Module3_Step3_Bounce_detector_final.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "402f4f9c",
   "metadata": {},
   "source": [
    "This notebook takes the <b> \"model2_inputdata.csv\" </b> file generated from pre-processing step and uses it to create Machine learning model for bounce detection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "44159c4f",
   "metadata": {},
   "outputs": [],
   "source": [
    "import mlflow.tensorflow\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')\n",
    "import pandas as pd\n",
    "import sklearn\n",
    "import re, numpy as np, tensorflow as tf\n",
    "from keras.layers import Bidirectional\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import LSTM, SimpleRNN, Bidirectional\n",
    "from tensorflow.keras.layers import Dense, Dropout\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "521e72b0",
   "metadata": {},
   "outputs": [],
   "source": [
    "# ground truth is 95 bounces.\n",
    "\n",
    "from numpy.random import seed\n",
    "seed(10)\n",
    "tf.random.set_seed(10)\n",
    "\n",
    "df = pd.read_csv(\"model2_inputdata.csv\", header=None)\n",
    "df.columns = ['file_name', 'top_left_x', 'top_left_y', 'width', 'height']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "f95819cc",
   "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>file_name</th>\n",
       "      <th>top_left_x</th>\n",
       "      <th>top_left_y</th>\n",
       "      <th>width</th>\n",
       "      <th>height</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>image1.txt</td>\n",
       "      <td>0.4216</td>\n",
       "      <td>0.4996</td>\n",
       "      <td>0.0091</td>\n",
       "      <td>0.0148</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>image2.txt</td>\n",
       "      <td>0.4233</td>\n",
       "      <td>0.4812</td>\n",
       "      <td>0.0083</td>\n",
       "      <td>0.0158</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>image3.txt</td>\n",
       "      <td>0.4249</td>\n",
       "      <td>0.4632</td>\n",
       "      <td>0.0088</td>\n",
       "      <td>0.0148</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>image4.txt</td>\n",
       "      <td>0.4259</td>\n",
       "      <td>0.4551</td>\n",
       "      <td>0.0096</td>\n",
       "      <td>0.0149</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>image5.txt</td>\n",
       "      <td>0.4277</td>\n",
       "      <td>0.4402</td>\n",
       "      <td>0.0094</td>\n",
       "      <td>0.0148</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    file_name  top_left_x  top_left_y   width  height\n",
       "0  image1.txt      0.4216      0.4996  0.0091  0.0148\n",
       "1  image2.txt      0.4233      0.4812  0.0083  0.0158\n",
       "2  image3.txt      0.4249      0.4632  0.0088  0.0148\n",
       "3  image4.txt      0.4259      0.4551  0.0096  0.0149\n",
       "4  image5.txt      0.4277      0.4402  0.0094  0.0148"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "478c5131",
   "metadata": {},
   "outputs": [],
   "source": [
    "def id_extractor(image_name):\n",
    "    return int(re.findall('\\d+', image_name)[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "be5a2553",
   "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>file_name</th>\n",
       "      <th>top_left_x</th>\n",
       "      <th>top_left_y</th>\n",
       "      <th>width</th>\n",
       "      <th>height</th>\n",
       "      <th>image_id</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>image1.txt</td>\n",
       "      <td>0.4216</td>\n",
       "      <td>0.4996</td>\n",
       "      <td>0.0091</td>\n",
       "      <td>0.0148</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>image2.txt</td>\n",
       "      <td>0.4233</td>\n",
       "      <td>0.4812</td>\n",
       "      <td>0.0083</td>\n",
       "      <td>0.0158</td>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>image3.txt</td>\n",
       "      <td>0.4249</td>\n",
       "      <td>0.4632</td>\n",
       "      <td>0.0088</td>\n",
       "      <td>0.0148</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>image4.txt</td>\n",
       "      <td>0.4259</td>\n",
       "      <td>0.4551</td>\n",
       "      <td>0.0096</td>\n",
       "      <td>0.0149</td>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>image5.txt</td>\n",
       "      <td>0.4277</td>\n",
       "      <td>0.4402</td>\n",
       "      <td>0.0094</td>\n",
       "      <td>0.0148</td>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    file_name  top_left_x  top_left_y   width  height  image_id  label\n",
       "0  image1.txt      0.4216      0.4996  0.0091  0.0148         1      0\n",
       "1  image2.txt      0.4233      0.4812  0.0083  0.0158         2      0\n",
       "2  image3.txt      0.4249      0.4632  0.0088  0.0148         3      0\n",
       "3  image4.txt      0.4259      0.4551  0.0096  0.0149         4      0\n",
       "4  image5.txt      0.4277      0.4402  0.0094  0.0148         5      0"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['image_id'] = df['file_name'].apply(id_extractor)\n",
    "df['label'] = 0\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "c83edd25",
   "metadata": {},
   "outputs": [],
   "source": [
    "ground_truth_bounce_frames = [36, 85, 126, 178, 415, 466, 504, 553, 592, 668, 947, 1000, 1054, 1135, 1312, 1557, 1600,\n",
    "                              1651, 1721, 1973, 2016, 2052, 2095, 2140, 2180, 2221, 2269, 2315, 2354, 2399, 2442, 2493,\n",
    "                              2535, 2574, 2615, 2658, 2700, 2738, 2774, 2825, 2862, 3080, 3115, 3164, 3202,\n",
    "                              3249, 3311, 3373, 3416, 3697, 3742, 3787, 3828, 3873, 3907, 3954, 3986, 4227, 4277, 4318,\n",
    "                              4361, 4402, 4446, 4499, 4531, 4580, 4610, 4673, 4732, 4767, 4801, 4846, 4890, 4931, 4964,\n",
    "                              5024, 5241, 5283, 5325, 5366, 5406, 5442, 5493, 5532, 5568, 5610, 5646, 5735, 5788, 5852,\n",
    "                              6201, 6248, 6286, 6350, 6413, 6671, 6770, 6807, 6867, 7066, 7113]\n",
    "\n",
    "def labeller(col_value):\n",
    "    if col_value in ground_truth_bounce_frames:\n",
    "        return 1\n",
    "    else:\n",
    "        return 0\n",
    "\n",
    "df['label'] = df['image_id'].apply(labeller)\n",
    "df = df.reset_index(drop=True)\n",
    "\n",
    "df['centroid_x'] = df['top_left_x'] + df['width'] / 2\n",
    "df['centroid_y'] = df['top_left_y'] + df['height'] / 2\n",
    "df = df.drop(columns=['file_name', 'top_left_x', 'top_left_y', 'width', 'height'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "3ecae71e",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "def windowing(window_size:int)->pd.DataFrame:\n",
    "    X_train = []\n",
    "    y_train = []\n",
    "    window = window_size\n",
    "    \"\"\"\n",
    "    valid means window is continous. invalid window is [image1,2,3,4,5,75,76,77,78,79] \n",
