importance sampling diagnostic.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "04e6c51d-ec2c-41f7-a453-a768c1ac7674",
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "064f0039-8dc6-4866-9dad-f0530ab3fae0",
   "metadata": {},
   "outputs": [],
   "source": [
    "from functools import partial\n",
    "from copy import deepcopy\n",
    "from pathlib import Path\n",
    "from typing import Optional, Callable\n",
    "\n",
    "from tqdm.notebook import trange, tqdm\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "from matplotlib.colors import BoundaryNorm\n",
    "from matplotlib.ticker import MaxNLocator\n",
    "import seaborn as sns\n",
    "\n",
    "import numpy as np\n",
    "\n",
    "import pandas as pd \n",
    "\n",
    "import scipy\n",
    "import scipy.stats\n",
    "import scipy.integrate\n",
    "\n",
    "import torch\n",
    "import torch.distributions\n",
    "import torch.nn.functional as F\n",
    "\n",
    "import math\n",
    "\n",
    "import pytorch_lightning as pl\n",
    "\n",
    "import sbi\n",
    "import sbi.analysis as analysis\n",
    "from sbi.inference import SNRE_A, SNRE_B, prepare_for_sbi\n",
    "from sbi.utils.get_nn_models import classifier_nn\n",
    "\n",
    "import cnre\n",
    "import cnre.data\n",
    "import cnre.data.benchmark\n",
    "import cnre.algorithms.cnre\n",
    "import cnre.posterior\n",
    "from corner import corner\n",
    "\n",
    "import sklearn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "830aab2e-e83d-422a-b228-43c73aa9092e",
   "metadata": {},
   "outputs": [],
   "source": [
    "factor = 2\n",
    "figsize = (1.83 * factor, 2.44 * factor * 0.75)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "9a28b8db-e2a0-44f0-afe7-0829e17c5a42",
   "metadata": {},
   "outputs": [],
   "source": [
    "torch.manual_seed(2)\n",
    "np.random.seed(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "8797f625-5ddb-4d5f-9115-79888d874caa",
   "metadata": {},
   "outputs": [],
   "source": [
    "color_truth = \"orange\"\n",
    "color_gilles = \"blue\"\n",
    "color_durkan = \"green\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "a517edea-09ab-4b43-bf99-b009aafb28cf",
   "metadata": {},
   "outputs": [],
   "source": [
    "num_total_samples = 40_000\n",
    "num_posterior_samples = 100_000\n",
    "\n",
    "train_kwargs = dict(\n",
    "    training_batch_size=512,\n",
    "    learning_rate=1e-4,\n",
    "    validation_fraction=0.25,\n",
    "    stop_after_epochs=2 ** 31 - 1,\n",
    "    max_num_epochs=60,\n",
    "    # max_num_epochs=10,\n",
    ")\n",
    "\n",
    "classifier_kwargs = dict(\n",
    "    # model='mlp', \n",
    "    model='resnet', \n",
    "    hidden_features=256, \n",
    "    # num_blocks=2,\n",
    "    # dropout_probability=0.0,\n",
    "    # use_batch_norm=True\n",
    ")\n",
    "sample_with = \"rejection\"\n",
    "\n",
    "num_atoms = 10\n",
    "K = num_atoms - 1\n",
    "\n",
    "device = \"cpu\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "9c414feb-6e2a-4c26-b87e-59a04129e62d",
   "metadata": {},
   "outputs": [],
   "source": [
    "dimension = 1\n",
    "\n",
    "if True:\n",
    "    low = -torch.ones(dimension).mul(3.).sqrt().to(device)\n",
    "    # low = -torch.zeros(dimension).to(device) + 0.001\n",
    "    high = torch.ones(dimension).mul(3.).sqrt().to(device)\n",
    "    limits = torch.stack([low.cpu(), high.cpu()], dim=-1).numpy()\n",
    "    \n",
    "    prior = sbi.utils.BoxUniform(low=low, high=high, device=device)\n",
    "else:\n",
    "    v = torch.ones(dimension).mul(4.).to(device)\n",
    "    limits = torch.stack([-v, v], dim=-1).numpy()\n",
    "    \n",
    "    # prior = torch.distributions.Normal(loc=torch.zeros(1, device=device), scale=torch.ones(1, device=device))\n",
    "    prior = torch.distributions.MultivariateNormal(loc=torch.zeros(dimension, device=device), covariance_matrix=torch.eye(dimension, device=device) )\n",
    "    \n",
    "task = cnre.Gaussian(scale=0.1)\n",
    "# task = cnre.Parabola(scale=0.9)\n",
    "# simulator, prior = prepare_for_sbi(expo_simulator, prior)\n",
    "simulator, prior = prepare_for_sbi(task.simulate, prior)\n",
    "\n",
    "# true_theta = torch.zeros(dimension).to(device).unsqueeze(0)\n",
    "true_theta = torch.ones(dimension).to(device).unsqueeze(0)\n",
    "observation = simulator(true_theta)\n",
    "\n",
    "theta = prior.sample((num_total_samples,)).to(device)\n",
    "x = simulator(theta).to(device)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a026ff14-d2ff-485b-923a-cfb052ac9403",
   "metadata": {},
   "source": [
    "## nre-c"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "c8349639-3135-4a13-8ff7-d5560d351592",
   "metadata": {},
   "outputs": [],
   "source": [
    "# get_classifier = classifier_nn(**classifier_kwargs)\n",
    "\n",
    "# inference = SNRE_A(\n",
    "#     prior, \n",
    "#     classifier=get_classifier,\n",
    "#     device=device,\n",
    "# )\n",
    "# _ = inference.append_simulations(theta, x).train(**train_kwargs)\n",
    "\n",
    "# inference._neural_net.load_state_dict(inference._best_model_state_dict)\n",
    "# density_estimator_nrea = deepcopy(inference._neural_net)\n",
    "\n",
    "# posterior = inference.build_posterior(\n",
    "#     density_estimator_nrea,\n",
    "#     sample_with=sample_with,\n",
    "# )\n",
    "\n",
    "# plt.plot(inference.summary[\"validation_log_probs\"])\n",
    "\n",
    "# samples_gilles = posterior.sample((num_posterior_samples,), x=observation.cpu())\n",
    "# name = \"gilles\"\n",
    "# fig, _ = analysis.pairplot(\n",
    "#     samples_gilles.cpu().numpy(), \n",
    "#     figsize=(6,6), \n",
    "#     points=true_theta.cpu().numpy(),\n",
    "#     title=name,\n",
    "#     limits=limits,\n",
    "# )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "1f9cae8b-0c70-4531-96a4-94d0d2e77737",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "                                                                                                                               \r"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7fef5234ea60>]"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "get_classifier = classifier_nn(**classifier_kwargs)\n",
    "\n",
    "dataset = torch.utils.data.TensorDataset(theta, x)\n",
    "train_loader, valid_loader = cnre.data.benchmark.get_dataloaders(\n",
    "    dataset, train_kwargs[\"training_batch_size\"], train_kwargs[\"validation_fraction\"]\n",
    ")\n",
    "\n",
    "for tt, xx in train_loader:\n",
    "    classifier = get_classifier(tt, xx)\n",
    "    break\n",
    "    \n",
    "optimizer = torch.optim.Adam(classifier.parameters(), lr=train_kwargs[\"learning_rate\"])\n",
    "\n",
    "results = cnre.algorithms.cnre.train(\n",
    "    classifier,\n",
    "    optimizer,\n",
    "    train_kwargs[\"max_num_epochs\"],\n",
    "    train_loader,\n",
    "    valid_loader,\n",
    "    None,\n",
    "    None,\n",
    "    K=K,\n",
    ")\n",
    "\n",
    "classifier.load_state_dict(results[\"best_network_state_dict\"])\n",
    "\n",
    "density_estimator_nrec = deepcopy(classifier)\n",
    "\n",
    "posterior_nrec = cnre.posterior.get_sbi_posterior(\n",
    "    ratio_estimator=classifier,\n",
    "    prior=prior,\n",
    "    sample_with=sample_with,\n",
    ")\n",
    "\n",
    "plt.plot(results[\"valid_losses\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "a0b7d5eb-b46c-4c08-938c-3b5ce0d803c6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1259d6c8b9674ca9b54172987779dfc5",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Drawing 100000 posterior samples:   0%|          | 0/100000 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/ben/sci/fix-cnre/remote/sbi/sbi/analysis/plot.py:176: UserWarning: Creating a tensor from a list of numpy.ndarrays is extremely slow. Please consider converting the list to a single numpy.ndarray with numpy.array() before converting to a tensor. (Triggered internally at /home/conda/feedstock_root/build_artifacts/pytorch-recipe_1673752754831/work/torch/csrc/utils/tensor_new.cpp:230.)\n",
