2. gp-vae.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "460bb457-44f4-4eaf-b7a7-62882bb6a266",
   "metadata": {},
   "source": [
    "<div style=\"text-align: right\"> \n",
    "<a href = \"https://nbviewer.jupyter.org/github/siddharthchaini/Improving-Feature-Extraction-RESSPECT/tree/main/\"> \n",
    "    View online on<br>\n",
    "    <img src=\"https://nbviewer.jupyter.org/static/img/nav_logo.svg\"\n",
    "         alt=\"nbviewer link\" style=\"height: 30px;\"/> \n",
    "</a>\n",
    "</div>"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2d1f3a19-20e0-40a7-97d8-1dbd9b41b45f",
   "metadata": {},
   "source": [
    "# 2. GP-VAE Approach for RESSPECT\n",
    "<a href = http://cosmostatistics-initiative.org/resspect/>\n",
    "<img src=\"https://cosmostatistics-initiative.org/wp-content/uploads/2019/04/coin_desc_3.png\" alt=\"Logo\" style=\"width: 250px;\"/>\n",
    "</a>\n",
    "\n",
    "**Authors**: Siddharth Chaini, Johann Cohen-Tanugi"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a87cbb6a-2314-4fe2-977c-17c7c7289b2d",
   "metadata": {},
   "source": [
    "This is the second Jupyter notebook (of a total of 2) describing the different feature extraction methods explored by Siddharth and Johann in June & July, 2021.\n",
    "\n",
    "For RESSPECT, a light curve is fit to a function and then the fit parameters are then used in the RESSPECT pipeline as features for training. You can read more about this here: https://arxiv.org/pdf/2010.05941.pdf\n",
    "\n",
    "In this notebook, we will look at a deep learning based approach for feature extraction, GP-VAE."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "5f5096bc-6b96-4270-9027-e5d5502680e5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Last Updated On: 30 July, 2021\n"
     ]
    }
   ],
   "source": [
    "import datetime\n",
    "print('Last Updated On:', datetime.datetime.now().strftime(\"%d %B, %Y\"))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "50095328-75ba-4d6c-ada2-05d64f1c838b",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Some prerequisites\n",
    "\n",
    "import numpy as np\n",
    "np.random.seed(42)\n",
    "np.set_printoptions(suppress=True)\n",
    "from scipy.optimize import least_squares, curve_fit\n",
    "from matplotlib import pylab as plt\n",
    "import pandas as pd\n",
    "import glob\n",
    "import time\n",
    "from tqdm.notebook import tqdm\n",
    "import os\n",
    "\n",
    "import warnings\n",
    "from scipy.optimize import OptimizeWarning\n",
    "warnings.simplefilter(\"error\", OptimizeWarning)\n",
    "\n",
    "import seaborn as sns\n",
    "sns.set()\n",
    "\n",
    "def plot_allbands_df(df,title=None,ax=plt):\n",
    "    ugrizYcolors = [\"violet\",\"green\",\"red\",\"darkred\",\"grey\",\"black\"]\n",
    "    for i,band in enumerate(\"ugrizY\"):\n",
    "        subdf = df[df[\"band\"]==band].reset_index(drop=True)\n",
    "        x = subdf[\"mjd\"]\n",
    "        y = subdf[\"flux\"]\n",
    "        yerr = subdf[\"fluxerr\"]\n",
    "#         ax.errorbar(x,y,yerr=yerr,ls = \"None\",color=ugrizYcolors[i])\n",
    "        ax.scatter(x,y,s=5,label=band,color=ugrizYcolors[i])\n",
    "        if title is not None:\n",
    "            if ax is not plt:\n",
    "                ax.set_title(title)\n",
    "            else:\n",
    "                ax.title(title)\n",
    "    return ax\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ed7ed7cd-8392-4f7a-a70e-18c6b7e9fb18",
   "metadata": {},
   "source": [
    "## 2.1 About GP-VAE\n",
    "\n",
    "Gaussian-Process Variational Autoencoders (GP-VAE ; [Fortuin et. al. 2020](https://arxiv.org/pdf/1907.04155.pdf)) is an approach developed for dimensionality reduction and data imputation. It combines Variational Autoencoders (VAEs) with Gaussian Processes (GPs) in the following way:\n",
    "\n",
    "- Deep VAEs are used to map the original time series data with missing values into a latent space.\n",
    "- A GP then utilizes latent representations to capture the temporal correlations in the time series.\n",
    "\n",
    "A detailed description of this is available in [Fortuin et. al. (2020)](https://arxiv.org/pdf/1907.04155.pdf).\n",
    "\n",
    "For our use case, we wish to use the latent space of the GP-VAE as features for the light curves in 6 passbands. We use code provided by the authors on [GitHub](https://github.com/ratschlab/GP-VAE)<sup>[*](#myfootnote1)</sup>.\n",
    "\n",
    "<a name=\"myfootnote1\">*</a> Note: The original code was based on TensorFlow 1.15 and so we made small modifications on a [fork](https://github.com/siddharthchaini/GP-VAE) using [TensorFlow's Upgrade Script](https://www.tensorflow.org/guide/upgrade) to make it compatible for TensorFlow 2+."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c48c32b4-4051-4c22-a606-62a899df6727",
   "metadata": {},
   "source": [
    "## 2.2 Data & Preprocessing\n",
    "\n",
    "PLAsTiCC/RESSPECT data consists of light-curve data in the form of CSVs with readings at a particular timestamp, indicating the mjd of observation, the passband in which it was observed, and the flux values and flux errors.\n",
    "\n",
    "For example:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "68d2ea2c-cc1a-4ed0-8a26-bd55cd6ac736",
   "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>mjd</th>\n",
       "      <th>band</th>\n",
       "      <th>flux</th>\n",
       "      <th>fluxerr</th>\n",
       "      <th>detected_bool</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>59710.4130</td>\n",
       "      <td>i</td>\n",
       "      <td>571.036438</td>\n",
       "      <td>38.534260</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>59715.3319</td>\n",
       "      <td>Y</td>\n",
       "      <td>749.963562</td>\n",
       "      <td>32.526573</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>59722.3088</td>\n",
       "      <td>z</td>\n",
       "      <td>811.698425</td>\n",
       "      <td>33.418995</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>59728.4313</td>\n",
       "      <td>z</td>\n",
       "      <td>791.284973</td>\n",
       "      <td>46.716145</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>59729.2258</td>\n",
       "      <td>i</td>\n",
       "      <td>834.375305</td>\n",
       "      <td>28.976799</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>113</th>\n",
       "      <td>60555.9838</td>\n",
       "      <td>z</td>\n",
       "      <td>-39.881969</td>\n",
       "      <td>46.477093</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>114</th>\n",
       "      <td>60560.0459</td>\n",
       "      <td>g</td>\n",
       "      <td>14.894439</td>\n",
       "      <td>18.947685</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>115</th>\n",
       "      <td>60571.0225</td>\n",
       "      <td>Y</td>\n",
       "      <td>30.593130</td>\n",
       "      <td>50.695290</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>116</th>\n",
       "      <td>60585.9974</td>\n",
       "      <td>z</td>\n",
       "      <td>-23.471439</td>\n",
       "      <td>44.819859</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>117</th>\n",
       "      <td>60588.0372</td>\n",
       "      <td>u</td>\n",
       "      <td>-41.214264</td>\n",
       "      <td>51.665123</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>118 rows × 5 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "            mjd band        flux    fluxerr  detected_bool\n",
       "0    59710.4130    i  571.036438  38.534260              1\n",
       "1    59715.3319    Y  749.963562  32.526573              1\n",
       "2    59722.3088    z  811.698425  33.418995              1\n",
       "3    59728.4313    z  791.284973  46.716145              1\n",
       "4    59729.2258    i  834.375305  28.976799              1\n",
       "..          ...  ...         ...        ...            ...\n",
       "113  60555.9838    z  -39.881969  46.477093              0\n",
       "114  60560.0459    g   14.894439  18.947685              0\n",
       "115  60571.0225    Y   30.593130  50.695290              0\n",
       "116  60585.9974    z  -23.471439  44.819859              0\n",
       "117  60588.0372    u  -41.214264  51.665123              0\n",
       "\n",
       "[118 rows x 5 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.read_csv(\"sample_data/plasticc/130779836_uLens.csv\")\n",
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "057374fc-bf20-40d0-9340-5f49fcb759eb",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "tempplt = plot_allbands_df(df)\n",
    "tempplt.title(\"Original Light Curve\")\n",
    "tempplt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7fbed61f-c6a3-4f57-b7bd-78094a0fbb94",
   "metadata": {},
   "source": [
    "This needs to be modified into a conventional time series having a constant length before being fed to the GP-VAE as input.\n",
    "\n",
    "We choose the constant length to be 200 days as a heuristic, because that covers almost all transient activity based on our testing (using ```detected_bool=1```) (ToDo: EXPLAIN MORE)\n",
    "\n",
    "If either of the following is true:\n",
    "- More than 3 observations in a span of 5 consecutive days have a ratio >=2\n",
    "- detected_bool = 1\n",
    "then the light curve is cropped to a length of 200 days roughly centred about the point when either of the above happens.    \n",
    "\n",
    "When neither of the above 2 criteria are fulfilled, the first 200 days are considered by default (ToDo: EXPLAIN MORE about why these criteria were chosen.)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "c3d4881e-46f9-442e-a533-171fc8a3cade",
   "metadata": {},
   "outputs": [],
   "source": [
    "def center_bigdf(df):\n",
    "    '''\n",
    "    If either of the following is true:\n",
    "        - More than 3 observations in a span of 5 consecutive days have a ratio >=2\n",
    "        - detected_bool = 1\n",
    "    then the light curve is cropped to a length of 200 days roughly centred about the point when either of the above happens.\n",
    "    \n",
    "    When neither of the above 2 criteria are fulfilled, the first 200 days are considered by default\n",
    "\n",
    "    '''\n",
    "    \n",
    "    # Start mjd from 0\n",
    "    df[\"mjd\"] = df[\"mjd\"] - df[\"mjd\"][0]\n",
    "    df[\"ratio\"] = abs(df[\"flux\"]/df[\"fluxerr\"])\n",
    "\n",
    "    df[\"ratio2bool\"] = 0\n",
    "\n",
    "    subdf_detected = df[df[\"detected_bool\"]==1]\n",
    "    subdf_ratio2 = df[df[\"ratio\"]>=2]\n",
    "\n",
    "    # Find the mjd where more than 3 observations in a span of 5 consecutive days have a flux-fluxerr ratio >=2\n",
    "    smalldf = subdf_ratio2.loc[abs(subdf_ratio2.mjd.diff(periods=-1))<=5]\n",
    "    for i in range(3):\n",
    "        smalldf = smalldf[abs(smalldf.mjd.diff(periods=-1))<=5]\n",
    "\n",
    "    \n",
    "    if len(subdf_detected)>0 and len(smalldf)>0:\n",
    "        detected_bool_minmjd = subdf_detected.mjd.values[0]\n",
    "        custom_bool_minmjd = smalldf.mjd.values[0]\n",
    "\n",
    "        diffmjd = abs(detected_bool_minmjd - custom_bool_minmjd)/2\n",
    "\n",
    "        minmjdstart = min(detected_bool_minmjd, custom_bool_minmjd)\n",
    "        maxmjdstart = max(detected_bool_minmjd, custom_bool_minmjd)\n",
    "\n",
    "        mjdlim_lower = minmjdstart - (100-diffmjd)\n",
    "        mjdlim_upper = maxmjdstart + (100-diffmjd)\n",
