GP_AL.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Gaussian processes and active learning\n",
    "\n",
    "This notebook explains and constructs a Gaussian Process metamodel for the simulation model using Active learning. It computes the prediction in 400 unlabeled input points."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Import necessary libraries\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import math\n",
    "import matplotlib.pyplot as plt\n",
    "import pickle\n",
    "import gpytorch\n",
    "import torch\n",
    "from Model import model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "###### Load initial data of 200 LHS samples\n",
    "with open('Data/output200.pickle', 'rb') as handle:\n",
    "    y = pickle.load(handle)\n",
    "with open('Data/X200.pickle', 'rb') as handle: #labeled points\n",
    "    X = pickle.load(handle)\n",
    "\n",
    "\n",
    "X_labeled = X\n",
    "y_labeled = y\n",
    "\n",
    "#Transform datatype to tensors\n",
    "y_labeled= y_labeled.astype(float)\n",
    "y_labeled = torch.from_numpy(y_labeled)\n",
    "X_labeled = torch.from_numpy(X_labeled)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We also need to load 800 LHS sample points in the input space that will define our unlabeled dataset.\n",
    "This dataset needs to be bigger than the initial labeled dataset."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open('Data/X800alone.pickle', 'rb') as handle: #unlabeled points\n",
    "    X_LHS = pickle.load(handle)\n",
    "\n",
    "X_LHS = torch.from_numpy(X_LHS)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define a Gaussian process regression model with GPYtorch\n",
    "class GPRegressionModel(gpytorch.models.ExactGP):\n",
    "    def __init__(self, train_x, train_y, likelihood):\n",
    "        super().__init__(train_x, train_y, likelihood)\n",
    "        self.mean_module = gpytorch.means.ConstantMean()\n",
    "        self.covar_module = gpytorch.kernels.ScaleKernel(gpytorch.kernels.RBFKernel())\n",
    "\n",
    "    def forward(self, x):\n",
    "        mean_x = self.mean_module(x)\n",
    "        covar_x = self.covar_module(x)\n",
    "        return gpytorch.distributions.MultivariateNormal(mean_x, covar_x)\n",
    "\n",
    "likelihood = gpytorch.likelihoods.GaussianLikelihood()\n",
    "model1 = GPRegressionModel(X_labeled, y_labeled, likelihood)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor(1194550.4603, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(1006478.8928, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(856905.9782, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(737299.6267, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(640901.7035, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(562474.8285, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(498020.1552, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(444510.2947, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(399658.5527, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(361732.6952, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(329411.3257, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(301676.0331, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(277732.4851, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(256953.0975, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(238835.5104, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(222972.6938, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(209031.0316, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(196734.5001, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(185852.7148, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(176192.0290, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(167588.7799, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(159903.8824, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(153018.8016, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(146832.0280, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(141256.5033, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(136217.2662, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(131649.6778, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(127497.8852, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(123713.4414, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(120254.3304, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(117083.9701, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(114170.4683, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(111485.9846, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(109006.0679, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(106709.3588, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(104576.9843, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(102592.3690, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(100740.8789, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(99009.5535, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(97386.8875, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(95862.6858, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(94427.8795, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(93074.3561, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(91794.8995, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(90583.0514, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(89432.9949, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(88339.5536, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(87298.0398, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(86304.2544, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(85354.3947, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(84445.0454, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(83573.0958, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(82735.7546, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(81930.4759, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(81154.9571, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(80407.0936, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(79684.9860, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(78986.9040, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(78311.2494, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(77656.5739, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(77021.5785, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(76405.0336, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(75805.8335, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(75222.9703, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(74655.5127, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(74102.5924, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(73563.4242, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(73037.2803, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(72523.4982, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(72021.4443, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(71530.5647, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(71050.3066, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(70580.1929, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(70119.7717, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(69668.6174, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(69226.3212, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