Baseline - LSTM-Skip AE.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "4b3c1f47-6e9c-4e56-a255-939f8178a82a",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import random\n",
    "\n",
    "os.environ[\"CUDA_DEVICE_ORDER\"] = \"PCI_BUS_ID\";\n",
    "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"1\"\n",
    "os.environ[\"TF_FORCE_GPU_ALLOW_GROWTH\"] = \"true\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "5ba60486-a465-4823-b8a5-d223f95fe93a",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import tensorflow as tf\n",
    "\n",
    "from evaluator import evaluate\n",
    "from data_loader import load_kdd_cup_urc, load_yahoo_A1, load_yahoo_A2, load_yahoo_A3, load_yahoo_A4, load_power_demand # Univariate Datasets\n",
    "from data_loader import load_nasa, load_ecg, load_gesture, load_smd # Multivariate Datasets\n",
    "\n",
    "from tqdm.notebook import tqdm\n",
    "from tensorflow.keras import Model\n",
    "from tensorflow.keras.layers import Dense, Input, LSTM, Reshape, Dropout, GaussianNoise, Concatenate, Lambda, RepeatVector\n",
    "\n",
    "# THESE LINES ARE FOR REPRODUCIBILITY\n",
    "random.seed(0)\n",
    "np.random.seed(0)\n",
    "tf.random.set_seed(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "ec0f1ab4-164a-4fab-82e3-e42045ec6cda",
   "metadata": {},
   "outputs": [],
   "source": [
    "def Skip_AE(seq_length, dim, N_RES=3, N_LAYERS=1, BATCH_SIZE=128):\n",
    "    tf.keras.backend.clear_session()\n",
    "    \n",
    "    en_inputs = []\n",
    "    shared_latents = []\n",
    "    for n in range(N_RES):\n",
    "        selected_t = [t for t in range(0, seq_length, 2**n)]\n",
    "        en_input = Input(shape=[len(selected_t), dim])\n",
    "        en_inputs.append(en_input)\n",
    "\n",
    "    de_outputs = []\n",
    "    for n in range(N_RES):\n",
    "        for l in range(N_LAYERS):\n",
    "            X = en_inputs[n] if l == 0 else X\n",
    "            X = GaussianNoise(0.5)(X) if l == 0 else X\n",
    "            X = LSTM(64, return_sequences=False if l + 1 == N_LAYERS else True)(X)\n",
    "        h = Dense(32, activation='relu', kernel_regularizer='l1')(X)\n",
    "        shared_latents.append(h)\n",
    "\n",
    "    for n in range(N_RES):\n",
    "        selected_t = [t for t in range(0, seq_length, 2**n)]\n",
    "        for l in range(N_LAYERS):\n",
    "            if l == 0:\n",
    "                X = Concatenate()(shared_latents)\n",
    "                X = Dense(128, activation='relu')(X)\n",
    "                X = Concatenate()([X, shared_latents[n]])\n",
    "                X = RepeatVector(len(selected_t))(X)\n",
    "            X = LSTM(64, return_sequences=False if l + 1 == N_LAYERS else True)(X)\n",
    "        rec_x = Dense(len(selected_t) * dim)(X)\n",
    "        rec_x = Reshape([len(selected_t), dim])(rec_x)\n",
    "        de_outputs.append(rec_x)\n",
    "\n",
    "    model = Model(inputs=en_inputs, outputs=de_outputs)\n",
    "    model.compile(optimizer=tf.keras.optimizers.Adam(learning_rate=0.001), loss=\"mse\")\n",
    "\n",
    "    return model"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "aceba82a-7e10-4058-92aa-9b12874ff37d",
   "metadata": {},
   "source": [
    "### Yahoo S5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "96ebab6a-0bb8-466c-bac2-1c15fb237ae3",
   "metadata": {},
   "outputs": [],
   "source": [
    "total_scores = {'dataset': [], 'f1': [], 'pr_auc': [], 'roc_auc': []}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "bfb543b5-1d55-4b47-ae21-e8bdeafaa8d9",
   "metadata": {
    "scrolled": true,
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "3b3fd64f0c9e436b8178d2c957c00f3d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/67 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "dbe5f11370c64957a959aa36f15448f5",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/56 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "yahoo_A1 0.999999948484851 0.9696969626562632 0.9999999981470107\n",
      "yahoo_A1 0.9999999416666693 0.9166666376815789 0.9999999913849765\n",
      "yahoo_A1 0.9999999456521764 0.956521720708995 0.9999999953614762\n",
      "yahoo_A1 0.999999930000003 0.7999999480000032 0.9999999797237573\n",
      "yahoo_A1 0.9230768591716004 0.7179109509467508 0.9009920503704727\n",
      "yahoo_A1 0.9230768591716004 0.717660255534233 0.9015872879775448\n",
      "yahoo_A1 0.999999948484851 0.969696962696227 0.9999999981601732\n",
      "yahoo_A1 0.9999999492063519 0.9920634872478045 0.9999999987844084\n",
      "yahoo_A1 0.9999999492063519 0.992063487180122 0.9999999987844084\n",
      "yahoo_A1 0.9999999487179513 0.9487179430329566 0.999999998371928\n",
      "yahoo_A1 0.9963099123650302 0.9925379201830041 0.9999673670680758\n",
      "yahoo_A1 0.9999999487654349 0.9876543142127767 0.9999999984095602\n",
      "yahoo_A1 0.9999999483870992 0.9677419277058845 0.9999999980537634\n",
      "yahoo_A1 0.999999944444447 0.8888888716321159 0.9999999941537466\n",
      "yahoo_A1 0.9999999487013013 0.9870129799630335 0.9999999983645983\n",
      "yahoo_A1 0.99999994761905 0.9285714200500658 0.9999999973104056\n",
      "yahoo_A1 0.559055077611758 0.24571750435146386 0.1623758540906115\n",
      "yahoo_A1 0.6879999494400036 0.7580728942309235 0.9193531678202171\n",
      "yahoo_A1 0.49795914607247255 0.14451960189825416 0.3530277536632994\n",
      "yahoo_A1 0.972111503011066 0.9771650749879865 0.9858651437662392\n",
      "yahoo_A1 0.999999942307695 0.9230768956557769 0.9999999920306839\n",
      "yahoo_A1 0.9999999388888916 0.888888853752206 0.9999999886149165\n",
      "yahoo_A1 0.8421052132077591 0.8839499576222681 0.9448888572448941\n",
      "yahoo_A1 0.950494998921677 0.9804748054122133 0.9883727154835323\n",
      "yahoo_A1 0.999999930000003 0.3999999816666673 0.9999999797237573\n",
      "yahoo_A1 0.9999999493288617 0.9932885865270504 0.9999999988637428\n",
      "yahoo_A1 0.682730875243951 0.5852011705015822 0.8460790381777903\n",