    "\n",
    "    elif len(subdf_detected)>0:\n",
    "        detected_bool_minmjd = subdf_detected.mjd.values[0]\n",
    "        mjdlim_lower = detected_bool_minmjd - 100\n",
    "        mjdlim_upper = detected_bool_minmjd + 100\n",
    "\n",
    "    elif len(smalldf)>0:\n",
    "        custom_bool_minmjd = smalldf.mjd.values[0]\n",
    "        mjdlim_lower = custom_bool_minmjd - 100\n",
    "        mjdlim_upper = custom_bool_minmjd + 100\n",
    "\n",
    "    else:\n",
    "        mjdlim_lower = 0\n",
    "        mjdlim_upper = 200\n",
    "    \n",
    "    # Lower limit can't be -ve\n",
    "    if mjdlim_lower<0:\n",
    "        mjdlim_upper = mjdlim_upper + abs(mjdlim_lower)\n",
    "        mjdlim_lower = 0\n",
    "\n",
    "    assert (mjdlim_upper-mjdlim_lower)==200\n",
    "    \n",
    "    df.loc[(df[\"mjd\"] >= mjdlim_lower) & (df[\"mjd\"] <= mjdlim_upper),\"ratio2bool\"] = 1\n",
    "    \n",
    "    newdf = df[df[\"ratio2bool\"]==1].reset_index(drop=True)\n",
    "    \n",
    "    return newdf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "4b47885c-98bb-445a-8ee3-09d4e066a2c2",
   "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>mjd</th>\n",
       "      <th>band</th>\n",
       "      <th>flux</th>\n",
       "      <th>fluxerr</th>\n",
       "      <th>detected_bool</th>\n",
       "      <th>ratio</th>\n",
       "      <th>ratio2bool</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.0000</td>\n",
       "      <td>i</td>\n",
       "      <td>571.036438</td>\n",
       "      <td>38.534260</td>\n",
       "      <td>1</td>\n",
       "      <td>14.818928</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>4.9189</td>\n",
       "      <td>Y</td>\n",
       "      <td>749.963562</td>\n",
       "      <td>32.526573</td>\n",
       "      <td>1</td>\n",
       "      <td>23.056950</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>11.8958</td>\n",
       "      <td>z</td>\n",
       "      <td>811.698425</td>\n",
       "      <td>33.418995</td>\n",
       "      <td>1</td>\n",
       "      <td>24.288535</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>18.0183</td>\n",
       "      <td>z</td>\n",
       "      <td>791.284973</td>\n",
       "      <td>46.716145</td>\n",
       "      <td>1</td>\n",
       "      <td>16.938148</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>18.8128</td>\n",
       "      <td>i</td>\n",
       "      <td>834.375305</td>\n",
       "      <td>28.976799</td>\n",
       "      <td>1</td>\n",
       "      <td>28.794599</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>42.7361</td>\n",
       "      <td>g</td>\n",
       "      <td>759.056091</td>\n",
       "      <td>23.736109</td>\n",
       "      <td>1</td>\n",
       "      <td>31.978960</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>58.7269</td>\n",
       "      <td>z</td>\n",
       "      <td>2148.493896</td>\n",
       "      <td>25.189192</td>\n",
       "      <td>1</td>\n",
       "      <td>85.294276</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>61.7802</td>\n",
       "      <td>Y</td>\n",
       "      <td>2285.122559</td>\n",
       "      <td>79.410194</td>\n",
       "      <td>1</td>\n",
       "      <td>28.776187</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>65.8392</td>\n",
       "      <td>z</td>\n",
       "      <td>2471.021973</td>\n",
       "      <td>27.277407</td>\n",
       "      <td>1</td>\n",
       "      <td>90.588595</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>66.8824</td>\n",
       "      <td>Y</td>\n",
       "      <td>2618.357910</td>\n",
       "      <td>46.559090</td>\n",
       "      <td>1</td>\n",
       "      <td>56.237309</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>67.6941</td>\n",
       "      <td>r</td>\n",
       "      <td>2269.778564</td>\n",
       "      <td>35.356678</td>\n",
       "      <td>1</td>\n",
       "      <td>64.196601</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>88.6689</td>\n",
       "      <td>Y</td>\n",
       "      <td>4463.832031</td>\n",
       "      <td>54.378235</td>\n",
       "      <td>1</td>\n",
       "      <td>82.088579</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>93.6982</td>\n",
       "      <td>z</td>\n",
       "      <td>4960.222656</td>\n",
       "      <td>41.600491</td>\n",
       "      <td>1</td>\n",
       "      <td>119.234714</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>94.5920</td>\n",
       "      <td>i</td>\n",
       "      <td>4972.901855</td>\n",
       "      <td>54.523323</td>\n",
       "      <td>1</td>\n",
       "      <td>91.206874</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>95.7639</td>\n",
       "      <td>i</td>\n",
       "      <td>4995.144043</td>\n",
       "      <td>50.181072</td>\n",
       "      <td>1</td>\n",
       "      <td>99.542394</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>96.7094</td>\n",
       "      <td>r</td>\n",
       "      <td>4531.691895</td>\n",
       "      <td>55.449078</td>\n",
       "      <td>1</td>\n",
       "      <td>81.727092</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>99.6729</td>\n",
       "      <td>r</td>\n",
       "      <td>4908.624512</td>\n",
       "      <td>66.093193</td>\n",
       "      <td>1</td>\n",
       "      <td>74.268231</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>99.7577</td>\n",
       "      <td>g</td>\n",
       "      <td>3088.283936</td>\n",
       "      <td>53.182644</td>\n",
       "      <td>1</td>\n",
       "      <td>58.069395</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>100.5904</td>\n",
       "      <td>r</td>\n",
       "      <td>5030.671387</td>\n",
       "      <td>60.416359</td>\n",
       "      <td>1</td>\n",
       "      <td>83.266709</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>101.6101</td>\n",
       "      <td>g</td>\n",
       "      <td>3320.936279</td>\n",
       "      <td>56.901775</td>\n",
       "      <td>1</td>\n",
       "      <td>58.362613</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>106.7293</td>\n",
       "      <td>u</td>\n",
       "      <td>1036.600708</td>\n",
       "      <td>57.762154</td>\n",
       "      <td>1</td>\n",
       "      <td>17.946019</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>107.6422</td>\n",
       "      <td>u</td>\n",
       "      <td>1081.982300</td>\n",
       "      <td>61.724331</td>\n",
       "      <td>1</td>\n",
       "      <td>17.529267</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>124.5715</td>\n",
       "      <td>Y</td>\n",
       "      <td>12149.348633</td>\n",
       "      <td>92.792801</td>\n",
       "      <td>1</td>\n",
       "      <td>130.929862</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>124.6396</td>\n",
       "      <td>r</td>\n",
       "      <td>10043.888672</td>\n",
       "      <td>120.556976</td>\n",
       "      <td>1</td>\n",
       "      <td>83.312381</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>134.5768</td>\n",
       "      <td>z</td>\n",
       "      <td>16366.390625</td>\n",
       "      <td>130.593475</td>\n",
       "      <td>1</td>\n",
       "      <td>125.323188</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>148.6704</td>\n",
       "      <td>Y</td>\n",
       "      <td>27343.818359</td>\n",
       "      <td>199.967590</td>\n",
       "      <td>1</td>\n",
       "      <td>136.741251</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>150.5826</td>\n",
       "      <td>Y</td>\n",
       "      <td>29058.212891</td>\n",
       "      <td>210.280746</td>\n",
       "      <td>1</td>\n",
       "      <td>138.187701</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>157.6125</td>\n",
       "      <td>r</td>\n",
       "      <td>31437.089844</td>\n",
       "      <td>362.050262</td>\n",
       "      <td>1</td>\n",
       "      <td>86.830734</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>166.5916</td>\n",
       "      <td>z</td>\n",
       "      <td>54982.875000</td>\n",
       "      <td>424.749237</td>\n",
       "      <td>1</td>\n",
       "      <td>129.447849</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>166.6447</td>\n",
       "      <td>u</td>\n",
       "      <td>7761.321777</td>\n",
       "      <td>167.561752</td>\n",
       "      <td>1</td>\n",
       "      <td>46.319173</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         mjd band          flux     fluxerr  detected_bool       ratio  \\\n",
       "0     0.0000    i    571.036438   38.534260              1   14.818928   \n",
       "1     4.9189    Y    749.963562   32.526573              1   23.056950   \n",
       "2    11.8958    z    811.698425   33.418995              1   24.288535   \n",
       "3    18.0183    z    791.284973   46.716145              1   16.938148   \n",
       "4    18.8128    i    834.375305   28.976799              1   28.794599   \n",
       "5    42.7361    g    759.056091   23.736109              1   31.978960   \n",
       "6    58.7269    z   2148.493896   25.189192              1   85.294276   \n",
       "7    61.7802    Y   2285.122559   79.410194              1   28.776187   \n",
       "8    65.8392    z   2471.021973   27.277407              1   90.588595   \n",
       "9    66.8824    Y   2618.357910   46.559090              1   56.237309   \n",
       "10   67.6941    r   2269.778564   35.356678              1   64.196601   \n",
       "11   88.6689    Y   4463.832031   54.378235              1   82.088579   \n",
       "12   93.6982    z   4960.222656   41.600491              1  119.234714   \n",
       "13   94.5920    i   4972.901855   54.523323              1   91.206874   \n",
       "14   95.7639    i   4995.144043   50.181072              1   99.542394   \n",
       "15   96.7094    r   4531.691895   55.449078              1   81.727092   \n",
       "16   99.6729    r   4908.624512   66.093193              1   74.268231   \n",
       "17   99.7577    g   3088.283936   53.182644              1   58.069395   \n",
       "18  100.5904    r   5030.671387   60.416359              1   83.266709   \n",
       "19  101.6101    g   3320.936279   56.901775              1   58.362613   \n",
       "20  106.7293    u   1036.600708   57.762154              1   17.946019   \n",
       "21  107.6422    u   1081.982300   61.724331              1   17.529267   \n",
       "22  124.5715    Y  12149.348633   92.792801              1  130.929862   \n",
       "23  124.6396    r  10043.888672  120.556976              1   83.312381   \n",
       "24  134.5768    z  16366.390625  130.593475              1  125.323188   \n",
       "25  148.6704    Y  27343.818359  199.967590              1  136.741251   \n",
       "26  150.5826    Y  29058.212891  210.280746              1  138.187701   \n",
       "27  157.6125    r  31437.089844  362.050262              1   86.830734   \n",
       "28  166.5916    z  54982.875000  424.749237              1  129.447849   \n",
       "29  166.6447    u   7761.321777  167.561752              1   46.319173   \n",
       "\n",
       "    ratio2bool  \n",
       "0            1  \n",
       "1            1  \n",
       "2            1  \n",
       "3            1  \n",
       "4            1  \n",
       "5            1  \n",
       "6            1  \n",
       "7            1  \n",
       "8            1  \n",
       "9            1  \n",
       "10           1  \n",
