HyDRA_example.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "8fcb284e",
   "metadata": {},
   "source": [
    "# HyDRA\n",
    "\n",
    "(Hybrid Deep learning models for Rapid Assesments)\n",
    "\n",
    "This notebook provides an example on how to train and evaluate a HyDRA model.  "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "34003b3d",
   "metadata": {},
   "source": [
    "### Libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "92ea64f8",
   "metadata": {},
   "outputs": [],
   "source": [
    "from utils.models import build_hydra,mae_loss,agg_Y,eval_r\n",
    "from utils.DataGenerators import DataGenerator\n",
    "import matplotlib.pyplot as plt\n",
    "import time\n",
    "import multiprocessing\n",
    "import numpy as np\n",
    "from tensorflow.keras import optimizers, metrics\n",
    "from tensorflow.keras.callbacks import EarlyStopping, ModelCheckpoint\n",
    "from tensorflow.keras.models import load_model"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4c32ee11",
   "metadata": {},
   "source": [
    "### Data generators"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "d5938ef4",
   "metadata": {},
   "outputs": [],
   "source": [
    "features='MW' # Put the desired feature set\n",
    "Data_dir='../Data' # Replace with the path of the main dataset directory for all folds\n",
    "fold_id=1          # A value between 1-6"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "3fffad53",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_generator=DataGenerator(Data_dir,'train',fold_id,features,shuffle=True)\n",
    "val_generator=DataGenerator(Data_dir,'val',fold_id,features,shuffle=False)\n",
    "test_generator=DataGenerator(Data_dir,'test',fold_id,features,shuffle=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bb12bbb9",
   "metadata": {},
   "source": [
    "Checking the generator data:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "5c24c855",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Input tesnors: \n",
      "(1436, 400, 48)\n",
      "(1436, 400, 48)\n",
      "Label tensors: \n",
      "(1436, 1)\n",
      "n_batch: 215\n"
     ]
    }
   ],
   "source": [
    "sample_input,sample_label = val_generator.__getitem__(0)\n",
    "print('Input tesnors: ')\n",
    "for X_i in sample_input:\n",
    "    print(X_i.shape)\n",
    "print('Label tensors: ')\n",
    "print(sample_label.shape)\n",
    "print('n_batch: %d'%(train_generator.__len__()))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "db4531c2",
   "metadata": {},
   "source": [
    "### Visualizing sample MFB and W feature tesnors"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "df9969e6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAl4AAAE+CAYAAABcPR5BAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8rg+JYAAAACXBIWXMAAAsTAAALEwEAmpwYAABFr0lEQVR4nO3de5hVZdkG8PthOI2n0Wg0wgOkRmoUCZrnNI95xBOfdtI8ZVmeMk+ZmlmSmqmfpVGamKVhmuIhzUwFzRMoiqCGiobEp0SJmgjMzPP9sTY5zHOv2e8Ms9eePXP/rssLeFxrr3evvfc7a/Z+7/2Yu0NEREREKq9PtQcgIiIi0lvowktERESkILrwEhERESmILrxERERECqILLxEREZGC6MJLREREpCBVufAysz3M7AUze9HMTq/GGEREOktzmIh0lhX9PV5mVgfgbwB2BfAagCcAHOrus3L3+WA/x9CBKxZfGR62W21h3PedjcgNNrwea/+3TqzlXZb2J7XVSI3dow+RXefGGh/3e7H25MBQ+jB5SP8xKNb6kfO1rJEcFwDW/3usLVg/lDYimzVgcahNW7ueHJwcdy1SWz2WRj39H7LhUlKJNzhjXbJrHDIAYCA5Z6uQ7ZpI7a1R5Hkn+aa99k93z3tGVk1H5zD74AccQ4e0qTaTLVmNTUJLErfLE+cMwEitL6mRZ3bLgFh7h+y6BntVsBcamRvoWBg2DwD8/LAa2//dWHqTTBprvkn2XZPU2ONH5vaWhljrs4Dsyx5P9kMKAFpILc6TADk29Wbidr3QtJdy56/UZ3NX2hLAi+7+MgCY2Y0A9gO/TMkMHQhM3XzF2pEPhs1GXhN3fegycnt7XhxrF5wSa6vmjCdebwDbktqnSI0cZuQJscbH/UIo1dXHC9Cvkdfwd/eLtQ+S8zX/QHJcALjya7F21ZWhdAnZbB/MDDU7ZLO44f+R47Lx7BRLU9d+nGwYrwLn4qBQW5+cfzxHagA2JOfsk2S7f5Ha3VPJ807y2bdfrfYQcnRsDhs6BJg6qU2RPUPY1Qr7AfoSqbHf/PJsknicD5AaGffiYbH2MNl1F3bRwE7ZprHUnHj9XfdIzv9g54fV2MCfjqVbL4q1MbfGWvOYWKubQ45BJpzFe8Za/VVkX/Z4sh9SAH+Okd/8m8mxmbpb07brjWz/3PmrGh81DsGKj/RrpZqISC3QHCYinVaNCy/2nnb4cMzMjjGzqWY2FQvYZ1AiIlVRdg5bcf5i726JSG9VjQuv1wCs1+rf6wL4R9uN3H28u49299Fo7FfY4EREyig7h604f7GP7ESkt6rGhdcTADY2s2Fm1h/AIQDaLoAQEemuNIeJSKcVvrje3ZvM7BsA7gFQB+Aad4+rr1vZ3N/Bo0snr1Drf/WaYbuHrmY3E1eVH4NpoTb+DBIjnDOmvWGtqDEu9sdrvyQbPhEqDx0/hWz3IqnFBZN92UL6z5Bdd4+l+SxkdyKpAbgHcVHn7sduEWr7Hht/uz/cxsQbPIochIVrRrEF6eeHyi9JuuZhbBlq1/4w3tq+p8RPjiZhFDkuMPPqm2KNbtmRxc5SSzo8hzUNABa0WYDO3gVbQJJkjSwJyBbCk7g0bubjWTQm7Sbr2WJ4tt2iWFuLpeISwwOLyEL6BnIeXiA/voaz4AHAF6DHuZhHZdaLpTHscSHHYNmBEWyiI7V6kqak53AkqbGkEkAX0qdih04NP8oKqpFqhLvfBeCuahxbRGRlaQ4Tkc7SN9eLiIiIFEQXXiIiIiIF0YWXiIiISEGqssaroxYZcHtYg8l66hweKr8gC+mPxl9D7VO2dag91YFmJQMWxBXtS84lq9zPIQsmJ8emM7ZXPPgUsriRNYl76IF7Y3HerrF2TCxt+TFygwB2v5P0ISLrN08+Ii5Uv+SZuK9tE/f1MeTAc8lX/S+NNbaM+G5yl3HG9aHEG/mQ3kcA+Blnq07Pytlfep2+LUBj29f99Lhd40iyM1udzVbCk1dA8/F8PA3zYu0F8v2vw0nAZ3GcJ+nievLl8/w1ReZxuoibLD4fztr7sMXxQH4LnbbIt9TT2/wTqZFF+GyIE2O3EYydQDaM4SXgUFK7ndTywj0b5tTbqCPPkf8jzxEtru8UveMlIiIiUhBdeImIiIgURBdeIiIiIgXRhZeIiIhIQXThJSIiIlKQmkg1cn8ktZNDhTWKYAmXp+4km+15Zs6xYxJnycQj4mbPsX1viaUddgolvygmSHb+Wtx1ice2HoOwW6htSwIpk958I9QeX5AT5WT35YiY0LwEI+J2IyaGkp85Nm5HeqEPiiFE2hzlbtIKCPG0Ang4VB7DGWQ78jgB6IP4RGkBSY/NI+mvIfHY0hu0IMbbWOPsx0mNxQPXIrV/JZUAAI0kuTectZhhCT/yvF5Mom31l8caS1nWkUQkDeSRCWgRaeuVm7Jj94Vh6Uf2WLHbIynErWbE2ifJHIntSI2ciGZyB+vYYxfby2VIqrE5pupRR8Y9nD3vpDP0jpeIiIhIQXThJSIiIlIQXXiJiIiIFEQXXiIiIiIF0YWXiIiISEFqItU4B6wLI+ndRXpqnYMYi/s8Rofav/ZMH8/dLNlDQnq8hxZrRLZOLB0ba0sOZA9XTJosXBB7I97TGHso0r6DjU+Q7QA0nhdrM8bE2n+eibX7ye2dwWKIMUm48JTfxNr+pNcZOYXYak6svXBlrM0l++7CeoEC38ORodYfF4Taaf8i948kS6U3cMT5is1fJK87jzxpWPCukezbGBPPGRIfphFIcpv1TWS7f5Maef2wqWUr0r+R9WVkGl4gxZdyNiZ9J2m/X5bcY2lk1r+R3UFyX+rZdiyrTc5NHYt8sp6OfP6i6dB3SDq0P0le1sd0enoPTGlN73iJiIiIFEQXXiIiIiIF0YWXiIiISEF04SUiIiJSkJpYXM9NILW4cP0+0sLm+4jtENj6arYMPsMWYLLFmwfG0ne+GGsnkl1ZFwi2rnICeQiXxdKSo+KCe7oOdSu26B1A85hQ2uwTcbPnB5Jd92U3eGGoDEJsH7KQrcz/Axs4W5BLwg3DSUhg+M1k37NJjS+BZTWMuIbuL73RMsQXNFucTRapDyGtW2g7GLYAfNuc8bB2N7uTGpsB2UL6e0htvVhia8DpWMi+GElqeT2RmNT9WXsgMqHSfdMCX3TRPH1MWTgrdSF93rkhP+n6k8X19Wz+SgyESFl6x0tERESkILrwEhERESmILrxERERECqILLxEREZGC6MJLREREpCA1kWpcH8D32tQOwx1kyz+T2lmhciNWDbVLyJ4su5i5K1SWktqmZM+bf/ClUGM5mimNl5HqdaGy+inTQo2FH+fj5VhsjEfuA55qbFl6ZqjNvJNsuCdLwxxCajFduBDnk+1ii4t38LVQm072jLlJ4I8kDbsMV5AtbyI14Ggyxj6IjwF/VDuSwpKeownhsV9Ept4GlowmLV5oBnsTUstr58LSiqltcVjcesO0Y9SxFj8sKcdmsFmJ+9LoJHh6kp0HVmPp0NTYOUtoslQjuz2WdGSPPUsWsu0Aev/q3yXbsecOe5xZylXK0TteIiIiIgXRhZeIiIhIQXThJSIiIlIQXXiJiIiIFEQXXiIiIiIFqYlU4ytYG4e1TcZNI6k/FnQ8IJZmjtg61M7HI6EWs4+Zk0iN5ZFYhu3rpMayOVNIJu8dzAu1VVkg75ux9DP/SKyRXZ9igwGwb72F2p/2jGfo6/hPqN2PI0NtJnYItcFYEGrzSbrmBjK+oxbG2kaDyIbUTqTG0mTA6pgcaseR7caxHpM04SQ9XzNC+qshvpZ5Qow9Z1jKjj1f81K0qb39niA1lp5kqTrWT5Ul7Vj6l6UDWeqSJe/ykpxsht6H1Nh57EdqbNwsEcnOAzsGGws7r+wxWYfUciZyNsbFq8RaPUu0sttUqrEz9I6XiIiISEF04SUiIiJSEF14iYiIiBREF14iIiIiBamJxfXAGgB2X7G0jGzGVsOPYO1cYmuhKXTBI1vcCNxN2s78BN8JNbb8ki0RZUtJWZUt197wG7G2CVk7y5bjstublbM2dUtSW40spL+ErL88n9zmTLIvW967GS4ONbY0+JtkIf09ZLud0Rxqd2M/clzW3oQ1oOKP81JcHmqX4CJ6m9LTkZZBdJE0ewWwRc5sFmETYt4Ca1ZPPc6fYmkOeV4PYw27WFCABVtYSx02MbHbe53UAH7/2CuXhQfYbbLHiv0MYS2M2M8Vdv/YYv288EBbrAURQBvh1cdQU/p4pDP0jpeIiIhIQXThJSIiIlIQXXiJiIiIFEQXXiIiIiIF0YWXiIiISEFqJNXYjJAi2eqquNlWXyb7svTQ+qRGEi4zhvDhXHdmKJ10UVq7jz6I+/6cJPzGkxziPqT/UV9cEmpNe2wfapuRJCfL9IzEKFIF0Dw11urifbmn/wWh9jYa+G22wXI0MzEi1HbCqWRLlh6KucZXcDLZLiarZua0WzmUJMI2xqJQY+dW5H3s+cra2jxNaiyVuDup3ZxzbJYLTm39MiuWaPibJQ5ZW5xbSI3N42TOXjw81up5GplnnNmrNDVxyB6r1BRiamKQ1Vh7INaWKG8GYj8PWWadpSJTE5VSjt7xEhERESmILrxERERECqILLxEREZGC6MJLREREpCAVu/Ays2vM7A0ze7ZV7QNmdq+ZzS79SfoXiIhUn+YwEamESqYarwVwBYDrWtVOB3Cfu48zs9NL/z6t/E29hbadBVfHNWGrt2kihSUYp8fSvF1Dqd/m5Ue23LKLSEppwdhQammMacWv0U6KLL0S70vT5PfiZiSQN3PMCeT2ppPaoaQG4HZSG3N0KL2NkWTDvUntulB5OzWtOG9YrLEQ4qYx9bSs7kay4SGkxvrFASwVNAcDQm02OvDkke7qWnTJHDYAwIZtaqkJOJYuY6nE2D8WmJsznsQ5kY6H9DysZ/1w2bFZoo69zlj3WvICz00wMuzYZN5d3Bhr7DTUsdQg65fJOuKyfVP7U7KfcaxHJEuVAvF5CPDnGEu5sucs7zQs7avYO17uPhnx1bIf3p8hJgAYU6nji4isDM1hIlIJRa/xWsfd5wNA6c+18zY0s2PMbKqZTcUC8q6OiEjxkuawFeev+D1vItJ7ddvF9e4+3t1Hu/toNA6s9nBERJKtOH+lfYGwiPQORV94vW5mgwGg9OcbBR9fRGRlaA4TkZVSdMugSQAOAzCu9OdtKTvVYSHWbLOYfiFpJYNFYzo/MtKFYxlbhwjwtalsAXnjvFibE1vyNA07m9zecaGyOtYMtbd3YKvK2SJ11uKCLYzchNQAjFlAivuRGlsQyk4kWTR/QU67orbOaIq1IXckHpe1W3mY1NhCUqAP3g21JsQWTbwljPQAnZrDIrYgelVSYwvp2fOazSGxFVaGLTRPC/PwFAubR1hgiLU/YgvN2euRSZ3T8pCF7/UkPDCN/JgcxeYHdm7YfWbnZmXaCLHnDXt+ATQckfx8YOebPT+lnEp+ncQNAB4BMNzMXjOzI5FNVrua2WwAu5b+LSLS7WgOE5FKqNg7Xu6e870E2LlSxxQR6Sqaw0SkErrt4noRERGRnkYXXiIiIiIF0YWXiIiISEGKTjV2SjPWwkK0belDEjsNc8jeLN0Rk3t9GmJKsuX5i3NGtE8s3Tok1saQtMhLZLthm5JjzAoVlnsBbiE1lkZie7ME43/oUXhLi7NIjaVuWL8hkmY6g7TDWEDOF00KTSc1lmBk94Pd3mGkBrTQlkjsPrPWHmw80vMtRWyhwxJwLHGWmnZjKTSezOXtZFgC8vW029w7tubCHey5zualn5LalxPHwtoS7U5qAPASqZE5qJn8SGTdc2iRtd5hjylLB7IUNHuOsPvcL3FfgM9BavtTNL3jJSIiIlIQXXiJiIiIFEQXXiIiIiIF0YWXiIiISEF04SUiIiJSkJpINXKXkhpLME4gtdherQV7ke1uzjn26bE05pK07bZgPXVJn0eSwlmIrcl2LDXD0kwkiUmTdyyZA2DGnrE2gp3vnyWO5+ukRvrNNcYE0CBMDrWFGEZuj/WsHE9qqT3RAN5nbTqpsXOT1/xTerY+iOlE9tpbRmrseciSaexL9lmCF+DJP9YzkaSy5xwba4+TXSeTXrprkdqIY2LtLjLPkbaKuIiksm8lCUsAGMPOI0lysraF95DaLjF1ztOBLEGamm5miVaWVmSPXd4x2PzOjsOei+rL2FX0jpeIiIhIQXThJSIiIlIQXXiJiIiIFEQXXiIiIiIFqZHF9WsgLghlqy3ZQka2+Jy1xYmL2fvRRdPAR9EcajNxA9mStTW6l2zHFjey+8daQ7D7TFaINl8da3UzyL6/IDUAIw5J3JYt6mRjZO2K2CLRuEh9JNnqPnoMtoCZtVZhi94vJDWAL25lrZPYQlQW9JCerw7xuc1ey+z5yhaFswXSHWlHxRaBk9r1ZCE9e9leQWqpHZHwm1DZgOSc2Mzw1HpkIT3rLAQAY0ibnsVHxBpZM78B6Rz36kWpbdlYqyL2WLETxrZLDQLltQFi42GBIdZGjc2xtJ+SlKF3vEREREQKogsvERERkYLowktERESkILrwEhERESmILrxEREREClIjqcYlANqmUlhqjCU5WGKDpdjWCpVlYMkVYCaN56SmUlhLHdbug6VPWOqJ5X1ItKduItmOJTFZMgcAbkncNjUh8y1SY62TjgqV+7AF2e6CWPrzKrEWH2ZgFEt3svsB8MeZRamuIzW1DOqdDDHFuCnZjs1fqQnGvBZXqUiK+saxsTY9lg6aF2u//wM5RMMJsXbNZaEUG7oBt5Pa9O/EmjeSDQHgB2SefDiWDj8y1n6Fa0PNFh0eN2x4hByYvebZ3M5+1rDHnv0MYJNa3jzO6h1JxEpX0DteIiIiIgXRhZeIiIhIQXThJSIiIlIQXXiJiIiIFEQXXiIiIiIFqZFUYzNiT6gDyHb7kBpL1JE4CxUTdQCAOWfG2jCWjEtNGrGE03RSI70fMZfUWCKPJWSYA3PqLPnC0oUsnZM6HpYYZKlNkmBkPcO2IKlG+pCwzFRe0zc2xtTkrPRO/RDTaKR3IH2us6Qj6zXL5KXV2HHuiaU74nyzL9YOtUPNQ+3358eb23nM5aF23xHxtXP+kTFNeRv5SbUReS2vljPlPoWvxyJ5ibJWlFTDHFIkicPFw2KtPjWVmjpnM3mpRvbYs+ci62mb+ryTcvSOl4iIiEhBdOElIj2amW1oZgNKf9/RzI43szWrPCwR6aV04SUiPd3NAJrNbCMAVwMYBuC31R2SiPRWSRdeZnawma1e+vtZZnaLmW1e2aGJiHSJFndvArA/gEvd/SQAg6s8JhHppVIX13/X3W8ys+0A7A7gYgBXAvh0xUa2gj6Ii0xZ6wS2IJAtHGQL7k8itZyllmx9Ii4kNbb4nLWLYEEBtmieLCBni2Lpglw26I4sCn8i8TZZLTVkwO7f30ntDlIjC1EbSFupBrY4nj1O55AaAKxHauzcsABHXhsiqbBlZnYogMPwfgKnbQ+fCnLEhe7suc6wRdKsRcwsUmPzXF49BkwG4ZpQu41MLfZDcnOkq9rd78bawlX+J9Q+9C5pVVQf+xLNpvNcXqBgv1DpM+yZUDttbJyrTrv2cHJ77E6TQE49C0SxMA+bV9hcysJL7LmU9/xiczFbNM+2oz/4pBNSP2psLv25F4Ar3f025McmRES6k68A2BrAD9x9jpkNA3B9lcckIr1U6oXXPDP7OYCxAO4qLVTV+jAR6fbcfRaA0wA8Wfr3HHcfV91RiUhvlXrxNBbZZ1p7uPubyN5z/HalBiUi0lXMbB9kX4x3d+nfI81sUlUHJSK9VtKFl7u/C+ANANuVSk0AZldqUCIiXehcAFsCeBMA3H06smSjiEjhUlON5yB7q/6MUqkftEZCRGpDk7svalOLX7cuIlKA1FTj/gA+hffXSPxj+ddLFGMZYmqNpe+uIzWW4ElNCrFEHYAGVkxt98HSbizVuDuppabnSKSIplRYkiYPO2csecnONzsOOzfsfLOUEkkr0rEwrA0He+zy0p3scUm9f6npTuliz5rZ5wHUmdnGAI4H8NfiDl+H+BxjzwWWREt9LbPnW574Gu2Lt0LtPrbrU7HUb+P4Y2Tgf/NY72Of7bJ70q8+3t4y+gHL50iNpxrfwbRQu4Fsd/TvtiFVlnpmaXmGzWlsHv8yqaX+nMpLcjKsFRo7jrJzlZS6xmupuztKvyWa2aqVG5KISJf6JoDNACxB9sWpiwCcWM0BiUjvlfqO18RSqnFNMzsawBEAflG5YYmIrDwzqwMwyd13AfCdao9HRKTshZeZGYDfAfgYgLcADAdwtrvfW+GxiYisFHdvNrN3zayBrPMSESlc2Qsvd3czu9XdRwHQxZaI1Jr3AMwws3vR6mu63f346g1JRHqr1I8aHzWzLdydrQoUEenO7iz9JyJSdakXXjsB+KqZvYrsN0ZD9mbYJyo2shX0RUwFpfY8ZOmM1ERKXlqEJfdYn0i2P0vL3U9qbNysd+ShpMbuX2pfRTYWgJ9vdl/YGNl9Se2zxpJeLJnI0kMsM8XGx/o3HkhqAMByJWyMSgV1F+4+ocojQHxupybJ2HOYvU5Y8o719QPY66eJvAZWI0vintiYjOa3McH4wOfjdjthl1AbjD+H2lYkETkFN8cbpPeZJ4c/T1KNbO8f4ZFQO43mMNh7EOxnAJtjWYJxJKmxx/k5UmPyHnt2mywpyeY51tNROiP1wovldkVEuj0zmwPyvV3u/pEqDEdEernUCy992aCI1KrRrf4+EMDB4G9FiIhUXOqF153ILr4M2cQ1DMALyL4bR0Sk23L3hW1Kl5rZQwDOrsZ4RKR3S7rwcvcRrf9tZpsD+Gp7+5jZesi+Sv5DAFoAjHf3y8zsA8i+nmIogFcAjHV3tkhGRGSllear5fogewes3c4bmr9EpFJS3/Fagbs/aWZsbWJrTQC+Vdp2dQDTSnHuwwHc5+7jzOx0AKcj6wPZjtUAbNumxhZTs8WpbDEh+5SBzZ15LV7uITV2OthC9Q+RGlu4zlqFsH3ZAvc7SC211cSNpAbwc9v2MQGyHEZbrDkHe/w2JTU2RhZuYOcr9fnAbi+vZRDbf2VbuEiF/bjV35sAzAEwtsw+XTh/NSMGONjziC1oZkGSZaTG5h/2eso7Tnzdb4StyXbxNe/3XBxqs8jierbgez45xnwcRfZl5yG1xpfCM3xJOpu/2Gv+EFJjjzM7SmqwgulIK7LUT9jZQnp2XxQi6oykCy8zO7nVP/sA2BzAgvb2cff5AOaX/v62mT0HYAiA/QDsWNpsAoAHUHbiEhHptCPd/eXWBTMb1t4Omr9EpFJSezWu3uq/AcjWfO2XehAzG4qsyfZjANYpTWrLJ7e1OzBeEZGO+n1ijdL8JSJdKfWjxlnuflPrgpkdDOCmnO1bb7cagJsBnOjub2UdiMozs2MAHAMAWP+DicMUEcmY2ceQBYAazKz1l/ytgSwklHIbXTB/se93EpHeKvUdrzMSaysws37IJq3fuPstpfLrZja49P8HA3iD7evu4919tLuPRuMaicMUEfmv4QD2BrAmgH1a/bc5gKPL7dx18xdbXykivVW773iZ2ecA7AlgiJld3up/rYFs8Wl7+xqAqwE85+6XtPpfkwAcBmBc6c/bOjFuEZF2ufttAG4zs63dPX4leTs0f4lIpZT7qPEfAKYC2BdYoefC2wBOKrPvtgC+hKw57fRS7UxkE9ZEMzsSWbTt4PLDrENM5bEkB2v9wlIubF+WLGTpH4C3bdiH1Fg7n8dJjY2RpU9YEpCl73YjNZZSuZ3U2DkEQJNGbFuWEGQpnrzjtMWSNOwcsnPN2gPNJTX2fMhL66SmZNnzSR85VclTZnYcso8d//sRo7sf0c4+XTh/GdLSX6nJavLambNnrOVF+dhL9IAvxtpLpDb2LrJzTDV+fVzc6rjT2WuHpZHZ6+nCUNkAL4Taq7iM7AtsgcmhxjLibISDcGuoLaTHSU1Rs3kprz1dynbsfOXd3sqkEJVg7CrtXni5+9MAnjaz37o7yzC3t+9DyGYcZueO3JaIyEr4NYDnAewO4DwAX0CZpneav0SkUlLXeA01s9+b2Swze3n5fxUdmYhI19jI3b8L4D+lhtl7ARhRZh8RkYpIvfD6FYArka3r2gnZNzr/ulKDEhHpQsvfrX/TzD4OoAHZN8+LiBQu9cKr3t3vA2Du/qq7nwvgs5UblohIlxlvZmsB+C6yxfGzwBYNiYgUIPV7vN4zsz4AZpvZNwDMg744UERqgLv/svTXBwF8pJpjERFJvfA6EcAqAI4H8H1kHzceVqExEc2IiR+W9mGJtdTeVGzfvOQdS4ywxOFGpMbSbiwtwnr9pd5n1i8xNRGZmq4BeDyK9bFkPR3Z45KaSmU5DzZudg5ZspAdN6/XYmqPUHabSjVWg5mtA+CHAD7s7p8zs00BbO3uVxczgmbE125qAo49j8g8wKICJ5MagH6vx9oy8tQ86IRY+/35MT1pz/wp1M5myeo5ZHXKsHfJCFmv2fNC5VXcTLbjLYQnYSqpsjc9HyY1klhfdHysNTxI9mXzc2o6kH3/G9uXpWF5z0qe5VyZnz8sLS/lJH3U6O5PuPs7AP7t7l9x9wPd/dEKj01EpCtci+w3gg+X/v03ZL9MiogULunCy8y2NrNZKP1eZWafNLOfVXRkIiJd44PuPhFACwC4exOyt6FERAqXurj+UmTfgbMQ+O/3e+1QoTGJiHSl/5jZIAAOAGa2FYBF1R2SiPRWqWu84O5z2zSI1W+MIlILTkaWZtzQzB4G0AjgoOoOSUR6q9QLr7lmtg0AN7P+yBbZt/vNz11rKeKiZrY4O7XNDsMWE+YtgmQLONn+eQsc22KL3Nm+qYu4XyQ1tgiSna+8ReVsASfDFpOS1bzJbXrYuPkC2rRjpI4l73mTupiUtULpSHBBVpaZHezuNyF78n4GWdNsA/BCRztxrJwmxOdd3ussBWnBxaaQ8XzvZWz9+KWxxDoO+QyyONsOjKXnPekYOHAVst3YWGPd237H5oGc19ijw2NtqxPJhmwhPWmdxHJFm3wm1tjDXB9bHfENU3+esZOzHqkBfI5NXVzPzq0W13dG6keNxwI4DsAQAK8BGFn6t4hId3VG6c+b3b3J3We6+7PFXnSJiKyo3Xe8zOxH7n4agJ3c/QsFjUlEpCssNLP7AQwzs0lt/6e771uFMYlIL1fuo8Y9zewsZL853lTAeEREuspeADZH1t7sx1Uei4gIgPIXXncD+CeAVc3sLWTrI3z5n+6+RoXHJyLSKe6+FMCjZraNuy+o9nhERIAya7zc/dvu3gDgTndfw91Xb/1nQWMUEek0XXSJSHeSlGp09/0qPZD2LUFstcOSaPeHyiBcH2oL0RhqByHOzbdiAB1NE2kr4QtHh9rAQXHfJc0x7bNBnYUaa7LDasctIumhhovJlrGVz8n4c6jlZWFOnBJrF24fa6fhSrJ3TGj+BN8LNRa2io8osBCnxOIFsYXH4WfGzQ4gt7fvXqR4xzWkCPCs1yaxNI+0FBnCHhfpnVhqmb36WLKNtJIZPo3UcoLne5KZ5LxhofRq3Q9DzXAuuUGSsgR5rl/JEtgxEYldWFs1ti9L2bE0MYBVSaqRzjhEw8RY24G182Hnm7XoYWNMTd+zFOEsUmOxS4A/x1hOjo3nOlIjKVApKzXVKCIiIiIrKfkLVEVEao2ZDQSwN4DtkfVqXAzgWWTLJ2ZWc2wi0jvpwktEeiQzOxfZZyEPAHgMwBsABgL4KIBxpYuyb7n7M9Uao4j0PmUvvMxsawBfRPYb42C0+o0RwPXurp5nItIdPeHu5+b8v0vMbG3wBUoiIhXT7hovM/sjgKOQrczeA9mF16YAzkL2m+NtZqYvIRSR7mietWkw25q7v+HuU4sckIiIuZNU3PL/afZBd/9nuzeQsM3KstFrOKZ+uk01pin64tRQawL71ovZsXRBQyjVncfH0zyYFF+O6UmWQ6yrj+mh5sVk36+R/mBXkuTSjqNi7UQyFBYPvIyl7FgvQ4CnXFiqiKVNWZ9H1mft67E0bUSsnRNLf7kz1nbCaeQYE0gtLvWxrUkkFQB+TmokhDVg/1hb4ko1doh9e5q7x7hw6u5mUwEMA/AksgjbXwE86u5vddEI08Yx+iOOqW0nE5KEZU8k2neVvUnHXk/stZh3HNabj6XlWEqPJTTZdqwnIJtDUucVth3Lfucdm02KrMaSe+wxuJnUdic19tjnpRDbYqlElnRkKVAgvUcoO1+p/XUFAGD7585f5b7Hq+wFVaUvukREOqM06a0H4AfIfkofD2C2mT1tZj+r6uBEpNcq16vxbWTfVP/fEvTN9SJSI9z9XQAPmNkTyBbYbwvgy8iWToiIFK7c4vr7kL1nfAuAG9099VveRESqysw+D2AbACORfQvz8ouv7dydfT4mIlJx7V54ufsYM2tA9oXfvyjFr3+H7CIs70NkEZHuYDyA5wFcBWCyu/+tyuMRESn/dRKlr4v4lZlNAPA/AP4XWaLxkgqPrZVGAEe3qcVFgsOxJNRmkoWRtnpcSN/yzkPkuNPpaF6f841QG2SxGU2/hlVCbdR78fYevzgupPer/hJqX7nqs6F2Oxnfwg1JkXWzuJ603mHrQwGsfmusvc26Ax27DSmeFSon2ZhQe4Dsye4KW+LJatvjR6H2OqnNZi2DbiI1gK6B3Y4spL+PLPzlDaikghqQ9WfZBsC5ZjYcwHwAjwB4xN3ji6wi+iDOV6xdzUakxsIubOE6W7K2U854UhfSs8XrrIUR+x2chQI2JbXUBdtsMTv7AIaFDADgfFJji9IPJTW28J2Nm93nQ0iNnVf2Bix7PrCxsHPDZk6AP/YXkhpbxE/aO2FuznGkPSnf47UNsmfj9gAeArC/u5POfSIi3Ye7NyNLND4J4AozWwfAQQBOAnAegLoqDk9Eeqlyi+tfAfAmgBsBHAOgqVTfHADc/cnKDk9EpHPM7BPI3u1a/l9/ZO92/S+SOySLiHStcu94vYIsxbg7gN2QpRmXcwDxsy8Rke7hWmTf3fVHAN9191erOxwRkfKL63csaBwiIl3K3Tev9hhERNoq1zJouzL/fw0z+3jXDklEZOWZ2SQz29vM+pH/9xEzO8/MjqjG2ESk9yr3UeOBZnYhgLsBTAOwAFmicSNkkZkNAHyroiMEALwB4KdtajHxMRN/IvvG5Iq/Hft622RyjflOu9edK/jsXjHByMJHj8eAH/Uz8inutexHBAu0sC4V4UcPeMAlBjYBAG+PIUUWUsL3kg70BNmKDZFllE4bEmv3kFbt/ibZue3TCOBhnfpPkCKA9Z4JpYfJeRwwTK00uoFjAJwM4DIz+xfen7+GIuubc4W731b5YTQhJv9YKo6lA9nziL0qOvLtPmxbNpGsRWosCciSjiyRx47L2hcxR5HaPR24PZYETb3PbNysxiZUNp5lpMYeZ5ZKZfeZpSTzsPRkahshdmz6Q0DKKPdR40lmthayJNDByJpkL0b25QQ/d3f2HQwiIlVX+pLUUwGcamZD8f789bfSN9qLiBQu5Xu8/g3gF6X/RERqjru/giwsJCJSVe2u8RIRERGRrqMLLxEREZGC6MJLRHokM7u22mMQEWkrpWXQBgD+4+7/NLOtAGwH4CV3/0PFR1cyDO/gh5i8Qu1Q/DtuOO3qWBs1j9wi6am1w6Vku/Qvt/7L8y/H4vAFZEuWhokdF49bjfRRPPZIsu+XY2nyZ2JtqyayXXz462K7SwBA8x/uJVXSAPKa42ONhG4ecjIesuHjE0ladNtY6js01pbVzYnF488OpT6IT+UWkAaMAHBDLHkMOgLDruL7S5FyoqlFa0Z83aemXlnPPZbQYynCdRKPAfBkHGvwyhKMDEsHst6K7LifJDXWo5Ddv/tzxsOOw3LU7D6npv4OIzXWTZclWtkx2H1hPz/YmPP6dLJjb5I4HnYc6YxyLYO+C+BwAG5mNwLYBVkv473M7DPufmKlBygi0kmrmNmnsGLHjf9SyzMRqYZy73gdiuxyeBVkv658yN3fNbO+AKZXeGwiIitjCIAfg194qeWZiFRFuQuv99x9KYClZvbS8u++cfcmM+vIt7aJiBTtRXfXxZWIdCvlLrzWNLMDkP3GuEbp7yj9u6GiIxMRERHpYcpdeE0GsA/5+/J/F2J1sKWCZPHfSLY3WyDPvgs2voHXB2/Q8exN2mZMuo5seF5jrE0ntVGkgQ7JDgDrkdrIWCILwLGUPNSke0Tz99lxAeBnsbSA5CvOJ7teEkv7koWtrI3Q/LGxP1A/MvBlT6Yt1me1FqyduC+AUx6MtWYSZph4bKyNvZjfplTKBWa2trvzF3Jh+iO2p5lFtmOLxdliahIOolIXwgN8sf90UmNtdthrhbXKYQu22SJwtjCfLexmgQJ2XgFgy8TxsOOweZfdZ3Z7LCjAtmOL3lkYgT1OqS2NAN4yiGGPc0fCGtKeci2DDi9oHCIiXW0sgJ+Y2bvIfgP7K4CH3X1mdYclIr1Zu9/j1fp7cMyMZWVFRLoldz/I3YcA2BXAn5B9vcR1ZrbAzO6q7uhEpLcq91Fj6/dJTwAwoYJjERHpcu7+ipkNBFBf+m/530VEClfuwssLGYWISBczszMBbA2gEcALAB4FcAWAY9y9uZpjE5Heq9yF17pmdjmyFOPyv/+Xu5OvKc+UfsOcDGBA6Ti/d/dzzOwDAH4HYCiAVwCMdXe6lFxEZCV8Gdmq5TuQre96zN0Xpeyo+UtEKqXchde3W/19agdvewmAz7r7O2bWD8BDZvZHAAcAuM/dx5nZ6QBOB3Baezf0DOrwoZAE2SJuyIIm9azlRuw50xfjQu1K8P45R82ONfsBS9V9IZZGxfFsgAtC7dUzWHIpJgvZuJuu/DjZN6ZcBiG21Fk4Le9NzvGxdHraZtjlmlDah2wWs6LAfJIA+hzimxUbjYjfkcmyq2/jIlI9NVQGYDTZDgAmhsqSuhFxs7FsX9LeSSrG3T9WulDaBsCOAE43s9WQ9Z/5q7v/qp3du2z+yj44aDs5sWc7awX0Oqmx5B1LseW1JWIJOpYkZEnA1PQjSyuyfdl9Zm12WBKQpezyUo2rklrqeWDpSZY4ZNi+7AcVa9vDct4sqcqSk3mPPTvf7Dyw88X2lc4ol2rs9Joud3e8/4j2K/3nAPZDNgkC2ZqxB1B24hIR6Th3/xeAO8zsbgCjAOwA4KsAjgCQe+Gl+UtEKqVcr8ZJ7f1/d9+3zP51AKYhu+z/qbs/ZmbruPv80v7zzYx9iZKIyEoxs32Rvdu1LYDNAMxE9pHjt0p/lttf85eIdLlyHzVuDWAusq/kfAw5zWbzlBawjjSzNQH8wczYZ2CUmR0D4BgAwPodOqyICAAcjuz7u04FMK3U/ixZ181fgzpyWBHp4dr9Hi9kH+qeCeDjAC5D9n04/3T3B92dfIU35+5vIntLfg8Ar5vZYAAo/Um/Vdrdx7v7aHcfjcZywxQRCU5DtqD+kdYXXWa2vZmxxZ/Uys9fa6zEXRCRnqbdKxp3b3b3u939MABbIetV8YCZfbPcDZtZY+k3RZhZPYBdADwPYBKA5V/GehiA2zo/fBGRXD8B8BapLwZwaXs7av4SkUop91EjzGwAgL0AHIosQn05gFsSbnswgAmldRJ9AEx09zvM7BEAE83sSGTRkIPL39RQAG373KV+akASG4vODaWmhths8UMk9QcAD2zMqvGUfJ8k4FhG6RBSuwPfCzWW4WFZn6NmTgu1hzeL27FuXENHsVvkWq5m3yZyR6h8GjNC7askNbMhTXrFx5kFJ9fZPdaeuCfWpuDGUNt+hfBuZvKr5CAAttog1h4jCafTEb+1IOZPpcKGuvszbYvuPtXMhpbZtwvnrz6Ir1T2httcUmPJtr0Tt8ubI1niLTWlx7CZhPVqZD0PWWNZlgTM6z2YMhaAN79N/RnC+huyGjs2u8/spwB77Bn22LFUI5tL8/a/n9RYgjH9Z4O0r9zi+gnIPmb8I4DvufuzqTdcmvA+ReoLAezcwXGKiHTUwHb+X7vfXK/5S0QqpdziqS8B+CiydkF/NbO3Sv+9bWbsLXwRke7iCTM7um2x9G5VfFtYRKQA5b7HS6vaRaRWnYgsjfgFvH+hNRrZ5y37V2tQItK7lV3jJSJSi9z9dQDbmNlOyJZMAMCd7v6XKg5LRHq5Grnw6oPYwmC3uFn9vWTfh2OpgS08jG2EDs9ZXM8bJ8Sl799FbCWzPVlozlrbtKAu1JrJkpVN2CJKspCenAWMJ2tLf9WfL8pkTSV+SVodsYYd0zAg1FoQH6vZZOE7W/j5NNlqtwNjjS1B3pZ8wvQTst05ZBE9ADyGK0k1Lvx9Ed/hNyCFc/f7wVcQVxFr/cLatLDF0GyhOGs5w14pAF+MzRaaswwVCwWwY7N9TyQ11v6ILVJnNRYoYM3IAD4DsttkrdrYfWZhBLYve0xZUIDV2GJ21tKIPUfY8yvvOKy9E7t/bDxacN8Z+ihRREREpCC68BIREREpiC68RERERAqiCy8RERGRgujCS0RERKQgNZJq7IeQJZyxStxs7q6xtidrz0DaPTT/OpQW1v2KjmYhWKvKM0KlL/4WalMwKtT6YXqofRTNoTYTW4fahvhzqB0N1sonWtr/8lD7LRkfAAwiacCF2IVsydp9kH4+JHG1MeJ4ZmM4ubUdQq3vMY+F2p5YEmrnkZGMI/fjYXJeMyxJFduC/B5H5OwvvY8jpgZTG4CxBCJLprGsNUvyAentbth2LN8cE3Abk/liNh0Pu8/suHkpvbZYShLg55G16dmU1FiKlCUJ2bjZfJiaDkxNtLLHJK8HPNt/C1JjqcbU+yfl6B0vERERkYLowktERESkILrwEhERESmILrxERERECqILLxEREZGC1ESq0TAL/fGJFWpLRpDU2NKryd6sd9fNSSX0zzk9Y8aRYuzJ1YQ7yHaxg+AyfDXUZtLWcrGf2iQ0xs2uuizWDo2lfzXEFCFI/0UAWEhTNy+R2gRSI6nGBceG0uzGs8m+RyUdt4k8zpNIwunfuIbc3qmhch9NYgK8px05udcPi7UvXpxzm9KzsVQjex6xGkussbQb25f1UATA+rvSZBubO1naLSYqZ4OlyW8nNdYnkCUY2XFZEnM6qQE8NcjSoew8svGwGnusWBIwpqC51ORkR3o1sseejZulZNmxpTP0jpeIiIhIQXThJSIiIlIQXXiJiIiIFEQXXiIiIiIFqYnF9Y46LAmLI8liSbbGkCx6B26IpbFsMWJe2wWyGPHPsZ0PdiGLKGeQUz6CtbN4kdQOJLVfxtKx88h28T7fjSFkuzys1QRre8IWpz4eS41sMS9bLLstqZHF7HShbDw3U9h9nkNaTQ1ji3EBfv+IL7LggvROLYjPG/b8Youc2XZscTYLurAawBfds/Y5bEJl42FttNjrm23Hbo+9xthczBbwsxoA/IjURpIaO7dsPKltjdi+bC5l26U+R9jPj7x5iu3PJM5z0il6x0tERESkILrwEhERESmILrxERERECqILLxEREZGC6MJLREREpCA1kWqsRzM+hkUr1J7CrLjhiEWx1twQa3UsHfgLUmNJH4AmbHY5jGz301ga8XuyXWxtw9M5qemab5AaSyh9OXEsAPA0qbHEIduOtd85OVQGYFqoLcEtZN8DSC0mpgbjhVCbjxFx12FxO+BgUgP4OWPPJ3Ye81Ky0rPVISbU2NySmlZk7VweJjX2WgR44jC2KOuLn4daEz5O9o1t0HhrmrwWRm2xlB5LPKemKYHBpFXYfDrGE0mNjZsdm83PLG2d2r6IpcbZvmxuz0svsrQ8Ow67Tfq1AdIJesdLREREpCC68BIREREpiC68RERERAqiCy8RERGRgujCS0RERKQgNZFqXIy+eAprJWx5eCzV3US2Y8mX2NdvMGbQo8ynKb2Y+OiDR0KthaYQY/podZLIe5smVViy8CxSY6knNhaWrgF4Qir2JuuL90Ktid7ePaGyBKTfJdkOmE5qMSU2H1eS7S4ltfNJjZ1XgCc0WeJqJ1L7e85tSs/miIk3lnBN7Y3I0nMsJZnXby+tD987zUtC7bW6mDzeiKaM2THYa4qllsnraRHpsdqwKdmXm096tPbB86HWQpOl7PXNtmNJ9NTEIdtuGanl9ZBti/WDzJPap5P1opTO0DteIiIiIgXRhZeIiIhIQXThJSIiIlIQXXiJiIiIFKQmFtcDHwZwRpsaW2RIFi83k7tYdzvZN7a9mE8XjQJ8cX5eq522JpBaXKj5Nl2cuiWpsUWQbN/UxfVsUTjA73Ns29SEz5HtYnCBL8Jn5/tnpMaCFqyVz1GkxtqtsPOVtxCebZu6KFp6pz6Izwe2+Jk9t9jziC2GZqGYvIDIRqQWF04P+EPcqv9B7PbY8z++5j+NBaH2GIbHXdlCetqxbEys1ZG2cQDYAvltMS/UprAwz+LPxFo9GxBrdcQeZ7a4ns2vrKUR2zevPRCT2q6IYc+btKCGrEjveImIiIgURBdeIiIiIgXRhZeIiIhIQXThJSIiIlIQXXiJiIiIFKRGUo2rICZGWIrwiViqi613eHKPJYXysGRITAW14NzEfVnSjqUQWZuRTUiNpTYPTNwury0ES6+wNAwbY0zDNGE3sh07Ny+RGktrsUQRjUKRGrtvPyW1PCNJrV8H9peezRCTf6ltY9hrLLUdD7s9ILVV2A4kwTgFo2LxqlNi7djrQunLNNVI5i82tbNpjkz3WKuBFAEMj/d5CvYkG54WS2eTzS5iA0pNMLK5IXVfVmOthVYlNYC3P2LPEzZGdhzpDL3jJSIiIlIQXXiJiIiIFEQXXiIiIiIF0YWXiIiISEEqfuFlZnVm9pSZ3VH69wfM7F4zm136k/V/ERGpOs1fItLVikg1noAsp7JG6d+nA7jP3ceZ2emlf5MoSWt9EZN/rIcfS7GxBA9LBbG0CNsXAF4kNZYGfD1n/7ZYopIl/Ni47yc1lgRk+7L7l3efU3uTsYTMdqTG+n7dQGqxHySvHU1qLOnFkpws5TqS1PK2ZT972f7sHEo31wXzVwvi/MJej6l9QNlcxeaQvNcyO058LU/BiFDrg+dDreXYe0NtM8wIta/PjEc9bjNy/7aaSMZHIox/vyjWhj9I9gX4fLp7LE0jqUh6GmMKsQ+uDrUWHE/2ZfMhS6+yGhsM+9nTkXQ6e46xPpFsnlOvxs6o6DteZrYugL2wYsfU/fB+p+gJAMZUcgwiIp2h+UtEKqHSHzVeCuBUZL/yLbeOu88HgNKfa7MdzewYM5tqZlOxILV7uohIl7kUXTJ/vVXxgYpI7ajYhZeZ7Q3gDXef1pn93X28u49299FozHvbVESk63Xt/LVG+R1EpNeo5BqvbQHsa2Z7AhgIYA0zux7A62Y22N3nm9lgAG9UcAwiIp2h+UtEKqJiF17ufgaAMwDAzHYEcIq7f9HMLgJwGIBxpT9vK39rSwH8vU2NLZZk7WrYAvebyx8SAHBoTp0tMmT9K9jic/axKVssm9paqPNte3hvjry2EOzYbKE5W9j/EKmxUMA6pDaS1Fi4gZ0HdnvrkRo7/+wc5mHnUe/S1rKunb/eQ3zObkq2S20PxBZDt50f8/bN258FP+Ic9GOy6PokHBxqz7K4Abt7v2CvvemhsjMuDrX7RrHXKOvvA7C5qh9ODrVlo7aIu/46tj9i57sFHyHbsfM6ktSY1J8BDDuvQHqog8lrQSUdVY3v8RoHYFczmw1g19K/RURqgeYvEVkphTTJdvcHADxQ+vtCADsXcVwRkZWl+UtEupK+uV5ERESkILrwEhERESmILrxEREREClLIGq+V54jJC5ZOY8keloBjaTyW2GCpuLxtWYottd0H2461hmDbsZRL6u2x1MwBpAbwtCLDkkZ5CZu2SKKIYqlS1jKFJU3XJzXWRiiv3VPquWXHlt6pH+LrlD2P2JwWW9Pw1xNLUOc9h9m8Frf9EWLrHpZ/OxmLYpFNxTeS2i/YefhlqKxKtuJzEk/esfvC9h6Pr4ZaX/wt1JqwH9mbnVeWeJ5OagxLvrL5PjUVD/Dzwx5VtQeqJL3jJSIiIlIQXXiJiIiIFEQXXiIiIiIF0YWXiIiISEF04SUiIiJSkBpJNTYhpm52J9ux1AWL0rB9WaIuLxXEsLTJrMTtWHKJ3ZcbEm/vcVJjPRjZ+PLSiyylx1JYLDXI+kSyJCDricZSQbuRGkv23EJq7HFmidS8nmhs3LF/HW9MJ71THWLKjOX0WOKMPbfYdiNJjUULAf7cjnPdqU/FreZ+KtZICZ8kU9WGbPoi88rGWJCwFcAS633xT7rlCc2x9nJdrI3HJqHWRH9epCa10841vz02b6bOK3mpxrx6W3k9e6Ur6B0vERERkYLowktERESkILrwEhERESmILrxEREREClIji