Explainable_OoD_detection_on_SNNs.ipynb

{"cells":[{"cell_type":"markdown","source":["# Instructions\n","---\n","This is the code for the paper *A Novel Explainable Out-of-Distribution Detection Approach for Spiking Neural Networks*, available as a preprint in https://arxiv.org/abs/2210.00894.\n","\n","First of all, run all the prerequisites. They are the installation and import of the needed packages, the definition of the functions and models of SNNs and several more details to make the code work. \n","\n","Then, if you want to run the benchmarks presented in the paper, run the _**Automated benchmark**_ section. Further instructions are presented in this section. For obtaining the figures of the paper of the Bayesian Analysis and CD Graph, a section is also provided after the benchmark.\n","\n","For following the step-by-step process, the other sections are provided. First the In- and Out-of-Distribution datasets must be selected, as well as the model to test. In the section _**Spike Counts Pattern based Detector (SCPD)**_ all the code to execute the detector is available.\n","\n","In the _**OoD detection benchmak**_ section, the results for the selected dataset and model for the other methods can be checked.\n","\n","Finally, in _**Explainability**_ the heatmap generation code is provided."],"metadata":{"id":"rkRz53tRGKNu"}},{"cell_type":"markdown","source":["# Prerequisites"],"metadata":{"id":"tnAt_S0yIOwB"}},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"wqCrVb8-V8ar"},"outputs":[],"source":["#@title #Installation of Norse and imports\n","!pip install --quiet norse\n","\n","# Install the libraries for the Bayesian tests and CD graph\n","!pip install --quiet Orange3 \n","!git clone https://github.com/BayesianTestsML/tutorial.git\n","import tutorial.Python.bayesiantests as bt\n","\n","import os\n","from pathlib import Path\n","import logging\n","import sys\n","from collections import OrderedDict\n","import datetime\n","import pytz\n","\n","import torch\n","import torchvision\n","from torchvision.datasets.vision import VisionDataset\n","import numpy as np\n","import pandas as pd\n","import matplotlib.pyplot as plt\n","import matplotlib.image as mpimg\n","from sklearn.metrics import pairwise_distances\n","from tqdm.notebook import tqdm, trange\n","\n","from norse.torch import ConstantCurrentLIFEncoder\n","from norse.torch import SpikeLatencyLIFEncoder\n","from norse.torch import PoissonEncoder\n","from norse.torch import LIFParameters, LIFState\n","from norse.torch.module.lif import LIFCell, LIFRecurrentCell\n","\n","from google.colab import files\n","\n","import ipywidgets as ipw\n","from IPython.core.display import HTML, display, clear_output\n","from ipywidgets import widgets,interact, interactive, fixed, interact_manual\n","\n","from scipy import spatial\n","import scipy.special as scysp\n","from sklearn.cluster import AgglomerativeClustering\n","from sklearn.metrics import silhouette_score\n","from sklearn.metrics import pairwise_distances\n","from skimage.transform import resize\n","from sklearn.utils import shuffle as skl_shuffle"]},{"cell_type":"markdown","metadata":{"id":"bRstJrMldjZV"},"source":["## Functions"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"hFGPVssSheYh"},"outputs":[],"source":["#@markdown ##Constants\n","\n","# Path for the datasets\n","root_path = Path(Path.cwd())\n","datasets_path = root_path / 'Datasets'\n","pretrained_weights_path = root_path / 'Pretrained_weights'\n","figures_path = root_path / 'Figures'\n","\n","# Create the directories\n","datasets_path.mkdir(exist_ok=True)\n","pretrained_weights_path.mkdir(exist_ok=True)\n","figures_path.mkdir(exist_ok=True)\n","\n","\n","# Constants for the correct visualization of the .csv in Spanish configuration\n","CSV_SEPARATOR = ';'\n","CSV_DECIMAL = ','\n","\n","# For the explainability part\n","WHICH_FPR = 80 "]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"E6nQ_IL9vyKa"},"outputs":[],"source":["#@markdown ##Utils\n","\n","def searchIndicesOfClass(searchedClass, labelsArray, n=0 ,posIni = 0):\n","  '''\n","  Function that outputs a list with all the indices of the searchedClass in the labelsList \n","  If n is provided, only the first n coincidences are outputted\n","  If posIni is provided, the search starts by this position.\n","  searchedClass, n, posIni -> integer\n","  labelsArray -> array\n","  '''\n","  indices = []\n","  if n is 0:\n","    # Case of searching for all the array\n","    for index, labels in enumerate(labelsArray[posIni:]):\n","      if labels == searchedClass:\n","        indices.append(posIni+index)\n","  else: \n","    # Case of searching only n number of indices\n","    i = 0\n","    for index, labels in enumerate(labelsArray[posIni:]):\n","      if i >= n: \n","          break\n","      if labels == searchedClass:\n","        indices.append(posIni+index)\n","        i += 1\n","  return indices\n","\n","\n","def indices_of_every_class_for_subset(\n","    train_loader, n_samples_per_class, dataset_name, init_pos=0):\n","  n_classes = len(train_loader.dataset.classes)\n","  selected_indices_per_class = []\n","  for cl_index in range(n_classes):\n","    # The +1 is beacuse the targets transformation does not happen\n","    # when accesing the dataset object, only in the dataloader\n","    if dataset_name == 'Letters':\n","      cl_index = cl_index + 1\n","    indices = searchIndicesOfClass(\n","        cl_index, \n","        