       "11           1  \n",
       "12           1  \n",
       "13           1  \n",
       "14           1  \n",
       "15           1  \n",
       "16           1  \n",
       "17           1  \n",
       "18           1  \n",
       "19           1  \n",
       "20           1  \n",
       "21           1  \n",
       "22           1  \n",
       "23           1  \n",
       "24           1  \n",
       "25           1  \n",
       "26           1  \n",
       "27           1  \n",
       "28           1  \n",
       "29           1  "
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "newdf = center_bigdf(df)\n",
    "newdf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "589d4476-81e1-4f4f-b9f9-956ef386c235",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "tempplt = plot_allbands_df(newdf)\n",
    "tempplt.title(\"New Cropped Light Curve\")\n",
    "tempplt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "79601a40-ebc6-401a-9a93-b8e61c10f6a3",
   "metadata": {},
   "source": [
    "\n",
    "---\n",
    "The full code for the above preprocessing and centering can be found on these GitHub repositories:\n",
    "- https://github.com/siddharthchaini/centering-perfect-sims-resspect: Centering of RESSPECT Perfect Simulation Light Curves\n",
    "- https://github.com/siddharthchaini/centering-time-series-plasticc: Centering of PLAsTiCC Data\n",
    "\n",
    "**(NOTE: These will be ported over to the current repository soon. ToDo)**"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c4b91bea-b97b-4de0-9510-9720b05525ac",
   "metadata": {},
   "source": [
    "## 2.3 Downloading the code for GP-VAE\n",
    "\n",
    "Based on the TensorFlow version, different code will be used."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "1f39708e-95d5-4c42-92ea-960d5b4f0600",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Install GP-VAE\n",
    "\n",
    "# import tensorflow as tf\n",
    "# import subprocess\n",
    "# tf_v = tf.__version__\n",
    "# print(f\"TF Version is: {tf_v}\")\n",
    "\n",
    "# if int(tf_v[0])==2:\n",
    "#     git_url = \"https://github.com/siddharthchaini/GP-VAE\"\n",
    "# elif int(tf_v[0])==1:\n",
    "#     git_url = \"https://github.com/ratschlab/GP-VAE\"\n",
    "    \n",
    "    \n",
    "# dl_loc = \"../.repos/GP-VAE\" # Directory where GP-VAE Code is saved / to be saved\n",
    "# if not os.path.isdir(dl_loc):\n",
    "#     print(\"Downloading ...\")\n",
    "#     # Download code\n",
    "#     process = subprocess.run([\"git\",\"clone\", git_url, dl_loc], stdout=subprocess.PIPE, universal_newlines=True)\n",
    "#     if process.returncode == 0:\n",
    "#         print(\"Downloaded successfully\")\n",
    "#     else:\n",
    "#         print(\"Error downloading.\")\n",
    "#         print(process.stdout)\n",
    "# else:\n",
    "#     print(\"Already found GP-VAE directory\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "fba71844-716a-4892-bb40-c66ab206227b",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Change working directory to GP-VAE Raw\n",
    "os.chdir(\"gpvae-raw\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "56d357c2-d35a-4479-b262-5afa33d73d90",
   "metadata": {},
   "source": [
    "This directory is based on https://github.com/siddharthchaini/gpvae-raw"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8dec5025-0781-47de-9bb5-afe5d8d09ed7",
   "metadata": {},
   "source": [
    "### 2.3.1 Let's now implement GP-VAE for the Physionet data as described in [Fortuin et. al. (2020)](https://arxiv.org/pdf/1907.04155.pdf)\n",
    "\n",
    "References: \n",
    "1. https://www.kaggle.com/siddharthchaini/gpvae-raw?scriptVersionId=67899219\n",
    "2. https://www.kaggle.com/siddharthchaini/gpvae-raw-visualisation?scriptVersionId=68152622"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "25602cdb-bbd2-4b68-b1e7-063c1238976f",
   "metadata": {},
   "outputs": [],
   "source": [
    "from lib.models import *"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a0f78a69-3b10-4ce2-8e89-556f8d74536a",
   "metadata": {
    "papermill": {
     "duration": 0.030107,
     "end_time": "2021-07-10T19:00:10.551188",
     "exception": false,
     "start_time": "2021-07-10T19:00:10.521081",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "#### 2.3.1.1 Download Physionet data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "e12f187c-a0af-4aa4-87ee-f4e5ca8965ff",
   "metadata": {
    "papermill": {
     "duration": 4.850243,
     "end_time": "2021-07-10T19:00:15.431471",
     "exception": false,
     "start_time": "2021-07-10T19:00:10.581228",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Download only the first time\n",
    "\n",
    "# !wget https://www.dropbox.com/s/651d86winb4cy9n/physionet.npz?dl=1 -O physionet.npz\n",
    "\n",
    "assert os.path.isfile(\"physionet.npz\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cdaf1d62-8246-4458-b1e0-436bc19abe38",
   "metadata": {
    "papermill": {
     "duration": 0.034606,
     "end_time": "2021-07-10T19:00:15.504120",
     "exception": false,
     "start_time": "2021-07-10T19:00:15.469514",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "#### 2.3.1.2 Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "39e7dcc1-6685-4376-b92a-4c68398a92ea",
   "metadata": {
    "papermill": {
     "duration": 5.484319,
     "end_time": "2021-07-10T19:00:21.023448",
     "exception": false,
     "start_time": "2021-07-10T19:00:15.539129",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import sys\n",
    "\n",
    "import os\n",
    "import time\n",
    "from datetime import datetime\n",
    "import numpy as np\n",
    "import matplotlib\n",
    "matplotlib.use(\"Agg\")\n",
    "from matplotlib import pyplot as plt\n",
    "import tensorflow as tf\n",
    "\n",
    "tf.compat.v1.enable_eager_execution()\n",
    "\n",
    "from sklearn.metrics import average_precision_score, roc_auc_score\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "\n",
    "import warnings\n",
    "warnings.simplefilter(action='ignore', category=FutureWarning)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "8d7a7459-234b-40f5-8da7-78aa78cb38b7",
   "metadata": {
    "papermill": {
     "duration": 0.754087,
     "end_time": "2021-07-10T19:00:21.813470",
     "exception": false,
     "start_time": "2021-07-10T19:00:21.059383",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "from lib.models import *"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "91aa026a-26e7-41e4-a548-326914ce52cb",
   "metadata": {
    "papermill": {
     "duration": 0.03395,
     "end_time": "2021-07-10T19:00:21.882305",
     "exception": false,
     "start_time": "2021-07-10T19:00:21.848355",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "#### 2.3.1.3 Flags"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "b9aa502d-e8e6-46e2-b818-2089f5915a3d",
   "metadata": {
    "papermill": {
     "duration": 0.044841,
     "end_time": "2021-07-10T19:00:21.961802",
     "exception": false,
     "start_time": "2021-07-10T19:00:21.916961",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "latent_dim = 35 # 'Dimensionality of the latent space'\n",
    "encoder_sizes = [128, 128] # 'Layer sizes of the encoder'\n",
    "decoder_sizes = [256, 256] # 'Layer sizes of the decoder'\n",
    "window_size = 24 # 'Window size for the inference CNN: Ignored if model_type is not gp-vae'\n",
    "sigma = 1.005 # 'Sigma value for the GP prior: Ignored if model_type is not gp-vae'\n",
    "length_scale = 7.0 # 'Length scale value for the GP prior: Ignored if model_type is not gp-vae'\n",
    "beta = 0.2 # 'Factor to weigh the KL term (similar to beta-VAE'\n",
    "num_epochs = 40 # 'Number of training epochs'\n",
    "\n",
    "# Flags with common default values for all three datasets\n",
    "learning_rate = 1e-3 # 'Learning rate for training'\n",
    "gradient_clip = 1e4 # 'Maximum global gradient norm for the gradient clipping during training'\n",
    "num_steps = 0 # 'Number of training steps: If non-zero it overwrites num_epochs'\n",
    "print_interval = 0 # 'Interval for printing the loss and saving the model during training'\n",
    "exp_name = \"reproduce_physionet\" # 'Name of the experiment'\n",
    "basedir = \"models\" # 'Directory where the models should be stored'\n",
    "data_dir = \"\" # 'Directory from where the data should be read in'\n",
    "data_type = 'physionet' # ['hmnist', 'physionet', 'sprites'], 'Type of data to be trained on'\n",
    "seed = 1337 # 'Seed for the random number generator'\n",
    "model_type = 'gp-vae' # ['vae', 'hi-vae', 'gp-vae'], 'Type of model to be trained'\n",
    "cnn_kernel_size = 3 # 'Kernel size for the CNN preprocessor'\n",
    "cnn_sizes = [256] # 'Number of filters for the layers of the CNN preprocessor'\n",
    "testing = True # 'Use the actual test set for testing'\n",
    "banded_covar = True # 'Use a banded covariance matrix instead of a diagonal one for the output of the inference network: Ignored if model_type is not gp-vae'\n",
    "batch_size = 64 # 'Batch size for training'\n",
    "\n",
    "M = 1 # 'Number of samples for ELBO estimation'\n",
    "K = 1 # 'Number of importance sampling weights'\n",
    "\n",
    "kernel = 'cauchy' # ['rbf', 'diffusion', 'matern', 'cauchy'], 'Kernel to be used for the GP prior: Ignored if model_type is not (mgp-vae'\n",
    "kernel_scales = 1 # 'Number of different length scales sigma for the GP prior: Ignored if model_type is not gp-vae'"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "57da1f78-56e0-4bb6-82b4-317ed8386c5d",
   "metadata": {
    "papermill": {
     "duration": 0.035186,
     "end_time": "2021-07-10T19:00:22.031270",
     "exception": false,
     "start_time": "2021-07-10T19:00:21.996084",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "#### 2.3.1.4 Prep"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "d63b3e90-faa8-4d50-9852-9a6b407c74b8",
   "metadata": {
    "papermill": {
     "duration": 0.044087,
     "end_time": "2021-07-10T19:00:22.109911",
     "exception": false,
     "start_time": "2021-07-10T19:00:22.065824",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Testing:  True \t Seed: 1337\n"
     ]
    }
   ],
   "source": [
    "np.random.seed(seed)\n",
    "tf.compat.v1.set_random_seed(seed)\n",
    "print(\"Testing: \", testing, f\"\\t Seed: {seed}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "bd0ddfdc-7a4d-4446-b3ac-9805bd1d5a6a",
   "metadata": {
    "papermill": {
     "duration": 0.041576,
     "end_time": "2021-07-10T19:00:22.186537",
     "exception": false,
     "start_time": "2021-07-10T19:00:22.144961",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "encoder_sizes = [int(size) for size in encoder_sizes]\n",
    "decoder_sizes = [int(size) for size in decoder_sizes]\n",
    "\n",
    "if 0 in encoder_sizes:\n",
    "    encoder_sizes.remove(0)\n",
    "if 0 in decoder_sizes:\n",
    "    decoder_sizes.remove(0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "09e34e49-74ca-491a-a766-f59b9754f282",
   "metadata": {
    "papermill": {
     "duration": 0.043822,
     "end_time": "2021-07-10T19:00:22.264885",
     "exception": false,
     "start_time": "2021-07-10T19:00:22.221063",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Full exp name:  210729_reproduce_physionet\n"
     ]
    }
   ],
   "source": [
    "# Make up full exp name\n",
    "timestamp = datetime.now().strftime(\"%y%m%d\")\n",
    "full_exp_name = \"{}_{}\".format(timestamp, exp_name)\n",