+sXIy4EZ4sR2YJ0thiRtRGKiyrzF1CyBd+sXQQbD2sDwRZss/GkLoxki8VZLfV8AXyMrK0EW/jLbpPdHjvf25IaW3TKFiGz9kUsEHAPqbGQAMDvCxsPO99acN87sZZnbIF16us7tXULew4Cqe1glnwibnUp2fNQUmPP9MMwPOm4s7E12S62MvsGvh1q+5A9AeB5spCenZ0+uDPUWrAX2ZK1aErF9mVnLHUBP3vepO4LrFxrO+kMveMlIiIiUhBdeImIiIgURBdeIiIiIgXRhZeIiIhIQXThJSIiIlKQGkk11iOmAVMTYiwdyLIvLCk0N+c2U9NHLGl3P6mxhB9LkLDtWEryQFJj7UNuJzXWygTgCUHWDoOdb3bs1FQq247VWFKLPU6ptbwEDxsjqykBJMs1Iz5nU58fqc9N9rplr9m824zbDqwbQraLad+f4pFQ4w23WOud6aTGxhdbfbE2QnkzM2uOxrbdjczZH8LEULsW48jeLO3OEtOpPz8YdmbZ3JfX7o49d1gtNTkrnaF3vEREREQKogsvERERkYLowktERESkILrwEhERESmILrxEREREClIjqcYWxJQFSzWy7ls8XxOx5B2rATwtxJIhqQk/liBJ7b/F0kzsGCxxw46bl4ZJ7WX5NKmx/BHrbcmO8W9SY48z63/GUqnsPLBznZeaZXWlgqQ9hrQUY2raLTWZxvqSAli8SqzVHxZrM06JNfLyPo6EqPvVsx8t7DXPko4sdf5SqDyGc0iNzTUAPz/xPu+L/UONZ0PZuWWP3/TE7VIfezYatm9en86O9HBMoXmuM/SOl4iIiEhBdOElIiIiUhBdeImIiIgURBdeIiIiIgWpocX1KS03WI0tMkxd3JjXcoPdJlu0mNoWhC0MZ4vKWQsihi1wZ4vK2f2IrTnyt52eOB62IJ0tTo3tSIBlpPZi4nFTWzGxxy5vsSs7D2yBKQt15IU1pGfrg/i8W5nWU6lhjrggHQBQT0Is0xpjbdQPY+0LZ4bSZ78UN/uLv0oOzMZ9QyzdNSLW1iO1Eaz9Wt6cS17jC8aE0lxyGvhMwM4tm2/YfJHa4ofNkalt1fLmr9RWbf1y9k/ZV8rRO14iIiIiBdGFl4iIiEhBdOElIiIiUhBdeImIiIgURBdeIiIiIgUxd6/2GMoyswUAXgXwQQD/rPJwuoruS/fUU+5LT7gfG7g7yZnVllbzF9AzHpflesp96Sn3A9B96U5y56+auPBazsymuvvoao+jK+i+dE895b70lPvR0/Skx6Wn3Jeecj8A3ZdaoY8aRURERAqiCy8RERGRgtTahdf4ag+gC+m+dE895b70lPvR0/Skx6Wn3Jeecj8A3ZeaUFNrvERERERqWa294yUiIiJSs2rmwsvM9jCzF8zsRTM7vdrj6Qgzu8bM3jCzZ1vVPmBm95rZ7NKfrHNqt2Jm65nZ/Wb2nJnNNLMTSvVavC8DzexxM3u6dF++V6rX3H0BADOrM7OnzOyO0r9r8n70VJq/uoeeMof1tPkL6F1zWE1ceJlZHYCfAvgcgE0BHGpmm1Z3VB1yLYA92tROB3Cfu28M4L7Sv7u7JgDfcvdNAGwF4LjS41CL92UJgM+6+ycBjASwh5lthdq8LwBwAoDnWv27Vu9Hj6P5q1vpKXNYT5u/gF40h9XEhReALQG86O4vu/tSADcC2K/KY0rm7pMB/KtNeT8AE0p/nwBgTJFj6gx3n+/uT5b+/jayF8kQ1OZ9cXd/p/TPfqX/HDV4X8xsXQB7Afhlq3LN3Y8eTPNXN9FT5rCeNH8BvW8Oq5ULryEA5rb692ulWi1bx93nA9lkAGDtKo+nQ8xsKIBPAXgMNXpfSm9tTwfwBoB73b1W78ulAE4F0NKqVov3o6fS/NUN1foc1oPmL6CXzWG1cuFlpKY4ZpWY2WoAbgZworu/Ve3xdJa7N7v7SADrAtjSzD5e5SF1mJntDeANd59W7bFILs1f3UxPmMN6wvwF9M45rFYuvF4DsF6rf68L4B9VGktXed3MBgNA6c83qjyeJGbWD9mE9Rt3v6VUrsn7spy7vwngAWTrWGrtvmwLYF8zewXZR1ifNbPrUXv3oyfT/NWN9LQ5rMbnL6AXzmG1cuH1BICNzWyYmfUHcAiASVUe08qaBOCw0t8PA3BbFceSxMwMwNUAnnP3S1r9r1q8L41mtmbp7/UAdgHwPGrsvrj7Ge6+rrsPRfa6+Iu7fxE1dj96OM1f3URPmcN6yvwF9M45rGa+QNXM9kT2OXAdgGvc/QfVHVE6M7sBwI7Iuq2/DuAcALcCmAhgfQB/B3Cwu7ddwNqtmNl2AKYAmIH3P4s/E9kaiVq7L59AtmCzDtkvIBPd/TwzG4Qauy/LmdmOAE5x971r+X70RJq/uoeeMof1xPkL6D1zWM1ceImIiIjUulr5qFFERESk5unCS0RERKQguvASERERKYguvEREREQKogsvERERkYLowktymdkgM5te+u//zGxe6e/vmNnPqjCefc2sQ41SzexaMzuoUmMSke5J85d0V32rPQDpvtx9IbLO9zCzcwG84+4XV3E8k1D7XzwpIgXQ/CXdld7xkg4zsx3N7I7S3881swlm9icze8XMDjCzC81shpndXWrPATMbZWYPmtk0M7tneSuIVrdZZ2YvW2ZNM2sxsx1K/2+KmW1kZoeb2RWl2rVmdrmZ/bW030GlupnZFWY2y8zuRKvGqma2s5k9VRrbNWY2wMy2NLNbSv9/PzNbbGb9zWygmb1cyAkVkcJo/pJq04WXdIUNAewFYD8A1wO4391HAFgMYK/S5PW/AA5y91EArgGwwjd3u3szgL8B2BTAdgCmAdjezAYAWNfdXyTHHVzadm8A40q1/QEMBzACwNEAtgEAMxsI4FoA/1MaW18AXwPwJIBPlfbdHsCzALYA8Glk32YtIj2b5i8plD5qlK7wR3dfZmYzkLWwuLtUnwFgKLKJ5OMA7jUzlLaZT25nCoAdAAwDcAGyiedBZL3umFvdvQXALDNbp1TbAcANpYnwH2b2l1J9OIA57v630r8nADjO3S81sxfNbBMAWwK4pHQbdaXxiEjPpvlLCqV3vKQrLAGA0iSyzN/vQ9WC7OLeAMx095Gl/0a4+27kdqYg+61tSwB3AVgTWY+4ye0dt8Ra/Z31wTJSa33czwFYBuDPyH4L3a6d44pIz6H5SwqlCy8pwgsAGs1sawAws35mthnZ7jFkb623uPt7AKYD+Co69pvbZACHlNZcDAawU6n+PIChZrZR6d9fQvbb6PJ9TgTwiLsvADAIwMcAzOzAcUWkZ9L8JV1KF15Sce6+FMBBAH5kZk8jm5C2IdstATAXwKOl0hQAqyN7yz/VHwDMLu1zJUqTU2ki/AqAm0ofKbQAuKq0z2MA1sH7vyE+A+CZVr/5ikgvpflLuprpsREREREpht7xEhERESmILrxERERECqILLxEREZGC6MJLREREpCC68BIREREpiC68RERERAqiCy8RERGRgujCS0RERKQg/w/cRymUgeDy1wAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 720x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "batch_obs_indx=6 # observation index inside each batch\n",
    "max_frame_indx=50 # last window indx in the plots\n",
    "fig,ax=plt.subplots(1,2,figsize=(10,5))\n",
    "if 'M' in features:\n",
    "    ax[0].imshow(sample_input[0][batch_obs_indx,0:max_frame_indx].swapaxes(0,1),aspect='auto',cmap='hsv')\n",
    "    ax[0].set_xlabel('Time window')\n",
    "    ax[0].set_ylabel('MFB (M) features')\n",
    "else:\n",
    "    ax[0].set_axis_off()\n",
    "    ax[0].get_xaxis().set_visible(False)\n",
    "    ax[0].get_yaxis().set_visible(False)\n",
    "if 'W' in features:\n",
    "    ax[1].imshow(sample_input[1][batch_obs_indx,0:max_frame_indx].swapaxes(0,1),aspect='auto',cmap='hsv')\n",
    "    ax[1].set_xlabel('Time window')\n",
    "    ax[1].set_ylabel('CWT (W) features')\n",
    "else:\n",
    "    ax[1].set_axis_off()\n",
    "    ax[1].get_xaxis().set_visible(False)\n",
    "    ax[1].get_yaxis().set_visible(False)\n",
    "plt.show()\n",