      "yahoo_A1 0.965909039837296 0.949167005695869 0.9566763537025791\n",
      "yahoo_A1 0.422163553960223 0.27260545513334145 0.5984858726469914\n",
      "yahoo_A1 0.9999999472222247 0.9444444322916667 0.9999999969154737\n",
      "yahoo_A1 0.8806940933498365 0.5949292927651182 0.6698875499451914\n",
      "yahoo_A1 0.9999999487179513 0.9615384560018861 0.9999999983707265\n",
      "yahoo_A1 0.9999999482758646 0.49999999788900257 0.9999999979532814\n",
      "yahoo_A1 0.408488030938094 0.13436015162707046 0.37320754425802694\n",
      "yahoo_A1 0.9999999484375026 0.9687499930224395 0.9999999981085526\n",
      "yahoo_A1 0.8813558817836536 0.9320057377109398 0.9520360445425892\n",
      "yahoo_A1 0.9999999488235322 0.9882352876324428 0.9999999984651065\n",
      "yahoo_A1 0.9999999333333361 0.8333332883333356 0.9999999830563252\n",
      "yahoo_A1 0.9999999488235322 0.9882352876346604 0.9999999984651065\n",
      "yahoo_A1 0.980842861144143 0.9832448556437186 0.9927460616216617\n",
      "yahoo_A1 0.9999999487013013 0.9870129799841798 0.9999999983645983\n",
      "yahoo_A1 0.6356588659665922 0.6746402076825738 0.7860966819637931\n",
      "yahoo_A1 0.7314814308727743 0.8283659577839725 0.8817793295187937\n",
      "yahoo_A1 0.49795914607247255 0.008000001056772411 -0.0\n",
      "yahoo_A1 0.99999994761905 0.9761904621430798 0.9999999973104056\n",
      "yahoo_A1 0.9230768591716004 0.7182144735237114 0.9156745889661677\n",
      "yahoo_A1 0.9863013185213011 0.9258615894641946 0.9977702685893419\n",
      "yahoo_A1 0.0 0.0 -0.0\n",
      "yahoo_A1 0.9999999428571453 0.9285714024576228 0.9999999921626985\n",
      "yahoo_A1 0.9999999357142885 0.8571428160204102 0.999999985436508\n",
      "yahoo_A1 0.9147981554585882 0.8831849333715565 0.9480777008367671\n",
      "yahoo_A1 0.9999999490384642 0.9727272677892265 0.9999999986508749\n",
      "yahoo_A1 0.9999999473684236 0.9473684106129373 0.999999997059779\n",
      "yahoo_A1 0.9999999490000026 0.9899999943215703 0.9999999986183205\n",
      "yahoo_A1 0.999999948571431 0.9571428510492783 0.9999999982289627\n",
      "yahoo_A1 0.9999999484615409 0.9692307623089043 0.9999999981379137\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "eb96de0db0ea415e9629cea01cc34352",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/100 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9e8afca699ed4a7aa6496799f6363704",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/100 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "yahoo_A2 0.9999999485074653 0.9850746190014044 0.9999999983492349\n",
      "yahoo_A2 0.9999999485074653 0.9850746189412825 0.9999999983492349\n",
      "yahoo_A2 0.9999999485074653 0.9701492469331415 0.9999999983492349\n",
      "yahoo_A2 0.9999999490384642 0.9903846098918182 0.9999999988797315\n",
      "yahoo_A2 0.9999998500000123 0.0 0.999999899784027\n",
      "yahoo_A2 0.9999999485074653 0.9701492470117672 0.9999999983492349\n",
      "yahoo_A2 0.9999999490384642 0.9903846098645308 0.9999999988797315\n",
      "yahoo_A2 0.9999998500000123 0.0 0.9999998997840271\n",
      "yahoo_A2 0.8311687809327065 0.7623752206265004 0.9793241056949096\n",
      "yahoo_A2 0.9999999490384642 0.9903846098915461 0.9999999988797315\n",
      "yahoo_A2 0.9999998500000123 0.0 0.9999998997840271\n",
      "yahoo_A2 0.9999999485074653 0.9850746190043864 0.999999998349235\n",
      "yahoo_A2 0.9999999490384642 0.9903846098471291 0.9999999988797315\n",
      "yahoo_A2 0.5549737718045055 0.607229213182199 0.8832395847471731\n",
      "yahoo_A2 0.9999998500000123 0.0 0.9999998997840271\n",
      "yahoo_A2 0.8133332828177808 0.7653282635156582 0.9814259383005461\n",
      "yahoo_A2 0.9999999490384642 0.9903846098911262 0.9999999988797315\n",
      "yahoo_A2 0.11764704636678214 0.03124999667968783 0.9676024948234216\n",
      "yahoo_A2 0.9999999485074653 0.9850746190038604 0.9999999983492349\n",
      "yahoo_A2 0.3597121920190506 0.3523007883857501 0.6737164308473812\n",
      "yahoo_A2 0.9999998500000123 0.0 0.9999998997840271\n",
      "yahoo_A2 0.9999999485074653 0.9701492469373142 0.9999999983492349\n",
      "yahoo_A2 0.9999999490384642 0.9903846098808915 0.9999999988797316\n",
      "yahoo_A2 0.9999998500000123 0.0 0.999999899784027\n",
      "yahoo_A2 0.9999998500000123 0.0 0.999999899784027\n",
      "yahoo_A2 0.9999999485074653 0.9850746190034834 0.9999999983492349\n",
      "yahoo_A2 0.6119873372289537 0.6087393461458541 0.9186355301064127\n",
      "yahoo_A2 0.9999998500000123 0.0 0.9999998997840271\n",
      "yahoo_A2 0.6168223863175852 0.534788764758384 0.9396254471547371\n",
      "yahoo_A2 0.9999999490384642 0.9807692259534122 0.9999999988797315\n",
      "yahoo_A2 0.008097165178908029 0.002032520121125012 0.47300211242484225\n",
      "yahoo_A2 0.9999999485074653 0.9701492469923376 0.9999999983492349\n",
      "yahoo_A2 0.5211267206308301 0.45311241378154016 0.6106771847458219\n",
      "yahoo_A2 0.9999998500000123 0.0 0.9999998997840271\n",
      "yahoo_A2 0.6702127193696275 0.7271647939088228 0.9625212528981618\n",
      "yahoo_A2 0.2143999807897617 0.11495849186512289 0.5527999291008708\n",
      "yahoo_A2 0.9999999490384642 0.9903846098835932 0.9999999988797315\n",
      "yahoo_A2 0.9999998500000123 0.0 0.9999998997840269\n",
      "yahoo_A2 0.9999999485074653 0.9850746190043695 0.999999998349235\n",
      "yahoo_A2 0.9999999490384642 0.9807692260024056 0.9999999988797316\n",
      "yahoo_A2 0.9999998500000123 0.0 0.999999899784027\n",
      "yahoo_A2 0.7924527804279923 0.8306478668067271 0.9827838638603678\n",
      "yahoo_A2 0.9999999490384642 0.9903846098914703 0.9999999988797315\n",
      "yahoo_A2 0.9999998500000123 0.0 0.9999998997840271\n",
      "yahoo_A2 0.9999999485074653 0.9701492469227789 0.9999999983492349\n",
      "yahoo_A2 0.9999999490384642 0.9903846098871513 0.9999999988797315\n",
      "yahoo_A2 0.38805967007128783 0.18106836582114197 0.6706578136564976\n",
      "yahoo_A2 0.9999998500000123 0.0 0.9999998997840273\n",
      "yahoo_A2 0.26746504658388015 0.12812265023015942 0.6378471555690177\n",
      "yahoo_A2 0.9999999490384642 0.9807692260067277 0.9999999988797315\n",
      "yahoo_A2 0.18181816198347142 0.049999994500000554 0.9805614568076637\n",
      "yahoo_A2 0.9999999485074653 0.9850746190033579 0.9999999983492348\n",
      "yahoo_A2 0.9999999490384642 0.9903846097151461 0.9999999988797315\n",
      "yahoo_A2 0.9999998500000123 0.0 0.9999998997840273\n",
      "yahoo_A2 0.9999999485074653 0.9701492470071083 0.9999999983492349\n",
      "yahoo_A2 0.9999999490384642 0.9903846098921186 0.9999999988797315\n",
      "yahoo_A2 0.9999998500000123 0.0 0.9999998997840271\n",
      "yahoo_A2 0.9999998500000123 0.0 0.9999998997840271\n",
      "yahoo_A2 0.9999999485074653 0.9850746190027357 0.9999999983492351\n",
      "yahoo_A2 0.9999999490384642 0.9903846098817632 0.9999999988797315\n",
      "yahoo_A2 0.9999998500000123 0.0 0.9999998997840271\n",
      "yahoo_A2 0.9999999485074653 0.9850746190034505 0.9999999983492349\n",
      "yahoo_A2 0.9999999490384642 0.990384609719219 0.9999999988797316\n",
      "yahoo_A2 0.9999998500000123 0.0 0.9999998997840271\n",
      "yahoo_A2 0.9999999485074653 0.9850746186763568 0.9999999983492349\n",
      "yahoo_A2 0.8128341743601506 0.8911485354366715 0.9773427949979099\n",
      "yahoo_A2 0.9999998500000123 0.0 0.9999998997840271\n",
      "yahoo_A2 0.9999999485074653 0.9701492470113248 0.999999998349235\n",
      "yahoo_A2 0.9999999485074653 0.9701492469227367 0.9999999983492349\n",
      "yahoo_A2 0.9999999490384642 0.9903846098879514 0.9999999988797316\n",
      "yahoo_A2 0.9999998500000123 0.0 0.999999899784027\n",
      "yahoo_A2 0.9999999485074653 0.9850746190040556 0.9999999983492349\n",
      "yahoo_A2 0.9999999490384642 0.9903846098717968 0.9999999988797315\n",
      "yahoo_A2 0.9999998500000123 0.0 0.9999998997840271\n",
      "yahoo_A2 0.9999999485074653 0.9850746190004646 0.9999999983492349\n",
      "yahoo_A2 0.9999999490384642 0.9711538417826937 0.9999999988797315\n",
      "yahoo_A2 0.049999994875000016 0.012820511505588559 0.9179264738838262\n",
      "yahoo_A2 0.9999999485074653 0.9402985018268922 0.9999999983492349\n",
      "yahoo_A2 0.8278688028688553 0.8951563700295077 0.9828373004862598\n",
      "yahoo_A2 0.5170067606275196 0.5377918073753934 0.7902854081500773\n",
      "yahoo_A2 0.9999998500000123 0.0 0.9999998997840271\n",
      "yahoo_A2 0.2718052502367855 0.12593509217795262 0.6479479875350103\n",
      "yahoo_A2 0.9999999490384642 0.9903846097017209 0.9999999988797315\n",
      "yahoo_A2 0.9999998500000123 0.0 0.9999998997840271\n",
      "yahoo_A2 0.9999999485074653 0.9850746190033 0.9999999983492347\n",
      "yahoo_A2 0.9999999490384642 0.9711538417486254 0.9999999988797316\n",
      "yahoo_A2 0.9999998500000123 0.0 0.999999899784027\n",
      "yahoo_A2 0.9999999485074653 0.9701492470105573 0.9999999983492349\n",
      "yahoo_A2 0.9999999490384642 0.9903846097300855 0.9999999988797316\n",
      "yahoo_A2 0.9999998500000123 0.0 0.999999899784027\n",
      "yahoo_A2 0.9999998500000123 0.0 0.9999998997840271\n",
      "yahoo_A2 0.9999999485074653 0.970149246914438 0.9999999983492349\n",
      "yahoo_A2 0.6204818843119496 0.31827584072486925 0.8509233812201196\n",
      "yahoo_A2 0.9999998500000123 0.0 0.9999998997840271\n",
      "yahoo_A2 0.49242420417527843 0.24585066073485246 0.8498134314329786\n",
      "yahoo_A2 0.9999999490384642 0.9711538417848923 0.9999999988797315\n",
      "yahoo_A2 0.18181816198347142 0.049999994500000554 0.9816413703063506\n",
      "yahoo_A2 0.9999999485074653 0.9850746190018934 0.999999998349235\n",
      "yahoo_A2 0.9999999490384642 0.9711538417878075 0.9999999988797315\n",
      "yahoo_A2 0.9999998500000123 0.0 0.999999899784027\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "039fe09f74aa48838667589fa8feddd6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/100 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "add308dba9554b55a4d7c87d0c2687a5",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/92 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "yahoo_A3 0.25396823169564314 0.002645503616392491 -0.0\n",
      "yahoo_A3 0.442477841491114 0.20191682533099842 0.32977436223327566\n",
      "yahoo_A3 0.9176470084429093 0.9618512565169975 0.9801721386285549\n",
      "yahoo_A3 0.25396823169564314 0.001326260929714058 -0.0\n",
      "yahoo_A3 0.7947367916925241 0.8802454593135991 0.878443968785417\n",
      "yahoo_A3 0.41726615387920135 0.10342506586890683 0.1382768966292481\n",
      "yahoo_A3 0.5852089616298454 0.2528003865626263 0.12137194905012542\n",
      "yahoo_A3 0.45825928955576367 0.08395758234720899 0.20264433160661557\n",
      "yahoo_A3 0.5210083646578659 0.5494926202049178 0.6856728943504886\n",
      "yahoo_A3 0.7134502462945892 0.5707750142263852 0.46812132669509215\n",
      "yahoo_A3 0.637770854195864 0.5764878928749207 0.4986632723539781\n",
      "yahoo_A3 0.5783521396394755 0.5691513732622765 0.5177003158525797\n",
      "yahoo_A3 0.24999997799202311 0.08685868109503675 0.41714038122710306\n",
      "yahoo_A3 0.6766916843462069 0.5293994525114056 0.2981902005618742\n",
      "yahoo_A3 0.11563168070466742 0.0012077301733057306 -0.0\n",
      "yahoo_A3 0.8495574722844418 0.930833193381485 0.9430931398461644\n",
      "yahoo_A3 0.25396823169564314 0.014908147401721975 0.2147356452102096\n",
      "yahoo_A3 0.2694736607805006 0.08808042747156497 0.15839837124866094\n",
      "yahoo_A3 0.2812499757202169 0.001355014555048782 -0.0\n",
      "yahoo_A3 0.26845635247661714 0.09458148573144066 0.2931002879205925\n",
      "yahoo_A3 0.23578945278581892 0.05814925216349081 0.09912261935245899\n",
      "yahoo_A3 0.45070419027970904 0.0015974454052876353 -0.0\n",
      "yahoo_A3 0.6075948942156735 0.002008033917585507 -0.0\n",
      "yahoo_A3 0.9753085907788475 0.9880379581332805 0.9991706490258453\n",