    "outdir = os.path.join(basedir, full_exp_name)\n",
    "if not os.path.exists(outdir): os.mkdir(outdir)\n",
    "checkpoint_prefix = os.path.join(outdir, \"ckpt\")\n",
    "print(\"Full exp name: \", full_exp_name)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2f25db62-6f3f-45cc-a64c-b914131691ea",
   "metadata": {
    "papermill": {
     "duration": 0.034568,
     "end_time": "2021-07-10T19:00:22.334299",
     "exception": false,
     "start_time": "2021-07-10T19:00:22.299731",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "#### 2.3.1.5 Define data specific parameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "14018387-20ba-4f61-8c2f-66e0427928ee",
   "metadata": {
    "papermill": {
     "duration": 0.044722,
     "end_time": "2021-07-10T19:00:22.413713",
     "exception": false,
     "start_time": "2021-07-10T19:00:22.368991",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'physionet'"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data_type"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "383a05b3-209e-4cc8-a045-94fb36242dfe",
   "metadata": {
    "papermill": {
     "duration": 0.043675,
     "end_time": "2021-07-10T19:00:22.492672",
     "exception": false,
     "start_time": "2021-07-10T19:00:22.448997",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "if data_type == \"hmnist\":\n",
    "    data_dir = \"data/hmnist/hmnist_mnar.npz\"\n",
    "    data_dim = 784\n",
    "    time_length = 10\n",
    "    num_classes = 10\n",
    "    decoder = BernoulliDecoder\n",
    "    img_shape = (28, 28, 1)\n",
    "    val_split = 50000\n",
    "elif data_type == \"physionet\":\n",
    "    if data_dir == \"\":\n",
    "        data_dir = \"physionet.npz\"\n",
    "    data_dim = 35\n",
    "    time_length = 48\n",
    "    num_classes = 2\n",
    "    decoder = GaussianDecoder\n",
    "elif data_type == \"sprites\":\n",
    "    if data_dir == \"\":\n",
    "        data_dir = \"data/sprites/sprites.npz\"\n",
    "    data_dim = 12288\n",
    "    time_length = 8\n",
    "    decoder = GaussianDecoder\n",
    "    img_shape = (64, 64, 3)\n",
    "    val_split = 8000\n",
    "else:\n",
    "    raise ValueError(\"Data type must be one of ['hmnist', 'physionet', 'sprites']\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "18739e39-ba79-4d33-b233-ed170e61c50a",
   "metadata": {
    "papermill": {
     "duration": 0.03487,
     "end_time": "2021-07-10T19:00:22.563187",
     "exception": false,
     "start_time": "2021-07-10T19:00:22.528317",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "#### 2.3.1.6 Load data\n",
    "\n",
    "#### A small description of the data:\n",
    "\n",
    "In Ratschlab's GP-VAE, the `testing` flag is `true` implies train = full train and test = test.\n",
    "If it's `false`, it implies train = part of full train, test = remaining part of full train (i.e val split)\n",
    "\n",
    "###### ToDo: Write more about TimeSeries and how for PhysioNet, x_train_full and x_val_full is the same\n",
    "\n",
    "1. Training Set:\n",
    "- x_train_full is the full data available, which already has missing values (0 instead of 1).\n",
    "- x_train_miss is same as x_train full FOR TRAINING SET\n",
    "- m_train_miss is mask which tells us which datapoints are 0 and so are to be imputed using GP-VAE. This includes the original points which were 0 (aka not available) as well as the ones which were manually dropped by us\n",
    "\n",
    "2. Testing Set\n",
    "- x_test_full is the full available time series data\n",
    "- x_test_miss is the one created by dropping X % (here 1.5%) of available datapoints\n",
    "- m_test_miss is the mask which tells us which datapoints are 0 and so are to be imputed using GP-VAE. This includes the original points which were 0 (aka not available) as well as the ones which were manually dropped by us"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "088cb26d-d4c1-4371-a627-60400856a599",
   "metadata": {
    "papermill": {
     "duration": 0.044624,
     "end_time": "2021-07-10T19:00:22.642979",
     "exception": false,
     "start_time": "2021-07-10T19:00:22.598355",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "data = np.load(data_dir)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "d01df4e1-c7a1-4038-9ab4-33dfab157444",
   "metadata": {
    "papermill": {
     "duration": 0.148632,
     "end_time": "2021-07-10T19:00:22.826850",
     "exception": false,
     "start_time": "2021-07-10T19:00:22.678218",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "x_train_full = data['x_train_full']\n",
    "x_train_miss = data['x_train_miss']\n",
    "m_train_miss = data['m_train_miss']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "88f649fa-ad49-4868-af89-5b8fd842e902",
   "metadata": {
    "papermill": {
     "duration": 0.042257,
     "end_time": "2021-07-10T19:00:22.904393",
     "exception": false,
     "start_time": "2021-07-10T19:00:22.862136",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "if data_type in ['hmnist', 'physionet']:\n",
    "    y_train = data['y_train']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "d5b508cc-df38-47e0-a9f6-590186250848",
   "metadata": {
    "papermill": {
     "duration": 0.189586,
     "end_time": "2021-07-10T19:00:23.128979",
     "exception": false,
     "start_time": "2021-07-10T19:00:22.939393",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "if testing:\n",
    "    if data_type in ['hmnist', 'sprites']:\n",
    "        x_val_full = data['x_test_full']\n",
    "        x_val_miss = data['x_test_miss']\n",
    "        m_val_miss = data['m_test_miss']\n",
    "    if data_type == 'hmnist':\n",
    "        y_val = data['y_test']\n",
    "    elif data_type == 'physionet':\n",
    "        x_val_full = data['x_train_full']\n",
    "        x_val_miss = data['x_train_miss']\n",
    "        m_val_miss = data['m_train_miss']\n",
    "        y_val = data['y_train']\n",
    "        m_val_artificial = data[\"m_train_artificial\"]\n",
    "elif data_type in ['hmnist', 'sprites']:\n",
    "    x_val_full = x_train_full[val_split:]\n",
    "    x_val_miss = x_train_miss[val_split:]\n",
    "    m_val_miss = m_train_miss[val_split:]\n",
    "    if data_type == 'hmnist':\n",
    "        y_val = y_train[val_split:]\n",
    "    x_train_full = x_train_full[:val_split]\n",
    "    x_train_miss = x_train_miss[:val_split]\n",
    "    m_train_miss = m_train_miss[:val_split]\n",
    "    y_train = y_train[:val_split]\n",
    "elif data_type == 'physionet':\n",
    "    x_val_full = data[\"x_val_full\"]  # full for artificial missings\n",
    "    x_val_miss = data[\"x_val_miss\"]\n",
    "    m_val_miss = data[\"m_val_miss\"]\n",
    "    m_val_artificial = data[\"m_val_artificial\"]\n",
    "    y_val = data[\"y_val\"]\n",
    "else:\n",
    "    raise ValueError(\"Data type must be one of ['hmnist', 'physionet', 'sprites']\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "fbec6145-d062-4648-8d84-1b221176fd2c",
   "metadata": {
    "papermill": {
     "duration": 2.153319,
     "end_time": "2021-07-10T19:00:25.318642",
     "exception": false,
     "start_time": "2021-07-10T19:00:23.165323",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "tf_x_train_miss = tf.data.Dataset.from_tensor_slices((x_train_miss, m_train_miss))\\\n",
    "                                 .shuffle(len(x_train_miss)).batch(batch_size).repeat()\n",
    "tf_x_val_miss = tf.data.Dataset.from_tensor_slices((x_val_miss, m_val_miss)).batch(batch_size).repeat()\n",
    "tf_x_val_miss = tf.compat.v1.data.make_one_shot_iterator(tf_x_val_miss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "3eef4f23-2456-447e-9956-49f99e21ec3f",
   "metadata": {
    "papermill": {
     "duration": 0.042695,
     "end_time": "2021-07-10T19:00:25.402421",
     "exception": false,
     "start_time": "2021-07-10T19:00:25.359726",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Build Conv2D preprocessor for image data\n",
    "if data_type in ['hmnist', 'sprites']:\n",
    "    print(\"Using CNN preprocessor\")\n",
    "    image_preprocessor = ImagePreprocessor(img_shape, cnn_sizes, cnn_kernel_size)\n",
    "elif data_type == 'physionet':\n",
    "    image_preprocessor = None\n",
    "else:\n",
    "    raise ValueError(\"Data type must be one of ['hmnist', 'physionet', 'sprites']\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "63a11f75-3f59-4bb6-995b-05c347d8903d",
   "metadata": {
    "papermill": {
     "duration": 0.034967,
     "end_time": "2021-07-10T19:00:25.472455",
     "exception": false,
     "start_time": "2021-07-10T19:00:25.437488",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "#### 2.3.1.7 Build Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "8021272f-9373-46eb-afb2-dd4590ae64ec",
   "metadata": {
    "papermill": {
     "duration": 0.081137,
     "end_time": "2021-07-10T19:00:25.596436",
     "exception": false,
     "start_time": "2021-07-10T19:00:25.515299",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "if model_type == \"vae\":\n",
    "    model = VAE(latent_dim=latent_dim, data_dim=data_dim, time_length=time_length,\n",
    "                encoder_sizes=encoder_sizes, encoder=DiagonalEncoder,\n",
    "                decoder_sizes=decoder_sizes, decoder=decoder,\n",
    "                image_preprocessor=image_preprocessor, window_size=window_size,\n",
    "                beta=beta, M=M, K=K)\n",
    "elif model_type == \"hi-vae\":\n",
    "    model = HI_VAE(latent_dim=latent_dim, data_dim=data_dim, time_length=time_length,\n",
    "                   encoder_sizes=encoder_sizes, encoder=DiagonalEncoder,\n",
    "                   decoder_sizes=decoder_sizes, decoder=decoder,\n",
    "                   image_preprocessor=image_preprocessor, window_size=window_size,\n",
    "                   beta=beta, M=M, K=K)\n",
    "elif model_type == \"gp-vae\":\n",
    "    encoder = BandedJointEncoder if banded_covar else JointEncoder\n",
    "    model = GP_VAE(latent_dim=latent_dim, data_dim=data_dim, time_length=time_length,\n",
    "                   encoder_sizes=encoder_sizes, encoder=encoder,\n",
    "                   decoder_sizes=decoder_sizes, decoder=decoder,\n",
    "                   kernel=kernel, sigma=sigma,\n",
    "                   length_scale=length_scale, kernel_scales = kernel_scales,\n",
    "                   image_preprocessor=image_preprocessor, window_size=window_size,\n",
    "                   beta=beta, M=M, K=K, data_type=data_type)\n",
    "else:\n",
    "    raise ValueError(\"Model type must be one of ['vae', 'hi-vae', 'gp-vae']\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "202941d4-2abc-438a-9c40-d5eb8c04b5bb",
   "metadata": {
    "papermill": {
     "duration": 0.035486,
     "end_time": "2021-07-10T19:00:25.668608",
     "exception": false,
     "start_time": "2021-07-10T19:00:25.633122",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "#### 2.3.1.8 Training preparation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "1d7c416c-8e38-4e49-ba91-17f66935f56b",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-07-10T19:00:25.748415Z",
     "iopub.status.busy": "2021-07-10T19:00:25.747713Z",
     "iopub.status.idle": "2021-07-10T19:00:25.753066Z",
     "shell.execute_reply": "2021-07-10T19:00:25.753483Z",
     "shell.execute_reply.started": "2021-06-28T13:04:16.577538Z"
    },
    "papermill": {
     "duration": 0.049454,
     "end_time": "2021-07-10T19:00:25.753622",
