GRU_Experiment3_Padding_and_Masking.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "python version:  3.7.3 (default, Mar 27 2019, 22:11:17) \n",
      "[GCC 7.3.0]\n",
      "pandas version:  0.24.2\n",
      "numpy version:  1.16.4\n",
      "scipy version:  1.3.0\n",
      "sklearn version:  0.21.2\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "keras version:  2.2.4\n",
      "keras backend:  tensorflow\n",
      "tensorflow version:  1.14.0\n"
     ]
    }
   ],
   "source": [
    "import sys\n",
    "print(\"python version: \", sys.version)\n",
    "\n",
    "import pandas as pd\n",
    "print(\"pandas version: \", pd.__version__)\n",
    "\n",
    "import numpy as np\n",
    "print('numpy version: ', np.__version__)\n",
    "\n",
    "import scipy as sp\n",
    "print('scipy version: ', sp.__version__)\n",
    "\n",
    "import sklearn as skl\n",
    "print('sklearn version: ', skl.__version__)\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import os\n",
    "os.environ[\"CUDA_DEVICE_ORDER\"] = \"PCI_BUS_ID\"\n",
    "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"\"\n",
    "\n",
    "import keras as ks\n",
    "print('keras version: ', ks.__version__)\n",
    "print('keras backend: ', ks.backend.backend())\n",
    "\n",
    "import tensorflow as tf\n",
    "print('tensorflow version: ', tf.__version__)\n",
    "\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')\n",
    "\n",
    "from keras import backend as K\n",
    "from sklearn.metrics import confusion_matrix, precision_score, recall_score\n",
    "from sklearn.metrics import roc_curve, auc, f1_score, accuracy_score, precision_recall_curve\n",
    "\n",
    "import random"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(7584, 208)\n"
     ]
    }
   ],
   "source": [
    "excel_file = r'geriomop-data-from-egems-paper-UPDATED-COL-FIX-ATC_2-and-behavior-change-022819.tsv'\n",
    "df = pd.DataFrame(pd.read_csv(excel_file, sep='\\t'))\n",
    "print(df.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Drop features\n",
    "f_drop = [\"Blank-1\", \"ws-sudden-stop\", \"ws-sudden-stop-drugs\", \"ws-prolonged-psychotropics\", \"mds-toilet-prgm-atmptd\"] + \\\n",
    "         ['Deleted-%d' % i for i in range(1, 13)] + ['Psychotropic-%d' % i for i in range(1, 6)] + \\\n",
    "         ['Psychotropic-%d-average-daily-dose' % i for i in range(1, 6)] + [\"Entry-discharge-type\"]\n",
    "\n",
    "f_drop += [\"acute_mental_change\", \"aphasia\", \"balance_toileting\", \"balance_turning_around\", \"bipolar\", \"bph\", \\\n",
    "           \"cam_motor_retardation\", \"cerebralpalsy\", \"cirrhosis\", \"cms-long-stay\", \"comatose\", \"conduct_pain_assmnt\",\\\n",
    "           \"dysrhythmias\", \"embolisms\", \"facility\", \"hemiplegia/hemiparesis\", \"hepatitis\", \"huntingtons\", \"hyperkalemia\",\\\n",
    "           \"hypertension\", \"hyperthyroidism\", \"hyponatremia\", \"hypotension\", \"hypothyroidism\", \\\n",
    "           \"limb_prosthesis_past_7_days\", \"marital-status\", \"mds-conduct-staff-assessment-mental-status\", \\\n",
    "           \"mds-dehydrated\", \"mds-delirium-scale\", \"mds-excess-weight-loss\", \"mds-long-term-memory-ok\", \\\n",
    "           \"mds-pressure-ulcer-stage-1\", \"mds-pressure-ulcer-stage-2\", \"mds-pressure-ulcer-stage-3\", \\\n",
    "           \"mds-pressure-ulcer-stage-4\", \"mds-short-term-memory-ok\", \"mds-staff-assess-pain\", \"multiple-sclerosis\", \\\n",
    "           \"neurogenic-bladder\", \"obstructive-uropathy\", \"paraplegia\", \"pneumonia\", \"ptsd\", \"pvd\", \"quadriplegia\", \\\n",
    "           \"renal-failure\", \"schizophrenia\", \"septicemia\", \"stroke\", \"tourettes\", \"traumatic-brain-injury\", \\\n",
    "           \"tuberculosis\", \"wound\", \"ws-antibiotic-anticoag-coexposure\", \"ws-diuretic-adl\", \\\n",
    "           \"ws-meclizine-psych-coexposure\", \"ws-meclizine-psych-coexposure\", \"ws-meclizine-psych-coexposure\",\\\n",
    "           \"ws-meclizine-psych-coexposure\", \"ws-meclizine-psych-coexposure\", \\\n",
    "           \"ws-psychotropic-with-not-ordered-weight-loss\", \"ws-tramadol-antidepressant-coexposure\"]\n",
    " \n",
    "# Missing value means not relevant\n",
    "f_one_hot = [\"Mds-fall-2-6-months-to-admission\", \"Mds-fall-last-month-to-admission\", \"mds-pain-last-five-days\", \\\n",
    "             \"recent-start-other-fall-risk-rx\"]\n",
    "cate_one_hot = [['Yes', 'No', 'Unable to answer']] * 3 + [['Yes', 'No']]\n",
    "\n",
    "f_period = [(\"Psychotropic-%d-Start-date\" % i, 'Psychotropic-%d-End-date' % i, 'Psychotropic-%d-Period' % i) for i in range(1, 6)]\n",
    "\n",
    "f_label = ['mds-cognitive-scale', 'mds-pain-freq-last-five-days', 'mds-pain-intensity']\n",
    "dic_label = [{'Independent': 0, 'Modified Independence': 1, 'Moderately Impaired': 2, 'Severely Impaired': 3}, \\\n",
    "             {'Continuous': 4, 'Frequent': 3, 'Occasional': 2, 'Rare': 1, 'Unable to respond': 0}, \\\n",
    "             {'None': 0, 'Mild': 1, 'Moderate': 2, 'Severe': 3, 'Very severe, horrible': 4}]\n",
    "\n",
    "f_transform = [('mds-cognitive-scale', 3, 0), ('mds-pain-freq-last-five-days', 4, 0), ('mds-pain-intensity', 4, 0)] + \\\n",
    "              [('Psychotropic-%d-Period' % i, 0, 0) for i in range(1, 6)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Encode features (originally) with no missing values\n",
    "f_drop += [\"predictor-date\", \"projected-date\", \"PPS-assessment\", \"Federal-assessment\"]\n",
    "\n",
    "f_period += [('episode-start-date', 'episode-end-date', 'episode-period'),\\\n",
    "             ('stay-start-date', 'stay-end-date', 'stay-period')]\n",
    "\n",
    "f_label += [\"mds-stay-trans-short-to-long\", \"mds-antianxiety-medication\", \\\n",
    "            \"mds-antidepressant-medication\", \"mds-antipsychotic-medication\", \"mds-antibiotic-medication\",\n",
    "            \"mds-anticoagulant-medication\", \"mds-diuretic-medication\", \\\n",
    "            \"mds-hypnotic-medication\", \"mds-behavioral-symptoms\", \"mds-behavioral-symptoms-to-others\", \\\n",
    "            \"mds-depression\", \"mds-fever\", \"mds-impaired-mobility\", \"mds-impaired-transfer\", \\\n",
    "            \"mds-internal-bleeding\", \"mds-malnutrition\", \"mds-no-problem-conditions\", \"mds-vomiting\", \\\n",
    "            \"mds-impaired-walk-in-room\", \"mds-impaired-walk-in-corridor\", \"mds-impaired-locomot-unit\", \\\n",
    "            \"mds-impaired-locomot-other\"] + [\"mds-pain-non-verbal\"]\n",
    "dic_label += [{'Yes': 1, 'No': 0}] * 22 + [{'None/Mild':0,'Moderate/Severe':1}]\n",
    "\n",
    "f_one_hot += [\"race\", \"gender\", \"mds-bims-summary-ranking\"]\n",
    "cate_one_hot += [['White', 'Black', 'Other'], \\\n",
    "                 ['Female', 'Male'], ['Intact or Moderately Intact', 'Moderate Impairment']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Features need imputation\n",
    "f_one_hot += [\"uti\", \"mdro\", \"anemia\", \"constipation\", \"alzheimers\", \"anxiety\", \"depression\", \"non-alz-dimentia\", \\\n",
    "              \"parkinsons\", \"psychosis\", \"seizure\", \"emphysema\", \"arthritis\", \"ashd\", \"cancer\", \"diabetes\", \"gerd\",\\\n",
    "              \"heart-failure\", \"osteoporosis\", \"thyroid-disorder\", \"transient-ischemic-attack\", \"mds-pain-medication\", \\\n",
    "              \"mds-received-pain-tx-non-pharm\", \"mds-received-prn-pain-medication\", \"mds-pressure-ulcer-prsnt\", \\\n",
    "              \"conduct_bims\", \"cane_or_crutch_past_7_days\", \"walker_past_7_days\", \"wheelchair_past_7_days\", \\\n",
    "              \"no_listed_mobility_device\", \"hyperlipidemia\"] + \\\n",
    "              [\"balance_while_standing\", \"balance_while_walking\", \"balance_bed_to_chair\"] + [\"mds-entered-from\"]\n",
    "cate_one_hot += [['Yes', 'No']] * 31 + \\\n",
    "                [[\"Steady at all times\", \"able to stabilize without assistance\", \\\n",
    "                  \"able to stabilize with assistance\", \"Activity did not occur\"]] * 3 + \\\n",
    "                [[\"Acute Hospital\", \"Psychiatric Hospital\", \"Another nursing home or swing bed\", \"Hospice\", \\\n",
    "                  \"Inpatient rehabilitation facility\",\n",
    "                  \"Community (private home/apt., board/care, assisted living, group home)\", \"Other\"]]\n",
    "\n",
    "f_label += [\"mds-urinary-incontinence\", \"functlimit_rom_upper\", \"functlimit_rom_lower\", \"does_resident_wander\"] + \\\n",
    "           [\"cam_inattention\", \"cam_disorganized_thought\", \"cam_altered_conc\"]\n",
    "dic_label += [{\"Always continent\": 0, \"Occasionally incontinent\": 1, \"Frequently incontinent\": 2,\\\n",
    "               \"Always incontinent\": 3, \"Catheter - not rated\": np.nan}] + \\\n",
    "             [{\"No impairment\": 0, \"Impairment on one side\": 1, \"Impairment on both sides\": 2}] * 2 + \\\n",
    "             [{\"Behavior not exhibited\": 0, \"1 to 3 days\": 1, \"4 to 6 days\": 2, \"occurred daily\": 3}] + \\\n",
    "             [{\"Behavior not present\": 0, \"Behavior present, fluctuates\": 1, \"Behavior continuously present\": 2}] * 3\n",
    "\n",
    "f_transform += [(\"mds-urinary-incontinence\", 3, 0), (\"cam_inattention\", 2, 0), (\"cam_disorganized_thought\", 2, 0), \\\n",
    "                (\"cam_altered_conc\", 2, 0), (\"functlimit_rom_upper\", 2, 0), \\\n",
    "                (\"functlimit_rom_lower\", 2, 0), (\"does_resident_wander\", 3, 0)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Set sdd=0 to NA\n",
    "for i in range(1, 6):\n",
    "    df.loc[df['psych_sdd_%d' % i] == 0, 'psych_sdd_%d' % i] = np.nan\n",
    "# These data were dropped in Experiment 1\n",
    "f_transform += [('psych_sdd_%d' % i, 0, 0) for i in range(1, 6)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Dropping min effective dose and one hot encoding ATC psychotropic drugs\n",
    "# These data were not used in Experiment 1 b/c of missing values\n",
    "f_drop += [\"min_effective_dose_%d\" % i for i in range(1, 6)]\n",
    "\n",
    "atc_list = []\n",
    "for i in range(1, 6):\n",
    "    atc_list.append(df.groupby('atc_%d' % i)['min_effective_dose_%d' % i].first())\n",
    "atc_codes = pd.concat(atc_list)\n",
    "atc_codes = atc_codes[~atc_codes.index.duplicated()].index.tolist()\n",
    "f_one_hot += [\"atc_%d\" % i for i in range(1, 6)]\n",
    "cate_one_hot += [atc_codes] * 5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# generate target\n",
    "target = df['outcome-Mds-fall-since-prior-assessment'].map({'None': 0, 'Yes': 1})\n",
    "dff = df.drop(['outcome-mds-fall-no-injury', 'outcome-mds-fall-minor-injury', 'outcome-mds-fall-major-injury', \\\n",
    "               'outcome-Mds-fall-since-prior-assessment', 'outcome-riskmaster-fall-incident'], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# train test split\n",
    "from sklearn.model_selection import GroupShuffleSplit\n",
    "def train_test_split(df, target, groupby, radio, random_state):\n",
    "    train_idx, test_idx = next(GroupShuffleSplit(train_size=radio, test_size=1-radio, random_state=random_state).split(df, target, df[groupby]))\n",
    "    return df.iloc[train_idx], df.iloc[test_idx], target.iloc[train_idx], target.iloc[test_idx]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total number of patients: 3133\n",
      "Number of patients in training set: 2193\n"
     ]
    }
   ],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(dff, target, 'patient-id', 0.7, 0)\n",
    "print('Total number of patients: %d' % len(df.groupby('patient-id')))\n",
    "print('Number of patients in training set: %d' % len(X_train.groupby('patient-id')))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# drop feature\n",
    "def drop_features(df, f):\n",
    "    df.drop(f, axis=1, inplace=True)\n",
    "    \n",
    "# one hot encoding\n",
    "def one_hot_encoding(df, features, categories):\n",
    "    for f, c in zip(features, categories):\n",
    "        df[f] = df[f].astype('category', categories=c)\n",
    "    return pd.get_dummies(df, columns=features, prefix=features)\n",
    "\n",
    "# compute date difference\n",
    "def date_diff(start, end):\n",
    "    return (pd.to_datetime(end) - pd.to_datetime(start)) / np.timedelta64(1, 'D')\n",
    "\n",
    "# generate date diff features\n",
    "def generate_period(df, f):\n",
    "    for start, end, period in f:\n",
    "        df[period] = date_diff(df[start], df[end])\n",
    "        df.drop([start, end], axis=1, inplace=True)\n",
    "        \n",
    "# label encoding\n",
    "def label_encoding(df, features, dictionaries):\n",
    "    for f, dic in zip(features, dictionaries):\n",
    "        df[f] = df[f].map(dic)\n",
    "        \n",
    "# sin&cos transform\n",
    "def col_transform(df, col, mmax, mmin, df_ref=None):\n",
    "    if(mmax <= mmin):\n",
    "        # for test set, if max&min not set, use training set value\n",
    "        if(df_ref is not None): \n",
    "            mmax = df_ref[col].max()\n",
    "            mmin = df_ref[col].max()\n",
    "        else:\n",
    "            mmax = df[col].max()\n",
    "            mmin = df[col].min()\n",
    "    angle = 0.25 * np.pi * (df[col] - mmin) / (mmax - mmin) + 0.125 * np.pi\n",
    "    return (np.cos(angle), np.sin(angle))\n",
    "\n",
    "def df_transform(df, f, df_ref=None):\n",
    "    for col, mmax, mmin in f:\n",
    "        df[col + '_x'], df[col + '_y'] = col_transform(df, col, mmax, mmin, df_ref)\n",
    "        df[col + '_x'].fillna(0, inplace=True)\n",
    "        df[col + '_y'].fillna(0, inplace=True)\n",
    "        df.drop(col, axis=1, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 1min 22s, sys: 1.36 s, total: 1min 23s\n",
      "Wall time: 25.6 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "## Training set data reshaping\n",
    "generate_period(X_train, f_period)\n",
    "label_encoding(X_train, f_label, dic_label)\n",
    "X_train_original = X_train.copy() # store value range\n",
    "df_transform(X_train, f_transform)\n",
    "X_train = one_hot_encoding(X_train, f_one_hot, cate_one_hot)\n",
    "drop_features(X_train, f_drop)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train_use = X_train.copy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Index([], dtype='object')\n"
     ]
    }
   ],
   "source": [
    "print(X_train.dtypes[X_train.dtypes == 'object'].index)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>count</th>\n",
       "      <th>percentage</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "Empty DataFrame\n",
       "Columns: [count, percentage]\n",
       "Index: []"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# missing data describe\n",
    "def missing_summary(df):\n",
    "    missing_cnt = df.isna().sum()\n",
    "    missing_data = pd.concat([missing_cnt, missing_cnt/df.shape[0]], axis=1, keys=['count', 'percentage'])\n",
    "    return missing_data[missing_data['percentage'] != 0].sort_values(by='percentage', ascending=False)\n",
    "\n",
    "missing_summary(X_train_use)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 22.3 s, sys: 83.8 ms, total: 22.4 s\n",
      "Wall time: 22.4 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "## Test set data reshaping\n",
    "generate_period(X_test, f_period)\n",
    "label_encoding(X_test, f_label, dic_label)\n",
    "df_transform(X_test, f_transform, X_train_original)\n",
    "X_test = one_hot_encoding(X_test, f_one_hot, cate_one_hot)\n",
    "drop_features(X_test, f_drop)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_test_use = X_test.copy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Index([], dtype='object')\n"
     ]
    }
   ],
   "source": [
    "print(X_test.dtypes[X_test.dtypes == 'object'].index)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>count</th>\n",
       "      <th>percentage</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "Empty DataFrame\n",
       "Columns: [count, percentage]\n",
       "Index: []"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "missing_summary(X_test_use)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(5301, 398)"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train_use.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2283, 398)"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_test_use.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>patient-id</th>\n",
       "      <th>current-stay-days</th>\n",
       "      <th>cumulative-days-in-facility</th>\n",
       "      <th>mds-stay-trans-short-to-long</th>\n",
       "      <th>age</th>\n",
       "      <th>mds-antianxiety-medication</th>\n",
       "      <th>mds-antidepressant-medication</th>\n",
       "      <th>mds-antipsychotic-medication</th>\n",
       "      <th>mds-antibiotic-medication</th>\n",
       "      <th>mds-anticoagulant-medication</th>\n",
       "      <th>...</th>\n",
       "      <th>atc_5_N06AB10</th>\n",
       "      <th>atc_5_N06AX05</th>\n",
       "      <th>atc_5_N06AX11</th>\n",
       "      <th>atc_5_N06AX12</th>\n",
       "      <th>atc_5_N06AX16</th>\n",
       "      <th>atc_5_N06AX21</th>\n",
       "      <th>atc_5_N06AX24</th>\n",
       "      <th>atc_5_N05AA01</th>\n",
       "      <th>atc_5_N05BA05</th>\n",
       "      <th>atc_5_N06AX23</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>10353</td>\n",
       "      <td>7</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>67</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>10364</td>\n",
       "      <td>7</td>\n",
       "      <td>7</td>\n",
       "      <td>1</td>\n",
       "      <td>84</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>10386</td>\n",
       "      <td>7</td>\n",
       "      <td>7</td>\n",
       "      <td>1</td>\n",
       "      <td>77</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>5000</td>\n",
       "      <td>7</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>90</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>5003</td>\n",
       "      <td>7</td>\n",
       "      <td>7</td>\n",
       "      <td>1</td>\n",
       "      <td>90</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 398 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   patient-id  current-stay-days  cumulative-days-in-facility  \\\n",
       "0       10353                  7                            7   \n",
       "1       10364                  7                            7   \n",
       "4       10386                  7                            7   \n",
       "5        5000                  7                            7   \n",
       "6        5003                  7                            7   \n",
       "\n",
       "   mds-stay-trans-short-to-long  age  mds-antianxiety-medication  \\\n",
       "0                             0   67                           1   \n",
       "1                             1   84                           0   \n",
       "4                             1   77                           1   \n",
       "5                             0   90                           0   \n",
       "6                             1   90                           0   \n",
       "\n",
       "   mds-antidepressant-medication  mds-antipsychotic-medication  \\\n",
       "0                              1                             0   \n",
       "1                              0                             1   \n",
       "4                              1                             1   \n",
       "5                              1                             1   \n",
       "6                              1                             1   \n",
       "\n",
       "   mds-antibiotic-medication  mds-anticoagulant-medication  ...  \\\n",
       "0                          0                             0  ...   \n",
       "1                          1                             0  ...   \n",
       "4                          0                             0  ...   \n",
       "5                          0                             0  ...   \n",
       "6                          1                             0  ...   \n",
       "\n",
       "   atc_5_N06AB10  atc_5_N06AX05  atc_5_N06AX11  atc_5_N06AX12  atc_5_N06AX16  \\\n",
       "0              0              0              0              0              0   \n",
       "1              0              0              0              0              0   \n",
       "4              0              0              0              0              0   \n",
       "5              0              0              0              0              0   \n",
       "6              0              0              0              0              0   \n",
       "\n",
       "   atc_5_N06AX21  atc_5_N06AX24  atc_5_N05AA01  atc_5_N05BA05  atc_5_N06AX23  \n",
       "0              0              0              0              0              0  \n",
       "1              0              0              0              0              0  \n",
       "4              0              0              0              0              0  \n",
       "5              0              0              0              0              0  \n",
       "6              0              0              0              0              0  \n",
       "\n",
       "[5 rows x 398 columns]"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train_use.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>patient-id</th>\n",
       "      <th>current-stay-days</th>\n",
       "      <th>cumulative-days-in-facility</th>\n",
       "      <th>mds-stay-trans-short-to-long</th>\n",
       "      <th>age</th>\n",
       "      <th>mds-antianxiety-medication</th>\n",
       "      <th>mds-antidepressant-medication</th>\n",
       "      <th>mds-antipsychotic-medication</th>\n",
       "      <th>mds-antibiotic-medication</th>\n",
       "      <th>mds-anticoagulant-medication</th>\n",
       "      <th>...</th>\n",
       "      <th>atc_5_N06AB10</th>\n",
       "      <th>atc_5_N06AX05</th>\n",
       "      <th>atc_5_N06AX11</th>\n",
       "      <th>atc_5_N06AX12</th>\n",
       "      <th>atc_5_N06AX16</th>\n",
       "      <th>atc_5_N06AX21</th>\n",
       "      <th>atc_5_N06AX24</th>\n",
       "      <th>atc_5_N05AA01</th>\n",
       "      <th>atc_5_N05BA05</th>\n",
       "      <th>atc_5_N06AX23</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>10379</td>\n",
       "      <td>7</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>84</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>10385</td>\n",
       "      <td>7</td>\n",
       "      <td>7</td>\n",
       "      <td>1</td>\n",
       "      <td>85</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>5032</td>\n",
       "      <td>7</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>85</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>5032</td>\n",
       "      <td>8</td>\n",
       "      <td>8</td>\n",
       "      <td>0</td>\n",
       "      <td>86</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>5032</td>\n",
       "      <td>7</td>\n",
       "      <td>7</td>\n",
       "      <td>0</td>\n",
       "      <td>86</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 398 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "    patient-id  current-stay-days  cumulative-days-in-facility  \\\n",
       "2        10379                  7                            7   \n",
       "3        10385                  7                            7   \n",
       "14        5032                  7                            7   \n",
       "15        5032                  8                            8   \n",
       "16        5032                  7                            7   \n",
       "\n",
       "    mds-stay-trans-short-to-long  age  mds-antianxiety-medication  \\\n",
       "2                              0   84                           0   \n",
       "3                              1   85                           0   \n",
       "14                             0   85                           1   \n",
       "15                             0   86                           1   \n",
       "16                             0   86                           1   \n",
       "\n",
       "    mds-antidepressant-medication  mds-antipsychotic-medication  \\\n",
       "2                               1                             0   \n",
       "3                               1                             0   \n",
       "14                              1                             0   \n",
       "15                              1                             0   \n",
       "16                              1                             0   \n",
       "\n",
       "    mds-antibiotic-medication  mds-anticoagulant-medication  ...  \\\n",
       "2                           0                             1  ...   \n",
       "3                           1                             0  ...   \n",
       "14                          1                             0  ...   \n",
       "15                          1                             0  ...   \n",
       "16                          1                             0  ...   \n",
       "\n",
       "    atc_5_N06AB10  atc_5_N06AX05  atc_5_N06AX11  atc_5_N06AX12  atc_5_N06AX16  \\\n",
       "2               0              0              0              0              0   \n",
       "3               0              0              0              0              0   \n",
       "14              0              0              0              0              0   \n",
       "15              0              0              0              0              0   \n",
       "16              0              0              0              0              0   \n",
       "\n",
       "    atc_5_N06AX21  atc_5_N06AX24  atc_5_N05AA01  atc_5_N05BA05  atc_5_N06AX23  \n",
       "2               0              0              0              0              0  \n",
       "3               0              0              0              0              0  \n",
       "14              0              0              0              0              0  \n",
       "15              0              0              0              0              0  \n",
       "16              0              0              0              0              0  \n",
       "\n",
       "[5 rows x 398 columns]"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_test_use.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "# The main difference in this experiment from experiment 2 is that it  \n",
    "# includes clincally important variables (based on domain knowledge), \n",
    "# imputing them if neede, and drops a bunch of likely non-relevant features \n",
    "# (TODO:recall if this is based on feature selection + expert knowledge?)\n",
    "from sklearn.preprocessing import MinMaxScaler\n",
    "from keras.models import Sequential\n",
    "from keras.layers import LSTM, GRU, Dense, Dropout, Masking, TimeDistributed\n",
    "from numpy.random import seed\n",
    "from tensorflow import set_random_seed\n",
    "from copy import deepcopy\n",
    "from keras.preprocessing.sequence import pad_sequences"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "def scaler_transform(X_train_use, X_test_use):    \n",
    "    scaler = MinMaxScaler(feature_range=(0, 1)).fit(X_train_use.drop('patient-id', axis=1))\n",
    "    X_train_transformed = pd.DataFrame(scaler.transform(X_train_use.drop('patient-id', axis=1)), \\\n",
    "                                       columns=X_train_use.columns[1:], index=X_train_use.index)                                   \n",
    "    X_train_transformed['patient-id'] = X_train_use['patient-id']\n",
    "    X_test_transformed = pd.DataFrame(scaler.transform(X_test_use.drop('patient-id', axis=1)), \\\n",
    "                                       columns=X_test_use.columns[1:], index=X_test_use.index) \n",
    "    X_test_transformed['patient-id'] = X_test_use['patient-id']\n",
    "    return X_train_transformed, X_test_transformed"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "def add_previous_outcome(X, Y):\n",
    "    X = X.copy()\n",
    "    x1, x2 = [], []\n",
    "    pids = X['patient-id'].unique()\n",
    "    for pid in pids:\n",
    "        df = Y[X[X['patient-id']==pid].index]\n",
    "        if(df.shape[0] == 1):\n",
    "            x1.append(pd.Series([0]))\n",
    "            x2.append(pd.Series([0]))\n",
    "        else:\n",
    "            # input 1=(1, 0), 0=(0, 1), NA=(0, 0)\n",
    "            x1.append(pd.Series([0]))\n",
    "            x1.append(df[:-1])\n",
    "            x2.append(pd.Series([0]))\n",
    "            x2.append(df[:-1].map({0: 1, 1: 0}))\n",
    "    X['input_outcome_x'] = pd.Series(pd.concat(x1, ignore_index=True).tolist(), index=X.index)\n",
    "    X['input_outcome_y'] = pd.Series(pd.concat(x2, ignore_index=True).tolist(), index=X.index)\n",
    "    return X"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "## This step converts the outcome to a point in x,y coordinates where fall (1) = (1,0); no fall (0) = (0,1); and NA = (0,0) \n",
    "X_train_use = add_previous_outcome(X_train_use, y_train)\n",
    "X_test_use = add_previous_outcome(X_test_use, y_test)\n",
    "\n",
    "## This step transforms the data to range between 0 and 1\n",
    "X_train_transformed, X_test_transformed = scaler_transform(X_train_use, X_test_use)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "def generate_nested_list(X, Y):\n",
    "    x_list, y_list = [], [] \n",
    "    pids = X['patient-id'].unique()\n",
    "    for pid in pids:\n",
    "        df = X[X['patient-id']==pid]\n",
    "        y_list.append(Y[df.index].values)\n",
    "        x_list.append([record for record in df.drop('patient-id', axis=1).values])\n",
    "    return x_list, y_list\n",
    "\n",
    "def mygenerator(x_list, y_list=None):\n",
    "    if(y_list is not None):\n",
    "        while True:\n",
    "            for x, y in zip(x_list, y_list):\n",