train_loader.dataset.targets, \n","        n=n_samples_per_class,\n","        posIni=init_pos)\n","    selected_indices_per_class += indices\n","  return selected_indices_per_class\n","\n","\n","def average_per_class_and_cluster(spikeFrencuencies, preds, \n","                                  clusterizationPerClass, option='median', \n","                                  n_samples=None):\n","  '''\n","  Function that receives the counts and the predictions of a subset and the \n","  clusterization objects per class of that subset and outputs a list of arrays \n","  with the mean or median hidden frecuency vector of each cluster of each class\n","  '''\n","  # For every class\n","  avgPerClass = []\n","  for class_index in range(len(train_data.classes)):\n","    # Calculation of the array of frecuencies that corresponds to a class\n","    if n_samples is None:\n","      indices = searchIndicesOfClass(class_index, preds)\n","    else:\n","      indices = searchIndicesOfClass(class_index,preds,n_samples)\n","    spikesFrecsOneClass = spikeFrencuencies[indices]\n","    # For every cluster of the class, compute the median\n","    avgPerCluster = []\n","    for cluster_index in np.unique(clusterizationPerClass[class_index].labels_):\n","      # We can compute the mean or the median of the neuron frecuencies for that cluster\n","      if option == 'median': \n","        avgPerCluster.append(np.median(spikesFrecsOneClass[np.where(clusterizationPerClass[class_index].labels_ == cluster_index)[0]],axis=0))\n","      elif option == 'mean':\n","        avgPerCluster.append(np.mean(spikesFrecsOneClass[np.where(clusterizationPerClass[class_index].labels_ == cluster_index)[0]],axis=0))\n","    avgPerCluster = np.array(avgPerCluster)\n","    # Save all the cluster averages to the list of the classes\n","    avgPerClass.append(avgPerCluster)\n","  return avgPerClass\n","\n","\n","def distance_to_clusters_averages(spikeFrencuencies, predictions, avgPerClass):\n","  '''\n","  Function that computes the distance of the introduced array to the \n","  cluster's averages of the predicted class\n","  Takes the frecuencies and the predictions of each sample of a subset (not ordered by class) and the averages of each cluster and class\n","  Outputs the distance of each sample for each class\n","  # TODO: Describe the dimensions of the inputs\n","  '''\n","  # Order array by predicted class\n","  spikeFrecuencies_per_class = []\n","  for class_index in range(len(train_data.classes)):\n","    spikeFrecuencies_per_class.append(spikeFrencuencies[searchIndicesOfClass(class_index,predictions)])\n","  # Compute the pairwise distances per predicted class\n","  distances_per_class = []\n","  closest_clusters_per_class = []\n","  for class_index, frecs_one_class in enumerate(spikeFrecuencies_per_class):\n","    if frecs_one_class.size == 0:\n","      distances_per_class.append([])\n","      closest_clusters_per_class.append([])\n","    else:\n","      parwiseDist = pairwise_distances(frecs_one_class, avgPerClass[class_index],metric='manhattan')\n","      distances_per_class.append(np.min(parwiseDist, axis=1))\n","      closest_clusters_per_class.append(np.argmin(parwiseDist, axis=1))\n","  return distances_per_class, closest_clusters_per_class\n","\n","\n","def parse_size_of_dataloader(dataloader_obj: torch.utils.data.DataLoader, batch_size):\n","  if len(dataloader_obj.dataset) > 10000:\n","    subset = torch.utils.data.Subset(dataloader_obj.dataset, [x for x in range(10000)])\n","    subset_loader = torch.utils.data.DataLoader(subset, batch_size=batch_size, shuffle=False)\n","    return subset_loader\n","\n","\n","def isolate_or_remove_mnistc(df, option):\n","  if option == 'Isolate':\n","    df = df.loc[\n","      df['Out-Distribution'].apply(lambda x: x.split('/')[0]) == 'MNIST-C'\n","    ]\n","  \n","  elif option == 'Remove' :\n","    df = df.loc[\n","      df['Out-Distribution'].apply(lambda x: x.split('/')[0]) != 'MNIST-C'\n","    ]\n","\n","  return df\n","\n","\n","def isolate_model(df, option):\n","  df = df.loc[df['Model'] == option]\n","  if df.size == 0:\n","    raise ValueError('Empty dataframe, probably wrong model name selected')\n","\n","  return df\n","\n","def print_created_clusters_per_class(clusters_per_class):\n","  print('')\n","  print('Created clusters:')\n","  print('-'*75)\n","  # Printing the cluster composition for each class\n","  for cl_ind, clusters_one_class in enumerate(clusters_per_class):\n","    unique, counts = np.unique(clusters_one_class.labels_, return_counts=True)\n","    print('Clase',train_data.classes[cl_ind].ljust(15),'\\t',dict(zip(unique, counts)))\n","    print('-'*75)"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"G2fiLI__C9c9"},"outputs":[],"source":["#@markdown ##Cluster creation\n","def create_clusters(subset_train_loader_clusters, preds_train_clusters,\n","                    spk_count_train_clusters,\n","                    size=1000, distance_for_clustering=None, verbose=2):\n","  \"\"\"\n","  Verbose = 0 -> No prints and plots neither loggin info\n","  verbose = 1 -> Returns loggin info only\n","  Verbose = 2 -> Prints and plots\n","  \"\"\"\n","  # Select a distance threshold for each class\n","  if distance_for_clustering is None:\n","    distance_for_clustering = (800,3000)\n","  opt_dist_thr_per_class = []\n","  opt_silh_score_values_per_class = []\n","  dist_thrs          = np.linspace(distance_for_clustering[0], \n","                                   