      "  limits = torch.as_tensor(limits)\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 600x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "samples_nrec = posterior_nrec.sample((num_posterior_samples,), x=observation.cpu())\n",
    "name = \"nre-c\"\n",
    "fig, _ = analysis.pairplot(\n",
    "    samples_nrec.cpu().numpy(), \n",
    "    figsize=(6,6), \n",
    "    points=true_theta.cpu().numpy(),\n",
    "    title=name,\n",
    "    limits=limits,\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b5c9f84d-d805-44ef-ada6-7cedcf7c8e56",
   "metadata": {},
   "source": [
    "## nre-b"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "57c18f3c-84b1-4598-90c1-318c55ee4e18",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " Training neural network. Epochs trained: 61"
     ]
    }
   ],
   "source": [
    "get_classifier = classifier_nn(**classifier_kwargs)\n",
    "\n",
    "inference = SNRE_B(\n",
    "    prior, \n",
    "    classifier=get_classifier,\n",
    "    device=device,\n",
    ")\n",
    "_ = inference.append_simulations(theta, x).train(\n",
    "    num_atoms=num_atoms, \n",
    "    **train_kwargs\n",
    ")\n",
    "\n",
    "inference._neural_net.load_state_dict(inference._best_model_state_dict)\n",
    "density_estimator_nreb = deepcopy(inference._neural_net)\n",
    "\n",
    "posterior_nreb = inference.build_posterior(\n",
    "    density_estimator_nreb,\n",
    "    sample_with=sample_with,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "ea0eb885-b146-4449-a421-0dcd3cdaf9f6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7fef51f23b50>]"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(inference.summary[\"validation_log_probs\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "db4ee3de-bb23-4f25-8c37-610ec6bd4e74",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a9eccbf3610e483eb2c4c6d6d321129b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Drawing 100000 posterior samples:   0%|          | 0/100000 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 600x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "samples_durkan = posterior_nreb.sample((num_posterior_samples,), x=observation.cpu())\n",
    "name = \"durkan\"\n",
    "fig, _ = analysis.pairplot(\n",
    "    samples_durkan.cpu().numpy(), \n",
    "    figsize=(6,6), \n",
    "    points=true_theta.cpu().numpy(),\n",
    "    title=name,\n",
    "    limits=limits,\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c40d74b8-6aa0-4a3e-aeb3-971b0dee740d",
   "metadata": {
    "tags": []
   },
   "source": [
    "## posteriors"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "91a2d1ff-6047-4a11-82c6-f0d3eeffe17c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAWQAAAFkCAYAAAAXG0EgAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/OQEPoAAAACXBIWXMAAA9hAAAPYQGoP6dpAABMlElEQVR4nO3deXxU9b3/8deZfbJDIBsECKIRQUWhalSWisIFryLSVmurQMWfeFGUNGLB1l2jXqq4IJSKIPW22tuo1YoW2groFRckKNWIioFESAghy2Sd9fz+OJkhIdtMMpOZTD7Px2MeZs6c5TOpffvN93zP96uoqqoihBAi7HThLkAIIYRGAlkIISKEBLIQQkQICWQhhIgQEshCCBEhJJCFECJCSCALIUSEkEAWQogIYQh3AX3N4/Fw5MgR4uPjURQl3OUIIaKcqqrU1dWRkZGBTtd1G3jABfKRI0fIzMwMdxlCiAGmtLSU4cOHd7nPgAvk+Ph4QPvlJCQkhLkaIUS0s9lsZGZm+rKnKwMukL3dFAkJCRLIQog+408XqdzUE0KICCGBLIQQEUICWQghIsSA60P2l9vtxul0hruMqGc0GtHr9eEuQ4iIIIF8ElVVKS8vp6amJtylDBhJSUmkpaXJuHAx4Ekgn8QbxikpKcTExEhIhJCqqjQ2NlJRUQFAenp6mCsSIrwiJpDz8/NZuXIlt99+O6tXr+50vx07dpCbm8sXX3xBRkYGy5cvZ/HixUGpwe12+8I4OTk5KOcUXbNarQBUVFSQkpIi3RdiQIuIm3qffPIJ69ev56yzzupyv+LiYmbPns3kyZMpLCxk5cqVLF26lIKCgqDU4e0zjomJCcr5hH+8v2/psxcDXdgDub6+np/97Gf8/ve/Z9CgQV3uu27dOkaMGMHq1asZO3YsixYt4he/+AWrVq0Kak3STdG35PcthCbsgbxkyRIuv/xyLr300m733bVrFzNmzGizbebMmezevVtaV0KIfi+sfcgvv/wye/bs4ZNPPvFr//LyclJTU9tsS01NxeVyUVlZ2eFNIbvdjt1u97232Ww9K7ahBOyVPTu2J8xDIHZE313PTwcPHiQrK4vCwkImTJgQ7nJEf1f+L4gZBnoLGBPBlBTuisIqbIFcWlrK7bffztatW7FYLH4fd/Kft6qqdrjdKz8/n/vvv7/nhYIWxn8bC+7G3p0nEPoY+M8iv0K5uz/558+fz6ZNmwIuYcGCBdTU1PD6668HfKwQ3ao7AP+arv0cNwaGXwnn/ja8NYVZ2AL5008/paKigokTJ/q2ud1udu7cybPPPovdbm93xz0tLY3y8vI22yoqKjAYDJ2OilixYgW5ubm+996ZlwJir9TCeNxKiBkZ2LE90XgIvnhEu64fgVxWVub7+ZVXXuGee+5h//79vm3ekQxeTqcTo9EYvHqF6ImafSd+rv8Wmso633eACFsgT58+nX379rXZtnDhQk4//XTuuuuuDoc/5eTk8Oabb7bZtnXrViZNmtRpwJjNZsxmc3CKjhkJCacF51xBlJaW5vs5MTERRVF82w4ePEh6ejqvvPIKzz33HB9++CFr167l0KFDvP766+zdu9d37OrVq1m9ejUHDx7kvvvu48UXXwROtMDfffddRo0aBcB3333HsmXL+Oijjzj11FNZt24dOTk5ffOFRXRoPimA7cfDU0cECdtNvfj4eMaPH9/mFRsbS3JyMuPHjwe01u0NN9zgO2bx4sUcOnSI3NxcioqKeOGFF9iwYQN5eXnh+hr9xl133cXSpUspKipi5syZ3e6fl5fHT37yE/7jP/6DsrIyysrKuPDCC32f33333eTl5bF3715OO+00fvrTn+JyuUL5FUS0aSht+15ayJHzYEhHysrKKCkp8b3Pyspiy5YtLFu2jDVr1pCRkcHTTz/NvHnzwlhl/3DHHXdw9dVX+71/XFwcVqsVu93epgXulZeXx+WXXw7A/fffz7hx4/j22285/fTTg1aziHKNJW3f9+VN8wgVUYG8ffv2Nu87uhE1depU9uzZ0zcFRZFJkyYF9XytH+Lxjm6pqKiQQBb+ayzF41EoOjIWw9Ai4jzVDAt3TWEW9nHIom/Exsa2ea/T6XwjVLwCGcvdus/e28fs8Xh6UaEYcBpKuWHdZsbnv8W4YiOnfdfMwaoD4a4qrCSQB6ihQ4dSXl7eJpRb3+ADMJlMuN3uPq5MDATfl3r4n22T+Z//+zmMfxm33kGjCn//+q/hLi2sJJAHqGnTpnHs2DEef/xxDhw4wJo1a3j77bfb7DNq1Cg+//xz9u/fT2VlpTwNKYKiqgrOOht+vkYbxcO4P/s++6z80zBVFRkiqg854jUeiprrjB07lueee45HHnmEBx98kHnz5pGXl8f69et9+9x0001s376dSZMmUV9f32bYmxA9tWMHVFe3tAWT90N6oe+z72tLOzlqYFDUkzsSo5zNZiMxMZHa2tp2q043NzdTXFxMVlZW26cHI/xJvf6u09+7iEr33Qfeh2en3jKFHanv+T47N3kMn976TXgKC5GuMudk0kL2R+wILRxlLgsheu3gQe2fD19zF8+kaPPYGD+6Fef5z3KodmAPfZNA9lfsCAlIIYLgyBHtn66h/6ZcaUYHnH5kKvt4lmpXHaqqDtgpWeWmnhCiT3kD+XiiNsTNpMAZCTUAeHDT4GwIU2XhJ4EshOhTLUsocsh8FAC3CqcPLQG3Nra9uqk6XKWFnQSyEKLPqCpUt+TtV0o9ALH6JEanHYSmwQBUN0sgCyFEyDU0gDYHlcohjzYZ1XDTKDKGHIUmbQm3qqaq8BUYZhLIQog+U9WStYaE72luGXA72jKGpEE2aPYGsrSQhRAi5LzdFfHpn/u2jbKcQkJCs6+FfKRKAlkIIULOG8jWlK9824aZMlFNRozOOAAOVhwLR2kRQcYh+6mkBCr7cMz6kCEwIoBhzwsWLODFF18kPz+fX/3qV77tr7/+OnPnzkVVVbZv384Pf/hD32eDBw/m7LPP5sEHH+Siiy7ybb/vvvs6XIcwOzubr776qt321goLC3nkkUfYuXMntbW1jBgxgqlTp3LnnXdy2mmRt9qK6FveQDYM/lb7JwZMOjNNxGFxW3ACB48fDl+BYSaB7IeSEhg7Fhr78MnpmBgoKgoslC0WC4899hg333wzgwYN6nS//fv3k5CQwLFjx3jooYe4/PLL+frrr0lJSfHtM27cOP7xj3+0Oc5g6Ppfl7/97W/MmzePmTNn8j//8z+ccsopVFRU8L//+7/85je/4ZVXXvH/y4io5A1kJVGbs8KsaMurNaiJxLjN1AFHao+Eqbrwk0D2Q2WlFsYrV8LIPljj9NAheOQR7bqBBPKll17Kt99+S35+Po8//nin+6WkpJCUlERaWhq//vWv+fOf/8xHH33EFVdc4dvHYDB0uFJIZxobG1m4cCGzZ8/mtdde823Pysri/PPPp6amxv8vIqKW96aeGqsNRrbqtQV4G0gkDiNHgcqGgfv4tARyAEaOhEj+q1uv1/PII49w3XXXsXTpUoYPH97l/o2NjWzcuBGg16tQ//3vf6eyspLly5d3+HlSUlKvzi+ig7eF7LJoyRyj0/qNG0kkTtHm3rY114WltkggN/WizNy5c5kwYQL33ntvp/sMHz6cuLg44uLiePLJJ5k4cSLTp09vs8++fft8+3hfixYt6vSc33yjzdAlSziJrngDudlcA0C8PhHQuiwS9NqKM/WugfvotLSQo9Bjjz3GJZdcwi9/+csOP3/vvfeIjY2lsLCQu+66i02bNrVrIWdnZ/PGG2+02RYfHw/AI488wiOPPOLb/uWXX7ZbDkqIjngDudGotYKTDNq9jgYSGWTQHqV2eJrDUlskkECOQlOmTGHmzJmsXLmSBQsWtPs8KyuLpKQkTjvtNJqbm5k7dy7//ve/MZvNvn1MJhNjxozp8PyLFy/mJz/5ie99RkaGbwTFV199RU5OTnC/kIga1dWAoZlmvQOAJL32uHQT8SQZtC4Lt27gBrJ0WUSpRx99lDfffJMPPvigy/2uv/56PB4Pzz33nN/nHjx4MGPGjPG9DAYDM2bMYMiQIZ3eTJSbegJabuq13NBTgEEtLWQVHUktf6SpeglkEWXOPPNMfvazn/HMM890uZ9Op+OOO+7g0UcfpbHVuD6Xy0V5eXmb19GjRzs9T2xsLM8//zxvvfUWV155Jf/4xz84ePAgu3fvZvny5SxevDho3030X9XV+AJZB8QZTqygkWDR5kBWDc3UNteGobrwky6LABzqoyX1gnWdBx98kD//+c/d7veLX/yCe++9l2effdY3SuKLL74gPT29zX5ms5nm5s5bL3PmzOGDDz4gPz+f6667DpvNRmZmJpdccgkPPfRQ776MiAq2WhUSjvvex7WMsgCIN+u1Hwx2SmpLONNyZl+XF3ZhDeS1a9eydu1aDras6TJu3DjuueceZs2a1eH+Jz9p5lVUVBTSu/tDhmgParS6jxVyMTHadf21adOmdttGjhzZJkCnTZvW4c232NhYqqpOzLB13333cd999wVSrs+kSZMoKCjo0bEi+tlsQKr275obiNXH+z6Lt5yIo8rGgTkWOayBPHz4cB599FHfzaMXX3yROXPmUFhYyLhx4zo9zvukmdfQoUNDWueIEdpTc5H86LQQkc7lgma7AtYT//GP059oIZstJnCZwOCQQA6H1k+GATz88MOsXbuWDz/8sMtA9j5p1pdGjJCAFKI36rzPe7QEsh4dBqXVcEuzHmriwXB8wAZyxNzUc7vdvPzyyzQ0NHQ7bOqcc84hPT2d6dOn8+677/ZRhUKI3rDZtH/qYrXZ3EyKqe0OFhPYtS6MstqBOUl92G/q7du3j5ycHJqbm4mLi+O1117jjDPO6HDf9PR01q9fz8SJE7Hb7fzhD39g+vTpbN++nSlTpnR4jN1ux263+97bvP9WCCH6lLeFbIwrxw6Ydda2O5gNYNe6Ir+vqOnT2iJF2AM5OzubvXv3UlNTQ0FBAfPnz2fHjh0dhnJ2djbZ2dm+9zk5OZSWlrJq1apOAzk/P7/DqSSFEH3LG8i6WK07wqqLafO5YtaDQ2shH60emJPUh73LwvtE2KRJk8jPz+fss8/mqaee8vv4Cy64wDePQkdWrFhBbW2t71VaWhqMsoUQAfL+capYtWFvrW/oAbgNZhSHtq2qbmBOMBT2FvLJVFVt08XQncLCwnbjZVszm81tHgkWQoSHN2NVaw0AcfqENp87FCsGlxUnUN84MLsWwxrIK1euZNasWWRmZlJXV8fLL7/M9u3beeeddwCtdXv48GE2b94MwOrVqxk1ahTjxo3D4XDw0ksvUVBQIONehegHvIHsNmtP4XlnevNyYMHYsmqI3dGHq0FEkLAG8tGjR7n++uspKysjMTGRs846i3feeYfLLrsMgLKyMkpKSnz7OxwO8vLyOHz4MFarlXHjxvHWW28xe/bscH0FIYSftC4LFWfLTG+DWiYWOkGH2W2mEXC6B+Z8FmEN5A0bNnT5+clPny1fvrzTCdCFEJGtrg4w21B12rzHg4wnBzKYPdpQOKfb/27LaBJxfciRqqS2pE8Hqw+JGcKIRP+fRAn3