      "yahoo_A3 0.25396823169564314 0.002645503410977035 -0.0\n",
      "yahoo_A3 0.8732393858008801 0.9374793287709823 0.9758888198759064\n",
      "yahoo_A3 0.8648648144124296 0.9144159624942634 0.9738853327067811\n",
      "yahoo_A3 0.25396823169564314 0.0013262608998209863 -0.0\n",
      "yahoo_A3 0.4133332890666713 0.218749949766052 0.5699843775037111\n",
      "yahoo_A3 0.845238045374507 0.8977102585638608 0.8786897498878892\n",
      "yahoo_A3 0.8916876077927048 0.7490479721931574 0.42101352738203496\n",
      "yahoo_A3 0.7726638296349493 0.6032002028009235 0.5589565721491722\n",
      "yahoo_A3 0.5357736710795402 0.12848993018211155 0.03678145205665377\n",
      "yahoo_A3 0.8151658280126262 0.8823885616594822 0.7899789511585751\n",
      "yahoo_A3 0.6075948942156735 0.5859796693175285 0.6809419972283167\n",
      "yahoo_A3 0.7356321371911775 0.604946756850655 0.6716031203806837\n",
      "yahoo_A3 0.7826086470197939 0.7952732180256625 0.7330283124062962\n",
      "yahoo_A3 0.7105262693234967 0.758935609604023 0.6881098803429669\n",
      "yahoo_A3 0.4671532345463318 0.2602764988443979 0.5154068103216489\n",
      "yahoo_A3 0.7978141594762488 0.8971920579339038 0.8602978368060883\n",
      "yahoo_A3 0.999999947674421 0.9767441748075016 0.9999999974225294\n",
      "yahoo_A3 0.8478802499860103 0.9300091160586706 0.9194943329746619\n",
      "yahoo_A3 0.25396823169564314 0.001326260909711886 -0.0\n",
      "yahoo_A3 0.32621356480573327 0.04506479594974547 0.39782376102744504\n",
      "yahoo_A3 0.7153284209963267 0.5070804976675628 0.3267268079520557\n",
      "yahoo_A3 0.7743732113825961 0.30583492704886445 0.22409622498659865\n",
      "yahoo_A3 0.9094269368684703 0.9642672867089358 0.9310901923873263\n",
      "yahoo_A3 0.6846038411666818 0.5514608605297446 0.5842293570458812\n",
      "yahoo_A3 0.25396823169564314 0.06773987749333843 0.3519312535451615\n",
      "yahoo_A3 0.4626865181666347 0.27164239232660276 0.6485036624093168\n",
      "yahoo_A3 0.9961088986419199 0.9921266967842028 0.9999749588341622\n",
      "yahoo_A3 0.5874798941145886 0.0024390263141050744 0.20622567976422532\n",
      "yahoo_A3 0.5568861843271574 0.33934868438246274 0.6463633139467323\n",
      "yahoo_A3 0.7268907092816569 0.5374721414557928 0.6699299686914028\n",
      "yahoo_A3 0.8737200859229606 0.9415095029262031 0.9618267168497812\n",
      "yahoo_A3 0.9999999491803305 0.9918032738707058 0.9999999988658624\n",
      "yahoo_A3 0.6258064017419395 0.7426581612512617 0.7947642769533606\n",
      "yahoo_A3 0.9999999484375026 0.9843749918323997 0.9999999981715426\n",
      "yahoo_A3 0.7283236529018707 0.3048980093765892 0.388548822580837\n",
      "yahoo_A3 0.7485379629014087 0.8294670497168102 0.8028548779768706\n",
      "yahoo_A3 0.7914893112322349 0.8504334427004927 0.8869212170268725\n",
      "yahoo_A3 0.9142856640000028 0.8877827187627392 0.848945973374245\n",
      "yahoo_A3 0.2935779564683128 0.10693702672752328 0.3606008969470396\n",
      "yahoo_A3 0.6943619722688439 0.736222987984001 0.6455717116131277\n",
      "yahoo_A3 0.700147664776179 0.333834454296343 0.30512354754429327\n",
      "yahoo_A3 0.7386934171788625 0.8322180697624961 0.7917697367296535\n",
      "yahoo_A3 0.8541666175075981 0.9467590583611669 0.8669384271856395\n",
      "yahoo_A3 0.7617187025413543 0.73888223858645 0.69029206015716\n",
      "yahoo_A3 0.9470587731124593 0.948703920658218 0.9474733815617395\n",
      "yahoo_A3 0.7433627822243138 0.7640124320436588 0.7105578792531556\n",
      "yahoo_A3 0.7935222165647723 0.8890880726058175 0.9468660667195257\n",
      "yahoo_A3 0.995433739046311 0.9904368217976183 0.9998347095318807\n",
      "yahoo_A3 0.9929077507167674 0.9880393150770327 0.9964416648020962\n",
      "yahoo_A3 0.3301707502968744 0.003464293392943838 0.19405098915097876\n",
      "yahoo_A3 0.7811633870811333 0.4156070251888201 0.2752943646574528\n",
      "yahoo_A3 0.3882783568463308 0.0014925385091889094 -0.0\n",
      "yahoo_A3 0.45070419027970904 0.17126612325852855 0.3469855588898277\n",
      "yahoo_A3 0.516020197625476 0.2331165801037688 0.4611756494928667\n",
      "yahoo_A3 0.8648648151100936 0.9005101567051795 0.9674569892593612\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9178f8583ee64fbfa6ff179d05577b3a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/100 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "17d477361a704ed99ff6cb1fc2061104",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/82 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "yahoo_A4 0.8395061234364709 0.8719433981570495 0.8918924267150741\n",
      "yahoo_A4 0.26890753963703323 0.001445087784987022 0.08510638245625994\n",
      "yahoo_A4 0.41726615387920135 0.12327932642155423 0.25545155534250813\n",
      "yahoo_A4 0.31067958525780215 0.06936637400920384 0.2252996998855609\n",
      "yahoo_A4 0.7987804392214485 0.8422668820505975 0.8346436903288543\n",
      "yahoo_A4 0.7927272242988459 0.8273889172319309 0.8673110560454775\n",
      "yahoo_A4 0.6483516038220051 0.582713660229045 0.623102983301391\n",
      "yahoo_A4 0.666666616021054 0.6480415155772209 0.7278587073850769\n",
      "yahoo_A4 0.6084033181634095 0.3038297415407498 0.40175080454891926\n",
      "yahoo_A4 0.5289255807117028 0.5615790172673154 0.6484249792535381\n",
      "yahoo_A4 0.6795095860606518 0.3964768449513648 0.48440195070593517\n",
      "yahoo_A4 0.43902435490051145 0.20030710985399902 0.5644750298921495\n",
      "yahoo_A4 0.45070419027970904 0.0031847144476671936 -0.0\n",
      "yahoo_A4 0.37551017342774085 0.19653340959560067 0.5187280294107938\n",
      "yahoo_A4 0.07439824225732532 0.0011792465267388328 -0.0\n",
      "yahoo_A4 0.903703652499317 0.9521150284767764 0.9940367335547866\n",
      "yahoo_A4 0.5942491593197877 0.3646800849601076 0.33955508220938563\n",
      "yahoo_A4 0.6075948942156735 0.1324138299602586 0.19683393238282004\n",
      "yahoo_A4 0.7675069552887851 0.5673601934358616 0.42923202293649865\n",
