - loss: 0.5354 - accuracy: 0.6926 - val_loss: 0.5660 - val_accuracy: 0.6693\n",
            "Epoch 151/500\n",
            "257/257 [==============================] - 0s 717us/step - loss: 0.5361 - accuracy: 0.7004 - val_loss: 0.5854 - val_accuracy: 0.6378\n",
            "Epoch 152/500\n",
            "257/257 [==============================] - 0s 719us/step - loss: 0.5404 - accuracy: 0.6654 - val_loss: 0.6570 - val_accuracy: 0.5591\n",
            "Epoch 153/500\n",
            "257/257 [==============================] - 0s 687us/step - loss: 0.6091 - accuracy: 0.5681 - val_loss: 0.6603 - val_accuracy: 0.5354\n",
            "Epoch 154/500\n",
            "257/257 [==============================] - 0s 685us/step - loss: 0.6023 - accuracy: 0.5798 - val_loss: 0.5984 - val_accuracy: 0.6142\n",
            "Epoch 155/500\n",
            "257/257 [==============================] - 0s 707us/step - loss: 0.5494 - accuracy: 0.6381 - val_loss: 0.5883 - val_accuracy: 0.6220\n",
            "Epoch 156/500\n",
            "257/257 [==============================] - 0s 707us/step - loss: 0.5318 - accuracy: 0.6732 - val_loss: 0.5866 - val_accuracy: 0.6772\n",
            "Epoch 157/500\n",
            "257/257 [==============================] - 0s 710us/step - loss: 0.5289 - accuracy: 0.6926 - val_loss: 0.5885 - val_accuracy: 0.6772\n",
            "Epoch 158/500\n",
            "257/257 [==============================] - 0s 710us/step - loss: 0.5305 - accuracy: 0.6965 - val_loss: 0.6025 - val_accuracy: 0.6772\n",
            "Epoch 159/500\n",
            "257/257 [==============================] - 0s 702us/step - loss: 0.5240 - accuracy: 0.6965 - val_loss: 0.5668 - val_accuracy: 0.7008\n",
            "Epoch 160/500\n",
            "257/257 [==============================] - 0s 672us/step - loss: 0.5160 - accuracy: 0.7082 - val_loss: 0.5535 - val_accuracy: 0.7087\n",
            "Epoch 161/500\n",
            "257/257 [==============================] - 0s 686us/step - loss: 0.5127 - accuracy: 0.7276 - val_loss: 0.5841 - val_accuracy: 0.7323\n",
            "Epoch 162/500\n",
            "257/257 [==============================] - 0s 739us/step - loss: 0.5163 - accuracy: 0.7237 - val_loss: 0.5579 - val_accuracy: 0.7402\n",
            "Epoch 163/500\n",
            "257/257 [==============================] - 0s 759us/step - loss: 0.5099 - accuracy: 0.7276 - val_loss: 0.5420 - val_accuracy: 0.7323\n",
            "Epoch 164/500\n",
            "257/257 [==============================] - 0s 685us/step - loss: 0.5047 - accuracy: 0.7315 - val_loss: 0.5324 - val_accuracy: 0.7165\n",
            "Epoch 165/500\n",
            "257/257 [==============================] - 0s 682us/step - loss: 0.5005 - accuracy: 0.7276 - val_loss: 0.5392 - val_accuracy: 0.7244\n",
            "Epoch 166/500\n",
            "257/257 [==============================] - 0s 692us/step - loss: 0.4975 - accuracy: 0.7237 - val_loss: 0.5723 - val_accuracy: 0.7244\n",
            "Epoch 167/500\n",
            "257/257 [==============================] - 0s 750us/step - loss: 0.5039 - accuracy: 0.7315 - val_loss: 0.6295 - val_accuracy: 0.7087\n",
            "Epoch 168/500\n",
            "257/257 [==============================] - 0s 669us/step - loss: 0.5184 - accuracy: 0.7237 - val_loss: 0.6775 - val_accuracy: 0.7008\n",
            "Epoch 169/500\n",
            "257/257 [==============================] - 0s 711us/step - loss: 0.5273 - accuracy: 0.7315 - val_loss: 0.5548 - val_accuracy: 0.7165\n",
            "Epoch 170/500\n",
            "257/257 [==============================] - 0s 690us/step - loss: 0.4974 - accuracy: 0.7471 - val_loss: 0.5214 - val_accuracy: 0.7480\n",
            "Epoch 171/500\n",
            "257/257 [==============================] - 0s 685us/step - loss: 0.4946 - accuracy: 0.7471 - val_loss: 0.5214 - val_accuracy: 0.7402\n",
            "Epoch 172/500\n",
            "257/257 [==============================] - 0s 670us/step - loss: 0.4936 - accuracy: 0.7549 - val_loss: 0.5291 - val_accuracy: 0.7402\n",
            "Epoch 173/500\n",
            "257/257 [==============================] - 0s 737us/step - loss: 0.4951 - accuracy: 0.7510 - val_loss: 0.5300 - val_accuracy: 0.7323\n",
            "Epoch 174/500\n",
            "257/257 [==============================] - 0s 705us/step - loss: 0.4862 - accuracy: 0.7393 - val_loss: 0.5431 - val_accuracy: 0.7402\n",
            "Epoch 175/500\n",
            "257/257 [==============================] - 0s 669us/step - loss: 0.4885 - accuracy: 0.7276 - val_loss: 0.6078 - val_accuracy: 0.7323\n",
            "Epoch 176/500\n",
            "257/257 [==============================] - 0s 701us/step - loss: 0.4998 - accuracy: 0.7432 - val_loss: 0.5679 - val_accuracy: 0.7323\n",
            "Epoch 177/500\n",
            "257/257 [==============================] - 0s 696us/step - loss: 0.4904 - accuracy: 0.7665 - val_loss: 0.5423 - val_accuracy: 0.7402\n",
            "Epoch 178/500\n",
            "257/257 [==============================] - 0s 751us/step - loss: 0.4834 - accuracy: 0.7510 - val_loss: 0.6250 - val_accuracy: 0.5748\n",
            "Epoch 179/500\n",
            "257/257 [==============================] - 0s 705us/step - loss: 0.6242 - accuracy: 0.5798 - val_loss: 0.6483 - val_accuracy: 0.5354\n",
            "Epoch 180/500\n",
            "257/257 [==============================] - 0s 706us/step - loss: 0.6301 - accuracy: 0.5798 - val_loss: 0.6538 - val_accuracy: 0.5512\n",
            "Epoch 181/500\n",
            "257/257 [==============================] - 0s 673us/step - loss: 0.6259 - accuracy: 0.5953 - val_loss: 0.6677 - val_accuracy: 0.5748\n",
            "Epoch 182/500\n",
            "257/257 [==============================] - 0s 745us/step - loss: 0.6177 - accuracy: 0.6109 - val_loss: 0.5959 - val_accuracy: 0.6378\n",
            "Epoch 183/500\n",
            "257/257 [==============================] - 0s 677us/step - loss: 0.5695 - accuracy: 0.6809 - val_loss: 0.5519 - val_accuracy: 0.6929\n",
            "Epoch 184/500\n",
            "257/257 [==============================] - 0s 708us/step - loss: 0.5403 - accuracy: 0.7043 - val_loss: 0.5296 - val_accuracy: 0.7480\n",
            "Epoch 185/500\n",
            "257/257 [==============================] - 0s 704us/step - loss: 0.5230 - accuracy: 0.7082 - val_loss: 0.5402 - val_accuracy: 0.7559\n",
            "Epoch 186/500\n",
            "257/257 [==============================] - 0s 679us/step - loss: 0.5201 - accuracy: 0.7237 - val_loss: 0.5461 - val_accuracy: 0.7559\n",
            "Epoch 187/500\n",
            "257/257 [==============================] - 0s 675us/step - loss: 0.5205 - accuracy: 0.7121 - val_loss: 0.5303 - val_accuracy: 0.7087\n",
            "Epoch 188/500\n",
            "257/257 [==============================] - 0s 678us/step - loss: 0.5228 - accuracy: 0.7082 - val_loss: 0.5284 - val_accuracy: 0.7559\n",
            "Epoch 189/500\n",
            "257/257 [==============================] - 0s 677us/step - loss: 0.5314 - accuracy: 0.7121 - val_loss: 0.5338 - val_accuracy: 0.7480\n",
            "Epoch 190/500\n",
            "257/257 [==============================] - 0s 699us/step - loss: 0.5340 - accuracy: 0.6965 - val_loss: 0.5389 - val_accuracy: 0.7638\n",
            "Epoch 191/500\n",
            "257/257 [==============================] - 0s 680us/step - loss: 0.5344 - accuracy: 0.6965 - val_loss: 0.5375 - val_accuracy: 0.7402\n",
            "Epoch 192/500\n",
            "257/257 [==============================] - 0s 755us/step - loss: 0.5328 - accuracy: 0.7121 - val_loss: 0.5348 - val_accuracy: 0.7559\n",
            "Epoch 193/500\n",
            "257/257 [==============================] - 0s 748us/step - loss: 0.5320 - accuracy: 0.7121 - val_loss: 0.5325 - val_accuracy: 0.7480\n",
            "Epoch 194/500\n",
            "257/257 [==============================] - 0s 734us/step - loss: 0.5287 - accuracy: 0.7043 - val_loss: 0.5314 - val_accuracy: 0.7638\n",
            "Epoch 195/500\n",
            "257/257 [==============================] - 0s 784us/step - loss: 0.5260 - accuracy: 0.7432 - val_loss: 0.5314 - val_accuracy: 0.7244\n",
            "Epoch 196/500\n",
            "257/257 [==============================] - 0s 766us/step - loss: 0.5216 - accuracy: 0.7121 - val_loss: 0.5327 - val_accuracy: 0.7323\n",
            "Epoch 197/500\n",
            "257/257 [==============================] - 0s 754us/step - loss: 0.5207 - accuracy: 0.7198 - val_loss: 0.5340 - val_accuracy: 0.7165\n",
            "Epoch 198/500\n",
            "257/257 [==============================] - 0s 770us/step - loss: 0.5169 - accuracy: 0.7082 - val_loss: 0.5324 - val_accuracy: 0.6929\n",
            "Epoch 199/500\n",
            "257/257 [==============================] - 0s 741us/step - loss: 0.5170 - accuracy: 0.6926 - val_loss: 0.5373 - val_accuracy: 0.6693\n",
            "Epoch 200/500\n",
            "257/257 [==============================] - 0s 750us/step - loss: 0.5097 - accuracy: 0.7082 - val_loss: 0.5482 - val_accuracy: 0.6929\n",
            "Epoch 201/500\n",
            "257/257 [==============================] - 0s 740us/step - loss: 0.5058 - accuracy: 0.7043 - val_loss: 0.5489 - val_accuracy: 0.7008\n",
            "Epoch 202/500\n",
            "257/257 [==============================] - 0s 746us/step - loss: 0.5131 - accuracy: 0.7004 - val_loss: 0.5802 - val_accuracy: 0.7087\n",
            "Epoch 203/500\n",
            "257/257 [==============================] - 0s 726us/step - loss: 0.5247 - accuracy: 0.7082 - val_loss: 0.6052 - val_accuracy: 0.7323\n",
            "Epoch 204/500\n",
            "257/257 [==============================] - 0s 745us/step - loss: 0.5319 - accuracy: 0.7121 - val_loss: 0.5654 - val_accuracy: 0.7244\n",
            "Epoch 205/500\n",
            "257/257 [==============================] - 0s 785us/step - loss: 0.5137 - accuracy: 0.7004 - val_loss: 0.5704 - val_accuracy: 0.7244\n",
            "Epoch 206/500\n",
            "257/257 [==============================] - 0s 747us/step - loss: 0.5148 - accuracy: 0.7043 - val_loss: 0.5510 - val_accuracy: 0.7008\n",
            "Epoch 207/500\n",
            "257/257 [==============================] - 0s 670us/step - loss: 0.5059 - accuracy: 0.7198 - val_loss: 0.5417 - val_accuracy: 0.7087\n",
            "Epoch 208/500\n",
            "257/257 [==============================] - 0s 697us/step - loss: 0.5012 - accuracy: 0.7160 - val_loss: 0.5663 - val_accuracy: 0.7087\n",
            "Epoch 209/500\n",
            "257/257 [==============================] - 0s 690us/step - loss: 0.5021 - accuracy: 0.7121 - val_loss: 0.5595 - val_accuracy: 0.7087\n",
            "Epoch 210/500\n",
            "257/257 [==============================] - 0s 700us/step - loss: 0.4968 - accuracy: 0.7121 - val_loss: 0.5494 - val_accuracy: 0.7244\n",
            "Epoch 211/500\n",
            "257/257 [==============================] - 0s 713us/step - loss: 0.4936 - accuracy: 0.7237 - val_loss: 0.5448 - val_accuracy: 0.7087\n",
            "Epoch 212/500\n",
            "257/257 [==============================] - 0s 713us/step - loss: 0.4888 - accuracy: 0.7276 - val_loss: 0.5545 - val_accuracy: 0.6535\n",
            "Epoch 213/500\n",
            "257/257 [==============================] - 0s 695us/step - loss: 0.5284 - accuracy: 0.6887 - val_loss: 0.5684 - val_accuracy: 0.6457\n",
            "Epoch 214/500\n",
            "257/257 [==============================] - 0s 673us/step - loss: 0.5323 - accuracy: 0.6848 - val_loss: 0.5785 - val_accuracy: 0.6299\n",
            "Epoch 215/500\n",
            "257/257 [==============================] - 0s 679us/step - loss: 0.5325 - accuracy: 0.6887 - val_loss: 0.5638 - val_accuracy: 0.6772\n",
            "Epoch 216/500\n",
            "257/257 [==============================] - 0s 728us/step - loss: 0.4895 - accuracy: 0.7198 - val_loss: 0.5957 - val_accuracy: 0.7008\n",
            "Epoch 217/500\n",
            "257/257 [==============================] - 0s 730us/step - loss: 0.4917 - accuracy: 0.7160 - val_loss: 0.5882 - val_accuracy: 0.6772\n",
            "Epoch 218/500\n",
            "257/257 [==============================] - 0s 676us/step - loss: 0.4967 - accuracy: 0.7082 - val_loss: 0.6223 - val_accuracy: 0.6929\n",
            "Epoch 219/500\n",
            "257/257 [==============================] - 0s 668us/step - loss: 0.4938 - accuracy: 0.7198 - val_loss: 0.6173 - val_accuracy: 0.6772\n",
            "Epoch 220/500\n",
            "257/257 [==============================] - 0s 695us/step - loss: 0.4939 - accuracy: 0.7160 - val_loss: 0.5947 - val_accuracy: 0.6929\n",
            "Epoch 221/500\n",
            "257/257 [==============================] - 0s 685us/step - loss: 0.4829 - accuracy: 0.7354 - val_loss: 0.5837 - val_accuracy: 0.6929\n",
            "Epoch 222/500\n",
            "257/257 [==============================] - 0s 711us/step - loss: 0.4802 - accuracy: 0.7315 - val_loss: 0.5959 - val_accuracy: 0.7165\n",
            "Epoch 223/500\n",
            "257/257 [==============================] - 0s 695us/step - loss: 0.4854 - accuracy: 0.7276 - val_loss: 0.5754 - val_accuracy: 0.7323\n",
            "Epoch 224/500\n",
            "257/257 [==============================] - 0s 716us/step - loss: 0.4857 - accuracy: 0.7354 - val_loss: 0.5556 - val_accuracy: 0.7244\n",
            "Epoch 225/500\n",
            "257/257 [==============================] - 0s 690us/step - loss: 0.4803 - accuracy: 0.7393 - val_loss: 0.5776 - val_accuracy: 0.7402\n",
            "Epoch 226/500\n",
            "257/257 [==============================] - 0s 760us/step - loss: 0.4790 - accuracy: 0.7549 - val_loss: 0.5832 - val_accuracy: 0.7244\n",
            "Epoch 227/500\n",
            "257/257 [==============================] - 0s 756us/step - loss: 0.4732 - accuracy: 0.7471 - val_loss: 0.5836 - val_accuracy: 0.6614\n",
            "Epoch 228/500\n",
            "257/257 [==============================] - 0s 666us/step - loss: 0.5235 - accuracy: 0.6809 - val_loss: 0.5914 - val_accuracy: 0.6457\n",
            "Epoch 229/500\n",
            "257/257 [==============================] - 0s 733us/step - loss: 0.5182 - accuracy: 0.6809 - val_loss: 0.5690 - val_accuracy: 0.6614\n",
            "Epoch 230/500\n",
            "257/257 [==============================] - 0s 774us/step - loss: 0.4941 - accuracy: 0.6926 - val_loss: 0.5875 - val_accuracy: 0.6850\n",
            "Epoch 231/500\n",
            "257/257 [==============================] - 0s 736us/step - loss: 0.4775 - accuracy: 0.7237 - val_loss: 0.6195 - val_accuracy: 0.6929\n",
            "Epoch 232/500\n",
            "257/257 [==============================] - 0s 746us/step - loss: 0.4731 - accuracy: 0.7354 - val_loss: 0.6149 - val_accuracy: 0.7165\n",
            "Epoch 233/500\n",
            "257/257 [==============================] - 0s 720us/step - loss: 0.4929 - accuracy: 0.7549 - val_loss: 0.5960 - val_accuracy: 0.7165\n",
            "Epoch 234/500\n",
            "257/257 [==============================] - 0s 686us/step - loss: 0.4928 - accuracy: 0.7471 - val_loss: 0.6308 - val_accuracy: 0.7559\n",
            "Epoch 235/500\n",
            "257/257 [==============================] - 0s 678us/step - loss: 0.4986 - accuracy: 0.7393 - val_loss: 0.6185 - val_accuracy: 0.7559\n",
            "Epoch 236/500\n",
            "257/257 [==============================] - 0s 668us/step - loss: 0.4945 - accuracy: 0.7354 - val_loss: 0.7195 - val_accuracy: 0.7559\n",
            "Epoch 237/500\n",
            "257/257 [==============================] - 0s 705us/step - loss: 0.5200 - accuracy: 0.7160 - val_loss: 0.6627 - val_accuracy: 0.7717\n",
            "Epoch 238/500\n",
            "257/257 [==============================] - 0s 728us/step - loss: 0.5011 - accuracy: 0.7237 - val_loss: 0.6524 - val_accuracy: 0.7559\n",
            "Epoch 239/500\n",
            "257/257 [==============================] - 0s 691us/step - loss: 0.4934 - accuracy: 0.7354 - val_loss: 0.5827 - val_accuracy: 0.6929\n",
            "Epoch 240/500\n",
            "257/257 [==============================] - 0s 681us/step - loss: 0.4802 - accuracy: 0.7510 - val_loss: 0.5795 - val_accuracy: 0.6929\n",
            "Epoch 241/500\n",
            "257/257 [==============================] - 0s 716us/step - loss: 0.4754 - accuracy: 0.7510 - val_loss: 0.6578 - val_accuracy: 0.7165\n",
            "Epoch 242/500\n",
            "257/257 [==============================] - 0s 702us/step - loss: 0.4907 - accuracy: 0.7471 - val_loss: 0.6232 - val_accuracy: 0.7008\n",
            "Epoch 243/500\n",
            "257/257 [==============================] - 0s 679us/step - loss: 0.4754 - accuracy: 0.7432 - val_loss: 0.5933 - val_accuracy: 0.6929\n",
            "Epoch 244/500\n",
            "257/257 [==============================] - 0s 804us/step - loss: 0.4889 - accuracy: 0.7471 - val_loss: 0.5838 - val_accuracy: 0.7087\n",
            "Epoch 245/500\n",
            "257/257 [==============================] - 0s 684us/step - loss: 0.4753 - accuracy: 0.7432 - val_loss: 0.6213 - val_accuracy: 0.7008\n",
            "Epoch 246/500\n",
            "257/257 [==============================] - 0s 689us/step - loss: 0.4886 - accuracy: 0.7510 - val_loss: 0.6058 - val_accuracy: 0.6929\n",
            "Epoch 247/500\n",
            "257/257 [==============================] - 0s 693us/step - loss: 0.4721 - accuracy: 0.7626 - val_loss: 0.6617 - val_accuracy: 0.7717\n",
            "Epoch 248/500\n",
            "257/257 [==============================] - 0s 701us/step - loss: 0.4961 - accuracy: 0.7393 - val_loss: 0.5839 - val_accuracy: 0.6850\n",
            "Epoch 249/500\n",
            "257/257 [==============================] - 0s 693us/step - loss: 0.4799 - accuracy: 0.7393 - val_loss: 0.5943 - val_accuracy: 0.6850\n",
            "Epoch 250/500\n",
            "257/257 [==============================] - 0s 726us/step - loss: 0.4805 - accuracy: 0.7354 - val_loss: 0.6066 - val_accuracy: 0.6772\n",
            "Epoch 251/500\n",
            "257/257 [==============================] - 0s 674us/step - loss: 0.4845 - accuracy: 0.7276 - val_loss: 0.6301 - val_accuracy: 0.6850\n",
            "Epoch 252/500\n",
            "257/257 [==============================] - 0s 735us/step - loss: 0.4863 - accuracy: 0.7354 - val_loss: 0.6453 - val_accuracy: 0.6929\n",
            "Epoch 253/500\n",
            "257/257 [==============================] - 0s 786us/step - loss: 0.4882 - accuracy: 0.7276 - val_loss: 0.6528 - val_accuracy: 0.6929\n",
            "Epoch 254/500\n",
            "257/257 [==============================] - 0s 730us/step - loss: 0.4889 - accuracy: 0.7276 - val_loss: 0.7256 - val_accuracy: 0.7087\n",
            "Epoch 255/500\n",
            "257/257 [==============================] - 0s 703us/step - loss: 0.4936 - accuracy: 0.7315 - val_loss: 0.6370 - val_accuracy: 0.5591\n",
            "Epoch 256/500\n",
            "257/257 [==============================] - 0s 693us/step - loss: 0.6250 - accuracy: 0.5914 - val_loss: 0.6496 - val_accuracy: 0.5276\n",
            "Epoch 257/500\n",
            "257/257 [==============================] - 0s 670us/step - loss: 0.6334 - accuracy: 0.5720 - val_loss: 0.6458 - val_accuracy: 0.5276\n",
            "Epoch 258/500\n",
            "257/257 [==============================] - 0s 692us/step - loss: 0.6314 - accuracy: 0.5798 - val_loss: 0.5910 - val_accuracy: 0.6457\n",
            "Epoch 259/500\n",
            "257/257 [==============================] - 0s 724us/step - loss: 0.5806 - accuracy: 0.6732 - val_loss: 0.5467 - val_accuracy: 0.7008\n",
            "Epoch 260/500\n",
            "257/257 [==============================] - 0s 677us/step - loss: 0.5450 - accuracy: 0.7354 - val_loss: 0.5135 - val_accuracy: 0.7402\n",
            "Epoch 261/500\n",
            "257/257 [==============================] - 0s 716us/step - loss: 0.5187 - accuracy: 0.7432 - val_loss: 0.5093 - val_accuracy: 0.7402\n",
            "Epoch 262/500\n",
            "257/257 [==============================] - 0s 682us/step - loss: 0.5062 - accuracy: 0.7432 - val_loss: 0.5214 - val_accuracy: 0.7323\n",
            "Epoch 263/500\n",
            "257/257 [==============================] - 0s 676us/step - loss: 0.4986 - accuracy: 0.7198 - val_loss: 0.5312 - val_accuracy: 0.7323\n",
            "Epoch 264/500\n",
            "257/257 [==============================] - 0s 707us/step - loss: 0.4901 - accuracy: 0.7354 - val_loss: 0.5686 - val_accuracy: 0.7087\n",
            "Epoch 265/500\n",
            "257/257 [==============================] - 0s 673us/step - loss: 0.4946 - accuracy: 0.7276 - val_loss: 0.5732 - val_accuracy: 0.7087\n",