    "                yield np.array(x).reshape((len(x), 1, x[0].shape[0])), y\n",
    "    else:\n",
    "        while True:\n",
    "            for x in x_list:\n",
    "                yield np.array(x).reshape((len(x), 1, x[0].shape[0])) \n",
    "                \n",
    "def fit_model(model, epoch, X_train, y_train, X_valid=None, y_valid=None, verbose=0):\n",
    "    history = model.fit(X_train, y_train, epochs=epoch, verbose=verbose)\n",
    "    if(verbose != 0):\n",
    "        plot_history(history)\n",
    "    return history\n",
    "\n",
    "def plot_roc_curve(fpr, tpr):\n",
    "    plt.figure()\n",
    "    plt.plot(fpr, tpr, color='darkorange', label='ROC curve (area = %0.2f)' % auc(fpr, tpr))\n",
    "    plt.plot([0, 1], [0, 1], color='navy', linestyle='--')\n",
    "    plt.xlim([0.0, 1.0])\n",
    "    plt.ylim([0.0, 1.05])\n",
    "    plt.xlabel('False Positive Rate')\n",
    "    plt.ylabel('True Positive Rate')\n",
    "    plt.legend(loc=\"lower right\")\n",
    "    plt.show()\n",
    "    \n",
    "def plot_history(history, loss=None, val_loss=None):\n",
    "    if(history is not None):\n",
    "        loss = history.history['loss']\n",
    "    plt.figure()\n",
    "    plt.plot(loss)\n",
    "    plt.title('model loss')\n",
    "    plt.ylabel('loss')\n",
    "    plt.xlabel('epoch')\n",
    "    if('val_loss' in history.history):\n",
    "        val_loss = history.history['val_loss']\n",
    "        plt.plot(val_loss)\n",
    "    plt.legend(['train', 'test'], loc='upper right')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "def pad_data(X, maxlen):\n",
    "    return pad_sequences(X, maxlen=maxlen, dtype='float64', padding='post')\n",
    "\n",
    "def pad_all(X_train, y_train, X_test, y_test, maxlen):\n",
    "    return pad_data(X_train, maxlen), pad_data(y_train, maxlen), pad_data(X_test, maxlen), pad_data(y_test, maxlen)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "def auroc(y_predict, y_true, plot=False):\n",
    "    fpr, tpr, thresholds = roc_curve(y_true, y_predict, pos_label=1)    \n",
    "    if(plot):\n",
    "        plot_roc_curve(fpr, tpr)\n",
    "    return auc(fpr, tpr)\n",
    "\n",
    "\n",
    "def find_best_f1(y_predict, y_true):\n",
    "    fpr, tpr, thresholds = roc_curve(y_true, y_predict, pos_label=1) \n",
    "    max_f1, th = 0, 0\n",
    "    for threshold in thresholds:\n",
    "        f1 = f1_score(y_true, y_predict > threshold)\n",
    "        if(f1 > max_f1):\n",
    "            max_f1 = f1\n",
    "            th = threshold\n",
    "    return max_f1, th\n",
    "\n",
    "def other_metrics(y_predict, y_true, threshold):\n",
    "    y_threshold = y_predict > threshold\n",
    "    tn, fp, fn, tp = confusion_matrix(y_true, y_threshold).ravel()\n",
    "    return precision_score(y_true, y_threshold), recall_score(y_true, y_threshold),\\\n",
    "            accuracy_score(y_true, y_threshold), float(tn) / (tn + fp)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import GroupKFold\n",
    "def kfold_validation(unit, ratio, epoch, X, y, groupby, n_features, k=5, verbose=0, max_timestamps=20):\n",
    "    auroc_list = []\n",
    "    other_metric_list = []\n",
    "    for train_idx, valid_idx in GroupKFold(n_splits=5).split(X, y, X[groupby]):\n",
    "        X_train, X_valid, y_train, y_valid = X.iloc[train_idx], X.iloc[valid_idx], y.iloc[train_idx], y.iloc[valid_idx]\n",
    "        X_train, y_train = generate_nested_list(X_train, y_train)\n",
    "        X_valid, y_valid = generate_nested_list(X_valid, y_valid)\n",
    "        X_train_pad, y_train_pad, X_valid_pad, y_valid_pad = pad_all(X_train, y_train, X_valid, y_valid, max_timestamps)\n",
    "        y_train_pad = y_train_pad.reshape(y_train_pad.shape[0], y_train_pad.shape[1], 1)\n",
    "        seed(1)\n",
    "        set_random_seed(2)\n",
    "        model = build_model(unit, ratio, max_timestamps, n_features)\n",
    "        fit_model(model, epoch, X_train_pad, y_train_pad, verbose=verbose)\n",
    "        y_valid_len = [len(y) for y in y_valid]\n",
    "        y_predict = model.predict(X_valid_pad)\n",
    "        y_predict = truncate_prediction(y_predict, y_valid_len)\n",
    "        y_true = np.concatenate(y_valid)\n",
    "        auroc_list.append(auroc(y_predict, y_true))\n",
    "        f1, th = find_best_f1(y_predict, y_true)\n",
    "        precision, recall, accuracy, specificity = other_metrics(y_predict, y_true, th)\n",
    "        other_metric_list.append((precision, recall, accuracy, specificity,f1,th))\n",
    "    return (auroc_list, other_metric_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "# shape: (num_of_patients, num_of_records, num_of_features)\n",
    "X_train_list, y_train_list = generate_nested_list(X_train_transformed, y_train)\n",
    "X_test_list, y_test_list = generate_nested_list(X_test_transformed, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "## comparing this to the experiment 1 transformed dataset to confirm that this has \n",
    "## diagnoses and ATC codes while the other does not\n",
    "X_train_transformed.to_csv (r'/tmp/exp3_train_transformed_df.csv', index = False, header=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(5301, 400)"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train_transformed.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(2283, 400)"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_test_transformed.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "19\n"
     ]
    }
   ],
   "source": [
    "max_timestamps = df.groupby(['patient-id']).size().max()\n",
    "print(max_timestamps)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "399\n"
     ]
    }
   ],
   "source": [
    "n_features = X_train_list[0][0].shape[0]\n",
    "print(n_features)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "## function to build an GRU\n",
    "from keras import regularizers\n",
    "from keras.optimizers import Adagrad, Adadelta\n",
    "def build_model(units, dropout_ratio, n_timesteps, n_dimensions):\n",
    "    model = Sequential()\n",
    "    model.add(Masking(input_shape=(n_timesteps, n_dimensions)))\n",
    "    model.add(GRU(units, return_sequences=True, recurrent_dropout=dropout_ratio, dropout=0.5))\n",
    "    model.add(TimeDistributed(Dense(1, activation='sigmoid')))\n",
    "    model.compile(loss='binary_crossentropy', optimizer='adagrad', metrics=['binary_accuracy'])\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "def truncate_prediction(y_, y_len):\n",
    "    y_concat = []\n",
    "    for i, y in enumerate(y_):\n",
    "        y_concat.append(np.concatenate(y[:y_len[i], :]))\n",
    "    y_concat = np.concatenate(y_concat)\n",
    "    return y_concat"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "## the data is padded so that the final shape is as follows with the outcome data array having the same\n",
    "## shape but only one value\n",
    "#  array([[patient 1 data record 1],  \n",
    "#         [patient 1 data record 2],\n",
    "#         ...\n",
    "#         [patient 1 data record 19],\n",
    "#        ],\n",
    "#        [[patient 2 data record 1],  \n",
    "#         [patient 2 data record 2],\n",
    "#         ...\n",
    "#         [patient 2 data record 19],\n",
    "#        ],\n",
    "#        ...\n",
    "#       )\n",
    "# \n",
    "X_train_pad, y_train_pad, X_test_pad, y_test_pad = pad_all(X_train_list, y_train_list, X_test_list, y_test_list, max_timestamps)\n",
    "y_train_pad = y_train_pad.reshape((2193, 19, 1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING: Logging before flag parsing goes to stderr.\n",
      "W0424 07:52:14.801159 140032764692288 deprecation_wrapper.py:119] From /usr/local/anaconda3/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:74: The name tf.get_default_graph is deprecated. Please use tf.compat.v1.get_default_graph instead.\n",
      "\n",
      "W0424 07:52:14.804247 140032764692288 deprecation_wrapper.py:119] From /usr/local/anaconda3/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:517: The name tf.placeholder is deprecated. Please use tf.compat.v1.placeholder instead.\n",
      "\n",
      "W0424 07:52:14.821575 140032764692288 deprecation_wrapper.py:119] From /usr/local/anaconda3/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:4138: The name tf.random_uniform is deprecated. Please use tf.random.uniform instead.\n",
      "\n",
      "W0424 07:52:14.896182 140032764692288 deprecation_wrapper.py:119] From /usr/local/anaconda3/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:133: The name tf.placeholder_with_default is deprecated. Please use tf.compat.v1.placeholder_with_default instead.\n",
      "\n",
      "W0424 07:52:14.906334 140032764692288 deprecation.py:506] From /usr/local/anaconda3/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:3445: calling dropout (from tensorflow.python.ops.nn_ops) with keep_prob is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use `rate` instead of `keep_prob`. Rate should be set to `rate = 1 - keep_prob`.\n",
      "W0424 07:52:15.179850 140032764692288 deprecation.py:323] From /usr/local/anaconda3/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:2974: add_dispatch_support.<locals>.wrapper (from tensorflow.python.ops.array_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use tf.where in 2.0, which has the same broadcast rule as np.where\n",
      "W0424 07:52:15.234236 140032764692288 deprecation_wrapper.py:119] From /usr/local/anaconda3/lib/python3.7/site-packages/keras/optimizers.py:790: The name tf.train.Optimizer is deprecated. Please use tf.compat.v1.train.Optimizer instead.\n",
      "\n",
      "W0424 07:52:15.252321 140032764692288 deprecation_wrapper.py:119] From /usr/local/anaconda3/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py:3376: The name tf.log is deprecated. Please use tf.math.log instead.\n",
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/175\n",
      "2193/2193 [==============================] - 3s 1ms/step - loss: 0.4590 - binary_accuracy: 0.8223\n",
      "Epoch 2/175\n",
      "2193/2193 [==============================] - 2s 910us/step - loss: 0.4424 - binary_accuracy: 0.8171\n",
      "Epoch 3/175\n",
      "2193/2193 [==============================] - 2s 910us/step - loss: 0.4398 - binary_accuracy: 0.8189\n",
      "Epoch 4/175\n",
      "2193/2193 [==============================] - 2s 917us/step - loss: 0.4279 - binary_accuracy: 0.8209\n",
      "Epoch 5/175\n",
      "2193/2193 [==============================] - 2s 891us/step - loss: 0.4200 - binary_accuracy: 0.8314\n",
      "Epoch 6/175\n",
      "2193/2193 [==============================] - 2s 881us/step - loss: 0.4251 - binary_accuracy: 0.8212\n",
      "Epoch 7/175\n",
      "2193/2193 [==============================] - 2s 915us/step - loss: 0.4215 - binary_accuracy: 0.8251\n",
      "Epoch 8/175\n",
      "2193/2193 [==============================] - 2s 933us/step - loss: 0.4170 - binary_accuracy: 0.8285\n",
      "Epoch 9/175\n",
      "2193/2193 [==============================] - 2s 893us/step - loss: 0.4147 - binary_accuracy: 0.8272\n",
      "Epoch 10/175\n",
      "2193/2193 [==============================] - 2s 898us/step - loss: 0.4143 - binary_accuracy: 0.8273\n",
      "Epoch 11/175\n",
      "2193/2193 [==============================] - 2s 885us/step - loss: 0.4145 - binary_accuracy: 0.8242\n",
      "Epoch 12/175\n",
      "2193/2193 [==============================] - 2s 886us/step - loss: 0.4188 - binary_accuracy: 0.8280\n",
      "Epoch 13/175\n",
      "2193/2193 [==============================] - 2s 888us/step - loss: 0.4195 - binary_accuracy: 0.8223\n",
      "Epoch 14/175\n",
      "2193/2193 [==============================] - 2s 895us/step - loss: 0.4146 - binary_accuracy: 0.8314\n",
      "Epoch 15/175\n",
      "2193/2193 [==============================] - 2s 907us/step - loss: 0.4126 - binary_accuracy: 0.8289\n",
      "Epoch 16/175\n",
      "2193/2193 [==============================] - 2s 895us/step - loss: 0.4112 - binary_accuracy: 0.8297\n",
      "Epoch 17/175\n",
      "2193/2193 [==============================] - 2s 894us/step - loss: 0.4112 - binary_accuracy: 0.8263\n",
      "Epoch 18/175\n",
      "2193/2193 [==============================] - 2s 906us/step - loss: 0.4138 - binary_accuracy: 0.8288\n",
      "Epoch 19/175\n",
      "2193/2193 [==============================] - 2s 921us/step - loss: 0.4115 - binary_accuracy: 0.8303\n",
      "Epoch 20/175\n",
      "2193/2193 [==============================] - 2s 907us/step - loss: 0.4086 - binary_accuracy: 0.8276\n",
      "Epoch 21/175\n",
      "2193/2193 [==============================] - 2s 893us/step - loss: 0.4067 - binary_accuracy: 0.8296\n",
      "Epoch 22/175\n",
      "2193/2193 [==============================] - 2s 892us/step - loss: 0.4097 - binary_accuracy: 0.8284\n",
      "Epoch 23/175\n",
      "2193/2193 [==============================] - 2s 894us/step - loss: 0.4127 - binary_accuracy: 0.8253\n",
      "Epoch 24/175\n",
      "2193/2193 [==============================] - 2s 900us/step - loss: 0.4122 - binary_accuracy: 0.8283\n",
      "Epoch 25/175\n",
      "2193/2193 [==============================] - 2s 884us/step - loss: 0.4040 - binary_accuracy: 0.8350\n",
      "Epoch 26/175\n",
      "2193/2193 [==============================] - 2s 897us/step - loss: 0.4131 - binary_accuracy: 0.8252\n",
      "Epoch 27/175\n",
      "2193/2193 [==============================] - 2s 889us/step - loss: 0.4080 - binary_accuracy: 0.8299\n",
      "Epoch 28/175\n",
      "2193/2193 [==============================] - 2s 892us/step - loss: 0.4038 - binary_accuracy: 0.8340\n",
      "Epoch 29/175\n",
      "2193/2193 [==============================] - 2s 891us/step - loss: 0.4092 - binary_accuracy: 0.8287\n",
      "Epoch 30/175\n",
      "2193/2193 [==============================] - 2s 917us/step - loss: 0.4110 - binary_accuracy: 0.8283\n",
      "Epoch 31/175\n",
      "2193/2193 [==============================] - 2s 912us/step - loss: 0.4109 - binary_accuracy: 0.8298\n",
      "Epoch 32/175\n",
      "2193/2193 [==============================] - 2s 892us/step - loss: 0.4112 - binary_accuracy: 0.8281\n",
      "Epoch 33/175\n",
      "2193/2193 [==============================] - 2s 903us/step - loss: 0.4071 - binary_accuracy: 0.8336\n",
      "Epoch 34/175\n",
      "2193/2193 [==============================] - 2s 884us/step - loss: 0.4082 - binary_accuracy: 0.8297\n",
      "Epoch 35/175\n",
      "2193/2193 [==============================] - 2s 892us/step - loss: 0.4043 - binary_accuracy: 0.8296\n",
      "Epoch 36/175\n",
      "2193/2193 [==============================] - 2s 885us/step - loss: 0.4064 - binary_accuracy: 0.8292\n",
      "Epoch 37/175\n",
      "2193/2193 [==============================] - 2s 905us/step - loss: 0.4053 - binary_accuracy: 0.8327\n",
      "Epoch 38/175\n",
      "2193/2193 [==============================] - 2s 898us/step - loss: 0.4056 - binary_accuracy: 0.8272\n",
      "Epoch 39/175\n",
      "2193/2193 [==============================] - 2s 907us/step - loss: 0.4045 - binary_accuracy: 0.8306\n",
      "Epoch 40/175\n",
      "2193/2193 [==============================] - 2s 889us/step - loss: 0.4056 - binary_accuracy: 0.8280\n",
      "Epoch 41/175\n",
      "2193/2193 [==============================] - 2s 894us/step - loss: 0.4044 - binary_accuracy: 0.8303\n",
      "Epoch 42/175\n",
      "2193/2193 [==============================] - 2s 903us/step - loss: 0.3969 - binary_accuracy: 0.8342\n",
      "Epoch 43/175\n",
      "2193/2193 [==============================] - 2s 910us/step - loss: 0.4078 - binary_accuracy: 0.8279\n",
      "Epoch 44/175\n",
      "2193/2193 [==============================] - 2s 898us/step - loss: 0.4037 - binary_accuracy: 0.8332\n",
      "Epoch 45/175\n",
      "2193/2193 [==============================] - 2s 900us/step - loss: 0.3992 - binary_accuracy: 0.8368\n",
      "Epoch 46/175\n",
      "2193/2193 [==============================] - 2s 887us/step - loss: 0.3978 - binary_accuracy: 0.8334\n",
      "Epoch 47/175\n",
      "2193/2193 [==============================] - 2s 883us/step - loss: 0.4001 - binary_accuracy: 0.8320\n",
      "Epoch 48/175\n",
      "2193/2193 [==============================] - 2s 890us/step - loss: 0.3981 - binary_accuracy: 0.8321\n",
      "Epoch 49/175\n",
      "2193/2193 [==============================] - 2s 875us/step - loss: 0.3971 - binary_accuracy: 0.8339\n",
      "Epoch 50/175\n",
      "2193/2193 [==============================] - 2s 880us/step - loss: 0.3971 - binary_accuracy: 0.8316\n",
      "Epoch 51/175\n",
      "2193/2193 [==============================] - 2s 884us/step - loss: 0.3949 - binary_accuracy: 0.8325\n",
      "Epoch 52/175\n",
      "2193/2193 [==============================] - 2s 887us/step - loss: 0.4009 - binary_accuracy: 0.8307\n",
      "Epoch 53/175\n",
      "2193/2193 [==============================] - 2s 886us/step - loss: 0.3994 - binary_accuracy: 0.8307\n",
      "Epoch 54/175\n",
      "2193/2193 [==============================] - 2s 900us/step - loss: 0.4011 - binary_accuracy: 0.8351\n",
      "Epoch 55/175\n",
      "2193/2193 [==============================] - 2s 886us/step - loss: 0.4029 - binary_accuracy: 0.8294\n",
      "Epoch 56/175\n",
      "2193/2193 [==============================] - 2s 885us/step - loss: 0.3999 - binary_accuracy: 0.8298\n",
      "Epoch 57/175\n",
      "2193/2193 [==============================] - 2s 886us/step - loss: 0.3972 - binary_accuracy: 0.8339\n",
      "Epoch 58/175\n",
      "2193/2193 [==============================] - 2s 896us/step - loss: 0.3954 - binary_accuracy: 0.8320\n",
      "Epoch 59/175\n",
      "2193/2193 [==============================] - 2s 906us/step - loss: 0.3988 - binary_accuracy: 0.8347\n",
      "Epoch 60/175\n",
      "2193/2193 [==============================] - 2s 892us/step - loss: 0.4036 - binary_accuracy: 0.8307\n",
      "Epoch 61/175\n",
      "2193/2193 [==============================] - 2s 886us/step - loss: 0.4005 - binary_accuracy: 0.8277\n",
      "Epoch 62/175\n",
      "2193/2193 [==============================] - 2s 897us/step - loss: 0.3959 - binary_accuracy: 0.8366\n",
      "Epoch 63/175\n",
      "2193/2193 [==============================] - 2s 896us/step - loss: 0.4040 - binary_accuracy: 0.8305\n",
      "Epoch 64/175\n",
      "2193/2193 [==============================] - 2s 890us/step - loss: 0.3906 - binary_accuracy: 0.8366\n",
      "Epoch 65/175\n",
      "2193/2193 [==============================] - 2s 908us/step - loss: 0.3943 - binary_accuracy: 0.8358\n",
      "Epoch 66/175\n",
      "2193/2193 [==============================] - 2s 892us/step - loss: 0.3933 - binary_accuracy: 0.8371\n",
      "Epoch 67/175\n",
      "2193/2193 [==============================] - 2s 883us/step - loss: 0.3952 - binary_accuracy: 0.8324\n",
      "Epoch 68/175\n",
      "2193/2193 [==============================] - 2s 888us/step - loss: 0.3972 - binary_accuracy: 0.8318\n",
      "Epoch 69/175\n",
      "2193/2193 [==============================] - 2s 894us/step - loss: 0.3934 - binary_accuracy: 0.8365\n",
      "Epoch 70/175\n",
      "2193/2193 [==============================] - 2s 892us/step - loss: 0.3970 - binary_accuracy: 0.8347\n",
      "Epoch 71/175\n",
      "2193/2193 [==============================] - 2s 888us/step - loss: 0.3956 - binary_accuracy: 0.8346\n",
      "Epoch 72/175\n",
      "2193/2193 [==============================] - 2s 897us/step - loss: 0.3964 - binary_accuracy: 0.8363\n",
      "Epoch 73/175\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2193/2193 [==============================] - 2s 882us/step - loss: 0.3918 - binary_accuracy: 0.8348\n",
      "Epoch 74/175\n",
      "2193/2193 [==============================] - 2s 898us/step - loss: 0.3922 - binary_accuracy: 0.8337\n",
      "Epoch 75/175\n",
      "2193/2193 [==============================] - 2s 918us/step - loss: 0.3953 - binary_accuracy: 0.8356\n",
      "Epoch 76/175\n",
      "2193/2193 [==============================] - 2s 893us/step - loss: 0.3947 - binary_accuracy: 0.8365\n",
      "Epoch 77/175\n",
      "2193/2193 [==============================] - 2s 896us/step - loss: 0.3951 - binary_accuracy: 0.8364\n",
      "Epoch 78/175\n",
      "2193/2193 [==============================] - 2s 933us/step - loss: 0.3974 - binary_accuracy: 0.8341\n",
      "Epoch 79/175\n",
      "2193/2193 [==============================] - 2s 897us/step - loss: 0.3922 - binary_accuracy: 0.8355\n",
      "Epoch 80/175\n",
      "2193/2193 [==============================] - 2s 880us/step - loss: 0.3966 - binary_accuracy: 0.8379\n",
      "Epoch 81/175\n",
      "2193/2193 [==============================] - 2s 889us/step - loss: 0.3922 - binary_accuracy: 0.8337\n",
      "Epoch 82/175\n",
      "2193/2193 [==============================] - 2s 886us/step - loss: 0.3941 - binary_accuracy: 0.8365\n",
      "Epoch 83/175\n",
      "2193/2193 [==============================] - 2s 898us/step - loss: 0.3974 - binary_accuracy: 0.8280\n",
      "Epoch 84/175\n",
      "2193/2193 [==============================] - 2s 889us/step - loss: 0.3939 - binary_accuracy: 0.8373\n",
      "Epoch 85/175\n",
      "2193/2193 [==============================] - 2s 886us/step - loss: 0.3914 - binary_accuracy: 0.8392\n",
      "Epoch 86/175\n",
      "2193/2193 [==============================] - 2s 891us/step - loss: 0.3937 - binary_accuracy: 0.8331\n",
      "Epoch 87/175\n",
      "2193/2193 [==============================] - 2s 921us/step - loss: 0.3914 - binary_accuracy: 0.8334\n",
      "Epoch 88/175\n",
      "2193/2193 [==============================] - 2s 895us/step - loss: 0.3950 - binary_accuracy: 0.8313\n",
      "Epoch 89/175\n",
      "2193/2193 [==============================] - 2s 897us/step - loss: 0.3970 - binary_accuracy: 0.8284\n",
      "Epoch 90/175\n",
      "2193/2193 [==============================] - 2s 884us/step - loss: 0.3997 - binary_accuracy: 0.8308\n",
      "Epoch 91/175\n",
      "2193/2193 [==============================] - 2s 883us/step - loss: 0.3945 - binary_accuracy: 0.8357\n",
      "Epoch 92/175\n",
      "2193/2193 [==============================] - 2s 880us/step - loss: 0.3955 - binary_accuracy: 0.8363\n",
      "Epoch 93/175\n",
      "2193/2193 [==============================] - 2s 891us/step - loss: 0.3917 - binary_accuracy: 0.8347\n",
      "Epoch 94/175\n",
      "2193/2193 [==============================] - 2s 892us/step - loss: 0.3977 - binary_accuracy: 0.8326\n",
      "Epoch 95/175\n",
      "2193/2193 [==============================] - 2s 885us/step - loss: 0.3989 - binary_accuracy: 0.8322\n",
      "Epoch 96/175\n",
      "2193/2193 [==============================] - 2s 889us/step - loss: 0.3927 - binary_accuracy: 0.8345\n",
      "Epoch 97/175\n",
      "2193/2193 [==============================] - 2s 885us/step - loss: 0.3933 - binary_accuracy: 0.8334\n",
      "Epoch 98/175\n",
      "2193/2193 [==============================] - 2s 885us/step - loss: 0.3909 - binary_accuracy: 0.8375\n",
      "Epoch 99/175\n",
      "2193/2193 [==============================] - 2s 894us/step - loss: 0.3864 - binary_accuracy: 0.8424\n",
      "Epoch 100/175\n",
      "2193/2193 [==============================] - 2s 913us/step - loss: 0.3898 - binary_accuracy: 0.8345\n",
      "Epoch 101/175\n",
      "2193/2193 [==============================] - 2s 891us/step - loss: 0.3868 - binary_accuracy: 0.8369\n",
      "Epoch 102/175\n",
      "2193/2193 [==============================] - 2s 887us/step - loss: 0.3891 - binary_accuracy: 0.8359\n",
      "Epoch 103/175\n",
      "2193/2193 [==============================] - 2s 896us/step - loss: 0.3896 - binary_accuracy: 0.8401\n",
      "Epoch 104/175\n",
      "2193/2193 [==============================] - 2s 884us/step - loss: 0.3934 - binary_accuracy: 0.8340\n",
      "Epoch 105/175\n",
      "2193/2193 [==============================] - 2s 889us/step - loss: 0.3876 - binary_accuracy: 0.8376\n",
      "Epoch 106/175\n",
      "2193/2193 [==============================] - 2s 896us/step - loss: 0.3907 - binary_accuracy: 0.8377\n",
      "Epoch 107/175\n",
      "2193/2193 [==============================] - 2s 892us/step - loss: 0.3880 - binary_accuracy: 0.8389\n",
      "Epoch 108/175\n",
      "2193/2193 [==============================] - 2s 885us/step - loss: 0.3893 - binary_accuracy: 0.8352\n",
      "Epoch 109/175\n",
      "2193/2193 [==============================] - 2s 898us/step - loss: 0.3916 - binary_accuracy: 0.8345\n",
      "Epoch 110/175\n",
      "2193/2193 [==============================] - 2s 891us/step - loss: 0.3925 - binary_accuracy: 0.8380\n",
      "Epoch 111/175\n",
      "2193/2193 [==============================] - 2s 890us/step - loss: 0.3906 - binary_accuracy: 0.8361\n",
      "Epoch 112/175\n",
      "2193/2193 [==============================] - 2s 899us/step - loss: 0.3844 - binary_accuracy: 0.8415\n",
      "Epoch 113/175\n",
      "2193/2193 [==============================] - 2s 892us/step - loss: 0.3941 - binary_accuracy: 0.8309\n",
      "Epoch 114/175\n",
      "2193/2193 [==============================] - 2s 884us/step - loss: 0.3855 - binary_accuracy: 0.8384\n",
      "Epoch 115/175\n",
      "2193/2193 [==============================] - 2s 886us/step - loss: 0.3862 - binary_accuracy: 0.8413\n",
      "Epoch 116/175\n",
      "2193/2193 [==============================] - 2s 892us/step - loss: 0.3941 - binary_accuracy: 0.8307\n",
      "Epoch 117/175\n",
      "2193/2193 [==============================] - 2s 886us/step - loss: 0.3846 - binary_accuracy: 0.8403\n",
      "Epoch 118/175\n",
      "2193/2193 [==============================] - 2s 885us/step - loss: 0.3895 - binary_accuracy: 0.8392\n",
      "Epoch 119/175\n",
      "2193/2193 [==============================] - 2s 882us/step - loss: 0.3853 - binary_accuracy: 0.8391\n",
      "Epoch 120/175\n",
      "2193/2193 [==============================] - 2s 911us/step - loss: 0.3937 - binary_accuracy: 0.8322\n",
      "Epoch 121/175\n",
      "2193/2193 [==============================] - 2s 891us/step - loss: 0.3845 - binary_accuracy: 0.8351\n",
      "Epoch 122/175\n",
      "2193/2193 [==============================] - 2s 882us/step - loss: 0.3891 - binary_accuracy: 0.8413\n",
      "Epoch 123/175\n",
      "2193/2193 [==============================] - 2s 885us/step - loss: 0.3882 - binary_accuracy: 0.8392\n",
      "Epoch 124/175\n",
      "2193/2193 [==============================] - 2s 897us/step - loss: 0.3849 - binary_accuracy: 0.8427\n",
      "Epoch 125/175\n",
      "2193/2193 [==============================] - 2s 901us/step - loss: 0.3853 - binary_accuracy: 0.8409\n",
      "Epoch 126/175\n",
      "2193/2193 [==============================] - 2s 896us/step - loss: 0.3898 - binary_accuracy: 0.8334\n",
      "Epoch 127/175\n",
      "2193/2193 [==============================] - 2s 896us/step - loss: 0.3882 - binary_accuracy: 0.8362\n",
      "Epoch 128/175\n",
      "2193/2193 [==============================] - 2s 890us/step - loss: 0.3871 - binary_accuracy: 0.8365\n",
      "Epoch 129/175\n",
      "2193/2193 [==============================] - 2s 890us/step - loss: 0.3831 - binary_accuracy: 0.8395\n",
      "Epoch 130/175\n",
      "2193/2193 [==============================] - 2s 882us/step - loss: 0.3876 - binary_accuracy: 0.8400\n",
      "Epoch 131/175\n",
      "2193/2193 [==============================] - 2s 879us/step - loss: 0.3859 - binary_accuracy: 0.8367\n",
      "Epoch 132/175\n",
      "2193/2193 [==============================] - 2s 883us/step - loss: 0.3870 - binary_accuracy: 0.8395\n",
      "Epoch 133/175\n",
      "2193/2193 [==============================] - 2s 896us/step - loss: 0.3849 - binary_accuracy: 0.8429\n",
      "Epoch 134/175\n",
      "2193/2193 [==============================] - 2s 890us/step - loss: 0.3881 - binary_accuracy: 0.8416\n",
      "Epoch 135/175\n",
      "2193/2193 [==============================] - 2s 894us/step - loss: 0.3886 - binary_accuracy: 0.8351\n",
      "Epoch 136/175\n",
      "2193/2193 [==============================] - 2s 896us/step - loss: 0.3890 - binary_accuracy: 0.8364\n",
      "Epoch 137/175\n",
      "2193/2193 [==============================] - 2s 885us/step - loss: 0.3844 - binary_accuracy: 0.8395\n",
      "Epoch 138/175\n",
      "2193/2193 [==============================] - 2s 892us/step - loss: 0.3880 - binary_accuracy: 0.8379\n",
      "Epoch 139/175\n",
      "2193/2193 [==============================] - 2s 899us/step - loss: 0.3896 - binary_accuracy: 0.8374\n",
      "Epoch 140/175\n",
      "2193/2193 [==============================] - 2s 890us/step - loss: 0.3845 - binary_accuracy: 0.8405\n",
      "Epoch 141/175\n",
      "2193/2193 [==============================] - 2s 896us/step - loss: 0.3805 - binary_accuracy: 0.8383\n",
      "Epoch 142/175\n",
      "2193/2193 [==============================] - 2s 920us/step - loss: 0.3854 - binary_accuracy: 0.8376\n",
      "Epoch 143/175\n",
      "2193/2193 [==============================] - 2s 908us/step - loss: 0.3824 - binary_accuracy: 0.8404\n",
      "Epoch 144/175\n",
      "2193/2193 [==============================] - 2s 889us/step - loss: 0.3791 - binary_accuracy: 0.8429\n",
      "Epoch 145/175\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2193/2193 [==============================] - 2s 878us/step - loss: 0.3807 - binary_accuracy: 0.8391\n",
      "Epoch 146/175\n",
      "2193/2193 [==============================] - 2s 892us/step - loss: 0.3839 - binary_accuracy: 0.8368\n",
      "Epoch 147/175\n",
      "2193/2193 [==============================] - 2s 883us/step - loss: 0.3862 - binary_accuracy: 0.8368\n",
      "Epoch 148/175\n",
      "2193/2193 [==============================] - 2s 886us/step - loss: 0.3776 - binary_accuracy: 0.8422\n",
      "Epoch 149/175\n",
      "2193/2193 [==============================] - 2s 882us/step - loss: 0.3847 - binary_accuracy: 0.8395\n",
      "Epoch 150/175\n",
      "2193/2193 [==============================] - 2s 886us/step - loss: 0.3871 - binary_accuracy: 0.8383\n",