(68792.5320, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(68366.9048, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(67949.1196, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(67538.8761, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(67135.8986, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(66739.9225, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(66350.7025, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(65968.0051, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(65591.6095, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(65221.3205, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(64856.9372, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(64498.2850, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(64145.1729, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(63797.4601, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(63454.9787, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(63117.5847, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(62785.1408, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(62457.5053, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(62134.5495, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(61816.1617, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(61502.2249, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(61192.6205, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(60887.2491, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(60586.0062, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(60288.7930, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(59995.5189, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(59706.0915, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(59420.4255, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(59138.4390, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(58860.0538, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(58585.1806, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(58313.7575, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(58045.7080, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(57780.9599, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(57519.4522, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(57261.1117, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(57005.8793, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(56753.6882, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(56504.4856, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(56258.2114, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(56014.8090, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(55774.2301, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(55536.4149, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(55301.3114, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(55068.8744, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(54839.0542, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(54611.8016, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(54387.0699, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(54164.8165, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(53944.9968, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(53727.5691, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(53512.4884, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(53299.7220, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(53089.2186, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(52880.9541, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(52674.8796, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(52470.9581, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(52269.1649, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(52069.4513, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(51871.7929, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(51676.1540, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(51482.5008, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(51290.7988, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(51101.0207, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(50913.1313, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(50727.1086, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(50542.9217, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(50360.5298, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(50179.9217, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(50001.0599, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(49823.9175, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(49648.4768, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(49474.7023, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(49302.5769, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(49132.0660, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(48963.1541, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(48795.8154, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(48630.0255, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(48465.7653, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(48303.0073, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(48141.7331, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(47981.9210, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(47823.5470, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(47666.5907, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(47511.0328, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(47356.8582, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(47204.0450, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(47052.5691, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(46902.4169, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(46753.5697, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(46606.0081, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(46459.7155, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(46314.6739, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(46170.8648, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(46028.2753, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(45886.8835, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(45746.6819, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(45607.6451, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(45469.7624, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(45333.0209, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(45197.4049, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(45062.8911, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(44929.4741, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(44797.1386, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(44665.8701, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(44535.6482, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(44406.4729, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(44278.3203, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(44151.1797, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(44025.0395, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(43899.8873, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(43775.7104, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(43652.4942, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(43530.2324, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(43408.9084, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(43288.5120, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(43169.0366, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(43050.4612, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(42932.7823, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(42815.9876, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(42700.0686, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(42585.0017, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(42470.7962, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(42357.4265, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(42244.8925, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(42133.1816, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(42022.2787, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(41912.1847, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(41802.8794, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(41694.3638, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(41586.6205, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(41479.6450, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(41373.4328, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(41267.9624, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(41163.2362, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(41059.2412, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(40955.9713, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(40853.4182, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(40751.5753, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(40650.4293, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(40549.9765, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(40450.2083, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(40351.1137, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(40252.6961, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(40154.9380, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(40057.8339, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(39961.3795, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(39865.5706, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(39770.3935, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(39675.8416, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(39581.9135, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(39488.5995, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(39395.8949, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(39303.7911, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(39212.2852, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(39121.3630, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(39031.0263, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(38941.2657, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(38852.0772, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(38763.4510, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(38675.3850, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(38587.8795, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(38500.9177, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(38414.4984, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(38328.6151, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(38243.2654, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(38158.4392, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(38074.1392, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(37990.3514, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(37907.0741, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(37824.3066, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(37742.0328, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(37660.2588, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(37578.9751, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(37498.1764, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(37417.8621, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(37338.0211, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(37258.6568, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(37179.7598, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(37101.3211, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(37023.3469, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(36945.8291, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(36868.7534, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(36792.1285, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(36715.9457, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(36640.2005, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(36564.8880, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(36490.0062, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(36415.5530, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(36341.5131, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(36267.8984, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(36194.6937, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(36121.9007, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(36049.5116, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(35977.5267, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(35905.9399, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(35834.7510, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(35763.9510, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(35693.5445, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(35623.5211, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(35553.8804, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(35484.6149, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(35415.7233, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(35347.2058, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(35279.0579, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(35211.2723, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(35143.8511, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(35076.7860, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(35010.0741, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(34943.7195, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(34877.7149, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(34812.0536, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(34746.7342, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(34681.7585, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(34617.1200, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(34552.8165, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(34488.8459, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(34425.2051, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(34361.8890, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(34298.8946, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(34236.2242, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(34173.8706, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(34111.8353, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(34050.1100, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(33988.6954, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(33927.5869, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(33866.7870, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(33806.2910, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(33746.0939, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