      "yahoo_A4 0.8036253284398676 0.9052063334604891 0.9152975528590284\n",
      "yahoo_A4 0.3468207803802355 0.11878113183068806 0.5327892436148424\n",
      "yahoo_A4 0.45070419027970904 0.11963041576569698 0.07471746981613442\n",
      "yahoo_A4 0.5714285304267189 0.5070362769890528 0.46675767625919284\n",
      "yahoo_A4 0.3431635102617019 0.12917780284647962 0.5320266219836097\n",
      "yahoo_A4 0.9999999492187527 0.9843749959419914 0.9999999988982372\n",
      "yahoo_A4 0.5454545056048113 0.0020080337933514594 0.09818181798365472\n",
      "yahoo_A4 0.8328911978596936 0.836410775533148 0.6077591839401297\n",
      "yahoo_A4 0.25396823169564314 0.029014689862370848 0.01032577655988163\n",
      "yahoo_A4 0.709459411776391 0.8073904018612831 0.808050131773425\n",
      "yahoo_A4 0.7726638296349493 0.6166143526502688 0.5779058996490797\n",
      "yahoo_A4 0.9999999452380977 0.9523809329785107 0.9999999949994317\n",
      "yahoo_A4 0.4086798950024387 0.10000088621949269 0.3253205491251946\n",
      "yahoo_A4 0.25396823169564314 0.0014577269749584312 0.09042553134962256\n",
      "yahoo_A4 0.6279069329367255 0.6339387529884357 0.7736954095177897\n",
      "yahoo_A4 0.6666666220202049 0.6541691319522129 0.5863899246355697\n",
      "yahoo_A4 0.6767240897796556 0.5398712913068509 0.643210667611601\n",
      "yahoo_A4 0.7133332839333366 0.7170465197382546 0.634250488705144\n",
      "yahoo_A4 0.6075948942156735 0.6122602079370207 0.7062279415414651\n",
      "yahoo_A4 0.5027522557562523 0.17312388687990382 0.43495180417308754\n",
      "yahoo_A4 0.4074073742112509 0.3683900649717326 0.5218207718335217\n",
      "yahoo_A4 0.8590603531012144 0.8978262017126899 0.9083221092290189\n",
      "yahoo_A4 0.03899721063927221 0.01134870359761948 0.34460573567711894\n",
      "yahoo_A4 0.5806451198959446 0.3973618829200167 0.6898670473774708\n",
      "yahoo_A4 0.3035019196626768 0.04003738243602085 0.16417103155043852\n",
      "yahoo_A4 0.3703703400548721 0.09998509236779254 0.35692286927784916\n",
      "yahoo_A4 0.9105690542930823 0.9268212564114054 0.9813621990034934\n",
      "yahoo_A4 0.6666666215277807 0.5988150446784359 0.6196202973871507\n",
      "yahoo_A4 0.2565130035879393 0.016306329858393233 0.41133944526544336\n",
      "yahoo_A4 0.46422334988015346 0.13450294505859517 0.12256556974940988\n",
      "yahoo_A4 0.45070419027970904 0.17967563859413743 0.31029186340710013\n",
      "yahoo_A4 0.8630136491743319 0.9197432933209804 0.8791359574472202\n",
      "yahoo_A4 0.8909090405179092 0.8942733988900422 0.8457180933191244\n",
      "yahoo_A4 0.25396823169564314 0.0043976729215489665 0.0017730495858168443\n",
      "yahoo_A4 0.9999999484375026 0.9843749918410563 0.9999999981715427\n",
      "yahoo_A4 0.731249950667972 0.7531196235167853 0.8536093576925967\n",
      "yahoo_A4 0.8310501778528417 0.9089981057984761 0.9683143032183102\n",
      "yahoo_A4 0.6444444000493856 0.5672883184680169 0.5746580160496177\n",
      "yahoo_A4 0.3833865458012261 0.1845052291632514 0.6097516222214568\n",
      "yahoo_A4 0.7071822739476847 0.4190123098837331 0.9021359690951637\n",
      "yahoo_A4 0.5919999581337628 0.27754594664198556 0.3175193920709166\n",
      "yahoo_A4 0.45070419027970904 0.0015974454070187608 -0.0\n",
      "yahoo_A4 0.6982248064178453 0.3831220832477199 0.28170961827283547\n",
      "yahoo_A4 0.6578946922610833 0.742472706228115 0.8465446697238175\n",
      "yahoo_A4 0.7485379644198245 0.7896199882373152 0.8484316516307625\n",
      "yahoo_A4 0.25396823169564314 0.07294852160167213 0.06010832111012726\n",
      "yahoo_A4 0.6528496925125541 0.7310166811110468 0.8110803946064234\n",
      "yahoo_A4 0.9999999491869945 0.9918699139164027 0.9999999988715345\n",
      "yahoo_A4 0.2565130035879393 0.078044350077912 0.035654920147572475\n",
      "yahoo_A4 0.5919999581337628 0.5026040370132656 0.602510867706253\n",
      "yahoo_A4 0.7857142355333746 0.8885523425343109 0.8681026287872197\n",
      "yahoo_A4 0.25049699588078084 0.001322752280439461 -0.0\n",
      "yahoo_A4 0.772727224376036 0.8530695693778956 0.7483027527515591\n",
      "yahoo_A4 0.7216035169051769 0.5996017946370302 0.7406284325502787\n",
      "yahoo_A4 0.6544342065351805 0.2996477090700349 0.28829705522691207\n",
      "yahoo_A4 0.45070419027970904 0.22684172310900774 0.48886826562544805\n",
      "yahoo_A4 0.6274509302575972 0.3897995069502271 0.5236308436085891\n",
      "yahoo_A4 0.45070419027970904 0.26395452043742546 0.6657043600502113\n",
      "yahoo_A4 0.36701027864215363 0.1070781642248383 0.27166348480791574\n",
      "yahoo_A4 0.46460976452646996 0.1673220288293463 0.1726226584904566\n",
      "yahoo_A4 0.8637680657945839 0.7277708647748674 0.8906893946783117\n",
      "yahoo_A4 0.6419752648529217 0.08081960565381982 0.08357967316761744\n",
      "yahoo_A4 0.2608695424299727 0.0013333343090302626 -0.0\n"
     ]
    }
   ],
   "source": [
    "for loader in [load_yahoo_A1, load_yahoo_A2, load_yahoo_A3, load_yahoo_A4]:\n",
    "    datasets = loader(64, 1)\n",
    "    x_trains, x_tests, y_tests = datasets['x_train'], datasets['x_test'], datasets['y_test']\n",
    "    \n",
    "    for i in tqdm(range(len(x_trains))):\n",
    "        X_train = x_trains[i]\n",
    "        X_test = x_tests[i]\n",
    "        \n",
    "        N_RES = 3\n",
    "\n",
    "        seq_length, dim = X_train.shape[1], X_train.shape[2]\n",
    "        X_train_reverse = np.flip(X_train, axis=1)\n",
    "        X_test_reverse = np.flip(X_test, axis=1)\n",
    "\n",
    "        X_train_by_res = []\n",
    "        X_test_by_res = []\n",
    "        X_train_by_res_reverse = []\n",
    "        \n",
    "        for n in range(N_RES):\n",
    "            selected_t = [t for t in range(0, seq_length, 2**n)]\n",
    "            X_train_by_res.append(X_train[:, selected_t, :])\n",
    "            X_test_by_res.append(X_test[:, selected_t, :])\n",
    "            X_train_by_res_reverse.append(X_train_reverse[:, selected_t, :])\n",