            "Epoch 266/500\n",
            "257/257 [==============================] - 0s 730us/step - loss: 0.4904 - accuracy: 0.7315 - val_loss: 0.6154 - val_accuracy: 0.7087\n",
            "Epoch 267/500\n",
            "257/257 [==============================] - 0s 742us/step - loss: 0.5013 - accuracy: 0.7393 - val_loss: 0.5732 - val_accuracy: 0.7087\n",
            "Epoch 268/500\n",
            "257/257 [==============================] - 0s 665us/step - loss: 0.4895 - accuracy: 0.7432 - val_loss: 0.5407 - val_accuracy: 0.7323\n",
            "Epoch 269/500\n",
            "257/257 [==============================] - 0s 698us/step - loss: 0.4885 - accuracy: 0.7315 - val_loss: 0.5303 - val_accuracy: 0.7008\n",
            "Epoch 270/500\n",
            "257/257 [==============================] - 0s 698us/step - loss: 0.5293 - accuracy: 0.7315 - val_loss: 0.5346 - val_accuracy: 0.6929\n",
            "Epoch 271/500\n",
            "257/257 [==============================] - 0s 674us/step - loss: 0.5358 - accuracy: 0.7276 - val_loss: 0.5234 - val_accuracy: 0.7323\n",
            "Epoch 272/500\n",
            "257/257 [==============================] - 0s 714us/step - loss: 0.5247 - accuracy: 0.7315 - val_loss: 0.5230 - val_accuracy: 0.7717\n",
            "Epoch 273/500\n",
            "257/257 [==============================] - 0s 677us/step - loss: 0.5171 - accuracy: 0.7354 - val_loss: 0.5606 - val_accuracy: 0.7638\n",
            "Epoch 274/500\n",
            "257/257 [==============================] - 0s 665us/step - loss: 0.5184 - accuracy: 0.7354 - val_loss: 0.5512 - val_accuracy: 0.7559\n",
            "Epoch 275/500\n",
            "257/257 [==============================] - 0s 674us/step - loss: 0.5156 - accuracy: 0.7393 - val_loss: 0.5543 - val_accuracy: 0.7717\n",
            "Epoch 276/500\n",
            "257/257 [==============================] - 0s 707us/step - loss: 0.5106 - accuracy: 0.7276 - val_loss: 0.5147 - val_accuracy: 0.7638\n",
            "Epoch 277/500\n",
            "257/257 [==============================] - 0s 760us/step - loss: 0.5012 - accuracy: 0.7354 - val_loss: 0.5182 - val_accuracy: 0.7165\n",
            "Epoch 278/500\n",
            "257/257 [==============================] - 0s 692us/step - loss: 0.5104 - accuracy: 0.7160 - val_loss: 0.5218 - val_accuracy: 0.7008\n",
            "Epoch 279/500\n",
            "257/257 [==============================] - 0s 682us/step - loss: 0.5036 - accuracy: 0.7237 - val_loss: 0.5350 - val_accuracy: 0.7638\n",
            "Epoch 280/500\n",
            "257/257 [==============================] - 0s 702us/step - loss: 0.4941 - accuracy: 0.7471 - val_loss: 0.5318 - val_accuracy: 0.7008\n",
            "Epoch 281/500\n",
            "257/257 [==============================] - 0s 666us/step - loss: 0.4954 - accuracy: 0.7198 - val_loss: 0.5468 - val_accuracy: 0.7008\n",
            "Epoch 282/500\n",
            "257/257 [==============================] - 0s 700us/step - loss: 0.4872 - accuracy: 0.7276 - val_loss: 0.5564 - val_accuracy: 0.6929\n",
            "Epoch 283/500\n",
            "257/257 [==============================] - 0s 740us/step - loss: 0.4802 - accuracy: 0.7276 - val_loss: 0.5771 - val_accuracy: 0.7087\n",
            "Epoch 284/500\n",
            "257/257 [==============================] - 0s 677us/step - loss: 0.4778 - accuracy: 0.7315 - val_loss: 0.5663 - val_accuracy: 0.7087\n",
            "Epoch 285/500\n",
            "257/257 [==============================] - 0s 698us/step - loss: 0.4757 - accuracy: 0.7393 - val_loss: 0.5506 - val_accuracy: 0.6772\n",
            "Epoch 286/500\n",
            "257/257 [==============================] - 0s 682us/step - loss: 0.5281 - accuracy: 0.6809 - val_loss: 0.5612 - val_accuracy: 0.6614\n",
            "Epoch 287/500\n",
            "257/257 [==============================] - 0s 743us/step - loss: 0.5375 - accuracy: 0.6770 - val_loss: 0.5515 - val_accuracy: 0.6772\n",
            "Epoch 288/500\n",
            "257/257 [==============================] - 0s 687us/step - loss: 0.5293 - accuracy: 0.6965 - val_loss: 0.5396 - val_accuracy: 0.6929\n",
            "Epoch 289/500\n",
            "257/257 [==============================] - 0s 714us/step - loss: 0.5127 - accuracy: 0.7121 - val_loss: 0.5385 - val_accuracy: 0.6929\n",
            "Epoch 290/500\n",
            "257/257 [==============================] - 0s 696us/step - loss: 0.4946 - accuracy: 0.7354 - val_loss: 0.5666 - val_accuracy: 0.7480\n",
            "Epoch 291/500\n",
            "257/257 [==============================] - 0s 692us/step - loss: 0.4960 - accuracy: 0.7354 - val_loss: 0.6271 - val_accuracy: 0.7638\n",
            "Epoch 292/500\n",
            "257/257 [==============================] - 0s 665us/step - loss: 0.5118 - accuracy: 0.7198 - val_loss: 0.5936 - val_accuracy: 0.7402\n",
            "Epoch 293/500\n",
            "257/257 [==============================] - 0s 693us/step - loss: 0.4858 - accuracy: 0.7237 - val_loss: 0.6016 - val_accuracy: 0.7323\n",
            "Epoch 294/500\n",
            "257/257 [==============================] - 0s 773us/step - loss: 0.4844 - accuracy: 0.7315 - val_loss: 0.6051 - val_accuracy: 0.7244\n",
            "Epoch 295/500\n",
            "257/257 [==============================] - 0s 697us/step - loss: 0.4830 - accuracy: 0.7315 - val_loss: 0.5911 - val_accuracy: 0.7087\n",
            "Epoch 296/500\n",
            "257/257 [==============================] - 0s 684us/step - loss: 0.4819 - accuracy: 0.7354 - val_loss: 0.5859 - val_accuracy: 0.7087\n",
            "Epoch 297/500\n",
            "257/257 [==============================] - 0s 699us/step - loss: 0.4714 - accuracy: 0.7432 - val_loss: 0.6102 - val_accuracy: 0.7402\n",
            "Epoch 298/500\n",
            "257/257 [==============================] - 0s 688us/step - loss: 0.4767 - accuracy: 0.7471 - val_loss: 0.5970 - val_accuracy: 0.7323\n",
            "Epoch 299/500\n",
            "257/257 [==============================] - 0s 668us/step - loss: 0.4666 - accuracy: 0.7432 - val_loss: 0.6005 - val_accuracy: 0.7402\n",
            "Epoch 300/500\n",
            "257/257 [==============================] - 0s 804us/step - loss: 0.4629 - accuracy: 0.7471 - val_loss: 0.6112 - val_accuracy: 0.7087\n",
            "Epoch 301/500\n",
            "257/257 [==============================] - 0s 691us/step - loss: 0.4664 - accuracy: 0.7471 - val_loss: 0.5809 - val_accuracy: 0.7087\n",
            "Epoch 302/500\n",
            "257/257 [==============================] - 0s 685us/step - loss: 0.4697 - accuracy: 0.7432 - val_loss: 0.5613 - val_accuracy: 0.7165\n",
            "Epoch 303/500\n",
            "257/257 [==============================] - 0s 669us/step - loss: 0.4639 - accuracy: 0.7354 - val_loss: 0.5799 - val_accuracy: 0.7953\n",
            "Epoch 304/500\n",
            "257/257 [==============================] - 0s 734us/step - loss: 0.4761 - accuracy: 0.7237 - val_loss: 0.5564 - val_accuracy: 0.7402\n",
            "Epoch 305/500\n",
            "257/257 [==============================] - 0s 752us/step - loss: 0.4503 - accuracy: 0.7432 - val_loss: 0.9141 - val_accuracy: 0.7008\n",
            "Epoch 306/500\n",
            "257/257 [==============================] - 0s 761us/step - loss: 0.5801 - accuracy: 0.7237 - val_loss: 0.5383 - val_accuracy: 0.7087\n",
            "Epoch 307/500\n",
            "257/257 [==============================] - 0s 735us/step - loss: 0.4899 - accuracy: 0.7432 - val_loss: 0.5458 - val_accuracy: 0.6850\n",
            "Epoch 308/500\n",
            "257/257 [==============================] - 0s 706us/step - loss: 0.4903 - accuracy: 0.7471 - val_loss: 0.5525 - val_accuracy: 0.7087\n",
            "Epoch 309/500\n",
            "257/257 [==============================] - 0s 716us/step - loss: 0.4709 - accuracy: 0.7510 - val_loss: 0.5740 - val_accuracy: 0.7008\n",
            "Epoch 310/500\n",
            "257/257 [==============================] - 0s 688us/step - loss: 0.4653 - accuracy: 0.7510 - val_loss: 0.6060 - val_accuracy: 0.7008\n",
            "Epoch 311/500\n",
            "257/257 [==============================] - 0s 712us/step - loss: 0.4762 - accuracy: 0.7510 - val_loss: 0.6611 - val_accuracy: 0.7008\n",
            "Epoch 312/500\n",
            "257/257 [==============================] - 0s 691us/step - loss: 0.4718 - accuracy: 0.7549 - val_loss: 0.6916 - val_accuracy: 0.7323\n",
            "Epoch 313/500\n",
            "257/257 [==============================] - 0s 677us/step - loss: 0.4760 - accuracy: 0.7549 - val_loss: 0.5784 - val_accuracy: 0.6772\n",
            "Epoch 314/500\n",
            "257/257 [==============================] - 0s 738us/step - loss: 0.5014 - accuracy: 0.7004 - val_loss: 0.5583 - val_accuracy: 0.6929\n",
            "Epoch 315/500\n",
            "257/257 [==============================] - 0s 679us/step - loss: 0.4869 - accuracy: 0.7354 - val_loss: 0.5719 - val_accuracy: 0.6929\n",
            "Epoch 316/500\n",
            "257/257 [==============================] - 0s 673us/step - loss: 0.4667 - accuracy: 0.7510 - val_loss: 0.6895 - val_accuracy: 0.7795\n",
            "Epoch 317/500\n",
            "257/257 [==============================] - 0s 695us/step - loss: 0.4887 - accuracy: 0.7354 - val_loss: 0.6263 - val_accuracy: 0.7165\n",
            "Epoch 318/500\n",
            "257/257 [==============================] - 0s 688us/step - loss: 0.4542 - accuracy: 0.7626 - val_loss: 0.6648 - val_accuracy: 0.7402\n",
            "Epoch 319/500\n",
            "257/257 [==============================] - 0s 772us/step - loss: 0.4889 - accuracy: 0.7549 - val_loss: 0.6423 - val_accuracy: 0.7559\n",
            "Epoch 320/500\n",
            "257/257 [==============================] - 0s 723us/step - loss: 0.4984 - accuracy: 0.7432 - val_loss: 0.6211 - val_accuracy: 0.7323\n",
            "Epoch 321/500\n",
            "257/257 [==============================] - 0s 665us/step - loss: 0.4995 - accuracy: 0.7510 - val_loss: 0.5520 - val_accuracy: 0.6929\n",
            "Epoch 322/500\n",
            "257/257 [==============================] - 0s 710us/step - loss: 0.5354 - accuracy: 0.6965 - val_loss: 0.5542 - val_accuracy: 0.6772\n",
            "Epoch 323/500\n",
            "257/257 [==============================] - 0s 682us/step - loss: 0.5381 - accuracy: 0.6887 - val_loss: 0.5540 - val_accuracy: 0.6772\n",
            "Epoch 324/500\n",
            "257/257 [==============================] - 0s 687us/step - loss: 0.5307 - accuracy: 0.6926 - val_loss: 0.5573 - val_accuracy: 0.6693\n",
            "Epoch 325/500\n",
            "257/257 [==============================] - 0s 729us/step - loss: 0.5210 - accuracy: 0.7004 - val_loss: 0.5700 - val_accuracy: 0.6929\n",
            "Epoch 326/500\n",
            "257/257 [==============================] - 0s 698us/step - loss: 0.5123 - accuracy: 0.7198 - val_loss: 0.5894 - val_accuracy: 0.7087\n",
            "Epoch 327/500\n",
            "257/257 [==============================] - 0s 704us/step - loss: 0.5056 - accuracy: 0.7121 - val_loss: 0.6049 - val_accuracy: 0.7165\n",
            "Epoch 328/500\n",
            "257/257 [==============================] - 0s 702us/step - loss: 0.5005 - accuracy: 0.7160 - val_loss: 0.6159 - val_accuracy: 0.7087\n",
            "Epoch 329/500\n",
            "257/257 [==============================] - 0s 684us/step - loss: 0.4972 - accuracy: 0.7121 - val_loss: 0.6705 - val_accuracy: 0.7087\n",
            "Epoch 330/500\n",
            "257/257 [==============================] - 0s 759us/step - loss: 0.4979 - accuracy: 0.7160 - val_loss: 0.7245 - val_accuracy: 0.7480\n",
            "Epoch 331/500\n",
            "257/257 [==============================] - 0s 697us/step - loss: 0.5055 - accuracy: 0.7276 - val_loss: 0.6537 - val_accuracy: 0.7402\n",
            "Epoch 332/500\n",
            "257/257 [==============================] - 0s 698us/step - loss: 0.4808 - accuracy: 0.7510 - val_loss: 0.6019 - val_accuracy: 0.7323\n",
            "Epoch 333/500\n",
            "257/257 [==============================] - 0s 691us/step - loss: 0.4665 - accuracy: 0.7665 - val_loss: 0.5839 - val_accuracy: 0.7244\n",
            "Epoch 334/500\n",
            "257/257 [==============================] - 0s 681us/step - loss: 0.4629 - accuracy: 0.7626 - val_loss: 0.6042 - val_accuracy: 0.7480\n",
            "Epoch 335/500\n",
            "257/257 [==============================] - 0s 696us/step - loss: 0.4603 - accuracy: 0.7626 - val_loss: 0.6047 - val_accuracy: 0.7480\n",
            "Epoch 336/500\n",
            "257/257 [==============================] - 0s 699us/step - loss: 0.4567 - accuracy: 0.7626 - val_loss: 0.5984 - val_accuracy: 0.7559\n",
            "Epoch 337/500\n",
            "257/257 [==============================] - 0s 706us/step - loss: 0.4565 - accuracy: 0.7549 - val_loss: 0.5764 - val_accuracy: 0.7480\n",
            "Epoch 338/500\n",
            "257/257 [==============================] - 0s 793us/step - loss: 0.4524 - accuracy: 0.7588 - val_loss: 0.5837 - val_accuracy: 0.7244\n",
            "Epoch 339/500\n",
            "257/257 [==============================] - 0s 665us/step - loss: 0.4513 - accuracy: 0.7626 - val_loss: 0.5930 - val_accuracy: 0.6929\n",
            "Epoch 340/500\n",
            "257/257 [==============================] - 0s 698us/step - loss: 0.4471 - accuracy: 0.7588 - val_loss: 0.6941 - val_accuracy: 0.7717\n",
            "Epoch 341/500\n",
            "257/257 [==============================] - 0s 695us/step - loss: 0.4668 - accuracy: 0.7588 - val_loss: 0.6367 - val_accuracy: 0.7402\n",
            "Epoch 342/500\n",
            "257/257 [==============================] - 0s 749us/step - loss: 0.4509 - accuracy: 0.7588 - val_loss: 0.5992 - val_accuracy: 0.7874\n",
            "Epoch 343/500\n",
            "257/257 [==============================] - 0s 679us/step - loss: 0.4525 - accuracy: 0.7665 - val_loss: 0.5769 - val_accuracy: 0.7874\n",
            "Epoch 344/500\n",
            "257/257 [==============================] - 0s 715us/step - loss: 0.4470 - accuracy: 0.7704 - val_loss: 0.5712 - val_accuracy: 0.7874\n",
            "Epoch 345/500\n",
            "257/257 [==============================] - 0s 694us/step - loss: 0.4436 - accuracy: 0.7704 - val_loss: 0.6149 - val_accuracy: 0.7874\n",
            "Epoch 346/500\n",
            "257/257 [==============================] - 0s 707us/step - loss: 0.4455 - accuracy: 0.7782 - val_loss: 0.6029 - val_accuracy: 0.7559\n",
            "Epoch 347/500\n",
            "257/257 [==============================] - 0s 682us/step - loss: 0.4404 - accuracy: 0.7743 - val_loss: 0.6544 - val_accuracy: 0.7953\n",
            "Epoch 348/500\n",
            "257/257 [==============================] - 0s 721us/step - loss: 0.4502 - accuracy: 0.7743 - val_loss: 0.6440 - val_accuracy: 0.7480\n",
            "Epoch 349/500\n",
            "257/257 [==============================] - 0s 678us/step - loss: 0.4551 - accuracy: 0.7626 - val_loss: 0.6351 - val_accuracy: 0.7559\n",
            "Epoch 350/500\n",
            "257/257 [==============================] - 0s 697us/step - loss: 0.4457 - accuracy: 0.7704 - val_loss: 0.5850 - val_accuracy: 0.8031\n",
            "Epoch 351/500\n",
            "257/257 [==============================] - 0s 705us/step - loss: 0.4435 - accuracy: 0.7782 - val_loss: 0.5989 - val_accuracy: 0.8031\n",
            "Epoch 352/500\n",
            "257/257 [==============================] - 0s 715us/step - loss: 0.4565 - accuracy: 0.7432 - val_loss: 0.5696 - val_accuracy: 0.7717\n",
            "Epoch 353/500\n",
            "257/257 [==============================] - 0s 686us/step - loss: 0.4292 - accuracy: 0.7743 - val_loss: 0.5744 - val_accuracy: 0.7717\n",
            "Epoch 354/500\n",
            "257/257 [==============================] - 0s 769us/step - loss: 0.4376 - accuracy: 0.7510 - val_loss: 0.5816 - val_accuracy: 0.7559\n",
            "Epoch 355/500\n",
            "257/257 [==============================] - 0s 678us/step - loss: 0.4319 - accuracy: 0.7665 - val_loss: 0.6528 - val_accuracy: 0.6693\n",
            "Epoch 356/500\n",
            "257/257 [==============================] - 0s 700us/step - loss: 0.4765 - accuracy: 0.7237 - val_loss: 0.6590 - val_accuracy: 0.6614\n",
            "Epoch 357/500\n",
            "257/257 [==============================] - 0s 724us/step - loss: 0.4777 - accuracy: 0.7198 - val_loss: 0.7517 - val_accuracy: 0.6693\n",
            "Epoch 358/500\n",
            "257/257 [==============================] - 0s 829us/step - loss: 0.4879 - accuracy: 0.7160 - val_loss: 0.6781 - val_accuracy: 0.6535\n",
            "Epoch 359/500\n",
            "257/257 [==============================] - 0s 705us/step - loss: 0.4662 - accuracy: 0.7276 - val_loss: 0.6706 - val_accuracy: 0.6457\n",
            "Epoch 360/500\n",
            "257/257 [==============================] - 0s 681us/step - loss: 0.4551 - accuracy: 0.7354 - val_loss: 0.8372 - val_accuracy: 0.7008\n",
            "Epoch 361/500\n",
            "257/257 [==============================] - 0s 685us/step - loss: 0.4771 - accuracy: 0.7510 - val_loss: 0.7627 - val_accuracy: 0.7008\n",
            "Epoch 362/500\n",
            "257/257 [==============================] - 0s 703us/step - loss: 0.4628 - accuracy: 0.7471 - val_loss: 0.8076 - val_accuracy: 0.6929\n",
            "Epoch 363/500\n",
            "257/257 [==============================] - 0s 758us/step - loss: 0.4646 - accuracy: 0.7471 - val_loss: 0.8146 - val_accuracy: 0.6929\n",
            "Epoch 364/500\n",
            "257/257 [==============================] - 0s 751us/step - loss: 0.4575 - accuracy: 0.7549 - val_loss: 0.6423 - val_accuracy: 0.6299\n",
            "Epoch 365/500\n",
            "257/257 [==============================] - 0s 783us/step - loss: 0.5539 - accuracy: 0.6965 - val_loss: 0.6634 - val_accuracy: 0.6299\n",
            "Epoch 366/500\n",
            "257/257 [==============================] - 0s 711us/step - loss: 0.5830 - accuracy: 0.6693 - val_loss: 0.6453 - val_accuracy: 0.6299\n",
            "Epoch 367/500\n",
            "257/257 [==============================] - 0s 691us/step - loss: 0.5704 - accuracy: 0.6732 - val_loss: 0.6077 - val_accuracy: 0.6693\n",
            "Epoch 368/500\n",
            "257/257 [==============================] - 0s 700us/step - loss: 0.5314 - accuracy: 0.7276 - val_loss: 0.5711 - val_accuracy: 0.7087\n",
            "Epoch 369/500\n",
            "257/257 [==============================] - 0s 704us/step - loss: 0.4796 - accuracy: 0.7510 - val_loss: 0.5932 - val_accuracy: 0.6929\n",
            "Epoch 370/500\n",
            "257/257 [==============================] - 0s 704us/step - loss: 0.4549 - accuracy: 0.7315 - val_loss: 0.6533 - val_accuracy: 0.7480\n",
            "Epoch 371/500\n",
            "257/257 [==============================] - 0s 678us/step - loss: 0.4545 - accuracy: 0.7315 - val_loss: 0.6815 - val_accuracy: 0.7480\n",
            "Epoch 372/500\n",
            "257/257 [==============================] - 0s 718us/step - loss: 0.4541 - accuracy: 0.7354 - val_loss: 0.6981 - val_accuracy: 0.7087\n",
            "Epoch 373/500\n",
            "257/257 [==============================] - 0s 700us/step - loss: 0.4509 - accuracy: 0.7471 - val_loss: 0.6288 - val_accuracy: 0.7087\n",
            "Epoch 374/500\n",
            "257/257 [==============================] - 0s 728us/step - loss: 0.4798 - accuracy: 0.7315 - val_loss: 0.6384 - val_accuracy: 0.7087\n",
            "Epoch 375/500\n",
            "257/257 [==============================] - 0s 705us/step - loss: 0.4962 - accuracy: 0.7198 - val_loss: 0.6280 - val_accuracy: 0.7087\n",
            "Epoch 376/500\n",
            "257/257 [==============================] - 0s 714us/step - loss: 0.4877 - accuracy: 0.7160 - val_loss: 0.6287 - val_accuracy: 0.6929\n",
            "Epoch 377/500\n",
            "257/257 [==============================] - 0s 774us/step - loss: 0.4826 - accuracy: 0.7237 - val_loss: 0.6435 - val_accuracy: 0.6929\n",
            "Epoch 378/500\n",
            "257/257 [==============================] - 0s 732us/step - loss: 0.4765 - accuracy: 0.7160 - val_loss: 0.6385 - val_accuracy: 0.7165\n",
            "Epoch 379/500\n",
            "257/257 [==============================] - 0s 698us/step - loss: 0.4631 - accuracy: 0.7315 - val_loss: 0.7081 - val_accuracy: 0.7008\n",
            "Epoch 380/500\n",
            "257/257 [==============================] - 0s 724us/step - loss: 0.4686 - accuracy: 0.7510 - val_loss: 0.6595 - val_accuracy: 0.7087\n",
            "Epoch 381/500\n",