      "Epoch 151/175\n",
      "2193/2193 [==============================] - 2s 880us/step - loss: 0.3858 - binary_accuracy: 0.8389\n",
      "Epoch 152/175\n",
      "2193/2193 [==============================] - 2s 883us/step - loss: 0.3846 - binary_accuracy: 0.8422\n",
      "Epoch 153/175\n",
      "2193/2193 [==============================] - 2s 888us/step - loss: 0.3847 - binary_accuracy: 0.8399\n",
      "Epoch 154/175\n",
      "2193/2193 [==============================] - 2s 894us/step - loss: 0.3834 - binary_accuracy: 0.8362\n",
      "Epoch 155/175\n",
      "2193/2193 [==============================] - 2s 897us/step - loss: 0.3838 - binary_accuracy: 0.8394\n",
      "Epoch 156/175\n",
      "2193/2193 [==============================] - 2s 899us/step - loss: 0.3821 - binary_accuracy: 0.8413\n",
      "Epoch 157/175\n",
      "2193/2193 [==============================] - 2s 890us/step - loss: 0.3814 - binary_accuracy: 0.8425\n",
      "Epoch 158/175\n",
      "2193/2193 [==============================] - 2s 887us/step - loss: 0.3852 - binary_accuracy: 0.8416\n",
      "Epoch 159/175\n",
      "2193/2193 [==============================] - 2s 899us/step - loss: 0.3802 - binary_accuracy: 0.8390\n",
      "Epoch 160/175\n",
      "2193/2193 [==============================] - 2s 895us/step - loss: 0.3850 - binary_accuracy: 0.8365\n",
      "Epoch 161/175\n",
      "2193/2193 [==============================] - 2s 885us/step - loss: 0.3857 - binary_accuracy: 0.8342\n",
      "Epoch 162/175\n",
      "2193/2193 [==============================] - 2s 918us/step - loss: 0.3875 - binary_accuracy: 0.8406\n",
      "Epoch 163/175\n",
      "2193/2193 [==============================] - 2s 946us/step - loss: 0.3831 - binary_accuracy: 0.8390\n",
      "Epoch 164/175\n",
      "2193/2193 [==============================] - 2s 923us/step - loss: 0.3825 - binary_accuracy: 0.8422\n",
      "Epoch 165/175\n",
      "2193/2193 [==============================] - 2s 915us/step - loss: 0.3808 - binary_accuracy: 0.8402\n",
      "Epoch 166/175\n",
      "2193/2193 [==============================] - 2s 890us/step - loss: 0.3789 - binary_accuracy: 0.8404\n",
      "Epoch 167/175\n",
      "2193/2193 [==============================] - 2s 889us/step - loss: 0.3805 - binary_accuracy: 0.8417\n",
      "Epoch 168/175\n",
      "2193/2193 [==============================] - 2s 889us/step - loss: 0.3760 - binary_accuracy: 0.8416\n",
      "Epoch 169/175\n",
      "2193/2193 [==============================] - 2s 898us/step - loss: 0.3750 - binary_accuracy: 0.8443\n",
      "Epoch 170/175\n",
      "2193/2193 [==============================] - 2s 901us/step - loss: 0.3810 - binary_accuracy: 0.8421\n",
      "Epoch 171/175\n",
      "2193/2193 [==============================] - 2s 915us/step - loss: 0.3835 - binary_accuracy: 0.8391\n",
      "Epoch 172/175\n",
      "2193/2193 [==============================] - 2s 904us/step - loss: 0.3844 - binary_accuracy: 0.8382\n",
      "Epoch 173/175\n",
      "2193/2193 [==============================] - 2s 902us/step - loss: 0.3792 - binary_accuracy: 0.8393\n",
      "Epoch 174/175\n",
      "2193/2193 [==============================] - 2s 905us/step - loss: 0.3799 - binary_accuracy: 0.8414\n",
      "Epoch 175/175\n",
      "2193/2193 [==============================] - 2s 886us/step - loss: 0.3818 - binary_accuracy: 0.8398\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "auroc:  0.7165782410586106\n",
      "f1:  0.43478260869565216\n",
      "threshold:  0.1696575\n",
      "precision:  0.33052884615384615\n",
      "recall:  0.6351039260969977\n",
      "accuracy:  0.6868155935173018\n",
      "specificity:  0.6989189189189189\n",
      "CPU times: user 21min 3s, sys: 4min 9s, total: 25min 12s\n",
      "Wall time: 5min 48s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "### A single run w/out cross-validation\n",
    "# with previous outcome added to the features\n",
    "# outcome is added before scaling\n",
    "seed(1)\n",
    "set_random_seed(2)\n",
    "model = build_model(32, 0.5, max_timestamps, n_features)\n",
    "fit_model(model, 175, X_train_pad, y_train_pad, verbose=1)\n",
    "y_predict = model.predict(X_test_pad)\n",
    "y_test_len = [len(y) for y in y_test_list]\n",
    "y_predict = truncate_prediction(y_predict, y_test_len)\n",
    "y_true = np.concatenate(y_test_list)\n",
    "print('auroc: ', auroc(y_predict, y_true))\n",
    "f1, th = find_best_f1(y_predict, y_true)\n",
    "print('f1: ', f1);\n",
    "print('threshold: ', th)\n",
    "precision, recall, accuracy, specificity = other_metrics(y_predict, y_true, th)\n",
    "print('precision: ', precision)\n",
    "print('recall: ', recall)\n",
    "print('accuracy: ', accuracy)\n",
    "print('specificity: ', specificity)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/225\n",
      "1755/1755 [==============================] - 3s 2ms/step - loss: 0.4618 - binary_accuracy: 0.8173\n",
      "Epoch 2/225\n",
      "1755/1755 [==============================] - 2s 941us/step - loss: 0.4406 - binary_accuracy: 0.8201\n",
      "Epoch 3/225\n",
      "1755/1755 [==============================] - 2s 912us/step - loss: 0.4321 - binary_accuracy: 0.8199\n",
      "Epoch 4/225\n",
      "1755/1755 [==============================] - 2s 912us/step - loss: 0.4182 - binary_accuracy: 0.8253\n",
      "Epoch 5/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.4163 - binary_accuracy: 0.8300\n",
      "Epoch 6/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.4248 - binary_accuracy: 0.8264\n",
      "Epoch 7/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.4226 - binary_accuracy: 0.8217\n",
      "Epoch 8/225\n",
      "1755/1755 [==============================] - 2s 915us/step - loss: 0.4227 - binary_accuracy: 0.8194\n",
      "Epoch 9/225\n",
      "1755/1755 [==============================] - 2s 910us/step - loss: 0.4131 - binary_accuracy: 0.8270\n",
      "Epoch 10/225\n",
      "1755/1755 [==============================] - 2s 906us/step - loss: 0.4129 - binary_accuracy: 0.8232\n",
      "Epoch 11/225\n",
      "1755/1755 [==============================] - 2s 926us/step - loss: 0.4167 - binary_accuracy: 0.8270\n",
      "Epoch 12/225\n",
      "1755/1755 [==============================] - 2s 912us/step - loss: 0.4093 - binary_accuracy: 0.8306\n",
      "Epoch 13/225\n",
      "1755/1755 [==============================] - 2s 910us/step - loss: 0.4109 - binary_accuracy: 0.8282\n",
      "Epoch 14/225\n",
      "1755/1755 [==============================] - 2s 913us/step - loss: 0.4156 - binary_accuracy: 0.8263\n",
      "Epoch 15/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.4101 - binary_accuracy: 0.8286\n",
      "Epoch 16/225\n",
      "1755/1755 [==============================] - 2s 923us/step - loss: 0.4052 - binary_accuracy: 0.8287\n",
      "Epoch 17/225\n",
      "1755/1755 [==============================] - 2s 928us/step - loss: 0.4055 - binary_accuracy: 0.8291\n",
      "Epoch 18/225\n",
      "1755/1755 [==============================] - 2s 929us/step - loss: 0.4097 - binary_accuracy: 0.8317\n",
      "Epoch 19/225\n",
      "1755/1755 [==============================] - 2s 929us/step - loss: 0.4090 - binary_accuracy: 0.8266\n",
      "Epoch 20/225\n",
      "1755/1755 [==============================] - 2s 928us/step - loss: 0.4083 - binary_accuracy: 0.8306\n",
      "Epoch 21/225\n",
      "1755/1755 [==============================] - 2s 932us/step - loss: 0.4085 - binary_accuracy: 0.8292\n",
      "Epoch 22/225\n",
      "1755/1755 [==============================] - 2s 934us/step - loss: 0.4071 - binary_accuracy: 0.8290\n",
      "Epoch 23/225\n",
      "1755/1755 [==============================] - 2s 931us/step - loss: 0.4024 - binary_accuracy: 0.8354\n",
      "Epoch 24/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.4053 - binary_accuracy: 0.8320\n",
      "Epoch 25/225\n",
      "1755/1755 [==============================] - 2s 920us/step - loss: 0.4032 - binary_accuracy: 0.8325\n",
      "Epoch 26/225\n",
      "1755/1755 [==============================] - 2s 917us/step - loss: 0.4020 - binary_accuracy: 0.8370\n",
      "Epoch 27/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.3974 - binary_accuracy: 0.8376\n",
      "Epoch 28/225\n",
      "1755/1755 [==============================] - 2s 943us/step - loss: 0.4017 - binary_accuracy: 0.8340\n",
      "Epoch 29/225\n",
      "1755/1755 [==============================] - 2s 920us/step - loss: 0.3998 - binary_accuracy: 0.8340\n",
      "Epoch 30/225\n",
      "1755/1755 [==============================] - 2s 924us/step - loss: 0.4054 - binary_accuracy: 0.8298\n",
      "Epoch 31/225\n",
      "1755/1755 [==============================] - 2s 920us/step - loss: 0.4042 - binary_accuracy: 0.8333\n",
      "Epoch 32/225\n",
      "1755/1755 [==============================] - 2s 923us/step - loss: 0.4123 - binary_accuracy: 0.8295\n",
      "Epoch 33/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.4044 - binary_accuracy: 0.8291\n",
      "Epoch 34/225\n",
      "1755/1755 [==============================] - 2s 939us/step - loss: 0.4005 - binary_accuracy: 0.8317\n",
      "Epoch 35/225\n",
      "1755/1755 [==============================] - 2s 958us/step - loss: 0.4028 - binary_accuracy: 0.8321\n",
      "Epoch 36/225\n",
      "1755/1755 [==============================] - 2s 929us/step - loss: 0.4011 - binary_accuracy: 0.8284\n",
      "Epoch 37/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.4064 - binary_accuracy: 0.8291\n",
      "Epoch 38/225\n",
      "1755/1755 [==============================] - 2s 913us/step - loss: 0.4052 - binary_accuracy: 0.8297\n",
      "Epoch 39/225\n",
      "1755/1755 [==============================] - 2s 911us/step - loss: 0.4047 - binary_accuracy: 0.8300\n",
      "Epoch 40/225\n",
      "1755/1755 [==============================] - 2s 917us/step - loss: 0.4028 - binary_accuracy: 0.8298\n",
      "Epoch 41/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.4048 - binary_accuracy: 0.8341\n",
      "Epoch 42/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.4029 - binary_accuracy: 0.8274\n",
      "Epoch 43/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.3989 - binary_accuracy: 0.8345\n",
      "Epoch 44/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.4023 - binary_accuracy: 0.8289\n",
      "Epoch 45/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.3980 - binary_accuracy: 0.8316\n",
      "Epoch 46/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3982 - binary_accuracy: 0.8326\n",
      "Epoch 47/225\n",
      "1755/1755 [==============================] - 2s 932us/step - loss: 0.3977 - binary_accuracy: 0.8376\n",
      "Epoch 48/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.4015 - binary_accuracy: 0.8335\n",
      "Epoch 49/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.4026 - binary_accuracy: 0.8296\n",
      "Epoch 50/225\n",
      "1755/1755 [==============================] - 2s 921us/step - loss: 0.3975 - binary_accuracy: 0.8321\n",
      "Epoch 51/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3963 - binary_accuracy: 0.8317\n",
      "Epoch 52/225\n",
      "1755/1755 [==============================] - 2s 938us/step - loss: 0.3966 - binary_accuracy: 0.8347\n",
      "Epoch 53/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3993 - binary_accuracy: 0.8327\n",
      "Epoch 54/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3960 - binary_accuracy: 0.8351\n",
      "Epoch 55/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.4008 - binary_accuracy: 0.8292\n",
      "Epoch 56/225\n",
      "1755/1755 [==============================] - 2s 939us/step - loss: 0.3938 - binary_accuracy: 0.8377\n",
      "Epoch 57/225\n",
      "1755/1755 [==============================] - 2s 950us/step - loss: 0.4007 - binary_accuracy: 0.8317\n",
      "Epoch 58/225\n",
      "1755/1755 [==============================] - 2s 935us/step - loss: 0.3953 - binary_accuracy: 0.8347\n",
      "Epoch 59/225\n",
      "1755/1755 [==============================] - 2s 923us/step - loss: 0.4008 - binary_accuracy: 0.8351\n",
      "Epoch 60/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.3942 - binary_accuracy: 0.8400\n",
      "Epoch 61/225\n",
      "1755/1755 [==============================] - 2s 927us/step - loss: 0.4024 - binary_accuracy: 0.8315\n",
      "Epoch 62/225\n",
      "1755/1755 [==============================] - 2s 923us/step - loss: 0.3942 - binary_accuracy: 0.8378\n",
      "Epoch 63/225\n",
      "1755/1755 [==============================] - 2s 929us/step - loss: 0.3970 - binary_accuracy: 0.8334\n",
      "Epoch 64/225\n",
      "1755/1755 [==============================] - 2s 913us/step - loss: 0.3865 - binary_accuracy: 0.8415\n",
      "Epoch 65/225\n",
      "1755/1755 [==============================] - 2s 928us/step - loss: 0.3960 - binary_accuracy: 0.8330\n",
      "Epoch 66/225\n",
      "1755/1755 [==============================] - 2s 923us/step - loss: 0.3986 - binary_accuracy: 0.8306\n",
      "Epoch 67/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.3917 - binary_accuracy: 0.8337\n",
      "Epoch 68/225\n",
      "1755/1755 [==============================] - 2s 927us/step - loss: 0.3955 - binary_accuracy: 0.8342\n",
      "Epoch 69/225\n",
      "1755/1755 [==============================] - 2s 931us/step - loss: 0.3917 - binary_accuracy: 0.8341\n",
      "Epoch 70/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.3936 - binary_accuracy: 0.8386\n",
      "Epoch 71/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3923 - binary_accuracy: 0.8390\n",
      "Epoch 72/225\n",
      "1755/1755 [==============================] - 2s 928us/step - loss: 0.3950 - binary_accuracy: 0.8382\n",
      "Epoch 73/225\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1755/1755 [==============================] - 2s 915us/step - loss: 0.3918 - binary_accuracy: 0.8383\n",
      "Epoch 74/225\n",
      "1755/1755 [==============================] - 2s 917us/step - loss: 0.3945 - binary_accuracy: 0.8391\n",
      "Epoch 75/225\n",
      "1755/1755 [==============================] - 2s 915us/step - loss: 0.3932 - binary_accuracy: 0.8343\n",
      "Epoch 76/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3919 - binary_accuracy: 0.8383\n",
      "Epoch 77/225\n",
      "1755/1755 [==============================] - 2s 917us/step - loss: 0.3990 - binary_accuracy: 0.8314\n",
      "Epoch 78/225\n",
      "1755/1755 [==============================] - 2s 927us/step - loss: 0.3888 - binary_accuracy: 0.8362\n",
      "Epoch 79/225\n",
      "1755/1755 [==============================] - 2s 931us/step - loss: 0.3927 - binary_accuracy: 0.8350\n",
      "Epoch 80/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3870 - binary_accuracy: 0.8379\n",
      "Epoch 81/225\n",
      "1755/1755 [==============================] - 2s 912us/step - loss: 0.3916 - binary_accuracy: 0.8405\n",
      "Epoch 82/225\n",
      "1755/1755 [==============================] - 2s 923us/step - loss: 0.3956 - binary_accuracy: 0.8372\n",
      "Epoch 83/225\n",
      "1755/1755 [==============================] - 2s 927us/step - loss: 0.3873 - binary_accuracy: 0.8437\n",
      "Epoch 84/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3958 - binary_accuracy: 0.8356\n",
      "Epoch 85/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3911 - binary_accuracy: 0.8362\n",
      "Epoch 86/225\n",
      "1755/1755 [==============================] - 2s 914us/step - loss: 0.3897 - binary_accuracy: 0.8382\n",
      "Epoch 87/225\n",
      "1755/1755 [==============================] - 2s 911us/step - loss: 0.3914 - binary_accuracy: 0.8340\n",
      "Epoch 88/225\n",
      "1755/1755 [==============================] - 2s 921us/step - loss: 0.3931 - binary_accuracy: 0.8346\n",
      "Epoch 89/225\n",
      "1755/1755 [==============================] - 2s 910us/step - loss: 0.3820 - binary_accuracy: 0.8418\n",
      "Epoch 90/225\n",
      "1755/1755 [==============================] - 2s 910us/step - loss: 0.3896 - binary_accuracy: 0.8381\n",
      "Epoch 91/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3893 - binary_accuracy: 0.8369\n",
      "Epoch 92/225\n",
      "1755/1755 [==============================] - 2s 914us/step - loss: 0.3862 - binary_accuracy: 0.8401\n",
      "Epoch 93/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3878 - binary_accuracy: 0.8389\n",
      "Epoch 94/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.3883 - binary_accuracy: 0.8392\n",
      "Epoch 95/225\n",
      "1755/1755 [==============================] - 2s 905us/step - loss: 0.3880 - binary_accuracy: 0.8384\n",
      "Epoch 96/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.3878 - binary_accuracy: 0.8381\n",
      "Epoch 97/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.3857 - binary_accuracy: 0.8394\n",
      "Epoch 98/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3911 - binary_accuracy: 0.8353\n",
      "Epoch 99/225\n",
      "1755/1755 [==============================] - 2s 911us/step - loss: 0.3839 - binary_accuracy: 0.8411\n",
      "Epoch 100/225\n",
      "1755/1755 [==============================] - 2s 920us/step - loss: 0.3839 - binary_accuracy: 0.8433\n",
      "Epoch 101/225\n",
      "1755/1755 [==============================] - 2s 932us/step - loss: 0.3908 - binary_accuracy: 0.8398\n",
      "Epoch 102/225\n",
      "1755/1755 [==============================] - 2s 937us/step - loss: 0.3924 - binary_accuracy: 0.8348\n",
      "Epoch 103/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3881 - binary_accuracy: 0.8362\n",
      "Epoch 104/225\n",
      "1755/1755 [==============================] - 2s 917us/step - loss: 0.3947 - binary_accuracy: 0.8387\n",
      "Epoch 105/225\n",
      "1755/1755 [==============================] - 2s 920us/step - loss: 0.3811 - binary_accuracy: 0.8398\n",
      "Epoch 106/225\n",
      "1755/1755 [==============================] - 2s 917us/step - loss: 0.3914 - binary_accuracy: 0.8387\n",
      "Epoch 107/225\n",
      "1755/1755 [==============================] - 2s 923us/step - loss: 0.3898 - binary_accuracy: 0.8368\n",
      "Epoch 108/225\n",
      "1755/1755 [==============================] - 2s 923us/step - loss: 0.3896 - binary_accuracy: 0.8336\n",
      "Epoch 109/225\n",
      "1755/1755 [==============================] - 2s 905us/step - loss: 0.3824 - binary_accuracy: 0.8379\n",
      "Epoch 110/225\n",
      "1755/1755 [==============================] - 2s 913us/step - loss: 0.3841 - binary_accuracy: 0.8405\n",
      "Epoch 111/225\n",
      "1755/1755 [==============================] - 2s 910us/step - loss: 0.3898 - binary_accuracy: 0.8344\n",
      "Epoch 112/225\n",
      "1755/1755 [==============================] - 2s 929us/step - loss: 0.3843 - binary_accuracy: 0.8366\n",
      "Epoch 113/225\n",
      "1755/1755 [==============================] - 2s 912us/step - loss: 0.3823 - binary_accuracy: 0.8433\n",
      "Epoch 114/225\n",
      "1755/1755 [==============================] - 2s 917us/step - loss: 0.3906 - binary_accuracy: 0.8418\n",
      "Epoch 115/225\n",
      "1755/1755 [==============================] - 2s 920us/step - loss: 0.3892 - binary_accuracy: 0.8402\n",
      "Epoch 116/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3886 - binary_accuracy: 0.8385\n",
      "Epoch 117/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.3852 - binary_accuracy: 0.8389\n",
      "Epoch 118/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3861 - binary_accuracy: 0.8375\n",
      "Epoch 119/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.3843 - binary_accuracy: 0.8390\n",
      "Epoch 120/225\n",
      "1755/1755 [==============================] - 2s 917us/step - loss: 0.3848 - binary_accuracy: 0.8408\n",
      "Epoch 121/225\n",
      "1755/1755 [==============================] - 2s 924us/step - loss: 0.3861 - binary_accuracy: 0.8374\n",
      "Epoch 122/225\n",
      "1755/1755 [==============================] - 2s 914us/step - loss: 0.3864 - binary_accuracy: 0.8405\n",
      "Epoch 123/225\n",
      "1755/1755 [==============================] - 2s 917us/step - loss: 0.3768 - binary_accuracy: 0.8445\n",
      "Epoch 124/225\n",
      "1755/1755 [==============================] - 2s 910us/step - loss: 0.3844 - binary_accuracy: 0.8404\n",
      "Epoch 125/225\n",
      "1755/1755 [==============================] - 2s 915us/step - loss: 0.3828 - binary_accuracy: 0.8403\n",
      "Epoch 126/225\n",
      "1755/1755 [==============================] - 2s 913us/step - loss: 0.3858 - binary_accuracy: 0.8413\n",
      "Epoch 127/225\n",
      "1755/1755 [==============================] - 2s 915us/step - loss: 0.3905 - binary_accuracy: 0.8361\n",
      "Epoch 128/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3792 - binary_accuracy: 0.8483\n",
      "Epoch 129/225\n",
      "1755/1755 [==============================] - 2s 915us/step - loss: 0.3841 - binary_accuracy: 0.8415\n",
      "Epoch 130/225\n",
      "1755/1755 [==============================] - 2s 910us/step - loss: 0.3840 - binary_accuracy: 0.8427\n",
      "Epoch 131/225\n",
      "1755/1755 [==============================] - 2s 912us/step - loss: 0.3846 - binary_accuracy: 0.8388\n",
      "Epoch 132/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.3809 - binary_accuracy: 0.8454\n",
      "Epoch 133/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3786 - binary_accuracy: 0.8406\n",
      "Epoch 134/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3849 - binary_accuracy: 0.8371\n",
      "Epoch 135/225\n",
      "1755/1755 [==============================] - 2s 920us/step - loss: 0.3822 - binary_accuracy: 0.8389\n",
      "Epoch 136/225\n",
      "1755/1755 [==============================] - 2s 926us/step - loss: 0.3823 - binary_accuracy: 0.8415\n",
      "Epoch 137/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.3822 - binary_accuracy: 0.8391\n",
      "Epoch 138/225\n",
      "1755/1755 [==============================] - 2s 911us/step - loss: 0.3775 - binary_accuracy: 0.8433\n",
      "Epoch 139/225\n",
      "1755/1755 [==============================] - 2s 932us/step - loss: 0.3824 - binary_accuracy: 0.8386\n",
      "Epoch 140/225\n",
      "1755/1755 [==============================] - 2s 940us/step - loss: 0.3858 - binary_accuracy: 0.8371\n",
      "Epoch 141/225\n",
      "1755/1755 [==============================] - 2s 923us/step - loss: 0.3822 - binary_accuracy: 0.8386\n",
      "Epoch 142/225\n",
      "1755/1755 [==============================] - 2s 921us/step - loss: 0.3809 - binary_accuracy: 0.8359\n",
      "Epoch 143/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.3845 - binary_accuracy: 0.8412\n",
      "Epoch 144/225\n",
      "1755/1755 [==============================] - 2s 921us/step - loss: 0.3773 - binary_accuracy: 0.8425\n",
      "Epoch 145/225\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1755/1755 [==============================] - 2s 938us/step - loss: 0.3850 - binary_accuracy: 0.8369\n",
      "Epoch 146/225\n",
      "1755/1755 [==============================] - 2s 940us/step - loss: 0.3751 - binary_accuracy: 0.8453\n",
      "Epoch 147/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.3812 - binary_accuracy: 0.8424\n",
      "Epoch 148/225\n",
      "1755/1755 [==============================] - 2s 915us/step - loss: 0.3805 - binary_accuracy: 0.8421\n",
      "Epoch 149/225\n",
      "1755/1755 [==============================] - 2s 928us/step - loss: 0.3838 - binary_accuracy: 0.8419\n",
      "Epoch 150/225\n",
      "1755/1755 [==============================] - 2s 914us/step - loss: 0.3800 - binary_accuracy: 0.8395\n",
      "Epoch 151/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3842 - binary_accuracy: 0.8402\n",
      "Epoch 152/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.3833 - binary_accuracy: 0.8437\n",
      "Epoch 153/225\n",
      "1755/1755 [==============================] - 2s 913us/step - loss: 0.3801 - binary_accuracy: 0.8403\n",
      "Epoch 154/225\n",
      "1755/1755 [==============================] - 2s 917us/step - loss: 0.3843 - binary_accuracy: 0.8387\n",
      "Epoch 155/225\n",
      "1755/1755 [==============================] - 2s 914us/step - loss: 0.3836 - binary_accuracy: 0.8412\n",
      "Epoch 156/225\n",
      "1755/1755 [==============================] - 2s 907us/step - loss: 0.3852 - binary_accuracy: 0.8411\n",
      "Epoch 157/225\n",
      "1755/1755 [==============================] - 2s 903us/step - loss: 0.3843 - binary_accuracy: 0.8426\n",
      "Epoch 158/225\n",
      "1755/1755 [==============================] - 2s 894us/step - loss: 0.3790 - binary_accuracy: 0.8406\n",
      "Epoch 159/225\n",
      "1755/1755 [==============================] - 2s 912us/step - loss: 0.3812 - binary_accuracy: 0.8400\n",
      "Epoch 160/225\n",
      "1755/1755 [==============================] - 2s 914us/step - loss: 0.3798 - binary_accuracy: 0.8456\n",
      "Epoch 161/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3818 - binary_accuracy: 0.8417\n",
      "Epoch 162/225\n",
      "1755/1755 [==============================] - 2s 910us/step - loss: 0.3777 - binary_accuracy: 0.8448\n",
      "Epoch 163/225\n",
      "1755/1755 [==============================] - 2s 903us/step - loss: 0.3775 - binary_accuracy: 0.8410\n",
      "Epoch 164/225\n",
      "1755/1755 [==============================] - 2s 910us/step - loss: 0.3790 - binary_accuracy: 0.8426\n",
      "Epoch 165/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3823 - binary_accuracy: 0.8429\n",
      "Epoch 166/225\n",
      "1755/1755 [==============================] - 2s 928us/step - loss: 0.3805 - binary_accuracy: 0.8446\n",
      "Epoch 167/225\n",
      "1755/1755 [==============================] - 2s 915us/step - loss: 0.3740 - binary_accuracy: 0.8424\n",
      "Epoch 168/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3768 - binary_accuracy: 0.8420\n",
      "Epoch 169/225\n",
      "1755/1755 [==============================] - 2s 909us/step - loss: 0.3765 - binary_accuracy: 0.8423\n",
      "Epoch 170/225\n",
      "1755/1755 [==============================] - 2s 912us/step - loss: 0.3792 - binary_accuracy: 0.8431\n",
      "Epoch 171/225\n",
      "1755/1755 [==============================] - 2s 914us/step - loss: 0.3781 - binary_accuracy: 0.8420\n",
      "Epoch 172/225\n",
      "1755/1755 [==============================] - 2s 921us/step - loss: 0.3763 - binary_accuracy: 0.8439\n",
      "Epoch 173/225\n",
      "1755/1755 [==============================] - 2s 909us/step - loss: 0.3782 - binary_accuracy: 0.8436\n",
      "Epoch 174/225\n",
      "1755/1755 [==============================] - 2s 910us/step - loss: 0.3780 - binary_accuracy: 0.8444\n",
      "Epoch 175/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3747 - binary_accuracy: 0.8431\n",
      "Epoch 176/225\n",
      "1755/1755 [==============================] - 2s 900us/step - loss: 0.3853 - binary_accuracy: 0.8416\n",
      "Epoch 177/225\n",
      "1755/1755 [==============================] - 2s 913us/step - loss: 0.3767 - binary_accuracy: 0.8431\n",
      "Epoch 178/225\n",
      "1755/1755 [==============================] - 2s 908us/step - loss: 0.3789 - binary_accuracy: 0.8390\n",
      "Epoch 179/225\n",
      "1755/1755 [==============================] - 2s 917us/step - loss: 0.3707 - binary_accuracy: 0.8482\n",
      "Epoch 180/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.3739 - binary_accuracy: 0.8462\n",
      "Epoch 181/225\n",
      "1755/1755 [==============================] - 2s 902us/step - loss: 0.3771 - binary_accuracy: 0.8458\n",
      "Epoch 182/225\n",
      "1755/1755 [==============================] - 2s 912us/step - loss: 0.3794 - binary_accuracy: 0.8413\n",
      "Epoch 183/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.3759 - binary_accuracy: 0.8416\n",
      "Epoch 184/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3734 - binary_accuracy: 0.8479\n",
      "Epoch 185/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3755 - binary_accuracy: 0.8477\n",
      "Epoch 186/225\n",
      "1755/1755 [==============================] - 2s 907us/step - loss: 0.3762 - binary_accuracy: 0.8374\n",
      "Epoch 187/225\n",
      "1755/1755 [==============================] - 2s 911us/step - loss: 0.3760 - binary_accuracy: 0.8451\n",
      "Epoch 188/225\n",
      "1755/1755 [==============================] - 2s 903us/step - loss: 0.3772 - binary_accuracy: 0.8410\n",
      "Epoch 189/225\n",
      "1755/1755 [==============================] - 2s 928us/step - loss: 0.3748 - binary_accuracy: 0.8436\n",
      "Epoch 190/225\n",
      "1755/1755 [==============================] - 2s 926us/step - loss: 0.3793 - binary_accuracy: 0.8440\n",
      "Epoch 191/225\n",
      "1755/1755 [==============================] - 2s 912us/step - loss: 0.3736 - binary_accuracy: 0.8452\n",
      "Epoch 192/225\n",
      "1755/1755 [==============================] - 2s 912us/step - loss: 0.3683 - binary_accuracy: 0.8472\n",
      "Epoch 193/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3690 - binary_accuracy: 0.8484\n",
      "Epoch 194/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3795 - binary_accuracy: 0.8425\n",
      "Epoch 195/225\n",
      "1755/1755 [==============================] - 2s 909us/step - loss: 0.3767 - binary_accuracy: 0.8490\n",
      "Epoch 196/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.3717 - binary_accuracy: 0.8420\n",
      "Epoch 197/225\n",
      "1755/1755 [==============================] - 2s 914us/step - loss: 0.3726 - binary_accuracy: 0.8477\n",
      "Epoch 198/225\n",
      "1755/1755 [==============================] - 2s 910us/step - loss: 0.3708 - binary_accuracy: 0.8462\n",
      "Epoch 199/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.3702 - binary_accuracy: 0.8456\n",
      "Epoch 200/225\n",
      "1755/1755 [==============================] - 2s 913us/step - loss: 0.3778 - binary_accuracy: 0.8414\n",