     "exception": false,
     "start_time": "2021-07-10T19:00:25.704168",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "GPU support:  True\n"
     ]
    }
   ],
   "source": [
    "print(\"GPU support: \", tf.test.is_gpu_available())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "24246a75-ea27-441f-ba5a-66bf2897a62a",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-07-10T19:00:25.830258Z",
     "iopub.status.busy": "2021-07-10T19:00:25.829682Z",
     "iopub.status.idle": "2021-07-10T19:00:30.824028Z",
     "shell.execute_reply": "2021-07-10T19:00:30.823227Z",
     "shell.execute_reply.started": "2021-06-28T13:04:16.588252Z"
    },
    "papermill": {
     "duration": 5.034784,
     "end_time": "2021-07-10T19:00:30.824166",
     "exception": false,
     "start_time": "2021-07-10T19:00:25.789382",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training...\n",
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv1d (Conv1D)              (1, 48, 128)              107648    \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (1, 48, 128)              16512     \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (1, 48, 105)              13545     \n",
      "=================================================================\n",
      "Total params: 137,705\n",
      "Trainable params: 137,705\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Encoder:  None\n",
      "Model: \"sequential_1\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_2 (Dense)              (1, 48, 256)              9216      \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (1, 48, 256)              65792     \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (1, 48, 35)               8995      \n",
      "=================================================================\n",
      "Total params: 84,003\n",
      "Trainable params: 84,003\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "Decoder:  None\n"
     ]
    }
   ],
   "source": [
    "print(\"Training...\")\n",
    "_ = tf.compat.v1.train.get_or_create_global_step()\n",
    "trainable_vars = model.get_trainable_vars()\n",
    "optimizer = tf.compat.v1.train.AdamOptimizer(learning_rate=learning_rate)\n",
    "\n",
    "print(\"Encoder: \", model.encoder.net.summary())\n",
    "print(\"Decoder: \", model.decoder.net.summary())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "3f184ce4-5828-493e-9fba-e6a83bece115",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-07-10T19:00:30.903242Z",
     "iopub.status.busy": "2021-07-10T19:00:30.902717Z",
     "iopub.status.idle": "2021-07-10T19:00:30.906018Z",
     "shell.execute_reply": "2021-07-10T19:00:30.906504Z",
     "shell.execute_reply.started": "2021-06-28T13:04:21.471394Z"
    },
    "papermill": {
     "duration": 0.046132,
     "end_time": "2021-07-10T19:00:30.906639",
     "exception": false,
     "start_time": "2021-07-10T19:00:30.860507",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "if model.preprocessor is not None:\n",
    "    print(\"Preprocessor: \", model.preprocessor.net.summary())\n",
    "    saver = tf.compat.v1.train.Checkpoint(optimizer=optimizer, encoder=model.encoder.net,\n",
    "                                          decoder=model.decoder.net, preprocessor=model.preprocessor.net,\n",
    "                                          optimizer_step=tf.compat.v1.train.get_or_create_global_step())\n",
    "else:\n",
    "    saver = tf.compat.v1.train.Checkpoint(optimizer=optimizer, encoder=model.encoder.net, decoder=model.decoder.net,\n",
    "                                          optimizer_step=tf.compat.v1.train.get_or_create_global_step())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "1f6dfafd-a229-429b-af40-9656ca384e68",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-07-10T19:00:30.982559Z",
     "iopub.status.busy": "2021-07-10T19:00:30.982068Z",
     "iopub.status.idle": "2021-07-10T19:00:30.991565Z",
     "shell.execute_reply": "2021-07-10T19:00:30.991130Z",
     "shell.execute_reply.started": "2021-06-28T13:04:21.482243Z"
    },
    "papermill": {
     "duration": 0.048945,
     "end_time": "2021-07-10T19:00:30.991674",
     "exception": false,
     "start_time": "2021-07-10T19:00:30.942729",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "summary_writer = tf.compat.v2.summary.create_file_writer(logdir=outdir, flush_millis=10000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "51f9f1bc-9a52-42e3-a5ed-8cfd28ad4c89",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-07-10T19:00:31.068605Z",
     "iopub.status.busy": "2021-07-10T19:00:31.068069Z",
     "iopub.status.idle": "2021-07-10T19:00:31.071509Z",
     "shell.execute_reply": "2021-07-10T19:00:31.071895Z",
     "shell.execute_reply.started": "2021-06-28T13:04:21.503080Z"
    },
    "papermill": {
     "duration": 0.043677,
     "end_time": "2021-07-10T19:00:31.072023",
     "exception": false,
     "start_time": "2021-07-10T19:00:31.028346",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "if num_steps == 0:\n",
    "    num_steps = num_epochs * len(x_train_miss) // batch_size\n",
    "else:\n",
    "    num_steps = num_steps"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "1f0ed49d-403e-496e-b21d-53e0b1736898",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-07-10T19:00:31.147703Z",
     "iopub.status.busy": "2021-07-10T19:00:31.146485Z",
     "iopub.status.idle": "2021-07-10T19:00:31.148796Z",
     "shell.execute_reply": "2021-07-10T19:00:31.149194Z",
     "shell.execute_reply.started": "2021-06-28T13:04:21.512369Z"
    },
    "papermill": {
     "duration": 0.041529,
     "end_time": "2021-07-10T19:00:31.149320",
     "exception": false,
     "start_time": "2021-07-10T19:00:31.107791",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "if print_interval == 0:\n",
    "    print_interval = num_steps // num_epochs"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "13959788-36b3-493a-a8fc-0a78055b82ec",
   "metadata": {
    "papermill": {
     "duration": 0.035984,
     "end_time": "2021-07-10T19:00:31.221680",
     "exception": false,
     "start_time": "2021-07-10T19:00:31.185696",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "#### 2.3.1.9 Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "0f42586c-f858-411c-8d8e-388ad45e17eb",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-07-10T19:00:31.297043Z",
     "iopub.status.busy": "2021-07-10T19:00:31.296421Z",
     "iopub.status.idle": "2021-07-10T19:00:31.299181Z",
     "shell.execute_reply": "2021-07-10T19:00:31.298794Z",
     "shell.execute_reply.started": "2021-06-28T13:04:21.518211Z"
    },
    "papermill": {
     "duration": 0.041707,
     "end_time": "2021-07-10T19:00:31.299294",
     "exception": false,
     "start_time": "2021-07-10T19:00:31.257587",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "losses_train = []\n",
    "losses_val = []"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "f1081b45-b608-4435-a681-dcd9d8356964",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-07-10T19:00:31.374238Z",
     "iopub.status.busy": "2021-07-10T19:00:31.373566Z",
     "iopub.status.idle": "2021-07-10T19:00:31.375685Z",
     "shell.execute_reply": "2021-07-10T19:00:31.376111Z",
     "shell.execute_reply.started": "2021-06-28T13:04:21.526898Z"
    },
    "papermill": {
     "duration": 0.041207,
     "end_time": "2021-07-10T19:00:31.376224",
     "exception": false,
     "start_time": "2021-07-10T19:00:31.335017",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "t0 = time.time()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "bec760b9-087c-49ee-a55a-697d133fd7af",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-07-10T19:00:31.464116Z",
     "iopub.status.busy": "2021-07-10T19:00:31.460497Z",
     "iopub.status.idle": "2021-07-10T19:24:37.697876Z",
     "shell.execute_reply": "2021-07-10T19:24:37.697309Z"
    },
    "papermill": {
     "duration": 1446.285872,
     "end_time": "2021-07-10T19:24:37.698025",
     "exception": false,
     "start_time": "2021-07-10T19:00:31.412153",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "================================================\n",
      "Learning rate: 0.001 | Global gradient norm: 9908.78\n",
      "Step 0) Time = 1.067125\n",
      "Train loss = 59556.738 | NLL = 433.687 | KL = 295615.250\n",
      "Validation loss = 59752.488 | NLL = 420.138 | KL = 296661.750\n",
      "MSE miss: 1.0416 | AUROC: 0.6671\n",
      "================================================\n",
      "Learning rate: 0.001 | Global gradient norm: 1077.27\n",
      "Step 62) Time = 15.317358\n",
      "Train loss = 22612.756 | NLL = 365.161 | KL = 111237.984\n",
      "Validation loss = 22203.000 | NLL = 357.304 | KL = 109228.469\n",
      "MSE miss: 0.7399 | AUROC: 0.6958\n",
      "================================================\n",
      "Learning rate: 0.001 | Global gradient norm: 1534.75\n",
      "Step 124) Time = 14.760513\n",
      "Train loss = 21929.746 | NLL = 444.877 | KL = 107424.344\n",
      "Validation loss = 21805.504 | NLL = 376.892 | KL = 107143.055\n",
      "MSE miss: 0.7357 | AUROC: 0.6947\n",
      "================================================\n",
      "Learning rate: 0.001 | Global gradient norm: 326.75\n",
      "Step 186) Time = 14.234811\n",
      "Train loss = 21742.395 | NLL = 403.115 | KL = 106696.391\n",
      "Validation loss = 21695.477 | NLL = 341.212 | KL = 106771.328\n",
      "MSE miss: 0.6932 | AUROC: 0.7000\n",
      "================================================\n",
      "Learning rate: 0.001 | Global gradient norm: 200.83\n",
      "Step 248) Time = 14.817866\n",
      "Train loss = 21611.592 | NLL = 352.494 | KL = 106295.492\n",
      "Validation loss = 21593.701 | NLL = 382.133 | KL = 106057.836\n",
      "MSE miss: 0.6809 | AUROC: 0.7067\n",
      "================================================\n",
      "Learning rate: 0.001 | Global gradient norm: 179.64\n",
      "Step 310) Time = 14.242649\n",
      "Train loss = 21600.891 | NLL = 359.603 | KL = 106206.445\n",
      "Validation loss = 21615.424 | NLL = 361.914 | KL = 106267.547\n",
      "MSE miss: 0.6699 | AUROC: 0.7071\n",
      "================================================\n",
      "Learning rate: 0.001 | Global gradient norm: 245.58\n",
      "Step 372) Time = 13.978181\n",
      "Train loss = 21561.254 | NLL = 368.231 | KL = 105965.109\n",
      "Validation loss = 21596.670 | NLL = 395.254 | KL = 106007.078\n",
      "MSE miss: 0.6719 | AUROC: 0.7087\n",
      "================================================\n",
      "Learning rate: 0.001 | Global gradient norm: 301.36\n",
      "Step 434) Time = 14.201883\n",
      "Train loss = 21561.367 | NLL = 347.071 | KL = 106071.477\n",
      "Validation loss = 21550.922 | NLL = 351.988 | KL = 105994.656\n",
      "MSE miss: 0.6611 | AUROC: 0.7102\n",
      "================================================\n",
      "Learning rate: 0.001 | Global gradient norm: 181.57\n",
      "Step 496) Time = 14.091882\n",
      "Train loss = 21544.611 | NLL = 338.139 | KL = 106032.359\n",
      "Validation loss = 21526.488 | NLL = 333.973 | KL = 105962.578\n",
      "MSE miss: 0.6490 | AUROC: 0.7147\n",
      "================================================\n",
      "Learning rate: 0.001 | Global gradient norm: 153.37\n",
      "Step 558) Time = 13.799201\n",
      "Train loss = 21539.354 | NLL = 353.010 | KL = 105931.719\n",
      "Validation loss = 21566.820 | NLL = 372.704 | KL = 105970.578\n",