            "257/257 [==============================] - 0s 698us/step - loss: 0.4585 - accuracy: 0.7471 - val_loss: 0.5712 - val_accuracy: 0.7087\n",
            "Epoch 382/500\n",
            "257/257 [==============================] - 0s 748us/step - loss: 0.4768 - accuracy: 0.7237 - val_loss: 0.5764 - val_accuracy: 0.6850\n",
            "Epoch 383/500\n",
            "257/257 [==============================] - 0s 720us/step - loss: 0.4819 - accuracy: 0.7198 - val_loss: 0.5783 - val_accuracy: 0.6850\n",
            "Epoch 384/500\n",
            "257/257 [==============================] - 0s 706us/step - loss: 0.4785 - accuracy: 0.7198 - val_loss: 0.5765 - val_accuracy: 0.7008\n",
            "Epoch 385/500\n",
            "257/257 [==============================] - 0s 716us/step - loss: 0.4654 - accuracy: 0.7237 - val_loss: 0.5814 - val_accuracy: 0.7087\n",
            "Epoch 386/500\n",
            "257/257 [==============================] - 0s 763us/step - loss: 0.4676 - accuracy: 0.7198 - val_loss: 0.5912 - val_accuracy: 0.6772\n",
            "Epoch 387/500\n",
            "257/257 [==============================] - 0s 794us/step - loss: 0.5406 - accuracy: 0.7198 - val_loss: 0.6234 - val_accuracy: 0.6614\n",
            "Epoch 388/500\n",
            "257/257 [==============================] - 0s 769us/step - loss: 0.5648 - accuracy: 0.7315 - val_loss: 0.6153 - val_accuracy: 0.6614\n",
            "Epoch 389/500\n",
            "257/257 [==============================] - 0s 709us/step - loss: 0.5527 - accuracy: 0.7237 - val_loss: 0.5622 - val_accuracy: 0.7087\n",
            "Epoch 390/500\n",
            "257/257 [==============================] - 0s 738us/step - loss: 0.5286 - accuracy: 0.7276 - val_loss: 0.5527 - val_accuracy: 0.7087\n",
            "Epoch 391/500\n",
            "257/257 [==============================] - 0s 702us/step - loss: 0.5032 - accuracy: 0.7432 - val_loss: 0.5432 - val_accuracy: 0.7480\n",
            "Epoch 392/500\n",
            "257/257 [==============================] - 0s 692us/step - loss: 0.4638 - accuracy: 0.7354 - val_loss: 0.6331 - val_accuracy: 0.7559\n",
            "Epoch 393/500\n",
            "257/257 [==============================] - 0s 752us/step - loss: 0.4543 - accuracy: 0.7237 - val_loss: 0.6620 - val_accuracy: 0.7008\n",
            "Epoch 394/500\n",
            "257/257 [==============================] - 0s 733us/step - loss: 0.4453 - accuracy: 0.7549 - val_loss: 0.6887 - val_accuracy: 0.7008\n",
            "Epoch 395/500\n",
            "257/257 [==============================] - 0s 765us/step - loss: 0.4420 - accuracy: 0.7393 - val_loss: 0.8142 - val_accuracy: 0.7087\n",
            "Epoch 396/500\n",
            "257/257 [==============================] - 0s 765us/step - loss: 0.4455 - accuracy: 0.7432 - val_loss: 0.8007 - val_accuracy: 0.7008\n",
            "Epoch 397/500\n",
            "257/257 [==============================] - 0s 770us/step - loss: 0.4353 - accuracy: 0.7471 - val_loss: 0.7085 - val_accuracy: 0.7008\n",
            "Epoch 398/500\n",
            "257/257 [==============================] - 0s 744us/step - loss: 0.4334 - accuracy: 0.7471 - val_loss: 0.6909 - val_accuracy: 0.6772\n",
            "Epoch 399/500\n",
            "257/257 [==============================] - 0s 691us/step - loss: 0.4339 - accuracy: 0.7626 - val_loss: 0.7900 - val_accuracy: 0.6850\n",
            "Epoch 400/500\n",
            "257/257 [==============================] - 0s 681us/step - loss: 0.4437 - accuracy: 0.7510 - val_loss: 0.8315 - val_accuracy: 0.7402\n",
            "Epoch 401/500\n",
            "257/257 [==============================] - 0s 745us/step - loss: 0.4492 - accuracy: 0.7432 - val_loss: 0.7686 - val_accuracy: 0.7480\n",
            "Epoch 402/500\n",
            "257/257 [==============================] - 0s 708us/step - loss: 0.4414 - accuracy: 0.7393 - val_loss: 0.7316 - val_accuracy: 0.7638\n",
            "Epoch 403/500\n",
            "257/257 [==============================] - 0s 701us/step - loss: 0.4301 - accuracy: 0.7510 - val_loss: 0.7150 - val_accuracy: 0.7323\n",
            "Epoch 404/500\n",
            "257/257 [==============================] - 0s 745us/step - loss: 0.4306 - accuracy: 0.7432 - val_loss: 0.6852 - val_accuracy: 0.7165\n",
            "Epoch 405/500\n",
            "257/257 [==============================] - 0s 719us/step - loss: 0.4230 - accuracy: 0.7393 - val_loss: 0.7766 - val_accuracy: 0.7087\n",
            "Epoch 406/500\n",
            "257/257 [==============================] - 0s 727us/step - loss: 0.4225 - accuracy: 0.7432 - val_loss: 0.8081 - val_accuracy: 0.7402\n",
            "Epoch 407/500\n",
            "257/257 [==============================] - 0s 707us/step - loss: 0.4315 - accuracy: 0.7393 - val_loss: 0.7634 - val_accuracy: 0.7559\n",
            "Epoch 408/500\n",
            "257/257 [==============================] - 0s 783us/step - loss: 0.4271 - accuracy: 0.7510 - val_loss: 0.7126 - val_accuracy: 0.7480\n",
            "Epoch 409/500\n",
            "257/257 [==============================] - 0s 744us/step - loss: 0.4188 - accuracy: 0.7432 - val_loss: 0.7428 - val_accuracy: 0.7480\n",
            "Epoch 410/500\n",
            "257/257 [==============================] - 0s 711us/step - loss: 0.4154 - accuracy: 0.7432 - val_loss: 0.7575 - val_accuracy: 0.7480\n",
            "Epoch 411/500\n",
            "257/257 [==============================] - 0s 694us/step - loss: 0.4510 - accuracy: 0.7393 - val_loss: 0.6988 - val_accuracy: 0.7559\n",
            "Epoch 412/500\n",
            "257/257 [==============================] - 0s 742us/step - loss: 0.4335 - accuracy: 0.7393 - val_loss: 0.6794 - val_accuracy: 0.7480\n",
            "Epoch 413/500\n",
            "257/257 [==============================] - 0s 692us/step - loss: 0.4203 - accuracy: 0.7432 - val_loss: 0.7832 - val_accuracy: 0.6850\n",
            "Epoch 414/500\n",
            "257/257 [==============================] - 0s 707us/step - loss: 0.4276 - accuracy: 0.7549 - val_loss: 0.7610 - val_accuracy: 0.6693\n",
            "Epoch 415/500\n",
            "257/257 [==============================] - 0s 720us/step - loss: 0.4182 - accuracy: 0.7665 - val_loss: 0.5462 - val_accuracy: 0.7008\n",
            "Epoch 416/500\n",
            "257/257 [==============================] - 0s 741us/step - loss: 0.4988 - accuracy: 0.7432 - val_loss: 0.5574 - val_accuracy: 0.6929\n",
            "Epoch 417/500\n",
            "257/257 [==============================] - 0s 802us/step - loss: 0.5303 - accuracy: 0.7354 - val_loss: 0.5558 - val_accuracy: 0.7165\n",
            "Epoch 418/500\n",
            "257/257 [==============================] - 0s 751us/step - loss: 0.5218 - accuracy: 0.7393 - val_loss: 0.5237 - val_accuracy: 0.7480\n",
            "Epoch 419/500\n",
            "257/257 [==============================] - 0s 789us/step - loss: 0.4882 - accuracy: 0.7471 - val_loss: 0.5257 - val_accuracy: 0.7795\n",
            "Epoch 420/500\n",
            "257/257 [==============================] - 0s 709us/step - loss: 0.4530 - accuracy: 0.7626 - val_loss: 0.6089 - val_accuracy: 0.7165\n",
            "Epoch 421/500\n",
            "257/257 [==============================] - 0s 872us/step - loss: 0.4378 - accuracy: 0.7276 - val_loss: 0.6879 - val_accuracy: 0.7165\n",
            "Epoch 422/500\n",
            "257/257 [==============================] - 0s 755us/step - loss: 0.4348 - accuracy: 0.7471 - val_loss: 0.7639 - val_accuracy: 0.7323\n",
            "Epoch 423/500\n",
            "257/257 [==============================] - 0s 701us/step - loss: 0.4322 - accuracy: 0.7549 - val_loss: 0.6351 - val_accuracy: 0.7087\n",
            "Epoch 424/500\n",
            "257/257 [==============================] - 0s 695us/step - loss: 0.4243 - accuracy: 0.7549 - val_loss: 0.6207 - val_accuracy: 0.7795\n",
            "Epoch 425/500\n",
            "257/257 [==============================] - 0s 706us/step - loss: 0.4493 - accuracy: 0.7510 - val_loss: 0.5901 - val_accuracy: 0.7480\n",
            "Epoch 426/500\n",
            "257/257 [==============================] - 0s 684us/step - loss: 0.4762 - accuracy: 0.7510 - val_loss: 0.6359 - val_accuracy: 0.7244\n",
            "Epoch 427/500\n",
            "257/257 [==============================] - 0s 723us/step - loss: 0.4602 - accuracy: 0.7510 - val_loss: 0.8000 - val_accuracy: 0.7244\n",
            "Epoch 428/500\n",
            "257/257 [==============================] - 0s 733us/step - loss: 0.4469 - accuracy: 0.7510 - val_loss: 0.9680 - val_accuracy: 0.7244\n",
            "Epoch 429/500\n",
            "257/257 [==============================] - 0s 748us/step - loss: 0.4435 - accuracy: 0.7393 - val_loss: 0.7368 - val_accuracy: 0.7402\n",
            "Epoch 430/500\n",
            "257/257 [==============================] - 0s 710us/step - loss: 0.4480 - accuracy: 0.7549 - val_loss: 0.7104 - val_accuracy: 0.7638\n",
            "Epoch 431/500\n",
            "257/257 [==============================] - 0s 666us/step - loss: 0.4461 - accuracy: 0.7471 - val_loss: 0.7319 - val_accuracy: 0.7559\n",
            "Epoch 432/500\n",
            "257/257 [==============================] - 0s 720us/step - loss: 0.4373 - accuracy: 0.7510 - val_loss: 0.7870 - val_accuracy: 0.7559\n",
            "Epoch 433/500\n",
            "257/257 [==============================] - 0s 704us/step - loss: 0.4358 - accuracy: 0.7549 - val_loss: 0.9190 - val_accuracy: 0.7480\n",
            "Epoch 434/500\n",
            "257/257 [==============================] - 0s 675us/step - loss: 0.4393 - accuracy: 0.7549 - val_loss: 0.8155 - val_accuracy: 0.7559\n",
            "Epoch 435/500\n",
            "257/257 [==============================] - 0s 683us/step - loss: 0.4211 - accuracy: 0.7315 - val_loss: 0.7578 - val_accuracy: 0.7402\n",
            "Epoch 436/500\n",
            "257/257 [==============================] - 0s 698us/step - loss: 0.4222 - accuracy: 0.7588 - val_loss: 0.7301 - val_accuracy: 0.7402\n",
            "Epoch 437/500\n",
            "257/257 [==============================] - 0s 737us/step - loss: 0.4115 - accuracy: 0.7626 - val_loss: 0.7601 - val_accuracy: 0.7402\n",
            "Epoch 438/500\n",
            "257/257 [==============================] - 0s 762us/step - loss: 0.3991 - accuracy: 0.7626 - val_loss: 0.7862 - val_accuracy: 0.7874\n",
            "Epoch 439/500\n",
            "257/257 [==============================] - 0s 767us/step - loss: 0.4013 - accuracy: 0.7743 - val_loss: 0.7916 - val_accuracy: 0.7795\n",
            "Epoch 440/500\n",
            "257/257 [==============================] - 0s 726us/step - loss: 0.3961 - accuracy: 0.7626 - val_loss: 1.0272 - val_accuracy: 0.7638\n",
            "Epoch 441/500\n",
            "257/257 [==============================] - 0s 717us/step - loss: 0.4022 - accuracy: 0.7665 - val_loss: 1.0770 - val_accuracy: 0.7638\n",
            "Epoch 442/500\n",
            "257/257 [==============================] - 0s 674us/step - loss: 0.3977 - accuracy: 0.7821 - val_loss: 1.1339 - val_accuracy: 0.7480\n",
            "Epoch 443/500\n",
            "257/257 [==============================] - 0s 684us/step - loss: 0.3997 - accuracy: 0.7743 - val_loss: 0.8609 - val_accuracy: 0.7559\n",
            "Epoch 444/500\n",
            "257/257 [==============================] - 0s 743us/step - loss: 0.3744 - accuracy: 0.7860 - val_loss: 0.8375 - val_accuracy: 0.7402\n",
            "Epoch 445/500\n",
            "257/257 [==============================] - 0s 679us/step - loss: 0.3856 - accuracy: 0.7821 - val_loss: 0.8671 - val_accuracy: 0.7638\n",
            "Epoch 446/500\n",
            "257/257 [==============================] - 0s 665us/step - loss: 0.3741 - accuracy: 0.7899 - val_loss: 1.0083 - val_accuracy: 0.7717\n",
            "Epoch 447/500\n",
            "257/257 [==============================] - 0s 690us/step - loss: 0.3817 - accuracy: 0.7743 - val_loss: 0.8710 - val_accuracy: 0.7717\n",
            "Epoch 448/500\n",
            "257/257 [==============================] - 0s 700us/step - loss: 0.3645 - accuracy: 0.7821 - val_loss: 0.8229 - val_accuracy: 0.7795\n",
            "Epoch 449/500\n",
            "257/257 [==============================] - 0s 670us/step - loss: 0.3695 - accuracy: 0.7821 - val_loss: 0.9232 - val_accuracy: 0.7795\n",
            "Epoch 450/500\n",
            "257/257 [==============================] - 0s 679us/step - loss: 0.3596 - accuracy: 0.7821 - val_loss: 0.8568 - val_accuracy: 0.7874\n",
            "Epoch 451/500\n",
            "257/257 [==============================] - 0s 731us/step - loss: 0.3508 - accuracy: 0.7821 - val_loss: 1.0714 - val_accuracy: 0.8031\n",
            "Epoch 452/500\n",
            "257/257 [==============================] - 0s 731us/step - loss: 0.3552 - accuracy: 0.7899 - val_loss: 0.9634 - val_accuracy: 0.7874\n",
            "Epoch 453/500\n",
            "257/257 [==============================] - 0s 813us/step - loss: 0.3481 - accuracy: 0.7899 - val_loss: 0.8706 - val_accuracy: 0.7323\n",
            "Epoch 454/500\n",
            "257/257 [==============================] - 0s 742us/step - loss: 0.4558 - accuracy: 0.7354 - val_loss: 0.6135 - val_accuracy: 0.6929\n",
            "Epoch 455/500\n",
            "257/257 [==============================] - 0s 764us/step - loss: 0.4706 - accuracy: 0.7082 - val_loss: 0.6284 - val_accuracy: 0.6693\n",
            "Epoch 456/500\n",
            "257/257 [==============================] - 0s 712us/step - loss: 0.4606 - accuracy: 0.7549 - val_loss: 0.8875 - val_accuracy: 0.6772\n",
            "Epoch 457/500\n",
            "257/257 [==============================] - 0s 714us/step - loss: 0.4338 - accuracy: 0.7665 - val_loss: 1.0261 - val_accuracy: 0.6850\n",
            "Epoch 458/500\n",
            "257/257 [==============================] - 0s 682us/step - loss: 0.4133 - accuracy: 0.7704 - val_loss: 1.0262 - val_accuracy: 0.6614\n",
            "Epoch 459/500\n",
            "257/257 [==============================] - 0s 705us/step - loss: 0.4049 - accuracy: 0.7782 - val_loss: 1.0028 - val_accuracy: 0.6614\n",
            "Epoch 460/500\n",
            "257/257 [==============================] - 0s 681us/step - loss: 0.3989 - accuracy: 0.7860 - val_loss: 1.0660 - val_accuracy: 0.6693\n",
            "Epoch 461/500\n",
            "257/257 [==============================] - 0s 704us/step - loss: 0.3954 - accuracy: 0.7977 - val_loss: 0.6978 - val_accuracy: 0.5118\n",
            "Epoch 462/500\n",
            "257/257 [==============================] - 0s 716us/step - loss: 0.5554 - accuracy: 0.6265 - val_loss: 0.6284 - val_accuracy: 0.7323\n",
            "Epoch 463/500\n",
            "257/257 [==============================] - 0s 711us/step - loss: 0.5185 - accuracy: 0.7588 - val_loss: 0.6126 - val_accuracy: 0.7165\n",
            "Epoch 464/500\n",
            "257/257 [==============================] - 0s 774us/step - loss: 0.4981 - accuracy: 0.7549 - val_loss: 0.6070 - val_accuracy: 0.7402\n",
            "Epoch 465/500\n",
            "257/257 [==============================] - 0s 749us/step - loss: 0.4483 - accuracy: 0.7588 - val_loss: 0.6190 - val_accuracy: 0.7402\n",
            "Epoch 466/500\n",
            "257/257 [==============================] - 0s 832us/step - loss: 0.4186 - accuracy: 0.7743 - val_loss: 0.7253 - val_accuracy: 0.7717\n",
            "Epoch 467/500\n",
            "257/257 [==============================] - 0s 726us/step - loss: 0.3897 - accuracy: 0.7821 - val_loss: 0.7055 - val_accuracy: 0.7087\n",
            "Epoch 468/500\n",
            "257/257 [==============================] - 0s 694us/step - loss: 0.5932 - accuracy: 0.6459 - val_loss: 0.5814 - val_accuracy: 0.7008\n",
            "Epoch 469/500\n",
            "257/257 [==============================] - 0s 714us/step - loss: 0.5344 - accuracy: 0.6265 - val_loss: 0.5616 - val_accuracy: 0.7008\n",
            "Epoch 470/500\n",
            "257/257 [==============================] - 0s 683us/step - loss: 0.5374 - accuracy: 0.6304 - val_loss: 0.5607 - val_accuracy: 0.6929\n",
            "Epoch 471/500\n",
            "257/257 [==============================] - 0s 715us/step - loss: 0.5388 - accuracy: 0.6304 - val_loss: 0.5601 - val_accuracy: 0.6929\n",
            "Epoch 472/500\n",
            "257/257 [==============================] - 0s 684us/step - loss: 0.5342 - accuracy: 0.6304 - val_loss: 0.5770 - val_accuracy: 0.6929\n",
            "Epoch 473/500\n",
            "257/257 [==============================] - 0s 682us/step - loss: 0.5283 - accuracy: 0.6304 - val_loss: 0.5671 - val_accuracy: 0.7559\n",
            "Epoch 474/500\n",
            "257/257 [==============================] - 0s 702us/step - loss: 0.5205 - accuracy: 0.7237 - val_loss: 0.5821 - val_accuracy: 0.7087\n",
            "Epoch 475/500\n",
            "257/257 [==============================] - 0s 717us/step - loss: 0.5163 - accuracy: 0.7237 - val_loss: 0.5982 - val_accuracy: 0.7008\n",
            "Epoch 476/500\n",
            "257/257 [==============================] - 0s 683us/step - loss: 0.5114 - accuracy: 0.7315 - val_loss: 0.6093 - val_accuracy: 0.7323\n",
            "Epoch 477/500\n",
            "257/257 [==============================] - 0s 695us/step - loss: 0.5050 - accuracy: 0.7393 - val_loss: 0.6220 - val_accuracy: 0.7323\n",
            "Epoch 478/500\n",
            "257/257 [==============================] - 0s 742us/step - loss: 0.8911 - accuracy: 0.7549 - val_loss: 0.6503 - val_accuracy: 0.7717\n",
            "Epoch 479/500\n",
            "257/257 [==============================] - 0s 702us/step - loss: 0.5106 - accuracy: 0.7121 - val_loss: 0.6220 - val_accuracy: 0.7638\n",
            "Epoch 480/500\n",
            "257/257 [==============================] - 0s 712us/step - loss: 0.5034 - accuracy: 0.7276 - val_loss: 0.5974 - val_accuracy: 0.7717\n",
            "Epoch 481/500\n",
            "257/257 [==============================] - 0s 675us/step - loss: 0.4944 - accuracy: 0.7315 - val_loss: 0.5881 - val_accuracy: 0.7559\n",
            "Epoch 482/500\n",
            "257/257 [==============================] - 0s 702us/step - loss: 0.4907 - accuracy: 0.7432 - val_loss: 0.5890 - val_accuracy: 0.7480\n",
            "Epoch 483/500\n",
            "257/257 [==============================] - 0s 710us/step - loss: 0.4875 - accuracy: 0.7432 - val_loss: 0.5671 - val_accuracy: 0.6772\n",
            "Epoch 484/500\n",
            "257/257 [==============================] - 0s 723us/step - loss: 0.5381 - accuracy: 0.7043 - val_loss: 0.5688 - val_accuracy: 0.6693\n",
            "Epoch 485/500\n",
            "257/257 [==============================] - 0s 692us/step - loss: 0.5161 - accuracy: 0.7276 - val_loss: 0.5663 - val_accuracy: 0.7244\n",
            "Epoch 486/500\n",
            "257/257 [==============================] - 0s 680us/step - loss: 0.4917 - accuracy: 0.7510 - val_loss: 0.6253 - val_accuracy: 0.7559\n",
            "Epoch 487/500\n",
            "257/257 [==============================] - 0s 687us/step - loss: 0.5006 - accuracy: 0.7315 - val_loss: 0.6855 - val_accuracy: 0.7323\n",
            "Epoch 488/500\n",
            "257/257 [==============================] - 0s 713us/step - loss: 0.5140 - accuracy: 0.7198 - val_loss: 0.7249 - val_accuracy: 0.7480\n",
            "Epoch 489/500\n",
            "257/257 [==============================] - 0s 690us/step - loss: 0.5173 - accuracy: 0.7237 - val_loss: 0.5997 - val_accuracy: 0.7244\n",
            "Epoch 490/500\n",
            "257/257 [==============================] - 0s 710us/step - loss: 0.4998 - accuracy: 0.7237 - val_loss: 0.6155 - val_accuracy: 0.7008\n",
            "Epoch 491/500\n",
            "257/257 [==============================] - 0s 680us/step - loss: 0.4932 - accuracy: 0.7471 - val_loss: 0.6140 - val_accuracy: 0.7402\n",
            "Epoch 492/500\n",
            "257/257 [==============================] - 0s 713us/step - loss: 0.4858 - accuracy: 0.7510 - val_loss: 0.6564 - val_accuracy: 0.7402\n",
            "Epoch 493/500\n",
            "257/257 [==============================] - 0s 679us/step - loss: 0.4833 - accuracy: 0.7432 - val_loss: 0.6657 - val_accuracy: 0.7874\n",
            "Epoch 494/500\n",
            "257/257 [==============================] - 0s 707us/step - loss: 0.4862 - accuracy: 0.7393 - val_loss: 0.6293 - val_accuracy: 0.7402\n",
            "Epoch 495/500\n",
            "257/257 [==============================] - 0s 704us/step - loss: 0.4706 - accuracy: 0.7588 - val_loss: 0.6416 - val_accuracy: 0.6929\n",
            "Epoch 496/500\n",