    "        \n",
    "        model = Skip_AE(seq_length, dim, N_RES, N_LAYERS=2, BATCH_SIZE=128)\n",
    "        history = model.fit(X_train_by_res, X_train_by_res_reverse, epochs=50, batch_size=128, validation_split=0.3, verbose=0,\n",
    "                            callbacks=[tf.keras.callbacks.EarlyStopping(monitor=\"val_loss\", patience=5, mode=\"min\", restore_best_weights=True)]) \n",
    "        \n",
    "        X_test_rec = [np.flip(rec, axis=1) for rec in model.predict(X_test_by_res)]\n",
    "        scores = evaluate(X_test, X_test_rec[0], y_tests[i], is_reconstructed=True)\n",
    "    \n",
    "        total_scores['dataset'].append(loader.__name__.replace('load_', ''))\n",
    "        total_scores['f1'].append(np.max(scores['f1']))\n",
    "        total_scores['pr_auc'].append(scores['pr_auc'])\n",
    "        total_scores['roc_auc'].append(scores['roc_auc'])\n",
    "        print(loader.__name__.replace('load_', ''), np.max(scores['f1']), scores['pr_auc'], scores['roc_auc'])   "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "ba871981-8985-4f94-ba02-f11bee44bf97",
   "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>f1</th>\n",
       "      <th>pr_auc</th>\n",
       "      <th>roc_auc</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>dataset</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>yahoo_A1</th>\n",
       "      <td>0.897565</td>\n",
       "      <td>0.806933</td>\n",
       "      <td>0.894195</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>yahoo_A2</th>\n",
       "      <td>0.877235</td>\n",
       "      <td>0.573224</td>\n",
       "      <td>0.962131</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>yahoo_A3</th>\n",
       "      <td>0.640469</td>\n",
       "      <td>0.516348</td>\n",
       "      <td>0.558906</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>yahoo_A4</th>\n",
       "      <td>0.581571</td>\n",
       "      <td>0.433789</td>\n",
       "      <td>0.518512</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                f1    pr_auc   roc_auc\n",
       "dataset                               \n",
       "yahoo_A1  0.897565  0.806933  0.894195\n",
       "yahoo_A2  0.877235  0.573224  0.962131\n",
       "yahoo_A3  0.640469  0.516348  0.558906\n",
       "yahoo_A4  0.581571  0.433789  0.518512"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "yahoo_results = pd.DataFrame(total_scores)\n",
    "yahoo_results.groupby('dataset').mean()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ee4b544c-01d5-40dc-8c82-17002ca08ae8",
   "metadata": {},
   "source": [
    "### NASA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "cf10eaa5-4d09-4617-8ec2-004069465222",
   "metadata": {},
   "outputs": [],
   "source": [
    "total_scores = {'dataset': [], 'f1': [], 'pr_auc': [], 'roc_auc': []}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "eb19d991-dde5-49d1-b4db-52af438c7712",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7211dee26dca41df980177e17a0a08ce",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/2 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "13dc532f03b747228fbe28e416bc0940",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/2 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "D1 0.25300017462469454 0.13597482822023738 0.3967071569961004\n",
      "D2 0.2699530282282525 0.5840009300457375 0.5173913038147389\n"
     ]
    }
   ],
   "source": [
    "for loader in [load_nasa]:\n",
    "    datasets = loader(100, 100)\n",
    "    x_trains, x_tests, y_tests = datasets['x_train'], datasets['x_test'], datasets['y_test']\n",
    "    \n",
    "    for i in tqdm(range(len(x_trains))):\n",
    "        X_train = x_trains[i]\n",
    "        X_test = x_tests[i]\n",
    "\n",
    "        N_RES = 3\n",
    "\n",
    "        seq_length, dim = X_train.shape[1], X_train.shape[2]\n",
    "        X_train_reverse = np.flip(X_train, axis=1)\n",
    "        X_test_reverse = np.flip(X_test, axis=1)\n",
    "\n",
    "        X_train_by_res = []\n",
    "        X_test_by_res = []\n",
    "        X_train_by_res_reverse = []\n",
    "        \n",
    "        for n in range(N_RES):\n",
    "            selected_t = [t for t in range(0, seq_length, 2**n)]\n",
    "            X_train_by_res.append(X_train[:, selected_t, :])\n",
    "            X_test_by_res.append(X_test[:, selected_t, :])\n",
    "            X_train_by_res_reverse.append(X_train_reverse[:, selected_t, :])\n",
    "        \n",
    "        model = Skip_AE(seq_length, dim, N_RES, N_LAYERS=2, BATCH_SIZE=128)\n",
    "        history = model.fit(X_train_by_res, X_train_by_res_reverse, epochs=50, batch_size=128, validation_split=0.3, verbose=0,\n",
    "                            callbacks=[tf.keras.callbacks.EarlyStopping(monitor=\"val_loss\", patience=5, mode=\"min\", restore_best_weights=True)]) \n",
    "        \n",
    "        X_test_rec = [np.flip(rec, axis=1) for rec in model.predict(X_test_by_res)]\n",
    "        scores = evaluate(X_test, X_test_rec[0], y_tests[i], is_reconstructed=True)\n",
    "        \n",
    "        total_scores['dataset'].append(f'D{i+1}')\n",
    "        total_scores['f1'].append(np.max(scores['f1']))\n",
    "        total_scores['pr_auc'].append(scores['pr_auc'])\n",
    "        total_scores['roc_auc'].append(scores['roc_auc'])\n",
    "        print(f'D{i+1}', np.max(scores['f1']), scores['pr_auc'], scores['roc_auc'])   "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4358182d-2dd7-48b7-9c45-6b863091e963",
   "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>f1</th>\n",
       "      <th>pr_auc</th>\n",
       "      <th>roc_auc</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>dataset</th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>D1</th>\n",
       "      <td>0.253000</td>\n",
       "      <td>0.135975</td>\n",
       "      <td>0.396707</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>D2</th>\n",