      "MSE miss: 0.6353 | AUROC: 0.7155\n",
      "================================================\n",
      "Learning rate: 0.001 | Global gradient norm: 209.40\n",
      "Step 620) Time = 13.935709\n",
      "Train loss = 21535.285 | NLL = 346.436 | KL = 105944.242\n",
      "Validation loss = 21607.729 | NLL = 399.739 | KL = 106039.953\n",
      "MSE miss: 0.6231 | AUROC: 0.7200\n",
      "================================================\n",
      "Learning rate: 0.0005 | Global gradient norm: 193.69\n",
      "Step 682) Time = 13.953329\n",
      "Train loss = 21523.266 | NLL = 340.273 | KL = 105914.953\n",
      "Validation loss = 21543.734 | NLL = 360.886 | KL = 105914.250\n",
      "MSE miss: 0.6139 | AUROC: 0.7203\n",
      "================================================\n",
      "Learning rate: 0.0005 | Global gradient norm: 219.39\n",
      "Step 744) Time = 14.810445\n",
      "Train loss = 21521.896 | NLL = 339.817 | KL = 105910.391\n",
      "Validation loss = 21511.289 | NLL = 330.818 | KL = 105902.352\n",
      "MSE miss: 0.6133 | AUROC: 0.7209\n",
      "================================================\n",
      "Learning rate: 0.0005 | Global gradient norm: 163.33\n",
      "Step 806) Time = 14.572141\n",
      "Train loss = 21517.719 | NLL = 337.748 | KL = 105899.859\n",
      "Validation loss = 21542.396 | NLL = 358.376 | KL = 105920.102\n",
      "MSE miss: 0.6050 | AUROC: 0.7246\n",
      "================================================\n",
      "Learning rate: 0.0005 | Global gradient norm: 178.13\n",
      "Step 868) Time = 14.677836\n",
      "Train loss = 21518.311 | NLL = 334.458 | KL = 105919.258\n",
      "Validation loss = 21531.691 | NLL = 347.951 | KL = 105918.711\n",
      "MSE miss: 0.6166 | AUROC: 0.7198\n",
      "================================================\n",
      "Learning rate: 0.0005 | Global gradient norm: 173.31\n",
      "Step 930) Time = 14.221939\n",
      "Train loss = 21511.471 | NLL = 330.095 | KL = 105906.875\n",
      "Validation loss = 21533.176 | NLL = 348.009 | KL = 105925.828\n",
      "MSE miss: 0.6003 | AUROC: 0.7298\n",
      "================================================\n",
      "Learning rate: 0.0005 | Global gradient norm: 119.09\n",
      "Step 992) Time = 13.865875\n",
      "Train loss = 21518.904 | NLL = 336.558 | KL = 105911.734\n",
      "Validation loss = 21525.404 | NLL = 341.717 | KL = 105918.430\n",
      "MSE miss: 0.5985 | AUROC: 0.7296\n",
      "================================================\n",
      "Learning rate: 0.0005 | Global gradient norm: 175.24\n",
      "Step 1054) Time = 13.873162\n",
      "Train loss = 21503.949 | NLL = 321.251 | KL = 105913.484\n",
      "Validation loss = 21529.523 | NLL = 347.935 | KL = 105907.945\n",
      "MSE miss: 0.5970 | AUROC: 0.7303\n",
      "================================================\n",
      "Learning rate: 0.0005 | Global gradient norm: 207.67\n",
      "Step 1116) Time = 13.752577\n",
      "Train loss = 21540.486 | NLL = 359.090 | KL = 105906.969\n",
      "Validation loss = 21528.828 | NLL = 349.166 | KL = 105898.312\n",
      "MSE miss: 0.6043 | AUROC: 0.7238\n",
      "================================================\n",
      "Learning rate: 0.0005 | Global gradient norm: 395.73\n",
      "Step 1178) Time = 13.597574\n",
      "Train loss = 21579.906 | NLL = 393.830 | KL = 105930.375\n",
      "Validation loss = 21528.012 | NLL = 347.617 | KL = 105901.969\n",
      "MSE miss: 0.5897 | AUROC: 0.7367\n",
      "================================================\n",
      "Learning rate: 0.0005 | Global gradient norm: 138.18\n",
      "Step 1240) Time = 13.582683\n",
      "Train loss = 21515.248 | NLL = 335.693 | KL = 105897.781\n",
      "Validation loss = 21518.551 | NLL = 338.559 | KL = 105899.953\n",
      "MSE miss: 0.5941 | AUROC: 0.7312\n",
      "================================================\n",
      "Learning rate: 0.00025 | Global gradient norm: 164.77\n",
      "Step 1302) Time = 14.000550\n",
      "Train loss = 21508.994 | NLL = 327.509 | KL = 105907.430\n",
      "Validation loss = 21516.113 | NLL = 338.750 | KL = 105886.812\n",
      "MSE miss: 0.5857 | AUROC: 0.7340\n",
      "================================================\n",
      "Learning rate: 0.00025 | Global gradient norm: 80.96\n",
      "Step 1364) Time = 14.071234\n",
      "Train loss = 21510.611 | NLL = 332.327 | KL = 105891.422\n",
      "Validation loss = 21553.148 | NLL = 368.996 | KL = 105920.766\n",
      "MSE miss: 0.5836 | AUROC: 0.7317\n",
      "================================================\n",
      "Learning rate: 0.00025 | Global gradient norm: 211.20\n",
      "Step 1426) Time = 14.536267\n",
      "Train loss = 21518.611 | NLL = 338.187 | KL = 105902.117\n",
      "Validation loss = 21514.031 | NLL = 336.940 | KL = 105885.453\n",
      "MSE miss: 0.5858 | AUROC: 0.7379\n",
      "================================================\n",
      "Learning rate: 0.00025 | Global gradient norm: 116.31\n",
      "Step 1488) Time = 14.205123\n",
      "Train loss = 21518.725 | NLL = 339.711 | KL = 105895.062\n",
      "Validation loss = 21541.861 | NLL = 359.604 | KL = 105911.281\n",
      "MSE miss: 0.5844 | AUROC: 0.7311\n",
      "================================================\n",
      "Learning rate: 0.00025 | Global gradient norm: 126.95\n",
      "Step 1550) Time = 14.763297\n",
      "Train loss = 21517.621 | NLL = 338.873 | KL = 105893.734\n",
      "Validation loss = 21542.506 | NLL = 363.971 | KL = 105892.672\n",
      "MSE miss: 0.5809 | AUROC: 0.7364\n",
      "================================================\n",
      "Learning rate: 0.00025 | Global gradient norm: 115.23\n",
      "Step 1612) Time = 14.020620\n",
      "Train loss = 21527.809 | NLL = 348.988 | KL = 105894.109\n",
      "Validation loss = 21532.965 | NLL = 354.022 | KL = 105894.719\n",
      "MSE miss: 0.5806 | AUROC: 0.7360\n",
      "================================================\n",
      "Learning rate: 0.00025 | Global gradient norm: 212.66\n",
      "Step 1674) Time = 13.873818\n",
      "Train loss = 21516.795 | NLL = 337.513 | KL = 105896.406\n",
      "Validation loss = 21523.248 | NLL = 343.689 | KL = 105897.789\n",
      "MSE miss: 0.5794 | AUROC: 0.7412\n",
      "================================================\n",
      "Learning rate: 0.00025 | Global gradient norm: 143.10\n",
      "Step 1736) Time = 14.045883\n",
      "Train loss = 21495.004 | NLL = 318.363 | KL = 105883.195\n",
      "Validation loss = 21520.062 | NLL = 341.661 | KL = 105892.008\n",
      "MSE miss: 0.5776 | AUROC: 0.7380\n",
      "================================================\n",
      "Learning rate: 0.00025 | Global gradient norm: 136.31\n",
      "Step 1798) Time = 14.022536\n",
      "Train loss = 21525.113 | NLL = 347.114 | KL = 105889.984\n",
      "Validation loss = 21509.289 | NLL = 330.166 | KL = 105895.609\n",
      "MSE miss: 0.5803 | AUROC: 0.7369\n",
      "================================================\n",
      "Learning rate: 0.00025 | Global gradient norm: 148.43\n",
      "Step 1860) Time = 14.171203\n",
      "Train loss = 21507.285 | NLL = 328.447 | KL = 105894.188\n",
      "Validation loss = 21576.836 | NLL = 397.755 | KL = 105895.391\n",
      "MSE miss: 0.5817 | AUROC: 0.7382\n",
      "================================================\n",
      "Learning rate: 0.000125 | Global gradient norm: 252.34\n",
      "Step 1922) Time = 14.609870\n",
      "Train loss = 21544.656 | NLL = 363.407 | KL = 105906.250\n",
      "Validation loss = 21507.742 | NLL = 331.914 | KL = 105879.141\n",
      "MSE miss: 0.5762 | AUROC: 0.7380\n",
      "================================================\n",
      "Learning rate: 0.000125 | Global gradient norm: 160.44\n",
      "Step 1984) Time = 14.002039\n",
      "Train loss = 21521.482 | NLL = 343.506 | KL = 105889.875\n",
      "Validation loss = 21514.584 | NLL = 336.097 | KL = 105892.430\n",
      "MSE miss: 0.5764 | AUROC: 0.7387\n",
      "================================================\n",
      "Learning rate: 0.000125 | Global gradient norm: 103.40\n",
      "Step 2046) Time = 14.610659\n",
      "Train loss = 21507.805 | NLL = 331.032 | KL = 105883.859\n",
      "Validation loss = 21494.094 | NLL = 316.191 | KL = 105889.516\n",
      "MSE miss: 0.5771 | AUROC: 0.7378\n",
      "================================================\n",
      "Learning rate: 0.000125 | Global gradient norm: 120.90\n",
      "Step 2108) Time = 13.805177\n",
      "Train loss = 21511.635 | NLL = 333.053 | KL = 105892.898\n",
      "Validation loss = 21505.586 | NLL = 328.256 | KL = 105886.656\n",
      "MSE miss: 0.5763 | AUROC: 0.7399\n",
      "================================================\n",
      "Learning rate: 0.000125 | Global gradient norm: 430.59\n",
      "Step 2170) Time = 14.054384\n",
      "Train loss = 21514.305 | NLL = 336.809 | KL = 105887.484\n",
      "Validation loss = 21520.699 | NLL = 342.852 | KL = 105889.234\n",
      "MSE miss: 0.5743 | AUROC: 0.7404\n",
      "================================================\n",
      "Learning rate: 0.000125 | Global gradient norm: 138.39\n",
      "Step 2232) Time = 14.314467\n",
      "Train loss = 21547.428 | NLL = 369.352 | KL = 105890.375\n",
      "Validation loss = 21517.281 | NLL = 339.001 | KL = 105891.391\n",
      "MSE miss: 0.5742 | AUROC: 0.7378\n",
      "================================================\n",
      "Learning rate: 0.000125 | Global gradient norm: 121.93\n",
      "Step 2294) Time = 14.265030\n",
      "Train loss = 21526.656 | NLL = 348.195 | KL = 105892.305\n",
      "Validation loss = 21524.441 | NLL = 346.332 | KL = 105890.547\n",
      "MSE miss: 0.5750 | AUROC: 0.7388\n",
      "================================================\n",
      "Learning rate: 0.000125 | Global gradient norm: 103.34\n",
      "Step 2356) Time = 14.404418\n",
      "Train loss = 21523.629 | NLL = 346.337 | KL = 105886.453\n",
      "Validation loss = 21516.553 | NLL = 338.984 | KL = 105887.844\n",
      "MSE miss: 0.5737 | AUROC: 0.7411\n",
      "================================================\n",
      "Learning rate: 0.000125 | Global gradient norm: 133.88\n",
      "Step 2418) Time = 14.295514\n",
      "Train loss = 21517.123 | NLL = 338.063 | KL = 105895.297\n",
      "Validation loss = 21502.221 | NLL = 324.429 | KL = 105888.953\n",
      "MSE miss: 0.5741 | AUROC: 0.7397\n",
      "================================================\n",
      "Learning rate: 0.000125 | Global gradient norm: 103.17\n",
      "Step 2480) Time = 14.692725\n",
      "Train loss = 21521.375 | NLL = 342.874 | KL = 105892.508\n",
      "Validation loss = 21509.994 | NLL = 332.793 | KL = 105886.000\n",
      "MSE miss: 0.5741 | AUROC: 0.7389\n"
     ]
    }
   ],
   "source": [
    "with summary_writer.as_default(), tf.compat.v2.summary.record_if(True):\n",
    "    for i, (x_seq, m_seq) in enumerate(tf_x_train_miss.take(num_steps)):\n",
    "        try:\n",
    "            with tf.GradientTape() as tape:\n",
    "                tape.watch(trainable_vars)\n",
    "                loss = model.compute_loss(x_seq, m_mask=m_seq)\n",
    "                losses_train.append(loss.numpy())\n",
    "            grads = tape.gradient(loss, trainable_vars)\n",
    "            grads = [np.nan_to_num(grad) for grad in grads]\n",
    "            grads, global_norm = tf.clip_by_global_norm(grads, gradient_clip)\n",