    "    window 10 --> 3247 sliding valid windows\n",
    "    window 20 --> 2222 sliding valid windows\n",
    "    window 30 --> 1589 sliding valid windows\n",
    "    window 40 --> 1124 sliding valid windows\n",
    "    window 50 --> 816 sliding valid windows\n",
    "    window 60 --> 615 sliding valid windows\n",
    "    \"\"\"\n",
    "\n",
    "    for start in range(len(df) - window):\n",
    "        window_end = df['image_id'][start + window - 1]\n",
    "        window_end_max = df['image_id'][start] + window - 1\n",
    "        if window_end <= window_end_max:\n",
    "            X_train.append(df[['centroid_x', 'centroid_y']][start:start + window])\n",
    "            y_train.append(np.max(df['label'][start:start + window], axis=0))\n",
    "\n",
    "    X_train = np.array(X_train)\n",
    "    y_train = np.array(y_train).reshape(len(y_train), 1)\n",
    "    bounce_no = len([0 for i in y_train if i == 0])\n",
    "    bounce_yes = len([0 for i in y_train if i == 1])\n",
    "    print('If window = {}, {} % windows have bounce'.format(window, np.round(bounce_yes*100/len(y_train),3)))\n",
    "    return X_train,y_train"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "e1fad757",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "If window = 10, 21.836 % windows have bounce\n"
     ]
    }
   ],
   "source": [
    "X_train,y_train = windowing(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "24ee1d47",
   "metadata": {},
   "source": [
    "<b> Modeling"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "88dceb54",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/150\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2022-04-23 15:19:32.964915: I tensorflow/core/platform/cpu_feature_guard.cc:151] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2 FMA\n",
      "To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "163/163 - 2s - loss: 0.5933 - auc: 0.4391 - val_loss: 0.4889 - val_auc: 0.4759 - 2s/epoch - 14ms/step\n",
      "Epoch 2/150\n",
      "163/163 - 1s - loss: 0.5525 - auc: 0.4690 - val_loss: 0.4746 - val_auc: 0.3974 - 659ms/epoch - 4ms/step\n",
      "Epoch 3/150\n",
      "163/163 - 1s - loss: 0.5354 - auc: 0.5690 - val_loss: 0.4892 - val_auc: 0.4940 - 673ms/epoch - 4ms/step\n",
      "Epoch 4/150\n",
      "163/163 - 1s - loss: 0.5217 - auc: 0.6281 - val_loss: 0.4654 - val_auc: 0.5125 - 695ms/epoch - 4ms/step\n",
      "Epoch 5/150\n",
      "163/163 - 1s - loss: 0.5151 - auc: 0.6491 - val_loss: 0.4734 - val_auc: 0.5677 - 709ms/epoch - 4ms/step\n",
      "Epoch 6/150\n",
      "163/163 - 1s - loss: 0.5111 - auc: 0.6562 - val_loss: 0.4720 - val_auc: 0.5727 - 685ms/epoch - 4ms/step\n",
      "Epoch 7/150\n",
      "163/163 - 1s - loss: 0.5027 - auc: 0.6790 - val_loss: 0.4114 - val_auc: 0.7152 - 832ms/epoch - 5ms/step\n",
      "Epoch 8/150\n",
      "163/163 - 1s - loss: 0.5050 - auc: 0.6754 - val_loss: 0.4367 - val_auc: 0.7012 - 766ms/epoch - 5ms/step\n",
      "Epoch 9/150\n",
      "163/163 - 1s - loss: 0.4983 - auc: 0.6904 - val_loss: 0.4324 - val_auc: 0.6980 - 659ms/epoch - 4ms/step\n",
      "Epoch 10/150\n",
      "163/163 - 1s - loss: 0.4970 - auc: 0.6936 - val_loss: 0.4488 - val_auc: 0.6430 - 716ms/epoch - 4ms/step\n",
      "Epoch 11/150\n",
      "163/163 - 1s - loss: 0.4945 - auc: 0.6942 - val_loss: 0.4205 - val_auc: 0.6876 - 654ms/epoch - 4ms/step\n",
      "Epoch 12/150\n",
      "163/163 - 1s - loss: 0.4864 - auc: 0.7092 - val_loss: 0.4590 - val_auc: 0.6767 - 676ms/epoch - 4ms/step\n",
      "Epoch 13/150\n",
      "163/163 - 1s - loss: 0.4913 - auc: 0.7023 - val_loss: 0.4049 - val_auc: 0.7213 - 654ms/epoch - 4ms/step\n",
      "Epoch 14/150\n",
      "163/163 - 1s - loss: 0.4817 - auc: 0.7234 - val_loss: 0.4243 - val_auc: 0.6937 - 664ms/epoch - 4ms/step\n",
      "Epoch 15/150\n",
      "163/163 - 1s - loss: 0.4821 - auc: 0.7143 - val_loss: 0.4196 - val_auc: 0.7172 - 654ms/epoch - 4ms/step\n",
      "Epoch 16/150\n",
      "163/163 - 1s - loss: 0.4864 - auc: 0.7069 - val_loss: 0.4090 - val_auc: 0.7219 - 670ms/epoch - 4ms/step\n",
      "Epoch 17/150\n",
      "163/163 - 1s - loss: 0.4749 - auc: 0.7303 - val_loss: 0.4559 - val_auc: 0.6869 - 723ms/epoch - 4ms/step\n",
      "Epoch 18/150\n",
      "163/163 - 1s - loss: 0.4822 - auc: 0.7176 - val_loss: 0.4306 - val_auc: 0.6906 - 688ms/epoch - 4ms/step\n",
      "Epoch 19/150\n",
      "163/163 - 1s - loss: 0.4775 - auc: 0.7268 - val_loss: 0.4219 - val_auc: 0.7051 - 703ms/epoch - 4ms/step\n",
      "Epoch 20/150\n",
      "163/163 - 1s - loss: 0.4725 - auc: 0.7359 - val_loss: 0.4219 - val_auc: 0.7099 - 668ms/epoch - 4ms/step\n",
      "Epoch 21/150\n",
      "163/163 - 1s - loss: 0.4639 - auc: 0.7436 - val_loss: 0.4198 - val_auc: 0.7253 - 660ms/epoch - 4ms/step\n",
      "Epoch 22/150\n",
      "163/163 - 1s - loss: 0.4650 - auc: 0.7392 - val_loss: 0.4057 - val_auc: 0.7335 - 672ms/epoch - 4ms/step\n",
      "Epoch 23/150\n",
      "163/163 - 1s - loss: 0.4597 - auc: 0.7489 - val_loss: 0.4347 - val_auc: 0.7077 - 688ms/epoch - 4ms/step\n",
      "Epoch 24/150\n",
      "163/163 - 1s - loss: 0.4609 - auc: 0.7459 - val_loss: 0.4375 - val_auc: 0.6964 - 694ms/epoch - 4ms/step\n",
      "Epoch 25/150\n",
      "163/163 - 1s - loss: 0.4534 - auc: 0.7593 - val_loss: 0.4031 - val_auc: 0.7494 - 682ms/epoch - 4ms/step\n",
      "Epoch 26/150\n",
      "163/163 - 1s - loss: 0.4625 - auc: 0.7475 - val_loss: 0.4158 - val_auc: 0.7239 - 695ms/epoch - 4ms/step\n",
      "Epoch 27/150\n",
      "163/163 - 1s - loss: 0.4596 - auc: 0.7523 - val_loss: 0.4036 - val_auc: 0.7390 - 678ms/epoch - 4ms/step\n",
      "Epoch 28/150\n",
      "163/163 - 1s - loss: 0.4556 - auc: 0.7556 - val_loss: 0.4433 - val_auc: 0.7179 - 691ms/epoch - 4ms/step\n",
      "Epoch 29/150\n",
      "163/163 - 1s - loss: 0.4575 - auc: 0.7561 - val_loss: 0.4241 - val_auc: 0.7134 - 686ms/epoch - 4ms/step\n",
      "Epoch 30/150\n",
      "163/163 - 1s - loss: 0.4406 - auc: 0.7692 - val_loss: 0.4200 - val_auc: 0.7224 - 677ms/epoch - 4ms/step\n",
      "Epoch 31/150\n",
      "163/163 - 1s - loss: 0.4506 - auc: 0.7566 - val_loss: 0.4332 - val_auc: 0.7224 - 693ms/epoch - 4ms/step\n",
      "Epoch 32/150\n",
      "163/163 - 1s - loss: 0.4401 - auc: 0.7754 - val_loss: 0.4228 - val_auc: 0.7251 - 670ms/epoch - 4ms/step\n",
      "Epoch 33/150\n",
      "163/163 - 1s - loss: 0.4439 - auc: 0.7703 - val_loss: 0.4027 - val_auc: 0.7550 - 680ms/epoch - 4ms/step\n",
      "Epoch 34/150\n",
      "163/163 - 1s - loss: 0.4445 - auc: 0.7670 - val_loss: 0.4014 - val_auc: 0.7347 - 719ms/epoch - 4ms/step\n",
      "Epoch 35/150\n",
      "163/163 - 1s - loss: 0.4376 - auc: 0.7751 - val_loss: 0.4336 - val_auc: 0.7145 - 652ms/epoch - 4ms/step\n",
      "Epoch 36/150\n",
      "163/163 - 1s - loss: 0.4342 - auc: 0.7853 - val_loss: 0.3879 - val_auc: 0.7634 - 650ms/epoch - 4ms/step\n",
      "Epoch 37/150\n",
      "163/163 - 1s - loss: 0.4421 - auc: 0.7662 - val_loss: 0.4499 - val_auc: 0.7102 - 648ms/epoch - 4ms/step\n",
      "Epoch 38/150\n",