            "257/257 [==============================] - 0s 688us/step - loss: 0.4719 - accuracy: 0.7393 - val_loss: 0.6902 - val_accuracy: 0.6614\n",
            "Epoch 497/500\n",
            "257/257 [==============================] - 0s 725us/step - loss: 0.5809 - accuracy: 0.6965 - val_loss: 0.6774 - val_accuracy: 0.7165\n",
            "Epoch 498/500\n",
            "257/257 [==============================] - 0s 690us/step - loss: 0.4797 - accuracy: 0.7432 - val_loss: 0.7974 - val_accuracy: 0.7559\n",
            "Epoch 499/500\n",
            "257/257 [==============================] - 0s 730us/step - loss: 0.5063 - accuracy: 0.7276 - val_loss: 0.7141 - val_accuracy: 0.7638\n",
            "Epoch 500/500\n",
            "257/257 [==============================] - 0s 694us/step - loss: 0.4860 - accuracy: 0.7276 - val_loss: 0.6650 - val_accuracy: 0.7480\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GG0HOo2G09i8",
        "colab_type": "text"
      },
      "source": [
        "## Evaluate LSTM ##"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VY3k5jUbSgDL",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 923
        },
        "outputId": "706d1a39-5f7c-441a-8a83-cda22a0811e2"
      },
      "source": [
        "plt.plot(history.history['accuracy'], label = 'train')\n",
        "plt.plot(history.history['val_accuracy'], label='validation')\n",
        "plt.title('model accuracy')\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'validation'], loc = 'upper left')\n",
        "plt.show()\n",
        "plt.plot(history.history['loss'])\n",
        "plt.plot(history.history['val_loss'])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'validation'],loc = 'upper left')\n",
        "plt.show()\n",
        "\n",
        "y_pred2=model.predict(X_tst3)\n",
        "y_pred2=(y_pred2>.5).astype(int)\n",
        "\n",
        "# todo: use boot-strap estimation\n",
        "probs = model.predict_proba(X_tst3)\n",
        "probs = probs[:,0]\n",
        "auc = roc_auc_score(y_test, probs)\n",
        "print('AUC: %.3f' % auc)\n",
        "fpr, tpr, thresholds = roc_curve(y_test, probs)\n",
        "pyplot.plot([0, 1], [0, 1], linestyle = '-')\n",
        "pyplot.plot(fpr, tpr, marker = '.')\n",
        "pyplot.show()\n",
        "\n",
        "\n",
        "cmat = metrics.confusion_matrix(y_test, y_pred2)\n",
        "print( cmat )\n",
        "print(\"Accuracy:\",metrics.accuracy_score(y_test, y_pred2))\n",
        "print(\"Precision:\",metrics.precision_score(y_test, y_pred2, pos_label=1)) \n",
        "print(\"Recall:\",metrics.recall_score(y_test, y_pred2, pos_label=1))\n"
      ],
      "execution_count": 304,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "AUC: 0.783\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "[[198  55]\n",
            " [ 51  80]]\n",
            "Accuracy: 0.7239583333333334\n",
            "Precision: 0.5925925925925926\n",
            "Recall: 0.6106870229007634\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CDNmUanwSwLv",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 367
        },
        "outputId": "ec2932da-951d-4649-9441-f063bbbb3ee6"
      },
      "source": [
        "model.load_weights('best.h5')\n",
        "\n",
        "y_pred2=model.predict(X_tst3)\n",
        "y_pred2=(y_pred2>.5).astype(int)\n",
        "\n",
        "# todo: use boot-strap estimation\n",
        "probs = model.predict_proba(X_tst3)\n",
        "probs = probs[:,0]\n",
        "auc = roc_auc_score(y_test, probs)\n",
        "print('AUC: %.3f' % auc)\n",
        "fpr, tpr, thresholds = roc_curve(y_test, probs)\n",
        "pyplot.plot([0, 1], [0, 1], linestyle = '-')\n",
        "pyplot.plot(fpr, tpr, marker = '.')\n",
        "pyplot.show()\n",
        "\n",
        "cmat = metrics.confusion_matrix(y_test, y_pred2)\n",
        "print( cmat )\n",
        "print(\"Accuracy:\",metrics.accuracy_score(y_test, y_pred2))\n",
        "print(\"Precision:\",metrics.precision_score(y_test, y_pred2, pos_label=1)) \n",
        "print(\"Recall:\",metrics.recall_score(y_test, y_pred2, pos_label=1))"
      ],
      "execution_count": 305,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "AUC: 0.770\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "[[206  47]\n",
            " [ 63  68]]\n",
            "Accuracy: 0.7135416666666666\n",
            "Precision: 0.591304347826087\n",
            "Recall: 0.5190839694656488\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "u7Ft9sZ41AW8",
        "colab_type": "text"
      },
      "source": [
        "## Train MLP ##"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dPQBxrhrzy-I",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "12e53c9f-4d9f-429b-bee1-060e5087e06b"
      },
      "source": [
        "model = Sequential()\n",
        "model.add(Dense(64, input_dim = (8), activation = 'relu'))\n",
        "model.add(Dense(64, activation = 'relu'))\n",
        "model.add(Dense(1, activation = 'sigmoid'))\n",
        "\n",
        "from keras.optimizers import Nadam\n",
        "\n",
        "model.compile(metrics = ['accuracy'], optimizer = Nadam(lr = 0.002, schedule_decay = 0.004), loss ='binary_crossentropy')\n",
        "model.summary()\n",
        "\n",
        "model.fit(X_train3, y_train, validation_split = 0.33, initial_epoch=0, epochs = 500, batch_size = 64, verbose = 1,callbacks=C)\n"
      ],
      "execution_count": 317,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"sequential_30\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "dense_121 (Dense)            (None, 64)                576       \n",
            "_________________________________________________________________\n",
            "dense_122 (Dense)            (None, 64)                4160      \n",
            "_________________________________________________________________\n",
            "dense_123 (Dense)            (None, 1)                 65        \n",
            "=================================================================\n",
            "Total params: 4,801\n",
            "Trainable params: 4,801\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n",
            "Train on 257 samples, validate on 127 samples\n",
            "Epoch 1/500\n",
            "257/257 [==============================] - 0s 659us/step - loss: 0.6868 - accuracy: 0.5447 - val_loss: 0.6529 - val_accuracy: 0.7008\n",
            "Epoch 2/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.6534 - accuracy: 0.6615 - val_loss: 0.6089 - val_accuracy: 0.7165\n",
            "Epoch 3/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.6242 - accuracy: 0.7237 - val_loss: 0.5722 - val_accuracy: 0.7559\n",
            "Epoch 4/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.6024 - accuracy: 0.7354 - val_loss: 0.5600 - val_accuracy: 0.7638\n",
            "Epoch 5/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.5849 - accuracy: 0.7121 - val_loss: 0.5638 - val_accuracy: 0.7165\n",
            "Epoch 6/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.5778 - accuracy: 0.6926 - val_loss: 0.5446 - val_accuracy: 0.7244\n",
            "Epoch 7/500\n",
            "257/257 [==============================] - 0s 59us/step - loss: 0.5597 - accuracy: 0.7276 - val_loss: 0.5275 - val_accuracy: 0.7323\n",
            "Epoch 8/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.5429 - accuracy: 0.7549 - val_loss: 0.5111 - val_accuracy: 0.7323\n",
            "Epoch 9/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.5248 - accuracy: 0.7588 - val_loss: 0.5033 - val_accuracy: 0.7244\n",
            "Epoch 10/500\n",
            "257/257 [==============================] - 0s 47us/step - loss: 0.5128 - accuracy: 0.7626 - val_loss: 0.4909 - val_accuracy: 0.7717\n",
            "Epoch 11/500\n",
            "257/257 [==============================] - 0s 47us/step - loss: 0.5003 - accuracy: 0.7549 - val_loss: 0.4800 - val_accuracy: 0.7874\n",
            "Epoch 12/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.4929 - accuracy: 0.7743 - val_loss: 0.4786 - val_accuracy: 0.7717\n",
            "Epoch 13/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.4864 - accuracy: 0.7626 - val_loss: 0.4800 - val_accuracy: 0.7717\n",
            "Epoch 14/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.4817 - accuracy: 0.7704 - val_loss: 0.4802 - val_accuracy: 0.7717\n",
            "Epoch 15/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.4758 - accuracy: 0.7626 - val_loss: 0.4803 - val_accuracy: 0.7795\n",
            "Epoch 16/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.4728 - accuracy: 0.7704 - val_loss: 0.4865 - val_accuracy: 0.7717\n",
            "Epoch 17/500\n",
            "257/257 [==============================] - 0s 47us/step - loss: 0.4727 - accuracy: 0.7510 - val_loss: 0.4911 - val_accuracy: 0.7717\n",
            "Epoch 18/500\n",
            "257/257 [==============================] - 0s 47us/step - loss: 0.4704 - accuracy: 0.7549 - val_loss: 0.4825 - val_accuracy: 0.7874\n",
            "Epoch 19/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.4579 - accuracy: 0.7626 - val_loss: 0.4764 - val_accuracy: 0.7717\n",
            "Epoch 20/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.4507 - accuracy: 0.7743 - val_loss: 0.4844 - val_accuracy: 0.7874\n",
            "Epoch 21/500\n",
            "257/257 [==============================] - 0s 65us/step - loss: 0.4481 - accuracy: 0.7588 - val_loss: 0.4804 - val_accuracy: 0.7717\n",
            "Epoch 22/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.4529 - accuracy: 0.7782 - val_loss: 0.4830 - val_accuracy: 0.7874\n",
            "Epoch 23/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.4523 - accuracy: 0.7899 - val_loss: 0.4765 - val_accuracy: 0.7874\n",
            "Epoch 24/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.4408 - accuracy: 0.7782 - val_loss: 0.4818 - val_accuracy: 0.7953\n",
            "Epoch 25/500\n",
            "257/257 [==============================] - 0s 61us/step - loss: 0.4369 - accuracy: 0.7938 - val_loss: 0.4843 - val_accuracy: 0.7795\n",
            "Epoch 26/500\n",
            "257/257 [==============================] - 0s 46us/step - loss: 0.4367 - accuracy: 0.7899 - val_loss: 0.4853 - val_accuracy: 0.7874\n",
            "Epoch 27/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.4462 - accuracy: 0.7821 - val_loss: 0.4873 - val_accuracy: 0.7874\n",
            "Epoch 28/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.4536 - accuracy: 0.7782 - val_loss: 0.4874 - val_accuracy: 0.7795\n",
            "Epoch 29/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.4497 - accuracy: 0.7821 - val_loss: 0.4912 - val_accuracy: 0.7953\n",
            "Epoch 30/500\n",
            "257/257 [==============================] - 0s 60us/step - loss: 0.4383 - accuracy: 0.7860 - val_loss: 0.4776 - val_accuracy: 0.8189\n",
            "Epoch 31/500\n",
            "257/257 [==============================] - 0s 47us/step - loss: 0.4216 - accuracy: 0.8016 - val_loss: 0.4792 - val_accuracy: 0.7953\n",
            "Epoch 32/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.4205 - accuracy: 0.8132 - val_loss: 0.4809 - val_accuracy: 0.7874\n",
            "Epoch 33/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.4166 - accuracy: 0.8210 - val_loss: 0.4926 - val_accuracy: 0.8031\n",
            "Epoch 34/500\n",
            "257/257 [==============================] - 0s 45us/step - loss: 0.4279 - accuracy: 0.7860 - val_loss: 0.5057 - val_accuracy: 0.7638\n",
            "Epoch 35/500\n",
            "257/257 [==============================] - 0s 45us/step - loss: 0.4362 - accuracy: 0.7782 - val_loss: 0.4960 - val_accuracy: 0.8031\n",
            "Epoch 36/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.4209 - accuracy: 0.7899 - val_loss: 0.4934 - val_accuracy: 0.7953\n",
            "Epoch 37/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.4148 - accuracy: 0.7899 - val_loss: 0.4922 - val_accuracy: 0.7953\n",
            "Epoch 38/500\n",
            "257/257 [==============================] - 0s 62us/step - loss: 0.4096 - accuracy: 0.8132 - val_loss: 0.4952 - val_accuracy: 0.8031\n",
            "Epoch 39/500\n",
            "257/257 [==============================] - 0s 46us/step - loss: 0.4070 - accuracy: 0.8132 - val_loss: 0.5089 - val_accuracy: 0.7874\n",
            "Epoch 40/500\n",
            "257/257 [==============================] - 0s 46us/step - loss: 0.4081 - accuracy: 0.7977 - val_loss: 0.5079 - val_accuracy: 0.7874\n",
            "Epoch 41/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.4036 - accuracy: 0.8132 - val_loss: 0.5044 - val_accuracy: 0.7953\n",
            "Epoch 42/500\n",
            "257/257 [==============================] - 0s 46us/step - loss: 0.3984 - accuracy: 0.8249 - val_loss: 0.5123 - val_accuracy: 0.7795\n",
            "Epoch 43/500\n",
            "257/257 [==============================] - 0s 47us/step - loss: 0.4016 - accuracy: 0.8054 - val_loss: 0.5073 - val_accuracy: 0.7874\n",
            "Epoch 44/500\n",
            "257/257 [==============================] - 0s 47us/step - loss: 0.3957 - accuracy: 0.8288 - val_loss: 0.5040 - val_accuracy: 0.7953\n",
            "Epoch 45/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.3944 - accuracy: 0.8054 - val_loss: 0.5073 - val_accuracy: 0.7874\n",
            "Epoch 46/500\n",
            "257/257 [==============================] - 0s 61us/step - loss: 0.4025 - accuracy: 0.8132 - val_loss: 0.5041 - val_accuracy: 0.8110\n",
            "Epoch 47/500\n",
            "257/257 [==============================] - 0s 67us/step - loss: 0.3938 - accuracy: 0.8054 - val_loss: 0.5045 - val_accuracy: 0.8031\n",
            "Epoch 48/500\n",
            "257/257 [==============================] - 0s 68us/step - loss: 0.3923 - accuracy: 0.8132 - val_loss: 0.5050 - val_accuracy: 0.8031\n",
            "Epoch 49/500\n",
            "257/257 [==============================] - 0s 68us/step - loss: 0.3895 - accuracy: 0.8171 - val_loss: 0.5093 - val_accuracy: 0.7953\n",
            "Epoch 50/500\n",
            "257/257 [==============================] - 0s 60us/step - loss: 0.3848 - accuracy: 0.8210 - val_loss: 0.5194 - val_accuracy: 0.7874\n",
            "Epoch 51/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.3869 - accuracy: 0.8132 - val_loss: 0.5139 - val_accuracy: 0.7953\n",
            "Epoch 52/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.3852 - accuracy: 0.8249 - val_loss: 0.5142 - val_accuracy: 0.7874\n",
            "Epoch 53/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.3828 - accuracy: 0.8288 - val_loss: 0.5130 - val_accuracy: 0.7874\n",
            "Epoch 54/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.3774 - accuracy: 0.8366 - val_loss: 0.5189 - val_accuracy: 0.7795\n",
            "Epoch 55/500\n",
            "257/257 [==============================] - 0s 46us/step - loss: 0.3737 - accuracy: 0.8249 - val_loss: 0.5372 - val_accuracy: 0.7717\n",
            "Epoch 56/500\n",
            "257/257 [==============================] - 0s 46us/step - loss: 0.3826 - accuracy: 0.7977 - val_loss: 0.5193 - val_accuracy: 0.7795\n",
            "Epoch 57/500\n",
            "257/257 [==============================] - 0s 62us/step - loss: 0.3751 - accuracy: 0.8171 - val_loss: 0.5119 - val_accuracy: 0.7874\n",
            "Epoch 58/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.3748 - accuracy: 0.7860 - val_loss: 0.5100 - val_accuracy: 0.7559\n",
            "Epoch 59/500\n",
            "257/257 [==============================] - 0s 47us/step - loss: 0.3703 - accuracy: 0.8132 - val_loss: 0.5177 - val_accuracy: 0.7717\n",
            "Epoch 60/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.3712 - accuracy: 0.8171 - val_loss: 0.5218 - val_accuracy: 0.7717\n",
            "Epoch 61/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.3710 - accuracy: 0.8132 - val_loss: 0.5248 - val_accuracy: 0.7795\n",
            "Epoch 62/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.3669 - accuracy: 0.8132 - val_loss: 0.5391 - val_accuracy: 0.7795\n",
            "Epoch 63/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.3744 - accuracy: 0.8054 - val_loss: 0.5236 - val_accuracy: 0.7717\n",
            "Epoch 64/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.3718 - accuracy: 0.8288 - val_loss: 0.5350 - val_accuracy: 0.7717\n",
            "Epoch 65/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.3720 - accuracy: 0.8171 - val_loss: 0.5355 - val_accuracy: 0.7795\n",
            "Epoch 66/500\n",
            "257/257 [==============================] - 0s 61us/step - loss: 0.3682 - accuracy: 0.8171 - val_loss: 0.5323 - val_accuracy: 0.7638\n",
            "Epoch 67/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.3629 - accuracy: 0.8288 - val_loss: 0.5299 - val_accuracy: 0.7480\n",
            "Epoch 68/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.3577 - accuracy: 0.8405 - val_loss: 0.5377 - val_accuracy: 0.7244\n",
            "Epoch 69/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.3585 - accuracy: 0.8210 - val_loss: 0.5380 - val_accuracy: 0.7323\n",
            "Epoch 70/500\n",
            "257/257 [==============================] - 0s 78us/step - loss: 0.3534 - accuracy: 0.8327 - val_loss: 0.5465 - val_accuracy: 0.7244\n",
            "Epoch 71/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.3549 - accuracy: 0.8249 - val_loss: 0.5621 - val_accuracy: 0.7559\n",
            "Epoch 72/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.3920 - accuracy: 0.8288 - val_loss: 0.5589 - val_accuracy: 0.7480\n",
            "Epoch 73/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.3812 - accuracy: 0.8249 - val_loss: 0.5626 - val_accuracy: 0.7559\n",
            "Epoch 74/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.3776 - accuracy: 0.8288 - val_loss: 0.5600 - val_accuracy: 0.7717\n",
            "Epoch 75/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.3667 - accuracy: 0.8327 - val_loss: 0.5575 - val_accuracy: 0.7874\n",
            "Epoch 76/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.3602 - accuracy: 0.8405 - val_loss: 0.5548 - val_accuracy: 0.7795\n",
            "Epoch 77/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.3520 - accuracy: 0.8288 - val_loss: 0.5515 - val_accuracy: 0.7717\n",
            "Epoch 78/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.3460 - accuracy: 0.8288 - val_loss: 0.5456 - val_accuracy: 0.7559\n",
            "Epoch 79/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.3511 - accuracy: 0.8249 - val_loss: 0.5502 - val_accuracy: 0.7717\n",
            "Epoch 80/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.3492 - accuracy: 0.8327 - val_loss: 0.5505 - val_accuracy: 0.7717\n",
            "Epoch 81/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.3472 - accuracy: 0.8327 - val_loss: 0.5570 - val_accuracy: 0.7717\n",
            "Epoch 82/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.3528 - accuracy: 0.8366 - val_loss: 0.5475 - val_accuracy: 0.7559\n",
            "Epoch 83/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.3431 - accuracy: 0.8366 - val_loss: 0.5446 - val_accuracy: 0.7480\n",
            "Epoch 84/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.3381 - accuracy: 0.8366 - val_loss: 0.5572 - val_accuracy: 0.7244\n",
            "Epoch 85/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.3628 - accuracy: 0.8210 - val_loss: 0.5573 - val_accuracy: 0.7244\n",
            "Epoch 86/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.3540 - accuracy: 0.8171 - val_loss: 0.5524 - val_accuracy: 0.7323\n",
            "Epoch 87/500\n",
            "257/257 [==============================] - 0s 61us/step - loss: 0.3387 - accuracy: 0.8327 - val_loss: 0.5529 - val_accuracy: 0.7323\n",
            "Epoch 88/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.3343 - accuracy: 0.8405 - val_loss: 0.5784 - val_accuracy: 0.6929\n",
            "Epoch 89/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.3593 - accuracy: 0.8210 - val_loss: 0.5693 - val_accuracy: 0.7323\n",
            "Epoch 90/500\n",
            "257/257 [==============================] - 0s 80us/step - loss: 0.3431 - accuracy: 0.8482 - val_loss: 0.5710 - val_accuracy: 0.7402\n",
            "Epoch 91/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.3407 - accuracy: 0.8599 - val_loss: 0.5718 - val_accuracy: 0.7402\n",
            "Epoch 92/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.3421 - accuracy: 0.8482 - val_loss: 0.5721 - val_accuracy: 0.7402\n",
            "Epoch 93/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.3390 - accuracy: 0.8560 - val_loss: 0.5701 - val_accuracy: 0.7402\n",