distance_for_clustering[1],50)\n","  silhScoresPerClass = []\n","  clusterLabels = []\n","\n","  n_classes = len(subset_train_loader_clusters.dataset.dataset.classes)\n","  for class_index in tqdm(range(n_classes), desc='Computing silhuette score for various distance thresholds'):\n","    dunnIndexes = []\n","    silh_scores = []\n","    for dist in dist_thrs:\n","      indices = searchIndicesOfClass(class_index, preds_train_clusters, size)\n","      cluster_model = AgglomerativeClustering(n_clusters=None,affinity='manhattan',linkage='average',distance_threshold=dist)\n","      try: # Handle the case that one class has no representation in the training samples\n","        cluster_model.fit(spk_count_train_clusters[indices])\n","        clusterLabels.append(cluster_model.labels_)\n","      except ValueError as e:\n","        print('Error probably caused by the lack of training samples for one specific class')\n","        raise(e)\n","      try:\n","        silh_scores.append(silhouette_score(spk_count_train_clusters[indices], cluster_model.labels_, metric='manhattan'))\n","      except ValueError:\n","        silh_scores.append(0)\n","    silhScoresPerClass.append(silh_scores)\n","\n","    # Iterate the inverted to catch the smallest distance value with the \n","    # greatest silhouette score\n","    max_score = 0\n","    max_index = 0\n","    for idx, current_score in enumerate(silh_scores):\n","      # Store the greatest value we encounter traveling the curve\n","      # Only update the value if it is greater, not if it equal\n","      if current_score > max_score: \n","        max_index = idx\n","        max_score = current_score\n","    # We append the distance treshold to a list where they are going to be\n","    # stored, one for each class\n","    opt_dist_thr_per_class.append(dist_thrs[max_index])\n","    opt_silh_score_values_per_class.append(silh_scores[max_index])\n","  \n","  # Plot the silhouette score for every distance threshold\n","  if verbose == 2:\n","    # Plot to see the silhouette scores\n","    print('Selected distance thresholds:\\n', opt_dist_thr_per_class)\n","    if n_classes == 10:\n","      fig, axes = plt.subplots(2,5,figsize=(6*n_classes/2, 12))\n","    elif n_classes == 26:\n","      fig, axes = plt.subplots(2,13,figsize=(6*n_classes/2, 12))\n","    else:\n","      raise NameError(f'The number of classes {n_classes} is not implemented for the plots')\n","\n","    for class_index, ax in enumerate(axes.flat):\n","      ax.plot(dist_thrs,silhScoresPerClass[class_index], color='blue')\n","      ax.plot(opt_dist_thr_per_class[class_index], opt_silh_score_values_per_class[class_index], 'ro')\n","      ax.set_title(subset_train_loader_clusters.dataset.dataset.classes[class_index])\n","    plt.savefig('silhouetteScores.pdf')\n","\n","  # Create the clusters by extracting the labels for every sample\n","  clusters_per_class = []\n","  for class_index in range(n_classes):\n","    indices = searchIndicesOfClass(class_index, preds_train_clusters, 1000)\n","    if isinstance(opt_dist_thr_per_class, list):\n","      cluster_model = AgglomerativeClustering(n_clusters=None,affinity='manhattan',linkage='complete',distance_threshold=opt_dist_thr_per_class[class_index])\n","    else:\n","      cluster_model = AgglomerativeClustering(n_clusters=None,affinity='manhattan',linkage='complete',distance_threshold=opt_dist_thr_per_class)\n","    \n","    cluster_model.fit(spk_count_train_clusters[indices])\n","    # Save the cluster models\n","    clusters_per_class.append(cluster_model)\n","\n","  if verbose == 2:\n","    # Plot the top three levels of the dendrogram\n","    if n_classes == 10:\n","      fig, axes = plt.subplots(2,5,figsize=(6*n_classes/2, 12))\n","    elif n_classes == 26:\n","      fig, axes = plt.subplots(2,13,figsize=(6*n_classes/2, 12))\n","    else:\n","      raise NameError(f'The number of classes {n_classes} is not implemented for the plots')\n","    fig.suptitle('Hierarchical Clustering Dendrogram', fontsize=22, y=0.94)\n","    #fig.supxlabel('X axis: Number of points in node (index of the number if not in parenthesis)',fontsize = h + w*0.1,y=0.065)\n","  \n","    for class_index, ax in tqdm(enumerate(axes.flat), desc='Create the clusters with the selected distance thresholds'):\n","      plot_dendrogram(cluster_model, truncate_mode='level', p=3, ax=ax)\n","      ax.set_title('Class {}'.format(subset_train_loader_clusters.dataset.dataset.classes[class_index]),fontsize=22)\n","      #ax[i,j].set_xlabel(\"Number of points in node\",fontsize=h)\n","\n","    plt.savefig(f'DendrogramPerClass.pdf')\n","    fig.show()\n","\n","    print_created_clusters_per_class(clusters_per_class)\n","\n","  if verbose == 1:\n","    string_for_logger = 'Created clusters:\\n' + '-'*75 + '\\n'\n","    for class_index in range(len(train_data.classes)):\n","      unique, counts = np.unique(clusters_per_class[class_index].labels_, return_counts=True)\n","      string_for_logger += f'Clase {train_data.classes[class_index].ljust(15)} \\t {dict(zip(unique, counts))}\\n' + '-'*75 + '\\n'\n","\n","    return clusters_per_class, string_for_logger\n","  \n","  return clusters_per_class"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"gVrOIeH7O4DM"},"outputs":[],"source":["#@markdown ##Plots\n","def plot_loss_history(training_losses, test_losses):\n","  fig = plt.figure(facecolor=\"w\", figsize=(10, 5))\n","  plt.plot(training_losses)\n","  plt.plot(test_losses)\n","  