      "163/163 - 1s - loss: 0.4350 - auc: 0.7795 - val_loss: 0.3988 - val_auc: 0.7639 - 660ms/epoch - 4ms/step\n",
      "Epoch 39/150\n",
      "163/163 - 1s - loss: 0.4361 - auc: 0.7740 - val_loss: 0.4128 - val_auc: 0.7402 - 690ms/epoch - 4ms/step\n",
      "Epoch 40/150\n",
      "163/163 - 1s - loss: 0.4382 - auc: 0.7738 - val_loss: 0.4334 - val_auc: 0.7254 - 685ms/epoch - 4ms/step\n",
      "Epoch 41/150\n",
      "163/163 - 1s - loss: 0.4336 - auc: 0.7814 - val_loss: 0.3969 - val_auc: 0.7414 - 704ms/epoch - 4ms/step\n",
      "Epoch 42/150\n",
      "163/163 - 1s - loss: 0.4342 - auc: 0.7825 - val_loss: 0.3986 - val_auc: 0.7595 - 666ms/epoch - 4ms/step\n",
      "Epoch 43/150\n",
      "163/163 - 1s - loss: 0.4291 - auc: 0.7843 - val_loss: 0.3992 - val_auc: 0.7521 - 662ms/epoch - 4ms/step\n",
      "Epoch 44/150\n",
      "163/163 - 1s - loss: 0.4258 - auc: 0.7933 - val_loss: 0.4060 - val_auc: 0.7445 - 674ms/epoch - 4ms/step\n",
      "Epoch 45/150\n",
      "163/163 - 1s - loss: 0.4419 - auc: 0.7685 - val_loss: 0.4234 - val_auc: 0.7465 - 687ms/epoch - 4ms/step\n",
      "Epoch 46/150\n",
      "163/163 - 1s - loss: 0.4221 - auc: 0.7971 - val_loss: 0.3834 - val_auc: 0.7700 - 654ms/epoch - 4ms/step\n",
      "Epoch 47/150\n",
      "163/163 - 1s - loss: 0.4266 - auc: 0.7919 - val_loss: 0.3894 - val_auc: 0.7625 - 652ms/epoch - 4ms/step\n",
      "Epoch 48/150\n",
      "163/163 - 1s - loss: 0.4185 - auc: 0.7969 - val_loss: 0.4219 - val_auc: 0.7458 - 649ms/epoch - 4ms/step\n",
      "Epoch 49/150\n",
      "163/163 - 1s - loss: 0.4205 - auc: 0.7936 - val_loss: 0.4334 - val_auc: 0.7521 - 653ms/epoch - 4ms/step\n",
      "Epoch 50/150\n",
      "163/163 - 1s - loss: 0.4226 - auc: 0.7951 - val_loss: 0.4088 - val_auc: 0.7496 - 662ms/epoch - 4ms/step\n",
      "Epoch 51/150\n",
      "163/163 - 1s - loss: 0.4304 - auc: 0.7817 - val_loss: 0.3993 - val_auc: 0.7728 - 683ms/epoch - 4ms/step\n",
      "Epoch 52/150\n",
      "163/163 - 1s - loss: 0.4151 - auc: 0.8013 - val_loss: 0.3959 - val_auc: 0.7586 - 673ms/epoch - 4ms/step\n",
      "Epoch 53/150\n",
      "163/163 - 1s - loss: 0.4213 - auc: 0.7951 - val_loss: 0.4091 - val_auc: 0.7442 - 698ms/epoch - 4ms/step\n",
      "Epoch 54/150\n",
      "163/163 - 1s - loss: 0.4167 - auc: 0.8049 - val_loss: 0.3919 - val_auc: 0.7608 - 685ms/epoch - 4ms/step\n",
      "Epoch 55/150\n",
      "163/163 - 1s - loss: 0.4172 - auc: 0.7999 - val_loss: 0.4078 - val_auc: 0.7484 - 678ms/epoch - 4ms/step\n",
      "Epoch 56/150\n",
      "163/163 - 1s - loss: 0.4178 - auc: 0.8000 - val_loss: 0.4892 - val_auc: 0.7336 - 690ms/epoch - 4ms/step\n",
      "Epoch 57/150\n",
      "163/163 - 1s - loss: 0.4171 - auc: 0.8001 - val_loss: 0.4116 - val_auc: 0.7575 - 713ms/epoch - 4ms/step\n",
      "Epoch 58/150\n",
      "163/163 - 1s - loss: 0.4121 - auc: 0.8042 - val_loss: 0.3952 - val_auc: 0.7500 - 726ms/epoch - 4ms/step\n",
      "Epoch 59/150\n",
      "163/163 - 1s - loss: 0.4064 - auc: 0.8055 - val_loss: 0.3816 - val_auc: 0.7688 - 695ms/epoch - 4ms/step\n",
      "Epoch 60/150\n",
      "163/163 - 1s - loss: 0.4149 - auc: 0.7960 - val_loss: 0.3842 - val_auc: 0.7683 - 691ms/epoch - 4ms/step\n",
      "Epoch 61/150\n",
      "163/163 - 1s - loss: 0.4112 - auc: 0.8041 - val_loss: 0.4605 - val_auc: 0.7527 - 688ms/epoch - 4ms/step\n",
      "Epoch 62/150\n",
      "163/163 - 1s - loss: 0.4189 - auc: 0.7975 - val_loss: 0.3880 - val_auc: 0.7614 - 699ms/epoch - 4ms/step\n",
      "Epoch 63/150\n",
      "163/163 - 1s - loss: 0.4054 - auc: 0.8082 - val_loss: 0.3941 - val_auc: 0.7568 - 671ms/epoch - 4ms/step\n",
      "Epoch 64/150\n",
      "163/163 - 1s - loss: 0.4022 - auc: 0.8126 - val_loss: 0.3921 - val_auc: 0.7472 - 708ms/epoch - 4ms/step\n",
      "Epoch 65/150\n",
      "163/163 - 1s - loss: 0.4078 - auc: 0.8073 - val_loss: 0.3886 - val_auc: 0.7737 - 747ms/epoch - 5ms/step\n",
      "Epoch 66/150\n",
      "163/163 - 1s - loss: 0.4083 - auc: 0.8083 - val_loss: 0.3838 - val_auc: 0.7743 - 699ms/epoch - 4ms/step\n",
      "Epoch 67/150\n",
      "163/163 - 1s - loss: 0.4092 - auc: 0.8094 - val_loss: 0.4054 - val_auc: 0.7508 - 688ms/epoch - 4ms/step\n",
      "Epoch 68/150\n",
      "163/163 - 1s - loss: 0.4037 - auc: 0.8139 - val_loss: 0.3853 - val_auc: 0.7790 - 697ms/epoch - 4ms/step\n",
      "Epoch 69/150\n",
      "163/163 - 1s - loss: 0.4075 - auc: 0.8104 - val_loss: 0.3810 - val_auc: 0.7763 - 675ms/epoch - 4ms/step\n",
      "Epoch 70/150\n",
      "163/163 - 1s - loss: 0.3973 - auc: 0.8190 - val_loss: 0.4173 - val_auc: 0.7498 - 675ms/epoch - 4ms/step\n",
      "Epoch 71/150\n",
      "163/163 - 1s - loss: 0.4084 - auc: 0.8110 - val_loss: 0.3866 - val_auc: 0.7574 - 688ms/epoch - 4ms/step\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 72/150\n",
      "163/163 - 1s - loss: 0.4030 - auc: 0.8098 - val_loss: 0.3707 - val_auc: 0.7814 - 700ms/epoch - 4ms/step\n",
      "Epoch 73/150\n",
      "163/163 - 1s - loss: 0.3985 - auc: 0.8135 - val_loss: 0.3990 - val_auc: 0.7511 - 654ms/epoch - 4ms/step\n",
      "Epoch 74/150\n",
      "163/163 - 1s - loss: 0.4108 - auc: 0.8083 - val_loss: 0.3803 - val_auc: 0.7809 - 676ms/epoch - 4ms/step\n",
      "Epoch 75/150\n",
      "163/163 - 1s - loss: 0.4038 - auc: 0.8100 - val_loss: 0.3757 - val_auc: 0.7805 - 689ms/epoch - 4ms/step\n",
      "Epoch 76/150\n",
      "163/163 - 1s - loss: 0.4099 - auc: 0.8102 - val_loss: 0.3785 - val_auc: 0.7965 - 669ms/epoch - 4ms/step\n",
      "Epoch 77/150\n",
      "163/163 - 1s - loss: 0.3968 - auc: 0.8240 - val_loss: 0.3883 - val_auc: 0.7543 - 684ms/epoch - 4ms/step\n",
      "Epoch 78/150\n",
      "163/163 - 1s - loss: 0.3942 - auc: 0.8239 - val_loss: 0.3876 - val_auc: 0.7694 - 665ms/epoch - 4ms/step\n",
      "Epoch 79/150\n",
      "163/163 - 1s - loss: 0.3911 - auc: 0.8245 - val_loss: 0.3614 - val_auc: 0.8023 - 695ms/epoch - 4ms/step\n",
      "Epoch 80/150\n",
      "163/163 - 1s - loss: 0.3975 - auc: 0.8183 - val_loss: 0.3870 - val_auc: 0.7677 - 698ms/epoch - 4ms/step\n",
      "Epoch 81/150\n",
      "163/163 - 1s - loss: 0.3920 - auc: 0.8280 - val_loss: 0.3726 - val_auc: 0.7924 - 693ms/epoch - 4ms/step\n",
      "Epoch 82/150\n",
      "163/163 - 1s - loss: 0.3964 - auc: 0.8187 - val_loss: 0.3683 - val_auc: 0.8003 - 730ms/epoch - 4ms/step\n",
      "Epoch 83/150\n",
      "163/163 - 1s - loss: 0.3879 - auc: 0.8339 - val_loss: 0.3769 - val_auc: 0.7817 - 715ms/epoch - 4ms/step\n",
      "Epoch 84/150\n",
      "163/163 - 1s - loss: 0.4001 - auc: 0.8186 - val_loss: 0.3846 - val_auc: 0.7759 - 699ms/epoch - 4ms/step\n",
      "Epoch 85/150\n",
      "163/163 - 1s - loss: 0.3938 - auc: 0.8286 - val_loss: 0.3659 - val_auc: 0.8027 - 725ms/epoch - 4ms/step\n",
      "Epoch 86/150\n",
      "163/163 - 1s - loss: 0.3907 - auc: 0.8309 - val_loss: 0.3815 - val_auc: 0.7883 - 665ms/epoch - 4ms/step\n",
      "Epoch 87/150\n",
      "163/163 - 1s - loss: 0.3975 - auc: 0.8169 - val_loss: 0.4037 - val_auc: 0.7525 - 731ms/epoch - 4ms/step\n",
      "Epoch 88/150\n",
      "163/163 - 1s - loss: 0.3859 - auc: 0.8325 - val_loss: 0.3691 - val_auc: 0.8178 - 722ms/epoch - 4ms/step\n",