IqebNm1i4cKFvvexsbFkZ2dz9913c/XVV/v9PUT0qquj1UMhkKhParePGS2Q3Tj7rrAIIoHsh5LaEsauGUujs+/6tWKMMRQtKQoolMO9yGlCQgL79+8HoK6ujo0bN/KTn/yEL774os1wRTEw2Wz4AtmgQFyreSy8LC2R5MbVl6VFDAlkP1Q2VtLobGTl5JWMTAz9KqeHag/xyHuPUNlYGVAgh3ORUwBFUXzHpKWl8dBDD7Fq1So+//xzCWTRpoWso5NA1kkgCz+NTBzJacmRu8ppOBc5PZnb7faNjjn33HODem7RP7UOZAUwK5Z2+1i9gdyy4OlAI4EcZVovctrZTVNvUDc2NqKqapeLnLZ27bXX8vzzz3d67draWt8xTU1NGI1G1q9fzymnnNKbrySihM2mQoz2UIiCHkVR2u1j1muR5NFJC1lEiXAscur9fM+ePYAW9v/4xz+4+eabSU5Objeznxh46mxuiPO2kE0d7mPRe1vIclNPRIlwLHIK2nJQrY8566yz2Lp1K4899pgEstBWCxmqBbJR1767AsBq0BoG7gHaQg77XBYiNPp6kdPO6PV6mpqa/D63iF62Vn3I1pNnemthaflLza0fmOOQJZCjVF8vcgraPCTefYuLi1m/fj1///vfmTNnTlC+k+jf6ut0vkCO7WCEBYDVF8gO3J6Bd2NPuiwCcKi2b1Y5DdZ1+nqRU5vN5jvGbDYzcuRIHnjgAe66665efAsRDdxuaGjS+wI5TpfQ4X4Wo9Ztpuqd2Ow2Blk7H08fjRS1o1Uvo5jNZiMxMZHa2to26/IBNDc3U1xcTFZWFhbLiT6u/vJgSH/V2e9dRI/aWkhKApaMhaFfMSVhOnOTr22338F9Op6KuxGA/Uv2c9qQyB1m6q+uMudk0kL2w4jEERQtKYroR6eFiGQnr6eXcNJMb14msxlUBRSVcltlVARyICSQ/TQicYQEpBA95J3p7UQgJ3W4n86ih6YYMDVQfvx7GN1nJUYEuaknhAi5ujrAVA96bTjbIEPHfcMGM+CIBeB4VUmH+0QzCWQhRMi1fmzaQPvJ6b3MFjc4tUCuqjnSR9VFDglkIUTI2Wz4Hps2K+0XOPUyWZzg1D47XjvwFjqVQO7AABt4Enby+45+dXWARQtYQxeBbDS6fF0WNQNwoVMJ5Fa88zm0fkBChJ739x3sGedE5NACuQbQQseo6/h/a50OdG5tLLKtYeD9/1BGWbSi1+tJSkqioqICgJiYmA5npBLBoaoqjY2NVFRUkJSUhF6vD3dJIkRsNsCiLW6q66YdaHCbcQANjoH3+LQE8km8E6x7Q1mEnneyfBG9tPX0/AxkjxbITY6BN+ObBPJJFEUhPT2dlJQUnM6B9y9EXzMajdIyHgBstaqvhazQddeUUdWm5mx2D7wZ3ySQO6HX6yUohAiSOpvL10JG6XguZC9TSyDb3Z5QlxVx5KaeECLk6mwu3009fQdLN7VmUrUWtEMdeLO9SSALIULOVuv2dVlYOhny5mVSWgJ5AE6/GdZAXrt2LWeddRYJCQkkJCSQk5PD22+/3eUxO3bsYOLEiVgsFkaPHs26dev6qFohRE/V1am+LgurPrbLfc0tgexEuiz61PDhw3n00UfZvXs3u3fv5pJLLmHOnDl88cUXHe5fXFzM7NmzmTx5MoWFhaxcuZKlS5dSUFDQx5ULIQJRV6f4Wsgx+rgu9zUr2q0t5wBceTqsN/VOXmft4YcfZu3atXz44YeMGzeu3f7r1q1jxIgRrF69GoCxY8eye/duVq1axbx58/qiZCFED9TVK74Wcqyu60C2tDw0MhADOWL6kN1uNy+//DINDQ3k5OR0uM+uXbuYMWNGm20zZ85k9+7dnQ5Rs9vt2Gy2Ni8hRN+qbzD4burF6buepN2i1wLZpQy8YW9hD+R9+/YRFxeH2Wxm8eLFvPbaa5xxxhkd7lteXk5qamqbbampqbhcLiorO548Pj8/n8TERN8rMzMz6N9BCNE5VYX6Rr2vy6Kzyem9LC2L5roG4MrTYQ/k7Oxs9u7dy4cffsgtt9zC/Pnz+fLLLzvd/+RHmb0T03T2iPOKFSuora31vUpLS4NXvBCiWw4HuBUn6LW/YrsP5JYWss4R8toiTdgfDDGZTIwZMwaASZMm8cknn/DUU0/xu9/9rt2+aWlplJeXt9lWUVGBwWAgOTm5w/ObzWbMZnPwCxdC+KW+Hl/rGCDRz0B26wbek7JhbyGfTFVV7PaOJxXJyclh27ZtbbZt3bqVSZMmyUxhQkQoLZBrgJa5kLsZ9mY1tQSyvnnATc0a1kBeuXIl7733HgcPHmTfvn3cfffdbN++nZ/97GeA1t1www03+PZfvHgxhw4dIjc3l6KiIl544QU2bNhAXl5euL6CEKIb9fX4RlgY6XwuZC+LUfuL1mOwU9fYHOLqIkvAXRbe5dqD4ejRo1x//fWUlZWRmJjIWWedxTvvvMNll10GQFlZGSUlJ9bVysrKYsuWLSxbtow1a9aQkZHB008/LUPehIhgrbssDAqYupnLwmLRPlcNdsqOHSUhdlSIK4wcAQfymDFjmDJlCjfeeCM/+tGPsFi6fi69Kxs2bOjy802bNrXbNnXqVPbs2dPjawoh+lbrFrIBpds5xq0mM7Tcz/u+opjsUaNCW2AECbjL4rPPPuOcc87hl7/8JWlpadx88818/PHHoahNCBEFWreQ9X5EjtlyogV9tPJgiKqKTAEH8vjx43niiSc4fPgwGzdupLy8nIsvvphx48bxxBNPcOzYsVDUKYTop1rf1NMp3f9RbrIADq2fuaJyYK083eObegaDgblz5/LnP/+Zxx57jAMHDpCXl8fw4cO54YYbKCsrC2adQoh+qnWXhU7pfjSUyewEpzYS42j1wFp5useBvHv3bv7rv/6L9PR0nnjiCfLy8jhw4AD/+te/OHz4MHPmzAlmnUKIfqq+ztNqPb2ub+gB6A0qOLUWcvUAW3k64Jt6TzzxBBs3bmT//v3Mnj2bzZs3M3v2bHQ6LduzsrL43e9+x+mnnx70YoUQ/U+9ze5rIesVq1/H6F0W3EBt08BaeTrgQF67di2/+MUvWLhwYacLU44YMaLbERRCiIGh3ub09SGbdF0/FOKlc2uB3NA0sFaeDjiQt23bxogRI3wtYi9VVSktLWXEiBGYTCbmz58ftCKFEP1XQ53L12Vh7WYuZC+D24wTaHQNrPksAu5DPuWUUzqcWa2qqipoD4wIIaJHfZ37xFzI+ni/jjF6tL7mJtfAmvEt4EDu7Nny+vr6Xj0kIoSITvX1J27qxfkZyAbvytMDbF09v7sscnNzAW2ay3vuuYeYmBPPo7vdbj766CMmTJgQ9AKFEP1bff2J9fTi/QxkU0sgD7SFTv0O5MLCQkBrIe/btw+T6cTwFZPJxNlnny2T/Agh2qmrV3039eL1SX4dY/Kuq4cEcofeffddABYuXMhTTz1FQkLXy7AIIQRArb0RdNoK0kmGQX4dY24ZrzzQVp4OeJTFxo0bQ1GHECJK2dzawx0GVel2tRAv38rTEsjtXX311WzatImEhASuvvrqLvd99dVXg1KYECI6NHi0QDaj73YuZC+zrmVdvQG20KlfgZyYmOibMi8x0b//wgkhBECjzhvICmbFv+XUzDptzgvpQ+5A624K6bIQQvjL7QaHvgHQlm/qbi5kr4G68nTA45CbmppobDzxfPmhQ4dYvXo1W7duDWphQoj+r6EBsFYBYNL5vz6eRe9deVoCuUtz5sxh8+bNANTU1HDeeefx29/+ljlz5rB27dqgFyiE6L/q6wGrNoWmOYC0sRq9LWR5dLpLe/bsYfLkyQD85S9/IS0tjUOHDrF582aefvrpoBcohOi/tMnptUA2+jE5vZdFrw17c+sd4HGGorSIFHAgNzY2Eh+vPW2zdetWrr76anQ6HRdccAGHDh0KeoFCiP5LayFrXRZmPyan9zIZtUD26O247Q2hKC0iBRzIY8aM4fXXX6e0tJS///3vzJgxA4CKigp5WEQI0UbrLovuVptuzWzQRmN4DM002upDUVpECjiQ77nnHvLy8hg1ahTnn38+OTk5gNZaPuecc4JeoBCi/2rdQjYp/k8+ZjZp+3oMTRyvOh6K0iJSwE/q/ehHP+Liiy+mrKyMs88+27d9+vTpzJ07N6jFCSH6t4YGfH3I/k5OD2AymMEBGBuprq1gVEiqizwBBzJAWlpau9VCzjvvvKAUJISIHvV1aqs+ZP8mpwcwma3QCCgqx46Xhqi6yBNwl0VDQwO/+c1vuPDCCxkzZgyjR49u8wpEfn4+P/jBD4iPjyclJYWrrrqK/fv3d3nM9u3bURSl3eurr74K9KsIIUKs3uZsNezNv4mFAKytZpMsqzoa9LoiVcAt5EWLFrFjxw6uv/560tPT/X7ypiM7duxgyZIl/OAHP8DlcnH33XczY8YMvvzyS2Jju/7zZv/+/W1uIg4dOrTHdQghQqPW1gBm7dHpGP1gv4+zmFVwmcDgoLK2JkTVRZ6AA/ntt9/mrbfe4qKLLur1xd9555027zdu3EhKSgqffvopU6ZM6fLYlJQUkpKSel2DECJ0KuuqoKWnItGQ7Pdxer0KzhgwOKiqqwtRdZEn4C6LQYMGMXiw//+lC0RtrbaqgD/nP+ecc0hPT2f69Om+uZo7YrfbsdlsbV5CiL5R1aitv2lym4n1c+pNL8VlBaC2qbGbPaNHwIH84IMPcs8997SZzyIYVFUlNzeXiy++mPHjx3e6X3p6OuvXr6egoIBXX32V7Oxspk+fzs6dOzvcPz8/n8TERN8rMzMzqHULITpXbddu6FndFqwG/2/qAehaArnO3hz0uiJVwF0Wv/3tbzlw4ACpqamMGjUKo7Ht0zd79uzpUSG33norn3/+Oe+//36X+2VnZ5Odne17n5OTQ2lpKatWreqwm2PFihW+9QABbDabhLIQfaTWXgOAxWPCavBvPT0vnduCG2h02oNfWIQKOJCvuuqqoBdx22238cYbb7Bz506GDx8e8PEXXHABL730Uoefmc1mzGb/5mAVQgSXzVkDgFU1YFL8m5zeS++y4ASaXANnLouAA/nee+8N2sVVVeW2227jtddeY/v27WRlZfXoPIWFhaSnpwetLiFEcNR7tPtCVvQBj8jSe7Shb/YBtPJ0jx4Mqamp4S9/+QsHDhzgzjvvZPDgwezZs4fU1FSGDRvm93mWLFnCH//4R/76178SHx9PeXk5oK1KYrVq/UcrVqzg8OHDvik/V69ezahRoxg3bhwOh4OXXnqJgoICCgoKevJVhBAh1KhqgRzTg+GxRo/2+LQEchc+//xzLr30UhITEzl48CA33XQTgwcP5rXXXvNNw+kv7/zJ06ZNa7N948aNLFiwAICysjJKSkp8nzkcDvLy8jh8+DBWq5Vx48bx1ltvMXv27EC/ihAixJqUngeyoaWF7FAlkDuVm5vLggULePzxx33TcALMmjWL6667LqBzqWr3Kwhs2rSpzfvly5ezfPnygK4jhAgPu04bZhoT8HguMDDwAjngX9Mnn3zCzTff3G77sGHDfF0OQggB4DRoLeTYAJZv8jKq2s14pyKB3CmLxdLhwxX79++Xx5eFED6qCi5TSwtZ7wn4eBPelacDP7a/6tGaeg888ABOpzYURVEUSkpK+NWvfsW8efOCXqAQon9qbsY39Wa8sSeB3NJCRlrInVq1ahXHjh0jJSWFpqYmpk6dypgxY4iPj+fhhx8ORY1CiH6o9VzI8YbAQ9XUsuSTawB1WQR8Uy8hIYH333+ff/3rX+zZswePx8O5557LpZdeGor6hBD9VOvlm6x6/5dv8jLqtGOcOlcwy4poPRqHDHDJJZdwySWXBLMWIUQUqbY5waQtUBqjWAM+3qRoXRYuxQVuB/Qg1PsbvwL56aef9vuES5cu7XExQojoUVZd7fs5Vq89Bh0Ik64lkHVOcNVDAPMp91d+BfKTTz7Z5v2xY8dobGz0zUdcU1NDTEwMKSkpEshCCACO1mqBrHPEYdTFBB7IhpMC2Rz9gezXTb3i4mLf6+GHH2bChAkUFRVRVVVFVVUVRUVFnHvuuTz44IOhrlcI0U9U1GmBbHTEQYATCwEYddqj0y6dA5wDY5L6gEdZ/OY3v+GZZ55pMwVmdnY2Tz75JL/+9a+DWpwQov+qrG9ZbdoZi04JbOpNAKNB63d26x3gqApqbZEq4EAuKyvzjUFuze12c/TowFmMUAjRtaom7Sk9s8eCThd4IHu7LNx6O87GgfEUcMCBPH36dG666SZ2797tm4ti9+7d3HzzzTL0TQjhY2vSuhnMHiP0oIVsaVn8wmNowlYjgdyhF154gWHDhnHeeedhsVgwm82cf/75pKen8/zzz4eiRiFEP2RrbnlsWjXg6sGwt1iTNuZANTg4XlMR1NoiVcDjkIcOHcqWLVv45ptvKCoqQlVVxo4dy2mnnRaK+oQQ/VSdoxYsEIMeJ4Hf1LMaDXiHZhyvlUDu0qmnnsqpp54azFqEEFGkwWkDC8QqOly6wFvIMSadL5CrG48HubrI1INZSoUQonuNrpYuC0VB1QW+rqXJ7AGH1rK2NdUEs7SIJYEshAiJJo83kFVQAv9j3GhygVML5KqGgbHytASyECIkmtV6oGW1EJ0x4OO1QI4FoLZ5YKw8HVAgu1wu7r//fkpLS0NVjxAiSjjQAjlW7+lRC1mvV8Gp9T3XNw+MKTgDCmSDwcB///d/43YPjF+OEKLnnDotkON0HlD0PTqHzqV1WTQ4B8aqIQF3WVx66aVs3749BKUIIaKJS98SyPqeN+B0bm0+iwb3wAjkgP+OmDVrFitWrODf//43EydOJDY2ts3nV155ZdCKE0L0X26D9qRebA9WC/HSuy24gGZ34Iuk9kcBB/Itt9wCwBNPPNHuM0VRpDtDCAGAatRayPEGNz1d80Pf0kK2D5BYCbjLwuPxdPoKNIzz8/P5wQ9+QHx8PCkpKVx11VXs37+/2+N27NjBxIkTsVgsjB49mnXr1gX6NYQQIWR3uMHUCECCsedLMOlVbfyyXR0YLeSwDnvbsWMHS5Ys4cMPP2Tbtm24XC5mzJhBQ0NDp8cUFxcze/ZsJk+eTGFhIStXrmTp0qUUFBT0YeVCiK6UV5+YvzjR1PNANni0FnLzwMjjnj06vWPHDlatWkVRURGKojB27FjuvPNOJk+eHNB53nnnnTbvN27cSEpKCp9++ilTpkzp8Jh169YxYsQIVq9eDcDYsWPZvXs3q1atYt68eT35OkKIIKuoaQlktwFLzwZYAGCQFnLXXnrpJS699FJiYmJYunQpt956K1arlenTp/PHP/6xV8XU1mrzpw4e3PlSLbt27WLGjBltts2cOZPdu3d3OE+zEKLvVdRqgaw44lB0PV+c1KBqLWQHKgyAUA64hfzwww/z+OOPs2zZMt+222+/nSeeeIIHH3yQ6667rkeFqKpKbm4uF198MePHj+90v/LyclJTU9tsS01NxeVyUVlZSXp6epvP7HY7dvuJxy5tNluP6hNC+K/SpgWyzhmHogT+lJ6XoWXlaTsqeOygtwSlvkgVcAv5u+++44orrmi3/corr6S4uLjHhdx66618/vnn/OlPf+p2X0VR2rz3TpR/8nbQbhwmJib6XpmZmT2uUQjhn8o6LZANLitKDx6b9jLibSF7wBH96+oFHMiZmZn885//bLf9n//8Z4/D7rbbbuONN97g3XffZfjw4V3um5aWRnl529UDKioqMBgMJCcnt9t/xYoV1NbW+l7y2LcQoVfV4A1kCx59j2f59S102qSqYI/+OZED/k398pe/ZOnSpezdu5cLL7wQRVF4//332bRpE0899VRA51JVldtuu43XXnuN7du3k5WV1e0xOTk5vPnmm222bd26lUmTJmE0tv8vsdlsxmwOfOo/IUTP1TRqgWxym3HR824GY8u0nXbFDc3HglJbJOvRgyFpaWn89re/5c9//jOgjXR45ZVXmDNnTkDnWrJkCX/84x/561//Snx8vK/lm5iYiNWqTSqyYsUKDh8+zObNmwFYvHgxzz77LLm5udx0003s2rWLDRs2+NXVIYToGzW+9fRMuOj5TT2zXjvWrrjALoHcoblz5zJ37txeX3zt2rUATJs2rc32jRs3smDBAkBb5bqkpMT3WVZWFlu2bGHZsmWsWbOGjIwMnn76aRnyJkQEsTVrgWxRjTiVnreQzXrtr16H4gJ7ZVBqi2Q979wJAtWPYSybNm1qt23q1Kns2bMnBBUJIYLBZm+ZnB59r7osLC2B7NQ5wVEVlNoimV+BPHjwYL7++muGDBnCoEGDOhzN4FVVFf2/NCFE1+qddWDQ1tNz9mDFaS9ry30hl86Baj9O58kTHfwK5CeffJL4+HgA3xNyQgjRmUaXFshxKDh700I2ao/5ufTNOO3He9Eb3T/4Fcjz588HtBVDQHsyLi0tLXRVCSH6tSa31occr6dXfcgxJi2iPAY79qZjUR/IAa8Ycsstt7R58k0IIU7W7NECOUHvwUXPuyxiTS0TYejc1DXWBKGyyBbwgyHnn38+hYWFoahFCBEl7C0LnCbp3Th1sd3s3Tmr8cQf8TVN0T/tQcCjLP7rv/6LX/7yl3z//fcdrhhy1llnBa04IUT/5FBauiwMblRdz7ssrBagzgB6F7XNjUGqLnIFHMjXXHMNAEuXLvVtUxQFVVVlxRAhBAAu7wKnehfoev6krMnshKpY0NdS74j+rtKAA7k3EwgJIQYGt05bZCLB4MLVi8mFzGYXOGPAUkvtAJilPuBAHjlyZCjqEEJEEbdR67KIMzipUXo+Q73JG8hAjT2sCxz1iR59wz/84Q9cdNFFZGRkcOjQIUAbn/zXv/41qMUJIfofl8cFhmYAEoxO6MXjHDqd6gvkOocEcjtr164lNzeX2bNnU1NT4+szTkpKkodGhBDU20+siRlv6P0qPjq3NmyuztWLtaD6iYAD+ZlnnuH3v/89d999N3r9iV/QpEmT2LdvX1CLE0L0P8dsJ9bTizE5en0+nUsL5AZXWKfe6RMBB3JxcTHnnHNOu+1ms7nL1aKFEANDRY02wgJHHEZj70dd6VtWnm5068Hd+4CPZAEHclZWFnv37m23/e233+aMM84IRk1CiH7smE0LZMURh6rrfatW79GGzTW6deCK7mWcAv5t3XnnnSxZsoTm5mZUVeXjjz/mT3/6E/n5+Tz//POhqFEI0Y8cbwlkncuKqvT+Rpxe1bosmjx6cFSDuf1SbdEi4EBeuHAhLpeL5cuX09jYyHXXXcewYcN46qmnuPbaa0NRoxCiHzlef2KBU4/S+7HDBlXrsqhzK9BUBvFjen3OSNWjvyduuukmbrrpJiorK/F4PKSkpAS7LiFEP1XdoLWQjW4LLqX3fciGluk7Gz0KNB/t9fkiWcB/T1xyySXU1NQAMGTIEF8Y22w2LrnkkqAWJ4Tof2oavYFs6tXUm15GtD7kBpWoX1cv4Bby9u3bcTja3+lsbm7mvffeC0pRQoj+q9a7wKlqxNGzZ8/aMCpaIDepQHN0r6vndyB//vnnvp+//PJL3wrRAG63m3feeYdhw4YFtzohRL9TZ9dayBbVgIOeTyzkZdJp09I3qR5wHO/1+SKZ34E8YcIEFEVBUZQOuyasVivPPPNMUIsTQvQ/3kCOQY9Tien1+UwtkxM1K25Ue1VUr6vndyAXFxejqiqjR4/m448/ZujQob7PTCYTKSkpbZ7cE0IMTPXOOlAgRlFw0PPJ6b0sOm3YW7PixN58vBcr9EU+vwPZO8ubx+MJWTFCiP6v0VkPJohTFBxK7wM5Rq91e9j1dprt0R3IAfe4v/jii7z11lu+98uXLycpKYkLL7zQN/ObEGLganJrXRbxOhVHL5Zv8rLotQh26Jtx2Gt7fb5IFnAgP/LII1it2p8Qu3bt4tlnn+Xxxx9nyJAhLFu2LKBz7dy5kyuuuIKMjAwUReH111/vcv/t27f7+rFbv7766qtAv4YQIkSaPNradwl6FacS3+vzxRi0FrLL0ITTFd3z5QQ87K20tJQxY7QnZV5//XV+9KMf8f/+3//joosuYtq0aQGdq6GhgbPPPpuFCxcyb948v4/bv38/CQkJvvet+7OFEOHlUFvW09O7g3JTL86kjbJwGRpxOky9Pl8kCziQ4+LiOH78OCNGjGDr1q2+VrHFYqGpqSmgc82aNYtZs2YFWgIpKSkkJSUFfJwQIvQctHRZ6N2g732Pb5y5ZQkonYc6h6vX54tkAXdZXHbZZSxatIhFixbx9ddfc/nllwPwxRdfMGrUqGDX16FzzjmH9PR0pk+fzrvvvtsn1xRC+MfZssBpjN4NSu9ne4s168GjRVWdq/cT3keygAN5zZo15OTkcOzYMQoKCkhO1mZe+vTTT/npT38a9AJbS09PZ/369RQUFPDqq6+SnZ3N9OnT2blzZ6fH2O12bDZbm5cQInRcei2QY/UeerN8k5fF4gK71kXZ6InuVe0D/s9XUlISzz77bLvt999/f1AK6kp2djbZ2dm+9zk5OZSWlrJq1SqmTJnS4TH5+fl9UpsQQuMxaDfe4g1OmoNwPrPFqQWytYZGtwRyOzU1NWzYsIGioiIURWHs2LHceOONJCYmBru+bl1wwQW89NJLnX6+YsUKcnNzfe9tNhuZmZl9UZoQA47D7QC9NtdNvNEV3EAGap0m8Dih5em9aBNwl8Xu3bs55ZRTePLJJ6mqqqKyspInn3ySU045hT179oSixi4VFhaSnp7e6edms5mEhIQ2LyFEaHgfmwYtkIPBaDrRZWFzmsEZvauGBNxCXrZsGVdeeSW///3vMRi0w10uF4sWLeKOO+7osj/3ZPX19Xz77be+98XFxezdu5fBgwczYsQIVqxYweHDh9m8eTMAq1evZtSoUYwbNw6Hw8FLL71EQUEBBQUFgX4NIUQIHK9vCWSXGWt8cG7A6XSgOONQgTqnERy1YB4clHNHmoADeffu3W3CGMBgMLB8+XImTZoU8Ll++MMf+t57uxbmz5/Ppk2bKCsro6SkxPe5w+EgLy+Pw4cPY7VaGTduHG+99RazZ88O9GsIIULgaHVL69URh8kUvBERelccLqDaaQR7JcRnBe3ckSTgQE5ISKCkpITTTz+9zfbS0lLi4wN7KmfatGmoaudLvGzatKnN++XLl7N8+fKAriGE6DvHarwLnMai6IPXz6t3twSyywDN5d3u318F3Id8zTXXcOONN/LKK69QWlrK999/z8svv8yiRYtCPuxNCBHZKuu8C5zG4FGC91Sd0aPNiVHt1kV1IAfcQl61ahWKonDDDTfgcmmd9kajkVtuuYVHH3006AUKIfqP43XeBU4tuIPwlJ6X0aM9gl3r1kFz9C7jFHAgm0wmnnrqKfLz8zlw4ACqqjJmzBhiYnr/zLoQon87scCpGVcQ1tPzMrbMq1znQetDjlJ+d1k0NjayZMkShg0bRkpKCosWLSI9PZ2zzjpLwlgIAUBVg/YkrNljxh6Eyem9zGgzTNapHtxRvK6e34F87733smnTJi6//HKuvfZatm3bxi233BLK2oQQ/UxNkzZfscVjxh6Eyem9zIoWyPWKi6YoDmS/uyxeffVVNmzYwLXXXgvAz3/+cy666CLcbrcs3SSEAKCmWQvkGI8hqC1kS8s0ng2KA7u9irignTmy+N1CLi0tZfLkyb735513HgaDgSNHjoSkMCFE/1Pn0AI5TtFjD2Jsxuq1QG7S2XE4onfVEL8D2e12YzK1HcZiMBh8Iy2EEKLeWQNAHHrsuuBNUxDTEsjNhiZczvpu9u6//O6yUFWVBQsWYDabfduam5tZvHgxsbEn/jR59dVXg1uhEKLfaHDVgAESFIKywKlXrEHrQ3brHTQ6o3cZJ78Def78+e22/fznPw9qMUKI/q3Jo3UnxOoAXfCGvcWZzaAqoKjYnMGYQy4y+R3IGzduDGUdQogo0Iw27E0L5OA9qWcxu6E5Eaw12KK4mzTgR6eFEKIzDkUL5BiFoM5ZbLY4oXkQADaXJ2jnjTQSyEKIoHHqW1rIQVq+yctscUKTFsj1HglkIYTokqqquA3aXBZx+s5ncewJs9kFTdocyI2qCq7AVrjvLySQhRBB0eBsAJ3Wek00BHd1aFOrLosGFXBUB/X8kUICWQgRFLUtT+nh0RNvCe6NN0urLgubW4Gm6JyCUwJZCBEUxxtaArk5kTirPajnbt1CPuYwQNP3QT1/pJBAFkIERVm1dkMPewIxMY6gnttodKM0JwFQ4TJAowSyEEJ0qqxKayErjnh0xuCNsABQFDC6tSXiKp36qF01RAJZCBEUR2u0QDY6Y4O6WoiXSdUmK6pyK9B0NOjnjwQSyEKIoKiw1QBgcluxK8GfINOiJgJQrXrwNFcE/fyRQAJZCBEUFXXHAbA6Y3CEIJCtijZ7XDVOGpskkIUQolPHGrWVPGI9FpqUpKCfP1bR+pAbFAf1TdG5aogEshAiKKqbtNWg41UTzUpi0M8fZ9DmRFYVKG+qCfr5I0FYA3nnzp1cccUVZGRkoCgKr7/+erfH7Nixg4kTJ2KxWBg9ejTr1q0LfaFCiG7V2LVuhEQMOPTBD+QYq+p7OKTCHp1zIoc1kBsaGjj77LN59tln/dq/uLiY2bNnM3nyZAoLC1m5ciVLly6loKAgxJUKIbpjc2ndCAnoQR/8leitMQ5oHAJAlTO445wjhd/zIYfCrFmzmDVrlt/7r1u3jhEjRrB69WoAxo4dy+7du1m1ahXz5s0LUZVCCH80oN3Ui1N0oDd3s3fgrDF2aBgKyd9Q5Y7OOZH7VR/yrl27mDFjRpttM2fOZPfu3TidHU9mYrfbsdlsbV5CiOBrVrQJfxIUBZTgzYXsZbE6oHEoADVuFdTom4azXwVyeXk5qampbbalpqbicrmorOz4rmt+fj6JiYm+V2ZmZl+UKsSA4vK4cBq1B0MSQ/R3d+suixo34Iy+xlW/CmQARWn7SKaqqh1u91qxYgW1tbW+V2lpachrFGKgqW46MR1mgj4017DGOLQuC8DmARxVoblQGIW1DzlQaWlplJe3fYa9oqICg8FAcnJyh8eYzeY2K2ULIYKvsmUMMk1JxFtCc8PN0qqFXOsB7FUQNzok1wqXftVCzsnJYdu2bW22bd26lUmTJmE0Br/PSgjhn+NN2g09GocQExua1TysrfqQj3uIyhnfwhrI9fX17N27l7179wLasLa9e/dSUlICaN0NN9xwg2//xYsXc+jQIXJzcykqKuKFF15gw4YN5OXlhaN8IUSLo3XeFnIyMfHNIbmGNfZEl8UxF9AYfd2PYe2y2L17Nz/84Q9973NzcwGYP38+mzZtoqyszBfOAFlZWWzZsoVly5axZs0aMjIyePrpp2XImxBhVlp5ooVsjQtRl4X1RJdFhUuHq6Gkf/W5+iGs32fatGm+m3Id2bRpU7ttU6dOZc+ePSGsSggRqNIqrYVsdMSjmoO7WoiXXq9icifiAI57VOpshxgUkiuFT7/qQxZCRKbyWq2FHOOKoallEqBQiNVp57ajUll/OGTXCRcJZCFErx21tZ5YKClk14kx68GpTX5/uCH6JqmXQBZC9NrxxhMTCzXrQteREBt7YqRFZVNtyK4TLhLIQoheq7ZrLeRERYdHH/zJ6b1i4+y+G3uVjuib8U0CWQjRa3Vu7am5RJ0SkomFvGLi7L6hbzWdzF/Tn0kgCyF6rUHVAjlBp4IudIEcG9d84mk9tztk1wkXCWQhRK+4PW7seq0/N0GvEspYiYmz+/qQbZ7Oh8z2VxLIQoheqWmuAUWbClML5NCJiT3Rh2zzAO7omqheAlkI0StHvcPPmpKIsYTmoRCv2LjmE33I3gmGoogEshCiV47WtwRyQypxsaEd+aC1kLVArnIDjSVdH9DPSCALIXrl+5qWQK5P1fp4Q6j1sLdjbqDhYEiv19ckkIUQvXKgXAtkpSEFc2xoRz60HvZW6QZX/XchvV5fk0AWQvTKwUotkGNdcbiMwV9tujVrjAOatMUoqj1QVb0/pNfraxLIQohe+b66DIBEjwXVHBvSa+l0KlYlFlRtybYj1dJCFkIIn6P1RwBIVvS4DIkhv15srAsak1uuXRby6/UlCWQhRK8ct2vrXCYrCh59aFvIcNJ8Ft6lo6KEBLIQoldq3dpMb8kGQjqPhVdMXLNv6FutM7omGJJAFkL0mKqqNOm1Vmqy3gN6a8ivGZ/QdGI+C6c8qSeEEID22LSq02ZdSzU5QW8J+TUTkpqgPhWAKnd0zWchgSyE6LHvbd9rPzQOJsnaCErol+lMTGqAumEAHHcDjuqQX7OvSCALIXqspLZU+6F2BNbY+j65ZkJSI9i0QD7iBmr29cl1+4IEshCix74+2jKXRO0IYhKa++SaiYMafS3kwy5Qq//dJ9ftCxLIQoge+6pMC2RjQyqGOH2fXFNrIQ8HtECuqdrbJ9ftCxLIQoge++64FsiJzkRcloQ+uWZCYiPUZQDa49NHKr/sk+v2hbAH8nPPPUdWVhYWi4WJEyfy3nvvdbrv9u3bURSl3eurr77qw4qFEF7f2w4BMMRjpUGX1ifX1BtUEqx6sGuLqRbXRs8UnGEN5FdeeYU77riDu+++m8LCQiZPnsysWbMoKen6F7x//37Kysp8r1NPPbWPKhZCtHa0WQvkVEzYDUP77LpDU+t8/chlTdEzSX1YA/mJJ57gxhtvZNGiRYwdO5bVq1eTmZnJ2rVruzwuJSWFtLQ030uv75u+KyHECW6PmxpVm0siTaeAMb7Prj0kpc430qLS0Tc3E/tC2ALZ4XDw6aefMmPGjDbbZ8yYwQcffNDlseeccw7p6elMnz6dd999t8t97XY7NputzUsI0Xvl9eWoigs8elJNLtCF/qEQr+Shdb4be9Wu6Fl9OmyBXFlZidvtJjU1tc321NRUysvLOzwmPT2d9evXU1BQwKuvvkp2djbTp09n586dnV4nPz+fxMRE3yszMzOo30OIgeo779SXNSNJjK0DlD67dvJQG9SOBLSJ6nH2zRjoUAv9YzXdUJS2/yOqqtpum1d2djbZ2dm+9zk5OZSWlrJq1SqmTJnS4TErVqwgNzfX995ms0koCxEEB6oPaD9UjyYurrZPr52cUgcfjgbgezeotV+hDJnUpzWEQthayEOGDEGv17drDVdUVLRrNXflggsu4Jtvvun0c7PZTEJCQpuXEKL3DlS1tJCrTyEusW9nXUtOsUG1FsjfORRqK7ru5uwvwhbIJpOJiRMnsm3btjbbt23bxoUXXuj3eQoLC0lPTw92eUKIbnxV0aqFnNzUp9eOjbNjbNBu6h1yqZQf/bBPrx8qYe2yyM3N5frrr2fSpEnk5OSwfv16SkpKWLx4MaB1Nxw+fJjNmzcDsHr1akaNGsW4ceNwOBy89NJLFBQUUFBQEM6vIcSA9M0xrYUc15iCLi60a+mdTFEgOSaecpcJl8HBgcovOL1PKwiNsAbyNddcw/Hjx3nggQcoKytj/PjxbNmyhZEjtc76srKyNmOSHQ4HeXl5HD58GKvVyrhx43jrrbeYPXt2uL6CEAPWQZvWQk5TrTSYMvr8+ikpDZTXjIIhX3PIdqTPrx8Kiqqq0TWhaDdsNhuJiYnU1tZKf7IQPVRnryPhUe3/P9P++hfm3FoMhqQ+reGNl8/j3cF3wanvcPcQMw8ticzxyIFkTtgfnRZC9D++IW+NyQyxNIG+7x4K8UoeWgc1WQBUOKJj5RAJZCFEwHyBXHUKifGVoPT907KtR1oci5KVQySQhRAB+7bqW+2H6tHEx4dnLonklDpfIB9xAfaasNQRTBLIQoiAFVUWaT9Unk5sYt8+FOI1KLkeqrRA/tqhUFv6t7DUEUwSyEKIgO0r/0L74dg4BqXVhKUGg8FDkjoMPDpqVJXPD/T/4a8SyEKIgKiqyhfHtEBOrB9GXIoxbLUMSXb6ui0+L/s0bHUEiwSyECIgJbUlNLkbwG3kDGszzTEjw1ZL2rBqOHYGAAfqjoatjmCRQBZCBMTbOub4aWQmf49q6ruJ6U82LLPKF8ilUTD0TQJZCBGQL4+1rGFXMY6hyd+DEr4HfjNGHofKsQB87wTsx8NWSzBIIAshAuJrIR8bR9LQyrDWkjasGqVSm5L3W4eOysNdL1gR6SSQhRABOTHC4gxSRoZ3BR6TyU1GTAoAlaqHz77+Y1jr6S0JZCGE35xuJ58f/QyAwfXD0KeEr//Y65SsBqgZAcCXRz4KczW9I4EshPBbUWURTtUBzQlMSKzGaRke7pLIOvWo78be4YaKMFfTOxLIQgi/7Snbo/1Qfg6j0r4FnSm8BeEN5HEAHGxSQPWEuaKek0AWQvjt0yMtgVx2LqkZh8JbTIvEQY3E2rQW8t4GK2rVZ2GuqOckkIUQftvxzW4AzMfOIPWUyBliNipmFADf0sCRrzeGt5hekEAWQvilydnEF9VaIF8c34x7SOSs3p49zASNybh1bj4p7r+TDEkgCyH88tHhj/AoTqhL5wdDD6OzpoS7JJ/Rp1bA4fMA+PL499BPF0KSQBZC+OXlj/4FgFJyMWNP/Qh01jBXdEJ6ZjWG8h8AsLUyAcr/EeaKekYCWQjhl7988SoAExxpKCmJYa6mLZ1OZYxpPAAfehqp/ey+8BbUQxLIQohufXXsa47rvwCPnutSj3E84eJwl9TOBafGQ9Mg7IYm/lH8Cbgic9HTrkggCyG6de+b6wDQfzed7FHFYIisFjLA+AmHMX4zF4DHStJoKrwrzBUFTgJZCNGlYw3HePXgegB+2DCW6kFnhrmijun1KudbLgPgE0spf/x0HXhcYa4qMGEP5Oeee46srCwsFgsTJ07kvffe63L/HTt2MHHiRCwWC6NHj2bdunV9VKkQA9Mv//oQLn0DHJnIj7L2U2WdFO6SOjXrYgf6j5YBcGuZh8933BjmigIT1kB+5ZVXuOOOO7j77rspLCxk8uTJzJo1i5KSkg73Ly4uZvbs2UyePJnCwkJWrlzJ0qVLKSjo/2tpCRGJ/vTBVv6w/1kAJhVfhTNtCCj6MFfVuZhYB5fF/hgOTqVZ5+LaD16i6dv+86CIoqrhG7B3/vnnc+6557J27VrftrFjx3LVVVeRn5/fbv+77rqLN954g6KiIt+2xYsX89lnn7Fr1y6/rmmz2UhMTKS2tpaEhITefwkhooyqqmzb/z6/+cvjfOx8G3RurF/+iP/OasB52kwwxIa7xC7Zmw08cM8UGq/PgbgKpikJ/HH6daT/4BEwDerzegLJnLBN9e9wOPj000/51a9+1Wb7jBkz+OCDDzo8ZteuXcyYMaPNtpkzZ7JhwwacTidGY4gWWzz2f9BU3sUO/vw3rZt9/PrvYm/PESF1+lNHr79LX1zDj30i7Pf1z/KvONpsY/+hOMoqzTQ5oMkBx+2N1LqraFCrKFWKaUr8RjtMB4NKJ7MkXYfz9MtBZ+6+ljAzW1zcuOjfrPnfTXiumcN2vY2Md14g/g+fMMidyiBjHINjrQyJs2A1xDBYyWD5paeQnurUTqAorc6mtP2n97P0WWAI/jjssAVyZWUlbreb1NTUNttTU1MpL+84/MrLyzvc3+VyUVlZSXp6ertj7HY7drvd9762thbQ/qvlt48fgLKt/u8vRIT69WH4sKPRYAbapkG9hZhvZzHRaOSwtZQ/NdbAB2/0TZFBMuzMTMrefALXWZsg41PqBn1KHeDrEPX+HkpzuLRax+Ts//P/5P9ZBDEZfu3qzRp/OiPCtxhWC6XNf420ok/e1t3+HW33ys/P5/7772+3PTMzcp7DFyLyNNPIa3R9i72f2NvdDrv4z4BPOjbgI+rq6khM7Hq4YNgCeciQIej1+nat4YqKinatYK+0tLQO9zcYDCQnJ3d4zIoVK8jNzfW993g8VFVVkZyc3GXwDxQ2m43MzExKS0ulT70b8rvyj/ye2lJVlbq6OjIyum9Rhy2QTSYTEydOZNu2bcydO9e3fdu2bcyZM6fDY3JycnjzzTfbbNu6dSuTJk3qtP/YbDZjNrft90pKSupd8VEoISFB/s/jJ/ld+Ud+Tyd01zL2Cuuwt9zcXJ5//nleeOEFioqKWLZsGSUlJSxevBjQWrc33HCDb//Fixdz6NAhcnNzKSoq4oUXXmDDhg3k5eWF6ysIIUTQhLUP+ZprruH48eM88MADlJWVMX78eLZs2cLIkSMBKCsrazMmOSsriy1btrBs2TLWrFlDRkYGTz/9NPPmzQvXVxBCiKAJ6zhkEX52u538/HxWrFjRrmtHtCW/K//I76nnJJCFECJChH0uCyGEEBoJZCGEiBASyEIIESEkkIXPww8/zIUXXkhMTIyM1T5JoNPEDlQ7d+7kiiuuICMjA0VReP3118NdUr8igSx8HA4HP/7xj7nlllvCXUpECXSa2IGsoaGBs88+m2effTbcpfRLMspCtLNp0ybuuOMOampqwl1KRAh0mlihURSF1157jauuuircpfQb0kIWogveaWJPnva1q2lihegpCWQhutCTaWKF6CkJ5Ch33333oShKl6/du3eHu8yIF+g0sUL0RNjnQxahdeutt3Lttdd2uc+oUaP6pph+qCfTxArRUxLIUW7IkCEMGTIk3GX0Wz2ZJlaInpJAFj4lJSVUVVVRUlKC2+1m7969AIwZM4a4uLjwFhdGubm5XH/99UyaNImcnBzWr1/fZppYcUJ9fT3ffvut731xcTF79+5l8ODBjBgxIoyV9ROqEC3mz5+voq2G2eb17rvvhru0sFuzZo06cuRI1WQyqeeee666Y8eOcJcUkd59990O/x2aP39+uEvrF2QcshBCRAgZZSGEEBFCAlkIISKEBLIQQkQICWQhhIgQEshCCBEhJJCFECJCSCALIUSEkEAWQogIIYEshBARQgJZiH5s06ZNTJgwAavVyuDBg5kzZw779+8Pd1mihySQRURYsGCBb35mo9HI6NGjycvLo6GhIdyl9di0adO44447Qnb+zZs3s3DhQj777DOSk5Npbm7mjTfeYOrUqRw7dixk1xWhI4EsIsZ//Md/UFZWxnfffcdDDz3Ec889R15eXo/O5XA4glxd+HT0XVRV5Z577gG031tJSQl79+4lLi6Oo0ePsmbNmr4uUwSBBLKIGGazmbS0NDIzM7nuuuv42c9+5ltG/p133uHiiy8mKSmJ5ORk/vM//5MDBw74jp02bRq33norubm5DBkyhMsuu8yv47zH3nbbbdxxxx0MGjSI1NRU1q9fT0NDAwsXLiQ+Pp5TTjmFt99+23eMqqo8/vjjjB49GqvVytlnn81f/vIX3+cLFixgx44dPPXUU76W/8GDB7s9rqvv0tqXX37JoUOHAG16UJ1Ox2mnncaVV14JwJYtW3rxv4QIFwlkEbGsVitOpxPQlpfPzc3lk08+4Z///Cc6nY65c+fi8Xh8+7/44osYDAb+7//+j9/97nd+H+c9dsiQIXz88cfcdttt3HLLLfz4xz/mwgsvZM+ePcycOZPrr7+exsZGAH7961+zceNG1q5dyxdffMGyZcv4+c9/zo4dOwB46qmnyMnJ4aabbqKsrIyysjIyMzO7Pa6r79La119/7fv5zjvv5IILLuCCCy5g+/bt7T4X/UhYJ/8UosX8+fPVOXPm+N5/9NFHanJysvqTn/ykw/0rKipUQN23b5+qqqo6depUdcKECd1e5+TjvMdefPHFvvcul0uNjY1Vr7/+et+2srIyFVB37dql1tfXqxaLRf3ggw/anPvGG29Uf/rTn7Y57+233+57H8hx3X2XzZs3dzjvsPel1+u7/V2IyCMtZBEx/va3vxEXF4fFYiEnJ4cpU6bwzDPPAHDgwAGuu+46Ro8eTUJCAllZWYC2yonXpEmT2p3Tn+MAzjrrLN/Per2e5ORkzjzzTN827/p5FRUVfPnllzQ3N3PZZZcRFxfne23evLldd0hrgRzX0XdpLSYmxvdzYWEhqqqiqiq33XYbALGxsV0eLyKTLOEkIsYPf/hD1q5di9FoJCMjA6PR6PvsiiuuIDMzk9///vdkZGTg8XgYP358mxteHYWQP8cBba4F+EZ7tH4P4PF4fN0db731FsOGDWtznNls7vT7BXJcd4F66qmn+n7+9ttvmTBhAoAv2MeMGdPl8SIySSCLiBEbG9thkBw/fpyioiJ+97vfMXnyZADef//9bs/X0+O6c8YZZ2A2mykpKWHq1Kmd7mcymXC73QEf548zzzyTYcOGcfjwYZ544gmuuOIK9u3bx7Zt2wBt5IXofySQRcQbNGgQycnJrF+/nvT0dEpKSvjVr34VsuO6Ex8fT15eHsuWLcPj8XDxxRdjs9n44IMPiIuLY/78+QCMGjWKjz76iIMHDxIXF8fgwYP9Os4fiqLw4IMP8otf/IJdu3Zx6qmnUllZidPpZOjQodx+++29/p6i70kgi4in0+l4+eWXWbp0KePHjyc7O5unn36aadOmheQ4fzz44IOkpKSQn5/Pd999R1JSEueeey4rV6707ZOXl8f8+fM544wzaGpqori42K/j/LVw4UIUReHJJ59k//79WCwWLrvsMh577DFSUlJ6/R1F35NFToUQIkLIKAshhIgQEshCCBEhJJCFECJCSCALIUSEkEAWQogIIYEshBARQgJZCCEihASyEEJECAlkIYSIEBLIQggRISSQhRAiQkggCyFEhPj/eOnaDNvShgkAAAAASUVORK5CYII=",
      "text/plain": [
       "<Figure size 366x366 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "ll = lambda theta: task.log_likelihood(theta, observation).exp()\n",
    "evidence, _ = scipy.integrate.quad(ll, low.item(), high.item())\n",
    "tt = prior.sample((num_posterior_samples,))\n",
    "\n",
    "kwargs = dict(\n",
    "    bins=100,\n",
    "    binrange=(low.item(), high.item()), \n",
    "    stat=\"density\",\n",
    "    kde=True,\n",
    "    element=\"poly\",\n",
    "    alpha=0.25,\n",
    ")\n",
    "\n",
    "fig, ax = plt.subplots(figsize=figsize)\n",
    "\n",
    "sns.histplot(\n",
    "    x=tt.squeeze().numpy(), \n",
    "    weights=(ll(tt) / evidence).squeeze().numpy(),\n",
    "    color=color_truth,\n",
    "    label=\"Truth\",\n",
    "    **kwargs,\n",
    ")\n",
    "sns.histplot(\n",
    "    x=samples_nrec.squeeze().numpy(),\n",
    "    color=color_gilles,\n",
    "    label=\"NRE-C\",\n",
    "    **kwargs,\n",
    ")\n",
    "sns.histplot(\n",
    "    x=samples_durkan.squeeze().numpy(),\n",
    "    color=color_durkan,\n",
    "    label=\"NRE-B\",\n",
    "    **kwargs,\n",
    ")\n",
    "ax.legend()\n",
    "\n",
    "ax.set_ylabel(\"Posterior density\")\n",
    "ax.set_xlabel(\"Parameter \" + r\"$\\mathbf{\\theta}$\")\n",
    "fig.tight_layout()\n",
    "\n",
    "plt.savefig(\"importance-diagnostic-posterior.pdf\", dpi=300, bbox_inches=\"tight\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "af86b073-1419-43d2-a933-38b202c8dc23",
   "metadata": {},
   "source": [
    "## likelihood"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "2f2fd62b-717c-4404-bf9c-25c04a865e83",
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_weigh_function(density_estimator: torch.nn.Module, theta_o: torch.Tensor) -> Callable:\n",
    "    def weigh(x):\n",
    "        with torch.no_grad():\n",
    "            return density_estimator([theta_o.expand(x.shape[0], -1).to(device), x.to(device)]).exp().to(device=device)\n",
    "    return weigh"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "a438bd9f-24c1-4491-809b-73f2c18f8f78",
   "metadata": {},
   "outputs": [],
   "source": [
    "# nrea = get_weigh_function(density_estimator_nrea, true_theta)\n",
    "nrec = get_weigh_function(density_estimator_nrec, true_theta)\n",
    "nreb = get_weigh_function(density_estimator_nreb, true_theta)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "fc12ed48-6580-4686-8b27-ac397002ed67",
   "metadata": {},
   "outputs": [],
   "source": [
    "# # corner\n",
    "# fig = corner(\n",
    "#     x_likelihood.squeeze().numpy(),\n",
    "#     range=limits,\n",
    "# )\n",
    "# _ = corner(\n",
    "#     x_marginal.squeeze().numpy(),\n",
    "#     weights=nrea(x_marginal).squeeze().numpy(),\n",
    "#     color=\"red\",\n",
    "#     range=limits,\n",
    "#     fig=fig,\n",
    "# )\n",
    "# _ = corner(\n",
    "#     x_marginal.squeeze().numpy(),\n",
    "#     weights=nreb(x_marginal).squeeze().numpy(),\n",
    "#     color=\"blue\",\n",
    "#     range=limits,\n",
    "#     fig=fig,\n",
    "# )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "d1afbffc-1d21-43c4-a3db-6e67cad7c6fe",
   "metadata": {},
   "outputs": [],
   "source": [
    "# x_likelihood = simulator(true_theta.expand((num_posterior_samples, -1))).to(device)\n",
    "# x_marginal = simulator(prior.sample((num_posterior_samples,))).to(device)\n",
    "\n",
    "# fig, ax = plt.subplots(figsize=(7,7))\n",
    "# kwargs = dict(\n",
    "#     bins=1_000, \n",
    "#     binrange=limits.squeeze().tolist(),\n",
    "#     # kde=True, \n",
    "#     ax=ax,\n",
    "#     stat=\"density\",\n",
    "#     alpha=0.25,\n",
    "#     element=\"poly\",\n",
    "# )\n",
    "# sns.histplot(x=x_likelihood.squeeze().numpy(), color=\"orange\", **kwargs)\n",
    "# # sns.histplot(\n",
    "# #     x=x_marginal.squeeze().numpy(),\n",