plt.title(\"Loss Curves\")\n","  plt.legend([\"Train Loss\", \"Test Loss\"])\n","  plt.xlabel(\"Iteration\")\n","  plt.ylabel(\"Loss\")\n","  plt.show()\n","\n","\n","from scipy.cluster.hierarchy import dendrogram\n","def plot_dendrogram(model, **kwargs):\n","    # Create linkage matrix and then plot the dendrogram\n","\n","    # create the counts of samples under each node\n","    counts = np.zeros(model.children_.shape[0])\n","    n_samples = len(model.labels_)\n","    for i, merge in enumerate(model.children_):\n","        current_count = 0\n","        for child_idx in merge:\n","            if child_idx < n_samples:\n","                current_count += 1  # leaf node\n","            else:\n","                current_count += counts[child_idx - n_samples]\n","        counts[i] = current_count\n","\n","    linkage_matrix = np.column_stack([model.children_, model.distances_,\n","                                      counts]).astype(float)\n","\n","    # Plot the corresponding dendrogram\n","    dendrogram(linkage_matrix, **kwargs)\n","\n","def plot_AUROC(fpr, tpr):\n","  # AUC\n","  auc = np.trapz(tpr, fpr)\n","  # Plot\n","  plt.figure(figsize=(15,12))\n","  roc = plt.plot(fpr,tpr,label='ROC curve', lw=3)\n","  rnd_roc = plt.plot(np.linspace(0,1,100),np.linspace(0,1,100),'k--', label='Random ROC curve')\n","  plt.xticks(fontsize=15)\n","  plt.yticks(fontsize=15)\n","  plt.xlabel('FPR',fontsize=20)\n","  plt.ylabel('TPR',fontsize=20)\n","  plt.title('ROC curve, AUC = %.3f'%auc,fontsize=25,pad=10)\n","  plt.fill_between(fpr, tpr, alpha=0.3)\n","  # Create empty plot with blank marker containing the extra label\n","  plt.plot([], [], ' ', label='FPR at 95% TPR = {}%'.format(round(fpr[95]*100,2)))\n","  plt.plot([], [], ' ', label='FPR at 80% TPR = {}%'.format(round(fpr[80]*100,2)))\n","  plt.legend(fontsize=20,loc='upper left')\n","\n","\n","def plot_AUPR(precision, tpr):\n","  # AUPR\n","  auc = np.trapz(precision, tpr)\n","  # Plot\n","  plt.figure(figsize=(15,12))\n","  roc = plt.plot(tpr, precision, label='ROC curve', lw=3)\n","  plt.xticks(fontsize=15)\n","  plt.yticks(fontsize=15)\n","  plt.xlabel('FPR',fontsize=20)\n","  plt.ylabel('Precision',fontsize=20)\n","  plt.title('PR curve, AUC = %.3f'%auc,fontsize=25,pad=10)\n","  plt.fill_between(tpr, precision, alpha=0.3)\n","  plt.legend(fontsize=20,loc='upper left')\n","\n","\n","def plot_histogram(train, test, ood):\n","  fig = plt.figure(figsize=(10,5),tight_layout=True)\n","  plt.hist(train,bins=50,color='blue',alpha=0.6,density=True, label='Train')\n","  plt.hist(test,bins=50,color='green',alpha=0.6,density=True, label='Test')\n","  plt.hist(ood,bins=50,color='darkorange',alpha=0.6,density=True, label='ood')\n","  #plt.ylim([0,10])\n","  plt.legend(fontsize=18)\n","  plt.show()\n","\n","\n","def plot_ax(ax, img, plt_range, cmap, alpha=1, title=None, fontsize=8, xlabel=None):\n","  ax.set_xticks([])\n","  ax.set_yticks([])\n","  im = ax.imshow(img, vmin=plt_range[0],vmax=plt_range[1], cmap=cmap, alpha=alpha)\n","  if title is not None:\n","    ax.set_title(title, fontsize=fontsize)\n","  if xlabel is not None:\n","    ax.set_xlabel(xlabel, fontsize=fontsize)\n","  return im"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"y3bsHgafNv8i"},"outputs":[],"source":["#@markdown ##Metrics\n","\n","### Common for all approaches ###\n","def tp_fn_fp_tn_computation(in_or_out_distribution_per_tpr):\n","  '''\n","  Function that creates an array with the number of values of tp and fp or fn and tn, depending on if the \n","  passed array is InD or OD.\n","  :in_or_out_distribution_per_tpr: array with True if predicted InD and False if predicted OD, for each TPR\n","  ::return: array with shape (tpr, 2) with the 2 dimensions being tp,fn if passed array is InD, and fp and tn if the passed array is OD\n","  '''\n","  tp_fn_fp_tn = np.zeros((len(in_or_out_distribution_per_tpr),2),dtype='uint16')\n","  length_array = in_or_out_distribution_per_tpr.shape[1]\n","  for index, element in enumerate(in_or_out_distribution_per_tpr):\n","    n_True = int(len(element.nonzero()[0]))\n","    tp_fn_fp_tn[index,0] = n_True\n","    tp_fn_fp_tn[index,1] = length_array - n_True\n","  return tp_fn_fp_tn\n","\n","### Metrics for Distances per class approach ###\n","\n","def thresholds_per_class_for_each_TPR(dist_per_class):\n","  # Creation of the array with the thresholds for each TPR (class, dist_per_TPR)\n","  sorted_distances_per_class = [np.sort(x) for x in dist_per_class]\n","  tpr_range = np.arange(0,1,0.01)\n","  tpr_range[-1] = 0.99999999 # For selecting the last item correctly\n","  distance_thresholds_test = np.zeros((len(train_data.classes), len(tpr_range)))\n","  for class_index in range(len(train_data.classes)):\n","    for index, tpr in enumerate(tpr_range):\n","      distance_thresholds_test[class_index, index] = sorted_distances_per_class[class_index][int(len(sorted_distances_per_class[class_index])*tpr)]\n","  return distance_thresholds_test\n","\n","def compare_distances_per_class_to_distance_thr_per_class(distances_list_per_class, thr_distances_array):\n","  '''\n","  Function that creates an array of shape (tpr, InD_or_OD), where tpr has the lenght of the number of steps of the TPR list\n","  and second dimensions has the total lenght of the distances_list_per_class, and cotains True if its InD and False if is OD\n","  :distances_list_per_class: list with each element being an array with the distances to avg clusters of one class [array(.), array(.)]