(33686.1937, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(33626.5912, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(33567.2817, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(33508.2622, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(33449.5291, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(33391.0850, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(33332.9219, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(33275.0419, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(33217.4428, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(33160.1197, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(33103.0703, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(33046.2952, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(32989.7938, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(32933.5549, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(32877.5855, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(32821.8810, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(32766.4416, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(32711.2579, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(32656.3334, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(32601.6665, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(32547.2538, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(32493.0948, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(32439.1856, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(32385.5248, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(32332.1105, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(32278.9415, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(32226.0156, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(32173.3285, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(32120.8848, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(32068.6764, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(32016.7035, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(31964.9677, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(31913.4641, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(31862.1894, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(31811.1440, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(31760.3251, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(31709.7325, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(31659.3631, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(31609.2170, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(31559.2889, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(31509.5833, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(31460.0965, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(31410.8220, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(31361.7653, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(31312.9183, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(31264.2851, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(31215.8627, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(31167.6462, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(31119.6391, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(31071.8373, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(31024.2368, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(30976.8420, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(30929.6466, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(30882.6490, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(30835.8514, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(30789.2512, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(30742.8458, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(30696.6363, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(30650.6181, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(30604.7914, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(30559.1565, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(30513.7105, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(30468.4523, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(30423.3807, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(30378.4942, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(30333.7923, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(30289.2717, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(30244.9324, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(30200.7717, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(30156.7912, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(30112.9878, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(30069.3619, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(30025.9131, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29982.6374, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29939.5351, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29896.6040, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29853.8457, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29811.2550, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29768.8329, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29726.5812, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29684.4934, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29642.5719, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29600.8126, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29559.2173, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29517.7857, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29476.5142, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29435.4031, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29394.4504, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29353.6597, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29313.0206, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29272.5419, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29232.2150, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29192.0424, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29152.0253, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29112.1583, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29072.4415, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(29032.8755, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28993.4596, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28954.1927, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28915.0718, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28876.0978, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28837.2700, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28798.5867, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28760.0480, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28721.6507, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28683.3952, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28645.2820, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28607.3087, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28569.4743, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28531.7787, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28494.2227, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28456.8013, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28419.5180, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28382.3662, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28345.3514, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