       "      <td>0.269953</td>\n",
       "      <td>0.584001</td>\n",
       "      <td>0.517391</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               f1    pr_auc   roc_auc\n",
       "dataset                              \n",
       "D1       0.253000  0.135975  0.396707\n",
       "D2       0.269953  0.584001  0.517391"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nasa_results = pd.DataFrame(total_scores)\n",
    "nasa_results.groupby('dataset').mean()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "96d211db-832e-4e04-98d2-f24cd026d8e4",
   "metadata": {},
   "source": [
    "### SMD"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "42a32fb1-9d33-401a-99ca-b5b88fec0ff2",
   "metadata": {},
   "outputs": [],
   "source": [
    "total_scores = {'dataset': [], 'f1': [], 'pr_auc': [], 'roc_auc': []}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "01e9f650-8648-4626-a041-52daf5aa1f0a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "09ffb296c40e44e69108f258d33fa533",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/28 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b828bd4a315348f1b674d1c90e2eabc0",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/28 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for loader in [load_smd]:\n",
    "    datasets = loader(64, 1)\n",
    "    x_trains, x_tests, y_tests = datasets['x_train'], datasets['x_test'], datasets['y_test']\n",
    "    \n",
    "    for i in tqdm(range(len(x_trains))):\n",
    "        X_train = x_trains[i]\n",
    "        X_test = x_tests[i]\n",
    "\n",
    "        N_RES = 3\n",
    "\n",
    "        seq_length, dim = X_train.shape[1], X_train.shape[2]\n",
    "        X_train_reverse = np.flip(X_train, axis=1)\n",
    "        X_test_reverse = np.flip(X_test, axis=1)\n",
    "\n",
    "        X_train_by_res = []\n",
    "        X_test_by_res = []\n",
    "        X_train_by_res_reverse = []\n",
    "        \n",
    "        for n in range(N_RES):\n",
    "            selected_t = [t for t in range(0, seq_length, 2**n)]\n",
    "            X_train_by_res.append(X_train[:, selected_t, :])\n",
    "            X_test_by_res.append(X_test[:, selected_t, :])\n",
    "            X_train_by_res_reverse.append(X_train_reverse[:, selected_t, :])\n",
    "        \n",
    "        model = Skip_AE(seq_length, dim, N_RES, N_LAYERS=2, BATCH_SIZE=128)\n",
    "        history = model.fit(X_train_by_res, X_train_by_res_reverse, epochs=50, batch_size=128, validation_split=0.3, verbose=0,\n",
    "                            callbacks=[tf.keras.callbacks.EarlyStopping(monitor=\"val_loss\", patience=5, mode=\"min\", restore_best_weights=True)]) \n",
    "        \n",
    "        X_test_rec = [np.flip(rec, axis=1) for rec in model.predict(X_test_by_res)]\n",
    "        scores = evaluate(X_test, X_test_rec[0], y_tests[i], is_reconstructed=True)\n",
    "      \n",
    "        total_scores['dataset'].append(loader.__name__.replace('load_', ''))\n",
    "        total_scores['f1'].append(np.max(scores['f1']))\n",
    "        total_scores['pr_auc'].append(scores['pr_auc'])\n",
    "        total_scores['roc_auc'].append(scores['roc_auc'])\n",
    "        print(loader.__name__.replace('load_', ''), np.max(scores['f1']), scores['pr_auc'], scores['roc_auc'])   "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8091c8f3-f45f-4ed8-bbb8-fa3fbf49431d",
   "metadata": {},
   "outputs": [],
   "source": [
    "smd_results = pd.DataFrame(total_scores)\n",
    "smd_results.groupby('dataset').mean()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c4741e23-8ab0-419f-a479-66c210cc4fe2",
   "metadata": {},
   "source": [
    "### ECG"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "0f33cfd9-fbd6-4a66-9201-22ba9c607f07",
   "metadata": {},
   "outputs": [],
   "source": [
    "total_scores = {'dataset': [], 'f1': [], 'pr_auc': [], 'roc_auc': []}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "52b8855a-3d98-479d-beb7-8ed2d4728a78",
   "metadata": {},
   "outputs": [],
   "source": [
    "for loader in [load_ecg]:\n",
    "    datasets = loader(32, 16)\n",
    "    x_trains, x_tests, y_tests = datasets['x_train'], datasets['x_test'], datasets['y_test']\n",
    "    \n",
    "    for i in tqdm(range(len(x_trains))):\n",
    "        X_train = x_trains[i]\n",
    "        X_test = x_tests[i]\n",
    "        \n",
    "        N_RES = 3\n",
    "\n",
    "        seq_length, dim = X_train.shape[1], X_train.shape[2]\n",
    "        X_train_reverse = np.flip(X_train, axis=1)\n",
    "        X_test_reverse = np.flip(X_test, axis=1)\n",
    "\n",
    "        X_train_by_res = []\n",
    "        X_test_by_res = []\n",
    "        X_train_by_res_reverse = []\n",
    "        \n",
    "        for n in range(N_RES):\n",
    "            selected_t = [t for t in range(0, seq_length, 2**n)]\n",
    "            X_train_by_res.append(X_train[:, selected_t, :])\n",
    "            X_test_by_res.append(X_test[:, selected_t, :])\n",
    "            X_train_by_res_reverse.append(X_train_reverse[:, selected_t, :])\n",
    "        \n",
    "        model = Skip_AE(seq_length, dim, N_RES, N_LAYERS=2, BATCH_SIZE=128)\n",
    "        history = model.fit(X_train_by_res, X_train_by_res_reverse, epochs=50, batch_size=128, validation_split=0.3, verbose=0,\n",
    "                            callbacks=[tf.keras.callbacks.EarlyStopping(monitor=\"val_loss\", patience=5, mode=\"min\", restore_best_weights=True)]) \n",
    "        \n",
    "        X_test_rec = [np.flip(rec, axis=1) for rec in model.predict(X_test_by_res)]\n",
    "        scores = evaluate(X_test, X_test_rec[0], y_tests[i], is_reconstructed=True)\n",
    "\n",
    "        total_scores['dataset'].append(f'D{i+1}')\n",
    "        total_scores['f1'].append(np.max(scores['f1']))\n",
    "        total_scores['pr_auc'].append(scores['pr_auc'])\n",
    "        total_scores['roc_auc'].append(scores['roc_auc'])\n",