\n","  :thr_distances_array: array of shape (class, dist_for_each_tpr), where first dimension is the class and the second is the distance for the TPR\n","   corresponding to that position. For example, the TPR = 0.85 corresponds to the 85th position.\n","  '''\n","  in_or_out_distribution_per_tpr = np.zeros((len(np.transpose(thr_distances_array)),len(np.concatenate(distances_list_per_class))),dtype=bool)\n","  for tpr_index ,thr_distances_per_class in enumerate(np.transpose(thr_distances_array)):\n","      in_or_out_distribution_per_tpr[tpr_index] = np.concatenate([dist_one_class < thr_distances_per_class[cls_index] for cls_index, dist_one_class in enumerate(distances_list_per_class)])\n","    \n","  return in_or_out_distribution_per_tpr\n","\n","def compute_precision_tpr_fpr_for_test_and_ood(dist_test_per_class, dist_ood_per_class,dist_thresholds):\n","  # Creation of the array with True if predicted InD (True) or OD (False)\n","  in_or_out_distribution_per_tpr_test = compare_distances_per_class_to_distance_thr_per_class(dist_test_per_class, dist_thresholds)\n","  in_or_out_distribution_per_tpr_test[0] = np.zeros((in_or_out_distribution_per_tpr_test.shape[1]),dtype=bool) # To fix that one element is True when TPR is 0\n","  in_or_out_distribution_per_tpr_test[-1] = np.ones((in_or_out_distribution_per_tpr_test.shape[1]),dtype=bool) # To fix that last element is True when TPR is 1\n","  in_or_out_distribution_per_tpr_ood = compare_distances_per_class_to_distance_thr_per_class(dist_ood_per_class, dist_thresholds)\n","\n","  # Creation of arrays with TP, FN and FP, TN\n","  tp_fn_test = tp_fn_fp_tn_computation(in_or_out_distribution_per_tpr_test)\n","  fp_tn_ood = tp_fn_fp_tn_computation(in_or_out_distribution_per_tpr_ood)\n","\n","  # Computing TPR, FPR and Precision\n","  tpr_values = tp_fn_test[:,0] / (tp_fn_test[:,0] + tp_fn_test[:,1])\n","  fpr_values = fp_tn_ood[:,0] / (fp_tn_ood[:,0] + fp_tn_ood[:,1])\n","  precision  = tp_fn_test[:,0] / (tp_fn_test[:,0] + fp_tn_ood[:,0])\n","\n","  # Eliminating NaN value at TPR = 1\n","  precision[0] = 1\n","  return precision, tpr_values, fpr_values\n","\n","\n","### Metrics for Distances all classes at the same time approach ###\n","\n","def thresholds_for_each_TPR_distances(distance):\n","  \"\"\"\n","  Creation of the array with the thresholds for each TPR\n","  \"\"\"\n","  sorted_distance = np.sort(distance)\n","  # Inverse the order to get it correctly (greater the threshold, lower the TPR)\n","  tpr_range = np.arange(0,1,0.01)\n","  tpr_range[0] = 0.99999999 # For selecting the first item correctly\n","  distance_thresholds = np.zeros(len(tpr_range))\n","  for index, tpr in enumerate(tpr_range):\n","    distance_thresholds[index] = sorted_distance[int(len(sorted_distance)*tpr)]\n","  return distance_thresholds\n","\n","def compare_distances_to_distance_thr_one_for_all_classes(distances_evaluating, thr_distances_array):\n","  '''\n","  Function that creates an array of shape (tpr, InD_or_OD), where tpr has the lenght of the number of steps of the TPR list\n","  and second dimensions has the total lenght of the distances_evaluating, and cotains True if its InD and False if is ood\n","  :distances_evaluating: list with each element being an array with the distances to avg clusters of one class [array(.), array(.)]\n","  :thr_distances_array: array containing the distance for the TPR\n","   corresponding to that position. For example, the TPR = 0.85 corresponds to the 85th position.\n","  '''\n","  in_or_out_distribution_per_tpr = np.zeros((len(thr_distances_array), len(distances_evaluating)), dtype=bool)\n","  for tpr_index, thr_for_one_tpr in enumerate(thr_distances_array):\n","      in_or_out_distribution_per_tpr[tpr_index] = np.where(distances_evaluating < thr_for_one_tpr, True, False)\n","\n","  return in_or_out_distribution_per_tpr\n","\n","def distance_method_compute_precision_tpr_fpr_for_test_and_ood(dist_test, dist_ood, dist_thresholds):\n","  # Creation of the array with True if predicted InD (True) or OD (False)\n","  in_or_out_distribution_per_tpr_test = compare_distances_to_distance_thr_one_for_all_classes(dist_test, dist_thresholds)\n","  in_or_out_distribution_per_tpr_test[0] = np.zeros((in_or_out_distribution_per_tpr_test.shape[1]),dtype=bool) # To fix that one element is True when TPR is 0\n","  in_or_out_distribution_per_tpr_test[-1] = np.ones((in_or_out_distribution_per_tpr_test.shape[1]),dtype=bool) # To fix that last element is True when TPR is 1\n","  in_or_out_distribution_per_tpr_ood = compare_distances_to_distance_thr_one_for_all_classes(dist_ood, dist_thresholds)\n","\n","  # Creation of arrays with TP, FN and FP, TN\n","  tp_fn_test = tp_fn_fp_tn_computation(in_or_out_distribution_per_tpr_test)\n","  fp_tn_ood = tp_fn_fp_tn_computation(in_or_out_distribution_per_tpr_ood)\n","\n","  # Computing TPR, FPR and Precision\n","  tpr_values = tp_fn_test[:,0] / (tp_fn_test[:,0] + tp_fn_test[:,1])\n","  fpr_values = fp_tn_ood[:,0] / (fp_tn_ood[:,0] + fp_tn_ood[:,1])\n","  precision  = tp_fn_test[:,0] / (tp_fn_test[:,0] + fp_tn_ood[:,0])\n","\n","  # Eliminating NaN value at TPR = 1\n","  precision[0] = 1\n","  return precision, tpr_values, fpr_values\n","\n","\n","### Metrics for Likelihood all classes at the same time approach ###\n","\n","def thresholds_for_each_TPR_likelihood(likelihood):\n","  \"\"\"\n","  Creation of the array with the thresholds for each TPR\n","  \"\"\"\n","  sorted_likelihood = np.sort(likelihood)\n","  # Inverse the order to get it correctly (greater the threshold, lower the TPR)\n","  tpr_range = np.arange(0,1,0.01)[::-1]\n","  tpr_range[0] = 0.99999999 # For selecting the first item correctly\n","  likelihood_thresholds = np.zeros(len(tpr_range))\n","  for index, tpr in enumerate(tpr_range):\n","    likelihood_thresholds[index] = sorted_likelihood[int(len(sorted_likelihood)*tpr)]\n","  return likelihood_thresholds\n","\n","def compare_likelihood_to_likelihood_thr_one_for_all_classes(distances_evaluating, thr_distances_array):\n","  '''\n","  Function that creates an array of shape (tpr, InD_or_OD), where tpr has the lenght of the number of steps of the TPR list\n","  and second dimensions has the total lenght of the distances_evaluating, and cotains True if its InD and False if is ood\n","  :distances_evaluating: list with each element being an array with the distances to avg clusters of one class [array(.), array(.)]\n","  :thr_distances_array: array containing the distance for the TPR\n","   corresponding to that position. For example, the TPR = 0.85 corresponds to the 85th position.\n","  '''\n","  in_or_out_distribution_per_tpr = np.zeros((len(thr_distances_array), len(distances_evaluating)), dtype=bool)\n","  for tpr_index, thr_for_one_tpr in enumerate(thr_distances_array):\n","      in_or_out_distribution_per_tpr[tpr_index] = np.where(distances_evaluating > thr_for_one_tpr, True, False)\n","\n","  return in_or_out_distribution_per_tpr\n","\n","def likelihood_method_compute_precision_tpr_fpr_for_test_and_ood(likelihood_test, likelihood_ood, likelihood_thresholds):\n","  # Creation of the array with True if predicted InD (True) or OD (False)\n","  in_or_out_distribution_per_tpr_test = compare_likelihood_to_likelihood_thr_one_for_all_classes(likelihood_test, likelihood_thresholds)\n","  in_or_out_distribution_per_tpr_test[0] = np.zeros((in_or_out_distribution_per_tpr_test.shape[1]),dtype=bool) # To fix that one element is True when TPR is 0\n","  in_or_out_distribution_per_tpr_test[-1] = np.ones((in_or_out_distribution_per_tpr_test.shape[1]),dtype=bool) # To fix that last element is True when TPR is 1\n","  in_or_out_distribution_per_tpr_ood = compare_likelihood_to_likelihood_thr_one_for_all_classes(likelihood_ood, likelihood_thresholds)\n","\n","  # Creation of arrays with TP, FN and FP, TN\n","  tp_fn_test = tp_fn_fp_tn_computation(in_or_out_distribution_per_tpr_test)\n","  fp_tn_ood = tp_fn_fp_tn_computation(in_or_out_distribution_per_tpr_ood)\n","\n","  # Computing TPR, FPR and Precision\n","  tpr_values = tp_fn_test[:,0] / (tp_fn_test[:,0] + tp_fn_test[:,1])\n","  fpr_values = fp_tn_ood[:,0] / (fp_tn_ood[:,0] + fp_tn_ood[:,1])\n","  precision  = tp_fn_test[:,0] / (tp_fn_test[:,0] + fp_tn_ood[:,0])\n","\n","  # Eliminating NaN value at TPR = 1 and other NaN values that may appear due to\n","  # precision = TP / (TP + FN) = 0 / (0 + 0)\n","  #precision[0] = 1\n","  np.nan_to_num(precision, nan=1, copy=False)\n","  return precision, tpr_values, fpr_values\n","\n","\n","### Metrics for Likelihood per class approach ###\n","\n","def thresholds_likelihood_per_class_for_each_TPR(likelihood_per_class):\n","  # Creation of the array with the thresholds for each TPR (class, dist_per_TPR)\n","  sorted_distances_per_class = [np.sort(x) for x in likelihood_per_class]\n","  tpr_range = np.arange(0,1,0.01)\n","  tpr_range[-1] = 0.99999999 # For selecting the last item correctly\n","  distance_thresholds_test = np.zeros((len(train_data.classes), len(tpr_range)))\n","  for class_index in range(len(train_data.classes)):\n","    for index, tpr in enumerate(tpr_range[::-1]):\n","      distance_thresholds_test[class_index, index] = sorted_distances_per_class[class_index][int(len(sorted_distances_per_class[class_index])*tpr)]\n","  return distance_thresholds_test\n","\n","def compare_likelihood_per_class_to_likelihood_thr_per_class(likelihood_list_per_class, thr_likelihood_array):\n","  '''\n","  Function that creates an array of shape (tpr, InD_or_OD), where tpr has the lenght of the number of steps of the TPR list\n","  and second dimensions has the total lenght of the distances_list_per_class, and cotains True if its InD and False if is OD\n","  :distances_list_per_class: list with each element being an array with the distances to avg clusters of one class [array(.), array(.)]\n","  :thr_distances_array: array of shape (class, dist_for_each_tpr), where first dimension is the class and the second is the distance for the TPR\n","   corresponding to that position. For example, the TPR = 0.85 corresponds to the 85th position.