            "Epoch 94/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.3333 - accuracy: 0.8560 - val_loss: 0.5774 - val_accuracy: 0.7402\n",
            "Epoch 95/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.3340 - accuracy: 0.8482 - val_loss: 0.5806 - val_accuracy: 0.7402\n",
            "Epoch 96/500\n",
            "257/257 [==============================] - 0s 66us/step - loss: 0.3322 - accuracy: 0.8482 - val_loss: 0.5677 - val_accuracy: 0.7323\n",
            "Epoch 97/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.3298 - accuracy: 0.8444 - val_loss: 0.5765 - val_accuracy: 0.7165\n",
            "Epoch 98/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.3269 - accuracy: 0.8521 - val_loss: 0.5808 - val_accuracy: 0.7165\n",
            "Epoch 99/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.3304 - accuracy: 0.8482 - val_loss: 0.5937 - val_accuracy: 0.7323\n",
            "Epoch 100/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.3502 - accuracy: 0.8171 - val_loss: 0.5584 - val_accuracy: 0.7244\n",
            "Epoch 101/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.3328 - accuracy: 0.8677 - val_loss: 0.5816 - val_accuracy: 0.7323\n",
            "Epoch 102/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.3341 - accuracy: 0.8482 - val_loss: 0.5579 - val_accuracy: 0.7323\n",
            "Epoch 103/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.3466 - accuracy: 0.8366 - val_loss: 0.5586 - val_accuracy: 0.7323\n",
            "Epoch 104/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.3401 - accuracy: 0.8405 - val_loss: 0.6325 - val_accuracy: 0.7008\n",
            "Epoch 105/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.3717 - accuracy: 0.8327 - val_loss: 0.6169 - val_accuracy: 0.7244\n",
            "Epoch 106/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.3407 - accuracy: 0.8288 - val_loss: 0.6062 - val_accuracy: 0.7244\n",
            "Epoch 107/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.3260 - accuracy: 0.8444 - val_loss: 0.6018 - val_accuracy: 0.7244\n",
            "Epoch 108/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.3169 - accuracy: 0.8560 - val_loss: 0.5945 - val_accuracy: 0.7244\n",
            "Epoch 109/500\n",
            "257/257 [==============================] - 0s 59us/step - loss: 0.3097 - accuracy: 0.8560 - val_loss: 0.5961 - val_accuracy: 0.7244\n",
            "Epoch 110/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.3062 - accuracy: 0.8677 - val_loss: 0.6042 - val_accuracy: 0.7244\n",
            "Epoch 111/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.3082 - accuracy: 0.8521 - val_loss: 0.6131 - val_accuracy: 0.7402\n",
            "Epoch 112/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.3080 - accuracy: 0.8599 - val_loss: 0.6053 - val_accuracy: 0.7480\n",
            "Epoch 113/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.3021 - accuracy: 0.8599 - val_loss: 0.5925 - val_accuracy: 0.7244\n",
            "Epoch 114/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.3120 - accuracy: 0.8560 - val_loss: 0.5954 - val_accuracy: 0.7165\n",
            "Epoch 115/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.3071 - accuracy: 0.8482 - val_loss: 0.5976 - val_accuracy: 0.7244\n",
            "Epoch 116/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.3009 - accuracy: 0.8638 - val_loss: 0.6057 - val_accuracy: 0.7244\n",
            "Epoch 117/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.3007 - accuracy: 0.8599 - val_loss: 0.6072 - val_accuracy: 0.7323\n",
            "Epoch 118/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.2982 - accuracy: 0.8560 - val_loss: 0.6014 - val_accuracy: 0.7323\n",
            "Epoch 119/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.3030 - accuracy: 0.8560 - val_loss: 0.6019 - val_accuracy: 0.7402\n",
            "Epoch 120/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.3010 - accuracy: 0.8599 - val_loss: 0.5985 - val_accuracy: 0.7402\n",
            "Epoch 121/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.2972 - accuracy: 0.8716 - val_loss: 0.6301 - val_accuracy: 0.7244\n",
            "Epoch 122/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.3088 - accuracy: 0.8638 - val_loss: 0.6378 - val_accuracy: 0.7244\n",
            "Epoch 123/500\n",
            "257/257 [==============================] - 0s 59us/step - loss: 0.3068 - accuracy: 0.8638 - val_loss: 0.6365 - val_accuracy: 0.7244\n",
            "Epoch 124/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.3034 - accuracy: 0.8482 - val_loss: 0.6397 - val_accuracy: 0.7244\n",
            "Epoch 125/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.3184 - accuracy: 0.8405 - val_loss: 0.6506 - val_accuracy: 0.7008\n",
            "Epoch 126/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.3159 - accuracy: 0.8405 - val_loss: 0.6368 - val_accuracy: 0.7087\n",
            "Epoch 127/500\n",
            "257/257 [==============================] - 0s 47us/step - loss: 0.3046 - accuracy: 0.8444 - val_loss: 0.6242 - val_accuracy: 0.7087\n",
            "Epoch 128/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.2926 - accuracy: 0.8599 - val_loss: 0.6149 - val_accuracy: 0.7244\n",
            "Epoch 129/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.2853 - accuracy: 0.8638 - val_loss: 0.6148 - val_accuracy: 0.7244\n",
            "Epoch 130/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.2801 - accuracy: 0.8677 - val_loss: 0.6127 - val_accuracy: 0.7559\n",
            "Epoch 131/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.2801 - accuracy: 0.8755 - val_loss: 0.6157 - val_accuracy: 0.7323\n",
            "Epoch 132/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.2820 - accuracy: 0.8755 - val_loss: 0.6239 - val_accuracy: 0.7559\n",
            "Epoch 133/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.2868 - accuracy: 0.8599 - val_loss: 0.6085 - val_accuracy: 0.7480\n",
            "Epoch 134/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.2835 - accuracy: 0.8794 - val_loss: 0.6064 - val_accuracy: 0.7323\n",
            "Epoch 135/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.2783 - accuracy: 0.8677 - val_loss: 0.6053 - val_accuracy: 0.7323\n",
            "Epoch 136/500\n",
            "257/257 [==============================] - 0s 68us/step - loss: 0.2869 - accuracy: 0.8794 - val_loss: 0.6042 - val_accuracy: 0.7402\n",
            "Epoch 137/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.2805 - accuracy: 0.8794 - val_loss: 0.5988 - val_accuracy: 0.7323\n",
            "Epoch 138/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.2775 - accuracy: 0.8794 - val_loss: 0.6035 - val_accuracy: 0.7402\n",
            "Epoch 139/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.2769 - accuracy: 0.8716 - val_loss: 0.6052 - val_accuracy: 0.7323\n",
            "Epoch 140/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.2720 - accuracy: 0.8716 - val_loss: 0.6089 - val_accuracy: 0.7244\n",
            "Epoch 141/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.2676 - accuracy: 0.8794 - val_loss: 0.6117 - val_accuracy: 0.7323\n",
            "Epoch 142/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.2650 - accuracy: 0.8794 - val_loss: 0.6140 - val_accuracy: 0.7323\n",
            "Epoch 143/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.2616 - accuracy: 0.8794 - val_loss: 0.6186 - val_accuracy: 0.7165\n",
            "Epoch 144/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.2623 - accuracy: 0.8794 - val_loss: 0.6208 - val_accuracy: 0.7165\n",
            "Epoch 145/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.2601 - accuracy: 0.8833 - val_loss: 0.6220 - val_accuracy: 0.7244\n",
            "Epoch 146/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.2571 - accuracy: 0.8872 - val_loss: 0.6379 - val_accuracy: 0.7559\n",
            "Epoch 147/500\n",
            "257/257 [==============================] - 0s 62us/step - loss: 0.2632 - accuracy: 0.8833 - val_loss: 0.6358 - val_accuracy: 0.7559\n",
            "Epoch 148/500\n",
            "257/257 [==============================] - 0s 59us/step - loss: 0.2589 - accuracy: 0.8833 - val_loss: 0.6338 - val_accuracy: 0.7480\n",
            "Epoch 149/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.2582 - accuracy: 0.8949 - val_loss: 0.6410 - val_accuracy: 0.7559\n",
            "Epoch 150/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.2558 - accuracy: 0.8988 - val_loss: 0.6443 - val_accuracy: 0.7480\n",
            "Epoch 151/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.2538 - accuracy: 0.8949 - val_loss: 0.6401 - val_accuracy: 0.7480\n",
            "Epoch 152/500\n",
            "257/257 [==============================] - 0s 69us/step - loss: 0.2511 - accuracy: 0.8988 - val_loss: 0.6396 - val_accuracy: 0.7480\n",
            "Epoch 153/500\n",
            "257/257 [==============================] - 0s 61us/step - loss: 0.2488 - accuracy: 0.9027 - val_loss: 0.6604 - val_accuracy: 0.7402\n",
            "Epoch 154/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.2709 - accuracy: 0.8911 - val_loss: 0.6539 - val_accuracy: 0.7480\n",
            "Epoch 155/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.2623 - accuracy: 0.8911 - val_loss: 0.6471 - val_accuracy: 0.7480\n",
            "Epoch 156/500\n",
            "257/257 [==============================] - 0s 77us/step - loss: 0.2555 - accuracy: 0.8911 - val_loss: 0.6537 - val_accuracy: 0.7323\n",
            "Epoch 157/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.2542 - accuracy: 0.8949 - val_loss: 0.6555 - val_accuracy: 0.7244\n",
            "Epoch 158/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.2506 - accuracy: 0.8988 - val_loss: 0.6803 - val_accuracy: 0.7323\n",
            "Epoch 159/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.3028 - accuracy: 0.8677 - val_loss: 0.6710 - val_accuracy: 0.7323\n",
            "Epoch 160/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.2966 - accuracy: 0.8716 - val_loss: 0.6567 - val_accuracy: 0.7323\n",
            "Epoch 161/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.2834 - accuracy: 0.8794 - val_loss: 0.6489 - val_accuracy: 0.7559\n",
            "Epoch 162/500\n",
            "257/257 [==============================] - 0s 75us/step - loss: 0.2691 - accuracy: 0.8872 - val_loss: 0.6515 - val_accuracy: 0.7402\n",
            "Epoch 163/500\n",
            "257/257 [==============================] - 0s 68us/step - loss: 0.2650 - accuracy: 0.8911 - val_loss: 0.6513 - val_accuracy: 0.7480\n",
            "Epoch 164/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.2600 - accuracy: 0.8911 - val_loss: 0.6492 - val_accuracy: 0.7402\n",
            "Epoch 165/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.2526 - accuracy: 0.8911 - val_loss: 0.6489 - val_accuracy: 0.7402\n",
            "Epoch 166/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.2473 - accuracy: 0.8911 - val_loss: 0.6577 - val_accuracy: 0.7480\n",
            "Epoch 167/500\n",
            "257/257 [==============================] - 0s 47us/step - loss: 0.2449 - accuracy: 0.8911 - val_loss: 0.6617 - val_accuracy: 0.7402\n",
            "Epoch 168/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.2435 - accuracy: 0.8872 - val_loss: 0.6641 - val_accuracy: 0.7323\n",
            "Epoch 169/500\n",
            "257/257 [==============================] - 0s 59us/step - loss: 0.2479 - accuracy: 0.8833 - val_loss: 0.6666 - val_accuracy: 0.7402\n",
            "Epoch 170/500\n",
            "257/257 [==============================] - 0s 62us/step - loss: 0.2450 - accuracy: 0.8911 - val_loss: 0.6892 - val_accuracy: 0.7559\n",
            "Epoch 171/500\n",
            "257/257 [==============================] - 0s 62us/step - loss: 0.2629 - accuracy: 0.8833 - val_loss: 0.6882 - val_accuracy: 0.7559\n",
            "Epoch 172/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.2573 - accuracy: 0.8833 - val_loss: 0.6922 - val_accuracy: 0.7480\n",
            "Epoch 173/500\n",
            "257/257 [==============================] - 0s 67us/step - loss: 0.2759 - accuracy: 0.8794 - val_loss: 0.7004 - val_accuracy: 0.7480\n",
            "Epoch 174/500\n",
            "257/257 [==============================] - 0s 76us/step - loss: 0.2681 - accuracy: 0.8911 - val_loss: 0.6772 - val_accuracy: 0.7480\n",
            "Epoch 175/500\n",
            "257/257 [==============================] - 0s 59us/step - loss: 0.2684 - accuracy: 0.8949 - val_loss: 0.6703 - val_accuracy: 0.7402\n",
            "Epoch 176/500\n",
            "257/257 [==============================] - 0s 62us/step - loss: 0.2625 - accuracy: 0.9027 - val_loss: 0.6699 - val_accuracy: 0.7559\n",
            "Epoch 177/500\n",
            "257/257 [==============================] - 0s 62us/step - loss: 0.2533 - accuracy: 0.9027 - val_loss: 0.6645 - val_accuracy: 0.7480\n",
            "Epoch 178/500\n",
            "257/257 [==============================] - 0s 61us/step - loss: 0.2544 - accuracy: 0.8988 - val_loss: 0.6663 - val_accuracy: 0.7638\n",
            "Epoch 179/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.2481 - accuracy: 0.9066 - val_loss: 0.6628 - val_accuracy: 0.7638\n",
            "Epoch 180/500\n",
            "257/257 [==============================] - 0s 59us/step - loss: 0.2397 - accuracy: 0.9066 - val_loss: 0.6789 - val_accuracy: 0.7244\n",
            "Epoch 181/500\n",
            "257/257 [==============================] - 0s 59us/step - loss: 0.2497 - accuracy: 0.8949 - val_loss: 0.6605 - val_accuracy: 0.7323\n",
            "Epoch 182/500\n",
            "257/257 [==============================] - 0s 63us/step - loss: 0.2665 - accuracy: 0.8872 - val_loss: 0.6614 - val_accuracy: 0.7323\n",
            "Epoch 183/500\n",
            "257/257 [==============================] - 0s 68us/step - loss: 0.2606 - accuracy: 0.9027 - val_loss: 0.6953 - val_accuracy: 0.7008\n",
            "Epoch 184/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.2767 - accuracy: 0.8911 - val_loss: 0.6908 - val_accuracy: 0.7165\n",
            "Epoch 185/500\n",
            "257/257 [==============================] - 0s 66us/step - loss: 0.2743 - accuracy: 0.8794 - val_loss: 0.6803 - val_accuracy: 0.7244\n",
            "Epoch 186/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.2645 - accuracy: 0.8911 - val_loss: 0.6671 - val_accuracy: 0.7323\n",
            "Epoch 187/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.2542 - accuracy: 0.8988 - val_loss: 0.6605 - val_accuracy: 0.7402\n",
            "Epoch 188/500\n",
            "257/257 [==============================] - 0s 68us/step - loss: 0.2463 - accuracy: 0.9027 - val_loss: 0.6579 - val_accuracy: 0.7480\n",
            "Epoch 189/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.2395 - accuracy: 0.9066 - val_loss: 0.6694 - val_accuracy: 0.7402\n",
            "Epoch 190/500\n",
            "257/257 [==============================] - 0s 59us/step - loss: 0.2452 - accuracy: 0.9027 - val_loss: 0.6677 - val_accuracy: 0.7480\n",
            "Epoch 191/500\n",
            "257/257 [==============================] - 0s 71us/step - loss: 0.2400 - accuracy: 0.9066 - val_loss: 0.6658 - val_accuracy: 0.7402\n",
            "Epoch 192/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.2324 - accuracy: 0.9144 - val_loss: 0.6731 - val_accuracy: 0.7402\n",
            "Epoch 193/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.2293 - accuracy: 0.9144 - val_loss: 0.6656 - val_accuracy: 0.7402\n",
            "Epoch 194/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.2243 - accuracy: 0.9105 - val_loss: 0.6662 - val_accuracy: 0.7323\n",
            "Epoch 195/500\n",
            "257/257 [==============================] - 0s 59us/step - loss: 0.2208 - accuracy: 0.9144 - val_loss: 0.6778 - val_accuracy: 0.7480\n",
            "Epoch 196/500\n",
            "257/257 [==============================] - 0s 70us/step - loss: 0.2181 - accuracy: 0.9144 - val_loss: 0.6906 - val_accuracy: 0.7480\n",
            "Epoch 197/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.2223 - accuracy: 0.9144 - val_loss: 0.6903 - val_accuracy: 0.7402\n",
            "Epoch 198/500\n",
            "257/257 [==============================] - 0s 62us/step - loss: 0.2172 - accuracy: 0.9066 - val_loss: 0.6873 - val_accuracy: 0.7402\n",
            "Epoch 199/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.2138 - accuracy: 0.9105 - val_loss: 0.6868 - val_accuracy: 0.7480\n",
            "Epoch 200/500\n",
            "257/257 [==============================] - 0s 80us/step - loss: 0.2119 - accuracy: 0.9105 - val_loss: 0.6887 - val_accuracy: 0.7480\n",
            "Epoch 201/500\n",
            "257/257 [==============================] - 0s 67us/step - loss: 0.2117 - accuracy: 0.9105 - val_loss: 0.6896 - val_accuracy: 0.7480\n",
            "Epoch 202/500\n",
            "257/257 [==============================] - 0s 69us/step - loss: 0.2103 - accuracy: 0.9144 - val_loss: 0.6846 - val_accuracy: 0.7480\n",
            "Epoch 203/500\n",
            "257/257 [==============================] - 0s 69us/step - loss: 0.2093 - accuracy: 0.9144 - val_loss: 0.7573 - val_accuracy: 0.7087\n",
            "Epoch 204/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.2524 - accuracy: 0.8872 - val_loss: 0.7543 - val_accuracy: 0.7087\n",
            "Epoch 205/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.2416 - accuracy: 0.8988 - val_loss: 0.7347 - val_accuracy: 0.7244\n",
            "Epoch 206/500\n",
            "257/257 [==============================] - 0s 64us/step - loss: 0.2288 - accuracy: 0.9105 - val_loss: 0.7214 - val_accuracy: 0.7402\n",
            "Epoch 207/500\n",
            "257/257 [==============================] - 0s 65us/step - loss: 0.2214 - accuracy: 0.9183 - val_loss: 0.7216 - val_accuracy: 0.7323\n",
            "Epoch 208/500\n",
            "257/257 [==============================] - 0s 64us/step - loss: 0.2179 - accuracy: 0.9222 - val_loss: 0.7165 - val_accuracy: 0.7480\n",
            "Epoch 209/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.2132 - accuracy: 0.9144 - val_loss: 0.7130 - val_accuracy: 0.7480\n",
            "Epoch 210/500\n",
            "257/257 [==============================] - 0s 60us/step - loss: 0.2086 - accuracy: 0.9261 - val_loss: 0.7118 - val_accuracy: 0.7480\n",
            "Epoch 211/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.2055 - accuracy: 0.9222 - val_loss: 0.7091 - val_accuracy: 0.7480\n",
            "Epoch 212/500\n",
            "257/257 [==============================] - 0s 59us/step - loss: 0.2025 - accuracy: 0.9183 - val_loss: 0.8058 - val_accuracy: 0.7008\n",
            "Epoch 213/500\n",
            "257/257 [==============================] - 0s 60us/step - loss: 0.3027 - accuracy: 0.8833 - val_loss: 0.7815 - val_accuracy: 0.7323\n",
            "Epoch 214/500\n",
            "257/257 [==============================] - 0s 72us/step - loss: 0.2845 - accuracy: 0.8833 - val_loss: 0.7590 - val_accuracy: 0.7559\n",
            "Epoch 215/500\n",
            "257/257 [==============================] - 0s 64us/step - loss: 0.2680 - accuracy: 0.9027 - val_loss: 0.7413 - val_accuracy: 0.7402\n",
            "Epoch 216/500\n",
            "257/257 [==============================] - 0s 73us/step - loss: 0.2527 - accuracy: 0.9105 - val_loss: 0.7044 - val_accuracy: 0.7402\n",
            "Epoch 217/500\n",
            "257/257 [==============================] - 0s 59us/step - loss: 0.2311 - accuracy: 0.9183 - val_loss: 0.6966 - val_accuracy: 0.7402\n",
            "Epoch 218/500\n",
            "257/257 [==============================] - 0s 67us/step - loss: 0.2237 - accuracy: 0.9027 - val_loss: 0.6944 - val_accuracy: 0.7402\n",
            "Epoch 219/500\n",
            "257/257 [==============================] - 0s 65us/step - loss: 0.2168 - accuracy: 0.9027 - val_loss: 0.7015 - val_accuracy: 0.7480\n",
            "Epoch 220/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.2123 - accuracy: 0.9144 - val_loss: 0.7013 - val_accuracy: 0.7402\n",
            "Epoch 221/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.2073 - accuracy: 0.9144 - val_loss: 0.7013 - val_accuracy: 0.7244\n",
            "Epoch 222/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.2035 - accuracy: 0.9144 - val_loss: 0.7018 - val_accuracy: 0.7165\n",
            "Epoch 223/500\n",
            "257/257 [==============================] - 0s 47us/step - loss: 0.1996 - accuracy: 0.9300 - val_loss: 0.7037 - val_accuracy: 0.7244\n",
            "Epoch 224/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.1971 - accuracy: 0.9300 - val_loss: 0.7409 - val_accuracy: 0.7323\n",
            "Epoch 225/500\n",