      "Epoch 89/150\n",
      "163/163 - 1s - loss: 0.3905 - auc: 0.8298 - val_loss: 0.3625 - val_auc: 0.8198 - 705ms/epoch - 4ms/step\n",
      "Epoch 90/150\n",
      "163/163 - 1s - loss: 0.3839 - auc: 0.8362 - val_loss: 0.3664 - val_auc: 0.8111 - 655ms/epoch - 4ms/step\n",
      "Epoch 91/150\n",
      "163/163 - 1s - loss: 0.3811 - auc: 0.8355 - val_loss: 0.3651 - val_auc: 0.7946 - 707ms/epoch - 4ms/step\n",
      "Epoch 92/150\n",
      "163/163 - 1s - loss: 0.3881 - auc: 0.8275 - val_loss: 0.3753 - val_auc: 0.8143 - 652ms/epoch - 4ms/step\n",
      "Epoch 93/150\n",
      "163/163 - 1s - loss: 0.3857 - auc: 0.8340 - val_loss: 0.4005 - val_auc: 0.7931 - 712ms/epoch - 4ms/step\n",
      "Epoch 94/150\n",
      "163/163 - 1s - loss: 0.3858 - auc: 0.8341 - val_loss: 0.3722 - val_auc: 0.7918 - 651ms/epoch - 4ms/step\n",
      "Epoch 95/150\n",
      "163/163 - 1s - loss: 0.3903 - auc: 0.8316 - val_loss: 0.3958 - val_auc: 0.7688 - 674ms/epoch - 4ms/step\n",
      "Epoch 96/150\n",
      "163/163 - 1s - loss: 0.3936 - auc: 0.8289 - val_loss: 0.3737 - val_auc: 0.7984 - 700ms/epoch - 4ms/step\n",
      "Epoch 97/150\n",
      "163/163 - 1s - loss: 0.3860 - auc: 0.8329 - val_loss: 0.3786 - val_auc: 0.8028 - 709ms/epoch - 4ms/step\n",
      "Epoch 98/150\n",
      "163/163 - 1s - loss: 0.3750 - auc: 0.8452 - val_loss: 0.4008 - val_auc: 0.7940 - 679ms/epoch - 4ms/step\n",
      "Epoch 99/150\n",
      "163/163 - 1s - loss: 0.3782 - auc: 0.8466 - val_loss: 0.3723 - val_auc: 0.7917 - 666ms/epoch - 4ms/step\n",
      "Epoch 100/150\n",
      "163/163 - 1s - loss: 0.3824 - auc: 0.8370 - val_loss: 0.3923 - val_auc: 0.7761 - 722ms/epoch - 4ms/step\n",
      "Epoch 101/150\n",
      "163/163 - 1s - loss: 0.3717 - auc: 0.8481 - val_loss: 0.3711 - val_auc: 0.8154 - 671ms/epoch - 4ms/step\n",
      "Epoch 102/150\n",
      "163/163 - 1s - loss: 0.3828 - auc: 0.8431 - val_loss: 0.3891 - val_auc: 0.7987 - 699ms/epoch - 4ms/step\n",
      "Epoch 103/150\n",
      "163/163 - 1s - loss: 0.3771 - auc: 0.8415 - val_loss: 0.3715 - val_auc: 0.7947 - 676ms/epoch - 4ms/step\n",
      "Epoch 104/150\n",
      "163/163 - 1s - loss: 0.3710 - auc: 0.8497 - val_loss: 0.3583 - val_auc: 0.8140 - 723ms/epoch - 4ms/step\n",
      "Epoch 105/150\n",
      "163/163 - 1s - loss: 0.3684 - auc: 0.8557 - val_loss: 0.3765 - val_auc: 0.8016 - 693ms/epoch - 4ms/step\n",
      "Epoch 106/150\n",
      "163/163 - 1s - loss: 0.3728 - auc: 0.8484 - val_loss: 0.3860 - val_auc: 0.7976 - 1s/epoch - 7ms/step\n",
      "Epoch 107/150\n",
      "163/163 - 1s - loss: 0.3762 - auc: 0.8441 - val_loss: 0.3637 - val_auc: 0.8021 - 1s/epoch - 7ms/step\n",
      "Epoch 108/150\n",
      "163/163 - 1s - loss: 0.3754 - auc: 0.8452 - val_loss: 0.3708 - val_auc: 0.7819 - 1s/epoch - 8ms/step\n",
      "Epoch 109/150\n",
      "163/163 - 1s - loss: 0.3706 - auc: 0.8489 - val_loss: 0.3731 - val_auc: 0.8066 - 1s/epoch - 8ms/step\n",
      "Epoch 110/150\n",
      "163/163 - 1s - loss: 0.3760 - auc: 0.8429 - val_loss: 0.4146 - val_auc: 0.7762 - 1s/epoch - 8ms/step\n",
      "Epoch 111/150\n",
      "163/163 - 1s - loss: 0.3670 - auc: 0.8520 - val_loss: 0.3530 - val_auc: 0.8385 - 1s/epoch - 8ms/step\n",
      "Epoch 112/150\n",
      "163/163 - 1s - loss: 0.3683 - auc: 0.8528 - val_loss: 0.3421 - val_auc: 0.8357 - 1s/epoch - 7ms/step\n",
      "Epoch 113/150\n",
      "163/163 - 1s - loss: 0.3748 - auc: 0.8471 - val_loss: 0.3764 - val_auc: 0.8193 - 1s/epoch - 7ms/step\n",
      "Epoch 114/150\n",
      "163/163 - 1s - loss: 0.3701 - auc: 0.8524 - val_loss: 0.3666 - val_auc: 0.8141 - 823ms/epoch - 5ms/step\n",
      "Epoch 115/150\n",
      "163/163 - 1s - loss: 0.3680 - auc: 0.8550 - val_loss: 0.3949 - val_auc: 0.8158 - 744ms/epoch - 5ms/step\n",
      "Epoch 116/150\n",
      "163/163 - 1s - loss: 0.3594 - auc: 0.8565 - val_loss: 0.3742 - val_auc: 0.8117 - 768ms/epoch - 5ms/step\n",
      "Epoch 117/150\n",
      "163/163 - 1s - loss: 0.3701 - auc: 0.8502 - val_loss: 0.3608 - val_auc: 0.8101 - 795ms/epoch - 5ms/step\n",
      "Epoch 118/150\n",
      "163/163 - 1s - loss: 0.3708 - auc: 0.8505 - val_loss: 0.3510 - val_auc: 0.8247 - 752ms/epoch - 5ms/step\n",
      "Epoch 119/150\n",
      "163/163 - 1s - loss: 0.3604 - auc: 0.8576 - val_loss: 0.3943 - val_auc: 0.8290 - 682ms/epoch - 4ms/step\n",
      "Epoch 120/150\n",
      "163/163 - 1s - loss: 0.3580 - auc: 0.8603 - val_loss: 0.3951 - val_auc: 0.8056 - 762ms/epoch - 5ms/step\n",
      "Epoch 121/150\n",
      "163/163 - 1s - loss: 0.3565 - auc: 0.8613 - val_loss: 0.3798 - val_auc: 0.8313 - 712ms/epoch - 4ms/step\n",
      "Epoch 122/150\n",
      "163/163 - 1s - loss: 0.3537 - auc: 0.8647 - val_loss: 0.3710 - val_auc: 0.8147 - 715ms/epoch - 4ms/step\n",
      "Epoch 123/150\n",
      "163/163 - 1s - loss: 0.3626 - auc: 0.8591 - val_loss: 0.3610 - val_auc: 0.8348 - 655ms/epoch - 4ms/step\n",
      "Epoch 124/150\n",
      "163/163 - 1s - loss: 0.3555 - auc: 0.8650 - val_loss: 0.3626 - val_auc: 0.8006 - 703ms/epoch - 4ms/step\n",
      "Epoch 125/150\n",
      "163/163 - 1s - loss: 0.3633 - auc: 0.8566 - val_loss: 0.3441 - val_auc: 0.8461 - 795ms/epoch - 5ms/step\n",
      "Epoch 126/150\n",
      "163/163 - 1s - loss: 0.3531 - auc: 0.8658 - val_loss: 0.3389 - val_auc: 0.8450 - 905ms/epoch - 6ms/step\n",
      "Epoch 127/150\n",
      "163/163 - 1s - loss: 0.3587 - auc: 0.8580 - val_loss: 0.3603 - val_auc: 0.8209 - 755ms/epoch - 5ms/step\n",
      "Epoch 128/150\n",
      "163/163 - 1s - loss: 0.3487 - auc: 0.8664 - val_loss: 0.3637 - val_auc: 0.8312 - 860ms/epoch - 5ms/step\n",
      "Epoch 129/150\n",
      "163/163 - 1s - loss: 0.3502 - auc: 0.8671 - val_loss: 0.3303 - val_auc: 0.8586 - 762ms/epoch - 5ms/step\n",
      "Epoch 130/150\n",
      "163/163 - 1s - loss: 0.3542 - auc: 0.8636 - val_loss: 0.3561 - val_auc: 0.8188 - 730ms/epoch - 4ms/step\n",
      "Epoch 131/150\n",
      "163/163 - 1s - loss: 0.3597 - auc: 0.8619 - val_loss: 0.3645 - val_auc: 0.8098 - 668ms/epoch - 4ms/step\n",
      "Epoch 132/150\n",
      "163/163 - 1s - loss: 0.3574 - auc: 0.8631 - val_loss: 0.3427 - val_auc: 0.8433 - 767ms/epoch - 5ms/step\n",
      "Epoch 133/150\n",
      "163/163 - 1s - loss: 0.3509 - auc: 0.8698 - val_loss: 0.3810 - val_auc: 0.8207 - 723ms/epoch - 4ms/step\n",
      "Epoch 134/150\n",
      "163/163 - 1s - loss: 0.3397 - auc: 0.8771 - val_loss: 0.3928 - val_auc: 0.8385 - 742ms/epoch - 5ms/step\n",
      "Epoch 135/150\n",
      "163/163 - 1s - loss: 0.3455 - auc: 0.8690 - val_loss: 0.5027 - val_auc: 0.8113 - 747ms/epoch - 5ms/step\n",
      "Epoch 136/150\n",
      "163/163 - 1s - loss: 0.3626 - auc: 0.8626 - val_loss: 0.3730 - val_auc: 0.8285 - 681ms/epoch - 4ms/step\n",
      "Epoch 137/150\n",
      "163/163 - 1s - loss: 0.3481 - auc: 0.8704 - val_loss: 0.3416 - val_auc: 0.8442 - 657ms/epoch - 4ms/step\n",
      "Epoch 138/150\n",
      "163/163 - 1s - loss: 0.3429 - auc: 0.8764 - val_loss: 0.3545 - val_auc: 0.8521 - 677ms/epoch - 4ms/step\n",
      "Epoch 139/150\n",