28308.4705, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28271.7230, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28235.1085, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28198.6258, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28162.2701, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28126.0448, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28089.9478, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28053.9816, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(28018.1405, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27982.4278, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27946.8401, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27911.3790, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27876.0421, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27840.8266, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27805.7358, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27770.7674, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27735.9203, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27701.1947, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27666.5888, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27632.1015, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27597.7328, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27563.4832, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27529.3493, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27495.3351, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27461.4358, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27427.6527, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27393.9834, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27360.4293, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27326.9878, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27293.6601, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27260.4446, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27227.3395, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27194.3476, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27161.4650, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27128.6923, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27096.0287, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27063.4733, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(27031.0255, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26998.6869, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26966.4535, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26934.3246, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26902.3032, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26870.3856, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26838.5729, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26806.8648, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26775.2601, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26743.7569, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26712.3587, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26681.0615, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26649.8640, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26618.7657, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26587.7713, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26556.8750, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26526.0769, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26495.3769, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26464.7763, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26434.2723, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26403.8631, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26373.5517, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26343.3376, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26313.2169, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26283.1928, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26253.2610, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26223.4252, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26193.6814, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26164.0299, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26134.4734, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26105.0053, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26075.6307, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26046.3478, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(26017.1539, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(25988.0509, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(25959.0375, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(25930.1111, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(25901.2753, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(25872.5272, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(25843.8672, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(25815.2944, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(25786.8071, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "tensor(25758.4079, dtype=torch.float64, grad_fn=<NegBackward0>)\n"
     ]
    }
   ],
   "source": [
    "# Train the model1 on the labeled dataset\n",
    "model1.train()\n",
    "likelihood.train()\n",
    "optimizer = torch.optim.Adam(model1.parameters(), lr=0.1)\n",
    "mll = gpytorch.mlls.ExactMarginalLogLikelihood(likelihood, model1)\n",
    "\n",
    "for i in range(500):\n",
    "    optimizer.zero_grad()\n",
    "    output = model1(X_labeled)\n",
    "    loss = -mll(output, y_labeled)\n",
    "    print(loss)\n",
    "    loss.backward()\n",
    "    optimizer.step()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Change the name of the labeled dataset to include more points in the active learning loop\n",
    "X_new_labeled = X_labeled\n",
    "y_new_labeled = y_labeled"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we define the parameters of the Active Learning loop."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "#### Active learning loop\n",
    "n_iter= 100 # number of iterations in the active learning loop\n",
    "k= 1 # number of elements to query in each loop"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In this example, our query criterion will be the entropy of the unlabeled points. The entropy is directly related with the estimated variance and it is measures the informativenss of the data points."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Function to calculate the entropy given the standard deviation\n",
    "def entropy_given_normal_std(std_arr):\n",
    "        \"\"\"Return entropy given an array of 1D normal standard deviations.\"\"\"\n",
    "        entropy = np.log(std_arr) + np.log(np.sqrt(2 * np.pi)) + 0.5\n",
    "        return entropy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "X_new_labeled size:  torch.Size([201, 3])\n",
      "X_LHS size:  torch.Size([799, 3])\n",
      "y_new_labeled size:  torch.Size([201])\n",
      "tensor(25690.5342, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([202, 3])\n",
      "X_LHS size:  torch.Size([798, 3])\n",
      "y_new_labeled size:  torch.Size([202])\n",
      "tensor(25053.0508, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([203, 3])\n",
      "X_LHS size:  torch.Size([797, 3])\n",
      "y_new_labeled size:  torch.Size([203])\n",
      "tensor(24442.8386, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([204, 3])\n",