      "Epoch 201/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3745 - binary_accuracy: 0.8435\n",
      "Epoch 202/225\n",
      "1755/1755 [==============================] - 2s 912us/step - loss: 0.3803 - binary_accuracy: 0.8427\n",
      "Epoch 203/225\n",
      "1755/1755 [==============================] - 2s 905us/step - loss: 0.3759 - binary_accuracy: 0.8421\n",
      "Epoch 204/225\n",
      "1755/1755 [==============================] - 2s 915us/step - loss: 0.3724 - binary_accuracy: 0.8453\n",
      "Epoch 205/225\n",
      "1755/1755 [==============================] - 2s 913us/step - loss: 0.3767 - binary_accuracy: 0.8433\n",
      "Epoch 206/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3697 - binary_accuracy: 0.8467\n",
      "Epoch 207/225\n",
      "1755/1755 [==============================] - 2s 938us/step - loss: 0.3712 - binary_accuracy: 0.8475\n",
      "Epoch 208/225\n",
      "1755/1755 [==============================] - 2s 926us/step - loss: 0.3752 - binary_accuracy: 0.8413\n",
      "Epoch 209/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3763 - binary_accuracy: 0.8449\n",
      "Epoch 210/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.3666 - binary_accuracy: 0.8452\n",
      "Epoch 211/225\n",
      "1755/1755 [==============================] - 2s 912us/step - loss: 0.3739 - binary_accuracy: 0.8453\n",
      "Epoch 212/225\n",
      "1755/1755 [==============================] - 2s 930us/step - loss: 0.3757 - binary_accuracy: 0.8451\n",
      "Epoch 213/225\n",
      "1755/1755 [==============================] - 2s 945us/step - loss: 0.3791 - binary_accuracy: 0.8479\n",
      "Epoch 214/225\n",
      "1755/1755 [==============================] - 2s 911us/step - loss: 0.3716 - binary_accuracy: 0.8457\n",
      "Epoch 215/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3769 - binary_accuracy: 0.8421\n",
      "Epoch 216/225\n",
      "1755/1755 [==============================] - 2s 907us/step - loss: 0.3736 - binary_accuracy: 0.8505\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 217/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3724 - binary_accuracy: 0.8427\n",
      "Epoch 218/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.3691 - binary_accuracy: 0.8484\n",
      "Epoch 219/225\n",
      "1755/1755 [==============================] - 2s 909us/step - loss: 0.3736 - binary_accuracy: 0.8465\n",
      "Epoch 220/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3749 - binary_accuracy: 0.8440\n",
      "Epoch 221/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3645 - binary_accuracy: 0.8513\n",
      "Epoch 222/225\n",
      "1755/1755 [==============================] - 2s 915us/step - loss: 0.3707 - binary_accuracy: 0.8458\n",
      "Epoch 223/225\n",
      "1755/1755 [==============================] - 2s 915us/step - loss: 0.3739 - binary_accuracy: 0.8434\n",
      "Epoch 224/225\n",
      "1755/1755 [==============================] - 2s 911us/step - loss: 0.3663 - binary_accuracy: 0.8508\n",
      "Epoch 225/225\n",
      "1755/1755 [==============================] - 2s 910us/step - loss: 0.3780 - binary_accuracy: 0.8434\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/225\n",
      "1755/1755 [==============================] - 3s 2ms/step - loss: 0.4659 - binary_accuracy: 0.8235\n",
      "Epoch 2/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.4354 - binary_accuracy: 0.8249\n",
      "Epoch 3/225\n",
      "1755/1755 [==============================] - 2s 924us/step - loss: 0.4350 - binary_accuracy: 0.8226\n",
      "Epoch 4/225\n",
      "1755/1755 [==============================] - 2s 923us/step - loss: 0.4297 - binary_accuracy: 0.8284\n",
      "Epoch 5/225\n",
      "1755/1755 [==============================] - 2s 911us/step - loss: 0.4194 - binary_accuracy: 0.8262\n",
      "Epoch 6/225\n",
      "1755/1755 [==============================] - 2s 948us/step - loss: 0.4192 - binary_accuracy: 0.8265\n",
      "Epoch 7/225\n",
      "1755/1755 [==============================] - 2s 933us/step - loss: 0.4130 - binary_accuracy: 0.8283\n",
      "Epoch 8/225\n",
      "1755/1755 [==============================] - 2s 920us/step - loss: 0.4206 - binary_accuracy: 0.8271\n",
      "Epoch 9/225\n",
      "1755/1755 [==============================] - 2s 928us/step - loss: 0.4112 - binary_accuracy: 0.8286\n",
      "Epoch 10/225\n",
      "1755/1755 [==============================] - 2s 921us/step - loss: 0.4122 - binary_accuracy: 0.8294\n",
      "Epoch 11/225\n",
      "1755/1755 [==============================] - 2s 921us/step - loss: 0.4143 - binary_accuracy: 0.8261\n",
      "Epoch 12/225\n",
      "1755/1755 [==============================] - 2s 935us/step - loss: 0.4038 - binary_accuracy: 0.8329\n",
      "Epoch 13/225\n",
      "1755/1755 [==============================] - 2s 927us/step - loss: 0.4144 - binary_accuracy: 0.8259\n",
      "Epoch 14/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.4146 - binary_accuracy: 0.8276\n",
      "Epoch 15/225\n",
      "1755/1755 [==============================] - 2s 929us/step - loss: 0.4100 - binary_accuracy: 0.8297\n",
      "Epoch 16/225\n",
      "1755/1755 [==============================] - 2s 920us/step - loss: 0.4171 - binary_accuracy: 0.8219\n",
      "Epoch 17/225\n",
      "1755/1755 [==============================] - 2s 935us/step - loss: 0.4109 - binary_accuracy: 0.8308\n",
      "Epoch 18/225\n",
      "1755/1755 [==============================] - 2s 929us/step - loss: 0.4073 - binary_accuracy: 0.8305\n",
      "Epoch 19/225\n",
      "1755/1755 [==============================] - 2s 920us/step - loss: 0.4051 - binary_accuracy: 0.8327\n",
      "Epoch 20/225\n",
      "1755/1755 [==============================] - 2s 924us/step - loss: 0.4031 - binary_accuracy: 0.8343\n",
      "Epoch 21/225\n",
      "1755/1755 [==============================] - 2s 935us/step - loss: 0.4155 - binary_accuracy: 0.8272\n",
      "Epoch 22/225\n",
      "1755/1755 [==============================] - 2s 926us/step - loss: 0.4070 - binary_accuracy: 0.8285\n",
      "Epoch 23/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.4072 - binary_accuracy: 0.8305\n",
      "Epoch 24/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.4015 - binary_accuracy: 0.8347\n",
      "Epoch 25/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.4054 - binary_accuracy: 0.8274\n",
      "Epoch 26/225\n",
      "1755/1755 [==============================] - 2s 946us/step - loss: 0.4061 - binary_accuracy: 0.8357\n",
      "Epoch 27/225\n",
      "1755/1755 [==============================] - 2s 936us/step - loss: 0.4043 - binary_accuracy: 0.8303\n",
      "Epoch 28/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.4012 - binary_accuracy: 0.8341\n",
      "Epoch 29/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.4042 - binary_accuracy: 0.8335\n",
      "Epoch 30/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.4042 - binary_accuracy: 0.8328\n",
      "Epoch 31/225\n",
      "1755/1755 [==============================] - 2s 931us/step - loss: 0.4019 - binary_accuracy: 0.8372\n",
      "Epoch 32/225\n",
      "1755/1755 [==============================] - 2s 920us/step - loss: 0.4018 - binary_accuracy: 0.8312\n",
      "Epoch 33/225\n",
      "1755/1755 [==============================] - 2s 928us/step - loss: 0.4027 - binary_accuracy: 0.8322\n",
      "Epoch 34/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.4003 - binary_accuracy: 0.8342\n",
      "Epoch 35/225\n",
      "1755/1755 [==============================] - 2s 914us/step - loss: 0.3974 - binary_accuracy: 0.8356\n",
      "Epoch 36/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.3995 - binary_accuracy: 0.8329\n",
      "Epoch 37/225\n",
      "1755/1755 [==============================] - 2s 928us/step - loss: 0.4032 - binary_accuracy: 0.8356\n",
      "Epoch 38/225\n",
      "1755/1755 [==============================] - 2s 932us/step - loss: 0.3975 - binary_accuracy: 0.8319\n",
      "Epoch 39/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.4014 - binary_accuracy: 0.8335\n",
      "Epoch 40/225\n",
      "1755/1755 [==============================] - 2s 920us/step - loss: 0.3966 - binary_accuracy: 0.8369\n",
      "Epoch 41/225\n",
      "1755/1755 [==============================] - 2s 945us/step - loss: 0.4004 - binary_accuracy: 0.8348\n",
      "Epoch 42/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.4000 - binary_accuracy: 0.8347\n",
      "Epoch 43/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.3977 - binary_accuracy: 0.8355\n",
      "Epoch 44/225\n",
      "1755/1755 [==============================] - 2s 930us/step - loss: 0.4011 - binary_accuracy: 0.8345\n",
      "Epoch 45/225\n",
      "1755/1755 [==============================] - 2s 948us/step - loss: 0.3954 - binary_accuracy: 0.8385\n",
      "Epoch 46/225\n",
      "1755/1755 [==============================] - 2s 952us/step - loss: 0.3934 - binary_accuracy: 0.8344\n",
      "Epoch 47/225\n",
      "1755/1755 [==============================] - 2s 932us/step - loss: 0.4009 - binary_accuracy: 0.8336\n",
      "Epoch 48/225\n",
      "1755/1755 [==============================] - 2s 923us/step - loss: 0.3972 - binary_accuracy: 0.8344\n",
      "Epoch 49/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.4025 - binary_accuracy: 0.8312\n",
      "Epoch 50/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.4027 - binary_accuracy: 0.8348\n",
      "Epoch 51/225\n",
      "1755/1755 [==============================] - 2s 921us/step - loss: 0.3961 - binary_accuracy: 0.8369\n",
      "Epoch 52/225\n",
      "1755/1755 [==============================] - 2s 920us/step - loss: 0.3952 - binary_accuracy: 0.8338\n",
      "Epoch 53/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.4023 - binary_accuracy: 0.8292\n",
      "Epoch 54/225\n",
      "1755/1755 [==============================] - 2s 921us/step - loss: 0.3968 - binary_accuracy: 0.8352\n",
      "Epoch 55/225\n",
      "1755/1755 [==============================] - 2s 932us/step - loss: 0.3943 - binary_accuracy: 0.8364\n",
      "Epoch 56/225\n",
      "1755/1755 [==============================] - 2s 923us/step - loss: 0.3928 - binary_accuracy: 0.8367\n",
      "Epoch 57/225\n",
      "1755/1755 [==============================] - 2s 928us/step - loss: 0.3916 - binary_accuracy: 0.8387\n",
      "Epoch 58/225\n",
      "1755/1755 [==============================] - 2s 932us/step - loss: 0.3935 - binary_accuracy: 0.8394\n",
      "Epoch 59/225\n",
      "1755/1755 [==============================] - 2s 924us/step - loss: 0.3983 - binary_accuracy: 0.8364\n",
      "Epoch 60/225\n",
      "1755/1755 [==============================] - 2s 923us/step - loss: 0.3931 - binary_accuracy: 0.8350\n",
      "Epoch 61/225\n",
      "1755/1755 [==============================] - 2s 931us/step - loss: 0.3968 - binary_accuracy: 0.8421\n",
      "Epoch 62/225\n",
      "1755/1755 [==============================] - 2s 932us/step - loss: 0.3971 - binary_accuracy: 0.8372\n",
      "Epoch 63/225\n",
      "1755/1755 [==============================] - 2s 924us/step - loss: 0.3876 - binary_accuracy: 0.8366\n",
      "Epoch 64/225\n",
      "1755/1755 [==============================] - 2s 926us/step - loss: 0.3962 - binary_accuracy: 0.8365\n",
      "Epoch 65/225\n",
      "1755/1755 [==============================] - 2s 931us/step - loss: 0.3970 - binary_accuracy: 0.8346\n",
      "Epoch 66/225\n",
      "1755/1755 [==============================] - 2s 907us/step - loss: 0.3872 - binary_accuracy: 0.8384\n",
      "Epoch 67/225\n",
      "1755/1755 [==============================] - 2s 913us/step - loss: 0.3942 - binary_accuracy: 0.8350\n",
      "Epoch 68/225\n",
      "1755/1755 [==============================] - 2s 942us/step - loss: 0.3931 - binary_accuracy: 0.8352\n",
      "Epoch 69/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3873 - binary_accuracy: 0.8394\n",
      "Epoch 70/225\n",
      "1755/1755 [==============================] - 2s 941us/step - loss: 0.3899 - binary_accuracy: 0.8419\n",
      "Epoch 71/225\n",
      "1755/1755 [==============================] - 2s 927us/step - loss: 0.3858 - binary_accuracy: 0.8405\n",
      "Epoch 72/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.3921 - binary_accuracy: 0.8331\n",
      "Epoch 73/225\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.3878 - binary_accuracy: 0.8363\n",
      "Epoch 74/225\n",
      "1755/1755 [==============================] - 2s 920us/step - loss: 0.3903 - binary_accuracy: 0.8364\n",
      "Epoch 75/225\n",
      "1755/1755 [==============================] - 2s 917us/step - loss: 0.3925 - binary_accuracy: 0.8381\n",
      "Epoch 76/225\n",
      "1755/1755 [==============================] - 2s 926us/step - loss: 0.3919 - binary_accuracy: 0.8387\n",
      "Epoch 77/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.3851 - binary_accuracy: 0.8367\n",
      "Epoch 78/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3888 - binary_accuracy: 0.8368\n",
      "Epoch 79/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.3916 - binary_accuracy: 0.8390\n",
      "Epoch 80/225\n",
      "1755/1755 [==============================] - 2s 924us/step - loss: 0.3843 - binary_accuracy: 0.8426\n",
      "Epoch 81/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.3849 - binary_accuracy: 0.8430\n",
      "Epoch 82/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.3923 - binary_accuracy: 0.8398\n",
      "Epoch 83/225\n",
      "1755/1755 [==============================] - 2s 912us/step - loss: 0.3879 - binary_accuracy: 0.8353\n",
      "Epoch 84/225\n",
      "1755/1755 [==============================] - 2s 921us/step - loss: 0.3920 - binary_accuracy: 0.8338\n",
      "Epoch 85/225\n",
      "1755/1755 [==============================] - 2s 920us/step - loss: 0.3900 - binary_accuracy: 0.8384\n",
      "Epoch 86/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.3904 - binary_accuracy: 0.8392\n",
      "Epoch 87/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.3847 - binary_accuracy: 0.8371\n",
      "Epoch 88/225\n",
      "1755/1755 [==============================] - 2s 946us/step - loss: 0.3902 - binary_accuracy: 0.8392\n",
      "Epoch 89/225\n",
      "1755/1755 [==============================] - 2s 938us/step - loss: 0.3843 - binary_accuracy: 0.8384\n",
      "Epoch 90/225\n",
      "1755/1755 [==============================] - 2s 914us/step - loss: 0.3869 - binary_accuracy: 0.8410\n",
      "Epoch 91/225\n",
      "1755/1755 [==============================] - 2s 915us/step - loss: 0.3796 - binary_accuracy: 0.8398\n",
      "Epoch 92/225\n",
      "1755/1755 [==============================] - 2s 913us/step - loss: 0.3907 - binary_accuracy: 0.8357\n",
      "Epoch 93/225\n",
      "1755/1755 [==============================] - 2s 921us/step - loss: 0.3894 - binary_accuracy: 0.8417\n",
      "Epoch 94/225\n",
      "1755/1755 [==============================] - 2s 913us/step - loss: 0.3883 - binary_accuracy: 0.8356\n",
      "Epoch 95/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.3863 - binary_accuracy: 0.8404\n",
      "Epoch 96/225\n",
      "1755/1755 [==============================] - 2s 926us/step - loss: 0.3839 - binary_accuracy: 0.8407\n",
      "Epoch 97/225\n",
      "1755/1755 [==============================] - 2s 921us/step - loss: 0.3865 - binary_accuracy: 0.8365\n",
      "Epoch 98/225\n",
      "1755/1755 [==============================] - 2s 930us/step - loss: 0.3894 - binary_accuracy: 0.8361\n",
      "Epoch 99/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.3866 - binary_accuracy: 0.8423\n",
      "Epoch 100/225\n",
      "1755/1755 [==============================] - 2s 934us/step - loss: 0.3911 - binary_accuracy: 0.8386\n",
      "Epoch 101/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.3866 - binary_accuracy: 0.8359\n",
      "Epoch 102/225\n",
      "1755/1755 [==============================] - 2s 927us/step - loss: 0.3832 - binary_accuracy: 0.8402\n",
      "Epoch 103/225\n",
      "1755/1755 [==============================] - 2s 931us/step - loss: 0.3745 - binary_accuracy: 0.8429\n",
      "Epoch 104/225\n",
      "1755/1755 [==============================] - 2s 928us/step - loss: 0.3804 - binary_accuracy: 0.8412\n",
      "Epoch 105/225\n",
      "1755/1755 [==============================] - 2s 914us/step - loss: 0.3844 - binary_accuracy: 0.8429\n",
      "Epoch 106/225\n",
      "1755/1755 [==============================] - 2s 927us/step - loss: 0.3836 - binary_accuracy: 0.8404\n",
      "Epoch 107/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.3854 - binary_accuracy: 0.8385\n",
      "Epoch 108/225\n",
      "1755/1755 [==============================] - 2s 929us/step - loss: 0.3850 - binary_accuracy: 0.8377\n",
      "Epoch 109/225\n",
      "1755/1755 [==============================] - 2s 929us/step - loss: 0.3853 - binary_accuracy: 0.8353\n",
      "Epoch 110/225\n",
      "1755/1755 [==============================] - 2s 923us/step - loss: 0.3860 - binary_accuracy: 0.8406\n",
      "Epoch 111/225\n",
      "1755/1755 [==============================] - 2s 914us/step - loss: 0.3897 - binary_accuracy: 0.8390\n",
      "Epoch 112/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.3859 - binary_accuracy: 0.8417\n",
      "Epoch 113/225\n",
      "1755/1755 [==============================] - 2s 963us/step - loss: 0.3832 - binary_accuracy: 0.8438\n",
      "Epoch 114/225\n",
      "1755/1755 [==============================] - 2s 931us/step - loss: 0.3790 - binary_accuracy: 0.8431\n",
      "Epoch 115/225\n",
      "1755/1755 [==============================] - 2s 928us/step - loss: 0.3856 - binary_accuracy: 0.8353\n",
      "Epoch 116/225\n",
      "1755/1755 [==============================] - 2s 915us/step - loss: 0.3820 - binary_accuracy: 0.8400\n",
      "Epoch 117/225\n",
      "1755/1755 [==============================] - 2s 930us/step - loss: 0.3795 - binary_accuracy: 0.8405\n",
      "Epoch 118/225\n",
      "1755/1755 [==============================] - 2s 920us/step - loss: 0.3829 - binary_accuracy: 0.8396\n",
      "Epoch 119/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.3792 - binary_accuracy: 0.8403\n",
      "Epoch 120/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.3898 - binary_accuracy: 0.8354\n",
      "Epoch 121/225\n",
      "1755/1755 [==============================] - 2s 924us/step - loss: 0.3909 - binary_accuracy: 0.8336\n",
      "Epoch 122/225\n",
      "1755/1755 [==============================] - 2s 921us/step - loss: 0.3852 - binary_accuracy: 0.8378\n",
      "Epoch 123/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3841 - binary_accuracy: 0.8403\n",
      "Epoch 124/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.3810 - binary_accuracy: 0.8402\n",
      "Epoch 125/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.3831 - binary_accuracy: 0.8395\n",
      "Epoch 126/225\n",
      "1755/1755 [==============================] - 2s 914us/step - loss: 0.3778 - binary_accuracy: 0.8420\n",
      "Epoch 127/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.3902 - binary_accuracy: 0.8398\n",
      "Epoch 128/225\n",
      "1755/1755 [==============================] - 2s 912us/step - loss: 0.3799 - binary_accuracy: 0.8442\n",
      "Epoch 129/225\n",
      "1755/1755 [==============================] - 2s 926us/step - loss: 0.3761 - binary_accuracy: 0.8454\n",
      "Epoch 130/225\n",
      "1755/1755 [==============================] - 2s 939us/step - loss: 0.3794 - binary_accuracy: 0.8434\n",
      "Epoch 131/225\n",
      "1755/1755 [==============================] - 2s 938us/step - loss: 0.3730 - binary_accuracy: 0.8454\n",
      "Epoch 132/225\n",
      "1755/1755 [==============================] - 2s 936us/step - loss: 0.3807 - binary_accuracy: 0.8406\n",
      "Epoch 133/225\n",
      "1755/1755 [==============================] - 2s 940us/step - loss: 0.3809 - binary_accuracy: 0.8386\n",
      "Epoch 134/225\n",
      "1755/1755 [==============================] - 2s 933us/step - loss: 0.3743 - binary_accuracy: 0.8423\n",
      "Epoch 135/225\n",
      "1755/1755 [==============================] - 2s 914us/step - loss: 0.3814 - binary_accuracy: 0.8411\n",
      "Epoch 136/225\n",
      "1755/1755 [==============================] - 2s 911us/step - loss: 0.3784 - binary_accuracy: 0.8427\n",
      "Epoch 137/225\n",
      "1755/1755 [==============================] - 2s 911us/step - loss: 0.3781 - binary_accuracy: 0.8459\n",
      "Epoch 138/225\n",
      "1755/1755 [==============================] - 2s 915us/step - loss: 0.3747 - binary_accuracy: 0.8465\n",
      "Epoch 139/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3828 - binary_accuracy: 0.8386\n",
      "Epoch 140/225\n",
      "1755/1755 [==============================] - 2s 943us/step - loss: 0.3754 - binary_accuracy: 0.8403\n",
      "Epoch 141/225\n",
      "1755/1755 [==============================] - 2s 939us/step - loss: 0.3787 - binary_accuracy: 0.8425\n",
      "Epoch 142/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3839 - binary_accuracy: 0.8384\n",
      "Epoch 143/225\n",
      "1755/1755 [==============================] - 2s 931us/step - loss: 0.3830 - binary_accuracy: 0.8383\n",
      "Epoch 144/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.3818 - binary_accuracy: 0.8396\n",
      "Epoch 145/225\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1755/1755 [==============================] - 2s 927us/step - loss: 0.3794 - binary_accuracy: 0.8435\n",
      "Epoch 146/225\n",
      "1755/1755 [==============================] - 2s 912us/step - loss: 0.3785 - binary_accuracy: 0.8439\n",
      "Epoch 147/225\n",
      "1755/1755 [==============================] - 2s 927us/step - loss: 0.3857 - binary_accuracy: 0.8391\n",
      "Epoch 148/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3776 - binary_accuracy: 0.8400\n",
      "Epoch 149/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.3803 - binary_accuracy: 0.8406\n",
      "Epoch 150/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3728 - binary_accuracy: 0.8496\n",
      "Epoch 151/225\n",
      "1755/1755 [==============================] - 2s 915us/step - loss: 0.3780 - binary_accuracy: 0.8436\n",
      "Epoch 152/225\n",
      "1755/1755 [==============================] - 2s 915us/step - loss: 0.3725 - binary_accuracy: 0.8465\n",
      "Epoch 153/225\n",
      "1755/1755 [==============================] - 2s 914us/step - loss: 0.3791 - binary_accuracy: 0.8449\n",
      "Epoch 154/225\n",
      "1755/1755 [==============================] - 2s 923us/step - loss: 0.3769 - binary_accuracy: 0.8449\n",
      "Epoch 155/225\n",
      "1755/1755 [==============================] - 2s 934us/step - loss: 0.3729 - binary_accuracy: 0.8490\n",
      "Epoch 156/225\n",
      "1755/1755 [==============================] - 2s 923us/step - loss: 0.3815 - binary_accuracy: 0.8435\n",
      "Epoch 157/225\n",
      "1755/1755 [==============================] - 2s 929us/step - loss: 0.3819 - binary_accuracy: 0.8390\n",
      "Epoch 158/225\n",
      "1755/1755 [==============================] - 2s 960us/step - loss: 0.3741 - binary_accuracy: 0.8429\n",
      "Epoch 159/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3772 - binary_accuracy: 0.8449\n",
      "Epoch 160/225\n",
      "1755/1755 [==============================] - 2s 917us/step - loss: 0.3802 - binary_accuracy: 0.8412\n",
      "Epoch 161/225\n",
      "1755/1755 [==============================] - 2s 923us/step - loss: 0.3698 - binary_accuracy: 0.8484\n",
      "Epoch 162/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.3789 - binary_accuracy: 0.8425\n",
      "Epoch 163/225\n",
      "1755/1755 [==============================] - 2s 910us/step - loss: 0.3808 - binary_accuracy: 0.8471\n",
      "Epoch 164/225\n",
      "1755/1755 [==============================] - 2s 913us/step - loss: 0.3728 - binary_accuracy: 0.8421\n",
      "Epoch 165/225\n",
      "1755/1755 [==============================] - 2s 924us/step - loss: 0.3809 - binary_accuracy: 0.8437\n",
      "Epoch 166/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3820 - binary_accuracy: 0.8364\n",
      "Epoch 167/225\n",
      "1755/1755 [==============================] - 2s 907us/step - loss: 0.3737 - binary_accuracy: 0.8438\n",
      "Epoch 168/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.3757 - binary_accuracy: 0.8415\n",
      "Epoch 169/225\n",
      "1755/1755 [==============================] - 2s 915us/step - loss: 0.3751 - binary_accuracy: 0.8422\n",
      "Epoch 170/225\n",
      "1755/1755 [==============================] - 2s 912us/step - loss: 0.3845 - binary_accuracy: 0.8400\n",
      "Epoch 171/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.3751 - binary_accuracy: 0.8387\n",
      "Epoch 172/225\n",
      "1755/1755 [==============================] - 2s 920us/step - loss: 0.3835 - binary_accuracy: 0.8401\n",
      "Epoch 173/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.3758 - binary_accuracy: 0.8438\n",
      "Epoch 174/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3740 - binary_accuracy: 0.8453\n",
      "Epoch 175/225\n",
      "1755/1755 [==============================] - 2s 914us/step - loss: 0.3705 - binary_accuracy: 0.8474\n",
      "Epoch 176/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3761 - binary_accuracy: 0.8419\n",
      "Epoch 177/225\n",
      "1755/1755 [==============================] - 2s 916us/step - loss: 0.3777 - binary_accuracy: 0.8387\n",
      "Epoch 178/225\n",
      "1755/1755 [==============================] - 2s 909us/step - loss: 0.3759 - binary_accuracy: 0.8411\n",
      "Epoch 179/225\n",
      "1755/1755 [==============================] - 2s 911us/step - loss: 0.3738 - binary_accuracy: 0.8477\n",
      "Epoch 180/225\n",
      "1755/1755 [==============================] - 2s 912us/step - loss: 0.3792 - binary_accuracy: 0.8402\n",
      "Epoch 181/225\n",
      "1755/1755 [==============================] - 2s 940us/step - loss: 0.3688 - binary_accuracy: 0.8507\n",
      "Epoch 182/225\n",
      "1755/1755 [==============================] - 2s 942us/step - loss: 0.3703 - binary_accuracy: 0.8471\n",
      "Epoch 183/225\n",
      "1755/1755 [==============================] - 2s 910us/step - loss: 0.3738 - binary_accuracy: 0.8470\n",
      "Epoch 184/225\n",
      "1755/1755 [==============================] - 2s 908us/step - loss: 0.3772 - binary_accuracy: 0.8424\n",
      "Epoch 185/225\n",
      "1755/1755 [==============================] - 2s 913us/step - loss: 0.3731 - binary_accuracy: 0.8458\n",
      "Epoch 186/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.3703 - binary_accuracy: 0.8453\n",
      "Epoch 187/225\n",
      "1755/1755 [==============================] - 2s 905us/step - loss: 0.3761 - binary_accuracy: 0.8425\n",
      "Epoch 188/225\n",
      "1755/1755 [==============================] - 2s 913us/step - loss: 0.3770 - binary_accuracy: 0.8394\n",
      "Epoch 189/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.3710 - binary_accuracy: 0.8456\n",
      "Epoch 190/225\n",
      "1755/1755 [==============================] - 2s 921us/step - loss: 0.3769 - binary_accuracy: 0.8475\n",
      "Epoch 191/225\n",
      "1755/1755 [==============================] - 2s 907us/step - loss: 0.3753 - binary_accuracy: 0.8424\n",
      "Epoch 192/225\n",
      "1755/1755 [==============================] - 2s 914us/step - loss: 0.3759 - binary_accuracy: 0.8455\n",
      "Epoch 193/225\n",
      "1755/1755 [==============================] - 2s 905us/step - loss: 0.3733 - binary_accuracy: 0.8446\n",
      "Epoch 194/225\n",
      "1755/1755 [==============================] - 2s 921us/step - loss: 0.3681 - binary_accuracy: 0.8494\n",
      "Epoch 195/225\n",
      "1755/1755 [==============================] - 2s 911us/step - loss: 0.3723 - binary_accuracy: 0.8471\n",
      "Epoch 196/225\n",
      "1755/1755 [==============================] - 2s 913us/step - loss: 0.3734 - binary_accuracy: 0.8441\n",
      "Epoch 197/225\n",
      "1755/1755 [==============================] - 2s 911us/step - loss: 0.3722 - binary_accuracy: 0.8465\n",
      "Epoch 198/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3735 - binary_accuracy: 0.8473\n",
      "Epoch 199/225\n",
      "1755/1755 [==============================] - 2s 908us/step - loss: 0.3684 - binary_accuracy: 0.8489\n",
      "Epoch 200/225\n",
      "1755/1755 [==============================] - 2s 911us/step - loss: 0.3739 - binary_accuracy: 0.8461\n",
      "Epoch 201/225\n",
      "1755/1755 [==============================] - 2s 926us/step - loss: 0.3690 - binary_accuracy: 0.8450\n",
      "Epoch 202/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3738 - binary_accuracy: 0.8472\n",
      "Epoch 203/225\n",
      "1755/1755 [==============================] - 2s 911us/step - loss: 0.3742 - binary_accuracy: 0.8419\n",
      "Epoch 204/225\n",
      "1755/1755 [==============================] - 2s 915us/step - loss: 0.3766 - binary_accuracy: 0.8416\n",
      "Epoch 205/225\n",
      "1755/1755 [==============================] - 2s 908us/step - loss: 0.3754 - binary_accuracy: 0.8478\n",
      "Epoch 206/225\n",
      "1755/1755 [==============================] - 2s 948us/step - loss: 0.3711 - binary_accuracy: 0.8477\n",
      "Epoch 207/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3744 - binary_accuracy: 0.8457\n",
      "Epoch 208/225\n",
      "1755/1755 [==============================] - 2s 931us/step - loss: 0.3719 - binary_accuracy: 0.8484\n",
      "Epoch 209/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.3651 - binary_accuracy: 0.8516\n",
      "Epoch 210/225\n",
      "1755/1755 [==============================] - 2s 923us/step - loss: 0.3652 - binary_accuracy: 0.8478\n",
      "Epoch 211/225\n",
      "1755/1755 [==============================] - 2s 923us/step - loss: 0.3697 - binary_accuracy: 0.8493\n",
      "Epoch 212/225\n",
      "1755/1755 [==============================] - 2s 931us/step - loss: 0.3744 - binary_accuracy: 0.8380\n",
      "Epoch 213/225\n",
      "1755/1755 [==============================] - 2s 927us/step - loss: 0.3753 - binary_accuracy: 0.8440\n",