      "163/163 - 1s - loss: 0.3406 - auc: 0.8748 - val_loss: 0.3529 - val_auc: 0.8516 - 723ms/epoch - 4ms/step\n",
      "Epoch 140/150\n",
      "163/163 - 1s - loss: 0.3473 - auc: 0.8657 - val_loss: 0.3553 - val_auc: 0.8224 - 782ms/epoch - 5ms/step\n",
      "Epoch 141/150\n",
      "163/163 - 1s - loss: 0.3470 - auc: 0.8739 - val_loss: 0.3749 - val_auc: 0.8380 - 659ms/epoch - 4ms/step\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 142/150\n",
      "163/163 - 1s - loss: 0.3543 - auc: 0.8652 - val_loss: 0.3586 - val_auc: 0.8337 - 653ms/epoch - 4ms/step\n",
      "Epoch 143/150\n",
      "163/163 - 1s - loss: 0.3390 - auc: 0.8793 - val_loss: 0.3936 - val_auc: 0.8217 - 753ms/epoch - 5ms/step\n",
      "Epoch 144/150\n",
      "163/163 - 1s - loss: 0.3297 - auc: 0.8859 - val_loss: 0.3592 - val_auc: 0.8408 - 667ms/epoch - 4ms/step\n",
      "Epoch 145/150\n",
      "163/163 - 1s - loss: 0.3356 - auc: 0.8828 - val_loss: 0.3349 - val_auc: 0.8543 - 666ms/epoch - 4ms/step\n",
      "Epoch 146/150\n",
      "163/163 - 1s - loss: 0.3386 - auc: 0.8807 - val_loss: 0.3446 - val_auc: 0.8238 - 736ms/epoch - 5ms/step\n",
      "Epoch 147/150\n",
      "163/163 - 1s - loss: 0.3380 - auc: 0.8779 - val_loss: 0.3531 - val_auc: 0.8524 - 1s/epoch - 7ms/step\n",
      "Epoch 148/150\n",
      "163/163 - 1s - loss: 0.3288 - auc: 0.8836 - val_loss: 0.3798 - val_auc: 0.8351 - 1s/epoch - 6ms/step\n",
      "Epoch 149/150\n",
      "163/163 - 1s - loss: 0.3305 - auc: 0.8828 - val_loss: 0.3661 - val_auc: 0.8196 - 1s/epoch - 7ms/step\n",
      "Epoch 150/150\n",
      "163/163 - 1s - loss: 0.3350 - auc: 0.8800 - val_loss: 0.3387 - val_auc: 0.8445 - 1s/epoch - 6ms/step\n"
     ]
    }
   ],
   "source": [
    "#mlflow.tensorflow.autolog()\n",
    "#with mlflow.start_run():\n",
    "model = Sequential()\n",
    "model.add(LSTM(32, activation='relu', input_shape=(X_train.shape[1], \n",
    "                                                                 X_train.shape[2]), return_sequences=False))\n",
    "#model.add(LSTM(32, return_sequences=False))\n",
    "model.add(Dropout(0.2))\n",
    "model.add(Dense(1, activation= 'sigmoid'))\n",
    "\n",
    "model.compile(optimizer='adam', loss=tf.keras.losses.BinaryCrossentropy(), \n",
    "              metrics = [tf.keras.metrics.AUC()])\n",
    "history = model.fit(X_train, y_train, epochs=150, validation_split= 0.2, verbose=2, batch_size=16)\n",
    "# fit the model\n",
    "# from keras.callbacks import EarlyStopping\n",
    "# early_stopping = EarlyStopping(monitor='val_auc', patience=3, verbose=1)\n",
    "# callbacks_list = [early_stopping]\n",
    "# history = model.fit(X_train, y_train, epochs=500, validation_split= 0.2,verbose=1,\n",
    "#                     batch_size = 16, callbacks=[callbacks_list]);\n",
    "\n",
    "#mlflow.log_metric('validation_auc',history.history['val_auc'][-1])\n",
    "  \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "7fcf4885",
   "metadata": {},
   "outputs": [],
   "source": [
    "# history.history.keys()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "575714c0",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7faa6fd65a90>"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAEICAYAAABPgw/pAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/MnkTPAAAACXBIWXMAAAsTAAALEwEAmpwYAABG8UlEQVR4nO3dd1zVZfvA8c/NYQ9RQRBFwY0TRTRnYjZcWY/maqkNs2zXUzZ+7Z6nnmw87cfKsiXasGFqjlLTNPcEJy5EAVGmbO7fH/cBDhsNBY7X+/U6Lzjfda5z0Ot7n+t73/dXaa0RQghR/znUdgBCCCFqhiR0IYSwE5LQhRDCTkhCF0IIOyEJXQgh7IQkdCGEsBOS0IXdUEqlK6Van+e+K5VSd9R0TNV87fOOWwhbktAvMdbEdUYp5VLO8jtKLYtQSsXaPFdKqfuVUruUUhlKqVil1DdKqa7nEUeJY5ezvq9SKlUpZbFZ9lEFyz4E0Fp7aq1jzjWW2lZX4lZKOSulvlVKHVZKaaVURKn1Sin1qlIqyfr4j1JK1U60ojyS0C8hSqlgYCCggVHncYj/Ag8A9wONgfbAD8CImomwhE2ABQizWTYQiCu17HJg9QV4/UvVGuBm4GQ566YC1wOhQDdgJHDXRYtMVEkS+qXlVmA98Bkw6Vx2VEq1A6YDE7XWv2mts7XWZ7XWX2mtX6lgnylKqWilVJpSKkYpdZd1uQewGGhmLTekK6Wa2e6rtc61xnq5dR8/wBmYV2pZe6wJ3dqqbGv9/TOl1HtKqV+sr/+XUqqNTWxXKaX2KKVSlFLvAspmnYNS6mml1BGlVIJS6nOllLd13Ryl1CPW35tbX/Me6/O2SqnT5bVaretWWV/vlFJqns06bV1v+3mkK6XOKqW0zXa3WT/PM0qpX5VSQdX521WX1jpHa/2W1noNkF/OJpOA17XWsVrr48DrwOSajEH8PZLQLy23Al9ZH9copfzPYd8hQKzWesM57JOAacU1AKYAbyqlwrTWGcAwIM5abvDUWseVs/9qrMnb+nON9WG77JDWuqLSzUTgeaARcAB4GUAp5Qt8BzwN+AIHgf42+022PgYDrQFP4F3rulVAhPX3QUCM9WdhPH/o8ufTeBFYao0lEHin9AZaa9vPwxNYAERaY74eeBIYDTQB/gDmVvC+UUolV/KYUdF+VegMbLd5vt26TNQRktAvEUqpAUAQMF9rvRmTxG48h0P4ACfO5TW11r9orQ9qYxUmoQ08h0OsAgZYW7wDMUlsHdDHZtmqSvb/Xmu9QWudhzmJdbcuHw5Eaa2/tX4TeIuSJYabgDe01jFa63TgCWCCUsrR+noDlVIOmAT+H4pPBoMqiScX8/k301pnWVvBFVJKPQ6EALdZF90F/FtrHW19P/8CulfUStdaN6zkUe43qmrwBFJsnqcAnlJHrzskoV86JgFLtdanrM+/pmTZJQ9wKrWPEyYRASQBAefygkqpYUqp9dYyRDImkfqewyHWY5JIF4pbv+nAMZtlldXPbZP0WeuxAJpZjwGAtUV9zGbbZsARm+dHAEfAX2t9EEjHnBwGAguBOKVUBypP6I9hyjoblFK7lVK3VbAdSqlhmGsV12utM62Lg4D/FraygdPW4zWv6DgXQDrm21ahBkB6Bd9IRC2QhH4JUEq5AeOAQUqpk0qpk8BDQKhSKtS62VEguNSurShObCuAQKVUeDVf0wVT1piJSYQNgUUU16qrTAJa6yxgI6ZsE6C13mNd9Yd1WTfO74LoCaCFTazK9jnmwqtty7cl5oQXb32+CrgBcLbWkldhylmNgG0VvJeTWus7tdbNMK3t9wvr/basJ4Y5wDitte1J5hhwV6mWtpvW+s/yXq9ULb7048kKP5nK7cZcEC0Ual0m6ghJ6JeG6zEXuTphWpbdgY6YxHirdZt5wBSlVG9r97T2mKQfCaC13g+8D8xVpsuhs1LKVSk1oYKarDPgAiQCedZW59U26+MBn8KLjZVYDTwI2CauNdZlJ60t5nP1C9BZKTXaWka5H2hqs34u8JBSqpVSyhNT3phnLXWASeD3UnwyWQncB6zRWpd3MRGl1FilVKD16RnMCS2/1DYNgB+Bp8spyXwIPKGU6mzd1lspNbaiN2hbiy/n8a+K9lNKuSilXK1PC//GhSfhz4GHrReDmwGPYC6wi7pCay0PO38ASzC9E0ovH4cpSzhan9+GaXGlYi4izgAcbLZXmFLAbkwJ4zjmRNC5gtedjkncycAXmJPDSzbrZ2NKOcmY2nJ5x7gGk/wetlnmb102t9S2Gmhr/f2zUq8VgbmoW/h8KLAPUwd+F5Ok77CucwCewbSKE4EvgUY2+3awvtYk63NvTAv+8Ur+Bv+xfl7pmOsXU0vHbY1RW7cpethsdwuw0/r3OQbMvgD/Vg5bY7B9BNv8/f+DKfectv6uavvftzyKH8r6hxJCCFHPSclFCCHsRJUJXSk12zq4YlcF65VS6m2l1AGl1A6lVFh52wkhhLiwqtNC/wxTb6zIMKCd9TEV+ODvhyWEEOJcVZnQtdarMRdAKnId8Lk21gMNlVLn1F9ZCCHE3+dYA8doTslBGbHWZWVGFSqlpmJa8Xh4ePQMCQmpgZcXQohLx+bNm09prZuUt64mEnp5w37L7TqjtZ4FzAIIDw/XmzZtqoGXF0KIS4dS6khF62qil0ssJUfZBWJG2gkhhLiIaiKh/wTcau3t0gdI0Vqf0yROQggh/r4qSy5KqbmYEWy+ytxh5lmskzhprT/EzM8xHDOy8CxmmlQhhBAXWZUJXWs9sYr1GjPE+2/Lzc0lNjaWrKysmjicuEBcXV0JDAzEyan05IxCiNpUExdFa0xsbCxeXl4EBwcjUyzXTVprkpKSiI2NpVWrVrUdjhDCRp0a+p+VlYWPj48k8zpMKYWPj498ixKiDqpTCR2QZF4PyN9IiLqpziV0IYQQ50cSuo2kpCS6d+9O9+7dadq0Kc2bNy96npOTU+m+mzZt4v7776/yNfr161dT4QohRAl16qJobfPx8WHbtm0APPfcc3h6evLoo48Wrc/Ly8PRsfyPLDw8nPDwqu/O9uef5d4xTAgh/jZpoVdh8uTJPPzwwwwePJjHH3+cDRs20K9fP3r06EG/fv3Yu3cvACtXrmTkyJGAORncdtttRERE0Lp1a95+++2i43l6ehZtHxERwQ033EBISAg33XRT4R1jWLRoESEhIQwYMID777+/6Li2Dh8+zMCBAwkLCyMsLKzoRGEbB8C9997LZ599BsDGjRvp168foaGh9O7dm7S0tJr/wIQQtabOttCf/3k3UXGpNXrMTs0a8Oy1nc95v3379rF8+XIsFgupqamsXr0aR0dHli9fzpNPPsl3331XZp89e/bw+++/k5aWRocOHbj77rvL9NveunUru3fvplmzZvTv35+1a9cSHh7OXXfdxerVq2nVqhUTJ5Y/DMDPz49ly5bh6urK/v37mThxIpXNjZOTk8P48eOZN28evXr1IjU1FTc3t3P+LIQQdVedTeh1ydixY7FYLACkpKQwadIk9u/fj1KK3NzccvcZMWIELi4uuLi44OfnR3x8PIGBgSW26d27d9Gy7t27c/jwYTw9PWndunVRH++JEycya9asMsfPzc3l3nvvZdu2bVgsFvbt21fpe9i7dy8BAQH06tULgAYNGpzbhyCEqPPqbEI/n5b0heLh4VH0+//93/8xePBgFixYwOHDh4mIiCh3HxcXl6LfLRYLeXl51dqmuvd4ffPNN/H392f79u0UFBTg6mpu1O7o6EhBQUHRdoX9xbXW0t1QCDsnNfRzlJKSQvPmzQGKatM1KSQkhJiYGA4fPgzAvHnzKowjICAABwcHvvjiC/Lz8wEICgoiKiqK7OxsUlJSWLFiRdFx4+Li2LhxIwBpaWnlnmSEEPWXJPRz9Nhjj/HEE0/Qv3//oiRak9zc3Hj//fcZOnQoAwYMwN/fH29v7zLb3XPPPcyZM4c+ffqwb9++om8RLVq0YNy4cXTr1o2bbrqJHj16AODs7My8efO47777CA0N5aqrrpLRnkLYGVXdr/g1rbwbXERHR9OxY8daiacuSU9Px9PTE60106dPp127djz00EO1HVYJ8rcSonYopTZrrcvtIy0t9Droo48+onv37nTu3JmUlBTuuuuu2g5JCFEP1NmLopeyhx56qM61yIUQdZ+00IUQwk5IQhdCCDshCV0IIeyEJHQhhDhPJ1OyeGv5PpZFxZOVW3435qzcfHLyCspdV9MkoduIiIjg119/LbHsrbfe4p577ql0n8Lul8OHDyc5ObnMNs899xwzZ86s9LV/+OEHoqKiip4/88wzLF++/ByiF0JcTIt2nuCat1bz1vL93Pn5JsJfWs6tszfw3+X72XL0DFm5+by9Yj+hzy+l/dOL6f7CUiZ/uoHP1x3meHLmBYlJernYmDhxIpGRkVxzzTVFyyIjI3nttdeqtf+iRYvO+7V/+OEHRo4cSadOnQB44YUXzvtYQoiytNZ8szmWxLRspg1qg8Wh7FQYUXGprNqXSCtfd7oFNqRZw/InsPtpexz3z91KaKA3M8eGEpeSxdLdJ9l85AxvrdjHm8v34WRR5OZrRnQNIKSpF3EpWaw7eIpnftzNsdNneWpEpwvzJmvj0bNnT11aVFRUmWUX06lTp7Svr6/OysrSWmt96NAh3aJFC11QUKCnTZume/bsqTt16qSfeeaZon0GDRqkN27cqLXWOigoSCcmJmqttX7ppZd0+/bt9ZAhQ/SECRP0a6+9prXWetasWTo8PFx369ZNjx49WmdkZOi1a9fqRo0a6eDgYB0aGqoPHDigJ02apL/55huttdbLly/X3bt31126dNFTpkwpii8oKEg/88wzukePHrpLly46Ojq6zHs6dOiQHjBggO7Ro4fu0aOHXrt2rdZa699//12PGDGiaLvp06frTz/9VGut9YYNG3Tfvn11t27ddK9evXRqamqZ49b230qIc5GSmaPv+WqzDnp8oQ56fKG+/bONOj0rt2h9Zk6efmVxtG79xC9F27R54he9cHuc1lrrT/6I0QNeXaHXHkjU8amZOvT5X/X1763ROXn5ZV7rTEa2/mFrrH56wU69am9CmfUHE9L0sdMZ5/1egE26grxad1voi2fAyZ01e8ymXWHYKxWu9vHxoXfv3ixZsoTrrruOyMhIxo8fj1KKl19+mcaNG5Ofn8+QIUPYsWMH3bp1K/c4mzdvJjIykq1bt5KXl0dYWBg9e/YEYPTo0dx5550APP3003zyySfcd999jBo1ipEjR3LDDTeUOFZWVhaTJ09mxYoVtG/fnltvvZUPPviABx98EABfX1+2bNnC+++/z8yZM/n4449L7C/T7Ir6rKBA42DTkk5My6ahuxNOFlMt3hGbzK7jqaRm5dLK14PBHfxwdixZSd4Rm8y9X2/leHImjw8Nwc3JgRcWRjHmgz95a0J3vN2cmPr5ZnYeT2FceCAPXdWehNRsXvolivvmbuG7LX78ticBd2cLk2dvpH1TTzJz8pk5NrQoDlsN3Z25rntzruvevNz31LqJZw1+QiXV3YReSwrLLoUJffbs2QDMnz+fWbNmkZeXx4kTJ4iKiqowof/xxx/84x//wN3dHYBRo0YVrdu1axdPP/00ycnJpKenlyjvlGfv3r20atWK9u3bAzBp0iTee++9ooQ+evRoAHr27Mn3339fZn+ZZlfUZVm5+SSkZnPwVDoLt59gfUwSw7s2ZfrgtnyzKZb/rtjP+F4teGp4R37fm8DdX26hU7MGzLqlJ0uj4nnmx10U2Mxe0sjdicvbN6FTQAOcLA4cTExn/qZjNPF0Yf5dfegZ1BiAYF8PHv1mO9e+swYvVyeyc/P56NZwrurkD0CAtxtzbuvNHXM28dueBO4Y0Ip7Brfljjkb2XI0maeGd6TNBUzM56vuJvRKWtIX0vXXX8/DDz/Mli1byMzMJCwsjEOHDjFz5kw2btxIo0aNmDx5cpUTW1U0Ve3kyZP54YcfCA0N5bPPPmPlypWVHkdXMddO4RS8FU3RK9PsiqpcqL/5mYwcsvMKaOrtWmJ5enYen687zG/RCWw5eqYoIXu5ONK9ZUM+XnOIT9ceJq9A097fk0/WHCL6RCobD58m2MeDvSfTuPKNVaRm5XFFiB8vXNcZ332RRGf5MjuuGRsOnebHbXEAeDhbuLpzU166rguNPJyLYojo4MevD17OMz/tZn98Gu/eGEZ7f68Scbo7O/LplF7sj0+nS3MzQd5Xd/RhXcwpBrX3q/HPqybU3YReSzw9PYmIiOC2224rultQamoqHh4eeHt7Ex8fz+LFiyucBx3g8ssvZ/LkycyYMYO8vDx+/vnnovlY0tLSCAgIIDc3l6+++qpoKl4vL69ybwkXEhLC4cOHOXDgAG3btuWLL75g0KBB1X4/KSkpBAYG4uDgwJw5c8qdZjcrK4sVK1YwYMCAEtPs9urVi7S0NNzc3Cq8l6qo337fk8CzP+2mvb8nM8eG0tDducp9zubkse1YMknpOfh4ONO3jQ9KKVLO5rJwZxyr9iay+cgZkjJyUAqm9GvFP6/pgJuzhegTqUz/agsxpzLo0rwB0wa1IdjXg2beboQHN8LVycL2Y8l89EcMw7sGMKxLUz5cFcOrS/bQo2VD5tzWm9jTmdwfuZWRoY15YVRnHLOT4dfH6NGyD+9MXgiYk0m+1vh4OFd4svLxdOG9G8NKLszPhfjd4OgKXv64uDUqSuYAbs4WrgjxL+dDOQ3L/g8ingRva6nl5E5oEgIWp7LbXyDyv7QcEydOZPTo0URGRgIQGhpKjx496Ny5M61bt6Z///6V7h8WFsb48ePp3r07QUFBDBw4sGjdiy++yGWXXUZQUBBdu3YtSuITJkzgzjvv5O233+bbb78t2t7V1ZVPP/2UsWPHkpeXR69evZg2bVq138s999zDmDFj+Oabbxg8eHC50+y2a9eu3Gl2MzMzcXNzY/ny5UX3QhV1X+kW918xSbT188THs/iGKpk5+Ty5YCcLth4nyMedVfsSGfnOGmaODaVPa5+i7TKy81i86yTuzhaGdWlKYlo2N378FwcS0ou2GRLiR5/WPry38gDJZ3Np3tCNK0L8aO/vxeGkDGavPcSP247j6erIieQsvN2diJzap8Tr2Apt0ZB3bRLt3RFtiOjQhGAfD9ycLXRq5sTyh20aNXsWQkEuxG6CvBxwdC7RGq+2tHiIvBGOW68xObnDg7vAo/w4i2gNPz8A0T+BdwuImAGn9sOHA6H/A3DV8+cey3mS6XPFeZG/Vd2SmJbNUwt2sjsuldMZOTw2tANT+rcicsNRZny/Ew9nC1Mvb8PI0AC8XByZ+sVmtscmc98V7Zg+uA1RcablHJeSRVjLhnQLbEjsmUzWHTxFRo75VtevjQ8nUrKIT83i1THd6NDUi9X7Enlj2T7O5uTTv60PM4Z2pEvzBiVOKH8ePEXkhmMoBY3cnZk+uC1NvFwqeitl5efCJ1dDTjo0DIIrnzUdHAp9fh3ErAI03L4cWvQqXpcaZxKud/kXKIsk7IEvx0DmabjqBSjIhyWPw3XvQY+bK99321z4YZpp1TfpAHethlWvwe8vgcUF7t8C3ja3n1w8A9oMhvaVXz+rSGXT50oLXYh6Li+/gPvnbmXL0TMMtbain/85ioOJ6URuOEb/tj54uTjx5nLTPxrA1cmBD27qydAuTQHo0bIRyx8ZxDebYpm99hDfbDpG80ZuDO8awPheLdgbn8Yri/YA8NUtIfTwToUmAbT39+La0GacTMmiW6B3ueWNfm186dfG9/zfYPxuiNsCgb0gdgMsfRpu/dGsS0+EQ6sh7BbY8jkcXVcyoc+/FTKTYfoGcHCAzDOQc7Zkgi/IhwVTIT8bpiyGZt3NSWDdexD9c3FCT08EzyYlY0uNg0X/hJb9oN1VsOJ5SImFqB/Btz2cOQy//wuuf99sf2o//PUBeDU974ReGUnoQtSgxLRs3J0teLhU/79WfoHm07WHWBoVj7ebE+39Pbnvina4OlnIys1nXUwSOXkFuDg6EBbUiAauJWuyby7fx7qYJGaODeWGnoHk5BVwz1db+HL9Udr5efLBzT1p4OrEvvg0dsamcOT0Wa7u5F+iNgzmIuCkfsFM6hdcpmwTHtyYoZ2bkpNfQMDKR2Hrl9CgOYTfhv/lj+LfoOSFz79Fa9j1nUl4Ll4mmQOM/gh2zIeV/zKJslEwRP0AugAuuxuO/GkSev/7zfYZSaYMg4ZDq6B1BMy9EVKOwQM7TIIH2PwZnNgOYz4xyRxAKeh4LWz8CLJSYcc8WPQotBoEAx4yLWyANW9CXiZc/545Max43pwI4nfCNf+CtBPw57vQ5x5o2gW2fQ3KAqETau7zslHnErr0sqj7aqtMV9elnM1l2H//oKm3Cz9OH4DFQZGVm8+mw2fYcCiJIB8PxvQ0X73/s2QPP26LY0BbXw6dymDD4dN0CmhAamYuy6Li+fNgEg8MaceLC6M4mJhR9BoWB0WPFg0Z2K4Jbfw8WLDlOCv2JDChVwtusB7b2dGB927qwRfrjjC8a0DRCaC9v1eZnhwVKe//YFEN/uQu0/r0CoDfXjQt59bVv1BPQYFJqo2CzPOsFFj2DAx6HBo0g31L4Lvb4crnTPI8vgXcGpsE3uMmWPlv2PoVRDwB2yOhSUfw7wQt+8CeReb4Dg4Q8zugwcEJNn4M+Tlw9E/zmkfXQXB/k/RXvADBA6HLmJJxdhoF698zJ6+V/wa/zpC4F764Hq7/ANpcAZvnQOhEaNza+iG1hb8+NL93HAUunmb/xY+bbxXbI6HtlaaFfgHUqYTu6upKUlISPj4+ktTrKK01SUlJRd0f7V1qVi6pmblYHBRnc/LJzMmnnb8nLo4W4lOzmPLpRkJbNOSl67vwypI9nErP5lR6Nl+sO8x13ZszYdZ69sabC98OClo38cDZ0YEPVx2krZ8ni3adAA2vjw1ldFhzlFIs3nmCh+ZvY/KnGwnwduXDm3sS2MiN1Kxc/jyQxB/7E3lrxT60Bl9PZ+4f0o57ItqUiNvF0cIdA1vX/AeiNSQdgO43mVrzu73g1ydN3djBUvm+Bfmw7StY+19zjCmLIagfbJ9nWslZqTD2U9PqBTj4m0nocVuhWQ/TavYONAlx21eQlWwuYI60bt+yr0mep/aBXwgcWA5ujaDHLbDuXUiINjX4jETY9a1J6L+/ZGrzw18zx7cV2Bs8/WHpU+DgCOO/MK//1Q2w8CHT4i/Ig4EPF+/TYTj8+TY07wkNW5hlQ56FhQ/CT/dBWhwM/fff/ztUoE4l9MDAQGJjY0lMTKztUEQlXF1dCQwMrHrDOuR8vvlFxaVyw4d/cjan5Cx6rZt48MSwjryyOJpjpzOJOpHKsdNnWXPgFLcPaMW++DRmLt3Hd1uOcygpg/9O6E7vVo0Z8/6fPPrNdhq4OdHI3ZlvpvXDPScJrSw4NyiuzQ7rGkCLxu4s3X2S2we2xtutuMTSr40vj17TgTMZOeyLT6N7y4a4OFaRSGtS2kmTAH3agpMrXP0CfDPZ1K/Dp1S+7+qZplwSEAou3qbVHNQPts8F5QC7vzfrjv0FXs3g6HrTgk6IhvZDi48TdoupjW+YBX3vhZ7W123Z1/w8us58gziwwrSie90Bf74Dpw/C6I9h7yJT4+51p2lhh98GfuVc4HdwgJARsGk29LsffKwnzTGzYdYg800i9Mbi1jlAyEiT0DtdZxPvJNj6BWz/2pxgOgw754+9uupUQndycqJVq1a1HYawI1pr3ly+n8/XHebVMd24pnP1vupm5+Xz8PxtuDs78uy1nSjQ4O5sITdf8+ayfdz5+SbcnCx8ecdlbDx8mtd+3UuAtysPX9WexLRsrn5rNVEnUvnfzT250jr68JUx3bh19gYAIvvG4v3hQ6b04OhmWn/trip6/S7NvcvUuG018nDmsgq6/ZVxOsa0TKtqQVdH0gHz07et+dnpenNB8NenIOOUSY7Jh82Fx6B+xa+ZsAdWv2bKGmM+gSUzTKI8ss7UyAc/ZZ4vfxbcfWDETNOF8K8PQOdDc5v+4u2HgW8HaHkZXPViccu6cWvTot4eCX6dICPBtOYbBZkEm3zEvL6zuzl5zJ1guiYOerzi99v7LlOqGfhI8TLPJubvtexZGPRYye1b9IYJc82JpJCDA4x4HWYNhm7jwfEcevicozrVbVGIqmRk5zF/0zHO5uRzS9+gMhcIDySk8f7vB9lxPIWrO/mTlJ7DvE3H8PFw5vTZHB65qj3XdG6Ko8WBBVuPsz4miTFhzbmhZwssDoqUzFyycvOZveYQ/1sdw8e3hhclZNsYPv7jEP3a+tAr2AwlX7TzBEE+7nRuZpLw73sScLI4MKBdyd4dry/dS8ypDN7Vr6DitpoLeDvmQ0IU/ON/0NU6l8/2efD7y3DXKtOqO1/pifBGR+gzDa5+qXh5fp4pHbQaWDJZVWXTbFNueHBXcUkh5Tgsfsz0B7fVsKVpnTbvaXp6JO2H6RtNQkyIhvf7mJZ4+kl4eA8cXAE/3G2Se7/74dUg0+0vO8WsbxBQfOzCOnlp276GH+4BlwZmv0f2mnp1Qb65eGpxgrxseK2dWT/4aRj0z+q//7/jxHZz0nGp3nWMilTWbbFaCV0pNRT4L2ABPtZav1JqvTfwJdAS0+qfqbX+tLJjSkIXpWXm5OPq5FBhaWT+xmO8vCialMxcABp7ODMqtBmnM3I4mZrFyZQsjp05i6ujhdAW3mw8fIb8As09EW2474p2PPrNdn7ZeaLoeEpBi0buHD19liAfd7Jy84lPzS5af0PPQGaODb0wb/adnqYVOf4Lc1Fw7kRTarjnL9OifDsMUo7Clc/DgAfP/3WifjTlCQcnuHdDcXlg+zzTVQ/gxvmVd6FLOW4G7jQKhiVPmqT+ZFzZhHp0velp4tvOtGo3fgJH1hav/8f/Svbu+OQaOLYe2l4FN39rkvS+xdBmiCnnfDHaJHmvAHhkT/Xf8475sGCauVA6bU352yx8GPYvhel/gbNH9Y9dB/ythK6UsgD7gKuAWGAjMFFrHWWzzZOAt9b6caVUE2Av0FRrnVPRcSWh26/EtGx8Pc2Q670n05j25WZ6tGzIw1e1J7CRe9F2BQWa3/cmMGfdEXYdT+F0Rg59W/vw8aRwnCwO/HtxNFrDY0M7sHJvItO/3sJlrRrz2NAQnC0OvPxLNJuPnqFpA1eaersS4O1KOz9PbrwsiMYeziSkZXEk6WxRK1przc7jKRw6lUFKZi5XhPjRvKEbi3ae5Ku/jtC0gSshAV54uDji4ezI0C5NcXUqVaY4vNbUW90bV+/DiN1kWoctLytelp8HL/tDv/tMTw6A9ASTxFsNNN3lfrjblA8cHOGB7RUPHz99yNSEu98Ebg3Lrl88AzZ/arrKtR1iTiAFBfD+ZebYDhZIPma+CTQKNvuc2G7KHoWDYT4caBL09L/gq3GQehzuXlv2tcqTnggJuyE73dSjbU/W2yNhwV2mBFP4zcTWn++aC5IdRsDEr6v3eoViN5lEXV5tHMxgpbxs0wulnvm7A4t6Awe01jHWg0UC1wFRNttowEuZppUncBooO1OUqHfyCzQHEtJp7++JUor8As3hpAxa+XiUmNa00PZjyYz+4E96BzdmWkQbHpm/nfyCAhbuOMHP2+Pw9XQhv0BToDXZuQWkZecR4O3KNZ398XRx5JM1h7jts404KMW6mCSUgt/2JHAyNYueLRvx2ZTeRUl27tQ+lV7s9PNyxc+ruDeOUopugWYUpK0R3QIY0S2g5M5ZKVBqGlbSE2HOSFNXtZ087vQh082u7ZUw+Mni5VrDt1Pg7BmYvr44QaYcNb0jGtv0TPH0g8sfgeXPmZauf1dzrMiJppVdOuHlnDU9J3Z+Y0oJ2+bCLQvKDnw5+qfpVtjqclPC2TQbnD1NT5AbPjUXIWdFwLu9zWskH4XDf4B/F9O6PbnDPABOHTBlE9tRmlXxbAKeEeWv6zrOnDjaDCl/fdshJqE371H91ysUWG6+K2ZxuqhzrFws1UnozYFjNs9jgctKbfMu8BMQB3gB47XWZW6ip5SaCkwFaNmy5fnEKy6w+NQsDiSk07+tqf1+uvYQL/0STd/WPozq3ozZaw6xPyGdpg1cGRziR5K1m96rY7rRzt+Ld347gJuThZ3HU5g0ewM+Hs58M60vHi6OfLr2MGcycrA4KBwcFI4OivDgxgzr0rRoXumugQ15MHIrjg4OvDk+lABvNx6Zv50Wjdz46NbwMi3mC9K9NW6rGWo+5hPTF7nQwRUmeR5cUbzs6Hpz8e5skpmMKezW4sQdu9EkSDB15xvnmxZqUoxZ5mOT0MEMjtn0qbl4N+J107OjcRtY+5Zprft1BA9rTX77XDPYpc890CzMdIn7dKhp9QcNMBcts1JNTJf/0/QG2bPQxAHg085cKHSwwJ2/m/7W2yPB1Ru63GC69R1YDvuXmXJNQS5ELYAzR6Dz6Jr5nB0cSlwILsOvI4z7wpyMRLVUJ6GX9z+mdJ3mGmAbcAXQBlimlPpDa51aYietZwGzwJRczjlacUHFJKZz08d/cSIli1X/jCDIx4Oft8cR4O3K3vg0nvh+J62bePD0iI6sj0nip23HaertSkJaNg9EbuPVMd1YHh3PA0PaMb5XCz7+4xDje7WgrZ+5CPTk8KrnfhkV2gw/Lxc8nB3pGmi9wPhoBAValy1/XAhaw6LHTIkhZmXJhL5/mfl5ap+pK7s3hq/Hm59jPoGvxsIfrxf3i975jZnfY+AjpnW8Yz6Ejjfd56BkCx1M3XjUO7BzvjXZOpheFAvuMt8MnD1h6iqTrKN/Nl0Hr/lXcf/sb28zk0SBGfji4WdOQC37mp4dd640g222zzUlmsIeKL5tTczX/Nss09p0/Vs9E07tNSWgpAOw4WPT48S33YX69Muy/fxFlaqT0GOBFjbPAzEtcVtTgFest0c6oJQ6BIQAG2okSlEtWms+WHWQNk08q909Lys3n8NJGeyMTeHVJXsp0BoHBd9tjmVcrxZsj01hxrAQbrysJVFxqYQHNcLR4lBi0MryqHju+HwTN3/yF+7OFqb0D6ahuzPPXHt+90wsPQtf6TvQXFA75pv5Qpy9zM9CBfmmZR7QHU5sM4nR4mIGt4z73IyUDLsFtnxhBsN4NYPdC8zFxoGPwN7FsPo/1oQeY5KzZzlzarceVHLUZegECOpveoVETjQlk0H/NGWRvvcW16SD+sLDUSbx/niv6UbY+XpTJ2/R22zj4GDKGG0rKHE42QwW6zvdDBgCMzozdpMZLQnmRCLqpOr8T9kItFNKtVJKOQMTMOUVW0eBIQBKKX+gAxBTk4GKqr26ZC//WbKXx7/bQXq2uYSxIzaZdQeTSmyXm1/AvxdF0+vl5YT83xKGvvUH//x2By6ODsy/qy8D2jXhuy3HWWTtETKsS1MauDrRp7UPjuXccuvKTv7ceFlLUjJzublPULXm1L6gstPN7HvZ6eWvT9wHn44wFyLBzGX99QTTwl4yw5Qw+kwzk0IVHuP4FjOxU7/7TPnj4G+w7UvTNS/YOj3ywEdMgl1wt1mXkQhdx5pWb9exJtmmxELSQdPbpLrlooYtoP3VpqW87SvY/YOpwXcs1XpVyrSeR75hrgFsmm1q5OfTiyNsErg2NPO1tB5sRkAWKl0qEnVGlS10rXWeUupe4FdMt8XZWuvdSqlp1vUfAi8CnymldmJKNI9rrU9dwLgvSVprvt5wlFa+HmVmr/tg5UE+XHWQiA5NWLk3kS/XH2FE1wBu+ugv0rLzuLKjP1P6B+NkceDNZWYyp2FdmtIxoAFBPu50CmhAK18PHC0OjO0ZyH1zt/Le7wfp3KwBQT5VJ4SnR5hbco0Nv4gjSHOzTI04KwW8/Isv1q3+jxlebnE2I/dGzyp5AWz9+3BkjWlBX3aXGWSybzE07WZm4Rv5hhmhqAvMoJdWl8OBZWY0Y5srTILb84sZMRkxo7j7nnegKV0seswc36WB6ZIHxa3umFWm5NK0/NsXVir8dhPzsmfNN4BmFVws9O9s3tf694tHT54rF0/zzaOwJ0zTrmau79zMv9cvXlxQ1RopqrVeBCwqtexDm9/jgKtrNjQBkJSeTWPrXVfWxSTx1IJdAAxq34Rpg9oQHtyIVxbv4ZM1h7g2tBlvje/OlM828tHqGJZHxaOB+65oyydrDrE8Oh4wJYw3xoUyOqz85HtVJ3+83ZxIycxl2oBAMz1p8MBKW5Tu6Ue5feMoaPaumePifOScNUnSpUHJr/8VWfxPM+QczGjLR6JNq3LXAtOzo0kHM7dH6ITiftY5GbDTegORPb+YxBe90JQR7lpd/B4zz5ifxzaYhL5/GTQPN/XyNoNhh7n5CaETS8bU42YzI9+K500SLHwffp3Ao4m50HjmCHT+x7l/PsEDzAjJU3uh27jyB9YUinjCfBv4O7P62ZZ+lDLHTI8//+OJC65ODf0XJa2PSeKmj/9iSr9gnh7ZibeW78e/gQu39W/Fe78fYOJH6/FwtpCRk8+U/sE8PaITFgfF/Ve05YYP15GUkcPrY0MZE9acSf2C2Xsyjdz8Alr7etLSx73C13V1snBd92Z8vu4IE87OhTlvmyHTo96p+Ov7xk/MMPYf74V71p37aLj8XPjfwOKh5WGTYNTbFW+fm2kSd8hIkxy/u93Uv5uHm26BETNMmWPPL2Z5YULfvQBy0kxd+shaMw3r4T9Mzdj2hOXWyMwHErvR9MuO2wpXPGXWFZ6wWl1ePGOgrYYtYMzHJZcpZbaPXmguLJa+IFodSkGv282ozKouFro2MH3Oa1KPm2r2eKLGSUKvJQUFutx+3IWSz+bw0LxtOCj4eM0hHBwUGw6d5rlrOzG5fytu7RvM8uh4lkXF06+NDxN6W7uB/jidcJ92jAkbjLuzhdFHXoCDWfiO+xzftuXMIaG1KS2Umufj4avaE9G2IY0WP2S+au/63tSeb1tcNlnnZpqWcNNupgSy7Jninh5gWrl+nSofxLE90iTz/g+YHiRb5piLem2uMN8QMhKh43Vgsf6T3bfEJObeU01Lct17Zsa+lGOmm13IcHB0Nsl+e6Sphbt4mha9b3szoOeTq+CXR009OuTasjEF9jaDdn5+wLSue91plns1haGvmulaz0XrCDPPN5x/HTr8dvPNo7BuL4SNi9h9QBT6ZccJery4jDX7S15myMsvYPORMyyLiufh+ds5lZ5N5NQ+dGnegFmrY/DzcilK3G4qh2u7NuXtiT2Kk3l+rmmN7l7A6+NCefH6LqiDv5uBKWcOlx/Mr0/B6x2st/Aq1tDdmSvUFvMVe/hMmDjXTNq/5s2yx9j9g+ntcc3LpqW7abZJwmCO+8lV8M0kM0LR1voPYN+vJu7Vr5keJFc+b2771bi1uRPMpk/NLca+vQ3e7WleC2DHN+DZ1JQhAHpONvOhbJxtEmdhnbfrOMg9a1rqsZvM8PqwW01L3tPf1MY9m5r5Rkpr0cvckixuq5ny1HYkZp9pxTdDqK5WNiWM82mhgzmhtY6o/gVVcUmRhH6RxSVnMuP7HaRk5nLf3C0cO32WY6fP8txPu+nz7xWM+eBP7vx8E7/tSeDxoSH0DGrMuxPDaObtyj+v6WD6YuflwNs9zFzOthL3mv7TiXtMN7uMU2bGOTDd6Uo7HQMb/mcuKn5xvRlqbTsVxObPTC+HtleaKT+7jjPbJB81x974iRkKv+kTU4MOHmgmVmrc2gx0yUo1Xd+c3E3teM0bJWNdMgO+HgdfjjaDaSKeMInKyRWGvWZa7AsfNIlw7BxTW/9mkolh/1IzsrHwm0WXMaYrYG5Gyfp0i8vAu6U5YcwZZeYFCb3R1J8LpzENGVF+PTrQ2t2v9eCyNz84H42CzPB6lwbFA4SEqEFScrkI1h44xf9Wx9CndWNW7U2kRcFx5jafwz9OTWPc/9ZxKj0bpRRXdvRjeNcAWjZ2p5G7My0amzp3sK8Ha2dcUTwq8vBqc2urv/5nutEVtkYLh2jnZZnh6KnHzXN3X1MSiXiiuGQBsPIVU56YtgZWPGeGWcdtMbXypIOma17EjOJ9hjxj7mz+3R1m/VmbbxiFA1yc3WHUu/DZcJh9jWk13zDb3Enm95dNr4vg/qb04eBkEuWOSNM6t50gqt2VcNk007oePtNMOdp+qOmLvdRay+46tnh7F08zNem2r0y5pZCDg0n8a94wF0rHfVF8F/fO/zAnrS4VjHz06wgj3jBd9mqqRRx+uzmRSgtbXACS0C+AdQeTWBYVz42XteRUeja3z9mIk8WB1fvMjTu+73EC7+htfNRtL9fvbMy48Bbcd0U7mnpX3LOjxBD3Pb+YLnk56aaVfPmjZvmJHcXbxO8ySR/MxbyFD8H+X01rFMxAlR3zzfStvm1NolvzprkdV/RCc8Nci0vJO543bGFOIKtfM/2bJ841fbkTo03Jo1Bwf1Pb3jDLtJA7j4Z2V8PxzfDTvTB1pZnmNGQ4jLZOGevTtmySG/ZqyedOrjDha5h3s2n9B5SaCfHqF6HP3WW71Q14EBq3KjsXdesIeHCn6Ute/oduLkLWpML7XQpxAch86DXsi/VHeO6n3eQXaFM9sDjQsrE786b2ITuvgJjEDPrvfha17UvwaYuevhGVnWpakP0frHoWv4ICM791y8vMhb6TO0xScnKDT4ebRJew28zfkXbCJOdH98NbXczFwFt/NInq6/GmXPLgjpKvGbMSon4yrdPWEWWHeefnmrp460GVT26UnW4mmup9p7mIV3jsz68zA3fitsDN35lyzvkoyK+ZGzYIUc/83dkWRTVorfn34j3MWh3DFSF+PHttJxb+uY2zR7cx6Zbbi26w26yhG/y+2wzYSDqAittibk+1+TOzbMgzlb/Q8c3mhgAhI01viznXmtZuzymmh0m3cabkEr/b3C7Mv7MpmfR/wNSsd39varj7lph7QpY+gbSOqLwfucXJlEOq4uJp7jpT+tihN5pbcXm3hNZXlLtrtUgyF6IMuSh6HhbvPMEby/ZR+O1Ga81zP+1m1uoYbu0bxEe3hhOUd4Tp+6fyz8Qn8Tu7v3jngnxzO67QCWbipl+fMsncyd30DsnJKP9FC+1ZaBJ/u6vNRchmYdb7JcZAdqrpOujfyZRcEqJNd0EwJZBmPczdxxc/bkocl919YT6gylzzsukG2efuygfGCCHOmfyPqkTy2RzSsnJLLMvLL+DZn3bz9or9fPXXUQoKNE8u2MWcdUe4c2Arnh/VGUvcZnNBsCDPXPjbHll8gDOHIS8TWvQxrexj66FBoKlhZ54xM+HZKsg3E/Fnp8G+paYfc/BA04VOKVMfPnMIfnvRbB/QDfw6m9fJzTDJHUyL9tq3zbwlpw/C0FdMP+2Lzb2xKRH1vefiv7YQdk5KLuVZPZMCDz/Grw4mOTOHr+/sQ5uzO2H1a8TpJiSkjaJ5QzdeWBjF6n2JLI2KZ/rgNjx6dQdUfq6524xbQ5iy2MzrsfMb07/a4mhazmBKIY2CTAlk+GtmBrxmYbDufeh5m2m9HvrDegEwuTg2J3fT86JQyEjTpznqB9Ny9+tkhnwX8utc/HtANxj+HzP0vLJ5qC806eEhxAVxaST0qB9Nq7Y6tw0ryIc1b5Hr4Mb+5DdwdHRk3fvTaKN/BhTNsBDsNYp59/Rj5DtrWBoVz0NXtuf+IW1NT5R175r5sm/8xkzWFDoe9v4Ch1aaC4DxUYCCJiGmi99jMcW9MvpON0PYIyeabZc+bfot97/f7BMQaoas285z4mAx639+AJp0NL04/G2SuF9IyffX646/9VEKIeou+0/oGafMTXLbDDG9KippHWZk5+F8Zh9OOWm4kMZQr0M8fOMoWs5ZzCLdH7ewsQze+iD3t0vAv4ErX9zem4MJGcW3Lztz2HTpCxlppjsF03fa1dvclLftlaYHSuPWJplDyS52nUeb8smf75qLlk27wS0/FPebrki3CbDqteKh6A2DwcnD7Pc37zAuhKg/7D+hn7XOBX5whelNEnZruZulZeUy9K0/GKV/53EgXyumN9lB2ziAPH5pdDPL17mxw8WJq1yjAQhp2oCQpg1MV8G1/4W/PgSUqU8XcnQxA1h2zIfUE6b3iX+X8mN1cDDdDS+bZobEt7vKnAyq4uQKd68xMw4WHieon5l/RAhxybD/i6KZyeanW2NY8iQc21hyeLvV60v3EZeSSeucvaRpN5bTm45nfjM9UFr25fXp47mxf3tOeHfHK3ZN8Y6FNwL+Y6ZJwFNXmgE4tvrcAyiYO8GM4LQtiZTHxcsMtqlOMi/k1qhkKWZipBnxKYS4ZNh/Qi+8oDj8NdNy/eRKeCfMTLhkte1YMnPWHebWPkH8w+8kJzw6Ygkdh0NmkukR0nMKrk4Wnr22M8G9hpuySZp1Xugd88w8JUNfhbGfQZP2ZWNo0sHcZOHENkBXndBrgsWx5DB/IYTds/+EXthCD+gO92+Da99GpycQs+Rd7pizkWveXM0tH/+Fv5crjw4JwjFxN+3DBnHltTeZyZ5cG5ace7rNYPPz0Coz7H3JDDO8vfedlcfRcaTp6eLgZHqzCCFEDbP/JlxhC92tIbg3JrH9BI7+/CEc3k60exqdmjVgTMN9XN47DK8ze0zf8eY9zVD6a/5lBv84uRUfr2moKW/88Tos/T8zEGjUO9UbuTjgQTPAp/CCqBBC1CD7T+iFLXRrPfqlX6LolR/IBNc/WfNYBEoXwCujYLFn8ex9hXNj95xU9ngODqbnyo555n6R/e4rnqukOiSZCyEuELtP6DozmVyLOyuiTuHqZOHHbXFc26kHjjFLIS3OzAWem2Ee694182U3aFb5QUe8AVe/JHNaCyHqFLuvoccnxpOY58bdX21hymcbaeXrwcD+l5uVCdFmsiswt0xzcCz/zjWlObtLMhdC1Dl230I/cfIEHsqTTyaFs2TXSW7qE4SLj/VWaAnRpheLa0MzW2HTUPCUvttCiPrJrhP68eRMctJP4+vdmCEd/RnS0b94pWdTk9BP7oTmYWYEaWA1WudCCFFH2XXJZe5fR/EmA98mfmVX+nU05ZaEqOqVWYQQoo6z24SelJ5N5MZj+Dll4dagnHq3Xyc4tRd0viR0IYRdsMuEvuHQaYa//QepWbl4qwxTIy/NdhZCGegjhLADdpfQ98enceNH63FzsrBgWi8seWfNoKLSCu/k490CvPzLrhdCiHrG7i6KfvRHDI4Wxbd398OXVLOwvBZ64WCg5tI6F0LYB7tK6Ilp2fywNY5xvQLx9XSBUylmRXktdBcvuPwxaHX5RY1RCCEuFLtK6F+sO0xuQQG3D2htFhTO41JeCx3giqcuRlhCCHFR2E0NPSs3ny/WH+HKjv608vUwC4vmQm9YW2EJIcRFYzcJffORM5w5m8uNvVsWLyxqoZ/DjSKEEKKespuEfiAhHYBOzRoUL8w8Y35WVHIRQgg7YjcJ/WBiOl4ujvh5uRQvtJ0LXQgh7Fy1ErpSaqhSaq9S6oBSakYF20QopbYppXYrpVbVbJhVO5CQThs/T5RSxQszk82Nkx1dKtxPCCHsRZW9XJRSFuA94CogFtiolPpJax1ls01D4H1gqNb6qFKqnMlTLqwDCekMbFdqpsSsZGmdCyEuGdVpofcGDmitY7TWOUAkcF2pbW4EvtdaHwXQWifUbJiVS83KJSEtm7Z+niVXZKVI/VwIccmoTkJvDhyzeR5rXWarPdBIKbVSKbVZKXVreQdSSk1VSm1SSm1KTEw8v4jLcdB6QbRNE4+SKzKTpYUuhLhkVCehq3KW6VLPHYGewAjgGuD/lFLty+yk9SytdbjWOrxJk5q7kURhD5eyLfRkaaELIS4Z1UnosUALm+eBQFw52yzRWmdorU8Bq4HQmgmxagcTM3CyKFo2LnUD5swU6YMuhLhkVCehbwTaKaVaKaWcgQnAT6W2+REYqJRyVEq5A5cB0TUbasUOJKQT7OOBo6XU25GLokKIS0iVvVy01nlKqXuBXwELMFtrvVspNc26/kOtdbRSagmwAygAPtZa77qQgds6mJhOSFOvkgsL8iE7VUouQohLRrUm59JaLwIWlVr2YannrwGv1Vxo1ZOdl8/R02cZ0TWg5Ir0ePNTWuhCiEtEvR8peiTpLPkFuuQF0b9mwQf9ze9Nu9VOYEIIcZHV+4Qee+YsAC0KL4ie3AmL/wn+nWHqSgjuX3vBCSHERVTv50NPTMsGKJ7D5aS1dD/yTfBtV0tRCSHExVfvW+iFCb1JYUJPiAKLMzRqVYtRCSHExWcXCb2BqyOuThbrgj3g2wEs9f7LhxBCnJP6n9DTs4tb5wAJ0eAXUnsBCSFELan/CT3NJqFnpULKMfDrWLtBCSFELbCThO5qfbLX/PTrVHsBCSFELbGPhO5pc0EUoImUXIQQl556ndAzsvPIyMm36eESDU7u0DCodgMTQohaUK8T+qn0Ul0WE6OhSQdwqNdvSwghzku9znxl+6BHS/1cCHHJso+E7ukCZ0+bCbmkh4sQ4hJVvxO6bcll72KzMOCi3VdDCCHqlPqd0NOysTgoGrsqWP0fM7Ni8MDaDksIIWpFvR4fn5iWjY+HM5ad8+DMYZgYCaq8W6AKIYT9q/ct9ABPB9M6b9YD2g+t7ZCEEKLW1O+Enp5NX+cDkHwUBjwkrXMhxCWtXif0hNRsghxTzBPpriiEuMTV24ReUKA5lZ5NgCXZLPD0r9V4hBCittW/hJ50EP58l+Tk0+QVaHxJNsP9XbxqOzIhhKhV9a+XS0IULH2KVK8wABoVnDatc6mfCyEucfWvhe4dCEB6wmEAvHKTwKtpLQYkhBB1Qz1M6C0BOHFkH04WhUfuKamfCyEE9TGhuzdGO7pxJu4g/dr4YklPkBa6EEJQHxO6UuR4Nscz+yTDOnhBTpq00IUQgvp4URQ4qZrQXMVzWaB1gbTQhRCiHrbQgeizDQiynMaHM2aBtNCFEKL+JfTjyZnsTPemoU42E3KBtNCFEIJ6mNDXH0ziuPY1T45vNj89JaELIUS9S+hjegby5MQrzZPYjeDgBO6NazcoIYSoA+pdQgfwa9HO/BK/W0aJCiGEVb1M6HgFgHKAgjzwkguiQggB9TWhW5zAq5n5XXq4CCEEUF8TOkDDFuanJHQhhACqmdCVUkOVUnuVUgeUUjMq2a6XUipfKXVDzYVYAeskXdJlUQghjCoTulLKArwHDAM6AROVUmVuD2Td7lXg15oOslze0kIXQghb1Wmh9wYOaK1jtNY5QCRwXTnb3Qd8ByTUYHwVkxa6EEKUUJ2E3hw4ZvM81rqsiFKqOfAP4MPKDqSUmqqU2qSU2pSYmHiusZYU0B0cHMG3/d87jhBC2InqJPTyOnnrUs/fAh7XWudXdiCt9SytdbjWOrxJkybVDLECgT3hiVjwafP3jiOEEHaiOrMtxgItbJ4HAnGltgkHIpUZ4OMLDFdK5Wmtf6iJICvk5HZBDy+EEPVJdRL6RqCdUqoVcByYANxou4HWulXh70qpz4CFFzyZCyGEKKHKhK61zlNK3YvpvWIBZmutdyulplnXV1o3F0IIcXFU6wYXWutFwKJSy8pN5FrryX8/LCGEEOeq/o4UFUIIUYIkdCGEsBOS0IUQwk5IQhdCCDshCV0IIeyEJHQhhLATktCFEMJOSEIXQgg7IQldCCHshCR0IYSwE5LQhRDCTkhCF0IIOyEJXQgh7IQkdCGEsBOS0IUQwk5IQhdCCDshCV0IIeyEJHQhhLATktCFEMJOSEIXQgg7IQldCCHshCR0IYSwE5LQhRDCTkhCF0IIOyEJXQgh7IQkdCGEsBOS0IUQwk5IQhdCCDshCV0IIeyEJHQhhLATktCFEMJOSEIXQgg7IQldCCHshCR0IYSwE9VK6EqpoUqpvUqpA0qpGeWsv0kptcP6+FMpFVrzoQohhKhMlQldKWUB3gOGAZ2AiUqpTqU2OwQM0lp3A14EZtV0oEIIISpXnRZ6b+CA1jpGa50DRALX2W6gtf5Ta33G+nQ9EFizYQohhKhKdRJ6c+CYzfNY67KK3A4sLm+FUmqqUmqTUmpTYmJi9aMUQghRpeokdFXOMl3uhkoNxiT0x8tbr7WepbUO11qHN2nSpPpRCiGEqJJjNbaJBVrYPA8E4kpvpJTqBnwMDNNaJ9VMeEIIIaqrOi30jUA7pVQrpZQzMAH4yXYDpVRL4HvgFq31vpoPUwghRFWqbKFrrfOUUvcCvwIWYLbWerdSapp1/YfAM4AP8L5SCiBPax1+4cIWQghRmtK63HL4BRceHq43bdpUK68thBD1lVJqc0UNZhkpKoQQdkISuhBC2AlJ6EIIYSckoQshhJ2QhC6EEHZCEroQQtgJSehCCGEnJKELIYSdkIQuhBB2QhK6EELYCUnoQghhJyShCyGEnZCELoQQdkISuhBC2AlJ6EIIYSckoQshhJ2QhC6EEHZCEroQQtgJSehCCGEnJKELIYSdkIQuhBB2QhK6EELYCUnoQghhJyShCyGEnZCELoQQdkISuhBC2AlJ6EIIYSckoQshhJ2QhC6EEHZCEroQQtgJSehCCGEnJKELIYSdkIQuhBB2QhK6EELYCUnoQghhJ6qV0JVSQ5VSe5VSB5RSM8pZr5RSb1vX71BKhdV8qEIIISpTZUJXSlmA94BhQCdgolKqU6nNhgHtrI+pwAc1HKcQQogqVKeF3hs4oLWO0VrnAJHAdaW2uQ74XBvrgYZKqYAajlUIIUQlHKuxTXPgmM3zWOCyamzTHDhhu5FSaiqmBQ+QrpTae07RFvMFTp3nvheLxFgzJMaaITH+fXUlvqCKVlQnoatylunz2Aat9SxgVjVes/KAlNqktQ7/u8e5kCTGmiEx1gyJ8e+r6/FB9UousUALm+eBQNx5bCOEEOICqk5C3wi0U0q1Uko5AxOAn0pt8xNwq7W3Sx8gRWt9ovSBhBBCXDhVlly01nlKqXuBXwELMFtrvVspNc26/kNgETAcOACcBaZcuJCBGijbXAQSY82QGGuGxPj31fX4UFqXKXULIYSoh2SkqBBC2AlJ6EIIYSfqXUKvahqC2qCUaqGU+l0pFa2U2q2UesC6vLFSaplSar/1Z6NajtOilNqqlFpYR+NrqJT6Vim1x/pZ9q2DMT5k/RvvUkrNVUq51naMSqnZSqkEpdQum2UVxqSUesL6/2evUuqaWozxNevfeodSaoFSqmFdi9Fm3aNKKa2U8q3NGKtSrxJ6NachqA15wCNa645AH2C6Na4ZwAqtdTtghfV5bXoAiLZ5Xtfi+y+wRGsdAoRiYq0zMSqlmgP3A+Fa6y6YTgIT6kCMnwFDSy0rNybrv8sJQGfrPu9b/1/VRozLgC5a627APuCJOhgjSqkWwFXAUZtltRVjpepVQqd60xBcdFrrE1rrLdbf0zCJqDkmtjnWzeYA19dKgIBSKhAYAXxss7guxdcAuBz4BEBrnaO1TqYOxWjlCLgppRwBd8x4i1qNUWu9GjhdanFFMV0HRGqts7XWhzA903rXRoxa66Va6zzr0/WY8St1KkarN4HHKDlYslZirEp9S+gVTTFQZyilgoEewF+Af2F/fOtPv1oM7S3MP8oCm2V1Kb7WQCLwqbUs9LFSyqMuxai1Pg7MxLTUTmDGWyytSzHaqCimuvp/6DZgsfX3OhOjUmoUcFxrvb3UqjoTo636ltCrNcVAbVFKeQLfAQ9qrVNrO55CSqmRQILWenNtx1IJRyAM+EBr3QPIoPZLQCVY69DXAa2AZoCHUurm2o3qnNW5/0NKqacwZcuvCheVs9lFj1Ep5Q48BTxT3upyltV6LqpvCb3OTjGglHLCJPOvtNbfWxfHF846af2ZUEvh9QdGKaUOY8pUVyilvqxD8YH528Zqrf+yPv8Wk+DrUoxXAoe01ola61zge6BfHYuxUEUx1an/Q0qpScBI4CZdPCimrsTYBnPy3m79vxMIbFFKNaXuxFhCfUvo1ZmG4KJTSilM7Tdaa/2GzaqfgEnW3ycBP17s2AC01k9orQO11sGYz+w3rfXNdSU+AK31SeCYUqqDddEQIIo6FCOm1NJHKeVu/ZsPwVwvqUsxFqoopp+ACUopF6VUK8w9DDbUQnwopYYCjwOjtNZnbVbViRi11ju11n5a62Dr/51YIMz6b7VOxFiG1rpePTBTDOwDDgJP1XY81pgGYL5u7QC2WR/DAR9MD4P91p+N60CsEcBC6+91Kj6gO7DJ+jn+ADSqgzE+D+wBdgFfAC61HSMwF1PTz8UkndsriwlTRjgI7AWG1WKMBzB16ML/Mx/WtRhLrT8M+NZmjFU9ZOi/EELYifpWchFCCFEBSehCCGEnJKELIYSdkIQuhBB2QhK6EELYCUnoQghhJyShCyGEnfh/oID5zIOFF8YAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history.history['auc'], label='Training auc')\n",
    "plt.plot(history.history['val_auc'], label='Validation auc')\n",
    "plt.ylim([0.0, 1.0])\n",
    "plt.title('AUC at Window size = 10')\n",
    "plt.legend()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4922762f",
   "metadata": {},
   "outputs": [],
   "source": [
    "#--------------the end------------"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.7"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}