      "X_LHS size:  torch.Size([796, 3])\n",
      "y_new_labeled size:  torch.Size([204])\n",
      "tensor(23881.5138, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([205, 3])\n",
      "X_LHS size:  torch.Size([795, 3])\n",
      "y_new_labeled size:  torch.Size([205])\n",
      "tensor(23346.2650, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([206, 3])\n",
      "X_LHS size:  torch.Size([794, 3])\n",
      "y_new_labeled size:  torch.Size([206])\n",
      "tensor(22819.4441, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([207, 3])\n",
      "X_LHS size:  torch.Size([793, 3])\n",
      "y_new_labeled size:  torch.Size([207])\n",
      "tensor(22308.7281, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([208, 3])\n",
      "X_LHS size:  torch.Size([792, 3])\n",
      "y_new_labeled size:  torch.Size([208])\n",
      "tensor(21827.9189, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([209, 3])\n",
      "X_LHS size:  torch.Size([791, 3])\n",
      "y_new_labeled size:  torch.Size([209])\n",
      "tensor(21399.2559, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([210, 3])\n",
      "X_LHS size:  torch.Size([790, 3])\n",
      "y_new_labeled size:  torch.Size([210])\n",
      "tensor(20946.6540, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([211, 3])\n",
      "X_LHS size:  torch.Size([789, 3])\n",
      "y_new_labeled size:  torch.Size([211])\n",
      "tensor(20521.4150, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([212, 3])\n",
      "X_LHS size:  torch.Size([788, 3])\n",
      "y_new_labeled size:  torch.Size([212])\n",
      "tensor(20111.1399, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([213, 3])\n",
      "X_LHS size:  torch.Size([787, 3])\n",
      "y_new_labeled size:  torch.Size([213])\n",
      "tensor(19716.6831, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([214, 3])\n",
      "X_LHS size:  torch.Size([786, 3])\n",
      "y_new_labeled size:  torch.Size([214])\n",
      "tensor(19351.3564, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([215, 3])\n",
      "X_LHS size:  torch.Size([785, 3])\n",
      "y_new_labeled size:  torch.Size([215])\n",
      "tensor(18979.0993, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([216, 3])\n",
      "X_LHS size:  torch.Size([784, 3])\n",
      "y_new_labeled size:  torch.Size([216])\n",
      "tensor(18641.4813, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([217, 3])\n",
      "X_LHS size:  torch.Size([783, 3])\n",
      "y_new_labeled size:  torch.Size([217])\n",
      "tensor(18313.9590, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([218, 3])\n",
      "X_LHS size:  torch.Size([782, 3])\n",
      "y_new_labeled size:  torch.Size([218])\n",
      "tensor(17977.7118, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([219, 3])\n",
      "X_LHS size:  torch.Size([781, 3])\n",
      "y_new_labeled size:  torch.Size([219])\n",
      "tensor(17662.1198, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([220, 3])\n",
      "X_LHS size:  torch.Size([780, 3])\n",
      "y_new_labeled size:  torch.Size([220])\n",
      "tensor(17355.2247, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([221, 3])\n",
      "X_LHS size:  torch.Size([779, 3])\n",
      "y_new_labeled size:  torch.Size([221])\n",
      "tensor(17059.9744, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([222, 3])\n",
      "X_LHS size:  torch.Size([778, 3])\n",
      "y_new_labeled size:  torch.Size([222])\n",
      "tensor(16765.6573, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([223, 3])\n",
      "X_LHS size:  torch.Size([777, 3])\n",
      "y_new_labeled size:  torch.Size([223])\n",
      "tensor(16488.7376, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([224, 3])\n",
      "X_LHS size:  torch.Size([776, 3])\n",
      "y_new_labeled size:  torch.Size([224])\n",
      "tensor(16215.0218, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([225, 3])\n",
      "X_LHS size:  torch.Size([775, 3])\n",
      "y_new_labeled size:  torch.Size([225])\n",
      "tensor(15946.9188, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([226, 3])\n",
      "X_LHS size:  torch.Size([774, 3])\n",
      "y_new_labeled size:  torch.Size([226])\n",
      "tensor(15686.3949, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([227, 3])\n",
      "X_LHS size:  torch.Size([773, 3])\n",
      "y_new_labeled size:  torch.Size([227])\n",
      "tensor(15437.2832, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([228, 3])\n",
      "X_LHS size:  torch.Size([772, 3])\n",
      "y_new_labeled size:  torch.Size([228])\n",
      "tensor(15191.5738, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([229, 3])\n",
      "X_LHS size:  torch.Size([771, 3])\n",
      "y_new_labeled size:  torch.Size([229])\n",
      "tensor(14956.0412, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([230, 3])\n",
      "X_LHS size:  torch.Size([770, 3])\n",
      "y_new_labeled size:  torch.Size([230])\n",
      "tensor(14725.3072, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([231, 3])\n",
      "X_LHS size:  torch.Size([769, 3])\n",
      "y_new_labeled size:  torch.Size([231])\n",
      "tensor(14505.6838, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([232, 3])\n",
      "X_LHS size:  torch.Size([768, 3])\n",
      "y_new_labeled size:  torch.Size([232])\n",
      "tensor(14285.7169, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([233, 3])\n",
      "X_LHS size:  torch.Size([767, 3])\n",
      "y_new_labeled size:  torch.Size([233])\n",
      "tensor(14070.9832, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([234, 3])\n",
      "X_LHS size:  torch.Size([766, 3])\n",
      "y_new_labeled size:  torch.Size([234])\n",
      "tensor(13861.5337, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([235, 3])\n",
      "X_LHS size:  torch.Size([765, 3])\n",
      "y_new_labeled size:  torch.Size([235])\n",
      "tensor(13670.5645, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([236, 3])\n",
      "X_LHS size:  torch.Size([764, 3])\n",
      "y_new_labeled size:  torch.Size([236])\n",
      "tensor(13471.6603, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([237, 3])\n",
      "X_LHS size:  torch.Size([763, 3])\n",
      "y_new_labeled size:  torch.Size([237])\n",
      "tensor(13277.5602, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([238, 3])\n",
      "X_LHS size:  torch.Size([762, 3])\n",
      "y_new_labeled size:  torch.Size([238])\n",
      "tensor(13092.3696, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([239, 3])\n",
      "X_LHS size:  torch.Size([761, 3])\n",
      "y_new_labeled size:  torch.Size([239])\n",
      "tensor(12907.7660, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([240, 3])\n",
      "X_LHS size:  torch.Size([760, 3])\n",
      "y_new_labeled size:  torch.Size([240])\n",
      "tensor(12728.1379, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([241, 3])\n",
      "X_LHS size:  torch.Size([759, 3])\n",
      "y_new_labeled size:  torch.Size([241])\n",
      "tensor(12552.2367, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([242, 3])\n",
      "X_LHS size:  torch.Size([758, 3])\n",
      "y_new_labeled size:  torch.Size([242])\n",
      "tensor(12380.2534, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([243, 3])\n",
      "X_LHS size:  torch.Size([757, 3])\n",