            "257/257 [==============================] - 0s 60us/step - loss: 0.2238 - accuracy: 0.9105 - val_loss: 0.7390 - val_accuracy: 0.7402\n",
            "Epoch 226/500\n",
            "257/257 [==============================] - 0s 47us/step - loss: 0.2170 - accuracy: 0.9066 - val_loss: 0.7519 - val_accuracy: 0.7244\n",
            "Epoch 227/500\n",
            "257/257 [==============================] - 0s 47us/step - loss: 0.2298 - accuracy: 0.9066 - val_loss: 0.7325 - val_accuracy: 0.7323\n",
            "Epoch 228/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.2095 - accuracy: 0.9144 - val_loss: 0.7113 - val_accuracy: 0.7323\n",
            "Epoch 229/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.2113 - accuracy: 0.9066 - val_loss: 0.7042 - val_accuracy: 0.7480\n",
            "Epoch 230/500\n",
            "257/257 [==============================] - 0s 59us/step - loss: 0.2059 - accuracy: 0.9222 - val_loss: 0.6952 - val_accuracy: 0.7323\n",
            "Epoch 231/500\n",
            "257/257 [==============================] - 0s 46us/step - loss: 0.1957 - accuracy: 0.9222 - val_loss: 0.6931 - val_accuracy: 0.7480\n",
            "Epoch 232/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.1906 - accuracy: 0.9339 - val_loss: 0.6955 - val_accuracy: 0.7323\n",
            "Epoch 233/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.1881 - accuracy: 0.9377 - val_loss: 0.7041 - val_accuracy: 0.7402\n",
            "Epoch 234/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.1892 - accuracy: 0.9339 - val_loss: 0.7080 - val_accuracy: 0.7480\n",
            "Epoch 235/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.1858 - accuracy: 0.9377 - val_loss: 0.7101 - val_accuracy: 0.7559\n",
            "Epoch 236/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.1833 - accuracy: 0.9377 - val_loss: 0.7157 - val_accuracy: 0.7480\n",
            "Epoch 237/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.1814 - accuracy: 0.9300 - val_loss: 0.7171 - val_accuracy: 0.7717\n",
            "Epoch 238/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.1802 - accuracy: 0.9416 - val_loss: 0.7189 - val_accuracy: 0.7717\n",
            "Epoch 239/500\n",
            "257/257 [==============================] - 0s 59us/step - loss: 0.1779 - accuracy: 0.9416 - val_loss: 0.7485 - val_accuracy: 0.7559\n",
            "Epoch 240/500\n",
            "257/257 [==============================] - 0s 47us/step - loss: 0.2075 - accuracy: 0.9183 - val_loss: 0.7475 - val_accuracy: 0.7559\n",
            "Epoch 241/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.2038 - accuracy: 0.9261 - val_loss: 0.7158 - val_accuracy: 0.7638\n",
            "Epoch 242/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.2088 - accuracy: 0.9183 - val_loss: 0.7141 - val_accuracy: 0.7402\n",
            "Epoch 243/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.2022 - accuracy: 0.9339 - val_loss: 0.7328 - val_accuracy: 0.7244\n",
            "Epoch 244/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.2161 - accuracy: 0.9105 - val_loss: 0.7280 - val_accuracy: 0.7402\n",
            "Epoch 245/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.2034 - accuracy: 0.9416 - val_loss: 0.7381 - val_accuracy: 0.7087\n",
            "Epoch 246/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.2051 - accuracy: 0.9339 - val_loss: 0.7501 - val_accuracy: 0.7244\n",
            "Epoch 247/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.2049 - accuracy: 0.9183 - val_loss: 0.7442 - val_accuracy: 0.7244\n",
            "Epoch 248/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.1975 - accuracy: 0.9222 - val_loss: 0.7488 - val_accuracy: 0.7165\n",
            "Epoch 249/500\n",
            "257/257 [==============================] - 0s 65us/step - loss: 0.2051 - accuracy: 0.9222 - val_loss: 0.7419 - val_accuracy: 0.7165\n",
            "Epoch 250/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.1951 - accuracy: 0.9222 - val_loss: 0.7407 - val_accuracy: 0.7008\n",
            "Epoch 251/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.1876 - accuracy: 0.9183 - val_loss: 0.7586 - val_accuracy: 0.7087\n",
            "Epoch 252/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.1978 - accuracy: 0.9105 - val_loss: 0.7662 - val_accuracy: 0.7087\n",
            "Epoch 253/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.1938 - accuracy: 0.9222 - val_loss: 0.7653 - val_accuracy: 0.7087\n",
            "Epoch 254/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.1874 - accuracy: 0.9339 - val_loss: 0.7640 - val_accuracy: 0.7165\n",
            "Epoch 255/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.1808 - accuracy: 0.9261 - val_loss: 0.7623 - val_accuracy: 0.7402\n",
            "Epoch 256/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.1764 - accuracy: 0.9261 - val_loss: 0.7781 - val_accuracy: 0.7402\n",
            "Epoch 257/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.1758 - accuracy: 0.9222 - val_loss: 0.7814 - val_accuracy: 0.7480\n",
            "Epoch 258/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.1741 - accuracy: 0.9222 - val_loss: 0.7817 - val_accuracy: 0.7559\n",
            "Epoch 259/500\n",
            "257/257 [==============================] - 0s 65us/step - loss: 0.1704 - accuracy: 0.9261 - val_loss: 0.7842 - val_accuracy: 0.7402\n",
            "Epoch 260/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.1771 - accuracy: 0.9339 - val_loss: 0.7771 - val_accuracy: 0.7402\n",
            "Epoch 261/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.1713 - accuracy: 0.9377 - val_loss: 0.7738 - val_accuracy: 0.7323\n",
            "Epoch 262/500\n",
            "257/257 [==============================] - 0s 59us/step - loss: 0.1682 - accuracy: 0.9377 - val_loss: 0.7758 - val_accuracy: 0.7402\n",
            "Epoch 263/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.1656 - accuracy: 0.9377 - val_loss: 0.7893 - val_accuracy: 0.7402\n",
            "Epoch 264/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.1717 - accuracy: 0.9416 - val_loss: 0.7933 - val_accuracy: 0.7402\n",
            "Epoch 265/500\n",
            "257/257 [==============================] - 0s 46us/step - loss: 0.1705 - accuracy: 0.9416 - val_loss: 0.7912 - val_accuracy: 0.7402\n",
            "Epoch 266/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.1781 - accuracy: 0.9377 - val_loss: 0.7751 - val_accuracy: 0.7402\n",
            "Epoch 267/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.1660 - accuracy: 0.9494 - val_loss: 0.7709 - val_accuracy: 0.7402\n",
            "Epoch 268/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.1618 - accuracy: 0.9611 - val_loss: 0.7843 - val_accuracy: 0.7165\n",
            "Epoch 269/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.1716 - accuracy: 0.9455 - val_loss: 0.7884 - val_accuracy: 0.7244\n",
            "Epoch 270/500\n",
            "257/257 [==============================] - 0s 59us/step - loss: 0.1679 - accuracy: 0.9377 - val_loss: 0.7827 - val_accuracy: 0.7323\n",
            "Epoch 271/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.1604 - accuracy: 0.9533 - val_loss: 0.7805 - val_accuracy: 0.7480\n",
            "Epoch 272/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.1566 - accuracy: 0.9572 - val_loss: 0.7832 - val_accuracy: 0.7402\n",
            "Epoch 273/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.1532 - accuracy: 0.9650 - val_loss: 0.7982 - val_accuracy: 0.7480\n",
            "Epoch 274/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.1552 - accuracy: 0.9572 - val_loss: 0.7993 - val_accuracy: 0.7480\n",
            "Epoch 275/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.1524 - accuracy: 0.9572 - val_loss: 0.7976 - val_accuracy: 0.7402\n",
            "Epoch 276/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.1512 - accuracy: 0.9494 - val_loss: 0.7970 - val_accuracy: 0.7480\n",
            "Epoch 277/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.1497 - accuracy: 0.9572 - val_loss: 0.7952 - val_accuracy: 0.7638\n",
            "Epoch 278/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.1479 - accuracy: 0.9611 - val_loss: 0.7953 - val_accuracy: 0.7638\n",
            "Epoch 279/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.1454 - accuracy: 0.9650 - val_loss: 0.7962 - val_accuracy: 0.7638\n",
            "Epoch 280/500\n",
            "257/257 [==============================] - 0s 68us/step - loss: 0.1445 - accuracy: 0.9650 - val_loss: 0.7980 - val_accuracy: 0.7638\n",
            "Epoch 281/500\n",
            "257/257 [==============================] - 0s 63us/step - loss: 0.1433 - accuracy: 0.9650 - val_loss: 0.8008 - val_accuracy: 0.7480\n",
            "Epoch 282/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.1443 - accuracy: 0.9572 - val_loss: 0.7995 - val_accuracy: 0.7638\n",
            "Epoch 283/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.1428 - accuracy: 0.9572 - val_loss: 0.8003 - val_accuracy: 0.7638\n",
            "Epoch 284/500\n",
            "257/257 [==============================] - 0s 45us/step - loss: 0.1409 - accuracy: 0.9572 - val_loss: 0.8061 - val_accuracy: 0.7559\n",
            "Epoch 285/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.1412 - accuracy: 0.9572 - val_loss: 0.8060 - val_accuracy: 0.7559\n",
            "Epoch 286/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.1390 - accuracy: 0.9611 - val_loss: 0.8191 - val_accuracy: 0.7480\n",
            "Epoch 287/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.1480 - accuracy: 0.9494 - val_loss: 0.8159 - val_accuracy: 0.7402\n",
            "Epoch 288/500\n",
            "257/257 [==============================] - 0s 60us/step - loss: 0.1439 - accuracy: 0.9455 - val_loss: 0.8124 - val_accuracy: 0.7480\n",
            "Epoch 289/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.1401 - accuracy: 0.9572 - val_loss: 0.8145 - val_accuracy: 0.7480\n",
            "Epoch 290/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.1382 - accuracy: 0.9611 - val_loss: 0.8278 - val_accuracy: 0.7559\n",
            "Epoch 291/500\n",
            "257/257 [==============================] - 0s 47us/step - loss: 0.1776 - accuracy: 0.9339 - val_loss: 0.8262 - val_accuracy: 0.7638\n",
            "Epoch 292/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.1756 - accuracy: 0.9455 - val_loss: 0.8481 - val_accuracy: 0.7323\n",
            "Epoch 293/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.2459 - accuracy: 0.9027 - val_loss: 0.8355 - val_accuracy: 0.7402\n",
            "Epoch 294/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.2454 - accuracy: 0.9066 - val_loss: 0.8137 - val_accuracy: 0.7638\n",
            "Epoch 295/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.2039 - accuracy: 0.9377 - val_loss: 0.8149 - val_accuracy: 0.7717\n",
            "Epoch 296/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.1898 - accuracy: 0.9377 - val_loss: 0.8186 - val_accuracy: 0.7717\n",
            "Epoch 297/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.1800 - accuracy: 0.9416 - val_loss: 0.8224 - val_accuracy: 0.7874\n",
            "Epoch 298/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.1700 - accuracy: 0.9494 - val_loss: 0.8254 - val_accuracy: 0.7795\n",
            "Epoch 299/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.1643 - accuracy: 0.9533 - val_loss: 0.8329 - val_accuracy: 0.7559\n",
            "Epoch 300/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.1660 - accuracy: 0.9455 - val_loss: 0.8367 - val_accuracy: 0.7559\n",
            "Epoch 301/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.1615 - accuracy: 0.9455 - val_loss: 0.8347 - val_accuracy: 0.7559\n",
            "Epoch 302/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.1559 - accuracy: 0.9572 - val_loss: 0.8356 - val_accuracy: 0.7480\n",
            "Epoch 303/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.1513 - accuracy: 0.9572 - val_loss: 0.8384 - val_accuracy: 0.7559\n",
            "Epoch 304/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.1484 - accuracy: 0.9611 - val_loss: 0.8452 - val_accuracy: 0.7559\n",
            "Epoch 305/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.1478 - accuracy: 0.9494 - val_loss: 0.8486 - val_accuracy: 0.7402\n",
            "Epoch 306/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.1735 - accuracy: 0.9300 - val_loss: 0.9039 - val_accuracy: 0.7480\n",
            "Epoch 307/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.1927 - accuracy: 0.9300 - val_loss: 0.9163 - val_accuracy: 0.7480\n",
            "Epoch 308/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.1903 - accuracy: 0.9261 - val_loss: 0.8917 - val_accuracy: 0.7795\n",
            "Epoch 309/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.1576 - accuracy: 0.9377 - val_loss: 0.9013 - val_accuracy: 0.7795\n",
            "Epoch 310/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.1536 - accuracy: 0.9416 - val_loss: 0.9039 - val_accuracy: 0.7874\n",
            "Epoch 311/500\n",
            "257/257 [==============================] - 0s 44us/step - loss: 0.1504 - accuracy: 0.9494 - val_loss: 0.9030 - val_accuracy: 0.7795\n",
            "Epoch 312/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.1468 - accuracy: 0.9494 - val_loss: 0.8998 - val_accuracy: 0.7717\n",
            "Epoch 313/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.1431 - accuracy: 0.9455 - val_loss: 0.8962 - val_accuracy: 0.7717\n",
            "Epoch 314/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.1399 - accuracy: 0.9533 - val_loss: 0.8917 - val_accuracy: 0.7717\n",
            "Epoch 315/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.1370 - accuracy: 0.9572 - val_loss: 0.9100 - val_accuracy: 0.7559\n",
            "Epoch 316/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.1548 - accuracy: 0.9416 - val_loss: 0.9023 - val_accuracy: 0.7638\n",
            "Epoch 317/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.1461 - accuracy: 0.9455 - val_loss: 0.8955 - val_accuracy: 0.7638\n",
            "Epoch 318/500\n",
            "257/257 [==============================] - 0s 61us/step - loss: 0.1413 - accuracy: 0.9455 - val_loss: 0.8904 - val_accuracy: 0.7795\n",
            "Epoch 319/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.1575 - accuracy: 0.9339 - val_loss: 0.8864 - val_accuracy: 0.7874\n",
            "Epoch 320/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.1550 - accuracy: 0.9455 - val_loss: 0.8781 - val_accuracy: 0.7795\n",
            "Epoch 321/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.1461 - accuracy: 0.9494 - val_loss: 0.8735 - val_accuracy: 0.7874\n",
            "Epoch 322/500\n",
            "257/257 [==============================] - 0s 63us/step - loss: 0.1441 - accuracy: 0.9494 - val_loss: 0.8651 - val_accuracy: 0.7795\n",
            "Epoch 323/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.1363 - accuracy: 0.9533 - val_loss: 0.8719 - val_accuracy: 0.7717\n",
            "Epoch 324/500\n",
            "257/257 [==============================] - 0s 62us/step - loss: 0.1454 - accuracy: 0.9377 - val_loss: 0.8795 - val_accuracy: 0.7717\n",
            "Epoch 325/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.1497 - accuracy: 0.9377 - val_loss: 0.8803 - val_accuracy: 0.7638\n",
            "Epoch 326/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.1458 - accuracy: 0.9416 - val_loss: 0.9009 - val_accuracy: 0.7638\n",
            "Epoch 327/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.1394 - accuracy: 0.9494 - val_loss: 0.9173 - val_accuracy: 0.7638\n",
            "Epoch 328/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.1418 - accuracy: 0.9611 - val_loss: 0.9305 - val_accuracy: 0.7559\n",
            "Epoch 329/500\n",
            "257/257 [==============================] - 0s 63us/step - loss: 0.1469 - accuracy: 0.9416 - val_loss: 0.9314 - val_accuracy: 0.7480\n",
            "Epoch 330/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.1465 - accuracy: 0.9455 - val_loss: 0.9259 - val_accuracy: 0.7480\n",
            "Epoch 331/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.1418 - accuracy: 0.9455 - val_loss: 0.9220 - val_accuracy: 0.7559\n",
            "Epoch 332/500\n",
            "257/257 [==============================] - 0s 61us/step - loss: 0.1372 - accuracy: 0.9494 - val_loss: 0.9173 - val_accuracy: 0.7559\n",
            "Epoch 333/500\n",
            "257/257 [==============================] - 0s 66us/step - loss: 0.1341 - accuracy: 0.9455 - val_loss: 0.9091 - val_accuracy: 0.7559\n",
            "Epoch 334/500\n",
            "257/257 [==============================] - 0s 64us/step - loss: 0.1424 - accuracy: 0.9416 - val_loss: 0.9291 - val_accuracy: 0.7559\n",
            "Epoch 335/500\n",
            "257/257 [==============================] - 0s 73us/step - loss: 0.1697 - accuracy: 0.9183 - val_loss: 0.9430 - val_accuracy: 0.7323\n",
            "Epoch 336/500\n",
            "257/257 [==============================] - 0s 63us/step - loss: 0.1713 - accuracy: 0.9183 - val_loss: 0.9367 - val_accuracy: 0.7323\n",
            "Epoch 337/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.1622 - accuracy: 0.9222 - val_loss: 0.9225 - val_accuracy: 0.7480\n",
            "Epoch 338/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.1496 - accuracy: 0.9339 - val_loss: 0.9107 - val_accuracy: 0.7480\n",
            "Epoch 339/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.1395 - accuracy: 0.9533 - val_loss: 0.9061 - val_accuracy: 0.7480\n",
            "Epoch 340/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.1341 - accuracy: 0.9455 - val_loss: 0.9034 - val_accuracy: 0.7559\n",
            "Epoch 341/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.1295 - accuracy: 0.9494 - val_loss: 0.8994 - val_accuracy: 0.7638\n",
            "Epoch 342/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.1255 - accuracy: 0.9494 - val_loss: 0.8748 - val_accuracy: 0.7638\n",
            "Epoch 343/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.1743 - accuracy: 0.9300 - val_loss: 0.8687 - val_accuracy: 0.7480\n",
            "Epoch 344/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.1791 - accuracy: 0.9339 - val_loss: 0.9028 - val_accuracy: 0.7323\n",
            "Epoch 345/500\n",
            "257/257 [==============================] - 0s 62us/step - loss: 0.2004 - accuracy: 0.9144 - val_loss: 0.8864 - val_accuracy: 0.7402\n",
            "Epoch 346/500\n",
            "257/257 [==============================] - 0s 93us/step - loss: 0.1762 - accuracy: 0.9416 - val_loss: 0.8703 - val_accuracy: 0.7559\n",
            "Epoch 347/500\n",
            "257/257 [==============================] - 0s 66us/step - loss: 0.1560 - accuracy: 0.9572 - val_loss: 0.8614 - val_accuracy: 0.7480\n",
            "Epoch 348/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.1420 - accuracy: 0.9572 - val_loss: 0.8597 - val_accuracy: 0.7559\n",
            "Epoch 349/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.1337 - accuracy: 0.9611 - val_loss: 0.8609 - val_accuracy: 0.7559\n",
            "Epoch 350/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.1271 - accuracy: 0.9611 - val_loss: 0.8637 - val_accuracy: 0.7559\n",
            "Epoch 351/500\n",
            "257/257 [==============================] - 0s 60us/step - loss: 0.1221 - accuracy: 0.9611 - val_loss: 0.8673 - val_accuracy: 0.7559\n",
            "Epoch 352/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.1187 - accuracy: 0.9611 - val_loss: 0.9013 - val_accuracy: 0.7559\n",
            "Epoch 353/500\n",
            "257/257 [==============================] - 0s 60us/step - loss: 0.1376 - accuracy: 0.9533 - val_loss: 0.9039 - val_accuracy: 0.7480\n",
            "Epoch 354/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.1281 - accuracy: 0.9650 - val_loss: 0.9003 - val_accuracy: 0.7638\n",
            "Epoch 355/500\n",
            "257/257 [==============================] - 0s 46us/step - loss: 0.1216 - accuracy: 0.9650 - val_loss: 0.9003 - val_accuracy: 0.7717\n",
            "Epoch 356/500\n",
            "257/257 [==============================] - 0s 46us/step - loss: 0.1273 - accuracy: 0.9494 - val_loss: 0.9038 - val_accuracy: 0.7717\n",
            "Epoch 357/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.1246 - accuracy: 0.9572 - val_loss: 0.8869 - val_accuracy: 0.7717\n",
            "Epoch 358/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.1351 - accuracy: 0.9533 - val_loss: 0.8721 - val_accuracy: 0.7638\n",
            "Epoch 359/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.1238 - accuracy: 0.9611 - val_loss: 0.8651 - val_accuracy: 0.7717\n",