      "Epoch 214/225\n",
      "1755/1755 [==============================] - 2s 924us/step - loss: 0.3708 - binary_accuracy: 0.8481\n",
      "Epoch 215/225\n",
      "1755/1755 [==============================] - 2s 946us/step - loss: 0.3709 - binary_accuracy: 0.8452\n",
      "Epoch 216/225\n",
      "1755/1755 [==============================] - 2s 909us/step - loss: 0.3696 - binary_accuracy: 0.8487\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 217/225\n",
      "1755/1755 [==============================] - 2s 925us/step - loss: 0.3718 - binary_accuracy: 0.8439\n",
      "Epoch 218/225\n",
      "1755/1755 [==============================] - 2s 915us/step - loss: 0.3672 - binary_accuracy: 0.8522\n",
      "Epoch 219/225\n",
      "1755/1755 [==============================] - 2s 918us/step - loss: 0.3701 - binary_accuracy: 0.8475\n",
      "Epoch 220/225\n",
      "1755/1755 [==============================] - 2s 919us/step - loss: 0.3699 - binary_accuracy: 0.8429\n",
      "Epoch 221/225\n",
      "1755/1755 [==============================] - 2s 920us/step - loss: 0.3713 - binary_accuracy: 0.8459\n",
      "Epoch 222/225\n",
      "1755/1755 [==============================] - 2s 922us/step - loss: 0.3681 - binary_accuracy: 0.8501\n",
      "Epoch 223/225\n",
      "1755/1755 [==============================] - 2s 910us/step - loss: 0.3671 - binary_accuracy: 0.8469\n",
      "Epoch 224/225\n",
      "1755/1755 [==============================] - 2s 923us/step - loss: 0.3738 - binary_accuracy: 0.8437\n",
      "Epoch 225/225\n",
      "1755/1755 [==============================] - 2s 927us/step - loss: 0.3685 - binary_accuracy: 0.8504\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/225\n",
      "1754/1754 [==============================] - 3s 2ms/step - loss: 0.4560 - binary_accuracy: 0.8241\n",
      "Epoch 2/225\n",
      "1754/1754 [==============================] - 2s 914us/step - loss: 0.4324 - binary_accuracy: 0.8189\n",
      "Epoch 3/225\n",
      "1754/1754 [==============================] - 2s 902us/step - loss: 0.4287 - binary_accuracy: 0.8249\n",
      "Epoch 4/225\n",
      "1754/1754 [==============================] - 2s 909us/step - loss: 0.4231 - binary_accuracy: 0.8242\n",
      "Epoch 5/225\n",
      "1754/1754 [==============================] - 2s 891us/step - loss: 0.4154 - binary_accuracy: 0.8270\n",
      "Epoch 6/225\n",
      "1754/1754 [==============================] - 2s 887us/step - loss: 0.4209 - binary_accuracy: 0.8258\n",
      "Epoch 7/225\n",
      "1754/1754 [==============================] - 2s 878us/step - loss: 0.4133 - binary_accuracy: 0.8284\n",
      "Epoch 8/225\n",
      "1754/1754 [==============================] - 2s 878us/step - loss: 0.4212 - binary_accuracy: 0.8232\n",
      "Epoch 9/225\n",
      "1754/1754 [==============================] - 2s 889us/step - loss: 0.4101 - binary_accuracy: 0.8307\n",
      "Epoch 10/225\n",
      "1754/1754 [==============================] - 2s 917us/step - loss: 0.4075 - binary_accuracy: 0.8302\n",
      "Epoch 11/225\n",
      "1754/1754 [==============================] - 2s 877us/step - loss: 0.4125 - binary_accuracy: 0.8307\n",
      "Epoch 12/225\n",
      "1754/1754 [==============================] - 2s 885us/step - loss: 0.4092 - binary_accuracy: 0.8285\n",
      "Epoch 13/225\n",
      "1754/1754 [==============================] - 2s 903us/step - loss: 0.4079 - binary_accuracy: 0.8295\n",
      "Epoch 14/225\n",
      "1754/1754 [==============================] - 2s 901us/step - loss: 0.4077 - binary_accuracy: 0.8291\n",
      "Epoch 15/225\n",
      "1754/1754 [==============================] - 2s 898us/step - loss: 0.4088 - binary_accuracy: 0.8320\n",
      "Epoch 16/225\n",
      "1754/1754 [==============================] - 2s 890us/step - loss: 0.4029 - binary_accuracy: 0.8316\n",
      "Epoch 17/225\n",
      "1754/1754 [==============================] - 2s 882us/step - loss: 0.4036 - binary_accuracy: 0.8296\n",
      "Epoch 18/225\n",
      "1754/1754 [==============================] - 2s 880us/step - loss: 0.4011 - binary_accuracy: 0.8305\n",
      "Epoch 19/225\n",
      "1754/1754 [==============================] - 2s 894us/step - loss: 0.4110 - binary_accuracy: 0.8298\n",
      "Epoch 20/225\n",
      "1754/1754 [==============================] - 2s 886us/step - loss: 0.4048 - binary_accuracy: 0.8274\n",
      "Epoch 21/225\n",
      "1754/1754 [==============================] - 2s 884us/step - loss: 0.3986 - binary_accuracy: 0.8358\n",
      "Epoch 22/225\n",
      "1754/1754 [==============================] - 2s 883us/step - loss: 0.3976 - binary_accuracy: 0.8336\n",
      "Epoch 23/225\n",
      "1754/1754 [==============================] - 2s 881us/step - loss: 0.3989 - binary_accuracy: 0.8357\n",
      "Epoch 24/225\n",
      "1754/1754 [==============================] - 2s 885us/step - loss: 0.3995 - binary_accuracy: 0.8304\n",
      "Epoch 25/225\n",
      "1754/1754 [==============================] - 2s 886us/step - loss: 0.4067 - binary_accuracy: 0.8270\n",
      "Epoch 26/225\n",
      "1754/1754 [==============================] - 2s 887us/step - loss: 0.4028 - binary_accuracy: 0.8315\n",
      "Epoch 27/225\n",
      "1754/1754 [==============================] - 2s 885us/step - loss: 0.4025 - binary_accuracy: 0.8313\n",
      "Epoch 28/225\n",
      "1754/1754 [==============================] - 2s 888us/step - loss: 0.3978 - binary_accuracy: 0.8311\n",
      "Epoch 29/225\n",
      "1754/1754 [==============================] - 2s 883us/step - loss: 0.4050 - binary_accuracy: 0.8282\n",
      "Epoch 30/225\n",
      "1754/1754 [==============================] - 2s 894us/step - loss: 0.3905 - binary_accuracy: 0.8332\n",
      "Epoch 31/225\n",
      "1754/1754 [==============================] - 2s 930us/step - loss: 0.3969 - binary_accuracy: 0.8329\n",
      "Epoch 32/225\n",
      "1754/1754 [==============================] - 2s 887us/step - loss: 0.4010 - binary_accuracy: 0.8336\n",
      "Epoch 33/225\n",
      "1754/1754 [==============================] - 2s 884us/step - loss: 0.3998 - binary_accuracy: 0.8339\n",
      "Epoch 34/225\n",
      "1754/1754 [==============================] - 2s 888us/step - loss: 0.3933 - binary_accuracy: 0.8365\n",
      "Epoch 35/225\n",
      "1754/1754 [==============================] - 2s 872us/step - loss: 0.3985 - binary_accuracy: 0.8320\n",
      "Epoch 36/225\n",
      "1754/1754 [==============================] - 2s 880us/step - loss: 0.3973 - binary_accuracy: 0.8326\n",
      "Epoch 37/225\n",
      "1754/1754 [==============================] - 2s 889us/step - loss: 0.3903 - binary_accuracy: 0.8380\n",
      "Epoch 38/225\n",
      "1754/1754 [==============================] - 2s 880us/step - loss: 0.3925 - binary_accuracy: 0.8361\n",
      "Epoch 39/225\n",
      "1754/1754 [==============================] - 2s 885us/step - loss: 0.3971 - binary_accuracy: 0.8330\n",
      "Epoch 40/225\n",
      "1754/1754 [==============================] - 2s 887us/step - loss: 0.3957 - binary_accuracy: 0.8337\n",
      "Epoch 41/225\n",
      "1754/1754 [==============================] - 2s 888us/step - loss: 0.3950 - binary_accuracy: 0.8335\n",
      "Epoch 42/225\n",
      "1754/1754 [==============================] - 2s 893us/step - loss: 0.3942 - binary_accuracy: 0.8352\n",
      "Epoch 43/225\n",
      "1754/1754 [==============================] - 2s 894us/step - loss: 0.3912 - binary_accuracy: 0.8365\n",
      "Epoch 44/225\n",
      "1754/1754 [==============================] - 2s 899us/step - loss: 0.3940 - binary_accuracy: 0.8366\n",
      "Epoch 45/225\n",
      "1754/1754 [==============================] - 2s 913us/step - loss: 0.3891 - binary_accuracy: 0.8373\n",
      "Epoch 46/225\n",
      "1754/1754 [==============================] - 2s 888us/step - loss: 0.3885 - binary_accuracy: 0.8377\n",
      "Epoch 47/225\n",
      "1754/1754 [==============================] - 2s 880us/step - loss: 0.3968 - binary_accuracy: 0.8354\n",
      "Epoch 48/225\n",
      "1754/1754 [==============================] - 2s 900us/step - loss: 0.3924 - binary_accuracy: 0.8364\n",
      "Epoch 49/225\n",
      "1754/1754 [==============================] - 2s 889us/step - loss: 0.3936 - binary_accuracy: 0.8339\n",
      "Epoch 50/225\n",
      "1754/1754 [==============================] - 2s 888us/step - loss: 0.3898 - binary_accuracy: 0.8378\n",
      "Epoch 51/225\n",
      "1754/1754 [==============================] - 2s 916us/step - loss: 0.3889 - binary_accuracy: 0.8346\n",
      "Epoch 52/225\n",
      "1754/1754 [==============================] - 2s 907us/step - loss: 0.3967 - binary_accuracy: 0.8286\n",
      "Epoch 53/225\n",
      "1754/1754 [==============================] - 2s 886us/step - loss: 0.3939 - binary_accuracy: 0.8380\n",
      "Epoch 54/225\n",
      "1754/1754 [==============================] - 2s 889us/step - loss: 0.3869 - binary_accuracy: 0.8378\n",
      "Epoch 55/225\n",
      "1754/1754 [==============================] - 2s 880us/step - loss: 0.3929 - binary_accuracy: 0.8365\n",
      "Epoch 56/225\n",
      "1754/1754 [==============================] - 2s 891us/step - loss: 0.3907 - binary_accuracy: 0.8349\n",
      "Epoch 57/225\n",
      "1754/1754 [==============================] - 2s 892us/step - loss: 0.3938 - binary_accuracy: 0.8355\n",
      "Epoch 58/225\n",
      "1754/1754 [==============================] - 2s 891us/step - loss: 0.3916 - binary_accuracy: 0.8358\n",
      "Epoch 59/225\n",
      "1754/1754 [==============================] - 2s 888us/step - loss: 0.3860 - binary_accuracy: 0.8383\n",
      "Epoch 60/225\n",
      "1754/1754 [==============================] - 2s 885us/step - loss: 0.3958 - binary_accuracy: 0.8357\n",
      "Epoch 61/225\n",
      "1754/1754 [==============================] - 2s 893us/step - loss: 0.3924 - binary_accuracy: 0.8382\n",
      "Epoch 62/225\n",
      "1754/1754 [==============================] - 2s 884us/step - loss: 0.3958 - binary_accuracy: 0.8339\n",
      "Epoch 63/225\n",
      "1754/1754 [==============================] - 2s 896us/step - loss: 0.3902 - binary_accuracy: 0.8400\n",
      "Epoch 64/225\n",
      "1754/1754 [==============================] - 2s 904us/step - loss: 0.3882 - binary_accuracy: 0.8378\n",
      "Epoch 65/225\n",
      "1754/1754 [==============================] - 2s 891us/step - loss: 0.3925 - binary_accuracy: 0.8383\n",
      "Epoch 66/225\n",
      "1754/1754 [==============================] - 2s 887us/step - loss: 0.3892 - binary_accuracy: 0.8345\n",
      "Epoch 67/225\n",
      "1754/1754 [==============================] - 2s 897us/step - loss: 0.3789 - binary_accuracy: 0.8375\n",
      "Epoch 68/225\n",
      "1754/1754 [==============================] - 2s 887us/step - loss: 0.3838 - binary_accuracy: 0.8339\n",
      "Epoch 69/225\n",
      "1754/1754 [==============================] - 2s 884us/step - loss: 0.3933 - binary_accuracy: 0.8328\n",
      "Epoch 70/225\n",
      "1754/1754 [==============================] - 2s 885us/step - loss: 0.3897 - binary_accuracy: 0.8371\n",
      "Epoch 71/225\n",
      "1754/1754 [==============================] - 2s 897us/step - loss: 0.3899 - binary_accuracy: 0.8351\n",
      "Epoch 72/225\n",
      "1754/1754 [==============================] - 2s 931us/step - loss: 0.3855 - binary_accuracy: 0.8336\n",
      "Epoch 73/225\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1754/1754 [==============================] - 2s 894us/step - loss: 0.3897 - binary_accuracy: 0.8338\n",
      "Epoch 74/225\n",
      "1754/1754 [==============================] - 2s 889us/step - loss: 0.3961 - binary_accuracy: 0.8338\n",
      "Epoch 75/225\n",
      "1754/1754 [==============================] - 2s 888us/step - loss: 0.3810 - binary_accuracy: 0.8426\n",
      "Epoch 76/225\n",
      "1754/1754 [==============================] - 2s 889us/step - loss: 0.3846 - binary_accuracy: 0.8422\n",
      "Epoch 77/225\n",
      "1754/1754 [==============================] - 2s 892us/step - loss: 0.3854 - binary_accuracy: 0.8378\n",
      "Epoch 78/225\n",
      "1754/1754 [==============================] - 2s 881us/step - loss: 0.3792 - binary_accuracy: 0.8410\n",
      "Epoch 79/225\n",
      "1754/1754 [==============================] - 2s 900us/step - loss: 0.3834 - binary_accuracy: 0.8360\n",
      "Epoch 80/225\n",
      "1754/1754 [==============================] - 2s 891us/step - loss: 0.3758 - binary_accuracy: 0.8420\n",
      "Epoch 81/225\n",
      "1754/1754 [==============================] - 2s 880us/step - loss: 0.3801 - binary_accuracy: 0.8432\n",
      "Epoch 82/225\n",
      "1754/1754 [==============================] - 2s 885us/step - loss: 0.3868 - binary_accuracy: 0.8368\n",
      "Epoch 83/225\n",
      "1754/1754 [==============================] - 2s 889us/step - loss: 0.3827 - binary_accuracy: 0.8405\n",
      "Epoch 84/225\n",
      "1754/1754 [==============================] - 2s 883us/step - loss: 0.3810 - binary_accuracy: 0.8339\n",
      "Epoch 85/225\n",
      "1754/1754 [==============================] - 2s 889us/step - loss: 0.3833 - binary_accuracy: 0.8425\n",
      "Epoch 86/225\n",
      "1754/1754 [==============================] - 2s 885us/step - loss: 0.3880 - binary_accuracy: 0.8359\n",
      "Epoch 87/225\n",
      "1754/1754 [==============================] - 2s 897us/step - loss: 0.3795 - binary_accuracy: 0.8409\n",
      "Epoch 88/225\n",
      "1754/1754 [==============================] - 2s 886us/step - loss: 0.3767 - binary_accuracy: 0.8392\n",
      "Epoch 89/225\n",
      "1754/1754 [==============================] - 2s 883us/step - loss: 0.3784 - binary_accuracy: 0.8448\n",
      "Epoch 90/225\n",
      "1754/1754 [==============================] - 2s 888us/step - loss: 0.3837 - binary_accuracy: 0.8382\n",
      "Epoch 91/225\n",
      "1754/1754 [==============================] - 2s 925us/step - loss: 0.3864 - binary_accuracy: 0.8354\n",
      "Epoch 92/225\n",
      "1754/1754 [==============================] - 2s 899us/step - loss: 0.3833 - binary_accuracy: 0.8392\n",
      "Epoch 93/225\n",
      "1754/1754 [==============================] - 2s 886us/step - loss: 0.3790 - binary_accuracy: 0.8408\n",
      "Epoch 94/225\n",
      "1754/1754 [==============================] - 2s 883us/step - loss: 0.3761 - binary_accuracy: 0.8425\n",
      "Epoch 95/225\n",
      "1754/1754 [==============================] - 2s 883us/step - loss: 0.3843 - binary_accuracy: 0.8380\n",
      "Epoch 96/225\n",
      "1754/1754 [==============================] - 2s 890us/step - loss: 0.3768 - binary_accuracy: 0.8426\n",
      "Epoch 97/225\n",
      "1754/1754 [==============================] - 2s 890us/step - loss: 0.3817 - binary_accuracy: 0.8409\n",
      "Epoch 98/225\n",
      "1754/1754 [==============================] - 2s 882us/step - loss: 0.3903 - binary_accuracy: 0.8375\n",
      "Epoch 99/225\n",
      "1754/1754 [==============================] - 2s 882us/step - loss: 0.3879 - binary_accuracy: 0.8385\n",
      "Epoch 100/225\n",
      "1754/1754 [==============================] - 2s 881us/step - loss: 0.3851 - binary_accuracy: 0.8375\n",
      "Epoch 101/225\n",
      "1754/1754 [==============================] - 2s 885us/step - loss: 0.3824 - binary_accuracy: 0.8397\n",
      "Epoch 102/225\n",
      "1754/1754 [==============================] - 2s 894us/step - loss: 0.3787 - binary_accuracy: 0.8412\n",
      "Epoch 103/225\n",
      "1754/1754 [==============================] - 2s 886us/step - loss: 0.3794 - binary_accuracy: 0.8398\n",
      "Epoch 104/225\n",
      "1754/1754 [==============================] - 2s 885us/step - loss: 0.3841 - binary_accuracy: 0.8380\n",
      "Epoch 105/225\n",
      "1754/1754 [==============================] - 2s 891us/step - loss: 0.3814 - binary_accuracy: 0.8444\n",
      "Epoch 106/225\n",
      "1754/1754 [==============================] - 2s 884us/step - loss: 0.3850 - binary_accuracy: 0.8384\n",
      "Epoch 107/225\n",
      "1754/1754 [==============================] - 2s 893us/step - loss: 0.3819 - binary_accuracy: 0.8372\n",
      "Epoch 108/225\n",
      "1754/1754 [==============================] - 2s 890us/step - loss: 0.3776 - binary_accuracy: 0.8428\n",
      "Epoch 109/225\n",
      "1754/1754 [==============================] - 2s 893us/step - loss: 0.3784 - binary_accuracy: 0.8447\n",
      "Epoch 110/225\n",
      "1754/1754 [==============================] - 2s 893us/step - loss: 0.3830 - binary_accuracy: 0.8384\n",
      "Epoch 111/225\n",
      "1754/1754 [==============================] - 2s 884us/step - loss: 0.3806 - binary_accuracy: 0.8393\n",
      "Epoch 112/225\n",
      "1754/1754 [==============================] - 2s 883us/step - loss: 0.3756 - binary_accuracy: 0.8481\n",
      "Epoch 113/225\n",
      "1754/1754 [==============================] - 2s 926us/step - loss: 0.3828 - binary_accuracy: 0.8376\n",
      "Epoch 114/225\n",
      "1754/1754 [==============================] - 2s 885us/step - loss: 0.3788 - binary_accuracy: 0.8418\n",
      "Epoch 115/225\n",
      "1754/1754 [==============================] - 2s 888us/step - loss: 0.3755 - binary_accuracy: 0.8452\n",
      "Epoch 116/225\n",
      "1754/1754 [==============================] - 2s 892us/step - loss: 0.3776 - binary_accuracy: 0.8405\n",
      "Epoch 117/225\n",
      "1754/1754 [==============================] - 2s 895us/step - loss: 0.3770 - binary_accuracy: 0.8395\n",
      "Epoch 118/225\n",
      "1754/1754 [==============================] - 2s 894us/step - loss: 0.3765 - binary_accuracy: 0.8444\n",
      "Epoch 119/225\n",
      "1754/1754 [==============================] - 2s 885us/step - loss: 0.3789 - binary_accuracy: 0.8416\n",
      "Epoch 120/225\n",
      "1754/1754 [==============================] - 2s 886us/step - loss: 0.3775 - binary_accuracy: 0.8425\n",
      "Epoch 121/225\n",
      "1754/1754 [==============================] - 2s 888us/step - loss: 0.3811 - binary_accuracy: 0.8449\n",
      "Epoch 122/225\n",
      "1754/1754 [==============================] - 2s 883us/step - loss: 0.3813 - binary_accuracy: 0.8389\n",
      "Epoch 123/225\n",
      "1754/1754 [==============================] - 2s 888us/step - loss: 0.3753 - binary_accuracy: 0.8435\n",
      "Epoch 124/225\n",
      "1754/1754 [==============================] - 2s 886us/step - loss: 0.3744 - binary_accuracy: 0.8416\n",
      "Epoch 125/225\n",
      "1754/1754 [==============================] - 2s 882us/step - loss: 0.3802 - binary_accuracy: 0.8415\n",
      "Epoch 126/225\n",
      "1754/1754 [==============================] - 2s 890us/step - loss: 0.3709 - binary_accuracy: 0.8466\n",
      "Epoch 127/225\n",
      "1754/1754 [==============================] - 2s 888us/step - loss: 0.3769 - binary_accuracy: 0.8450\n",
      "Epoch 128/225\n",
      "1754/1754 [==============================] - 2s 884us/step - loss: 0.3714 - binary_accuracy: 0.8442\n",
      "Epoch 129/225\n",
      "1754/1754 [==============================] - 2s 887us/step - loss: 0.3768 - binary_accuracy: 0.8419\n",
      "Epoch 130/225\n",
      "1754/1754 [==============================] - 2s 889us/step - loss: 0.3763 - binary_accuracy: 0.8419\n",
      "Epoch 131/225\n",
      "1754/1754 [==============================] - 2s 891us/step - loss: 0.3708 - binary_accuracy: 0.8446\n",
      "Epoch 132/225\n",
      "1754/1754 [==============================] - 2s 901us/step - loss: 0.3733 - binary_accuracy: 0.8433\n",
      "Epoch 133/225\n",
      "1754/1754 [==============================] - 2s 922us/step - loss: 0.3714 - binary_accuracy: 0.8400\n",
      "Epoch 134/225\n",
      "1754/1754 [==============================] - 2s 893us/step - loss: 0.3759 - binary_accuracy: 0.8374\n",
      "Epoch 135/225\n",
      "1754/1754 [==============================] - 2s 884us/step - loss: 0.3810 - binary_accuracy: 0.8424\n",
      "Epoch 136/225\n",
      "1754/1754 [==============================] - 2s 886us/step - loss: 0.3733 - binary_accuracy: 0.8453\n",
      "Epoch 137/225\n",
      "1754/1754 [==============================] - 2s 883us/step - loss: 0.3726 - binary_accuracy: 0.8449\n",
      "Epoch 138/225\n",
      "1754/1754 [==============================] - 2s 887us/step - loss: 0.3694 - binary_accuracy: 0.8467\n",
      "Epoch 139/225\n",
      "1754/1754 [==============================] - 2s 891us/step - loss: 0.3729 - binary_accuracy: 0.8413\n",
      "Epoch 140/225\n",
      "1754/1754 [==============================] - 2s 893us/step - loss: 0.3766 - binary_accuracy: 0.8406\n",
      "Epoch 141/225\n",
      "1754/1754 [==============================] - 2s 892us/step - loss: 0.3750 - binary_accuracy: 0.8396\n",
      "Epoch 142/225\n",
      "1754/1754 [==============================] - 2s 887us/step - loss: 0.3754 - binary_accuracy: 0.8444\n",
      "Epoch 143/225\n",
      "1754/1754 [==============================] - 2s 890us/step - loss: 0.3700 - binary_accuracy: 0.8447\n",
      "Epoch 144/225\n",
      "1754/1754 [==============================] - 2s 890us/step - loss: 0.3691 - binary_accuracy: 0.8474\n",
      "Epoch 145/225\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1754/1754 [==============================] - 2s 890us/step - loss: 0.3732 - binary_accuracy: 0.8444\n",
      "Epoch 146/225\n",
      "1754/1754 [==============================] - 2s 884us/step - loss: 0.3731 - binary_accuracy: 0.8453\n",
      "Epoch 147/225\n",
      "1754/1754 [==============================] - 2s 887us/step - loss: 0.3810 - binary_accuracy: 0.8421\n",
      "Epoch 148/225\n",
      "1754/1754 [==============================] - 2s 887us/step - loss: 0.3759 - binary_accuracy: 0.8429\n",
      "Epoch 149/225\n",
      "1754/1754 [==============================] - 2s 883us/step - loss: 0.3723 - binary_accuracy: 0.8461\n",
      "Epoch 150/225\n",
      "1754/1754 [==============================] - 2s 914us/step - loss: 0.3714 - binary_accuracy: 0.8452\n",
      "Epoch 151/225\n",
      "1754/1754 [==============================] - 2s 894us/step - loss: 0.3700 - binary_accuracy: 0.8477\n",
      "Epoch 152/225\n",
      "1754/1754 [==============================] - 2s 897us/step - loss: 0.3758 - binary_accuracy: 0.8455\n",
      "Epoch 153/225\n",
      "1754/1754 [==============================] - 2s 888us/step - loss: 0.3760 - binary_accuracy: 0.8444\n",
      "Epoch 154/225\n",
      "1754/1754 [==============================] - 2s 897us/step - loss: 0.3685 - binary_accuracy: 0.8410\n",
      "Epoch 155/225\n",
      "1754/1754 [==============================] - 2s 912us/step - loss: 0.3672 - binary_accuracy: 0.8485\n",
      "Epoch 156/225\n",
      "1754/1754 [==============================] - 2s 892us/step - loss: 0.3697 - binary_accuracy: 0.8407\n",
      "Epoch 157/225\n",
      "1754/1754 [==============================] - 2s 891us/step - loss: 0.3753 - binary_accuracy: 0.8395\n",
      "Epoch 158/225\n",
      "1754/1754 [==============================] - 2s 884us/step - loss: 0.3659 - binary_accuracy: 0.8447\n",
      "Epoch 159/225\n",
      "1754/1754 [==============================] - 2s 890us/step - loss: 0.3731 - binary_accuracy: 0.8465\n",
      "Epoch 160/225\n",
      "1754/1754 [==============================] - 2s 902us/step - loss: 0.3747 - binary_accuracy: 0.8403\n",
      "Epoch 161/225\n",
      "1754/1754 [==============================] - 2s 899us/step - loss: 0.3751 - binary_accuracy: 0.8443\n",
      "Epoch 162/225\n",
      "1754/1754 [==============================] - 2s 891us/step - loss: 0.3694 - binary_accuracy: 0.8446\n",
      "Epoch 163/225\n",
      "1754/1754 [==============================] - 2s 885us/step - loss: 0.3718 - binary_accuracy: 0.8450\n",
      "Epoch 164/225\n",
      "1754/1754 [==============================] - 2s 893us/step - loss: 0.3666 - binary_accuracy: 0.8453\n",
      "Epoch 165/225\n",
      "1754/1754 [==============================] - 2s 889us/step - loss: 0.3744 - binary_accuracy: 0.8480\n",
      "Epoch 166/225\n",
      "1754/1754 [==============================] - 2s 891us/step - loss: 0.3654 - binary_accuracy: 0.8466\n",
      "Epoch 167/225\n",
      "1754/1754 [==============================] - 2s 883us/step - loss: 0.3746 - binary_accuracy: 0.8419\n",
      "Epoch 168/225\n",
      "1754/1754 [==============================] - 2s 885us/step - loss: 0.3718 - binary_accuracy: 0.8457\n",
      "Epoch 169/225\n",
      "1754/1754 [==============================] - 2s 876us/step - loss: 0.3651 - binary_accuracy: 0.8482\n",
      "Epoch 170/225\n",
      "1754/1754 [==============================] - 2s 903us/step - loss: 0.3693 - binary_accuracy: 0.8457\n",
      "Epoch 171/225\n",
      "1754/1754 [==============================] - 2s 923us/step - loss: 0.3723 - binary_accuracy: 0.8413\n",
      "Epoch 172/225\n",
      "1754/1754 [==============================] - 2s 899us/step - loss: 0.3713 - binary_accuracy: 0.8432\n",
      "Epoch 173/225\n",
      "1754/1754 [==============================] - 2s 887us/step - loss: 0.3756 - binary_accuracy: 0.8426\n",
      "Epoch 174/225\n",
      "1754/1754 [==============================] - 2s 886us/step - loss: 0.3786 - binary_accuracy: 0.8384\n",
      "Epoch 175/225\n",
      "1754/1754 [==============================] - 2s 884us/step - loss: 0.3732 - binary_accuracy: 0.8403\n",
      "Epoch 176/225\n",
      "1754/1754 [==============================] - 2s 881us/step - loss: 0.3634 - binary_accuracy: 0.8478\n",
      "Epoch 177/225\n",
      "1754/1754 [==============================] - 2s 890us/step - loss: 0.3677 - binary_accuracy: 0.8460\n",
      "Epoch 178/225\n",
      "1754/1754 [==============================] - 2s 890us/step - loss: 0.3778 - binary_accuracy: 0.8379\n",
      "Epoch 179/225\n",
      "1754/1754 [==============================] - 2s 887us/step - loss: 0.3753 - binary_accuracy: 0.8414\n",
      "Epoch 180/225\n",
      "1754/1754 [==============================] - 2s 891us/step - loss: 0.3725 - binary_accuracy: 0.8385\n",
      "Epoch 181/225\n",
      "1754/1754 [==============================] - 2s 898us/step - loss: 0.3654 - binary_accuracy: 0.8462\n",
      "Epoch 182/225\n",
      "1754/1754 [==============================] - 2s 888us/step - loss: 0.3607 - binary_accuracy: 0.8489\n",
      "Epoch 183/225\n",
      "1754/1754 [==============================] - 2s 891us/step - loss: 0.3682 - binary_accuracy: 0.8454\n",
      "Epoch 184/225\n",
      "1754/1754 [==============================] - 2s 902us/step - loss: 0.3695 - binary_accuracy: 0.8458\n",
      "Epoch 185/225\n",
      "1754/1754 [==============================] - 2s 892us/step - loss: 0.3729 - binary_accuracy: 0.8437\n",
      "Epoch 186/225\n",
      "1754/1754 [==============================] - 2s 878us/step - loss: 0.3673 - binary_accuracy: 0.8438\n",
      "Epoch 187/225\n",
      "1754/1754 [==============================] - 2s 890us/step - loss: 0.3743 - binary_accuracy: 0.8399\n",
      "Epoch 188/225\n",
      "1754/1754 [==============================] - 2s 879us/step - loss: 0.3696 - binary_accuracy: 0.8480\n",
      "Epoch 189/225\n",
      "1754/1754 [==============================] - 2s 880us/step - loss: 0.3667 - binary_accuracy: 0.8495\n",
      "Epoch 190/225\n",
      "1754/1754 [==============================] - 2s 908us/step - loss: 0.3696 - binary_accuracy: 0.8444\n",
      "Epoch 191/225\n",
      "1754/1754 [==============================] - 2s 908us/step - loss: 0.3753 - binary_accuracy: 0.8416\n",
      "Epoch 192/225\n",
      "1754/1754 [==============================] - 2s 889us/step - loss: 0.3674 - binary_accuracy: 0.8453\n",
      "Epoch 193/225\n",
      "1754/1754 [==============================] - 2s 883us/step - loss: 0.3787 - binary_accuracy: 0.8403\n",
      "Epoch 194/225\n",
      "1754/1754 [==============================] - 2s 884us/step - loss: 0.3650 - binary_accuracy: 0.8496\n",
      "Epoch 195/225\n",
      "1754/1754 [==============================] - 2s 890us/step - loss: 0.3701 - binary_accuracy: 0.8442\n",
      "Epoch 196/225\n",
      "1754/1754 [==============================] - 2s 874us/step - loss: 0.3665 - binary_accuracy: 0.8436\n",
      "Epoch 197/225\n",
      "1754/1754 [==============================] - 2s 882us/step - loss: 0.3691 - binary_accuracy: 0.8447\n",
      "Epoch 198/225\n",
      "1754/1754 [==============================] - 2s 874us/step - loss: 0.3677 - binary_accuracy: 0.8471\n",
      "Epoch 199/225\n",
      "1754/1754 [==============================] - 2s 890us/step - loss: 0.3629 - binary_accuracy: 0.8467\n",
      "Epoch 200/225\n",
      "1754/1754 [==============================] - 2s 891us/step - loss: 0.3706 - binary_accuracy: 0.8445\n",
      "Epoch 201/225\n",
      "1754/1754 [==============================] - 2s 905us/step - loss: 0.3643 - binary_accuracy: 0.8433\n",
      "Epoch 202/225\n",
      "1754/1754 [==============================] - 2s 886us/step - loss: 0.3712 - binary_accuracy: 0.8439\n",
      "Epoch 203/225\n",
      "1754/1754 [==============================] - 2s 882us/step - loss: 0.3660 - binary_accuracy: 0.8470\n",
      "Epoch 204/225\n",
      "1754/1754 [==============================] - 2s 892us/step - loss: 0.3687 - binary_accuracy: 0.8477\n",
      "Epoch 205/225\n",
      "1754/1754 [==============================] - 2s 894us/step - loss: 0.3682 - binary_accuracy: 0.8459\n",
      "Epoch 206/225\n",
      "1754/1754 [==============================] - 2s 889us/step - loss: 0.3689 - binary_accuracy: 0.8465\n",
      "Epoch 207/225\n",
      "1754/1754 [==============================] - 2s 887us/step - loss: 0.3685 - binary_accuracy: 0.8444\n",
      "Epoch 208/225\n",
      "1754/1754 [==============================] - 2s 885us/step - loss: 0.3648 - binary_accuracy: 0.8449\n",
      "Epoch 209/225\n",
      "1754/1754 [==============================] - 2s 885us/step - loss: 0.3663 - binary_accuracy: 0.8452\n",
      "Epoch 210/225\n",
      "1754/1754 [==============================] - 2s 889us/step - loss: 0.3678 - binary_accuracy: 0.8452\n",