    "# #     weights=task.log_likelihood(true_theta.expand(num_posterior_samples, -1).squeeze(), x_marginal.squeeze()).exp(),\n",
    "# #     color=\"red\", \n",
    "# #     **kwargs\n",
    "# # )\n",
    "# sns.histplot(\n",
    "#     x=x_marginal.squeeze().numpy(), \n",
    "#     weights=nrea(x_marginal).squeeze().numpy(), \n",
    "#     color=\"blue\", \n",
    "#     **kwargs\n",
    "# )\n",
    "# sns.histplot(\n",
    "#     x=x_marginal.squeeze().numpy(),\n",
    "#     weights=nreb(x_marginal).squeeze().numpy(),\n",
    "#     color=\"green\",\n",
    "#     **kwargs\n",
    "# )\n",
    "# plt.xlim(*limits.squeeze().tolist())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "f898fc06-2ec6-428a-9445-198dfa2c5696",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 366x366 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "to_check = torch.linspace(low.item() / 2, high.item() / 2, 3)\n",
    "fig, axes = plt.subplots(nrows=len(to_check), figsize=figsize, sharex=True, sharey=True)\n",
    "\n",
    "for i, (tt, ax) in enumerate(zip(to_check, axes.flatten())):\n",
    "    tt = tt.unsqueeze(-1)\n",
    "\n",
    "    x_likelihood = simulator(tt.expand((num_posterior_samples, -1))).to(device)\n",
    "    x_marginal = simulator(prior.sample((num_posterior_samples,))).to(device)\n",
    "    \n",
    "    kwargs = dict(\n",
    "        bins=1_000, \n",
    "        binrange=limits.squeeze().tolist(),\n",
    "        # kde=True, \n",
    "        ax=ax,\n",
    "        stat=\"density\",\n",
    "        alpha=0.25,\n",
    "        element=\"poly\",\n",
    "    )\n",
    "    sns.histplot(\n",
    "        x=x_likelihood.squeeze().numpy(), \n",
    "        color=color_truth, \n",
    "        label=\"Truth\",\n",
    "        **kwargs,\n",
    "    )\n",
    "    # sns.histplot(\n",
    "    #     x=x_marginal.squeeze().numpy(),\n",
    "    #     weights=task.log_likelihood(true_theta.expand(num_posterior_samples, -1).squeeze(), x_marginal.squeeze()).exp(),\n",
    "    #     color=\"red\", \n",
    "    #     **kwargs\n",
    "    # )\n",
    "    # sns.histplot(\n",
    "    #     x=x_marginal.squeeze().numpy(), \n",
    "    #     weights=get_weigh_function(density_estimator_nrea, tt)(x_marginal).squeeze().numpy(), \n",
    "    #     color=color_gilles, \n",
    "    #     **kwargs\n",
    "    # )\n",
    "    sns.histplot(\n",
    "        x=x_marginal.squeeze().numpy(), \n",
    "        weights=get_weigh_function(density_estimator_nrec, tt)(x_marginal).squeeze().numpy(), \n",
    "        label=\"NRE-C\",\n",
    "        color=color_gilles, \n",
    "        **kwargs,\n",
    "    )\n",
    "    sns.histplot(\n",
    "        x=x_marginal.squeeze().numpy(),\n",
    "        weights=get_weigh_function(density_estimator_nreb, tt)(x_marginal).squeeze().numpy(),\n",
    "        color=color_durkan,\n",
    "        label=\"NRE-B\",\n",
    "        **kwargs,\n",
    "    )\n",
    "    \n",
    "    if i == 0:\n",
    "        ax.legend()\n",
    "    \n",
    "    if i in [0, 2]:\n",
    "        ax.set_ylabel(\"\")\n",
    "    else:\n",
    "        ax.set_ylabel(\"Approximate likelihood density, various \" + r\"$\\mathbf{\\theta}$\")\n",
    "    \n",
    "    if i == 2:\n",
    "        ax.set_xlabel(\"Data $\\mathbf{x}$\")\n",
    "plt.xlim(*limits.squeeze().tolist())\n",
    "plt.tight_layout()\n",
    "\n",
    "plt.savefig(\"importance-diagnostic-likelihoods.pdf\", dpi=300, bbox_inches=\"tight\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6607d881-9236-489f-942b-04b9f9d8ebb2",
   "metadata": {},
   "source": [
    "## importance sampling diagnostic"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "905ffbcd-99fe-4182-9934-f35a48c317e6",
   "metadata": {},
   "outputs": [],
   "source": [
    "class WeightedLossClassifier(pl.LightningModule):\n",
    "    def __init__(self, weigh: Callable, ndim: int):\n",
    "        super().__init__()\n",
    "        self.weigh = weigh\n",
    "        self.classifier = torch.nn.Sequential(\n",
    "            torch.nn.Linear(ndim, 10*ndim),\n",
    "            torch.nn.ReLU(),\n",
    "            torch.nn.Linear(10*ndim, 10*ndim),\n",
    "            torch.nn.ReLU(),\n",
    "            torch.nn.Linear(10*ndim, 1),\n",
    "        )\n",
    "\n",
    "    def forward(self, x):\n",
    "        return self.classifier(x).sigmoid()\n",
    "\n",
    "    def _get_loss(self, batch, batch_idx):\n",
    "        x_ind, x_dep = batch\n",
    "        # ind goes to zero, dep goes to one\n",
    "        t = torch.cat(\n",
    "            [torch.zeros((x_ind.shape[0], 1), device=self.device), torch.ones((x_dep.shape[0], 1), device=self.device)],\n",
    "            dim=0,\n",
    "        )\n",
    "        y_ind = self.classifier(x_ind)\n",
    "        y_dep = self.classifier(x_dep)\n",
    "        y = torch.cat([y_ind, y_dep], dim=0)\n",
    "        w = torch.cat([self.weigh(x_ind), torch.ones((x_dep.shape[0], 1), device=self.device)], dim=0)\n",
    "        return F.binary_cross_entropy_with_logits(y, t, w)\n",
    "    \n",
    "    def training_step(self, batch, batch_idx):\n",
    "        loss = self._get_loss(batch, batch_idx)\n",
    "        self.log(\"train_loss\", loss)\n",
    "        return loss\n",
    "    \n",
    "    def validation_step(self, batch, batch_idx):\n",
    "        loss = self._get_loss(batch, batch_idx)\n",
    "        self.log(\"val_loss\", loss)\n",
    "\n",
    "    def configure_optimizers(self):\n",
    "        optimizer = torch.optim.Adam(self.parameters(), lr=1e-3)\n",
    "        return optimizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "5763b3e7-3aa1-48d0-bf79-21fce3e37005",
   "metadata": {},
   "outputs": [],
   "source": [
    "def importance_diagnostic(\n",
    "    X: torch.Tensor,\n",
    "    Y: torch.Tensor,\n",
    "    weigh: Callable,\n",
    "    seed: int = 1,\n",
    "    n_folds: int = 5,\n",
    "    scoring: str = \"accuracy\",\n",
    "    z_score: bool = True,\n",
    "    noise_scale: Optional[float] = None,\n",
    ") -> torch.Tensor:\n",
    "    \"\"\"X - independent, Y - dependent, classifier returns prob sample is dependent.\n",
    "    \n",
    "    Classifier-based 2-sample test returning accuracy\n",
    "\n",
    "    Trains classifiers with N-fold cross-validation [1]. Scikit learn MLPClassifier are\n",
    "    used, with 2 hidden layers of 10x dim each, where dim is the dimensionality of the\n",
    "    samples X and Y.\n",
    "\n",
    "    Args:\n",
    "        X: Sample 1\n",
    "        Y: Sample 2\n",
    "        seed: Seed for sklearn\n",
    "        n_folds: Number of folds\n",
    "        z_score: Z-scoring using X\n",
    "        noise_scale: If passed, will add Gaussian noise with std noise_scale to samples\n",
    "\n",
    "    References:\n",
    "        [1]: https://scikit-learn.org/stable/modules/cross_validation.html\n",
    "    \"\"\"\n",
    "    if z_score:\n",
    "        X_mean = torch.mean(X, dim=0)\n",
    "        X_std = torch.std(X, dim=0)\n",
    "        X = (X - X_mean) / X_std\n",
    "        Y = (Y - X_mean) / X_std\n",
    "\n",
    "    if noise_scale is not None:\n",
    "        X += noise_scale * torch.randn(X.shape)\n",
    "        Y += noise_scale * torch.randn(Y.shape)\n",
    "\n",
    "    X = X.cpu()\n",
    "    Y = Y.cpu()\n",
    "\n",
    "    ndim = X.shape[1]\n",
    "\n",
    "    # data = np.concatenate((X, Y))\n",
    "    # target = np.concatenate(\n",
    "    #     (\n",
    "    #         np.zeros((X.shape[0],)),\n",
    "    #         np.ones((Y.shape[0],)),\n",
    "    #     )\n",
    "    # )\n",
    "    \n",
    "    validation_percentage = 0.1\n",
    "    dataset = torch.utils.data.TensorDataset(X, Y)  # for this X is ind and Y is dep\n",
    "    train_dataset, test_dataset = torch.utils.data.random_split(\n",
    "        dataset, \n",
    "        [math.floor(len(dataset) * (1 - validation_percentage)), math.ceil(len(dataset) * validation_percentage)],\n",
    "    )\n",
    "    train_dataloader = torch.utils.data.DataLoader(train_dataset, batch_size=256, shuffle=True)\n",
    "    test_dataloader = torch.utils.data.DataLoader(test_dataset, batch_size=256)\n",