            "Epoch 360/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.1183 - accuracy: 0.9689 - val_loss: 0.9245 - val_accuracy: 0.7402\n",
            "Epoch 361/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.1700 - accuracy: 0.9300 - val_loss: 0.9053 - val_accuracy: 0.7559\n",
            "Epoch 362/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.1616 - accuracy: 0.9416 - val_loss: 0.8782 - val_accuracy: 0.7795\n",
            "Epoch 363/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.1374 - accuracy: 0.9533 - val_loss: 0.8689 - val_accuracy: 0.7717\n",
            "Epoch 364/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.1263 - accuracy: 0.9650 - val_loss: 0.8760 - val_accuracy: 0.7717\n",
            "Epoch 365/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.1191 - accuracy: 0.9650 - val_loss: 0.8747 - val_accuracy: 0.7638\n",
            "Epoch 366/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.1135 - accuracy: 0.9650 - val_loss: 0.8735 - val_accuracy: 0.7638\n",
            "Epoch 367/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.1084 - accuracy: 0.9728 - val_loss: 0.9015 - val_accuracy: 0.7480\n",
            "Epoch 368/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.1157 - accuracy: 0.9650 - val_loss: 0.9021 - val_accuracy: 0.7480\n",
            "Epoch 369/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.1135 - accuracy: 0.9689 - val_loss: 0.8991 - val_accuracy: 0.7480\n",
            "Epoch 370/500\n",
            "257/257 [==============================] - 0s 60us/step - loss: 0.1084 - accuracy: 0.9728 - val_loss: 0.8970 - val_accuracy: 0.7638\n",
            "Epoch 371/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.1058 - accuracy: 0.9767 - val_loss: 0.8974 - val_accuracy: 0.7717\n",
            "Epoch 372/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.1041 - accuracy: 0.9767 - val_loss: 0.8987 - val_accuracy: 0.7638\n",
            "Epoch 373/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.1026 - accuracy: 0.9767 - val_loss: 0.9021 - val_accuracy: 0.7717\n",
            "Epoch 374/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.1019 - accuracy: 0.9805 - val_loss: 0.9031 - val_accuracy: 0.7717\n",
            "Epoch 375/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.1013 - accuracy: 0.9805 - val_loss: 0.9027 - val_accuracy: 0.7717\n",
            "Epoch 376/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.0986 - accuracy: 0.9767 - val_loss: 0.9021 - val_accuracy: 0.7638\n",
            "Epoch 377/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.0969 - accuracy: 0.9805 - val_loss: 0.9022 - val_accuracy: 0.7638\n",
            "Epoch 378/500\n",
            "257/257 [==============================] - 0s 68us/step - loss: 0.0953 - accuracy: 0.9844 - val_loss: 0.9251 - val_accuracy: 0.7638\n",
            "Epoch 379/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.1025 - accuracy: 0.9728 - val_loss: 0.9294 - val_accuracy: 0.7638\n",
            "Epoch 380/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.1001 - accuracy: 0.9689 - val_loss: 0.9291 - val_accuracy: 0.7559\n",
            "Epoch 381/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.0971 - accuracy: 0.9689 - val_loss: 0.9280 - val_accuracy: 0.7480\n",
            "Epoch 382/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.0952 - accuracy: 0.9728 - val_loss: 0.9280 - val_accuracy: 0.7480\n",
            "Epoch 383/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.0933 - accuracy: 0.9767 - val_loss: 0.9277 - val_accuracy: 0.7480\n",
            "Epoch 384/500\n",
            "257/257 [==============================] - 0s 60us/step - loss: 0.0917 - accuracy: 0.9767 - val_loss: 0.9271 - val_accuracy: 0.7559\n",
            "Epoch 385/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.0902 - accuracy: 0.9767 - val_loss: 0.9273 - val_accuracy: 0.7559\n",
            "Epoch 386/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.0887 - accuracy: 0.9767 - val_loss: 0.9275 - val_accuracy: 0.7559\n",
            "Epoch 387/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.0878 - accuracy: 0.9805 - val_loss: 0.9295 - val_accuracy: 0.7559\n",
            "Epoch 388/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.0870 - accuracy: 0.9805 - val_loss: 0.9318 - val_accuracy: 0.7638\n",
            "Epoch 389/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.0869 - accuracy: 0.9844 - val_loss: 0.9318 - val_accuracy: 0.7638\n",
            "Epoch 390/500\n",
            "257/257 [==============================] - 0s 45us/step - loss: 0.0865 - accuracy: 0.9805 - val_loss: 0.9324 - val_accuracy: 0.7559\n",
            "Epoch 391/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.0856 - accuracy: 0.9844 - val_loss: 0.9330 - val_accuracy: 0.7559\n",
            "Epoch 392/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.0842 - accuracy: 0.9844 - val_loss: 0.9418 - val_accuracy: 0.7402\n",
            "Epoch 393/500\n",
            "257/257 [==============================] - 0s 63us/step - loss: 0.0869 - accuracy: 0.9844 - val_loss: 0.9398 - val_accuracy: 0.7559\n",
            "Epoch 394/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.0860 - accuracy: 0.9805 - val_loss: 0.9421 - val_accuracy: 0.7402\n",
            "Epoch 395/500\n",
            "257/257 [==============================] - 0s 82us/step - loss: 0.0852 - accuracy: 0.9805 - val_loss: 0.9589 - val_accuracy: 0.7244\n",
            "Epoch 396/500\n",
            "257/257 [==============================] - 0s 42us/step - loss: 0.1053 - accuracy: 0.9689 - val_loss: 0.9555 - val_accuracy: 0.7244\n",
            "Epoch 397/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.1013 - accuracy: 0.9650 - val_loss: 0.9487 - val_accuracy: 0.7480\n",
            "Epoch 398/500\n",
            "257/257 [==============================] - 0s 47us/step - loss: 0.0950 - accuracy: 0.9805 - val_loss: 0.9449 - val_accuracy: 0.7559\n",
            "Epoch 399/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.0893 - accuracy: 0.9767 - val_loss: 0.9441 - val_accuracy: 0.7559\n",
            "Epoch 400/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.0857 - accuracy: 0.9844 - val_loss: 0.9451 - val_accuracy: 0.7638\n",
            "Epoch 401/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.0836 - accuracy: 0.9883 - val_loss: 0.9477 - val_accuracy: 0.7559\n",
            "Epoch 402/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.0821 - accuracy: 0.9883 - val_loss: 0.9464 - val_accuracy: 0.7638\n",
            "Epoch 403/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.0823 - accuracy: 0.9844 - val_loss: 0.9483 - val_accuracy: 0.7638\n",
            "Epoch 404/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.0815 - accuracy: 0.9844 - val_loss: 0.9592 - val_accuracy: 0.7559\n",
            "Epoch 405/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.0866 - accuracy: 0.9805 - val_loss: 0.9622 - val_accuracy: 0.7559\n",
            "Epoch 406/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.0852 - accuracy: 0.9844 - val_loss: 0.9654 - val_accuracy: 0.7559\n",
            "Epoch 407/500\n",
            "257/257 [==============================] - 0s 79us/step - loss: 0.0834 - accuracy: 0.9883 - val_loss: 0.9617 - val_accuracy: 0.7638\n",
            "Epoch 408/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.0810 - accuracy: 0.9883 - val_loss: 0.9738 - val_accuracy: 0.7638\n",
            "Epoch 409/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.0833 - accuracy: 0.9805 - val_loss: 0.9707 - val_accuracy: 0.7638\n",
            "Epoch 410/500\n",
            "257/257 [==============================] - 0s 76us/step - loss: 0.0805 - accuracy: 0.9805 - val_loss: 0.9666 - val_accuracy: 0.7638\n",
            "Epoch 411/500\n",
            "257/257 [==============================] - 0s 63us/step - loss: 0.0784 - accuracy: 0.9883 - val_loss: 0.9666 - val_accuracy: 0.7559\n",
            "Epoch 412/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.0778 - accuracy: 0.9844 - val_loss: 0.9688 - val_accuracy: 0.7559\n",
            "Epoch 413/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.0769 - accuracy: 0.9883 - val_loss: 0.9705 - val_accuracy: 0.7559\n",
            "Epoch 414/500\n",
            "257/257 [==============================] - 0s 47us/step - loss: 0.0762 - accuracy: 0.9883 - val_loss: 0.9718 - val_accuracy: 0.7559\n",
            "Epoch 415/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.0754 - accuracy: 0.9844 - val_loss: 0.9823 - val_accuracy: 0.7480\n",
            "Epoch 416/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.0809 - accuracy: 0.9844 - val_loss: 0.9892 - val_accuracy: 0.7402\n",
            "Epoch 417/500\n",
            "257/257 [==============================] - 0s 47us/step - loss: 0.0835 - accuracy: 0.9844 - val_loss: 0.9805 - val_accuracy: 0.7480\n",
            "Epoch 418/500\n",
            "257/257 [==============================] - 0s 46us/step - loss: 0.0824 - accuracy: 0.9805 - val_loss: 0.9727 - val_accuracy: 0.7480\n",
            "Epoch 419/500\n",
            "257/257 [==============================] - 0s 47us/step - loss: 0.0806 - accuracy: 0.9844 - val_loss: 0.9712 - val_accuracy: 0.7559\n",
            "Epoch 420/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.0803 - accuracy: 0.9844 - val_loss: 0.9696 - val_accuracy: 0.7638\n",
            "Epoch 421/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.0789 - accuracy: 0.9883 - val_loss: 1.0022 - val_accuracy: 0.7559\n",
            "Epoch 422/500\n",
            "257/257 [==============================] - 0s 65us/step - loss: 0.0878 - accuracy: 0.9844 - val_loss: 1.0107 - val_accuracy: 0.7638\n",
            "Epoch 423/500\n",
            "257/257 [==============================] - 0s 63us/step - loss: 0.0861 - accuracy: 0.9844 - val_loss: 1.0112 - val_accuracy: 0.7638\n",
            "Epoch 424/500\n",
            "257/257 [==============================] - 0s 67us/step - loss: 0.0808 - accuracy: 0.9883 - val_loss: 1.0113 - val_accuracy: 0.7559\n",
            "Epoch 425/500\n",
            "257/257 [==============================] - 0s 64us/step - loss: 0.0778 - accuracy: 0.9922 - val_loss: 1.0272 - val_accuracy: 0.7480\n",
            "Epoch 426/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.0793 - accuracy: 0.9805 - val_loss: 1.0479 - val_accuracy: 0.7559\n",
            "Epoch 427/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.0880 - accuracy: 0.9689 - val_loss: 1.0371 - val_accuracy: 0.7559\n",
            "Epoch 428/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.0792 - accuracy: 0.9883 - val_loss: 1.0282 - val_accuracy: 0.7638\n",
            "Epoch 429/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.0752 - accuracy: 0.9922 - val_loss: 1.0236 - val_accuracy: 0.7480\n",
            "Epoch 430/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.0737 - accuracy: 0.9922 - val_loss: 1.0187 - val_accuracy: 0.7402\n",
            "Epoch 431/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.0718 - accuracy: 0.9922 - val_loss: 1.0165 - val_accuracy: 0.7323\n",
            "Epoch 432/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.0706 - accuracy: 0.9922 - val_loss: 1.0176 - val_accuracy: 0.7480\n",
            "Epoch 433/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.0697 - accuracy: 0.9922 - val_loss: 1.0151 - val_accuracy: 0.7323\n",
            "Epoch 434/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.0684 - accuracy: 0.9922 - val_loss: 1.0170 - val_accuracy: 0.7323\n",
            "Epoch 435/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.0677 - accuracy: 0.9922 - val_loss: 1.0183 - val_accuracy: 0.7323\n",
            "Epoch 436/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.0678 - accuracy: 0.9883 - val_loss: 1.0174 - val_accuracy: 0.7323\n",
            "Epoch 437/500\n",
            "257/257 [==============================] - 0s 65us/step - loss: 0.0753 - accuracy: 0.9883 - val_loss: 1.0140 - val_accuracy: 0.7323\n",
            "Epoch 438/500\n",
            "257/257 [==============================] - 0s 58us/step - loss: 0.0753 - accuracy: 0.9883 - val_loss: 1.0100 - val_accuracy: 0.7323\n",
            "Epoch 439/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.0729 - accuracy: 0.9922 - val_loss: 1.0074 - val_accuracy: 0.7323\n",
            "Epoch 440/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.0696 - accuracy: 0.9922 - val_loss: 1.0066 - val_accuracy: 0.7323\n",
            "Epoch 441/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.0678 - accuracy: 0.9922 - val_loss: 1.0079 - val_accuracy: 0.7323\n",
            "Epoch 442/500\n",
            "257/257 [==============================] - 0s 46us/step - loss: 0.0660 - accuracy: 0.9961 - val_loss: 1.0102 - val_accuracy: 0.7402\n",
            "Epoch 443/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.0655 - accuracy: 0.9961 - val_loss: 1.0135 - val_accuracy: 0.7402\n",
            "Epoch 444/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.0668 - accuracy: 0.9922 - val_loss: 1.0189 - val_accuracy: 0.7480\n",
            "Epoch 445/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.0661 - accuracy: 0.9961 - val_loss: 1.0222 - val_accuracy: 0.7480\n",
            "Epoch 446/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.0649 - accuracy: 0.9961 - val_loss: 1.0250 - val_accuracy: 0.7480\n",
            "Epoch 447/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.0636 - accuracy: 0.9961 - val_loss: 1.0286 - val_accuracy: 0.7244\n",
            "Epoch 448/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.0669 - accuracy: 0.9961 - val_loss: 1.0322 - val_accuracy: 0.7323\n",
            "Epoch 449/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.0657 - accuracy: 0.9922 - val_loss: 1.0342 - val_accuracy: 0.7402\n",
            "Epoch 450/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.0646 - accuracy: 0.9922 - val_loss: 1.0369 - val_accuracy: 0.7402\n",
            "Epoch 451/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.0635 - accuracy: 0.9961 - val_loss: 1.0385 - val_accuracy: 0.7323\n",
            "Epoch 452/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.0625 - accuracy: 0.9922 - val_loss: 1.0391 - val_accuracy: 0.7402\n",
            "Epoch 453/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.0614 - accuracy: 0.9961 - val_loss: 1.0400 - val_accuracy: 0.7402\n",
            "Epoch 454/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.0603 - accuracy: 0.9961 - val_loss: 1.0411 - val_accuracy: 0.7323\n",
            "Epoch 455/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.0598 - accuracy: 0.9961 - val_loss: 1.0432 - val_accuracy: 0.7323\n",
            "Epoch 456/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.0594 - accuracy: 0.9961 - val_loss: 1.0449 - val_accuracy: 0.7323\n",
            "Epoch 457/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.0584 - accuracy: 1.0000 - val_loss: 1.0467 - val_accuracy: 0.7244\n",
            "Epoch 458/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.0589 - accuracy: 1.0000 - val_loss: 1.0593 - val_accuracy: 0.7402\n",
            "Epoch 459/500\n",
            "257/257 [==============================] - 0s 68us/step - loss: 0.0650 - accuracy: 0.9922 - val_loss: 1.0480 - val_accuracy: 0.7402\n",
            "Epoch 460/500\n",
            "257/257 [==============================] - 0s 62us/step - loss: 0.0619 - accuracy: 0.9961 - val_loss: 1.0502 - val_accuracy: 0.7402\n",
            "Epoch 461/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.0677 - accuracy: 0.9922 - val_loss: 1.0468 - val_accuracy: 0.7480\n",
            "Epoch 462/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.0657 - accuracy: 0.9922 - val_loss: 1.0451 - val_accuracy: 0.7480\n",
            "Epoch 463/500\n",
            "257/257 [==============================] - 0s 71us/step - loss: 0.0622 - accuracy: 0.9961 - val_loss: 1.0457 - val_accuracy: 0.7559\n",
            "Epoch 464/500\n",
            "257/257 [==============================] - 0s 76us/step - loss: 0.0602 - accuracy: 0.9961 - val_loss: 1.0455 - val_accuracy: 0.7480\n",
            "Epoch 465/500\n",
            "257/257 [==============================] - 0s 61us/step - loss: 0.0579 - accuracy: 0.9961 - val_loss: 1.0462 - val_accuracy: 0.7480\n",
            "Epoch 466/500\n",
            "257/257 [==============================] - 0s 43us/step - loss: 0.0566 - accuracy: 0.9961 - val_loss: 1.0476 - val_accuracy: 0.7480\n",
            "Epoch 467/500\n",
            "257/257 [==============================] - 0s 60us/step - loss: 0.0556 - accuracy: 0.9961 - val_loss: 1.0495 - val_accuracy: 0.7480\n",
            "Epoch 468/500\n",
            "257/257 [==============================] - 0s 78us/step - loss: 0.0554 - accuracy: 0.9961 - val_loss: 1.0509 - val_accuracy: 0.7480\n",
            "Epoch 469/500\n",
            "257/257 [==============================] - 0s 51us/step - loss: 0.0547 - accuracy: 0.9961 - val_loss: 1.0549 - val_accuracy: 0.7559\n",
            "Epoch 470/500\n",
            "257/257 [==============================] - 0s 47us/step - loss: 0.0552 - accuracy: 0.9961 - val_loss: 1.0600 - val_accuracy: 0.7480\n",
            "Epoch 471/500\n",
            "257/257 [==============================] - 0s 47us/step - loss: 0.0543 - accuracy: 0.9961 - val_loss: 1.0617 - val_accuracy: 0.7480\n",
            "Epoch 472/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.0539 - accuracy: 0.9961 - val_loss: 1.0639 - val_accuracy: 0.7480\n",
            "Epoch 473/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.0532 - accuracy: 0.9961 - val_loss: 1.0708 - val_accuracy: 0.7480\n",
            "Epoch 474/500\n",
            "257/257 [==============================] - 0s 60us/step - loss: 0.0532 - accuracy: 0.9961 - val_loss: 1.0736 - val_accuracy: 0.7402\n",
            "Epoch 475/500\n",
            "257/257 [==============================] - 0s 64us/step - loss: 0.0525 - accuracy: 0.9961 - val_loss: 1.0600 - val_accuracy: 0.7480\n",
            "Epoch 476/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.0629 - accuracy: 0.9961 - val_loss: 1.0565 - val_accuracy: 0.7480\n",
            "Epoch 477/500\n",
            "257/257 [==============================] - 0s 59us/step - loss: 0.0607 - accuracy: 1.0000 - val_loss: 1.0531 - val_accuracy: 0.7480\n",
            "Epoch 478/500\n",
            "257/257 [==============================] - 0s 64us/step - loss: 0.0598 - accuracy: 0.9961 - val_loss: 1.0785 - val_accuracy: 0.7717\n",
            "Epoch 479/500\n",
            "257/257 [==============================] - 0s 63us/step - loss: 0.0660 - accuracy: 0.9961 - val_loss: 1.1107 - val_accuracy: 0.7559\n",
            "Epoch 480/500\n",
            "257/257 [==============================] - 0s 60us/step - loss: 0.0746 - accuracy: 0.9883 - val_loss: 1.1178 - val_accuracy: 0.7480\n",
            "Epoch 481/500\n",
            "257/257 [==============================] - 0s 57us/step - loss: 0.0871 - accuracy: 0.9844 - val_loss: 1.1047 - val_accuracy: 0.7638\n",
            "Epoch 482/500\n",
            "257/257 [==============================] - 0s 45us/step - loss: 0.0726 - accuracy: 0.9883 - val_loss: 1.1006 - val_accuracy: 0.7638\n",
            "Epoch 483/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.0656 - accuracy: 0.9883 - val_loss: 1.1142 - val_accuracy: 0.7559\n",
            "Epoch 484/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.0649 - accuracy: 0.9844 - val_loss: 1.1800 - val_accuracy: 0.7480\n",
            "Epoch 485/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.0787 - accuracy: 0.9689 - val_loss: 1.1664 - val_accuracy: 0.7638\n",
            "Epoch 486/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.0690 - accuracy: 0.9805 - val_loss: 1.1481 - val_accuracy: 0.7559\n",
            "Epoch 487/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.0610 - accuracy: 0.9883 - val_loss: 1.1353 - val_accuracy: 0.7402\n",
            "Epoch 488/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.0567 - accuracy: 0.9883 - val_loss: 1.1100 - val_accuracy: 0.7402\n",
            "Epoch 489/500\n",
            "257/257 [==============================] - 0s 46us/step - loss: 0.0581 - accuracy: 0.9961 - val_loss: 1.1960 - val_accuracy: 0.7480\n",
            "Epoch 490/500\n",
            "257/257 [==============================] - 0s 54us/step - loss: 0.1102 - accuracy: 0.9494 - val_loss: 1.2148 - val_accuracy: 0.7244\n",
            "Epoch 491/500\n",
            "257/257 [==============================] - 0s 55us/step - loss: 0.1016 - accuracy: 0.9689 - val_loss: 1.1628 - val_accuracy: 0.7323\n",