    "        print(f'D{i+1}', np.max(scores['f1']), scores['pr_auc'], scores['roc_auc'])  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "f626d0e8-fe11-4214-aba8-27941e07496c",
   "metadata": {},
   "outputs": [],
   "source": [
    "ecg_results = pd.DataFrame(total_scores)\n",
    "ecg_results.groupby('dataset').mean()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5cf8f433-88bf-46fc-a479-42f3f4186f74",
   "metadata": {},
   "source": [
    "### Power Demand"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8cbbc69b-5280-4d6d-9569-0b2627597f2c",
   "metadata": {},
   "outputs": [],
   "source": [
    "total_scores = {'dataset': [], 'f1': [], 'pr_auc': [], 'roc_auc': []}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d0b968b0-53cc-4aef-ba7a-6cdf94a34c5b",
   "metadata": {},
   "outputs": [],
   "source": [
    "for loader in [load_power_demand]:\n",
    "    datasets = loader(64, 1)\n",
    "    x_trains, x_tests, y_tests = datasets['x_train'], datasets['x_test'], datasets['y_test']\n",
    "    \n",
    "    for i in tqdm(range(len(x_trains))):\n",
    "        X_train = x_trains[i]\n",
    "        X_test = x_tests[i]\n",
    "        \n",
    "        N_RES = 3\n",
    "\n",
    "        seq_length, dim = X_train.shape[1], X_train.shape[2]\n",
    "        X_train_reverse = np.flip(X_train, axis=1)\n",
    "        X_test_reverse = np.flip(X_test, axis=1)\n",
    "\n",
    "        X_train_by_res = []\n",
    "        X_test_by_res = []\n",
    "        X_train_by_res_reverse = []\n",
    "        \n",
    "        for n in range(N_RES):\n",
    "            selected_t = [t for t in range(0, seq_length, 2**n)]\n",
    "            X_train_by_res.append(X_train[:, selected_t, :])\n",
    "            X_test_by_res.append(X_test[:, selected_t, :])\n",
    "            X_train_by_res_reverse.append(X_train_reverse[:, selected_t, :])\n",
    "        \n",
    "        model = Skip_AE(seq_length, dim, N_RES, N_LAYERS=2, BATCH_SIZE=128)\n",
    "        history = model.fit(X_train_by_res, X_train_by_res_reverse, epochs=50, batch_size=128, validation_split=0.3, verbose=0,\n",
    "                            callbacks=[tf.keras.callbacks.EarlyStopping(monitor=\"val_loss\", patience=5, mode=\"min\", restore_best_weights=True)]) \n",
    "        \n",
    "        X_test_rec = [np.flip(rec, axis=1) for rec in model.predict(X_test_by_res)]\n",
    "        scores = evaluate(X_test, X_test_rec[0], y_tests[i], is_reconstructed=True)\n",
    "\n",
    "        total_scores['dataset'].append(loader.__name__.replace('load_', ''))\n",
    "        total_scores['f1'].append(np.max(scores['f1']))\n",
    "        total_scores['pr_auc'].append(scores['pr_auc'])\n",
    "        total_scores['roc_auc'].append(scores['roc_auc'])\n",
    "        print(loader.__name__.replace('load_', ''), np.max(scores['f1']), scores['pr_auc'], scores['roc_auc']) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "2c4f3ff4-7927-48fd-8754-32e051613044",
   "metadata": {},
   "outputs": [],
   "source": [
    "power_results = pd.DataFrame(total_scores)\n",
    "power_results.groupby('dataset').mean()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a8c6d607-50ac-423f-8b38-2110977e4e49",
   "metadata": {},
   "source": [
    "### 2D Gesture"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3fb89b02-542d-4519-8c99-dbafeb5b633f",
   "metadata": {},
   "outputs": [],
   "source": [
    "total_scores = {'dataset': [], 'f1': [], 'pr_auc': [], 'roc_auc': []}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "405d2e4a-59b2-4567-92ed-fec74be40dd7",
   "metadata": {},
   "outputs": [],
   "source": [
    "for loader in [load_gesture]:\n",
    "    datasets = loader(64, 1)\n",
    "    x_trains, x_tests, y_tests = datasets['x_train'], datasets['x_test'], datasets['y_test']\n",
    "    \n",
    "    for i in tqdm(range(len(x_trains))):\n",
    "        X_train = x_trains[i]\n",
    "        X_test = x_tests[i]\n",
    "        \n",
    "        N_RES = 3\n",
    "\n",
    "        seq_length, dim = X_train.shape[1], X_train.shape[2]\n",
    "        X_train_reverse = np.flip(X_train, axis=1)\n",
    "        X_test_reverse = np.flip(X_test, axis=1)\n",
    "\n",
    "        X_train_by_res = []\n",
    "        X_test_by_res = []\n",
    "        X_train_by_res_reverse = []\n",
    "        \n",
    "        for n in range(N_RES):\n",
    "            selected_t = [t for t in range(0, seq_length, 2**n)]\n",
    "            X_train_by_res.append(X_train[:, selected_t, :])\n",
    "            X_test_by_res.append(X_test[:, selected_t, :])\n",
    "            X_train_by_res_reverse.append(X_train_reverse[:, selected_t, :])\n",
    "        \n",
    "        model = Skip_AE(seq_length, dim, N_RES, N_LAYERS=2, BATCH_SIZE=128)\n",
    "        history = model.fit(X_train_by_res, X_train_by_res_reverse, epochs=50, batch_size=128, validation_split=0.3, verbose=0,\n",
    "                            callbacks=[tf.keras.callbacks.EarlyStopping(monitor=\"val_loss\", patience=5, mode=\"min\", restore_best_weights=True)]) \n",
    "        \n",
    "        X_test_rec = [np.flip(rec, axis=1) for rec in model.predict(X_test_by_res)]\n",
    "        scores = evaluate(X_test, X_test_rec[0], y_tests[i], is_reconstructed=True)\n",
    "\n",
    "        total_scores['dataset'].append(loader.__name__.replace('load_', ''))\n",
    "        total_scores['f1'].append(np.max(scores['f1']))\n",
    "        total_scores['pr_auc'].append(scores['pr_auc'])\n",
    "        total_scores['roc_auc'].append(scores['roc_auc'])\n",
    "        print(loader.__name__.replace('load_', ''), np.max(scores['f1']), scores['pr_auc'], scores['roc_auc'])  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b2d18eea-9b0d-44fc-bc61-6e3c91f35b40",
   "metadata": {},
   "outputs": [],
   "source": [
    "gesture_results = pd.DataFrame(total_scores)\n",
    "gesture_results.groupby('dataset').mean()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}