      "y_new_labeled size:  torch.Size([243])\n",
      "tensor(12212.6031, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([244, 3])\n",
      "X_LHS size:  torch.Size([756, 3])\n",
      "y_new_labeled size:  torch.Size([244])\n",
      "tensor(12048.6070, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([245, 3])\n",
      "X_LHS size:  torch.Size([755, 3])\n",
      "y_new_labeled size:  torch.Size([245])\n",
      "tensor(11888.5476, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([246, 3])\n",
      "X_LHS size:  torch.Size([754, 3])\n",
      "y_new_labeled size:  torch.Size([246])\n",
      "tensor(11732.3013, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([247, 3])\n",
      "X_LHS size:  torch.Size([753, 3])\n",
      "y_new_labeled size:  torch.Size([247])\n",
      "tensor(11581.4513, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([248, 3])\n",
      "X_LHS size:  torch.Size([752, 3])\n",
      "y_new_labeled size:  torch.Size([248])\n",
      "tensor(11431.9147, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([249, 3])\n",
      "X_LHS size:  torch.Size([751, 3])\n",
      "y_new_labeled size:  torch.Size([249])\n",
      "tensor(11284.9692, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([250, 3])\n",
      "X_LHS size:  torch.Size([750, 3])\n",
      "y_new_labeled size:  torch.Size([250])\n",
      "tensor(11141.1165, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([251, 3])\n",
      "X_LHS size:  torch.Size([749, 3])\n",
      "y_new_labeled size:  torch.Size([251])\n",
      "tensor(10999.9202, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([252, 3])\n",
      "X_LHS size:  torch.Size([748, 3])\n",
      "y_new_labeled size:  torch.Size([252])\n",
      "tensor(10861.6486, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([253, 3])\n",
      "X_LHS size:  torch.Size([747, 3])\n",
      "y_new_labeled size:  torch.Size([253])\n",
      "tensor(10728.9090, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([254, 3])\n",
      "X_LHS size:  torch.Size([746, 3])\n",
      "y_new_labeled size:  torch.Size([254])\n",
      "tensor(10604.1082, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([255, 3])\n",
      "X_LHS size:  torch.Size([745, 3])\n",
      "y_new_labeled size:  torch.Size([255])\n",
      "tensor(10473.9216, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([256, 3])\n",
      "X_LHS size:  torch.Size([744, 3])\n",
      "y_new_labeled size:  torch.Size([256])\n",
      "tensor(10346.8875, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([257, 3])\n",
      "X_LHS size:  torch.Size([743, 3])\n",
      "y_new_labeled size:  torch.Size([257])\n",
      "tensor(10226.3838, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([258, 3])\n",
      "X_LHS size:  torch.Size([742, 3])\n",
      "y_new_labeled size:  torch.Size([258])\n",
      "tensor(10105.3293, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([259, 3])\n",
      "X_LHS size:  torch.Size([741, 3])\n",
      "y_new_labeled size:  torch.Size([259])\n",
      "tensor(9985.8800, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([260, 3])\n",
      "X_LHS size:  torch.Size([740, 3])\n",
      "y_new_labeled size:  torch.Size([260])\n",
      "tensor(9869.1890, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([261, 3])\n",
      "X_LHS size:  torch.Size([739, 3])\n",
      "y_new_labeled size:  torch.Size([261])\n",
      "tensor(9761.3857, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([262, 3])\n",
      "X_LHS size:  torch.Size([738, 3])\n",
      "y_new_labeled size:  torch.Size([262])\n",
      "tensor(9648.5168, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([263, 3])\n",
      "X_LHS size:  torch.Size([737, 3])\n",
      "y_new_labeled size:  torch.Size([263])\n",
      "tensor(9541.0902, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([264, 3])\n",
      "X_LHS size:  torch.Size([736, 3])\n",
      "y_new_labeled size:  torch.Size([264])\n",
      "tensor(9433.9343, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([265, 3])\n",
      "X_LHS size:  torch.Size([735, 3])\n",
      "y_new_labeled size:  torch.Size([265])\n",
      "tensor(9326.3963, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([266, 3])\n",
      "X_LHS size:  torch.Size([734, 3])\n",
      "y_new_labeled size:  torch.Size([266])\n",
      "tensor(9223.4314, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([267, 3])\n",
      "X_LHS size:  torch.Size([733, 3])\n",
      "y_new_labeled size:  torch.Size([267])\n",
      "tensor(9121.2319, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([268, 3])\n",
      "X_LHS size:  torch.Size([732, 3])\n",
      "y_new_labeled size:  torch.Size([268])\n",
      "tensor(9020.0931, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([269, 3])\n",
      "X_LHS size:  torch.Size([731, 3])\n",
      "y_new_labeled size:  torch.Size([269])\n",
      "tensor(8919.5420, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([270, 3])\n",
      "X_LHS size:  torch.Size([730, 3])\n",
      "y_new_labeled size:  torch.Size([270])\n",
      "tensor(8821.1517, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([271, 3])\n",
      "X_LHS size:  torch.Size([729, 3])\n",
      "y_new_labeled size:  torch.Size([271])\n",
      "tensor(8724.8070, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([272, 3])\n",
      "X_LHS size:  torch.Size([728, 3])\n",
      "y_new_labeled size:  torch.Size([272])\n",
      "tensor(8628.9500, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([273, 3])\n",
      "X_LHS size:  torch.Size([727, 3])\n",
      "y_new_labeled size:  torch.Size([273])\n",
      "tensor(8535.3430, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([274, 3])\n",
      "X_LHS size:  torch.Size([726, 3])\n",
      "y_new_labeled size:  torch.Size([274])\n",
      "tensor(8443.2792, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([275, 3])\n",
      "X_LHS size:  torch.Size([725, 3])\n",
      "y_new_labeled size:  torch.Size([275])\n",
      "tensor(8352.2254, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([276, 3])\n",
      "X_LHS size:  torch.Size([724, 3])\n",
      "y_new_labeled size:  torch.Size([276])\n",
      "tensor(8267.9156, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([277, 3])\n",
      "X_LHS size:  torch.Size([723, 3])\n",
      "y_new_labeled size:  torch.Size([277])\n",
      "tensor(8179.9317, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([278, 3])\n",
      "X_LHS size:  torch.Size([722, 3])\n",
      "y_new_labeled size:  torch.Size([278])\n",
      "tensor(8093.2114, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([279, 3])\n",
      "X_LHS size:  torch.Size([721, 3])\n",
      "y_new_labeled size:  torch.Size([279])\n",
      "tensor(8009.0618, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([280, 3])\n",
      "X_LHS size:  torch.Size([720, 3])\n",
      "y_new_labeled size:  torch.Size([280])\n",
      "tensor(7926.5072, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([281, 3])\n",
      "X_LHS size:  torch.Size([719, 3])\n",
      "y_new_labeled size:  torch.Size([281])\n",
      "tensor(7843.8375, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([282, 3])\n",
      "X_LHS size:  torch.Size([718, 3])\n",
      "y_new_labeled size:  torch.Size([282])\n",
      "tensor(7766.0379, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([283, 3])\n",