            "Epoch 492/500\n",
            "257/257 [==============================] - 0s 59us/step - loss: 0.0813 - accuracy: 0.9805 - val_loss: 1.1337 - val_accuracy: 0.7323\n",
            "Epoch 493/500\n",
            "257/257 [==============================] - 0s 56us/step - loss: 0.0777 - accuracy: 0.9844 - val_loss: 1.2490 - val_accuracy: 0.6929\n",
            "Epoch 494/500\n",
            "257/257 [==============================] - 0s 52us/step - loss: 0.1815 - accuracy: 0.9222 - val_loss: 1.1550 - val_accuracy: 0.7165\n",
            "Epoch 495/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.1371 - accuracy: 0.9533 - val_loss: 1.1118 - val_accuracy: 0.7244\n",
            "Epoch 496/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.1114 - accuracy: 0.9650 - val_loss: 1.0507 - val_accuracy: 0.7480\n",
            "Epoch 497/500\n",
            "257/257 [==============================] - 0s 48us/step - loss: 0.0831 - accuracy: 0.9805 - val_loss: 1.0265 - val_accuracy: 0.7480\n",
            "Epoch 498/500\n",
            "257/257 [==============================] - 0s 50us/step - loss: 0.0699 - accuracy: 0.9883 - val_loss: 1.0323 - val_accuracy: 0.7559\n",
            "Epoch 499/500\n",
            "257/257 [==============================] - 0s 49us/step - loss: 0.0665 - accuracy: 0.9883 - val_loss: 1.0370 - val_accuracy: 0.7480\n",
            "Epoch 500/500\n",
            "257/257 [==============================] - 0s 53us/step - loss: 0.0619 - accuracy: 0.9922 - val_loss: 1.0417 - val_accuracy: 0.7323\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<keras.callbacks.callbacks.History at 0x7eff7c1b94e0>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 317
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "HB9taL891hHd",
        "colab_type": "text"
      },
      "source": [
        "## Evaluate MLP ##"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YtUfl3fv0r99",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 923
        },
        "outputId": "4de64eca-5005-49a9-de21-1d63c9588265"
      },
      "source": [
        "plt.plot(history.history['accuracy'], label = 'train')\n",
        "plt.plot(history.history['val_accuracy'], label='validation')\n",
        "plt.title('model accuracy')\n",
        "plt.ylabel('accuracy')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'validation'], loc = 'upper left')\n",
        "plt.show()\n",
        "plt.plot(history.history['loss'])\n",
        "plt.plot(history.history['val_loss'])\n",
        "plt.title('model loss')\n",
        "plt.ylabel('loss')\n",
        "plt.xlabel('epoch')\n",
        "plt.legend(['train', 'validation'],loc = 'upper left')\n",
        "plt.show()\n",
        "\n",
        "y_pred2=model.predict(X_test3)\n",
        "y_pred2=(y_pred2>.5).astype(int)\n",
        "\n",
        "# todo: use boot-strap estimation\n",
        "probs = model.predict_proba(X_test3)\n",
        "probs = probs[:,0]\n",
        "auc = roc_auc_score(y_test, probs)\n",
        "print('AUC: %.3f' % auc)\n",
        "fpr, tpr, thresholds = roc_curve(y_test, probs)\n",
        "pyplot.plot([0, 1], [0, 1], linestyle = '-')\n",
        "pyplot.plot(fpr, tpr, marker = '.')\n",
        "pyplot.show()\n",
        "\n",
        "\n",
        "cmat = metrics.confusion_matrix(y_test, y_pred2)\n",
        "print( cmat )\n",
        "print(\"Accuracy:\",metrics.accuracy_score(y_test, y_pred2))\n",
        "print(\"Precision:\",metrics.precision_score(y_test, y_pred2, pos_label=1)) \n",
        "print(\"Recall:\",metrics.recall_score(y_test, y_pred2, pos_label=1))"
      ],
      "execution_count": 318,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "AUC: 0.783\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAD4CAYAAAD8Zh1EAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+j8jraAAAgAElEQVR4nO3de1zUVf7H8dcB5OIFvF8B8Y4IWopopWVl5S3NrC3dSrtZ7fbb3WpTyy6WlWZbbbtrteaatttlS8xITbtp2RVxS0C8oaLgDUUFlTtzfn98IYcRYoRhvjPf+TwfDx8xM1+dzzfx7fHMOZ+jtNYIIYTwfn5mFyCEEMI1JNCFEMIiJNCFEMIiJNCFEMIiJNCFEMIiAsx647Zt2+qoqCiz3l4IIbzS5s2bj2mt29X0mmmBHhUVRUpKillvL4QQXkkpta+212TKRQghLEICXQghLEICXQghLEICXQghLEICXQghLKLOQFdKLVFK5Sql0mt5XSml/qaUylRKpSqlBrq+TCGEEHVxZoS+FBj1K6+PBnpV/pgOvNbwsoQQPi87GTa+aPzXShrxvupch661/lopFfUrl0wA3tJGH94flFItlVKdtNaHXFSjEMLXZCfDm2PAVgbKDzrEQlCo2VU1WEVxPupIOgpQAcEwNQkiElz267tiDr0LkG33OKfyuXMopaYrpVKUUilHjx51wVsLIbyOMyPUrI1GmANoGxTnu6e2RpRfXMaR3CMorVFoqCg17tOF3LpTVGu9CFgEEB8fLydrCOErspON8AppA2seAls5oKBjXM0j71MO/8Af9iDET3NHpS5XUFzGvDXbeHdHNmNa7ufvZXPwt5WBfyBEDXfpe7ki0A8AEXaPwyufE0KI6tMn1Wg4nVtzoNvK7R74QVFeY1bYaD7LOMJjK9M4eqqEey7rzgMjR+F/eJDxl1vUcJdOt4BrAj0JuF8p9R4wBMiX+XMhLKZqhF2fELKfPnEUPRrG/bXm91s23piWaISRbGM7drqEOUlbWZV6iOiOLXjjtnj6h7c0XoxIcHmQV6kz0JVS7wIjgLZKqRzgSaAJgNb6dWANMAbIBAqB2xulUiFEw9Q3lLOT4c3RdU+T1MZx+kT5G/Pi/k1gwJSaf05EgvGBYSONZBuL1pqPfj7IUx9v5UxJBQ9d1Zt7LutBYIB7tvw4s8plch2va+D3LqtICOF6DVk1kpdpNwXyK9MktXGcPhl0G4RF1B3UjTiSbQwHTxbx2Mp0vtyey4WRLVkwqT+9OrRwaw2mtc8VQtRDfUfZNa0aqe8ywNqmSWrjOH0yYIpXBXVdbDbNO8n7mf/JdipsmifGxTD14ij8/ZTba5FAF6IhGjK3XJ/3qu/UR0NWjWQnw9KxUFH269MktfHS6RNn7D12hpmJqSTvPc6wnm2Zd30cEa2bmlaPBLoQ9eXuzS8NmfpoyKqRiASYtrphgexl0yd1Ka+wsfibvbz82U4CA/xYMKk/N8aHo5T7R+X2JNCFOB/2I3JXTmPUx/lMfTR01YjFArkhMg4WMDMxlbQD+Vwd04G518XSITTY7LIACXQhnOc4Im/eqfrrjb35pSFTHxae9nCXkvIK/vFlJq9t2E3Lpk1YOGUgY+I6mj4qtyeBLoSzHEfkZWfsXnTD5peGTn3IKLveNu87wczEVDJzT3P9wC48PjaGVs0CzS7rHBLoQjgrargxMtc2CAiBkU/B2lnu3fwioexWhaXlvLBuB0u/y6JzWAhLbx/MiD7tzS6rVhLoQjgrIsH44LM4HyYtrnwcI9MYFvXNrmPMWpFKzokibruoKzNGRdM8yLMj07OrE8LTBIUaP6rCW0bMlpNfWMazazJ4PyWH7m2b8f49F5HQrbXZZTlFAl0IZ1Stbjl1yFgCmJ0sQW5Ba9MP8/hH6Rw/U8p9I3rwxyt7EdzE3+yynCaBLoQjx81C1Tb0VFo23uWHEwjzHD1lNNNanXaImE6hvDltMLFdwswu67xJoAthr6bNQqdzHTbmcPZwAgl0r6a1ZsX/DvD0qgyKSit4+Jo+TL+0O0383dNMy9Uk0IWwt+VdJ07KUV7Z0lVUd+BkEY+uSOOrnUcZ1LUVz0/qT8/2zc0uq0Ek0IWokrIUUpZUf27Yg8ZKlqoNPX7+MPA2GDBZRudeymbT/OfHfTz/yXY08NT4ftw6tCt+JjTTcjUJdCHAmGpZ9adzny/Kc00vE+ERdh89zazEVDZlnWB4r7Y8N9HcZlquJoEuRHYybJgHOBxzq/zPTqvI8kSvVlZh442Ne/jr57sIaeLPX24cwKSBXTxq274rSKAL31TjocX2FIx9SULcAtIP5DMzMZWtBwsYHduRpyb0o30Lz2im5WoS6ML31HpocRVlNNny0lPmhaG4rIK/f7mL17/aQ6umgbz224GMjutU90/0YhLowndUjcrzc349zAOCz/8QB+FRUrKOMyMxlT1Hz3DjoHBmj+1Ly6ae10zL1STQhW9IWQqrHzCWIjqqOrRYVrB4vdMl5bywdjtv/bCPzmEhvHVHApf2bmd2WW4jgS6s75cVLLqGF8/j0GLh0b7aeZRHV6RxML+IqRdF8fA1fWjm4c20XM237lZYW23ne2ZtpMYVLGDJQ4t9zcnCUuau2kbi/3Lo0a4ZH9xzEfFR3tFMy9Uk0IV3+tV+Kw4HKDsekKz8jRUsRXkyKvdyn6Qd4vGPtnKisJT7L+/J/Vf09KpmWq4mgS68T539VhwOUK62JFEZUyyygsWr5RYU88RHW1m79TCxXUJZdsdg+nX2vmZariaBLrxL1Saguvqt2B+g7HhAsqxg8Vpaa5ZvzmHuqgyKy23MHBXN3cO7EeClzbRcTQJdmKe2Oe9fu96xjS2c22/F8QBlOSDZErKPF/Loh2ls3HWMhKjWzJ8UR/d23t1My9Uk0IU5apo2qZoiqU1eZg07Ov2c67ciW/e9VoVN89b3WbywbgcKmDuhH78dYo1mWq4mgS7cx35EnrXx3GmTugL9HAoCgqTfioVl5p5iZmIam/ed4LLe7Xju+ji6tAwxuyyPJYEu3MNxRN7cYQv2sAfr/qAyO1na2PqIsgob//xqN3/7IpOmQf689JsBTLzQes20XE0CXbiH44i87Izdi5XTJnWRNrY+IS0nnxmJqWw7VMDY/p2Yc20/2rUIMrssryCBLtwjpM3Zr/2DYORTsHbW2ZUnzp7+I9MqllVcVsFfP9/FGxv30KZZIP+8dRDX9OtodllexalAV0qNAl4B/IHFWuv5Dq9HAsuAlpXXzNJar3FxrcJbpSytfniEthmrUmTliaj04548Zq1IY++xM9wUH8GjY/sSFtLE7LK8Tp2BrpTyBxYCVwE5wCalVJLWOsPusseA97XWrymlYoA1QFQj1Cu8TU19VGzlRpAPf0iC3MedKi5jwdod/PuHfUS0DuHtu4ZwSc+2ZpfltZwZoScAmVrrPQBKqfeACYB9oGugaolCGHDQlUUKL2TfqvacPip+csCyYP2OXGavSONQQTF3XNKNP1/Tm6aBMgvcEM783+sCZNs9zgGGOFwzB/hUKfV/QDNgZE2/kFJqOjAdIDIy8nxrFZ7qV/uqOJKTgHzdiTOlzF2VwYqfDtCrfXMS77uYgZGtzC7LElz11+FkYKnW+kWl1EXAv5VSsVpXbz6ttV4ELAKIj4+vqZep8DZ19lWxJycB+TKtNavTDvHkR1vJLyrjD1f24veX9yAowHebabmaM4F+AIiwexxe+Zy9O4FRAFrr75VSwUBbINcVRQoPVtMGIUeOrWqFzzlSUMxjK9P5LOMI/cPD+M9dQ+jb6Xw3kom6OBPom4BeSqluGEF+M+D4p3I/cCWwVCnVFwgGjrqyUOGhooYbI3Ntg4AQmLTYeN6+r8roF6RVrY/SWvN+SjbPrN5GabmNR8dEc8cl0kyrsdQZ6FrrcqXU/cA6jCWJS7TWW5VSTwMpWusk4CHgDaXUAxifgE3TWsuUiq8IbmmsJ7/62bOBLRuAfN7+vEJmrUjlu915DOnWmucn9SeqbTOzy7I0ZVbuxsfH65SUFFPeW7jIL+vLK7+H/INg2ioJcB9XYdMs/S6Lv6zbgb+f4pEx0UweHCnNtFxEKbVZax1f02uyRkjUT03ryytKjVG5BLrP2nnkFDOWp/Jz9kmuiG7PsxNj6RQmzbTcRQJd1E+N53TK+nJfVVpu47UNu/nH+l00DwrglZsvYPyAztJMy80k0IXz7NebOwa38pP15T5qS/ZJZiamsv3wKcYP6MyT18bQprk00zKDBLpwjuN681bdzr5WdeiyrC/3KUWlFbz8+U4Wb9xD+xbBLL4tnpExHcwuy6dJoAvnOK43L3Rod+tM+1thGd/vzuORFalk5RUyOSGSR8ZEExoszbTMJoEunOO43ry+7W+FVysoLmP+J9t558f9dG3TlHfuHsLFPaSZlqeQQBfOiUgwtvUX5xubhyISjBa4stbcZ3yx7QizP0wn91Qxdw/vxoNX9SEkULbtexIJdHGWY5OtushhEz4h73QJT32cQdKWg/Tp0ILXbx3EBREtzS5L1EAC3ZfZBzjYdUhU0DGu+qHNJQVwONX4etl443AKCXNL01qTtOUgT32cwaniMh4Y2Zv7RvQgMEC27XsqCXRf5bhqJbilXYdEbXRMtA90+6ZbsoHI8g7lF/HYh+l8sT2XAREtWTCpP306tjC7LFEHCXRflJ0MG+ZVX7VSUVr9mujRMO6v1X/OsvHyIajF2Wya9zZlM2/NNspsNh4b25fbL+mGv2zb9woS6L7GfmRub/Bd8MOrZzskOra5jUiQM0AtLuvYGWatSOWHPce5qHsb5k+Ko2sbaablTSTQfY39evJf+EFwaN0dEuVDUEsqr7Dx5rdZvPjZDpr4+TH/+jhuGhwh2/a9kAS6ldW0aiVqOKD4pQ+L8jO6JFZdI4HtU7YfLmDm8lS25OQzsm8Hnrkulo5hwWaXJepJAt2qajoaLijUWK1SFeZ+ATDwNhgwWYLcx5SUV7Bw/W5eXZ9JWEgT/j75Qsb17ySjci8ngW5VNR0NFxRafbWK1hAWLmHuY37af4KZiansPHKaiRd24fFxMbRuFmh2WcIFJNCtqqaj4SISZLWKDyssLefFT3ey5Nu9dAwNZsm0eK6IlmZaViKBblU1bdWvel5Wq/ic7zKPMWtFGvuPF3LL0EhmjoqmhTTTshwJdCsLCjV+OIa2fPjpM/KLypi3ZhvvbcqmW9tmvDd9KEO7tzG7LNFIJNCtxn5lS0mBMULPTpYA90Gfbj3MYyvTOXa6hHsu684DI3sT3ESaaVmZBLqVpCyF1Q8Y8+b2pPeKTzl2uoQ5SVtZlXqI6I4tWDw1nv7h0kzLF0igW0VNhzZXkd4rPkFrzcqfD/DUxxkUllTw0FW9uXdED5r4SzMtXyGBbhU1Htpc+c9rWc1ieQdPFjH7wzTW7zjKhZFGM61eHaSZlq+RQPdWjrtAHQPbrwmM+YtxNJysZrEsm03zdvJ+nv9kOxU2zRPjYph6cZQ00/JREujeqKZdoPaUvxHmcmizpe05eppZiWkkZx1nWM+2zLs+jojWTc0uS5hIAt0bbXn33F2gjuTQZssqr7Cx+Ju9vPzZToIC/FhwQ39uHBQu2/aFBLrXqJpiCWljrGaxN+xB43xP2QFqeRkHC5iRuIX0AwVc068DcyfE0j5UmmkJgwS6N6ith3mVojzZAWpxJeUV/OPLTF7bsJuWTZvw6m8HMjq2o4zKRTUS6N6gxh7mlZT/2dG47AC1pM37jjMzMY3M3NNcP7ALj4+NoZU00xI1kED3BvaNtvwDjS6JtjIjzMe+JCFuUWdKynlh3Q6WfZ9F57AQlt4+mBF92ptdlvBgTgW6UmoU8ArgDyzWWs+v4ZrfAHMwFkNv0VpPcbxG1JNjoy2QqRWL27jrKI+sSCPnRBFTL+rKw6OiaR4k4y/x6+r8DlFK+QMLgauAHGCTUipJa51hd00v4BHgEq31CaWUDCNczbHRlgS5JeUXlvHM6gw+2JxD93bN+ODeixgc1drssoSXcOav/AQgU2u9B0Ap9R4wAciwu+ZuYKHW+gSA1jrX1YUKYXVr0w/z+EfpHD9Tyu9G9OAPV/aSZlrivDgT6F2AbLvHOcAQh2t6AyilvsWYlpmjtV7r+AsppaYD0wEiIyPrU68QlpN7qpg5SVtZk3aYmE6hvDltMLFdwswuS3ghV03KBQC9gBFAOPC1UipOa33S/iKt9SJgEUB8fHwNXaQEUPPhztIK13K01iT+7wBzV2VQVFbBw9f0Yfql3aWZlqg3ZwL9ABBh9zi88jl7OcCPWusyYK9SaidGwG9ySZW+pLZt/YdTjf9KK1xLyDlRyKMfpvP1zqPEd23F/En96dm+udllCS/nTKBvAnoppbphBPnNgOMKlpXAZOBNpVRbjCmYPa4s1GfUdLizPWmF69VsNs2/f9jH82u3A/DU+H7cOrQrftJMS7hAnYGutS5XSt0PrMOYH1+itd6qlHoaSNFaJ1W+drVSKgOoAB7WWkszkfqo6XBnkG39FrD76GlmLk8lZd8JLu3djucmxhLeSpppCddRWpszlR0fH69TUlJMeW+P9/rwcw93rmleXXiFsgobi77ewytf7CKkiT+Pj4th0sAusm1f1ItSarPWOr6m12Sngieq6XBn2dbvldIP5DNjeSoZhwoYE9eROeP70b6FNNMSjUMC3dNkJ0Ne5tmvJcS9UnFZBa98sYtFX++hVdNAXr9lIKNiO5ldlrA4CXRPkZ0MW96BlGVA5SHPS8fBtFUS6l5mU9ZxZi5PZc+xM9w4KJzHxsYQ1rSJ2WUJHyCB7glqa48rK1q8yumSchas3c5b3+8jvFUI/74zgeG92pldlvAhEuieoLb2uMpPVrR4ia92HuXRFWkczC9i2sVRPHxNH5pJMy3hZvId5wnslypWkda4XuFkYSlPr8pgxf8O0KNdM5bfexGDukozLWEOCXRPYN8ed9iDxglEsjzRo2mt+ST9ME98lM7JwjLuv7wn91/RU5ppCVNJoHuKqqWK8dPMrkTUIbegmMc/Smfd1iPEdgll2R0J9OsszbSE+STQzVa1YejUIbCVy1JFD6a15oPNOTyzKoOSchuzRkdz17BuBEgzLeEhJNDNVNPqFmm+5ZGyjxfyyIo0vsk8RkJUa+ZPiqN7O2mmJTyLBLqZtrwrSxU9XIVN89b3WSxYuwM/BXOvi+W3CZHSTEt4JAl0s6QshZQlDk/6SfMtD5KZe4oZy1P53/6TjOjTjmcnxtGlZYjZZQlRKwl0M2Qnw6o/nft8jxEw4hEZnZusrMLG6xt28/cvM2ka5M/LNw3gugukmZbwfBLo7padDBvmAQ5dLpW/hLkHSMvJ5+HlW9h++BTj+ndizvh+tG0eZHZZQjhFAt2datvij5JNRCYrLqvg5c938sbXe2jbPIhFtw7i6n4dzS5LiPMige5ONW7xV8bac1l/bpof9+Qxa0Uae4+d4ebBETwypi9hIdJMS3gfCXR3ctzir/zAPwgGOJ7oJ9zhVHEZz6/dzn9+2E9E6xDevmsIl/Rsa3ZZQtSbBLo7yRZ/j7F+ey6PfpjG4YJi7hzWjYeu7k3TQPnjILybfAe7m2zxN9XxM6U8/fFWVv58kF7tm5N438UMjGxldllCuIQEujvJaUSm0VqzKvUQc5K2kl9Uxh+u7MXvL+9BUIA00xLWIYHuLtnJsGQU6ArjsZxG5DZHCoqZ/WE6n287Qv/wMN6+ewjRHUPNLksIl5NAd5esjWfDHGSLvxtorfnvpmyeXbON0nIbs8f05fZLoqSZlrAsCXR3CWlT/bFs8W9U+/MKmbUile925zGkW2uen9SfqLbNzC5LiEYlge4OKUurb/VXfjB6gYzOG0GFTfPmt3v5y6c7CPDz47mJcdw8OEKaaQmfIIHe2H7p22K31V9rY8micKkdh08xIzGVLdknuSK6Pc9OjKVTmDTTEr5DAr2xZW3k3L4tcvizK5WW23h1QyYL12fSIrgJr9x8AeMHdJZmWsLnSKA3Nse5c+UnfVtcaEv2SWYsT2XHkVNMuKAzT4yLoY000xI+SgK9MWUnw+oHzz5WfjD2ZdlU5AJFpRW89NkO/vXNXtq3CGbxbfGMjOlgdllCmEoCvTE5LlWUuXOX+G73MR5Zkca+vEKmDIlk1uhoQoOlmZYQEuiNKWo4oPhlDl2WKjZIQXEZ89Zs593k/XRt05R37h7CxT2kmZYQVSTQG1NEAnSMg9O5ED0GBkyWufN6+jzjCLNXpnH0VAnTL+3OAyN7ExIo2/aFsOdUoCulRgGvAP7AYq31/FqumwQsBwZrrVNcVqW3yU42pluqRuMBQRLm9ZR3uoSnPs4gactBoju2YNGt8QyIaGl2WUJ4pDoDXSnlDywErgJygE1KqSStdYbDdS2APwI/NkahXqPaqUR20y3LxsPUJAl1J2mtSdpykDlJWzldUs4DI3tz34geBAbItn0hauPMCD0ByNRa7wFQSr0HTAAyHK6bCzwPPOzSCr1NtVOJ7NafS+8Wpx3KL+KxD9P5YnsuF0S0ZMEN/endoYXZZQnh8ZwJ9C5Att3jHGCI/QVKqYFAhNZ6tVKq1kBXSk0HpgNERkaef7XewP5UIv9AQIGtXD4QdYLNpnl3037mrdlOuc3GY2P7cvsl3fCXbftCOKXBH4oqpfyAl4BpdV2rtV4ELAKIj4/XdVzunexPJZq02Hiuaj5dRue12nvsDLMSU/lx73Eu7tGG+df3J7JNU7PLEsKrOBPoB4AIu8fhlc9VaQHEAhsqt1p3BJKUUuN99oPRqlOJqgJcgrxW5RU2lny7lxc/3UlggB/PT4rjN/ERsm1fiHpwJtA3Ab2UUt0wgvxm4JdTjbXW+cAvi4GVUhuAP/tsmMupRE7bdqiAmYmppObkc1VMB565LpYOocFmlyWE16oz0LXW5Uqp+4F1GMsWl2ittyqlngZStNZJjV2k15BTiZxSUl7BwvW7eXV9JmEhTfjHlAsZG9dJRuVCNJBTc+ha6zXAGofnnqjl2hENL8tLyalEdfrf/hPMXJ7KrtzTTLywC0+Mi6FVs0CzyxLCEmSnqCvJqUS1Kiwt5y/rdvLmd3vpGBrMm9MGc3l0e7PLEsJSJNBdpabOinIqEQDfZh5j1opUso8XccvQSGaOiqaFNNMSwuUk0F1ly7vSWdFBflEZz63exn9TsunWthn/nT6UId3b1P0ThRD1IoHeEFU9W0LaGOeG2vML8Onplk+3HuaxlenknSnl3st68KeRvQhuIs20hGhMEuj1Va1nSw0G3uKT0y1HT5Uw5+OtrE49RN9Oofxr6mDiwsPMLksInyCBXh/ZybBhXu1hrvxhwJSaX7MorTUf/nSAp1dlUFhSwZ+v7s09l/Wgib800xLCXSTQz1fKUlj9gNGrxZ7yN+bQlb/PnRl64GQRsz9MY8OOowyMNJpp9WwvzbSEcDcJ9PORnQyr/kS1LooA+MGg2yAswqd6tthsmrd/3Mf8T7Zj0/DktTHcdlGUNNMSwiQS6OcjayPnhrmqPMBiis8EOcCeo6eZlZhGctZxhvdqy3MT44hoLc20hDCTBLqzspMhP7v6c8ofBk31qdOIyitsvLFxLy9/vpPgAD9euKE/NwwKl237QngACXRn1LSipWquPH6aaWW5W8bBAmYkbiH9QAHX9OvA3AmxtJdmWkJ4DAl0Z1Q7hciOj2wcKi6r4B9fZvL6V7tp2TSQ1347kNFxncwuSwjhQALdGY49WpSfz/Rp2bzvODOWp7L76BkmDQzn8XF9adlUmmkJ4Ykk0OtyTo8W35g3P1NSzgvrdrDs+yw6h4Ww7I4ELuvdzuyyhBC/QgK9Lo4tcbUNwsItHeZf7zzKIyvSOJhfxG1Du/LwqGiaB8m3ihCeTv6U2qvqzWK/ltxxWsXCUy35hWXMXZ3B8s05dG/XjPfvuYjBUa3NLksI4SQJ9Cr2K1mUn3HQc1AolBScvUb5W7Yl7tr0Qzz+0VaOnynldyN68IcrpZmWEN5GAr3KlnfPrmTRNijONwK9OL/6dRZb2ZJ7qpgnP9rKJ+mHiekUypvTBhPbRZppCeGNJNCzk2HLO+e2vx32oLHGPDsZlo03jpOz0HSL1prlm3N4ZvU2isoqePiaPky/tLs00xLCi/l2oP9aC9yqkXhEAkxNOndu3YtlHy/k0Q/T2LjrGPFdWzF/Un96tm9udllCiAby7UCvbcOQ8q8+Eo9IsESQ22yat77PYsG6HSjg6Qn9uGVIV/ykmZYQluDbgR413PgA1L4VrkXb32bmnmZWYiop+05wae92PDcxlvBW0kxLCCvx7UCPSDBWsxTnG3PmRXmWmVapUlZhY9HXe3jl812EBPrz4o0DuH5gF2mmJYQF+Xagg7GSJSjUkk220g/kM2N5KhmHChgT15GnxsfSrkWQ2WUJIRqJbwa6/QaikgJjhJ6dbJmReXFZBa98sYtFX++hdbNAXr9lIKNipZmWEFbne4FebWWL4pcDK5aNN1azeHmob8o6zszlqew5dobfxIcze0wMYU2bmF2WEMINfC/Qq61ssTt9qKLUeM1LA/10STkL1m7nre/3Ed4qhP/cOYRhvdqaXZYQwo18L9DtV7b4BwIKbOVevWlo/Y5cZq9I41BBMbdfEsWfr+5DM2mmJYTP8b0/9fYrWyYtNp7z0k1DJ86UMndVBit+OkDP9s1Zfu/FDOrayuyyhBAm8b1Ah7MrW6oC3MuCXGvNmrTDPJmUzsnCMv7vip7cf0VPggKkmZYQvsypQFdKjQJeAfyBxVrr+Q6vPwjcBZQDR4E7tNb7XFyrAHILinlsZTqfZhwhrksYb90xhJjOoWaXJYTwAHUGulLKH1gIXAXkAJuUUkla6wy7y34C4rXWhUqp+4AFwE2NUbCv0lrzQUoOc1dnUFpu45HR0dw5rBsB0kxLCFHJmRF6ApCptd4DoJR6D5gA/BLoWuv1dtf/ANziyiJ9XfbxQh5ZkcY3mcdI6Naa+dfH0b2dNNMSQlTnTKB3AbLtHucAQ37l+juBT2p6QSk1HZgOEBkZ6WSJLpadDHmZZ7/24PnzCptm2XdZvLBuB/5+imeui2VKQqQ00xJC1MilHwJdYaUAAAtYSURBVIoqpW4B4oHLanpda70IWAQQHx+va7qmUWUnw5JRZ88IXToOpq3yyFDfdeQUMxJT+Wn/SUb0acdzE+Po3DLE7LKEEB7MmUA/AETYPQ6vfK4apdRIYDZwmda6xDXludiWd6sf+OyBm4lKy228/tVu/vFlJs2C/PnrTRcw4YLO0kxLCFEnZwJ9E9BLKdUNI8hvBqbYX6CUuhD4JzBKa53r8ipdITv53FOJ/AI8ajNRas5JZixPZfvhU1w7oDNPXhtD2+bSTEsI4Zw6A11rXa6Uuh9Yh7FscYnWeqtS6mkgRWudBLwANAc+qBxJ7tdaj2/Eus9f1kbAVv25gbd4xOi8uKyClz/byRsb99CuRRBv3BbPVTEdzC5LCOFlnJpD11qvAdY4PPeE3dcjXVyX60UNp1ozLv8gGDDl136GW/ywJ49Zialk5RUyOSGCWaP7EhYizbSEEOfPt3aKhrSCsiLoeSVc8kdTR+enisuY/8l23v5xP5Gtm/LOXUO4uKc00xJC1J9vBLrj6pZdnxmBbpIvtx9h9ofpHCko5q5h3Xjw6t40DfSN3wohROPxjRTJ2ugRq1uOnynl6Y+3svLng/Rq35xX77uYCyOlmZYQwjV8I9DPmT93b6tcrTUfpx5iTtJWThWX8ccre/G7y3tIMy0hhEv5RqBHJEDHODidC9FjYMBkt43OD+cbzbQ+33aEAeFhPH/DEKI7SjMtIYTrWTvQ7c8OBQgIcluYa615b1M2z63eRpnNxuwxfbljWDf8Zdu+EKKRWDfQTTw7dF/eGWYlpvH9njyGdm/N/Ov7E9W2WaO9nxBCgJUD3YSzQytsmje/3ctfPt1BEz8/npsYx82DI6SZlhDCLawb6G4+O3THYaOZ1pbsk1wZ3Z5nJsbSKUyaaQkh3Me6gQ4Q3NIYkV/9LHSIaZSzQ0vLbby6IZOF6zNpEdyEv02+kGv7d5JmWkIIt7NmoDtuJPpkhtEmd/hDLn2bn7NPMnN5KjuOnGLCBZ158tp+tG4W6NL3EEIIZ1kz0Bu5TW5RaQUvfrqDJd/upX2LYP41NZ4r+0ozLSGEuawT6FVLFEPaNGqb3O92H2NWYhr7jxcyZUgks0ZHExoszbSEEOazRqBXW6JYAxe0yS0oLmPemm28m5xN1zZNeffuoVzUo02Dfk0hhHAlawR6tSWKDpR/g9vkfp5xhNkr0zh6qoTpl3bngZG9CQmUbftCCM9ijUB3XKKotRHwyh/GvlTv0Xne6RLmfJzBx1sOEt2xBYtujWdAREsXFy+EEK5hjUCPSIAOsVCcD5MWG881YImi1pqPfj7IUx9v5XRJOQ9e1Zt7L+tBYICfiwsXQgjXsUagAwSFGj+qAryeo/KDJ4t4bGU6X27P5YKIliy4oT+9O7RwYaFCCNE4rBPoDWSzad5J3s/8T7ZTYdM8Pi6GaRdHSTMtIYTXkEAH9h47w6zEVH7ce5xLerZh3sT+RLZpanZZQghxXnw60MsrbPzrm7289NlOAgP8eH5SHL+Jj5Bt+0IIr+Szgb7tUAEzE1NJzcnnqpgOPHNdLB1Cg80uSwgh6s3nAr2kvIKFX2by6obdtGzahIVTBjImrqOMyoUQXs8agZ6dDHmZZ7+uZYXL5n0nmJmYSmbuaa6/sAuPj4uhlTTTEkJYhPcHespSWPUnfjnEYuk4o7OiXagXlpbzwrodLP0ui06hwbx5+2Au79PelHKFEKKxeHegZydXD3M4p7PiN7uOMWtFKjknirh1aFdmjOpDC2mmJYSwIO8O9KyNVAtzMFoARA0nv6iMZ1dn8H5KDt3aNuO/04cypLs00xJCWJd3B3rUcKodAK38YOxLrCuI5PG3viLvTCn3jejBH6/sRXATaaYlhLA27w70iAToGAencyF6DCd6TeKxlKasTttM306h/GvqYOLCw8yuUggh3MK7Ax0gKBQd1IIVnR7i6fcyKCo9xcPX9GH6pd1p4i/NtIQQvsP7Ar3qZKLKToqlZ05y8sQx3t75AT0iElhwQ396tpdmWkII3+NUoCulRgGvAP7AYq31fIfXg4C3gEFAHnCT1jrLtaVS7WQirfwoCIkg9Mw+2gHvh8xDjf0YfwlzIYSPqnNOQinlDywERgMxwGSlVIzDZXcCJ7TWPYGXgeddXShQ/WQibUOfPgYKlIIAXY7//m8a5W2FEMIbODPJnABkaq33aK1LgfeACQ7XTACWVX69HLhSNcZe+pA2aIwDiQAO9rwJAkKMk4n8A112ELQQQngjZ6ZcugDZdo9zgCG1XaO1LldK5QNtgGP2FymlpgPTASIjI8+/2qK8yl8HNH7EdAuHy5MadDqREEJYhVs/FNVaLwIWAcTHx+s6Lj9X1HBUQAhUlKKqRuQRCRLkQgiBc4F+AIiwexxe+VxN1+QopQKAMIwPR10rIgGmyohcCCFq4kygbwJ6KaW6YQT3zcAUh2uSgKnA98ANwJda6/MfgTtDRuRCCFGjOgO9ck78fmAdxrLFJVrrrUqpp4EUrXUS8C/g30qpTOA4RugLIYRwI6fm0LXWa4A1Ds89Yfd1MXCja0sTQghxPmRvvBBCWIQEuhBCWIQEuhBCWIQEuhBCWIRqrNWFdb6xUkeBffX86W1x2IXqA+SefYPcs29oyD131Vq3q+kF0wK9IZRSKVrreLPrcCe5Z98g9+wbGuueZcpFCCEsQgJdCCEswlsDfZHZBZhA7tk3yD37hka5Z6+cQxdCCHEubx2hCyGEcCCBLoQQFuHRga6UGqWU2qGUylRKzarh9SCl1H8rX/9RKRXl/ipdy4l7flAplaGUSlVKfaGU6mpGna5U1z3bXTdJKaWVUl6/xM2Ze1ZK/aby93qrUuodd9foak58b0cqpdYrpX6q/P4eY0adrqKUWqKUylVKpdfyulJK/a3y/0eqUmpgg99Ua+2RPzBa9e4GugOBwBYgxuGa3wGvV359M/Bfs+t2wz1fDjSt/Po+X7jnyutaAF8DPwDxZtftht/nXsBPQKvKx+3NrtsN97wIuK/y6xggy+y6G3jPlwIDgfRaXh8DfAIoYCjwY0Pf05NH6J5zOLX71HnPWuv1WuvCyoc/YJwg5c2c+X0GmAs8DxS7s7hG4sw93w0s1FqfANBa57q5Rldz5p41EFr5dRhw0I31uZzW+muM8yFqMwF4Sxt+AFoqpTo15D09OdBrOpy6S23XaK3LgarDqb2VM/ds706Mv+G9WZ33XPlP0Qit9Wp3FtaInPl97g30Vkp9q5T6QSk1ym3VNQ5n7nkOcItSKgfj/IX/c09ppjnfP+91cush0cJ1lFK3APHAZWbX0piUUn7AS8A0k0txtwCMaZcRGP8K+1opFae1PmlqVY1rMrBUa/2iUuoijFPQYrXWNrML8xaePEI/n8OpadTDqd3HmXtGKTUSmA2M11qXuKm2xlLXPbcAYoENSqksjLnGJC//YNSZ3+ccIElrXaa13gvsxAh4b+XMPd8JvA+gtf4eCMZoYmVVTv15Px+eHOi/HE6tlArE+NAzyeGaqsOpobEPp3aPOu9ZKXUh8E+MMPf2eVWo45611vla67Za6yitdRTG5wbjtdYp5pTrEs58b6/EGJ2jlGqLMQWzx51Fupgz97wfuBJAKdUXI9CPurVK90oCbqtc7TIUyNdaH2rQr2j2J8F1fEo8BmNkshuYXfnc0xh/oMH4Df8AyASSge5m1+yGe/4cOAL8XPkjyeyaG/ueHa7dgJevcnHy91lhTDVlAGnAzWbX7IZ7jgG+xVgB8zNwtdk1N/B+3wUOAWUY/+K6E7gXuNfu93hh5f+PNFd8X8vWfyGEsAhPnnIRQghxHiTQhRDCIiTQhRDCIiTQhRDCIiTQhRDCIiTQhRDCIiTQhRDCIv4fE11xTlCC9e8AAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "[[202  51]\n",
            " [ 51  80]]\n",
            "Accuracy: 0.734375\n",
            "Precision: 0.6106870229007634\n",
            "Recall: 0.6106870229007634\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AwjaOM_r1DaU",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 367
        },
        "outputId": "c3a5341c-40c5-4503-8f15-0125829e24cd"
      },
      "source": [
        "model.load_weights('best.h5')\n",
        "\n",
        "y_pred2=model.predict(X_test3)\n",
        "y_pred2=(y_pred2>.5).astype(int)\n",
        "\n",
        "# todo: use boot-strap estimation\n",
        "probs = model.predict_proba(X_test3)\n",
        "probs = probs[:,0]\n",
        "auc = roc_auc_score(y_test, probs)\n",
        "print('AUC: %.3f' % auc)\n",
        "fpr, tpr, thresholds = roc_curve(y_test, probs)\n",
        "pyplot.plot([0, 1], [0, 1], linestyle = '-')\n",
        "pyplot.plot(fpr, tpr, marker = '.')\n",
        "pyplot.show()\n",
        "\n",
        "cmat = metrics.confusion_matrix(y_test, y_pred2)\n",
        "print( cmat )\n",
        "print(\"Accuracy:\",metrics.accuracy_score(y_test, y_pred2))\n",
        "print(\"Precision:\",metrics.precision_score(y_test, y_pred2, pos_label=1)) \n",
        "print(\"Recall:\",metrics.recall_score(y_test, y_pred2, pos_label=1))"
      ],
      "execution_count": 320,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "AUC: 0.832\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "[[218  35]\n",
            " [ 57  74]]\n",
            "Accuracy: 0.7604166666666666\n",
            "Precision: 0.6788990825688074\n",
            "Recall: 0.5648854961832062\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "g2yb1O_c1lXW",
        "colab_type": "text"
      },
      "source": [
        "\n",
        "LSTM:\n",
        "- Accuracy: 0.7135416666666666\n",
        "- Precision: 0.591304347826087\n",
        "- Recall: 0.5190839694656488\n",
        "\n",
        "MLP:\n",
        "- Accuracy: 0.7604166666666666\n",
        "- Precision: 0.6788990825688074\n",
        "- Recall: 0.5648854961832062\n"
      ]
    }
  ]
}