    "            optimizer.apply_gradients(zip(grads, trainable_vars),\n",
    "                                      global_step=tf.compat.v1.train.get_or_create_global_step())\n",
    "\n",
    "            # Print intermediate results\n",
    "            if i % print_interval == 0:\n",
    "                print(\"================================================\")\n",
    "                print(\"Learning rate: {} | Global gradient norm: {:.2f}\".format(optimizer._lr, global_norm))\n",
    "                print(\"Step {}) Time = {:2f}\".format(i, time.time() - t0))\n",
    "                loss, nll, kl = model.compute_loss(x_seq, m_mask=m_seq, return_parts=True)\n",
    "                print(\"Train loss = {:.3f} | NLL = {:.3f} | KL = {:.3f}\".format(loss, nll, kl))\n",
    "\n",
    "                saver.save(checkpoint_prefix)\n",
    "                tf.compat.v2.summary.scalar(name=\"loss_train\", data=loss, step=tf.compat.v1.train.get_or_create_global_step())\n",
    "                tf.compat.v2.summary.scalar(name=\"kl_train\", data=kl, step=tf.compat.v1.train.get_or_create_global_step())\n",
    "                tf.compat.v2.summary.scalar(name=\"nll_train\", data=nll, step=tf.compat.v1.train.get_or_create_global_step())\n",
    "\n",
    "                # Validation loss\n",
    "                x_val_batch, m_val_batch = tf_x_val_miss.get_next()\n",
    "                val_loss, val_nll, val_kl = model.compute_loss(x_val_batch, m_mask=m_val_batch, return_parts=True)\n",
    "                losses_val.append(val_loss.numpy())\n",
    "                print(\"Validation loss = {:.3f} | NLL = {:.3f} | KL = {:.3f}\".format(val_loss, val_nll, val_kl))\n",
    "\n",
    "                tf.compat.v2.summary.scalar(name=\"loss_val\", data=val_loss, step=tf.compat.v1.train.get_or_create_global_step())\n",
    "                tf.compat.v2.summary.scalar(name=\"kl_val\", data=val_kl, step=tf.compat.v1.train.get_or_create_global_step())\n",
    "                tf.compat.v2.summary.scalar(name=\"nll_val\", data=val_nll, step=tf.compat.v1.train.get_or_create_global_step())\n",
    "\n",
    "                if data_type in [\"hmnist\", \"sprites\"]:\n",
    "                    # Draw reconstructed images\n",
    "                    x_hat = model.decode(model.encode(x_seq).sample()).mean()\n",
    "                    tf.compat.v2.summary.image(name=\"input_train\", data=tf.reshape(x_seq, [-1]+list(img_shape)), step=tf.compat.v1.train.get_or_create_global_step())\n",
    "                    tf.compat.v2.summary.image(name=\"reconstruction_train\", data=tf.reshape(x_hat, [-1]+list(img_shape)), step=tf.compat.v1.train.get_or_create_global_step())\n",
    "                elif data_type == 'physionet':\n",
    "                    # Eval MSE and AUROC on entire val set\n",
    "                    x_val_miss_batches = np.array_split(x_val_miss, batch_size, axis=0)\n",
    "                    x_val_full_batches = np.array_split(x_val_full, batch_size, axis=0)\n",
    "                    m_val_artificial_batches = np.array_split(m_val_artificial, batch_size, axis=0)\n",
    "                    get_val_batches = lambda: zip(x_val_miss_batches, x_val_full_batches, m_val_artificial_batches)\n",
    "\n",
    "                    n_missings = m_val_artificial.sum()\n",
    "                    mse_miss = np.sum([model.compute_mse(x, y=y, m_mask=m).numpy()\n",
    "                                       for x, y, m in get_val_batches()]) / n_missings\n",
    "\n",
    "                    x_val_imputed = np.vstack([model.decode(model.encode(x_batch).mean()).mean().numpy()\n",
    "                                               for x_batch in x_val_miss_batches])\n",
    "                    x_val_imputed[m_val_miss == 0] = x_val_miss[m_val_miss == 0]  # impute gt observed values\n",
    "\n",
    "                    x_val_imputed = x_val_imputed.reshape([-1, time_length * data_dim])\n",
    "                    val_split = len(x_val_imputed) // 2\n",
    "                    cls_model = LogisticRegression(solver='liblinear', tol=1e-10, max_iter=10000)\n",
    "                    cls_model.fit(x_val_imputed[:val_split], y_val[:val_split])\n",
    "                    probs = cls_model.predict_proba(x_val_imputed[val_split:])[:, 1]\n",
    "                    auroc = roc_auc_score(y_val[val_split:], probs)\n",
    "                    print(\"MSE miss: {:.4f} | AUROC: {:.4f}\".format(mse_miss, auroc))\n",
    "\n",
    "                    # Update learning rate (used only for physionet with decay=0.5)\n",
    "                    if i > 0 and i % (10*print_interval) == 0:\n",
    "                        optimizer._lr = max(0.5 * optimizer._lr, 0.1 * learning_rate)\n",
    "                t0 = time.time()\n",
    "        except KeyboardInterrupt as e:\n",
    "            print(\"KeyboardInterrupt\")\n",
    "            saver.save(checkpoint_prefix)\n",
    "#             if debug:\n",
    "#                 import ipdb\n",
    "#                 ipdb.set_trace()\n",
    "            break"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "48644ef0-6225-460b-a647-7b38d3bb25cd",
   "metadata": {
    "papermill": {
     "duration": 0.066757,
     "end_time": "2021-07-10T19:24:37.832252",
     "exception": false,
     "start_time": "2021-07-10T19:24:37.765495",
     "status": "completed"
    },
    "tags": []
   },
   "source": [
    "#### 2.3.1.10 Evaluation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "ce62df31-d76c-44e4-bf2d-a4f4f26aed79",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-07-10T19:24:37.970809Z",
     "iopub.status.busy": "2021-07-10T19:24:37.970100Z",
     "iopub.status.idle": "2021-07-10T19:24:37.973819Z",
     "shell.execute_reply": "2021-07-10T19:24:37.973344Z"
    },
    "papermill": {
     "duration": 0.074863,
     "end_time": "2021-07-10T19:24:37.973926",
     "exception": false,
     "start_time": "2021-07-10T19:24:37.899063",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Split data on batches\n",
    "x_val_miss_batches = np.array_split(x_val_miss, batch_size, axis=0)\n",
    "x_val_full_batches = np.array_split(x_val_full, batch_size, axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "3379f344-b3f7-4fc6-8536-5484d23563fa",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-07-10T19:24:38.112701Z",
     "iopub.status.busy": "2021-07-10T19:24:38.112183Z",
     "iopub.status.idle": "2021-07-10T19:24:38.116188Z",
     "shell.execute_reply": "2021-07-10T19:24:38.115694Z"
    },
    "papermill": {
     "duration": 0.074825,
     "end_time": "2021-07-10T19:24:38.116290",
     "exception": false,
     "start_time": "2021-07-10T19:24:38.041465",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "if data_type == 'physionet':\n",
    "    m_val_batches = np.array_split(m_val_artificial, batch_size, axis=0)\n",
    "else:\n",
    "    m_val_batches = np.array_split(m_val_miss, batch_size, axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "efc63881-a92f-47ca-b4c8-1fe08d3fb087",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-07-10T19:24:38.254850Z",
     "iopub.status.busy": "2021-07-10T19:24:38.253639Z",
     "iopub.status.idle": "2021-07-10T19:24:38.255874Z",
     "shell.execute_reply": "2021-07-10T19:24:38.256348Z"
    },
    "papermill": {
     "duration": 0.073223,
     "end_time": "2021-07-10T19:24:38.256479",
     "exception": false,
     "start_time": "2021-07-10T19:24:38.183256",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "get_val_batches = lambda: zip(x_val_miss_batches, x_val_full_batches, m_val_batches)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "1a18eeb4-23a8-438c-bcdb-679568aee753",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-07-10T19:24:38.396011Z",
     "iopub.status.busy": "2021-07-10T19:24:38.395213Z",
     "iopub.status.idle": "2021-07-10T19:24:54.347141Z",
     "shell.execute_reply": "2021-07-10T19:24:54.346670Z"
    },
    "papermill": {
     "duration": 16.023699,
     "end_time": "2021-07-10T19:24:54.347273",
     "exception": false,
     "start_time": "2021-07-10T19:24:38.323574",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Compute NLL and MSE on missing values\n",
    "n_missings = m_val_artificial.sum() if data_type == 'physionet' else m_val_miss.sum()\n",
    "nll_miss = np.sum([model.compute_nll(x, y=y, m_mask=m).numpy()\n",
    "                   for x, y, m in get_val_batches()]) / n_missings\n",
    "mse_miss = np.sum([model.compute_mse(x, y=y, m_mask=m, binary=data_type==\"hmnist\").numpy()\n",
    "                   for x, y, m in get_val_batches()]) / n_missings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "0ba6e402-53a8-46af-9fa2-330154202097",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-07-10T19:24:54.494220Z",
     "iopub.status.busy": "2021-07-10T19:24:54.492269Z",
     "iopub.status.idle": "2021-07-10T19:24:54.507845Z",
     "shell.execute_reply": "2021-07-10T19:24:54.507037Z"
    },
    "papermill": {
     "duration": 0.092712,
     "end_time": "2021-07-10T19:24:54.508012",
     "exception": false,
     "start_time": "2021-07-10T19:24:54.415300",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "NLL miss: 1.2348\n",
      "MSE miss: 0.5735\n"
     ]
    }
   ],
   "source": [
    "print(\"NLL miss: {:.4f}\".format(nll_miss))\n",
    "print(\"MSE miss: {:.4f}\".format(mse_miss))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "b2edcfa6-ae35-43ad-b4a7-58656e4c6154",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-07-10T19:24:54.747202Z",
     "iopub.status.busy": "2021-07-10T19:24:54.746452Z",
     "iopub.status.idle": "2021-07-10T19:25:03.163653Z",
     "shell.execute_reply": "2021-07-10T19:25:03.163180Z"
    },
    "papermill": {
     "duration": 8.537638,
     "end_time": "2021-07-10T19:25:03.163786",
     "exception": false,
     "start_time": "2021-07-10T19:24:54.626148",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# Save imputed values\n",
    "z_mean = [model.encode(x_batch).mean().numpy() for x_batch in x_val_miss_batches]\n",
    "np.save(os.path.join(outdir, \"z_mean\"), np.vstack(z_mean))\n",
    "x_val_imputed = np.vstack([model.decode(z_batch).mean().numpy() for z_batch in z_mean])\n",
    "x_val_imputed_std = np.vstack([model.decode(z_batch).stddev().numpy() for z_batch in z_mean])\n",
    "np.save(os.path.join(outdir, \"imputed_no_gt\"), x_val_imputed)\n",
    "np.save(os.path.join(outdir, \"x_val_imputed_std\"), x_val_imputed_std)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "6497142e-6503-4f7b-a1a4-8831732fda85",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-07-10T19:25:03.304150Z",
     "iopub.status.busy": "2021-07-10T19:25:03.303312Z",
     "iopub.status.idle": "2021-07-10T19:25:03.365593Z",
     "shell.execute_reply": "2021-07-10T19:25:03.365112Z"
    },
    "papermill": {
     "duration": 0.133635,
     "end_time": "2021-07-10T19:25:03.365727",
     "exception": false,
     "start_time": "2021-07-10T19:25:03.232092",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "# impute gt observed values\n",
    "x_val_imputed[m_val_miss == 0] = x_val_miss[m_val_miss == 0]\n",
    "np.save(os.path.join(outdir, \"imputed\"), x_val_imputed)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "e5bb488e-3081-419a-9b64-1a040414c346",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-07-10T19:25:03.512083Z",
     "iopub.status.busy": "2021-07-10T19:25:03.511252Z",
     "iopub.status.idle": "2021-07-10T19:25:09.439807Z",