      "X_LHS size:  torch.Size([717, 3])\n",
      "y_new_labeled size:  torch.Size([283])\n",
      "tensor(7685.9687, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([284, 3])\n",
      "X_LHS size:  torch.Size([716, 3])\n",
      "y_new_labeled size:  torch.Size([284])\n",
      "tensor(7609.0208, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([285, 3])\n",
      "X_LHS size:  torch.Size([715, 3])\n",
      "y_new_labeled size:  torch.Size([285])\n",
      "tensor(7532.4805, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([286, 3])\n",
      "X_LHS size:  torch.Size([714, 3])\n",
      "y_new_labeled size:  torch.Size([286])\n",
      "tensor(7456.8989, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([287, 3])\n",
      "X_LHS size:  torch.Size([713, 3])\n",
      "y_new_labeled size:  torch.Size([287])\n",
      "tensor(7382.7069, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([288, 3])\n",
      "X_LHS size:  torch.Size([712, 3])\n",
      "y_new_labeled size:  torch.Size([288])\n",
      "tensor(7309.8591, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([289, 3])\n",
      "X_LHS size:  torch.Size([711, 3])\n",
      "y_new_labeled size:  torch.Size([289])\n",
      "tensor(7239.5340, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([290, 3])\n",
      "X_LHS size:  torch.Size([710, 3])\n",
      "y_new_labeled size:  torch.Size([290])\n",
      "tensor(7172.7210, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([291, 3])\n",
      "X_LHS size:  torch.Size([709, 3])\n",
      "y_new_labeled size:  torch.Size([291])\n",
      "tensor(7105.3785, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([292, 3])\n",
      "X_LHS size:  torch.Size([708, 3])\n",
      "y_new_labeled size:  torch.Size([292])\n",
      "tensor(7035.6648, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([293, 3])\n",
      "X_LHS size:  torch.Size([707, 3])\n",
      "y_new_labeled size:  torch.Size([293])\n",
      "tensor(6968.2246, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([294, 3])\n",
      "X_LHS size:  torch.Size([706, 3])\n",
      "y_new_labeled size:  torch.Size([294])\n",
      "tensor(6901.5452, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([295, 3])\n",
      "X_LHS size:  torch.Size([705, 3])\n",
      "y_new_labeled size:  torch.Size([295])\n",
      "tensor(6835.2174, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([296, 3])\n",
      "X_LHS size:  torch.Size([704, 3])\n",
      "y_new_labeled size:  torch.Size([296])\n",
      "tensor(6769.5122, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([297, 3])\n",
      "X_LHS size:  torch.Size([703, 3])\n",
      "y_new_labeled size:  torch.Size([297])\n",
      "tensor(6704.9229, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([298, 3])\n",
      "X_LHS size:  torch.Size([702, 3])\n",
      "y_new_labeled size:  torch.Size([298])\n",
      "tensor(6643.8540, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([299, 3])\n",
      "X_LHS size:  torch.Size([701, 3])\n",
      "y_new_labeled size:  torch.Size([299])\n",
      "tensor(6581.0426, dtype=torch.float64, grad_fn=<NegBackward0>)\n",
      "X_new_labeled size:  torch.Size([300, 3])\n",
      "X_LHS size:  torch.Size([700, 3])\n",
      "y_new_labeled size:  torch.Size([300])\n",
      "tensor(6519.1791, dtype=torch.float64, grad_fn=<NegBackward0>)\n"
     ]
    }
   ],
   "source": [
    "### Active learning loop\n",
    "for i in range(n_iter):\n",
    "\n",
    "    # Use the model1 to make predictions on the unlabeled dataset\n",
    "    model1.eval()\n",
    "    likelihood.eval()\n",
    "    with torch.no_grad():\n",
    "        pred_dist = likelihood(model1(X_LHS))\n",
    "\n",
    "    # Select the most informative data points using uncertainty sampling\n",
    "    uncertainty = entropy_given_normal_std(np.sqrt(pred_dist.variance.numpy()))\n",
    "    uncertainty = torch.from_numpy(uncertainty)\n",
    "    top_indices = uncertainty.topk(k).indices\n",
    "\n",
    "    # Compute the value of the new point in the true model\n",
    "    new_X_point = X_LHS[top_indices,:].numpy()[0]\n",
    "    new_Y_point = model(new_X_point)\n",
    "\n",
    "    # Remove the new point from the unlabeled dataset\n",
    "    X_LHS = X_LHS.numpy()\n",
    "    X_mask= (X_LHS == new_X_point)\n",
    "    X_LHS = X_LHS[~X_mask].reshape((-1, 3))\n",
    "    X_LHS = torch.from_numpy(X_LHS)\n",
    "\n",
    "   #Update the train datasets with the new point\n",
    "    new_X_point = np.array([new_X_point])\n",
    "    new_X_point = torch.from_numpy(new_X_point)\n",
    "    X_new_labeled = torch.cat([X_new_labeled, new_X_point])\n",
    "    print('X_new_labeled size: ', X_new_labeled.size())\n",
    "    print('X_LHS size: ', X_LHS.size())\n",
    "\n",
    "    y_new_labeled = y_new_labeled.numpy()\n",
    "    y_new_labeled = np.append(y_new_labeled, new_Y_point)\n",
    "    y_new_labeled = torch.from_numpy(y_new_labeled)\n",
    "    print('y_new_labeled size: ', y_new_labeled.size())\n",
    "\n",
    "    # Re train the model1 on the updated labeled dataset\n",
    "    model1.set_train_data(X_new_labeled, y_new_labeled, strict=False)\n",
    "    model1.train()\n",
    "    likelihood.train()\n",
    "    optimizer = torch.optim.Adam(model1.parameters(), lr=0.1)\n",
    "    mll = gpytorch.mlls.ExactMarginalLogLikelihood(likelihood, model1)\n",
    "\n",
    "    for i in range(1):\n",
    "        optimizer.zero_grad()\n",
    "        output = model1(X_new_labeled)\n",
    "        loss = -mll(output, y_new_labeled)\n",
    "        print(loss)\n",
    "        loss.backward()\n",
    "        optimizer.step()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open('Data/X400.2.pickle', 'rb') as handle:\n",
    "   X_unlabeled = pickle.load(handle)\n",
    "\n",
    "X_unlabeled = torch.from_numpy(X_unlabeled)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Final loss: tensor(6519.1791, dtype=torch.float64, grad_fn=<NegBackward0>)\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "MultivariateNormal(loc: torch.Size([400]))"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Get the final value of the posterior for the another LHS samples\n",
    "model1.eval()\n",
    "likelihood.eval()\n",
    "print('Final loss:',loss)\n",
    "\n",
    "with torch.no_grad():\n",
    "    pred_dist = likelihood(model1(X_unlabeled))\n",
    "\n",
    "pred_dist"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 900x700 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "with torch.no_grad():\n",
    "    # Initialize plot\n",
    "    f, ax = plt.subplots(1, 1, figsize=(9, 7))\n",
    "    # Plot training data as black stars\n",
    "    ax.plot(X_new_labeled.numpy()[:200,2], y_new_labeled.numpy()[:200], 'k*')\n",
    "    ax.plot(X_new_labeled.numpy()[:100,2], y_new_labeled.numpy()[:100], 'g*')\n",
    "    # Plot predictive means as blue line\n",
    "    ax.plot(X_unlabeled.numpy()[:,2], pred_dist.mean.numpy(), 'b*')\n",
    "    ax.legend(['Observed Data', 'Observed query data', 'Mean'])\n",
    "    ax.set_title(\"Predictive mean and observed values for one dimension of the input space\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "f.savefig('figure.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "GP1",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.17"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}