      "Epoch 211/225\n",
      "1754/1754 [==============================] - 2s 883us/step - loss: 0.3621 - binary_accuracy: 0.8496\n",
      "Epoch 212/225\n",
      "1754/1754 [==============================] - 2s 900us/step - loss: 0.3667 - binary_accuracy: 0.8477\n",
      "Epoch 213/225\n",
      "1754/1754 [==============================] - 2s 900us/step - loss: 0.3678 - binary_accuracy: 0.8476\n",
      "Epoch 214/225\n",
      "1754/1754 [==============================] - 2s 882us/step - loss: 0.3688 - binary_accuracy: 0.8439\n",
      "Epoch 215/225\n",
      "1754/1754 [==============================] - 2s 890us/step - loss: 0.3611 - binary_accuracy: 0.8495\n",
      "Epoch 216/225\n",
      "1754/1754 [==============================] - 2s 882us/step - loss: 0.3598 - binary_accuracy: 0.8515\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 217/225\n",
      "1754/1754 [==============================] - 2s 880us/step - loss: 0.3687 - binary_accuracy: 0.8511\n",
      "Epoch 218/225\n",
      "1754/1754 [==============================] - 2s 895us/step - loss: 0.3641 - binary_accuracy: 0.8463\n",
      "Epoch 219/225\n",
      "1754/1754 [==============================] - 2s 886us/step - loss: 0.3722 - binary_accuracy: 0.8441\n",
      "Epoch 220/225\n",
      "1754/1754 [==============================] - 2s 889us/step - loss: 0.3721 - binary_accuracy: 0.8433\n",
      "Epoch 221/225\n",
      "1754/1754 [==============================] - 2s 882us/step - loss: 0.3618 - binary_accuracy: 0.8467\n",
      "Epoch 222/225\n",
      "1754/1754 [==============================] - 2s 884us/step - loss: 0.3635 - binary_accuracy: 0.8459\n",
      "Epoch 223/225\n",
      "1754/1754 [==============================] - 2s 885us/step - loss: 0.3569 - binary_accuracy: 0.8494\n",
      "Epoch 224/225\n",
      "1754/1754 [==============================] - 2s 882us/step - loss: 0.3590 - binary_accuracy: 0.8492\n",
      "Epoch 225/225\n",
      "1754/1754 [==============================] - 2s 893us/step - loss: 0.3649 - binary_accuracy: 0.8479\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAEWCAYAAAB8LwAVAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjAsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+17YcXAAAgAElEQVR4nOydd3hcZ5m373f6aIq6ZDVbLnLc4hI7TiEkIY0UUiBAwlK2fWTZb/koCwthK3WXDSxtNwHCLrC7LMkGQgkkYAjBkB47iZ24V9lW75qiadK83x+naEYaySWeSIqe+7q4MnPOe868I5Lzm6crrTWCIAiCMBHHTG9AEARBmJ2IQAiCIAgFEYEQBEEQCiICIQiCIBREBEIQBEEoiAiEIAiCUBARCEF4hSilvquU+uwprm1VSl31Su8jCK8GIhCCIAhCQUQgBEEQhIKIQAjzAtO181dKqZeUUnGl1H8opWqVUr9QSkWVUo8qpcpz1t+klNqtlBpSSm1VSq3MObdBKfWCed3/Ar4Jn/UmpdQO89qnlFJrz3DP71VKHVJKDSilHlJK1ZvHlVLqy0qpHqXUsPmd1pjnrldK7TH31q6U+ugZ/cEEAREIYX5xK3A1sBy4EfgF8NdAFcZ/Cx8AUEotB+4DPgRUA48AP1NKeZRSHuAnwH8DFcAPzPtiXnse8G3gz4BK4JvAQ0op7+lsVCl1BfBPwNuBOuAYcL95+hrgUvN7lAG3Af3muf8A/kxrHQLWAI+dzucKQi4iEMJ84l+11t1a63bgceBZrfWLWusU8GNgg7nuNuBhrfWvtdYZ4IuAH7gYuBBwA1/RWme01j8EtuV8xnuBb2qtn9Vaj2mt/xNImdedDu8Evq21fsHc3yeAi5RSzUAGCAErAKW13qu17jSvywCrlFJhrfWg1vqF0/xcQbARgRDmE905rxMF3gfN1/UYv9gB0FpngRNAg3muXed3uTyW83oR8BHTvTSklBoCmszrToeJe4hhWAkNWuvHgH8D7ga6lVL3KqXC5tJbgeuBY0qp3ymlLjrNzxUEGxEIQZhMB8aDHjB8/hgP+XagE2gwj1kszHl9Avic1ros538lWuv7XuEeAhguq3YArfXXtNYbgdUYrqa/Mo9v01rfDNRguMIeOM3PFQQbEQhBmMwDwA1KqSuVUm7gIxhuoqeAp4FR4ANKKZdS6i3A5pxrvwW8Tyl1gRlMDiilblBKhU5zD98H/lgptd6MX/wjhkusVSl1vnl/NxAHksCYGSN5p1Kq1HSNRYCxV/B3EOY5IhCCMAGt9X7gXcC/An0YAe0btdZprXUaeAvwR8AgRrziRznXbseIQ/ybef6QufZ09/Ab4O+ABzGslqXA7ebpMIYQDWK4ofox4iQA7wZalVIR4H3m9xCEM0LJwCBBEAShEGJBCIIgCAURgRAEQRAKIgIhCIIgFKSoAqGUulYptd9sF3DnNOveqpTSSqlNOcfWKqWeNtsdvKyU8k11vSAIgnD2KVqQWinlBA5gtDZow6g2fYfWes+EdSHgYcADvF9rvV0p5QJeAN6ttd6plKoEhrTWU6bsVVVV6ebm5qJ8F0EQhNcqzz//fJ/WurrQOVcRP3czcEhrfQRAKXU/cDOwZ8K6zwB3AblNxa4BXtJa7wTQWvdzEpqbm9m+ffvZ2LcgCMK8QSl1bKpzxXQxNWBUlVq0mcdslFIbgCat9c8nXLsc0EqpLWbXzI8V+gCl1B1Kqe1Kqe29vb1nc++CIAjznmIKhCpwzPZnKaUcwJcxqlQn4gIuwWhYdgnwZqXUlZNupvW9WutNWutN1dUFLSRBEAThDCmmQLRh9K+xaMToL2NhtSPeqpRqxeh2+ZAZqG4Dfqe17tNaj2C0Wz6viHsVBEEQJlDMGMQ2oEUptRijwdjtwB9YJ7XWwxh9+AFQSm0FPmoGqQ8DH1NKlQBp4DIMa0MQBOGskslkaGtrI5lMzvRWiorP56OxsRG3233K1xRNILTWo0qp9wNbACdGb/vdSqlPA9u11g9Nc+2gUupLGCKjgUe01g8Xa6+CIMxf2traCIVCNDc3k9+k97WD1pr+/n7a2tpYvHjxKV9XTAsCrfUjGO6h3GN/P8Xayye8/x7wvaJtThAEAUgmk69pcQBQSlFZWcnpJvNIJbUgCPOe17I4WJzJd5z3AtE5nOBLv9rPkd7YTG9FEARhVjHvBaI3muJrjx3iaF98prciCMI8ZGhoiHvuuee0r7v++usZGhoqwo7GmfcC4XU5AUiNZmd4J4IgzEemEoixsemHAT7yyCOUlZUVa1tAkYPUcwGPy9DItAiEIAgzwJ133snhw4dZv349brebYDBIXV0dO3bsYM+ePdxyyy2cOHGCZDLJBz/4Qe644w5gvL1QLBbjuuuu45JLLuGpp56ioaGBn/70p/j9/le8t3kvEF5TIFKjMrpXEOY7n/rZbvZ0RM7qPVfVh/mHG1dPef7zn/88u3btYseOHWzdupUbbriBXbt22emo3/72t6moqCCRSHD++edz6623UllZmXePgwcPct999/Gtb32Lt7/97Tz44IO8612vfNrsvBcIsSAEQZhNbN68Oa9W4Wtf+xo//vGPAThx4gQHDx6cJBCLFy9m/fr1AGzcuJHW1tazshcRCNuCEIEQhPnOdL/0Xy0CgYD9euvWrTz66KM8/fTTlJSUcPnllxes+PZ6vfZrp9NJIpE4K3uRILUIhCAIM0goFCIajRY8Nzw8THl5OSUlJezbt49nnnnmVd2bWBBOcTEJgjBzVFZW8rrXvY41a9bg9/upra21z1177bV84xvfYO3atZxzzjlceOGFr+re5r1AKKXwOB2kx0QgBEGYGb7//e8XPO71evnFL35R8JwVZ6iqqmLXrl328Y9+9KMF158J897FBIabKZURgRAEQchFBAIjUJ0+SVGKIAjCfEMEArEgBGG+o7U++aI5zpl8RxEILAtCBEIQ5iM+n4/+/v7XtEhY8yB8Pt9pXTfvg9RgCoRkMQnCvKSxsZG2trbTnpUw17Amyp0OIhAYDfukDkIQ5idut/u0pqzNJ8TFhFgQgiAIhRCBwCiWk2Z9giAI+YhAAF63WBCCIAgTEYHAsiBEIARBEHIRgQC8bqdYEIIgCBMQgUAsCEEQhEKIQCCFcoIgCIUQgcBqtSFZTIIgCLmIQGAIhFgQgiAI+YhAYLiYUqPZ13QvFkEQhNNFBALDgtAaRrMiEIIgCBYiEBgWBMjYUUEQhFxEIDCa9QGS6ioIgpCDCARiQQiCIBSiqAKhlLpWKbVfKXVIKXXnNOveqpTSSqlNE44vVErFlFJnbwp3ATxO488gDfsEQRDGKZpAKKWcwN3AdcAq4B1KqVUF1oWADwDPFrjNl4FfFGuPFl63WBCCIAgTKaYFsRk4pLU+orVOA/cDNxdY9xngLiCZe1ApdQtwBNhdxD0CuRaECIQgCIJFMQWiATiR877NPGajlNoANGmtfz7heAD4OPCp6T5AKXWHUmq7Umr7KxkXaMUgRCAEQRDGKaZAqALH7EIDpZQDw4X0kQLrPgV8WWsdm+4DtNb3aq03aa03VVdXn/FGrSwmcTEJgiCMU8yZ1G1AU877RqAj530IWANsVUoBLAAeUkrdBFwAvFUpdRdQBmSVUkmt9b8VY6N2FpO02xAEQbAppkBsA1qUUouBduB24A+sk1rrYaDKeq+U2gp8VGu9HXh9zvFPArFiiQMYldSANOwTBEHIoWguJq31KPB+YAuwF3hAa71bKfVp00qYNXjFghAEQZhEMS0ItNaPAI9MOPb3U6y9fIrjnzzrG5uAHaTOiEAIgiBYSCU1OUFqsSAEQRBsRCCQVhuCIAiFEIEgtw5CgtSCIAgWIhDkBKnFghAEQbARgQBcDoVSUkktCIKQiwgEoJQy5lKLQAiCINiIQJh4nA6xIARBEHIQgTDxuJwiEIIgCDmIQJiIi0kQBCEfEQiTkM/FcCI909sQBEGYNYhAmLTUhtjbGZ3pbQiCIMwaRCBMVtWFaR9KMDySmemtCIIgzApEIExW1YcB2NsVmeGdCIIgzA5EIExW1oUA2NMhAiEIggAiEDY1IR9VQS97OkUgBEEQQAQij1X1YbEgBEEQTEQgclhZF+JgT5RRmQshCIIgApFLc2WAzJimK5Kc6a0IgiDMOCIQOTSU+QFoH0zM8E4EQRBmHhGIHBrKTYEYEoEQBEEQgchBLAhBEIRxRCBy8LmdVAU9YkEIgiAgAjGJhjK/CIQgCAIiEJNoKPeLi0kQBAERiElYFoTWeqa3IgiCMKOIQEygocxPajRLX0xmQwiCML8RgZhAY3kJIKmugiAIIhATsGshJA4hCMI8RwRiArVhHwA9UWm3IQjC/EYEYgJlfjdOh6IvlprprQiCIMwoIhATcDgUlQEPfdH8IHV6NEsyMzZDuxIEQXj1KapAKKWuVUrtV0odUkrdOc26tyqltFJqk/n+aqXU80qpl81/XlHMfU6kMuilL5ZieCTDQzs7APi7n+zijv9+/tXchiAIwoziKtaNlVJO4G7gaqAN2KaUekhrvWfCuhDwAeDZnMN9wI1a6w6l1BpgC9BQrL1OpCrooS+W4gfPn+CzD+9lc3MFB3qi9EvqqyAI84hiWhCbgUNa6yNa6zRwP3BzgXWfAe4C7Kiw1vpFrXWH+XY34FNKeYu41zyqg176YmlODIwARsC6P5Ymmsy8WlsQBEGYcYopEA3AiZz3bUywApRSG4AmrfXPp7nPrcCLWutJUWOl1B1Kqe1Kqe29vb1nY88AVJoWxAkz1bUvlqI/liKSHJUKa0EQ5g3FFAhV4Jj9dFVKOYAvAx+Z8gZKrQb+GfizQue11vdqrTdprTdVV1e/wu2OUxX0khrNsr8rCkDbYIJ4eoyxrGYkLYFqQRDmB8UUiDagKed9I9CR8z4ErAG2KqVagQuBh3IC1Y3Aj4H3aK0PF3Gfk6gKGt4sq5p6nykUANHk6Ku5FUEQhBmjmAKxDWhRSi1WSnmA24GHrJNa62GtdZXWullr3Qw8A9yktd6ulCoDHgY+obV+soh7LEhVKD/ccSBHICIShxAEYZ5QNIHQWo8C78fIQNoLPKC13q2U+rRS6qaTXP5+YBnwd0qpHeb/aoq114lUBjx57/fnWRAiEIIgzA+KluYKoLV+BHhkwrG/n2Lt5TmvPwt8tph7m47qHAsi4HESTY27lSIJcTEJgjA/kErqAlTkWBBrG8vyzomLSRCE+YIIRAHcTgflJW5CPhdLqgN55yISpBYEYZ5QVBfTXKYy6MXtdNgZTdUhL73RFJGEWBCCIMwPRCCm4IZz6/C5nQR9xp+ooczPcCIjLiZBEOYNIhBT8OGrlwPwy12dgNGfKexzSR2EIAjzBolBnIRK08VUGfAS9rnFxSQIwrxBBOIkWDGIyqCHkN8tQWpBEOYNIhAnYUHYR8jrYml10HQxiQUhCML8QATiJPg9Tp78xBW8eUNDnovpJy+285mf7znJ1YIgCHMXEYhTIOxz43Aown6X7WL6yY52vvtUq4whFQThNYsIxGkQ8rltF9PxgRHGspo9nZEZ3pUgCEJxEIE4DcI+F8lMlmRmjLYBoxX4y23DJ73uP59q5btPHi329gRBEM4qIhCnQdjvBuBgd4z0WBaAl05BIH70Qhv/u72tqHsTBEE420ih3GkQ9hkCsaNtCICQ18VL5uvpGEpkpH5CEIQ5h1gQp8GKuhAAP9thDMa7enUth3pjxFPT10YMjWQYHMmQkHGlgiDMIU5JIJRSH1RKhZXBfyilXlBKXVPszc02zqkNURX08FzrAC6H4vo1dWgNTxzqA0BrzYvHBxnL2qO3yWa13b+pK5KckX0LgiCcCadqQfyJ1joCXANUA38MfL5ou5qlKKW4aGkVAA3lfi47p5qFFSV87TcHyWY1P9jexpvveYrH9vXY10STo2hTLzqHEzOxbUEQhDPiVAVCmf+8HviO1npnzrF5xSXLKgFYWFGC2+ngg1e2sLsjwud/uc8unDvWH7fXDyXS9uuuYbEgBEGYO5yqQDyvlPoVhkBsUUqFgGzxtjV7udi0IBZWlABwy4YGNjdXcO/vjzCmNV6Xg/ahcUthOCc43SkCIQjCHOJUs5j+FFgPHNFajyilKjDcTPOOpooSPnBlC1esqAHA6VD8759dSE80xVhW84fffo6OHIEYGhkXiKksiEgyw/BIhiZTdARBEGYDp2pBXATs11oPKaXeBfwtcPICgNcof3n1ctY3jc+qVkpRG/ZRX+anrsxPx9C4EAyZFoTH5ZjSgvjSrw7wtm88jda64HlBEISZ4FQF4uvAiFJqHfAx4BjwX0Xb1RymocyXF4y2XEzLa4N0RQoHqY8PjNAVSdIbS70qexQEQTgVTlUgRrXx8/Zm4Kta668CoeJta+5SX+qnL5a2m/gNjxhB6nNqw1O6mHqjhjDs74q+OpsUBEE4BU5VIKJKqU8A7wYeVko5AXfxtjV3qS/zA+MB6aGRDH63k0WVJfTF0qRGJxfLFRKIaDLD7fc+zeHe2Kuwa0EQhMmcqkDcBqQw6iG6gAbgC0Xb1RzGEggrUD2cyFBW4mZBqQ+Anki+Gymb1fSZrqUD3eMCcaA7xjNHBthx/OStPARBEIrBKQmEKQr/A5Qqpd4EJLXWEoMoQH2ZIQRWqutQIkOp302dKRATA9VDiQyjZuV1rgVhWRUjaRlxKgjCzHCqrTbeDjwHvA14O/CsUuqtxdzYXMWyFO757SGu/crv6YkkJwhEfqDaEoLasJcD3TGyplhYAeu49G8SBGGGONU6iL8Bztda9wAopaqBR4EfFmtjcxWvy0l1yEtr/wgASsE1q2pZUGq4niYGqi2BeN2yKn70Qjt7OiOsaSgdtyBO0ghQEAShWJxqDMJhiYNJ/2lcO+9484YG3vv6xXicDrSGUr+boNdFyOua5GLqjRnv37S2Do/LwZvveZKfvNhuC4RYEIIgzBSnakH8Uim1BbjPfH8b8EhxtjT3+evrVwKwuyPCU4f7KSvxAIb7aSoL4vzmCh77yGX8wbee5Sc72nE5DP2VGIQgCDPFqQap/wq4F1gLrAPu1Vp/vJgbey1w6fJqwLAgwBCIzshkgfC5HQS9LhrLSzi3sZTWvvh4DCIlFoQgCDPDKbuJtNYPaq3/Umv9Ya31j0/lGqXUtUqp/UqpQ0qpO6dZ91allFZKbco59gnzuv1KqTee6j5nE5e2GAJRFTQsiLpSH10TgtQ90RTVIS9KGc1xmytLODGYoNPMgrIsiMxYlo88sPOUJtgJgiCcDaZ1MSmlokChBkEK0Frr8DTXOoG7gauBNmCbUuohrfWeCetCwAeAZ3OOrQJuB1YD9cCjSqnlWus59XN6VX2Y//yTzZzfXA7AglI/PdEUmbEsbqehzb3RFNVBr31Nc2WAsaymJ5pvQRzsjvHgC2283D7Ewx94vX29IAhCsZj2KaO1DmmtwwX+F5pOHEw2A4e01ke01mngfoxWHRP5DHAXkOt7uRm4X2ud0lofBQ6Z95tzXLa8mhKPocN1pT60Ho87gCkQoXGBWFwVyLvesiAO9hg1Ege6Y3znyaPF3rYgCEJRM5EagBM579vMYzZKqQ1Ak9b656d7rXn9HUqp7Uqp7b29vWdn10VkwYRiud5oio6hRJ5ANOcIhEONZzEd6I7icijWN5XxU3MmtiAIQjEppkAUmjhnu6uUUg7gy8BHTvda+4DW92qtN2mtN1VXV5/xRl8trGK5ruEk3ZEkt379KbLaSIu1qAx4CHkNi6Oh3G/XQRzojtFcFWBlXYjuiHR9FQSh+BRTINqAppz3jUDuT98QsAbYqpRqBS4EHjID1Se7dk5SFzaK5XZ1DPPFLfvpiiT5/nsvYOOiCnuNUsq2IporA7YFcbA7yvLaILVhH/3xFKnRMW775tP8dEf7q/9FBEGYFxRTILYBLUqpxUopD0bQ+SHrpNZ6WGtdpbVu1lo3A88AN2mtt5vrbldKeZVSi4EWjFYfc5rSEjfXrVnAvz9+hAdfaOM9Fy5iw8LySessgVhUWcJIepRkZoxjAyO01ISoDRtxjN0dEZ49OsCnfraHSDIz6R6CIAivlKIJhNZ6FHg/sAXYCzygtd6tlPq0Uuqmk1y7G3gA2AP8EviLuZbBNBWfe/O5lJd4KPG4+L9vWFZwzcaFZTSU+VkQ9pEZ0+ztjKA1LK8NsSBsuKmebx0EYCCe5p7fHn7V9i8IwvzhVCupzwit9SNMqLjWWv/9FGsvn/D+c8Dnira5GaIi4OEH77uIaHKUioCn4Jo/vLiZd1/UzH893QrAjhNG7cPy2iDpsSwA21oHAGipCfKznR3ced0KANM6aeenf/E6PK7p9f9Ad5RkZoy1jWWTzrUNjlBW4iHoLeq/IoIgzGIkmX4GWFQZYE1D6ZTnlVI4HYqAmR67tzMCwMLKknEL4phhQbxuWRW90ZQ9z/rpw/3s7Yzwk1OITfzjI3v5yAM7C5677ZvP8LmH9576lzpLpEezMptbEGYJIhCzmBKvE4AjvXGqgh68LiflJR7cTkV/PE1t2EtjuZ/0WNaefX20Lw7A17ceZiw7/YO2J5KitT/OqGmVWIyOZekYTvDU4b5T2ufjB3uJJDOkR7Ns2d11xg/4zFiWiz//GD94vu2MrhcE4ewiAjGLsSyIw70xak3LweFQ1ISM103lJdSYx3vNCu3jAyMsrw1ytC/Ov/xq/7T3H4inyYxpTgzmt/8YGEmjNRzrH6E7UniOtkUsNcoffvs5/v3xo/xsZwd/9t/Ps7sjckbftyeaoi+W4lh//IyuFwTh7CICMYsp8RgWxOBIxq6hAGO4EEBTRYndpqMnmqJtMMFoVvN/Xr+Ed2xu4p6th/n61sIBbK01A/E0AId78ude98fS9uvnjg5Mu8fBeJqshpfahuw+UccHRk7na9pYnW6lQaEgzA4kAjmLCeQEiC0LAsYrspvK/dSELYFIkh41XEVLqwPcel4j8dQY//zLfdSX+bh5fX4hejw9Zge8j/TFgFr7XK5AbGsd4MZ19VPu0XJt7WofJmK+bp9gkZwqlkAkZAaGIMwKxIKYxVgWBJBnQVgupsaKEmrMNh290RRHzPjD4qogTofiC29by/nN5dz54MskM2P8y6/2883fGRbFQI4IHOnNj0P0ma3GG8r8BS2Ixw/2kswYD/GhkYx5TZqdbcPA+Dzu06XLdGfFZQaGIMwKRCBmMSe3IEoIel343A56IimO9sUo9bspLzHmT3hdTu64dCmJzBgvHB/kO0+28hOzj1N/3BABpeDF40Nc/PnH+K7ZBNASiMvOqeZgz/icbIBDPVHe/R/P8Z0nWwEYSowLjRUUbzuJBfH04X5eNsUkFyveIRaEIMwORCBmMbkWxIIcC2J1fZgSj5PltUGUMoLWPdEUR/viNFcF7NkSABsXGZXa33vmGLHUKMf643nxhxULwuzvjtITTfHMEcNa6IulcTsVLTVBxrKagZE09z13nEf3dLPNLND75e4uYNzFZLGosuSkFsQ/PLSLzzy8Z9Jxq4mhWBCCMDsQgZjFWG3CId/F9PqWal76h2uoNAPUNSGv4WLqjbNkQrvwioCHJdUBfrHLeKCPpMfoi6XpNwXCmlUBsL/baCneH0tRGfDaVktPJMVXHz3IP/5ir11/sfPEEB1DCdvFtKiyhBKPk9e3VNE+aASpO4YS3PxvT9A2mB+0Hohn2NsRmZQO2y0xCEGYVYhAzGKcDoXPbfxflOtiAnDlDAyqCXt5uX2YzuEk6xonF+BtWlRO7rP4WH/ctiBu2dDA5uYK3rF5Ia39cUbSo/TH01QGPXZ8o3M4QU80yZHeOFt2d7G8NgjAr3Z3MZzI4HM7eNcFi7jt/CaaykuIJEeJJjM8c6SfnW3D/O7AeCt2rTXDiTTR1OgkV9R4DEIEQhBmAyIQs5yAx0XA4yTkc0+5pjroJWa2Bb9qVe2k85abaXOz0TW2tX+EgXgar8vBhqYyHnjfRVy2vAqtjcl1fbEUlUGvHQx/uX0YKwwRTY7y5g2NLKsJ8tj+XoZG0pT63bz30iX8w42raSg3Ota2DyU43Gukz+bGG0bSY2TGjJvt7hg/rrW2BcJqcS4IwswiAjHLKfE68+IPhbCK5VbVhWksL5l0/sIllTgdindc0ITToTjWH6c/lqYy4LHjFecsMAYE7uuK0B9LUxX02Cm0O81eUG6nsXbjonJWLAhxvD/OcCJDmX+8p1RDmSkQgwkOmfUVL+UIxFBOzGJPTkHd4IhRie10KEYyYkEIwmxABGKWE/C4qCv1T7vGmkh3dQHrAYzeT49/7A3csr6BxnK/aUGkqAiOP9gXVpTgdzvZ2xmlL5aiKujF53YS9rns9NXbz19IRcDD2sZSGsr9dAwlGYxnKC0Zt24sC6JtMMHhXiPt1moKCEZhncWeznGBsGogmsr9jMxgodxIepQnD51aixFBeK0jAjHL+evrV/Lhq5dPu2ZVXRif28Gb1tZNuaa+zI9SikWVASMGMZKhIjA+6tTpUCxfEOL5Y4OkRrNUmp1ma8I+O17xsWvP4elPXIHP7aShzOgBdajXSK21qAp48bud7OmI0NoXZ2l1gNGstsXAynqqK/WxrXWQP/nuNvZ2RuwU18VVAdJjWTIT+kO9Wvz4xXbe+e/PcsAM2AvCfEYEYpZz6fJqO4YwFWsaStn9qWtpqQ2d9H7NlSUc7YszEE/ZImCxcWE5L7cb1kJVToYUQMjnIuRz43UZqbeWK2kgnqYsRyAcDsUbVlTz4x3tjGY1bzmvERiPQ1hZT1etrGU4keGxfT1s2d1Fm5kau6zGCICPzFCguscc5/rYvp4Z+XxBmE2IQLxGcDoKjfGezNLqINHkKO2DiUnzKD58dQtLqo002UrT/WQJRP0EN5flSgIoK8kPoN+4tt5u+3Hx0koqAh67ZblVWPfnly/l2b++kqYKP4d74xzuiRHwOFlcZQjETKW6WtaSCIQgiEDMO962qZHNzRVkNZMEIuRzc++7N3LNqlrWmUOErAB4XVl+oNyyIADKSvLv84YVNfagoaU1QaqCHvvBa1kQFQEPtWEfS6qCHOmNcagnxtKaIAGzxflMFcsNjBj7fP7Y4KQiQPbHFusAACAASURBVEGYb4hAzDNKPC6+/cfn80cXN3NNgaD2spoQ975nE+WBfAtiYqA85HMT8hkiEPbnWxA+t5Mb19WzpCpA2Oem1O+2H7ZDI2l8bgc+tyEES6oDHOmNc7AnyrLqoF0ceKYWxJd+tZ9PPrT7jK4Fo0dVwONkLKt5/GDvyS8QhNcw0s11HhL0uvjkTatPaW217WKanGrbUOZnX1c0LwZh8cmbVpHMGG6mUr+b9iEjCD00kp8Wu6Q6SCIzRiIzxrLaIAGzvUj8DGshtuzuJpM98wD34Eiai5ZW8uyRAZ463M+b1k7dyVYQXuuIBSFMi1UsV1c2OdW20YxDTIxBgNEo0MpuCvvddivwoUQmb/3S6vHWIMuqg/hNgThZkFprzQPbTuQJSTarac2pEj8TBuJpqoJeNjWX8+yR/jO+jyC8FhCBEKZlbWMpt6yv59KWqknnrDhErkVQiFwX0/DIRIEI2q+X1QTtDrYnE4j93VE+9uBL/PjF8dnb3dEkqVFj/OrEMaoAn3xoN7/c1TnlPbXWDI6kqQh42Ly4ksO9cbuzrSDMR0QghGkJeF185fYNdrA6l3pTIEoLuJhyKfW7iaVGGR3LMpRI5wlKTchLwOPE43TYxXpw8iC11fn1YE69gjWPW+v8im0wWph/96lWvv/ciSnvGU2NkhnTVAQ8XLDEaEtysol6gvBaRgRCOGNuXt/Ah65qoali+kpvS0AiyVEGJ1gQSimW1gRprirB5XTYFkQiPUY2q7ln6yG71UcuPWZh3YHu8XGprX3jXWMHJ7iZnjVbme88MTSpi6yFNUSpvMTDuQ2llHic4mYS5jUiEMIZs6DUx4euWp43f6IQlkAMJzKmiynfJfU316/kkzcaQXNrBkY8PcpnH97LXb/cz33PHZ90z26zoO1gz7gF0doft1/3TxCIZ8wH/XAiY1saE7FSXCuCHtxOBxsXlfPsNBbE9589zo3/+sSUgiMIcx0RCKHoWALROZwgPZadFNS+YEklFy8zYhxelwOHgt/s7eHbTx5FqXF3Ui5Wa46+WNoOSh/ti+MyCwYnWhDPHOm3YyY7ciySnmiSLebwI+uaClPANjdXsL87ytBI4aD3Y/u6ebl9mNb+kYLnBWGuIwIhFB1LII6bD9JCabEWSikCHhcvHjcGE128tNJu5JeLZUEAdt+k1r44q+uNrrS5FkRfLMXBnhh/cMFCAh5nnkD8zzPHed/3nrfnYMB4AeEFSyrROj8Ocagnyrn/sIX9XVG7G+2OE4On+qcQhDmFCIRQdCyBsOZDTKzgnojf4ySrjb5Ry6qDdAwniCYzfOyHO+1f+T3RJOeYvacOdkfJZjXHBkbYsNDoW5VrQVgxjAsWV3BuY2meQPTGUmgNfdH0uAVh7m9dUykelyNPIF44NkQ0NcoPtp+gwxSuHccnx0gE4bWACIRQdCyBsNqGL6oMTLfcDlSvri9lQamfaHKUX+/p5oHtbXYsoTuSZG1jKSGviwPdMfZ1RUmPZlnTYBzLtSCODxiWS3NVgHWNZeztjNhpsFZgujeWZCCexuNy2HEQr8vJhqayvDjEUTPOcf82IxvK7863SJKZMf70u9vyhiT9dEe77RIThLmECIRQdKxWHLvMTrELKyYPNcrFSnVdVR+2Z3E/cdCY0dAXTzOW1fRGUywo9bG6IcwzR/r57X6jud6ly6soD3gYzIkbtA0m8LudVAY8LKsJkhnTtmhY8YveqBHLqCjx5AXdL1hSye6OYaJJI2221QxwWxP8blxXx57OiD3v4kB3lN/s6+H7zx0DIJLM8MH7d/CdJ1tP++8mCDONCIRQdHxuJ16Xg5H0GAvCPrtaeiqshn2rcwTicXOIT180RX8sRVYbjQRvXt/AwZ4Y33mylXMbSqkJ+agIePKqqU8MjNBYbszDsNqJW9Pu+uNGLKMvlrKL5HLZtKicrMZug360L263A1kQ9nHFiloyY+PzLqwMqd/u60VrQ8gAmS8hzElEIIRXBcvNtKhyeusBsBv2ra4vtZsEWg/avljKnl1dG/LyprV1+N1O+mIprlhRAzBJINoGEzSZVstSSyB6LYFI2/ftjabsNucWK+uMoPfezqjdyuNNa+txOhSr6sOsabDOGwJxxJyi1xVJsqczctoCsbtjmO89cyxvXrcgzBRFFQil1LVKqf1KqUNKqTsLnH+fUuplpdQOpdQTSqlV5nG3Uuo/zXN7lVKfKOY+heJjCUTzSeIPYAwnqgl5qQ557bnYFn2xlJ3BVBv2EfK5uf5cY5KeJRDlJZ68IHXb4IjdNyrsc1Mb9nKoJ2ZUdpvtx3ujKY4NjExyf1Wb+9jbGaE7miSZybKmsZS/euM5/NHFzdSX+vG7nRzuMYThSF/cTuP97b4eWyDaBhO2W2oqjvbFueFrT/C3P9nFPz2y76R/p+mIJjO86V8fz4uFnC6f+NHL3LP10CvahzC3KVo3V6WUE7gbuBpoA7YppR7SWu/JWfZ9rfU3zPU3AV8CrgXeBni11ucqpUqAPUqp+7TWrcXar1BcbAui6uQWxAevbGHQfHD7zNjB+C/9tB3wXWC6nz50lTHo6NyGUsAYdtQfT6O1JpIcJZIctQUCjJ5Ph3ti9meAkWE1NJIpKGAr68Ls6YjY7qPFlQEuyelNtbQmYFskR/tirG0sYzCe5olDfbY1BEa2lZVlVYiX2oxg95LqAB3mhD0wWqQXmt8xHfu7ouxqj7D92ADnNpae8nW5/HZfD/VlPv7v5cvO6Hph7lNMC2IzcEhrfURrnQbuB27OXaC1juS8DQBWSaoGAkopF+AH0kDuWmGOcToWREttiM2LK+z31rCiulKf4WIaTuJQ2CNTmypK+Is3LMNhFsmVl3hIjWZJZMZoGzSC0U3l48K0rDrI4d64HX8A2HnCDKAXcIGtrAtxqCfGQbOtx+Lq/O+wrNoQHK01R3vjLKkKcM6CEK19I/TmNPs7mNMWpBB7OiN4nA4ubammYzhhV2h/7Icv8YH7Xpz22olYQfgzbTaotaY/nuL4QOLki4XXLMUUiAYgtzNam3ksD6XUXyilDgN3AR8wD/8QiAOdwHHgi1rrST0PlFJ3KKW2K6W29/bKcJfZzOnEICayIGz8+r9oSSV90RRH+mIsqgzgchb+19cSjr5omhPmA64xVyBqgsRSo3ahW23YS8LMQiokYKvqwqTHsjz8cidel4O6CY0Ll1YHaR9K0No/Qjw9xpLqAAsrSuiKJDkxMEJt2IvP7WD/SeIQezoitNQGaaooIZnJ2hbOoZ6YLXSniiUQ/bEza30eSRiNC/tiqTOezSHMfYopEIUa9ExqWqO1vltrvRT4OPC35uHNwBhQDywGPqKUWlLg2nu11pu01puqq6vP3s6Fs07YFoiTWxATWVkXMormaoPE02Psao/kzZGYyPIFRgHdS+1D4xZETkNBK1Bt1VQsNwvuoHAK7iozUP3c0QGuWlVrWyoWVmbUb/Z2A7CkKmgL4YvHh1gQ9tFSE5oyUP3N3x3m13u62dsZYVVd2B7O1DFkWBHtQ4k8d9ipYFWtn6kF0ZdjXZ04TXEqJm2DI2Szxe19pbWesl/XfKOYAtEGNOW8bwQ6pll/P3CL+foPgF9qrTNa6x7gSWBTUXYpvCrcsLaO91221J5VfTp84MoWHv7A66kKGAHr4wMjeXMkJrK6PkyJx8lzRwdoG0wQ9LryWpKvaShFKXhsn2F1WhXZU6XgLq4KEPa5uGBxBf/ytnWTzlsC8bOdxr/ei6sDdtZU+1CC6pCXFQtC7OmITGrsF0lmuGvLfj76g530xdKsrAvbw5k6h5P0x9OkRrNEkhnGTuPBaFkQvWdoQeRaHsdmSa+pY/1xLvvCVn61p6uon/P4wT6u+JettsjOZ4opENuAFqXUYqWUB7gdeCh3gVKqJeftDcBB8/Vx4AplEAAuBF5ZWocwo5zfXMGd1604o2vdZhvwqtB4kNayAqZav3FRuV1At6YhnFf8Fva5Oac2RF8shVLjFsRU7i+X08GWD1/Kf//pBfYs7VwWVQZwOhQ724bZsLCM+lJfniVSHfKytrGU/nia9qFEnkg8cbCPsay2Byqtqh+3IDqHE3awWmvsqXynwriL6cwsiNzrTgzMjgfls0cGGMvqvBbvYFh2v93Xc9Y+x7DcjOr6+U7RBEJrPQq8H9gC7AUe0FrvVkp92sxYAni/Umq3UmoH8JfAH5rH7waCwC4MofmO1vqlYu1VmBtUBcdTXqezIMDou3SgO8ax/hHesXnhpPMbFxnZRGV+t51KO10Ava7Uj8dV+D8Xj8vBkqoAtWEv33zXRpRSVAY8dkFdddDL2sYyAF44PsSVX/odf/yd5xiIp9m6v4ewz8WGhcb5lQvCVAW9uJ2KjqFkXjbT4BRdZSeSSI/RE03hUIaL6UzakVuuKYcaF5uZ5gWzgePEeMy/PnaQv/rhS2fN9RQxq+ZjqemnGs4HipbmCqC1fgR4ZMKxv895/cEprothpLoKgk2+QEwfy9i8uBIwAtbXrlkw6fym5nL+59njVAQ89n0LZTCdKl9/10b8Hqc9eU8pRVNFCfu6ooaLqS6Ex+ngW78/wpHeOEd649zwtcdJZsa4dHk1H7yyhWeO9FNq1lDUhn10DieoyincGxzJ8NEf7OQt5zWwaVEF/7v9BG86t47yCemv1gN0xYIwezojxNNjea49q+BvyTQi2xdL29bVbHExWQLRPpSfWTU4kqYvlmJfV5RVZjffV0I0aQTlJTgvldTCHMKqcq4KeiYNHZrI2sZSykvcvOvCRXhdk91CmxYZabSVAS9LqgOc31zOZcvPPNFhWU3QnjdhYbmsqkNevC4nK+tCvNw+TMDj5ME/vwi308HgSIbLz6mhpTbEuy9qtq+tL/XTOZSkY2jczdHaF+eHz7exZVcXv9nbzd/9ZBfXffXxSVXX1i/+8xYZVslEN9OW3V1c8S+/sx+4heiPpygv8bCkOjArXEzDiQwHzfYobYOJSecAHj94djIZLYE4WWHjfEAEQpgzeF1Owj7XtL98LXxuJ7//2Bv44JUtBc83lvtZEPZRW+qjxOPiB++7mDUNZ1ZQNhVWHKI6ZFgolpvpqlW1bFxUwc8/cAl33bqWm9bVT7q2rsxHhxmD8LmN/0ytfk/tQwk7syiazPCfT7XmXWsJxIYmw402MZPpRbP77H8/fWzKvffH0lQGPCysCHBicGRSgPzu3x7i9Xc9xp9/7/mT/BXODjtODKE1rGsspWMokbefYTPD6wmzX9crxYr1jIhAiEAIc4srV9ZyzaraU1ob8rknpaRaKKX47p+cf8aB81PBipNY/aTWNxkCcYPZGiTsc/P285sKxjbqy/x0DSc5MThi94Oy6jbaBhOcGEhQ6nezoi486Rd1a1+coNfFOWa6b280P3ZhWRwPv9TJVx49wL8/fmTS5/fH0lQGx7vf5o5zHYyn+cKW/bQNJnj+WL4VMpIe5XMP75nknslm9SuKEew8MYRScP25dWTGND1Rw7IayxrV8k6H4tmjA3ZX3TPhf549RsdQgojlYkpLDEIEQphTfPm29fyf108qiTkjViwIT3ILnU3efF4D/3vHhdSbn/GmdXV85bb1XLXy5AK3tqGU0axmd0eEFQtCOB2KvV2mBTFoWBCN5X4ay/2TBOJIX5wl1QHbcsmtGNdas6cjwvnN5aTHsnzl0YN85dGDkwLZffEUlUGvXQOyu2O8kYEVA1haHaQ/ns578D99uJ9vPX6UZ4/225/3748fYeNnf83f/GTXlN83mszYfasKcaA7SmO53xa9dvM7W23YL15aSXo0ywvHzmy6X18sxd/8eBc/fL7Nvqe4mEQgBKFoeF1OLlhSmff+lg0NU1o1uVy7ZgE3rzdcTw1lfkr9bruxYNSsAm8s99NQ5p/kcjlitvuoyKkot+iKJBkcyXDTunq+evt63nJeA7HUqO3Ht+iLpqgy52e4ncq2XmBcINY1ljGW1XnZVdY5a6+P7u3hsw/vJZ4es3tNWezvivLu/3iWRHqMzz28l9vufXrKv8fh3jjLqoN2Rbwlita+r1xRg0ORN9xpKg71xLj7t4fyRNG6X38sZVsQ4mISgRCEWYlSis+/ZS3vuWgRb1y9wO4Qa9ETTdFUXkJjeQmjWW03MEykx2gfSrCkOojb6aCsxJ0Xg9jdbjzoV9WHuXl9A9esMjK8TuT0XEqPZokkR6kMevG4HLTUhOz4B4z/el/XZMRs+nKK6qxzVuX3d548Sl2pj7dsaMhL2QV46nAfjx/so7U/TttggiO98UlrwHAjHemNsawmaDddtDK1LIFoLC9hVX04bzzsVPzohTa+sGV/3r6tz+2Lp3MsCHExiUAIwizF73Hy6ZvX0FIbotzM2grkVHpbLiYwrIZ/e+ygPdhoiZkGXBX05o073dMZQSnDvWbdA8YfuO1DCb5lxiSs9N9V9eE8C8IKnLfUWDGOcQFqMx+0wyNp9nVFeOpwP+++aBFNFSUMjmQYSY//KreqtQdH0rYVsq118gO+fTBBajTL0uogPreTqqBnkqVSWuJmc3MlLxwfJDU6/YO9x9xvblylPdeCMEVH0lxFIARhTlBuWhDnLRpvF95UUWI/4L/71FG++KsDfPZho5u+FSDfuLCcX+/t5ofPtwFGgHpxZcCe+221BLGyor6+9RBf2LIfGM/CWlUXpi+WsgPDHcMJ6sv8dowj10LJtSB+vrMTp0PxjvMX2vtsz4mXWLGRwXjGftAXchEdNlupWy1NGsr8tJvpv5YFUeZ3s3lxBanRbMEZGP2xFG/7xlO09sVtgcjtt2QJTl8sbcce4mkRCBEIQZgDlPoNC2JVndFnCgy3ihUAf8xsNfFS2zBKGf2jAD5182ouXlrJxx98id5oij2dkbxislK/m5DPZfvgD3THWNdYyqN/eRmvW2bET6z11kzx9sEEDWV+qoOTBcJy1QwlMnQOJ6kNeSkPeOxkgLYcF5KVXZVrQeS6iFKjY9yz9ZCdKWWJXnnAw5C53hKIUlMgALa1Tg5UP3W4n22tgzx7tJ8e06JqzREI6/ufGBjBCufMFgviqUN9/HJXZ54l+GohAiEIcwDLgqgr9dkP28ZyPz63k5qQl6zGrpeoL/XbPaN8bicfvmo5Y1nN7w70cmIgManauLG8xH5AHuqJsao+zLKaoN2/6twGo+jwi1sOkBodo30oSUOZn7DfhcfpsGdepEbH7F/nQyNpemMp28poyLEg7n/uOP2xlG1B9ESSjKTHKPW7OdQTswv7fruvl7t+uZ97th6iMuCxK8bL/G5bGKx/hv1uKgIeykvctA9NLuzLFbfeAi4mS9hSo1n7WHwWxCAyY1ne8+3neN/3XuA9//Hcq/75IhCCMAewHo51ZUbcoSLgsd1E1sP3/13Rgsuh7PiDxZqGUlwOxX3PHQfG25dbNJX7OTEwQn8sxUA8PanPVcDr4p9vXcuezgj/9Mg++mIp6sv8KKWoCnrsB25nTtX30IiRtmoJRE3Ih8uh2LK7izt/9DIPvtBmxyCOmL/kL15qWCxWM77njxnWRFbn994qK/HYLqnhRAaf22ELYnXIS09kcrqsFZs5NjDCgGl9HO0bF5L2oQRu53h2mc/tOK001x0nhk4a+zgTBuNpRrMat1PZs9hfTUQgBGEOYGUx1Zf6+cOLm/nwVeMV4lbq503r6vnkTat574Q6EZ/byYq6kO2qWV2fXzFuWRBWK4uWnPkYFtesXsDbNjbyXbNq23JtVYW8djaQ5cevK/UxOJLOEwinQ7Gg1MfjB41q5+OmIMF4LMAKnFuxjuePDbK+qYxrVtXyxpx+WqV+t93+fHgkk9fKvSbky5viB0YthmVBvNQ2jNZQ4nFyrD+O1tpO87U+3/gO/pPGINKjWTJjWXqiSd58z5M8+Hz7tOvPBGvU7qLKAJFkpuizMCZS1GZ9giCcHS5tqebW8xpZviCI15X/gL9+zQJ8LgdNFSW868JFBa/f0FTOrvYI1SGv/dC2aKrwk8iM2f7/lilaqX/o6uX8dEcH6bGs7eaqDnrpHDYe6JZArK4P89ThfpKZMTtOAUZw2Y51dMXsSmVLIKwiuJ5IimTGGAz1x5c084nrVubto9TvRmujSG44kS8Q1SEvR4/mD/s5PjBCJDmK26nsz9q4qJzHD/bRE03Zbqq1jaW2pbEg7KN9MEEyM0YyM1aw99ef/uc2Gsv93Hb+QrQeD6afTSwra3FVgEM9MaKp0bzvW2zEghCEOUBTRQn/8vZ1BRsPXnduHV8oMMgoF6vNx+oC3U6ted0/29lBwOOkrtQ3aQ0YD/jbNxszwKyspKqg1w5Stw8m7BTakfQYWU2eGFmuMDCm/VmMmEKxsKIEv9tJdyTJy+3DpMeybFw4nrVlYVlTQyMZhhLpCRaEl94JLc6th/7FS6vsYxeYAe2jffHxug6zVxYYvbDSY1n+8ZG9vPmepwr+PfZ3RdnTGbWD3sVoamjFaSy34enMBDkbiEAIwjxgvTlvYmL8AeDiZZUsrgpwsCeWF5wuxMevXcG33rPJTo+tCnnsdhuHemLUl/rt+RqQLxCNptVx4ZIKkhkjGBzKaUNeHnBTG/bSHU2x3cxE2rhoGoFIZBhOjNoZXtbnpUezRBLj7qHnjg7gdiquWFFjH7Mq3A/2xGyrYl1TjkCYIvnc0QGO9sXpjaboGk7acYlsVtMfT9M1nLAD88WYm2FZEEvMrLSJFe/FRgRCEOYBS6oC/O0NK/mDCyYPTyrxuPja7RtwO5Xt5pmKgNfF1TnNEquDXsaymr54iicO9XHR0spJLh+LG9fV82eXLuGGtePda5fVjruzyks81IR9dEeS7O4YprHcT2Uw3x0G4ym/QyNpIgVcTDAex+iNpnhg+wluXFtPc9V48H5dYxlVQQ8vHhvkheOD1JX6aKkJ4jTboCwwGyxabqPdHcPccveTfOj+HcZnJ4wYSE80ZWdAtQ0mGB3L8sLxQbTW3Pv7w3z0BzvpHJ5cHT6RHz7fxmP7uu33h3qi7OmIMBBP43QoW5BfbYGQGIQgzAOUUtM2OTy3sZQf/fnrqC2d/ECeDmtA0k9f7GA4keH1LVV21TcYQWOLltoQn7h+Zd7chmXVQV48PmRnItWGfbzcNkQkkbFHwU7EsiCGE4VjEGAIQ0ttiHt/f5j0aJb/d2WL3a+qIuDB43KwaVEFz7UOMDqm2dRcjsOhqAh4GB7JUGF+h8yYcc2PX2ynK5KkK5JkT0cEl5nxpPV4I8NYapRv/O4wX/zVAVYsCLGvK4pSsGVXF7/+y8tYMIXr7rmjA/zVD3dyweIKrlhhiO9nfr6XnmiK9U1GirH1NxULQhCEGeHcxtK8B/qp8PqWKkr9bu7asg+l4PUt1Xl9o6oKWAC587qt6mjrAVgb8tIVSXKkL26fm0iZKQi90RSx1Gje51n7t9w+P3qhnevW1LG4KmAH1mtMEdnUXE7bYIKuSJLzm60BUh7CfhcBb36s5+cvdQJG+uvdWw/Rl9NeJLcJ4X3PnSDkdXGkL87N6+u5/70XEk2N8pzZQiSazPBPj+y1x5qOpEf5yA92oHV+T6veaIojvTF6o2kqA15bBEUgBEGYM4R8bu64dAmZMc25DaVmsZrxsA95Xfg9k4Pq9WV+HMroK1VnPrStLKGasJdkJkt6NMuyKQZDWQ9Ly/1TkTNy1Yp/9EZTpEez9MfTtiXi9xh9nKptgaiwr9vUbMQ6KoMeQj63XWMChmiMZTWLqwK8Y/NCtuzqyqtJGBzJUG9aB+1DCd64ZgEv/N3VfOW29WxYWJ7XDfe/nj7GN39/hCfMdN97fnuYEwMJ1jSE8yb/DY6kSY1m2dMxTEXAIwIhCMLc5I8ubqahzG8PQrLmak9Mp7VwOx3UlRrxBatCvDxnFrfF0iksCJfTQcjrmtSCAwxR8roc9ESTdgZQVWhcQN514SJuWd8AGBldPreDoNdl10C8bWMT77xgIQHPuEBcs9pw+1y8tJLV9cacjh0n8luX5/bIunBJJUGvC6UUHpeDZTUh9nZGSI2O8Z0nWwGjcvt4/wj3Pn6EW9bXc9XKWgZHMmTGjOC91XqkYzhJZdBDiceJy6EkBiEIwtwi4HXx+MfeYM+5CHldOB2KqikEAoyah9TomG1tWP/MdXFN5WICQ4SsiuulNePBZ6UUNWEvvdGUPQcjtxbjQ1ctt1+7nQ4uX16Dy6ns4PQtGwzxON5vZCR5nA6uXFHLfc+d4JJlVbbobW8dxOUwrkuNZu35GwPxNBcuGbdMAFbWhXjiYB8/fbHDTgluG0xw/7bjZLOaO69bya/3GgHqgXiasM9tZ3mBYcEopSjNaTHyaiECIQjCKyZ3CJJSijK/e0oLAuCLb1tHVmu791GZbUEY11SHvNMWhJWVuGkbTBD2ufIEAAxB6Imm6I0ZbqDphOqed55X8HiJGYNorPDzhhU1/Os7NnDN6gX2L/u9XRFqQl78biet/SNUh300VZQQ8DrtynaLVXVhfvRCO1/9zUFW1oUZy2ZpH0rgULCosoQFpT6qg+Zwp1iK0QnV0lYmlwiEIAivCT5x/UoWV5VMed6KGyTMIjnbgjBdTFPFHyzKzFTXQnUbNSEfh3pjBS2IiUw13S9oxiAWVpTgdChuXGek5lYGPIS8LqKpUaqCXoJeF639I9SEvNx57Yq8yX4WVu1J+1CCj75xOQ/t6KBjKEF6NMsS83tW2p1x00yY/mr/rcJ+d8FCOa31tLUrrwSJQQiCcNZ568ZGNi6qOOk6v8fJXbeu5bbzjQrtoNdFVdBbsOI7FyvOMbGxIBhB8LbBEbsnU6FMqpPhdTlwO1VexhUY1tEiU/iqgl67oK4m5OWipZVc0lI16V4rTYFYEPbxprX1NJjNEY/1j9gFcNYe+6Ip20qxjlWZ1sVUFsQtdz/Jx3/40ml/x1NBLAhBEGaUt5viYPGjP7+YiuDk3ke5WO6nhRYITgAACmlJREFUQnGKxVUlJDNZdrUPE5wik+pkKKX4ym0bWNMwWaiaKwPsao9QFRzva5UbXJ9IecDDzevrufycatxOB/Vlfnvu9fjkP+P79sdTuF3G7/bzFpbxqz3dVASMzwj73Xktyi1a+0dYm9Mm5GwiAiEIwqxiYeXUrimLsmkFwji2rXVw2jjIybhhbV3B49YwpqqQh0tbqtjdMWzXVkzFV2/fYL+26jFy9xr0uvC4HPTF0na/rYuWVvKrPd22lVLqdzGcyDA6lsXpUCiliJudaOvKTq9+5VQRgRAEYc5hxSwKuZiaTRdQXyw1bRzkTFlUaQpEwMvFy6q4eNlkt9J05AuEcS+lFNVBL33RlD0x8J0XLOL85gq7zUapGYO4+su/59bzGnj/FS12G4/ce55NRCAEQZhz3LiuHqdDsaiAtVFf6sfjcpAezb4iC2IqrIf6md7b6mob8rls1xIYRXp98TQhn4tSvxuPy8GahvHW7qV+N1ltdKC12nt0mEOa6kqLIxASpBYEYc6xoNTHn1yyuGD2jsOhWFQxHkg+22xoKuPv37Qqr2nh6WBN11tSFcjbf5VpQQyMZOzCwVxy036t+dRWo8D6IrmYRCAEQXjN0TwhO+hs4nAo/uSSxXntOE4Hp0OxrCbI6ob8wU9VQQ/98RRDI2l7xGwulkA4FHSbY1U7hpMoNX2Q/JUgLiZBEF5zLHmFbqBi8z//54JJ2VWVQS/9sTT9gXTBzq/nLSrnujULCHpd/GRHO1prOoYS1IZ8uJ3F+a1fVAtCKXWtUmq/UuqQUurOAuffp5R6WSm1Qyn1hFJqVc65tUqpp5VSu801xZFIQRBecxTTgjgbVAa9lHjyf5/Xl/oYzWoO98byWqZb1IR8fP1dG1lVHyYzphmIp+kcThQtgwmKKBBKKSdwN3AdsAp4R64AmHxfa32u1no9cBfwJfNaF/A94H1a69XA5cCrW2MuCMKcZdOicioCHlacZADSbOKmdQ2EfC5So9mCMQgLy53UHUnRMZSkvkgZTFBcC2IzcEhrfURrnQbuB27OXaC1juS8DQBWkfk1wEta653mun6t9VgR9yoIwmuIltoQL/zd1XaK6FygtMTN+y5bClAwBmFh9avqjiTpGErYrcaLQTFjEA3AiZz3bcAFExcppf4C+EvAA1xhHl4OaKXUFqAauF9rfVeBa+8A7gBYuHDyKEVBEIS5xB+/rpmdJ4Z43TS1FZYFsa8rSmo0O2ctiELdoyZ1stJa3621Xgp8HPhb87ALuAR4p/nPNyulrixw7b1a601a603V1dVnb+eCIAgzQInHxb3v2cT6pqlbZ1iB92eP9gPMWYFoA3KbrDQCHdOsvx+4Jefa32mt+7TWI8AjQOG+vIIgCPMIr8tJRcDD7w/04nQoLlh88qaIZ0oxBWIb0KKUWqyU8gC3Aw/lLlBKteS8vQE4aL7eAqxVSpWYAevLgD1F3KsgCMKcoSbkJavhgsUV9rjWYlC0GITWelQp9X6Mh70T+LbWerdS6tPAdq31Q8D7lVJXYWQoDQJ/aF47qJT6EobIaOARrfXDxdqrIAjCXKI27GNfV5Q3rl5Q1M8paqGc1voRDPdQ7rH/3969hVpRxXEc//7SktJIw7Qwb6kPKtTRQiQpjKDUFw2M7GISgT0oKPSQRjd66sWCwMwiUdEyKyWJ6CZh+GDesLwliVmdFC0KyyJL/fcwy9zbZssx3XvO2fP7wGHPXmf2nP/8WcP/zJrZa56qWJ51ls8uI7vV1czMKlydLlT/3+k+2srfpDYz62CmjOrL4F7d6nqBGlwgzMw6nBH9ejCiX4+6/x1P1mdmZrlcIMzMLJcLhJmZ5XKBMDOzXC4QZmaWywXCzMxyuUCYmVkuFwgzM8uliP/MwN0hSfoR+PY8NtET+OkChdPRORfVnI9qzke1jp6P/hGR+7yEpikQ50vS5oi4qeg42gPnoprzUc35qNbM+fAQk5mZ5XKBMDOzXC4Qp71SdADtiHNRzfmo5nxUa9p8+BqEmZnl8hmEmZnlcoEwM7NcpS8QksZJ2iNpr6Q5RcdTBEn7JW2XtE3S5tR2paSPJX2dXuv/dJKCSFok6bCkHRVtufuvzIupv3wpaWRxkddHjXw8I+mH1Ee2SZpQ8bu5KR97JN1ZTNT1I6mvpE8l7Za0U9Ks1N70faTUBUJSJ2A+MB4YBtwraVixURXmtohoqbifew6wNiKGAGvT+2a1GBh3Rlut/R8PDEk/04EFDYqxkRbz33wAvJD6SEt63jzpeJkCDE+feSkdV83kOPBoRAwFRgMz0n43fR8pdYEARgF7I2JfRPwFrAAmFhxTezERWJKWlwCTCoylriLiM+DnM5pr7f9EYGlkNgDdJV3TmEgbo0Y+apkIrIiIYxHxDbCX7LhqGhFxMCK2puXfgN1AH0rQR8peIPoA31e8b01tZRPAR5K2SJqe2npHxEHIDhCgV2HRFaPW/pe5z8xMQyaLKoYcS5UPSQOAEcDnlKCPlL1AKKetjPf9jomIkWSnxjMk3Vp0QO1YWfvMAmAQ0AIcBOal9tLkQ1I34B1gdkT8erZVc9o6ZE7KXiBagb4V768FDhQUS2Ei4kB6PQysJhsiOHTqtDi9Hi4uwkLU2v9S9pmIOBQRJyLiJPAqp4eRSpEPSReTFYflEbEqNTd9Hyl7gdgEDJE0UNIlZBfb1hQcU0NJ6irp8lPLwB3ADrI8TEurTQPeLSbCwtTa/zXAg+lOldHAkVPDDM3sjDH0u8j6CGT5mCKpi6SBZBdmNzY6vnqSJOA1YHdEPF/xq6bvI52LDqBIEXFc0kzgQ6ATsCgidhYcVqP1BlZnxwCdgdcj4gNJm4CVkh4GvgPuLjDGupL0BjAW6CmpFXgaeI78/X8fmEB2MfYP4KGGB1xnNfIxVlIL2VDJfuARgIjYKWklsIvsbp8ZEXGiiLjraAwwFdguaVtqe5wS9BFPtWFmZrnKPsRkZmY1uECYmVkuFwgzM8vlAmFmZrlcIMzMLJcLhFk7IGmspPeKjsOskguEmZnlcoEwOweSHpC0MT0TYaGkTpKOSponaauktZKuSuu2SNqQJrhbXfG8gMGSPpH0RfrMoLT5bpLelvSVpOXpG7xmhXGBMGsjSUOBe8gmN2wBTgD3A12BrWnCw3Vk3zwGWAo8FhHXA9sr2pcD8yPiBuBmssnvIJsldDbZs0muI/sGr1lhSj3Vhtk5uh24EdiU/rm/lGyCtpPAm2mdZcAqSVcA3SNiXWpfAryV5r3qExGrASLiT4C0vY0R0ZrebwMGAOvrv1tm+VwgzNpOwJKImFvVKD15xnpnm7/mbMNGxyqWT+Dj0wrmISaztlsLTJbUC/59JnF/suNoclrnPmB9RBwBfpF0S2qfCqxLzxFolTQpbaOLpMsauhdmbeT/UMzaKCJ2SXqC7Ol7FwF/AzOA34HhkrYAR8iuU0A2BfTLqQDs4/SsnlOBhZKeTdto2plyrWPzbK5m50nS0YjoVnQcZheah5jMzCyXzyDMzCyXzyDMzCyXC4SZmeVygTAzs1wuEGZmlssFwszMcv0D7f9M7OV4YKsAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/225\n",
      "1754/1754 [==============================] - 3s 2ms/step - loss: 0.4590 - binary_accuracy: 0.8170\n",
      "Epoch 2/225\n",
      "1754/1754 [==============================] - 2s 934us/step - loss: 0.4370 - binary_accuracy: 0.8219\n",
      "Epoch 3/225\n",
      "1754/1754 [==============================] - 2s 946us/step - loss: 0.4295 - binary_accuracy: 0.8196\n",
      "Epoch 4/225\n",
      "1754/1754 [==============================] - 2s 937us/step - loss: 0.4309 - binary_accuracy: 0.8227\n",
      "Epoch 5/225\n",
      "1754/1754 [==============================] - 2s 918us/step - loss: 0.4187 - binary_accuracy: 0.8273\n",
      "Epoch 6/225\n",
      "1754/1754 [==============================] - 2s 910us/step - loss: 0.4191 - binary_accuracy: 0.8298\n",
      "Epoch 7/225\n",
      "1754/1754 [==============================] - 2s 915us/step - loss: 0.4273 - binary_accuracy: 0.8215\n",
      "Epoch 8/225\n",
      "1754/1754 [==============================] - 2s 934us/step - loss: 0.4262 - binary_accuracy: 0.8225\n",
      "Epoch 9/225\n",
      "1754/1754 [==============================] - 2s 936us/step - loss: 0.4111 - binary_accuracy: 0.8279\n",
      "Epoch 10/225\n",
      "1754/1754 [==============================] - 2s 921us/step - loss: 0.4227 - binary_accuracy: 0.8247\n",
      "Epoch 11/225\n",
      "1754/1754 [==============================] - 2s 917us/step - loss: 0.4142 - binary_accuracy: 0.8254\n",
      "Epoch 12/225\n",
      "1754/1754 [==============================] - 2s 921us/step - loss: 0.4206 - binary_accuracy: 0.8273\n",
      "Epoch 13/225\n",
      "1754/1754 [==============================] - 2s 919us/step - loss: 0.4202 - binary_accuracy: 0.8241\n",
      "Epoch 14/225\n",
      "1754/1754 [==============================] - 2s 928us/step - loss: 0.4170 - binary_accuracy: 0.8268\n",
      "Epoch 15/225\n",
      "1754/1754 [==============================] - 2s 917us/step - loss: 0.4132 - binary_accuracy: 0.8280\n",
      "Epoch 16/225\n",
      "1754/1754 [==============================] - 2s 928us/step - loss: 0.4144 - binary_accuracy: 0.8250\n",
      "Epoch 17/225\n",
      "1754/1754 [==============================] - 2s 917us/step - loss: 0.4151 - binary_accuracy: 0.8264\n",
      "Epoch 18/225\n",
      "1754/1754 [==============================] - 2s 916us/step - loss: 0.4116 - binary_accuracy: 0.8326\n",
      "Epoch 19/225\n",
      "1754/1754 [==============================] - 2s 912us/step - loss: 0.4123 - binary_accuracy: 0.8274\n",