    "\n",
    "    clf = WeightedLossClassifier(weigh, ndim)\n",
    "    trainer = pl.Trainer(max_epochs=100)\n",
    "    trainer.fit(clf, train_dataloader, test_dataloader)\n",
    "    return clf\n",
    "    \n",
    "    # shuffle_k_fold = KFold(n_splits=n_folds, shuffle=True, random_state=seed)\n",
    "    # shuffle_k_fold.get_n_splits()\n",
    "    \n",
    "#     scores = cross_val_score(clf, data, target, cv=shuffle, scoring=scoring)\n",
    "\n",
    "#     scores = np.asarray(np.mean(scores)).astype(np.float32)\n",
    "#     return torch.from_numpy(np.atleast_1d(scores))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "af3e8731-24f2-45a8-aaa6-545ff79a7e6b",
   "metadata": {},
   "outputs": [],
   "source": [
    "def draw_ind_and_dep_samples(n: int, theta_o: torch.Tensor):\n",
    "    x_ind = simulator(prior.sample((n,)))\n",
    "    x_dep = simulator(theta_o.expand(n, -1))\n",
    "    return (x_ind, torch.zeros((n,))), (x_dep, torch.ones((n,)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "399b6f25-29ee-40dd-9f37-b2482c2fbb35",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "GPU available: True, used: False\n",
      "TPU available: False, using: 0 TPU cores\n",
      "IPU available: False, using: 0 IPUs\n",
      "HPU available: False, using: 0 HPUs\n",
      "/home/ben/mambaforge/envs/fixcnre/lib/python3.8/site-packages/pytorch_lightning/trainer/trainer.py:1812: PossibleUserWarning: GPU available but not used. Set `accelerator` and `devices` using `Trainer(accelerator='gpu', devices=3)`.\n",
      "  rank_zero_warn(\n",
      "\n",
      "  | Name       | Type       | Params\n",
      "------------------------------------------\n",
      "0 | classifier | Sequential | 141   \n",
      "------------------------------------------\n",
      "141       Trainable params\n",
      "0         Non-trainable params\n",
      "141       Total params\n",
      "0.001     Total estimated model params size (MB)\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Sanity Checking: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/ben/mambaforge/envs/fixcnre/lib/python3.8/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:240: PossibleUserWarning: The dataloader, val_dataloader 0, does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` (try 24 which is the number of cpus on this machine) in the `DataLoader` init to improve performance.\n",
      "  rank_zero_warn(\n",
      "/home/ben/mambaforge/envs/fixcnre/lib/python3.8/site-packages/pytorch_lightning/trainer/connectors/data_connector.py:240: PossibleUserWarning: The dataloader, train_dataloader, does not have many workers which may be a bottleneck. Consider increasing the value of the `num_workers` argument` (try 24 which is the number of cpus on this machine) in the `DataLoader` init to improve performance.\n",
      "  rank_zero_warn(\n",
      "/home/ben/mambaforge/envs/fixcnre/lib/python3.8/site-packages/pytorch_lightning/trainer/trainer.py:1927: PossibleUserWarning: The number of training batches (36) is smaller than the logging interval Trainer(log_every_n_steps=50). Set a lower value for log_every_n_steps if you want to see logs for the training epoch.\n",
      "  rank_zero_warn(\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ef8bda4780ad4d3cb8c3b1b57cf595a2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Training: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "368bd5c4a5954398a594afba42d8d6af",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b963571e230741658e092d11b94ae210",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9c3de621d95a460f8b1f152d9c0e1bd5",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "463842614c234410a8ae2545c1a9b9ba",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b13838d34781447c851ba006f4fb1094",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0586b4153e2841b6be9ef44de48146a6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a825f508f0064fd88b71933c1257c32e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "cf8e2a81d7314dc3b779eba3e84d6abf",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b4cb0667999c41ba964c22443760ef66",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "969d8b79d53545db820d13ee2836f7cd",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "72286510350c482694f9adac9256604a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a4e4d07f4971482abb59fda4f0905684",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "e85e5cb2a7d447048ad661b5d68589a4",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c543b2b238344573b4c2a1922b777a2d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4fd0a16f76a44645b0870487e40440fa",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "edbdf93538cc49f2940f7aaf85be12d7",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d4986e3a345c470093ec8e65914e1f90",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8bb824a7620f4a38874c66c11bd0b0d7",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "51ef1230fae64859845f512f14819ed4",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "bb7a73d4b9c4400da96a7d43bf972cf9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4c0093fbbbcf4a6aa35105065ce29786",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "93f92ca3d9684662a5ff18b2d71d469d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ae4370da905b45b1b3aa19efe17c45b6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "979112cceb1e4b7e9a1e33181207ca8b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d3e365d3534e45a88c51c3d1691ee041",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d2d38445e54d4af6bb4757df7cc4e74e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4729990fa2b045438ba86d66c550b390",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "feb8e639befd4b0fad86f43cbbafbc09",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1dadf7bf97054220bf1e5acbdc3a732d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c097eccc1dad4d7c8136e11b1933e5c2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "637dbc507f3a4fdba763f3323c7934fe",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b66b1860bdc340298282ae6dad8627de",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1cfb2fbb5cbf4b63b835d5ce060b4c62",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "864650af1c2241a09349a50730cfe3d6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "916a34e5de1c4385965f5fa9258f43e9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a2e5b3a9b91845158cb4996561e3c6f2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "18d7998157384591bc6324d722a665f8",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "e8482c0d8a464211847a893a37ecc35a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f2d99013628b4b9f8aa55f17f313dfc0",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "dbd61c4f37b9495a8e151f5d67411104",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1defda2036b8457392c804c6b5da4a27",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9a82b6fb382f4d989821d1d9a0a63367",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "363f12f1d7bf4f1595da76bd0f642ddb",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "891b2a44351046799f39c6d2f1890945",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6e5bd90751a34203983a89716dbead2c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "540138c68f04403290877fb7c9bf68ec",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "70e9992ca1c44e3b88211f2b44989dc0",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c2e0ae096e514834a0b57a1647da9063",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "767ffcdfbd0d43d39567b206d3e461b6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0fc242566e1049d093bd1f48a027eec4",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b26d330643b54a99a1fd86fd9962121c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "242050f62b8145298d287e7d08afc390",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "516242ff80ff4ba3bb673a3851080ef4",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "e057b256b82e493da877a83d02f37efd",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "(x_ind, _), (x_dep, _) = draw_ind_and_dep_samples(10_000, true_theta)\n",