     "shell.execute_reply": "2021-07-10T19:25:09.440860Z"
    },
    "papermill": {
     "duration": 6.007191,
     "end_time": "2021-07-10T19:25:09.441094",
     "exception": false,
     "start_time": "2021-07-10T19:25:03.433903",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AUROC: 0.7388\n",
      "AUPRC: 0.3402\n"
     ]
    }
   ],
   "source": [
    "if data_type == \"hmnist\":\n",
    "    # AUROC evaluation using Logistic Regression\n",
    "    x_val_imputed = np.round(x_val_imputed)\n",
    "    x_val_imputed = x_val_imputed.reshape([-1, time_length * data_dim])\n",
    "\n",
    "    cls_model = LogisticRegression(solver='lbfgs', multi_class='multinomial', tol=1e-10, max_iter=10000)\n",
    "    val_split = len(x_val_imputed) // 2\n",
    "\n",
    "    cls_model.fit(x_val_imputed[:val_split], y_val[:val_split])\n",
    "    probs = cls_model.predict_proba(x_val_imputed[val_split:])\n",
    "\n",
    "    auprc = average_precision_score(np.eye(num_classes)[y_val[val_split:]], probs)\n",
    "    auroc = roc_auc_score(np.eye(num_classes)[y_val[val_split:]], probs)\n",
    "    print(\"AUROC: {:.4f}\".format(auroc))\n",
    "    print(\"AUPRC: {:.4f}\".format(auprc))\n",
    "\n",
    "elif data_type == \"sprites\":\n",
    "    auroc, auprc = 0, 0\n",
    "\n",
    "elif data_type == \"physionet\":\n",
    "    # Uncomment to preserve some z_samples and their reconstructions\n",
    "    # for i in range(5):\n",
    "    #     z_sample = [model.encode(x_batch).sample().numpy() for x_batch in x_val_miss_batches]\n",
    "    #     np.save(os.path.join(outdir, \"z_sample_{}\".format(i)), np.vstack(z_sample))\n",
    "    #     x_val_imputed_sample = np.vstack([model.decode(z_batch).mean().numpy() for z_batch in z_sample])\n",
    "    #     np.save(os.path.join(outdir, \"imputed_sample_{}_no_gt\".format(i)), x_val_imputed_sample)\n",
    "    #     x_val_imputed_sample[m_val_miss == 0] = x_val_miss[m_val_miss == 0]\n",
    "    #     np.save(os.path.join(outdir, \"imputed_sample_{}\".format(i)), x_val_imputed_sample)\n",
    "\n",
    "    # AUROC evaluation using Logistic Regression\n",
    "    x_val_imputed = x_val_imputed.reshape([-1, time_length * data_dim])\n",
    "    val_split = len(x_val_imputed) // 2\n",
    "    cls_model = LogisticRegression(solver='liblinear', tol=1e-10, max_iter=10000)\n",
    "    cls_model.fit(x_val_imputed[:val_split], y_val[:val_split])\n",
    "    probs = cls_model.predict_proba(x_val_imputed[val_split:])[:, 1]\n",
    "    auprc = average_precision_score(y_val[val_split:], probs)\n",
    "    auroc = roc_auc_score(y_val[val_split:], probs)\n",
    "\n",
    "    print(\"AUROC: {:.4f}\".format(auroc))\n",
    "    print(\"AUPRC: {:.4f}\".format(auprc))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "ebfdbe2a-ee26-406b-91c7-648993115951",
   "metadata": {
    "execution": {
     "iopub.execute_input": "2021-07-10T19:25:09.644992Z",
     "iopub.status.busy": "2021-07-10T19:25:09.644222Z",
     "iopub.status.idle": "2021-07-10T19:25:09.650225Z",
     "shell.execute_reply": "2021-07-10T19:25:09.649837Z"
    },
    "papermill": {
     "duration": 0.08602,
     "end_time": "2021-07-10T19:25:09.650342",
     "exception": false,
     "start_time": "2021-07-10T19:25:09.564322",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training finished.\n"
     ]
    }
   ],
   "source": [
    "# Visualize reconstructions\n",
    "if data_type in [\"hmnist\", \"sprites\"]:\n",
    "    img_index = 0\n",
    "    if data_type == \"hmnist\":\n",
    "        img_shape = (28, 28)\n",
    "        cmap = \"gray\"\n",
    "    elif data_type == \"sprites\":\n",
    "        img_shape = (64, 64, 3)\n",
    "        cmap = None\n",
    "\n",
    "    fig, axes = plt.subplots(nrows=3, ncols=x_val_miss.shape[1], figsize=(2*x_val_miss.shape[1], 6))\n",
    "\n",
    "    x_hat = model.decode(model.encode(x_val_miss[img_index: img_index+1]).mean()).mean().numpy()\n",
    "    seqs = [x_val_miss[img_index:img_index+1], x_hat, x_val_full[img_index:img_index+1]]\n",
    "\n",
    "    for axs, seq in zip(axes, seqs):\n",
    "        for ax, img in zip(axs, seq[0]):\n",
    "            ax.imshow(img.reshape(img_shape), cmap=cmap)\n",
    "            ax.axis('off')\n",
    "\n",
    "    suptitle = model_type + f\" reconstruction, NLL missing = {mse_miss}\"\n",
    "    fig.suptitle(suptitle, size=18)\n",
    "    fig.savefig(os.path.join(outdir, data_type + \"_reconstruction.pdf\"))\n",
    "\n",
    "results_all = [seed, model_type, data_type, kernel, beta, latent_dim,\n",
    "               num_epochs, batch_size, learning_rate, window_size,\n",
    "               kernel_scales, sigma, length_scale,\n",
    "               len(encoder_sizes), encoder_sizes[0] if len(encoder_sizes) > 0 else 0,\n",
    "               len(decoder_sizes), decoder_sizes[0] if len(decoder_sizes) > 0 else 0,\n",
    "               cnn_kernel_size, cnn_sizes,\n",
    "               nll_miss, mse_miss, losses_train[-1], losses_val[-1], auprc, auroc, testing, data_dir]\n",
    "\n",
    "with open(os.path.join(outdir, \"results.tsv\"), \"w\") as outfile:\n",
    "    outfile.write(\"seed\\tmodel\\tdata\\tkernel\\tbeta\\tz_size\\tnum_epochs\"\n",
    "                  \"\\tbatch_size\\tlearning_rate\\twindow_size\\tkernel_scales\\t\"\n",
    "                  \"sigma\\tlength_scale\\tencoder_depth\\tencoder_width\\t\"\n",
    "                  \"decoder_depth\\tdecoder_width\\tcnn_kernel_size\\t\"\n",
    "                  \"cnn_sizes\\tNLL\\tMSE\\tlast_train_loss\\tlast_val_loss\\tAUPRC\\tAUROC\\ttesting\\tdata_dir\\n\")\n",
    "    outfile.write(\"\\t\".join(map(str, results_all)))\n",
    "\n",
    "with open(os.path.join(outdir, \"training_curve.tsv\"), \"w\") as outfile:\n",
    "    outfile.write(\"\\t\".join(map(str, losses_train)))\n",
    "    outfile.write(\"\\n\")\n",
    "    outfile.write(\"\\t\".join(map(str, losses_val)))\n",
    "\n",
    "print(\"Training finished.\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "44f48b23-db8b-4c68-83a1-82f9e654654e",
   "metadata": {},
   "source": [
    "#### 2.3.1.10 Visaualising Results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "4d2cecfd-6808-46ce-a41a-fc42859ba4a8",
   "metadata": {},
   "outputs": [],
   "source": [
    "x_val_full[x_val_full == 0] = 'nan'\n",
    "imputed = np.load(\"models/210710_reproduce_physionet/imputed.npy\")\n",
    "x_val_imputed_std = np.load(\"models/210710_reproduce_physionet/x_val_imputed_std.npy\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "bcb29a83-6902-4b4a-9eca-a48c6f00ff21",
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "87b1da46-cb46-4cef-937b-c62f1be93bfb",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "k=0\n",
    "while k<=5: # 5 random non-nan plots\n",
    "    i = np.random.randint(low=0,high=x_val_full.shape[0])\n",
    "    j = np.random.randint(low=0,high=x_val_full.shape[2])\n",
    "    y1 = x_val_full[i,:,j]\n",
    "    if np.isnan(y1).all():\n",
    "        continue # Continue if full input was NA\n",
    "    y2 = imputed[i,:,j]\n",
    "    yerr = x_val_imputed_std[i,:,j]\n",
    "    dyfit = yerr\n",
    "    x = np.arange(48)\n",
    "    plt.plot(x, y2, '-', label='imputed')\n",
    "    plt.plot(x, y1,'or', label='full')\n",
    "    plt.fill_between(x, y2 - dyfit, y2 + dyfit,\n",
    "                 color='blue', alpha=0.2)\n",
    "    plt.title(f\"Sample #{i}, Channel #{j}\")\n",
    "    plt.legend()\n",
    "    plt.show()\n",
    "    k = k+1\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cfd08059-3d05-415b-84a8-b560b973c71b",
   "metadata": {},
   "source": [
    "### 2.3.2 The above can be run once again for RESSPECT Simulation Data\n",
    "\n",
    "All that needs to be done is that x_train_full, x_train_miss etc needs to be replaced by RESSPECT data.\n",
    "\n",
    "References: \n",
    "1. https://www.kaggle.com/siddharthchaini/gpvae-raw-plasticc?scriptVersionId=66968645\n",
    "2. https://www.kaggle.com/siddharthchaini/visualisation-gpvae-raw-plasticc?scriptVersionId=67006906"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "dc15ce6f-63ec-4cc7-b2c4-491f94471018",
   "metadata": {},
   "source": [
    "## 2.4 Some observations and issues:\n",
    "\n",
    "We had thought that the latent space for the GP-VAE compresses the time series by reducing the number of time series points. However, the compression actually happens across the channels. Because for the LSST, we're already working with 6 passbands, the dimensionality reduction over channels/passbands is not really of importance here.\n",
    "\n",
    "Thus, the latent space cannot be directly used for classification as part of RESSPECT, as we were earlier hoping, \n",
    "unfortunately."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "7f9f807b-1a00-4808-8e5b-02506d99a3f1",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The shape here is: (#Samples, Latent Space Dimensionaity, #Time Steps)\n",
      "(3997, 35, 48)\n"
     ]
    }
   ],
   "source": [
    "latent_mean = np.load(\"gpvae-raw/models/210710_reproduce_physionet/z_mean.npy\")\n",
    "\n",
    "print(\"The shape here is: (#Samples, Latent Space Dimensionaity, #Time Steps)\")\n",
    "print(latent_mean.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5ece86e5-2790-4fcc-82ca-c525162f7b90",
   "metadata": {},
   "source": [
    "Further, we also see that the latent space is a time series too. The correlations over all the channels are taken into account in this latent space, and moreover, no data is missing. This then begs the question, if imputation occurs occurs before decoding, what is the advantage of using the GP?\n",
    "\n",
    "The authors from [Fortuin et. al. (2020)](https://arxiv.org/pdf/1907.04155.pdf) say,\n",
    "   > We propose a VAE architecture for multivariate time series imputation with a GP prior in the latent space to capture temporal dynamics. We propose a Cauchy kernel to allow the time series to display dynamics at multiple scales in a reduced dimensionality.\n",
    "\n",
    "In other words, the VAE consists of an encoder and decoder. The encoder takes the time series as the input and encodes it into a latent space. In this latent space, a GP Kernel (Cauchy) is applied.Finally, after this application, the decoder decodes the latent space back into the original time series.\n",
    "\n",
    "Thus, while the latent space of the encoder is a time series whose channels are correlated, it does not capture the temporal dynamics of the time series well. By using the Cauchy Kernel GP, the temporal dynamics are more useful. However, because this is still inside the latent space, it still has to be decoded back to make sense in the real space.\n",
    "\n",
    "\n",
    "Thus, there are still some ways in which GP-VAE can be used for RESSPECT:\n",
    "Instead of the latent space, the GP Model (kernel and mean function) can be used:\n",
    "1. Can the model itself be used as the first step of a recommender and provide possible values for the next measurements?\n",
    "2. Instead of fitting the Bazin function individually to the 6 passband light curves, can we instead fit it to the GP's time series where the channels are correlated? (Note: We will have to explore a bit to find out what dimensionality of the latent space use)"
   ]
  }
 ],
 "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": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}