      "Epoch 20/225\n",
      "1754/1754 [==============================] - 2s 906us/step - loss: 0.4113 - binary_accuracy: 0.8266\n",
      "Epoch 21/225\n",
      "1754/1754 [==============================] - 2s 906us/step - loss: 0.4085 - binary_accuracy: 0.8298\n",
      "Epoch 22/225\n",
      "1754/1754 [==============================] - 2s 909us/step - loss: 0.4059 - binary_accuracy: 0.8320\n",
      "Epoch 23/225\n",
      "1754/1754 [==============================] - 2s 913us/step - loss: 0.4102 - binary_accuracy: 0.8265\n",
      "Epoch 24/225\n",
      "1754/1754 [==============================] - 2s 925us/step - loss: 0.4081 - binary_accuracy: 0.8284\n",
      "Epoch 25/225\n",
      "1754/1754 [==============================] - 2s 910us/step - loss: 0.4137 - binary_accuracy: 0.8290\n",
      "Epoch 26/225\n",
      "1754/1754 [==============================] - 2s 909us/step - loss: 0.4075 - binary_accuracy: 0.8277\n",
      "Epoch 27/225\n",
      "1754/1754 [==============================] - 2s 914us/step - loss: 0.4122 - binary_accuracy: 0.8308\n",
      "Epoch 28/225\n",
      "1754/1754 [==============================] - 2s 941us/step - loss: 0.4070 - binary_accuracy: 0.8299\n",
      "Epoch 29/225\n",
      "1754/1754 [==============================] - 2s 935us/step - loss: 0.4068 - binary_accuracy: 0.8263\n",
      "Epoch 30/225\n",
      "1754/1754 [==============================] - 2s 922us/step - loss: 0.4050 - binary_accuracy: 0.8307\n",
      "Epoch 31/225\n",
      "1754/1754 [==============================] - 2s 918us/step - loss: 0.4094 - binary_accuracy: 0.8273\n",
      "Epoch 32/225\n",
      "1754/1754 [==============================] - 2s 906us/step - loss: 0.4069 - binary_accuracy: 0.8270\n",
      "Epoch 33/225\n",
      "1754/1754 [==============================] - 2s 927us/step - loss: 0.4028 - binary_accuracy: 0.8343\n",
      "Epoch 34/225\n",
      "1754/1754 [==============================] - 2s 924us/step - loss: 0.4073 - binary_accuracy: 0.8241\n",
      "Epoch 35/225\n",
      "1754/1754 [==============================] - 2s 916us/step - loss: 0.4033 - binary_accuracy: 0.8329\n",
      "Epoch 36/225\n",
      "1754/1754 [==============================] - 2s 917us/step - loss: 0.4067 - binary_accuracy: 0.8255\n",
      "Epoch 37/225\n",
      "1754/1754 [==============================] - 2s 928us/step - loss: 0.4054 - binary_accuracy: 0.8325\n",
      "Epoch 38/225\n",
      "1754/1754 [==============================] - 2s 912us/step - loss: 0.4036 - binary_accuracy: 0.8307\n",
      "Epoch 39/225\n",
      "1754/1754 [==============================] - 2s 911us/step - loss: 0.4025 - binary_accuracy: 0.8304\n",
      "Epoch 40/225\n",
      "1754/1754 [==============================] - 2s 916us/step - loss: 0.4027 - binary_accuracy: 0.8333\n",
      "Epoch 41/225\n",
      "1754/1754 [==============================] - 2s 908us/step - loss: 0.4001 - binary_accuracy: 0.8265\n",
      "Epoch 42/225\n",
      "1754/1754 [==============================] - 2s 922us/step - loss: 0.3964 - binary_accuracy: 0.8352\n",
      "Epoch 43/225\n",
      "1754/1754 [==============================] - 2s 914us/step - loss: 0.3970 - binary_accuracy: 0.8320\n",
      "Epoch 44/225\n",
      "1754/1754 [==============================] - 2s 925us/step - loss: 0.3940 - binary_accuracy: 0.8374\n",
      "Epoch 45/225\n",
      "1754/1754 [==============================] - 2s 919us/step - loss: 0.4014 - binary_accuracy: 0.8302\n",
      "Epoch 46/225\n",
      "1754/1754 [==============================] - 2s 932us/step - loss: 0.4024 - binary_accuracy: 0.8311\n",
      "Epoch 47/225\n",
      "1754/1754 [==============================] - 2s 948us/step - loss: 0.4041 - binary_accuracy: 0.8299\n",
      "Epoch 48/225\n",
      "1754/1754 [==============================] - 2s 911us/step - loss: 0.4004 - binary_accuracy: 0.8284\n",
      "Epoch 49/225\n",
      "1754/1754 [==============================] - 2s 919us/step - loss: 0.3982 - binary_accuracy: 0.8322\n",
      "Epoch 50/225\n",
      "1754/1754 [==============================] - 2s 928us/step - loss: 0.4022 - binary_accuracy: 0.8336\n",
      "Epoch 51/225\n",
      "1754/1754 [==============================] - 2s 936us/step - loss: 0.3989 - binary_accuracy: 0.8334\n",
      "Epoch 52/225\n",
      "1754/1754 [==============================] - 2s 930us/step - loss: 0.4055 - binary_accuracy: 0.8314\n",
      "Epoch 53/225\n",
      "1754/1754 [==============================] - 2s 916us/step - loss: 0.3984 - binary_accuracy: 0.8306\n",
      "Epoch 54/225\n",
      "1754/1754 [==============================] - 2s 922us/step - loss: 0.3954 - binary_accuracy: 0.8332\n",
      "Epoch 55/225\n",
      "1754/1754 [==============================] - 2s 919us/step - loss: 0.4009 - binary_accuracy: 0.8351\n",
      "Epoch 56/225\n",
      "1754/1754 [==============================] - 2s 914us/step - loss: 0.3960 - binary_accuracy: 0.8343\n",
      "Epoch 57/225\n",
      "1754/1754 [==============================] - 2s 920us/step - loss: 0.3952 - binary_accuracy: 0.8362\n",
      "Epoch 58/225\n",
      "1754/1754 [==============================] - 2s 924us/step - loss: 0.4000 - binary_accuracy: 0.8318\n",
      "Epoch 59/225\n",
      "1754/1754 [==============================] - 2s 913us/step - loss: 0.3977 - binary_accuracy: 0.8329\n",
      "Epoch 60/225\n",
      "1754/1754 [==============================] - 2s 912us/step - loss: 0.4015 - binary_accuracy: 0.8313\n",
      "Epoch 61/225\n",
      "1754/1754 [==============================] - 2s 919us/step - loss: 0.4018 - binary_accuracy: 0.8364\n",
      "Epoch 62/225\n",
      "1754/1754 [==============================] - 2s 909us/step - loss: 0.4004 - binary_accuracy: 0.8293\n",
      "Epoch 63/225\n",
      "1754/1754 [==============================] - 2s 927us/step - loss: 0.3987 - binary_accuracy: 0.8286\n",
      "Epoch 64/225\n",
      "1754/1754 [==============================] - 2s 932us/step - loss: 0.3935 - binary_accuracy: 0.8341\n",
      "Epoch 65/225\n",
      "1754/1754 [==============================] - 2s 923us/step - loss: 0.3967 - binary_accuracy: 0.8330\n",
      "Epoch 66/225\n",
      "1754/1754 [==============================] - 2s 915us/step - loss: 0.3976 - binary_accuracy: 0.8310\n",
      "Epoch 67/225\n",
      "1754/1754 [==============================] - 2s 988us/step - loss: 0.3941 - binary_accuracy: 0.8422\n",
      "Epoch 68/225\n",
      "1754/1754 [==============================] - 2s 932us/step - loss: 0.3954 - binary_accuracy: 0.8336\n",
      "Epoch 69/225\n",
      "1754/1754 [==============================] - 2s 929us/step - loss: 0.3890 - binary_accuracy: 0.8413\n",
      "Epoch 70/225\n",
      "1754/1754 [==============================] - 2s 912us/step - loss: 0.3909 - binary_accuracy: 0.8315\n",
      "Epoch 71/225\n",
      "1754/1754 [==============================] - 2s 908us/step - loss: 0.3942 - binary_accuracy: 0.8389\n",
      "Epoch 72/225\n",
      "1754/1754 [==============================] - 2s 923us/step - loss: 0.4009 - binary_accuracy: 0.8296\n",
      "Epoch 73/225\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1754/1754 [==============================] - 2s 936us/step - loss: 0.3984 - binary_accuracy: 0.8343\n",
      "Epoch 74/225\n",
      "1754/1754 [==============================] - 2s 922us/step - loss: 0.3898 - binary_accuracy: 0.8370\n",
      "Epoch 75/225\n",
      "1754/1754 [==============================] - 2s 919us/step - loss: 0.3897 - binary_accuracy: 0.8380\n",
      "Epoch 76/225\n",
      "1754/1754 [==============================] - 2s 910us/step - loss: 0.3889 - binary_accuracy: 0.8343\n",
      "Epoch 77/225\n",
      "1754/1754 [==============================] - 2s 927us/step - loss: 0.3961 - binary_accuracy: 0.8336\n",
      "Epoch 78/225\n",
      "1754/1754 [==============================] - 2s 923us/step - loss: 0.3867 - binary_accuracy: 0.8384\n",
      "Epoch 79/225\n",
      "1754/1754 [==============================] - 2s 916us/step - loss: 0.3922 - binary_accuracy: 0.8368\n",
      "Epoch 80/225\n",
      "1754/1754 [==============================] - 2s 915us/step - loss: 0.3956 - binary_accuracy: 0.8330\n",
      "Epoch 81/225\n",
      "1754/1754 [==============================] - 2s 913us/step - loss: 0.3904 - binary_accuracy: 0.8379\n",
      "Epoch 82/225\n",
      "1754/1754 [==============================] - 2s 921us/step - loss: 0.3920 - binary_accuracy: 0.8310\n",
      "Epoch 83/225\n",
      "1754/1754 [==============================] - 2s 916us/step - loss: 0.3907 - binary_accuracy: 0.8369\n",
      "Epoch 84/225\n",
      "1754/1754 [==============================] - 2s 918us/step - loss: 0.3864 - binary_accuracy: 0.8372\n",
      "Epoch 85/225\n",
      "1754/1754 [==============================] - 2s 937us/step - loss: 0.3866 - binary_accuracy: 0.8364\n",
      "Epoch 86/225\n",
      "1754/1754 [==============================] - 2s 921us/step - loss: 0.3913 - binary_accuracy: 0.8385\n",
      "Epoch 87/225\n",
      "1754/1754 [==============================] - 2s 930us/step - loss: 0.3924 - binary_accuracy: 0.8315\n",
      "Epoch 88/225\n",
      "1754/1754 [==============================] - 2s 928us/step - loss: 0.3955 - binary_accuracy: 0.8326\n",
      "Epoch 89/225\n",
      "1754/1754 [==============================] - 2s 935us/step - loss: 0.3939 - binary_accuracy: 0.8337\n",
      "Epoch 90/225\n",
      "1754/1754 [==============================] - 2s 920us/step - loss: 0.3948 - binary_accuracy: 0.8362\n",
      "Epoch 91/225\n",
      "1754/1754 [==============================] - 2s 925us/step - loss: 0.3926 - binary_accuracy: 0.8359\n",
      "Epoch 92/225\n",
      "1754/1754 [==============================] - 2s 927us/step - loss: 0.3958 - binary_accuracy: 0.8347\n",
      "Epoch 93/225\n",
      "1754/1754 [==============================] - 2s 947us/step - loss: 0.3914 - binary_accuracy: 0.8332\n",
      "Epoch 94/225\n",
      "1754/1754 [==============================] - 2s 940us/step - loss: 0.3967 - binary_accuracy: 0.8307\n",
      "Epoch 95/225\n",
      "1754/1754 [==============================] - 2s 927us/step - loss: 0.3932 - binary_accuracy: 0.8383\n",
      "Epoch 96/225\n",
      "1754/1754 [==============================] - 2s 925us/step - loss: 0.3872 - binary_accuracy: 0.8350\n",
      "Epoch 97/225\n",
      "1754/1754 [==============================] - 2s 930us/step - loss: 0.3971 - binary_accuracy: 0.8352\n",
      "Epoch 98/225\n",
      "1754/1754 [==============================] - 2s 933us/step - loss: 0.3876 - binary_accuracy: 0.8361\n",
      "Epoch 99/225\n",
      "1754/1754 [==============================] - 2s 919us/step - loss: 0.3860 - binary_accuracy: 0.8364\n",
      "Epoch 100/225\n",
      "1754/1754 [==============================] - 2s 916us/step - loss: 0.3917 - binary_accuracy: 0.8382\n",
      "Epoch 101/225\n",
      "1754/1754 [==============================] - 2s 919us/step - loss: 0.3824 - binary_accuracy: 0.8446\n",
      "Epoch 102/225\n",
      "1754/1754 [==============================] - 2s 921us/step - loss: 0.3878 - binary_accuracy: 0.8401\n",
      "Epoch 103/225\n",
      "1754/1754 [==============================] - 2s 916us/step - loss: 0.3898 - binary_accuracy: 0.8404\n",
      "Epoch 104/225\n",
      "1754/1754 [==============================] - 2s 911us/step - loss: 0.3916 - binary_accuracy: 0.8349\n",
      "Epoch 105/225\n",
      "1754/1754 [==============================] - 2s 913us/step - loss: 0.3848 - binary_accuracy: 0.8372\n",
      "Epoch 106/225\n",
      "1754/1754 [==============================] - 2s 911us/step - loss: 0.3845 - binary_accuracy: 0.8375\n",
      "Epoch 107/225\n",
      "1754/1754 [==============================] - 2s 912us/step - loss: 0.3857 - binary_accuracy: 0.8390\n",
      "Epoch 108/225\n",
      "1754/1754 [==============================] - 2s 915us/step - loss: 0.3884 - binary_accuracy: 0.8382\n",
      "Epoch 109/225\n",
      "1754/1754 [==============================] - 2s 918us/step - loss: 0.3921 - binary_accuracy: 0.8364\n",
      "Epoch 110/225\n",
      "1754/1754 [==============================] - 2s 928us/step - loss: 0.3831 - binary_accuracy: 0.8399\n",
      "Epoch 111/225\n",
      "1754/1754 [==============================] - 2s 919us/step - loss: 0.3951 - binary_accuracy: 0.8362\n",
      "Epoch 112/225\n",
      "1754/1754 [==============================] - 2s 922us/step - loss: 0.3910 - binary_accuracy: 0.8361\n",
      "Epoch 113/225\n",
      "1754/1754 [==============================] - 2s 929us/step - loss: 0.3916 - binary_accuracy: 0.8348\n",
      "Epoch 114/225\n",
      "1754/1754 [==============================] - 2s 912us/step - loss: 0.3883 - binary_accuracy: 0.8379\n",
      "Epoch 115/225\n",
      "1754/1754 [==============================] - 2s 918us/step - loss: 0.3904 - binary_accuracy: 0.8390\n",
      "Epoch 116/225\n",
      "1754/1754 [==============================] - 2s 916us/step - loss: 0.3940 - binary_accuracy: 0.8336\n",
      "Epoch 117/225\n",
      "1754/1754 [==============================] - 2s 915us/step - loss: 0.3879 - binary_accuracy: 0.8328\n",
      "Epoch 118/225\n",
      "1754/1754 [==============================] - 2s 925us/step - loss: 0.3850 - binary_accuracy: 0.8376\n",
      "Epoch 119/225\n",
      "1754/1754 [==============================] - 2s 954us/step - loss: 0.3844 - binary_accuracy: 0.8371\n",
      "Epoch 120/225\n",
      "1754/1754 [==============================] - 2s 910us/step - loss: 0.3849 - binary_accuracy: 0.8382\n",
      "Epoch 121/225\n",
      "1754/1754 [==============================] - 2s 910us/step - loss: 0.3877 - binary_accuracy: 0.8412\n",
      "Epoch 122/225\n",
      "1754/1754 [==============================] - 2s 928us/step - loss: 0.3836 - binary_accuracy: 0.8396\n",
      "Epoch 123/225\n",
      "1754/1754 [==============================] - 2s 928us/step - loss: 0.3863 - binary_accuracy: 0.8396\n",
      "Epoch 124/225\n",
      "1754/1754 [==============================] - 2s 955us/step - loss: 0.3925 - binary_accuracy: 0.8335\n",
      "Epoch 125/225\n",
      "1754/1754 [==============================] - 2s 969us/step - loss: 0.3867 - binary_accuracy: 0.8335\n",
      "Epoch 126/225\n",
      "1754/1754 [==============================] - 2s 928us/step - loss: 0.3858 - binary_accuracy: 0.8423\n",
      "Epoch 127/225\n",
      "1754/1754 [==============================] - 2s 926us/step - loss: 0.3841 - binary_accuracy: 0.8418\n",
      "Epoch 128/225\n",
      "1754/1754 [==============================] - 2s 929us/step - loss: 0.3836 - binary_accuracy: 0.8377\n",
      "Epoch 129/225\n",
      "1754/1754 [==============================] - 2s 920us/step - loss: 0.3893 - binary_accuracy: 0.8357\n",
      "Epoch 130/225\n",
      "1754/1754 [==============================] - 2s 937us/step - loss: 0.3820 - binary_accuracy: 0.8409\n",
      "Epoch 131/225\n",
      "1754/1754 [==============================] - 2s 966us/step - loss: 0.3892 - binary_accuracy: 0.8330\n",
      "Epoch 132/225\n",
      "1754/1754 [==============================] - 2s 958us/step - loss: 0.3883 - binary_accuracy: 0.8356\n",
      "Epoch 133/225\n",
      "1754/1754 [==============================] - 2s 955us/step - loss: 0.3870 - binary_accuracy: 0.8365\n",
      "Epoch 134/225\n",
      "1754/1754 [==============================] - 2s 956us/step - loss: 0.3822 - binary_accuracy: 0.8398\n",
      "Epoch 135/225\n",
      "1754/1754 [==============================] - 2s 944us/step - loss: 0.3870 - binary_accuracy: 0.8363\n",
      "Epoch 136/225\n",
      "1754/1754 [==============================] - 2s 948us/step - loss: 0.3808 - binary_accuracy: 0.8436\n",
      "Epoch 137/225\n",
      "1754/1754 [==============================] - 2s 935us/step - loss: 0.3838 - binary_accuracy: 0.8352\n",
      "Epoch 138/225\n",
      "1754/1754 [==============================] - 2s 946us/step - loss: 0.3907 - binary_accuracy: 0.8317\n",
      "Epoch 139/225\n",
      "1754/1754 [==============================] - 2s 926us/step - loss: 0.3784 - binary_accuracy: 0.8432\n",
      "Epoch 140/225\n",
      "1754/1754 [==============================] - 2s 921us/step - loss: 0.3861 - binary_accuracy: 0.8367\n",
      "Epoch 141/225\n",
      "1754/1754 [==============================] - 2s 915us/step - loss: 0.3821 - binary_accuracy: 0.8418\n",
      "Epoch 142/225\n",
      "1754/1754 [==============================] - 2s 931us/step - loss: 0.3832 - binary_accuracy: 0.8401\n",
      "Epoch 143/225\n",
      "1754/1754 [==============================] - 2s 935us/step - loss: 0.3810 - binary_accuracy: 0.8437\n",
      "Epoch 144/225\n",
      "1754/1754 [==============================] - 2s 931us/step - loss: 0.3829 - binary_accuracy: 0.8433\n",
      "Epoch 145/225\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1754/1754 [==============================] - 2s 953us/step - loss: 0.3857 - binary_accuracy: 0.8378\n",
      "Epoch 146/225\n",
      "1754/1754 [==============================] - 2s 940us/step - loss: 0.3820 - binary_accuracy: 0.8405\n",
      "Epoch 147/225\n",
      "1754/1754 [==============================] - 2s 918us/step - loss: 0.3841 - binary_accuracy: 0.8364\n",
      "Epoch 148/225\n",
      "1754/1754 [==============================] - 2s 947us/step - loss: 0.3860 - binary_accuracy: 0.8376\n",
      "Epoch 149/225\n",
      "1754/1754 [==============================] - 2s 921us/step - loss: 0.3801 - binary_accuracy: 0.8447\n",
      "Epoch 150/225\n",
      "1754/1754 [==============================] - 2s 911us/step - loss: 0.3812 - binary_accuracy: 0.8433\n",
      "Epoch 151/225\n",
      "1754/1754 [==============================] - 2s 910us/step - loss: 0.3794 - binary_accuracy: 0.8435\n",
      "Epoch 152/225\n",
      "1754/1754 [==============================] - 2s 918us/step - loss: 0.3793 - binary_accuracy: 0.8378\n",
      "Epoch 153/225\n",
      "1754/1754 [==============================] - 2s 913us/step - loss: 0.3799 - binary_accuracy: 0.8439\n",
      "Epoch 154/225\n",
      "1754/1754 [==============================] - 2s 910us/step - loss: 0.3847 - binary_accuracy: 0.8361\n",
      "Epoch 155/225\n",
      "1754/1754 [==============================] - 2s 914us/step - loss: 0.3812 - binary_accuracy: 0.8388\n",
      "Epoch 156/225\n",
      "1754/1754 [==============================] - 2s 908us/step - loss: 0.3763 - binary_accuracy: 0.8450\n",
      "Epoch 157/225\n",
      "1754/1754 [==============================] - 2s 919us/step - loss: 0.3858 - binary_accuracy: 0.8373\n",
      "Epoch 158/225\n",
      "1754/1754 [==============================] - 2s 917us/step - loss: 0.3807 - binary_accuracy: 0.8414\n",
      "Epoch 159/225\n",
      "1754/1754 [==============================] - 2s 918us/step - loss: 0.3760 - binary_accuracy: 0.8426\n",
      "Epoch 160/225\n",
      "1754/1754 [==============================] - 2s 916us/step - loss: 0.3838 - binary_accuracy: 0.8395\n",
      "Epoch 161/225\n",
      "1754/1754 [==============================] - 2s 923us/step - loss: 0.3782 - binary_accuracy: 0.8415\n",
      "Epoch 162/225\n",
      "1754/1754 [==============================] - 2s 921us/step - loss: 0.3860 - binary_accuracy: 0.8392\n",
      "Epoch 163/225\n",
      "1754/1754 [==============================] - 2s 933us/step - loss: 0.3784 - binary_accuracy: 0.8413\n",
      "Epoch 164/225\n",
      "1754/1754 [==============================] - 2s 920us/step - loss: 0.3783 - binary_accuracy: 0.8428\n",
      "Epoch 165/225\n",
      "1754/1754 [==============================] - 2s 947us/step - loss: 0.3783 - binary_accuracy: 0.8450\n",
      "Epoch 166/225\n",
      "1754/1754 [==============================] - 2s 919us/step - loss: 0.3805 - binary_accuracy: 0.8410\n",
      "Epoch 167/225\n",
      "1754/1754 [==============================] - 2s 926us/step - loss: 0.3770 - binary_accuracy: 0.8441\n",
      "Epoch 168/225\n",
      "1754/1754 [==============================] - 2s 932us/step - loss: 0.3770 - binary_accuracy: 0.8434\n",
      "Epoch 169/225\n",
      "1754/1754 [==============================] - 2s 954us/step - loss: 0.3739 - binary_accuracy: 0.8453\n",
      "Epoch 170/225\n",
      "1754/1754 [==============================] - 2s 930us/step - loss: 0.3772 - binary_accuracy: 0.8412\n",
      "Epoch 171/225\n",
      "1754/1754 [==============================] - 2s 917us/step - loss: 0.3779 - binary_accuracy: 0.8432\n",
      "Epoch 172/225\n",
      "1754/1754 [==============================] - 2s 921us/step - loss: 0.3783 - binary_accuracy: 0.8412\n",
      "Epoch 173/225\n",
      "1754/1754 [==============================] - 2s 924us/step - loss: 0.3812 - binary_accuracy: 0.8400\n",
      "Epoch 174/225\n",
      "1754/1754 [==============================] - 2s 919us/step - loss: 0.3782 - binary_accuracy: 0.8443\n",
      "Epoch 175/225\n",
      "1754/1754 [==============================] - 2s 916us/step - loss: 0.3724 - binary_accuracy: 0.8471\n",
      "Epoch 176/225\n",
      "1754/1754 [==============================] - 2s 922us/step - loss: 0.3757 - binary_accuracy: 0.8414\n",
      "Epoch 177/225\n",
      "1754/1754 [==============================] - 2s 919us/step - loss: 0.3778 - binary_accuracy: 0.8390\n",
      "Epoch 178/225\n",
      "1754/1754 [==============================] - 2s 930us/step - loss: 0.3810 - binary_accuracy: 0.8389\n",
      "Epoch 179/225\n",
      "1754/1754 [==============================] - 2s 931us/step - loss: 0.3763 - binary_accuracy: 0.8427\n",
      "Epoch 180/225\n",
      "1754/1754 [==============================] - 2s 926us/step - loss: 0.3792 - binary_accuracy: 0.8417\n",
      "Epoch 181/225\n",
      "1754/1754 [==============================] - 2s 917us/step - loss: 0.3755 - binary_accuracy: 0.8385\n",
      "Epoch 182/225\n",
      "1754/1754 [==============================] - 2s 932us/step - loss: 0.3751 - binary_accuracy: 0.8407\n",
      "Epoch 183/225\n",
      "1754/1754 [==============================] - 2s 933us/step - loss: 0.3731 - binary_accuracy: 0.8449\n",
      "Epoch 184/225\n",
      "1754/1754 [==============================] - 2s 925us/step - loss: 0.3760 - binary_accuracy: 0.8475\n",
      "Epoch 185/225\n",
      "1754/1754 [==============================] - 2s 945us/step - loss: 0.3755 - binary_accuracy: 0.8433\n",
      "Epoch 186/225\n",
      "1754/1754 [==============================] - 2s 938us/step - loss: 0.3769 - binary_accuracy: 0.8431\n",
      "Epoch 187/225\n",
      "1754/1754 [==============================] - 2s 957us/step - loss: 0.3764 - binary_accuracy: 0.8450\n",
      "Epoch 188/225\n",
      "1754/1754 [==============================] - 2s 956us/step - loss: 0.3765 - binary_accuracy: 0.8409\n",
      "Epoch 189/225\n",
      "1754/1754 [==============================] - 2s 1000us/step - loss: 0.3728 - binary_accuracy: 0.8429\n",
      "Epoch 190/225\n",
      "1754/1754 [==============================] - 2s 964us/step - loss: 0.3737 - binary_accuracy: 0.8425\n",
      "Epoch 191/225\n",
      "1754/1754 [==============================] - 2s 955us/step - loss: 0.3809 - binary_accuracy: 0.8391\n",
      "Epoch 192/225\n",
      "1754/1754 [==============================] - 2s 946us/step - loss: 0.3763 - binary_accuracy: 0.8412\n",
      "Epoch 193/225\n",
      "1754/1754 [==============================] - 2s 958us/step - loss: 0.3786 - binary_accuracy: 0.8402\n",
      "Epoch 194/225\n",
      "1754/1754 [==============================] - 2s 965us/step - loss: 0.3707 - binary_accuracy: 0.8424\n",
      "Epoch 195/225\n",
      "1754/1754 [==============================] - 2s 953us/step - loss: 0.3781 - binary_accuracy: 0.8434\n",
      "Epoch 196/225\n",
      "1754/1754 [==============================] - 2s 954us/step - loss: 0.3720 - binary_accuracy: 0.8420\n",
      "Epoch 197/225\n",
      "1754/1754 [==============================] - 2s 950us/step - loss: 0.3745 - binary_accuracy: 0.8428\n",
      "Epoch 198/225\n",
      "1754/1754 [==============================] - 2s 959us/step - loss: 0.3712 - binary_accuracy: 0.8471\n",
      "Epoch 199/225\n",
      "1754/1754 [==============================] - 2s 947us/step - loss: 0.3716 - binary_accuracy: 0.8393\n",
      "Epoch 200/225\n",
      "1754/1754 [==============================] - 2s 945us/step - loss: 0.3756 - binary_accuracy: 0.8422\n",
      "Epoch 201/225\n",
      "1754/1754 [==============================] - 2s 962us/step - loss: 0.3769 - binary_accuracy: 0.8453\n",
      "Epoch 202/225\n",
      "1754/1754 [==============================] - 2s 962us/step - loss: 0.3778 - binary_accuracy: 0.8405\n",
      "Epoch 203/225\n",
      "1754/1754 [==============================] - 2s 964us/step - loss: 0.3721 - binary_accuracy: 0.8418\n",
      "Epoch 204/225\n",
      "1754/1754 [==============================] - 2s 1ms/step - loss: 0.3679 - binary_accuracy: 0.8471\n",
      "Epoch 205/225\n",
      "1754/1754 [==============================] - 2s 931us/step - loss: 0.3671 - binary_accuracy: 0.8413\n",
      "Epoch 206/225\n",
      "1754/1754 [==============================] - 2s 955us/step - loss: 0.3762 - binary_accuracy: 0.8431\n",
      "Epoch 207/225\n",
      "1754/1754 [==============================] - 2s 949us/step - loss: 0.3728 - binary_accuracy: 0.8469\n",
      "Epoch 208/225\n",
      "1754/1754 [==============================] - 2s 966us/step - loss: 0.3724 - binary_accuracy: 0.8460\n",
      "Epoch 209/225\n",
      "1754/1754 [==============================] - 2s 967us/step - loss: 0.3794 - binary_accuracy: 0.8395\n",
      "Epoch 210/225\n",
      "1754/1754 [==============================] - 2s 961us/step - loss: 0.3706 - binary_accuracy: 0.8441\n",
      "Epoch 211/225\n",
      "1754/1754 [==============================] - 2s 930us/step - loss: 0.3765 - binary_accuracy: 0.8446\n",
      "Epoch 212/225\n",
      "1754/1754 [==============================] - 2s 956us/step - loss: 0.3851 - binary_accuracy: 0.8398\n",
      "Epoch 213/225\n",
      "1754/1754 [==============================] - 2s 956us/step - loss: 0.3671 - binary_accuracy: 0.8457\n",
      "Epoch 214/225\n",
      "1754/1754 [==============================] - 2s 955us/step - loss: 0.3722 - binary_accuracy: 0.8449\n",
      "Epoch 215/225\n",
      "1754/1754 [==============================] - 2s 948us/step - loss: 0.3716 - binary_accuracy: 0.8444\n",
      "Epoch 216/225\n",
      "1754/1754 [==============================] - 2s 956us/step - loss: 0.3751 - binary_accuracy: 0.8416\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 217/225\n",
      "1754/1754 [==============================] - 2s 962us/step - loss: 0.3740 - binary_accuracy: 0.8424\n",
      "Epoch 218/225\n",
      "1754/1754 [==============================] - 2s 941us/step - loss: 0.3680 - binary_accuracy: 0.8418\n",
      "Epoch 219/225\n",
      "1754/1754 [==============================] - 2s 978us/step - loss: 0.3703 - binary_accuracy: 0.8474\n",
      "Epoch 220/225\n",
      "1754/1754 [==============================] - 2s 973us/step - loss: 0.3675 - binary_accuracy: 0.8414\n",
      "Epoch 221/225\n",
      "1754/1754 [==============================] - 2s 954us/step - loss: 0.3756 - binary_accuracy: 0.8443\n",
      "Epoch 222/225\n",
      "1754/1754 [==============================] - 2s 963us/step - loss: 0.3718 - binary_accuracy: 0.8449\n",
      "Epoch 223/225\n",
      "1754/1754 [==============================] - 2s 959us/step - loss: 0.3686 - binary_accuracy: 0.8498\n",
      "Epoch 224/225\n",
      "1754/1754 [==============================] - 2s 934us/step - loss: 0.3671 - binary_accuracy: 0.8455\n",
      "Epoch 225/225\n",
      "1754/1754 [==============================] - 2s 949us/step - loss: 0.3734 - binary_accuracy: 0.8428\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/225\n",
      "1754/1754 [==============================] - 4s 2ms/step - loss: 0.4809 - binary_accuracy: 0.8134\n",
      "Epoch 2/225\n",
      "1754/1754 [==============================] - 2s 964us/step - loss: 0.4418 - binary_accuracy: 0.8189\n",
      "Epoch 3/225\n",
      "1754/1754 [==============================] - 2s 949us/step - loss: 0.4347 - binary_accuracy: 0.8218\n",
      "Epoch 4/225\n",
      "1754/1754 [==============================] - 2s 935us/step - loss: 0.4271 - binary_accuracy: 0.8256\n",
      "Epoch 5/225\n",
      "1754/1754 [==============================] - 2s 925us/step - loss: 0.4321 - binary_accuracy: 0.8194\n",
      "Epoch 6/225\n",
      "1754/1754 [==============================] - 2s 970us/step - loss: 0.4253 - binary_accuracy: 0.8229\n",
      "Epoch 7/225\n",
      "1754/1754 [==============================] - 2s 921us/step - loss: 0.4267 - binary_accuracy: 0.8204\n",
      "Epoch 8/225\n",
      "1754/1754 [==============================] - 2s 917us/step - loss: 0.4222 - binary_accuracy: 0.8216\n",
      "Epoch 9/225\n",
      "1754/1754 [==============================] - 2s 918us/step - loss: 0.4203 - binary_accuracy: 0.8214\n",
      "Epoch 10/225\n",
      "1754/1754 [==============================] - 2s 929us/step - loss: 0.4167 - binary_accuracy: 0.8216\n",
      "Epoch 11/225\n",
      "1754/1754 [==============================] - 2s 937us/step - loss: 0.4232 - binary_accuracy: 0.8184\n",
      "Epoch 12/225\n",
      "1754/1754 [==============================] - 2s 931us/step - loss: 0.4176 - binary_accuracy: 0.8193\n",
      "Epoch 13/225\n",
      "1754/1754 [==============================] - 2s 926us/step - loss: 0.4160 - binary_accuracy: 0.8263\n",
      "Epoch 14/225\n",
      "1754/1754 [==============================] - 2s 959us/step - loss: 0.4137 - binary_accuracy: 0.8227\n",
      "Epoch 15/225\n",
      "1754/1754 [==============================] - 2s 947us/step - loss: 0.4195 - binary_accuracy: 0.8249\n",
      "Epoch 16/225\n",
      "1754/1754 [==============================] - 2s 933us/step - loss: 0.4182 - binary_accuracy: 0.8261\n",
      "Epoch 17/225\n",