    "# aaa = importance_diagnostic(x_ind, x_dep, nrea)\n",
    "ccc = importance_diagnostic(x_ind, x_dep, nrec)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "6737bf43-7973-4372-b002-f672acfe207b",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "GPU available: True, used: False\n",
      "TPU available: False, using: 0 TPU cores\n",
      "IPU available: False, using: 0 IPUs\n",
      "HPU available: False, using: 0 HPUs\n",
      "\n",
      "  | Name       | Type       | Params\n",
      "------------------------------------------\n",
      "0 | classifier | Sequential | 141   \n",
      "------------------------------------------\n",
      "141       Trainable params\n",
      "0         Non-trainable params\n",
      "141       Total params\n",
      "0.001     Total estimated model params size (MB)\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Sanity Checking: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "2e69cd27811b4c499a3631c1a1cb420b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Training: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "369de72b727d441a8a3a1ff32f5979f4",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "aba68adc16c4403b8b7f7cf796400de9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b6db837b3f6a4adea04332899dc3c93e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ef32fd33af7a4594854c1083860ff160",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "3a2f3c7c682144e390dba4842b35512e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "59c83fcb30474644b390f74960e0d42e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "436245d1543a435da26162821133a6fe",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5e2a24dc782a433b8f8c16d36678cd67",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7db4be572a3c4aaeb273abb8a29bbbca",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "33c78f200c2b45a9ac057f337d67c03e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d67abf7cea3240fdbceadf2b0d502a97",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5f1303fb359c44ec919180af0e1f211b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "301c34bda5cb4ee29ebc5d01f470fd95",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6c8ffe39321b46f1916e70619913fbb2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "50d2e6edd7fa4b009b410766860cca21",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ef849ea634f44840aa40d42ffedd292a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "235e0c7e618a452eb6be3acf700c7e5d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "397e568065fb44e3a407a1d65b903477",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "2b2a7669baed4fb0b59994ae7da52ab3",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "db3083cf3fec4acc9657307414ab5c60",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "fff32f17a47744aeaa206abd889662ef",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "2f1da68e07ca4ae9b85b6f795f7167f1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "524e5be3b01548279375d158be8b4ec5",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8becd9ad629e48bf8d86f42db7d46ae1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d90e2d8677914e35a5cc1015774dce9a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7b0be7cb50c441bea96007cb0ff994ed",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6ecd2955fba743edbf4cfa406758f618",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c88a38c3682d48cda039497b89b50435",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "208b53504d98469fa5c0d10813583be9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "19394f0b7c614841b0bad51dd17a8383",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0f5454f9447f40f3beb778ab4523a91a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5a9c3756b0d24002bf2b03bc3bb9709a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8a2329357e014fb29aff2a0d3e8b56f3",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8252d163b0ea4b0c987e39e14a94d3d1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "957d3561362b450aa9c8812dba0101b7",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d32a01bdab6e48238bfe3ca7d338b0fd",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a11f693cc58d40719e550bef4f5093ef",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a68b0e1217524cb9838713ceda0f375a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ac28f3194ccc43d6bef8280deaf755cf",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "da333febc7ac43979e851f568a6b2332",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d2d19edf10bc46e0ba738cbfd08cfd47",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4aa94bcd3888495aaee71e7258117d16",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5263039a9dd14c3e8f9620d03087f0f6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c2fe89516ec04628991acffe477c3b6a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "373fcf64ca0b448ebbf4b89929f86e92",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "90a1419929224087acf55dcdce395d1a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5fc35686f595442bb171bd3e54bbdfdc",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "2acd8c9dda664e5bad30684f94de9b14",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0ce5d85ae2714678850a0d4f70a21f0c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0d920b3b97c146de8dd7c988e6d2c028",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d4bd60bceb9840aaa716bbccb686979b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "135e1245faaa4987adda47798431d04a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "98cfafae6786443eb943e7deaab5cff4",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Validation: 0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "bbb = importance_diagnostic(x_ind, x_dep, nreb)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "db97c903-ef12-4032-becb-366e6fb368b5",
   "metadata": {},
   "source": [
    "#### weighted ROC\n",
    "https://mran.microsoft.com/snapshot/2017-06-26/web/packages/WeightedROC/vignettes/Definition.pdf  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "dd82b940-4ad0-4e5d-a04c-836bbc3921fe",
   "metadata": {},
   "outputs": [],
   "source": [
    "(x_ind, y_ind), (x_dep, y_dep) = draw_ind_and_dep_samples(num_posterior_samples, true_theta)\n",
    "x = torch.cat([x_ind, x_dep])\n",
    "y = torch.cat([y_ind, y_dep])\n",
    "# w_a = torch.cat([nrea(x_ind), torch.ones(x_dep.shape[0], 1)])\n",
    "w_c = torch.cat([nrec(x_ind), torch.ones(x_dep.shape[0], 1)])\n",
    "w_b = torch.cat([nreb(x_ind), torch.ones(x_dep.shape[0], 1)])\n",
    "\n",
    "with torch.no_grad():\n",
    "    fpr_c, tpr_c, thresholds_c = sklearn.metrics.roc_curve(y, ccc(x).squeeze().sigmoid(), sample_weight=w_c.squeeze())\n",
    "    fpr_b, tpr_b, thresholds_b = sklearn.metrics.roc_curve(y, bbb(x).squeeze().sigmoid(), sample_weight=w_b.squeeze())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "0babe2c7-958e-4ffe-89c3-df833b513a70",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 366x366 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(figsize=figsize)\n",
    "lw = 2\n",
    "plt.plot([0, 1], [0, 1], color=color_truth, lw=lw, linestyle=\"--\", label=\"Diagonal\")\n",
    "plt.plot(\n",
    "    fpr_c,\n",
    "    tpr_c,\n",
    "    color=color_gilles,\n",
    "    lw=lw,\n",
    "    label=\"NRE-C\"\n",
    "    # label=\"ROC curve (area = %0.2f)\" % roc_auc[2],\n",
    ")\n",
    "plt.plot(\n",
    "    fpr_b,\n",
    "    tpr_b,\n",
    "    color=color_durkan,\n",
    "    lw=lw,\n",
    "    label=\"NRE-B\"\n",
    "    # label=\"ROC curve (area = %0.2f)\" % roc_auc[2],\n",
    ")\n",
    "plt.xlim([0.0, 1.0])\n",
    "plt.ylim([0.0, 1.05])\n",
    "plt.xlabel(\"False Positive Rate\")\n",
    "plt.ylabel(\"True Positive Rate\")\n",
    "# plt.title(\"Receiver operating characteristic\")\n",
    "plt.legend(loc=\"lower right\")\n",
    "plt.tight_layout()\n",
    "\n",
    "plt.savefig(\"importance-diagnostic-roc.pdf\", dpi=300, bbox_inches=\"tight\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b0ad6a23-690e-49f6-8e13-aff682542c8c",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.19"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}