    "plt.close()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2e57a502",
   "metadata": {},
   "source": [
    "### Building the HyDRA model\n",
    "A tensorflow model object can be created by passing the hydra_id (e.g, 'M','MZ','WZ',...) to `build_hydra`. You can access to the function's documentation by pressing `shift`+`tab` twice in the following cell."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "78206345",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model\"\n",
      "__________________________________________________________________________________________________\n",
      " Layer (type)                   Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      " input_1 (InputLayer)           [(None, None, 48)]   0           []                               \n",
      "                                                                                                  \n",
      " input_2 (InputLayer)           [(None, None, 48)]   0           []                               \n",
      "                                                                                                  \n",
      " masking (Masking)              (None, None, 48)     0           ['input_1[0][0]']                \n",
      "                                                                                                  \n",
      " masking_1 (Masking)            (None, None, 48)     0           ['input_2[0][0]']                \n",
      "                                                                                                  \n",
      " gru (GRU)                      (None, None, 300)    315000      ['masking[0][0]']                \n",
      "                                                                                                  \n",
      " gru_2 (GRU)                    (None, None, 300)    315000      ['masking_1[0][0]']              \n",
      "                                                                                                  \n",
      " gru_1 (GRU)                    (None, 50)           52800       ['gru[0][0]']                    \n",
      "                                                                                                  \n",
      " gru_3 (GRU)                    (None, 50)           52800       ['gru_2[0][0]']                  \n",
      "                                                                                                  \n",
      " concatenate (Concatenate)      (None, 100)          0           ['gru_1[0][0]',                  \n",
      "                                                                  'gru_3[0][0]']                  \n",
      "                                                                                                  \n",
      " dense (Dense)                  (None, 500)          50500       ['concatenate[0][0]']            \n",
      "                                                                                                  \n",
      " dense_1 (Dense)                (None, 500)          250500      ['dense[0][0]']                  \n",
      "                                                                                                  \n",
      " dense_2 (Dense)                (None, 1)            501         ['dense_1[0][0]']                \n",
      "                                                                                                  \n",
      "==================================================================================================\n",
      "Total params: 1,037,101\n",
      "Trainable params: 1,037,101\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "model_HyDRA=build_hydra(features,show_model_summary=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "893a5be6",
   "metadata": {},
   "source": [
    "### Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "a83a9eb0",
   "metadata": {},
   "outputs": [],
   "source": [
    "n_epoch=10\n",
    "patience=20\n",
    "model_filename_save='example_'+features+'_model.h5'\n",
    "verbose=1\n",
    "# The following parameters will affect how fast your model is trained\n",
    "n_workers=multiprocessing.cpu_count() # adjust based on your cpu specs\n",
    "max_queue_size=10                      # adjust based on your cpu, ram and gpu specs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "6cd3ce7d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "215/215 [==============================] - 39s 128ms/step - loss: 0.2955 - mae_loss: 0.1693 - val_loss: 0.2343 - val_mae_loss: 0.1190\n",
      "Epoch 2/10\n",
      "215/215 [==============================] - 24s 109ms/step - loss: 0.1750 - mae_loss: 0.0701 - val_loss: 0.1553 - val_mae_loss: 0.0607\n",
      "Epoch 3/10\n",
      "215/215 [==============================] - 23s 104ms/step - loss: 0.1398 - mae_loss: 0.0546 - val_loss: 0.1336 - val_mae_loss: 0.0574\n",
      "Epoch 4/10\n",
      "215/215 [==============================] - 24s 108ms/step - loss: 0.1173 - mae_loss: 0.0491 - val_loss: 0.1148 - val_mae_loss: 0.0542\n",
      "Epoch 5/10\n",
      "215/215 [==============================] - 24s 110ms/step - loss: 0.0986 - mae_loss: 0.0444 - val_loss: 0.1106 - val_mae_loss: 0.0624\n",
      "Epoch 6/10\n",
      "215/215 [==============================] - 24s 111ms/step - loss: 0.0835 - mae_loss: 0.0406 - val_loss: 0.0932 - val_mae_loss: 0.0552\n",
      "Epoch 7/10\n",
      "215/215 [==============================] - 24s 111ms/step - loss: 0.0716 - mae_loss: 0.0377 - val_loss: 0.0823 - val_mae_loss: 0.0522\n",
      "Epoch 8/10\n",
      "215/215 [==============================] - 24s 111ms/step - loss: 0.0626 - mae_loss: 0.0356 - val_loss: 0.0829 - val_mae_loss: 0.0588\n",
      "Epoch 9/10\n",
      "215/215 [==============================] - 23s 108ms/step - loss: 0.0553 - mae_loss: 0.0336 - val_loss: 0.0707 - val_mae_loss: 0.0512\n",
      "Epoch 10/10\n",
      "215/215 [==============================] - 23s 106ms/step - loss: 0.0497 - mae_loss: 0.0319 - val_loss: 0.0687 - val_mae_loss: 0.0525\n",
      "training took 253.1 seconds or 4.22 minutes\n"
     ]
    }
   ],
   "source": [
    "model_HyDRA.compile(optimizer='Nadam', loss='mae',metrics=[mae_loss])\n",
    "\n",
    "model_callbacks = [EarlyStopping(monitor='val_loss', patience=patience),\n",
    "         ModelCheckpoint(filepath=model_filename_save,\n",
    "                         monitor='val_mae_loss', save_best_only=True,mode='min')]  #plot_losses\n",
    "\n",
    "start_time = time.time()\n",
    "\n",
    "model_hist = model_HyDRA.fit(train_generator,epochs=n_epoch,verbose=verbose,\n",
    "                           callbacks=model_callbacks,\n",
    "                           validation_data=val_generator,\n",
    "                           workers=n_workers,\n",
    "                           max_queue_size=max_queue_size)\n",
    "end_time = time.time()  \n",
    "\n",
    "train_dur=end_time-start_time\n",
    "\n",
    "print('training took %1.1f seconds or %1.2f minutes'%(train_dur, train_dur/60))\n",
    "\n",
    "# Loss log plot      \n",
    "train_loss=model_hist.history['loss']\n",
    "val_loss=model_hist.history['val_loss']\n",
    "train_loss_pure=model_hist.history['mae_loss']\n",
    "val_loss_pure=model_hist.history['val_mae_loss']"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "41a280fd",
   "metadata": {},
   "source": [
    "## Evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "21ef096a",
   "metadata": {},
   "outputs": [],
   "source": [
    "normFact=0.075\n",
    "Drift_limit=normFact\n",
    "\n",
    "best_model=load_model(model_filename_save,custom_objects={'mae_loss': mae_loss})\n",
    "\n",
    "Y_pred_val=best_model.predict(val_generator)*normFact\n",
    "Y_pred_test=best_model.predict(test_generator)*normFact\n",
    "\n",
    "Y_val_opt=agg_Y(val_generator)\n",
    "Y_test_opt=agg_Y(test_generator)\n",
    "\n",
    "Y_val_opt=Y_val_opt*normFact\n",
    "Y_test_opt=Y_test_opt*normFact\n",
    "\n",
    "MAE_val,MSE_val=eval_r(Y_val_opt,Y_pred_val)\n",
    "MAE_test,MSE_test=eval_r(Y_test_opt,Y_pred_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "d17175d0",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x360 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Regression plots\n",
    "fig,ax=plt.subplots(1,2,figsize=(10,5))\n",
    "\n",
    "plt.subplots_adjust(wspace=0.25, hspace=0)\n",
    "\n",
    "ax[0].plot(Y_val_opt,Y_pred_val,'bx',markersize=1)\n",
    "ax[0].plot([0,1],[0,1],'-r')\n",
    "ax[0].set_xlim(0,Drift_limit)\n",
    "ax[0].set_ylim(0,Drift_limit)\n",
    "ax[0].set_title('Validation-MAE='+str(np.round(MAE_val,6)))\n",
    "\n",
    "ax[1].plot(Y_test_opt,Y_pred_test,'bx',markersize=1)\n",
    "ax[1].plot([0,1],[0,1],'-r')\n",
    "ax[1].set_xlim(0,Drift_limit)\n",
    "ax[1].set_ylim(0,Drift_limit)\n",
    "ax[1].set_title('Testing-MAE='+str(np.round(MAE_test,6)))\n",
    "\n",
    "for i in range(2):\n",
    "    ax[i].set_xlabel('Ground truth')\n",
    "    ax[i].set_ylabel('Prediction')\n",
    "plt.show()\n",
    "plt.close()"
   ]
  }
 ],
 "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": "3.8.11"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}