      "1754/1754 [==============================] - 2s 922us/step - loss: 0.4158 - binary_accuracy: 0.8228\n",
      "Epoch 18/225\n",
      "1754/1754 [==============================] - 2s 932us/step - loss: 0.4091 - binary_accuracy: 0.8293\n",
      "Epoch 19/225\n",
      "1754/1754 [==============================] - 2s 925us/step - loss: 0.4128 - binary_accuracy: 0.8254\n",
      "Epoch 20/225\n",
      "1754/1754 [==============================] - 2s 927us/step - loss: 0.4055 - binary_accuracy: 0.8294\n",
      "Epoch 21/225\n",
      "1754/1754 [==============================] - 2s 984us/step - loss: 0.4166 - binary_accuracy: 0.8256\n",
      "Epoch 22/225\n",
      "1754/1754 [==============================] - 2s 937us/step - loss: 0.4067 - binary_accuracy: 0.8312\n",
      "Epoch 23/225\n",
      "1754/1754 [==============================] - 2s 919us/step - loss: 0.4101 - binary_accuracy: 0.8251\n",
      "Epoch 24/225\n",
      "1754/1754 [==============================] - 2s 937us/step - loss: 0.4108 - binary_accuracy: 0.8287\n",
      "Epoch 25/225\n",
      "1754/1754 [==============================] - 2s 936us/step - loss: 0.4142 - binary_accuracy: 0.8247\n",
      "Epoch 26/225\n",
      "1754/1754 [==============================] - 2s 938us/step - loss: 0.4106 - binary_accuracy: 0.8247\n",
      "Epoch 27/225\n",
      "1754/1754 [==============================] - 2s 940us/step - loss: 0.4085 - binary_accuracy: 0.8319\n",
      "Epoch 28/225\n",
      "1754/1754 [==============================] - 2s 933us/step - loss: 0.4114 - binary_accuracy: 0.8258\n",
      "Epoch 29/225\n",
      "1754/1754 [==============================] - 2s 952us/step - loss: 0.4142 - binary_accuracy: 0.8239\n",
      "Epoch 30/225\n",
      "1754/1754 [==============================] - 2s 949us/step - loss: 0.4067 - binary_accuracy: 0.8277\n",
      "Epoch 31/225\n",
      "1754/1754 [==============================] - 2s 940us/step - loss: 0.4036 - binary_accuracy: 0.8342\n",
      "Epoch 32/225\n",
      "1754/1754 [==============================] - 2s 958us/step - loss: 0.4093 - binary_accuracy: 0.8268\n",
      "Epoch 33/225\n",
      "1754/1754 [==============================] - 2s 940us/step - loss: 0.4063 - binary_accuracy: 0.8305\n",
      "Epoch 34/225\n",
      "1754/1754 [==============================] - 2s 936us/step - loss: 0.4017 - binary_accuracy: 0.8299\n",
      "Epoch 35/225\n",
      "1754/1754 [==============================] - 2s 937us/step - loss: 0.4085 - binary_accuracy: 0.8251\n",
      "Epoch 36/225\n",
      "1754/1754 [==============================] - 2s 946us/step - loss: 0.4028 - binary_accuracy: 0.8280\n",
      "Epoch 37/225\n",
      "1754/1754 [==============================] - 2s 934us/step - loss: 0.4103 - binary_accuracy: 0.8262\n",
      "Epoch 38/225\n",
      "1754/1754 [==============================] - 2s 936us/step - loss: 0.4084 - binary_accuracy: 0.8257\n",
      "Epoch 39/225\n",
      "1754/1754 [==============================] - 2s 922us/step - loss: 0.4034 - binary_accuracy: 0.8302\n",
      "Epoch 40/225\n",
      "1754/1754 [==============================] - 2s 931us/step - loss: 0.4053 - binary_accuracy: 0.8292\n",
      "Epoch 41/225\n",
      "1754/1754 [==============================] - 2s 943us/step - loss: 0.4069 - binary_accuracy: 0.8262\n",
      "Epoch 42/225\n",
      "1754/1754 [==============================] - 2s 920us/step - loss: 0.4046 - binary_accuracy: 0.8305\n",
      "Epoch 43/225\n",
      "1754/1754 [==============================] - 2s 917us/step - loss: 0.4036 - binary_accuracy: 0.8334\n",
      "Epoch 44/225\n",
      "1754/1754 [==============================] - 2s 949us/step - loss: 0.4002 - binary_accuracy: 0.8319\n",
      "Epoch 45/225\n",
      "1754/1754 [==============================] - 2s 940us/step - loss: 0.4048 - binary_accuracy: 0.8261\n",
      "Epoch 46/225\n",
      "1754/1754 [==============================] - 2s 931us/step - loss: 0.3965 - binary_accuracy: 0.8347\n",
      "Epoch 47/225\n",
      "1754/1754 [==============================] - 2s 923us/step - loss: 0.4002 - binary_accuracy: 0.8273\n",
      "Epoch 48/225\n",
      "1754/1754 [==============================] - 2s 918us/step - loss: 0.4020 - binary_accuracy: 0.8311\n",
      "Epoch 49/225\n",
      "1754/1754 [==============================] - 2s 932us/step - loss: 0.3983 - binary_accuracy: 0.8337\n",
      "Epoch 50/225\n",
      "1754/1754 [==============================] - 2s 945us/step - loss: 0.3991 - binary_accuracy: 0.8320\n",
      "Epoch 51/225\n",
      "1754/1754 [==============================] - 2s 938us/step - loss: 0.4001 - binary_accuracy: 0.8299\n",
      "Epoch 52/225\n",
      "1754/1754 [==============================] - 2s 926us/step - loss: 0.4014 - binary_accuracy: 0.8291\n",
      "Epoch 53/225\n",
      "1754/1754 [==============================] - 2s 927us/step - loss: 0.3996 - binary_accuracy: 0.8309\n",
      "Epoch 54/225\n",
      "1754/1754 [==============================] - 2s 935us/step - loss: 0.3957 - binary_accuracy: 0.8320\n",
      "Epoch 55/225\n",
      "1754/1754 [==============================] - 2s 927us/step - loss: 0.3970 - binary_accuracy: 0.8288\n",
      "Epoch 56/225\n",
      "1754/1754 [==============================] - 2s 928us/step - loss: 0.3976 - binary_accuracy: 0.8336\n",
      "Epoch 57/225\n",
      "1754/1754 [==============================] - 2s 933us/step - loss: 0.4023 - binary_accuracy: 0.8259\n",
      "Epoch 58/225\n",
      "1754/1754 [==============================] - 2s 938us/step - loss: 0.3993 - binary_accuracy: 0.8291\n",
      "Epoch 59/225\n",
      "1754/1754 [==============================] - 2s 942us/step - loss: 0.3938 - binary_accuracy: 0.8324\n",
      "Epoch 60/225\n",
      "1754/1754 [==============================] - 2s 937us/step - loss: 0.3949 - binary_accuracy: 0.8310\n",
      "Epoch 61/225\n",
      "1754/1754 [==============================] - 2s 932us/step - loss: 0.3960 - binary_accuracy: 0.8340\n",
      "Epoch 62/225\n",
      "1754/1754 [==============================] - 2s 923us/step - loss: 0.3976 - binary_accuracy: 0.8310\n",
      "Epoch 63/225\n",
      "1754/1754 [==============================] - 2s 933us/step - loss: 0.3913 - binary_accuracy: 0.8331\n",
      "Epoch 64/225\n",
      "1754/1754 [==============================] - 2s 955us/step - loss: 0.3952 - binary_accuracy: 0.8323\n",
      "Epoch 65/225\n",
      "1754/1754 [==============================] - 2s 946us/step - loss: 0.3979 - binary_accuracy: 0.8349\n",
      "Epoch 66/225\n",
      "1754/1754 [==============================] - 2s 934us/step - loss: 0.3923 - binary_accuracy: 0.8368\n",
      "Epoch 67/225\n",
      "1754/1754 [==============================] - 2s 928us/step - loss: 0.3967 - binary_accuracy: 0.8322\n",
      "Epoch 68/225\n",
      "1754/1754 [==============================] - 2s 934us/step - loss: 0.3907 - binary_accuracy: 0.8332\n",
      "Epoch 69/225\n",
      "1754/1754 [==============================] - 2s 932us/step - loss: 0.3978 - binary_accuracy: 0.8322\n",
      "Epoch 70/225\n",
      "1754/1754 [==============================] - 2s 929us/step - loss: 0.3960 - binary_accuracy: 0.8293\n",
      "Epoch 71/225\n",
      "1754/1754 [==============================] - 2s 943us/step - loss: 0.3941 - binary_accuracy: 0.8338\n",
      "Epoch 72/225\n",
      "1754/1754 [==============================] - 2s 965us/step - loss: 0.3966 - binary_accuracy: 0.8341\n",
      "Epoch 73/225\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1754/1754 [==============================] - 2s 932us/step - loss: 0.3920 - binary_accuracy: 0.8303\n",
      "Epoch 74/225\n",
      "1754/1754 [==============================] - 2s 925us/step - loss: 0.3878 - binary_accuracy: 0.8370\n",
      "Epoch 75/225\n",
      "1754/1754 [==============================] - 2s 923us/step - loss: 0.3927 - binary_accuracy: 0.8362\n",
      "Epoch 76/225\n",
      "1754/1754 [==============================] - 2s 925us/step - loss: 0.3933 - binary_accuracy: 0.8349\n",
      "Epoch 77/225\n",
      "1754/1754 [==============================] - 2s 944us/step - loss: 0.3950 - binary_accuracy: 0.8319\n",
      "Epoch 78/225\n",
      "1754/1754 [==============================] - 2s 936us/step - loss: 0.3855 - binary_accuracy: 0.8387\n",
      "Epoch 79/225\n",
      "1754/1754 [==============================] - 2s 929us/step - loss: 0.3951 - binary_accuracy: 0.8318\n",
      "Epoch 80/225\n",
      "1754/1754 [==============================] - 2s 948us/step - loss: 0.3949 - binary_accuracy: 0.8290\n",
      "Epoch 81/225\n",
      "1754/1754 [==============================] - 2s 927us/step - loss: 0.3940 - binary_accuracy: 0.8351\n",
      "Epoch 82/225\n",
      "1754/1754 [==============================] - 2s 926us/step - loss: 0.3974 - binary_accuracy: 0.8329\n",
      "Epoch 83/225\n",
      "1754/1754 [==============================] - 2s 927us/step - loss: 0.3902 - binary_accuracy: 0.8343\n",
      "Epoch 84/225\n",
      "1754/1754 [==============================] - 2s 934us/step - loss: 0.3890 - binary_accuracy: 0.8364\n",
      "Epoch 85/225\n",
      "1754/1754 [==============================] - 2s 930us/step - loss: 0.3923 - binary_accuracy: 0.8386\n",
      "Epoch 86/225\n",
      "1754/1754 [==============================] - 2s 938us/step - loss: 0.3904 - binary_accuracy: 0.8377\n",
      "Epoch 87/225\n",
      "1754/1754 [==============================] - 2s 931us/step - loss: 0.3923 - binary_accuracy: 0.8317\n",
      "Epoch 88/225\n",
      "1754/1754 [==============================] - 2s 932us/step - loss: 0.3905 - binary_accuracy: 0.8348\n",
      "Epoch 89/225\n",
      "1754/1754 [==============================] - 2s 931us/step - loss: 0.3928 - binary_accuracy: 0.8377\n",
      "Epoch 90/225\n",
      "1754/1754 [==============================] - 2s 923us/step - loss: 0.3895 - binary_accuracy: 0.8342\n",
      "Epoch 91/225\n",
      "1754/1754 [==============================] - 2s 928us/step - loss: 0.3906 - binary_accuracy: 0.8342\n",
      "Epoch 92/225\n",
      "1754/1754 [==============================] - 2s 921us/step - loss: 0.3855 - binary_accuracy: 0.8375\n",
      "Epoch 93/225\n",
      "1754/1754 [==============================] - 2s 917us/step - loss: 0.3899 - binary_accuracy: 0.8362\n",
      "Epoch 94/225\n",
      "1754/1754 [==============================] - 2s 919us/step - loss: 0.3847 - binary_accuracy: 0.8379\n",
      "Epoch 95/225\n",
      "1754/1754 [==============================] - 2s 923us/step - loss: 0.3931 - binary_accuracy: 0.8351\n",
      "Epoch 96/225\n",
      "1754/1754 [==============================] - 2s 953us/step - loss: 0.3937 - binary_accuracy: 0.8338\n",
      "Epoch 97/225\n",
      "1754/1754 [==============================] - 2s 927us/step - loss: 0.3918 - binary_accuracy: 0.8337\n",
      "Epoch 98/225\n",
      "1754/1754 [==============================] - 2s 921us/step - loss: 0.3894 - binary_accuracy: 0.8365\n",
      "Epoch 99/225\n",
      "1754/1754 [==============================] - 2s 932us/step - loss: 0.3922 - binary_accuracy: 0.8293\n",
      "Epoch 100/225\n",
      "1754/1754 [==============================] - 2s 926us/step - loss: 0.3939 - binary_accuracy: 0.8326\n",
      "Epoch 101/225\n",
      "1754/1754 [==============================] - 2s 929us/step - loss: 0.3837 - binary_accuracy: 0.8389\n",
      "Epoch 102/225\n",
      "1754/1754 [==============================] - 2s 926us/step - loss: 0.3889 - binary_accuracy: 0.8323\n",
      "Epoch 103/225\n",
      "1754/1754 [==============================] - 2s 942us/step - loss: 0.3914 - binary_accuracy: 0.8334\n",
      "Epoch 104/225\n",
      "1754/1754 [==============================] - 2s 933us/step - loss: 0.3853 - binary_accuracy: 0.8360\n",
      "Epoch 105/225\n",
      "1754/1754 [==============================] - 2s 928us/step - loss: 0.3908 - binary_accuracy: 0.8342\n",
      "Epoch 106/225\n",
      "1754/1754 [==============================] - 2s 930us/step - loss: 0.3870 - binary_accuracy: 0.8324\n",
      "Epoch 107/225\n",
      "1754/1754 [==============================] - 2s 932us/step - loss: 0.3909 - binary_accuracy: 0.8322\n",
      "Epoch 108/225\n",
      "1754/1754 [==============================] - 2s 929us/step - loss: 0.3905 - binary_accuracy: 0.8371\n",
      "Epoch 109/225\n",
      "1754/1754 [==============================] - 2s 936us/step - loss: 0.3927 - binary_accuracy: 0.8352\n",
      "Epoch 110/225\n",
      "1754/1754 [==============================] - 2s 931us/step - loss: 0.3802 - binary_accuracy: 0.8406\n",
      "Epoch 111/225\n",
      "1754/1754 [==============================] - 2s 933us/step - loss: 0.3917 - binary_accuracy: 0.8373\n",
      "Epoch 112/225\n",
      "1754/1754 [==============================] - 2s 928us/step - loss: 0.3867 - binary_accuracy: 0.8376\n",
      "Epoch 113/225\n",
      "1754/1754 [==============================] - 2s 927us/step - loss: 0.3846 - binary_accuracy: 0.8383\n",
      "Epoch 114/225\n",
      "1754/1754 [==============================] - 2s 933us/step - loss: 0.3866 - binary_accuracy: 0.8365\n",
      "Epoch 115/225\n",
      "1754/1754 [==============================] - 2s 936us/step - loss: 0.3874 - binary_accuracy: 0.8380\n",
      "Epoch 116/225\n",
      "1754/1754 [==============================] - 2s 938us/step - loss: 0.3856 - binary_accuracy: 0.8401\n",
      "Epoch 117/225\n",
      "1754/1754 [==============================] - 2s 931us/step - loss: 0.3882 - binary_accuracy: 0.8360\n",
      "Epoch 118/225\n",
      "1754/1754 [==============================] - 2s 932us/step - loss: 0.3934 - binary_accuracy: 0.8306\n",
      "Epoch 119/225\n",
      "1754/1754 [==============================] - 2s 947us/step - loss: 0.3886 - binary_accuracy: 0.8371\n",
      "Epoch 120/225\n",
      "1754/1754 [==============================] - 2s 962us/step - loss: 0.3849 - binary_accuracy: 0.8405\n",
      "Epoch 121/225\n",
      "1754/1754 [==============================] - 2s 927us/step - loss: 0.3916 - binary_accuracy: 0.8316\n",
      "Epoch 122/225\n",
      "1754/1754 [==============================] - 2s 930us/step - loss: 0.3815 - binary_accuracy: 0.8428\n",
      "Epoch 123/225\n",
      "1754/1754 [==============================] - 2s 939us/step - loss: 0.3883 - binary_accuracy: 0.8403\n",
      "Epoch 124/225\n",
      "1754/1754 [==============================] - 2s 938us/step - loss: 0.3869 - binary_accuracy: 0.8352\n",
      "Epoch 125/225\n",
      "1754/1754 [==============================] - 2s 940us/step - loss: 0.3908 - binary_accuracy: 0.8317\n",
      "Epoch 126/225\n",
      "1754/1754 [==============================] - 2s 933us/step - loss: 0.3905 - binary_accuracy: 0.8386\n",
      "Epoch 127/225\n",
      "1754/1754 [==============================] - 2s 939us/step - loss: 0.3891 - binary_accuracy: 0.8361\n",
      "Epoch 128/225\n",
      "1754/1754 [==============================] - 2s 933us/step - loss: 0.3911 - binary_accuracy: 0.8389\n",
      "Epoch 129/225\n",
      "1754/1754 [==============================] - 2s 929us/step - loss: 0.3835 - binary_accuracy: 0.8397\n",
      "Epoch 130/225\n",
      "1754/1754 [==============================] - 2s 929us/step - loss: 0.3829 - binary_accuracy: 0.8421\n",
      "Epoch 131/225\n",
      "1754/1754 [==============================] - 2s 935us/step - loss: 0.3908 - binary_accuracy: 0.8349\n",
      "Epoch 132/225\n",
      "1754/1754 [==============================] - 2s 924us/step - loss: 0.3803 - binary_accuracy: 0.8406\n",
      "Epoch 133/225\n",
      "1754/1754 [==============================] - 2s 923us/step - loss: 0.3865 - binary_accuracy: 0.8344\n",
      "Epoch 134/225\n",
      "1754/1754 [==============================] - 2s 934us/step - loss: 0.3901 - binary_accuracy: 0.8351\n",
      "Epoch 135/225\n",
      "1754/1754 [==============================] - 2s 930us/step - loss: 0.3885 - binary_accuracy: 0.8393\n",
      "Epoch 136/225\n",
      "1754/1754 [==============================] - 2s 924us/step - loss: 0.3832 - binary_accuracy: 0.8380\n",
      "Epoch 137/225\n",
      "1754/1754 [==============================] - 2s 931us/step - loss: 0.3843 - binary_accuracy: 0.8382\n",
      "Epoch 138/225\n",
      "1754/1754 [==============================] - 2s 929us/step - loss: 0.3846 - binary_accuracy: 0.8404\n",
      "Epoch 139/225\n",
      "1754/1754 [==============================] - 2s 933us/step - loss: 0.3885 - binary_accuracy: 0.8341\n",
      "Epoch 140/225\n",
      "1754/1754 [==============================] - 2s 946us/step - loss: 0.3847 - binary_accuracy: 0.8348\n",
      "Epoch 141/225\n",
      "1754/1754 [==============================] - 2s 947us/step - loss: 0.3829 - binary_accuracy: 0.8382\n",
      "Epoch 142/225\n",
      "1754/1754 [==============================] - 2s 930us/step - loss: 0.3788 - binary_accuracy: 0.8385\n",
      "Epoch 143/225\n",
      "1754/1754 [==============================] - 2s 937us/step - loss: 0.3832 - binary_accuracy: 0.8393\n",
      "Epoch 144/225\n",
      "1754/1754 [==============================] - 2s 935us/step - loss: 0.3834 - binary_accuracy: 0.8411\n",
      "Epoch 145/225\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1754/1754 [==============================] - 2s 937us/step - loss: 0.3840 - binary_accuracy: 0.8389\n",
      "Epoch 146/225\n",
      "1754/1754 [==============================] - 2s 963us/step - loss: 0.3814 - binary_accuracy: 0.8403\n",
      "Epoch 147/225\n",
      "1754/1754 [==============================] - 2s 928us/step - loss: 0.3870 - binary_accuracy: 0.8344\n",
      "Epoch 148/225\n",
      "1754/1754 [==============================] - 2s 936us/step - loss: 0.3877 - binary_accuracy: 0.8348\n",
      "Epoch 149/225\n",
      "1754/1754 [==============================] - 2s 931us/step - loss: 0.3843 - binary_accuracy: 0.8390\n",
      "Epoch 150/225\n",
      "1754/1754 [==============================] - 2s 930us/step - loss: 0.3870 - binary_accuracy: 0.8342\n",
      "Epoch 151/225\n",
      "1754/1754 [==============================] - 2s 920us/step - loss: 0.3835 - binary_accuracy: 0.8416\n",
      "Epoch 152/225\n",
      "1754/1754 [==============================] - 2s 929us/step - loss: 0.3791 - binary_accuracy: 0.8390\n",
      "Epoch 153/225\n",
      "1754/1754 [==============================] - 2s 930us/step - loss: 0.3787 - binary_accuracy: 0.8426\n",
      "Epoch 154/225\n",
      "1754/1754 [==============================] - 2s 926us/step - loss: 0.3807 - binary_accuracy: 0.8401\n",
      "Epoch 155/225\n",
      "1754/1754 [==============================] - 2s 924us/step - loss: 0.3797 - binary_accuracy: 0.8404\n",
      "Epoch 156/225\n",
      "1754/1754 [==============================] - 2s 930us/step - loss: 0.3782 - binary_accuracy: 0.8393\n",
      "Epoch 157/225\n",
      "1754/1754 [==============================] - 2s 927us/step - loss: 0.3774 - binary_accuracy: 0.8446\n",
      "Epoch 158/225\n",
      "1754/1754 [==============================] - 2s 929us/step - loss: 0.3814 - binary_accuracy: 0.8356\n",
      "Epoch 159/225\n",
      "1754/1754 [==============================] - 2s 924us/step - loss: 0.3867 - binary_accuracy: 0.8326\n",
      "Epoch 160/225\n",
      "1754/1754 [==============================] - 2s 934us/step - loss: 0.3795 - binary_accuracy: 0.8424\n",
      "Epoch 161/225\n",
      "1754/1754 [==============================] - 2s 934us/step - loss: 0.3791 - binary_accuracy: 0.8433\n",
      "Epoch 162/225\n",
      "1754/1754 [==============================] - 2s 922us/step - loss: 0.3787 - binary_accuracy: 0.8408\n",
      "Epoch 163/225\n",
      "1754/1754 [==============================] - 2s 919us/step - loss: 0.3841 - binary_accuracy: 0.8370\n",
      "Epoch 164/225\n",
      "1754/1754 [==============================] - 2s 917us/step - loss: 0.3827 - binary_accuracy: 0.8421\n",
      "Epoch 165/225\n",
      "1754/1754 [==============================] - 2s 928us/step - loss: 0.3850 - binary_accuracy: 0.8356\n",
      "Epoch 166/225\n",
      "1754/1754 [==============================] - 2s 930us/step - loss: 0.3833 - binary_accuracy: 0.8371\n",
      "Epoch 167/225\n",
      "1754/1754 [==============================] - 2s 923us/step - loss: 0.3873 - binary_accuracy: 0.8393\n",
      "Epoch 168/225\n",
      "1754/1754 [==============================] - 2s 933us/step - loss: 0.3860 - binary_accuracy: 0.8357\n",
      "Epoch 169/225\n",
      "1754/1754 [==============================] - 2s 933us/step - loss: 0.3799 - binary_accuracy: 0.8432\n",
      "Epoch 170/225\n",
      "1754/1754 [==============================] - 2s 927us/step - loss: 0.3749 - binary_accuracy: 0.8396\n",
      "Epoch 171/225\n",
      "1754/1754 [==============================] - 2s 936us/step - loss: 0.3755 - binary_accuracy: 0.8385\n",
      "Epoch 172/225\n",
      "1754/1754 [==============================] - 2s 929us/step - loss: 0.3816 - binary_accuracy: 0.8416\n",
      "Epoch 173/225\n",
      "1754/1754 [==============================] - 2s 942us/step - loss: 0.3765 - binary_accuracy: 0.8419\n",
      "Epoch 174/225\n",
      "1754/1754 [==============================] - 2s 955us/step - loss: 0.3771 - binary_accuracy: 0.8449\n",
      "Epoch 175/225\n",
      "1754/1754 [==============================] - 2s 932us/step - loss: 0.3824 - binary_accuracy: 0.8367\n",
      "Epoch 176/225\n",
      "1754/1754 [==============================] - 2s 925us/step - loss: 0.3832 - binary_accuracy: 0.8366\n",
      "Epoch 177/225\n",
      "1754/1754 [==============================] - 2s 923us/step - loss: 0.3742 - binary_accuracy: 0.8433\n",
      "Epoch 178/225\n",
      "1754/1754 [==============================] - 2s 936us/step - loss: 0.3745 - binary_accuracy: 0.8445\n",
      "Epoch 179/225\n",
      "1754/1754 [==============================] - 2s 925us/step - loss: 0.3764 - binary_accuracy: 0.8461\n",
      "Epoch 180/225\n",
      "1754/1754 [==============================] - 2s 928us/step - loss: 0.3747 - binary_accuracy: 0.8446\n",
      "Epoch 181/225\n",
      "1754/1754 [==============================] - 2s 940us/step - loss: 0.3852 - binary_accuracy: 0.8401\n",
      "Epoch 182/225\n",
      "1754/1754 [==============================] - 2s 961us/step - loss: 0.3751 - binary_accuracy: 0.8397\n",
      "Epoch 183/225\n",
      "1754/1754 [==============================] - 2s 933us/step - loss: 0.3851 - binary_accuracy: 0.8354\n",
      "Epoch 184/225\n",
      "1754/1754 [==============================] - 2s 955us/step - loss: 0.3725 - binary_accuracy: 0.8410\n",
      "Epoch 185/225\n",
      "1754/1754 [==============================] - 2s 925us/step - loss: 0.3746 - binary_accuracy: 0.8408\n",
      "Epoch 186/225\n",
      "1754/1754 [==============================] - 2s 926us/step - loss: 0.3743 - binary_accuracy: 0.8415\n",
      "Epoch 187/225\n",
      "1754/1754 [==============================] - 2s 933us/step - loss: 0.3756 - binary_accuracy: 0.8457\n",
      "Epoch 188/225\n",
      "1754/1754 [==============================] - 2s 923us/step - loss: 0.3699 - binary_accuracy: 0.8437\n",
      "Epoch 189/225\n",
      "1754/1754 [==============================] - 2s 937us/step - loss: 0.3746 - binary_accuracy: 0.8444\n",
      "Epoch 190/225\n",
      "1754/1754 [==============================] - 2s 922us/step - loss: 0.3668 - binary_accuracy: 0.8457\n",
      "Epoch 191/225\n",
      "1754/1754 [==============================] - 2s 923us/step - loss: 0.3790 - binary_accuracy: 0.8338\n",
      "Epoch 192/225\n",
      "1754/1754 [==============================] - 2s 926us/step - loss: 0.3764 - binary_accuracy: 0.8428\n",
      "Epoch 193/225\n",
      "1754/1754 [==============================] - 2s 929us/step - loss: 0.3730 - binary_accuracy: 0.8453\n",
      "Epoch 194/225\n",
      "1754/1754 [==============================] - 2s 940us/step - loss: 0.3796 - binary_accuracy: 0.8377\n",
      "Epoch 195/225\n",
      "1754/1754 [==============================] - 2s 961us/step - loss: 0.3763 - binary_accuracy: 0.8432\n",
      "Epoch 196/225\n",
      "1754/1754 [==============================] - 2s 966us/step - loss: 0.3818 - binary_accuracy: 0.8402\n",
      "Epoch 197/225\n",
      "1754/1754 [==============================] - 2s 929us/step - loss: 0.3773 - binary_accuracy: 0.8412\n",
      "Epoch 198/225\n",
      "1754/1754 [==============================] - 2s 922us/step - loss: 0.3730 - binary_accuracy: 0.8402\n",
      "Epoch 199/225\n",
      "1754/1754 [==============================] - 2s 912us/step - loss: 0.3756 - binary_accuracy: 0.8436\n",
      "Epoch 200/225\n",
      "1754/1754 [==============================] - 2s 930us/step - loss: 0.3729 - binary_accuracy: 0.8444\n",
      "Epoch 201/225\n",
      "1754/1754 [==============================] - 2s 931us/step - loss: 0.3664 - binary_accuracy: 0.8439\n",
      "Epoch 202/225\n",
      "1754/1754 [==============================] - 2s 931us/step - loss: 0.3722 - binary_accuracy: 0.8409\n",
      "Epoch 203/225\n",
      "1754/1754 [==============================] - 2s 925us/step - loss: 0.3760 - binary_accuracy: 0.8393\n",
      "Epoch 204/225\n",
      "1754/1754 [==============================] - 2s 930us/step - loss: 0.3755 - binary_accuracy: 0.8460\n",
      "Epoch 205/225\n",
      "1754/1754 [==============================] - 2s 931us/step - loss: 0.3742 - binary_accuracy: 0.8426\n",
      "Epoch 206/225\n",
      "1754/1754 [==============================] - 2s 932us/step - loss: 0.3731 - binary_accuracy: 0.8395\n",
      "Epoch 207/225\n",
      "1754/1754 [==============================] - 2s 919us/step - loss: 0.3709 - binary_accuracy: 0.8477\n",
      "Epoch 208/225\n",
      "1754/1754 [==============================] - 2s 929us/step - loss: 0.3760 - binary_accuracy: 0.8448\n",
      "Epoch 209/225\n",
      "1754/1754 [==============================] - 2s 931us/step - loss: 0.3720 - binary_accuracy: 0.8422\n",
      "Epoch 210/225\n",
      "1754/1754 [==============================] - 2s 929us/step - loss: 0.3713 - binary_accuracy: 0.8453\n",
      "Epoch 211/225\n",
      "1754/1754 [==============================] - 2s 931us/step - loss: 0.3809 - binary_accuracy: 0.8363\n",
      "Epoch 212/225\n",
      "1754/1754 [==============================] - 2s 924us/step - loss: 0.3715 - binary_accuracy: 0.8468\n",
      "Epoch 213/225\n",
      "1754/1754 [==============================] - 2s 920us/step - loss: 0.3753 - binary_accuracy: 0.8391\n",
      "Epoch 214/225\n",
      "1754/1754 [==============================] - 2s 930us/step - loss: 0.3721 - binary_accuracy: 0.8437\n",
      "Epoch 215/225\n",
      "1754/1754 [==============================] - 2s 944us/step - loss: 0.3765 - binary_accuracy: 0.8441\n",
      "Epoch 216/225\n",
      "1754/1754 [==============================] - 2s 975us/step - loss: 0.3702 - binary_accuracy: 0.8456\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 217/225\n",
      "1754/1754 [==============================] - 2s 947us/step - loss: 0.3753 - binary_accuracy: 0.8409\n",
      "Epoch 218/225\n",
      "1754/1754 [==============================] - 2s 935us/step - loss: 0.3771 - binary_accuracy: 0.8415\n",
      "Epoch 219/225\n",
      "1754/1754 [==============================] - 2s 923us/step - loss: 0.3767 - binary_accuracy: 0.8356\n",
      "Epoch 220/225\n",
      "1754/1754 [==============================] - 2s 919us/step - loss: 0.3796 - binary_accuracy: 0.8401\n",
      "Epoch 221/225\n",
      "1754/1754 [==============================] - 2s 918us/step - loss: 0.3696 - binary_accuracy: 0.8448\n",
      "Epoch 222/225\n",
      "1754/1754 [==============================] - 2s 920us/step - loss: 0.3685 - binary_accuracy: 0.8454\n",
      "Epoch 223/225\n",
      "1754/1754 [==============================] - 2s 934us/step - loss: 0.3708 - binary_accuracy: 0.8433\n",
      "Epoch 224/225\n",
      "1754/1754 [==============================] - 2s 931us/step - loss: 0.3694 - binary_accuracy: 0.8453\n",
      "Epoch 225/225\n",
      "1754/1754 [==============================] - 2s 935us/step - loss: 0.3701 - binary_accuracy: 0.8458\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "## Run 5-fold cross-validation to get average performance characteristics\n",
    "unit = 32\n",
    "ratio = 0.5\n",
    "epoch = 225\n",
    "(auroc_list, other_metric_list) = kfold_validation(unit, ratio, epoch, X_train_transformed, y_train, 'patient-id', \\\n",
    "                                                   n_features=n_features, verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "average precision:0.3812003577835038\n",
      "average recall:0.5559366446076757\n",
      "average specificity:0.7860004664198115\n",
      "average F1:0.4413213371046768\n"
     ]
    }
   ],
   "source": [
    "## other_metric_list has precision, recall, accuracy, specificity, f1, threshold for metrics\n",
    "ave_prec = np.mean([x[0] for x in other_metric_list])\n",
    "print(\"average precision:\" + str(ave_prec))\n",
    "\n",
    "ave_recall = np.mean([x[1] for x in other_metric_list])\n",
    "print(\"average recall:\" + str(ave_recall))\n",
    "\n",
    "ave_specificity = np.mean([x[3] for x in other_metric_list])\n",
    "print(\"average specificity:\" + str(ave_specificity))\n",
    "\n",
    "ave_f1 = np.mean([x[4] for x in other_metric_list])\n",
    "print(\"average F1:\" + str(ave_f1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "max F1: 0.46194225721784776\n"
     ]
    }
   ],
   "source": [
    "max_f1 = np.max([x[4] for x in other_metric_list])\n",
    "print(\"max F1: \" + str(max_f1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "unit: 32\tepoch: 225\tauroc_mean: 0.718560\tauroc_std: 0.015827\n"
     ]
    }
   ],
   "source": [
    "## Average metrics performance\n",
    "print(\"unit: %d\\tepoch: %d\\tauroc_mean: %f\\tauroc_std: %f\"  % (unit, epoch, np.mean(auroc_list), np.std(auroc_list)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}