\n","  '''\n","  in_or_out_distribution_per_tpr = np.zeros((len(np.transpose(thr_likelihood_array)),len(np.concatenate(likelihood_list_per_class))),dtype=bool)\n","  for tpr_index ,thr_likelihood_per_class in enumerate(np.transpose(thr_likelihood_array)):\n","      in_or_out_distribution_per_tpr[tpr_index] = np.concatenate([likelihood_one_class > thr_likelihood_per_class[cls_index] for cls_index, likelihood_one_class in enumerate(likelihood_list_per_class)])\n","    \n","  return in_or_out_distribution_per_tpr\n","\n","def likelihood_method_per_class_compute_precision_tpr_fpr_for_test_and_ood(likelihood_test_per_class, likelihood_ood_per_class, likelihood_thresholds_per_class):\n","  # Creation of the array with True if predicted InD (True) or OD (False)\n","  in_or_out_distribution_per_tpr_test = compare_likelihood_per_class_to_likelihood_thr_per_class(likelihood_test_per_class, likelihood_thresholds_per_class)\n","  in_or_out_distribution_per_tpr_test[0] = np.zeros((in_or_out_distribution_per_tpr_test.shape[1]),dtype=bool) # To fix that one element is True when TPR is 0\n","  in_or_out_distribution_per_tpr_test[-1] = np.ones((in_or_out_distribution_per_tpr_test.shape[1]),dtype=bool) # To fix that last element is True when TPR is 1\n","  in_or_out_distribution_per_tpr_ood = compare_likelihood_per_class_to_likelihood_thr_per_class(likelihood_ood_per_class, likelihood_thresholds_per_class)\n","\n","  # Creation of arrays with TP, FN and FP, TN\n","  tp_fn_test = tp_fn_fp_tn_computation(in_or_out_distribution_per_tpr_test)\n","  fp_tn_ood = tp_fn_fp_tn_computation(in_or_out_distribution_per_tpr_ood)\n","\n","  # Computing TPR, FPR and Precision\n","  tpr_values = tp_fn_test[:,0] / (tp_fn_test[:,0] + tp_fn_test[:,1])\n","  fpr_values = fp_tn_ood[:,0] / (fp_tn_ood[:,0] + fp_tn_ood[:,1])\n","  precision  = tp_fn_test[:,0] / (tp_fn_test[:,0] + fp_tn_ood[:,0])\n","\n","  # Eliminating NaN value at TPR = 1\n","  precision[0] = 1\n","  return precision, tpr_values, fpr_values"]},{"cell_type":"code","source":["#@markdown ##Explainability\n","def order_array_of_samples_to_per_class_list(array, preds, n_samples=0):\n","  list_per_class = []\n","  for cl_index in range(len(train_data.classes)):\n","    indices = searchIndicesOfClass(cl_index, preds, n_samples)\n","    list_per_class.append(array[indices])\n","  return list_per_class\n","\n","def compute_reconstruction(spk_counts, weights):\n","  if isinstance(weights, list):\n","    for l, w_l in enumerate(weights):\n","      if l == 0:\n","        reconst = np.matmul(spk_counts, w_l)\n","      else:\n","        reconst = np.matmul(reconst, w_l)\n","  else:\n","    reconst = np.matmul(spk_counts, weights)\n","  return reconst\n","\n","\n","def compute_reconstruction_per_class(spikeFrec_per_class, weights):\n","  reconst_per_class = []\n","  for spikes_one_class in spikeFrec_per_class:\n","    reconst = compute_reconstruction_n_layers(spikes_one_class, weights)\n","    reconst_per_class.append(reconst)\n","  return reconst_per_class\n","\n","\n","def compute_reconstruction_n_layers(spk_counts, weights):\n","  if isinstance(weights, list):\n","    for l, w_l in enumerate(weights):\n","      if l == 0:\n","        reconst = np.matmul(spk_counts, w_l)\n","      else:\n","        reconst = np.matmul(reconst, w_l)\n","  else:\n","    reconst = np.matmul(spk_counts, weights)\n","\n","  return reconst\n","\n","\n","def extract_positive_part_per_class(list_per_class):\n","  positive_part_list = []\n","  for array_one_class in list_per_class:\n","    positive_part_list.append(np.where(array_one_class > 0, array_one_class, 0))\n","  return positive_part_list\n","\n","\n","def rearrange_to_ftmaps(spk_count, ftmaps_shape=(50, 8, 8)):\n","  ch, h, w = ftmaps_shape\n","  unflatten = torch.nn.Unflatten(-1, (ch, h, w))\n","  return unflatten(torch.from_numpy(spk_count)).numpy()\n","\n","\n","def rearrange_to_ftmaps_per_class(spk_count_per_class, ftmaps_shape=(50, 8, 8)):\n","  ftmaps_per_class = []\n","  for spk_one_class in spk_count_per_class:\n","    ftmaps_per_class.append(rearrange_to_ftmaps(spk_one_class, ftmaps_shape))\n","  return ftmaps_per_class\n","\n","def auroc_aupr(d_train_per_class, d_test_per_class, d_ood_per_class):\n","  # Creation of the array with the thresholds for each TPR (class, dist_per_TPR)\n","  d_thresholds_train = thresholds_per_class_for_each_TPR(d_train_per_class)\n","  # Conmputing precision, tpr and fpr\n","  precision, tpr_values, fpr_values = compute_precision_tpr_fpr_for_test_and_ood(d_test_per_class, d_ood_per_class, d_thresholds_train)\n","  # Appending that when FPR = 1 the TPR is also 1:\n","  tpr_val_auroc = np.append(tpr_values,1)\n","  fpr_val_auroc = np.append(fpr_values,1)\n","\n","  #AUROC\n","  auroc = np.trapz(tpr_val_auroc, fpr_val_auroc)\n","\n","  # AUPR\n","  aupr = np.trapz(precision, tpr_values)\n","\n","  print('-'*30)\n","  print(f'Results')\n","  print('-'*30)\n","  print(f'AUROC = {auroc*100:.3f} %')\n","  print(f'AUPR  = {aupr*100:.3f} %')\n","  print('-'*30)\n","  print(f'FPR at 95% TPR: {round(fpr_val_auroc[95]*100,2)}%')\n","  print(f'Threshold mean and std for all classes at 95% TPR: {d_thresholds_train[:, 95].mean():.2f}, {d_thresholds_train[:, 95].std():.2f}')\n","  print('-'*30,'\\n')\n","\n","  return d_thresholds_train\n","\n","\n","def create_spk_count(model, device, ood_dict, od_dataset, mnist_c_opt='zigzag', conv_spikes=False):\n","  if od_dataset == 'MNIST-C':\n","    test_loader_ood = ood_dict[od_dataset](\n","        BATCH_SIZE,\n","        test_only=True,\n","        option=mnist_c_opt\n","        )\n","  else:\n","    test_loader_ood = ood_dict[od_dataset](BATCH_SIZE, test_only=True)\n","    \n","  # Extract logits and hidden spikes\n","  if selected_model == 'ConvNet':\n","    if conv_spikes is True:\n","      accuracy_ood, preds_ood, logits_ood, _spk_count_ood, _conv_spk_ood = test(model, device, test_loader_ood, return_logits=True, return_conv_spikes=True)\n","      print(f'Accuracy for the ood dataset {od_dataset} is {accuracy_ood:.3f} %')\n","      return test_loader_ood, preds_ood, np.sum(_spk_count_ood,axis=0, dtype='uint16'), np.sum(_conv_spk_ood, axis=0, dtype='uint16')\n","    else:\n","      accuracy_ood, preds_ood, logits_ood, _spk_count_ood = test(model, device, test_loader_ood, return_logits=True, return_conv_spikes=False)\n","  else:\n","    accuracy_ood, preds_ood, logits_ood, _spk_count_ood = test(model, device, test_loader_ood, return_logits=True, return_conv_spikes=False)\n","  print(f'Accuracy for the ood dataset {od_dataset} is {accuracy_ood:.3f} %')\n","\n","  return test_loader_ood, preds_ood, np.sum(_spk_count_ood,axis=0, dtype='uint16')"],"metadata":{"id":"6_RTwryyyfE8"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"T3EcN3dbO-AG"},"source":["## Datasets loading"]},{"cell_type":"markdown","metadata":{"id":"dOBdw3VlDdfR"},"source":["### MNIST\n"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"g5Vf8J2sNl24"},"outputs":[],"source":["def load_MNIST(batch_size, test_only=False):\n","  transform = torchvision.transforms.Compose(\n","      [\n","          torchvision.transforms.ToTensor(),\n","      ]\n","  )\n","  test_data_MNIST = torchvision.datasets.MNIST(\n","      root=datasets_path,\n","      train=False,\n","      download=True,\n","      transform=transform,\n","  )\n","  test_loader_MNIST = torch.utils.data.DataLoader(\n","      torchvision.datasets.MNIST(\n","          root=datasets_path,\n","          train=False,\n","          transform=transform,\n","      ),\n","      batch_size=batch_size\n","  )\n","  if test_only is False:\n","    train_data_MNIST = torchvision.datasets.MNIST(\n","        root=datasets_path,\n","        train=True,\n","        download=True,\n","        transform=transform,\n","    )\n","    train_loader_MNIST = torch.utils.data.DataLoader(\n","        train_data_MNIST,\n","        batch_size=batch_size,\n","        shuffle=True\n","    )\n","    return train_data_MNIST, train_loader_MNIST, test_loader_MNIST\n","  else:\n","    return test_loader_MNIST"]},{"cell_type":"markdown","metadata":{"id":"tF9WS8JC7xZv"},"source":["### Fashion MNIST"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"EgDPfdy9Q_yy"},"outputs":[],"source":["def load_Fashion_MNIST(batch_size, test_only=False):\n","  transform = torchvision.transforms.Compose(\n","    [\n","        torchvision.transforms.ToTensor(),\n","    ]\n","  )\n","  test_data_fashion_MNIST = torchvision.datasets.FashionMNIST(\n","      root=datasets_path,\n","      train=False,\n","      download=True,\n","      transform=transform,\n","  )\n","  test_loader_fashion_MNIST = torch.utils.data.DataLoader(\n","      test_data_fashion_MNIST,\n","      batch_size=batch_size\n","  )\n","\n","  if test_only is False:\n","    train_data_fashion_MNIST = torchvision.datasets.FashionMNIST(\n","        root=datasets_path,\n","        train=True,\n","        download=True,\n","        transform=transform,\n","    )\n","    train_loader_fashion_MNIST = torch.utils.data.DataLoader(\n","        train_data_fashion_MNIST,\n","        batch_size=batch_size,\n","        shuffle=True\n","    )\n","    return train_data_fashion_MNIST, train_loader_fashion_MNIST, test_loader_fashion_MNIST\n","  else:\n","    return test_loader_fashion_MNIST"]},{"cell_type":"markdown","metadata":{"id":"yXCmg_x5OdKV"},"source":["### KMNIST"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"X8HcRLmfOdKV","scrolled":false},"outputs":[],"source":["def load_KMNIST(batch_size, test_only=False):\n","  BATCH_SIZE = 256\n","\n","  transform = torchvision.transforms.Compose(\n","      [\n","          torchvision.transforms.ToTensor(),\n","      ]\n","  )\n","\n","  test_data_KMNIST = torchvision.datasets.KMNIST(\n","      root=datasets_path,\n","      train=False,\n","      download=True,\n","      transform=transform,\n","  )\n","  test_loader_KMNIST = torch.utils.data.DataLoader(\n","      torchvision.datasets.KMNIST(\n","          root=datasets_path,\n","          train=False,\n","          transform=transform,\n","      ),\n","      batch_size=batch_size\n","  )\n","  if test_only is False:\n","    train_data_KMNIST = torchvision.datasets.KMNIST(\n","        root=datasets_path,\n","        train=True,\n","        download=True,\n","        transform=transform,\n","    )\n","\n","    train_loader_KMNIST = torch.utils.data.DataLoader(\n","        train_data_KMNIST,\n","        batch_size=batch_size,\n","        shuffle=True\n","    )\n","    return train_data_KMNIST, train_loader_KMNIST, test_loader_KMNIST\n","  else:\n","    return test_loader_KMNIST"]},{"cell_type":"markdown","metadata":{"id":"x0vIsj9fuZwA"},"source":["### MNIST with a square added"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"K2KTyXYvujZu"},"outputs":[],"source":["import random\n","def square_creation(input_tensor: torch.Tensor):\n","  posible_values = [2,20]\n","  mean = int(torch.mean(input_tensor[0])*100)\n","  random.seed(mean)\n","  x_rnd = random.randint(0, 1)\n","  x_start = posible_values[x_rnd]\n","  random.seed(mean-1)\n","  y_rnd = random.randint(0, 1)\n","  y_start = posible_values[y_rnd]\n","  input_tensor[:, x_start:x_start+6, y_start:y_start+6] = torch.ones((1,6,6), dtype=torch.float32)\n","  return input_tensor\n","\n","def load_MNIST_square(batch_size, test_only=True, *args):\n","  transform = torchvision.transforms.Compose(\n","      [\n","          torchvision.transforms.ToTensor(),\n","          torchvision.transforms.Lambda(square_creation)\n","      ]\n","  )\n","  test_data_MNIST_square = torchvision.datasets.MNIST(\n","      root=datasets_path,\n","      train=False,\n","      download=True,\n","      transform=transform,\n","  )\n","  test_loader_MNIST_square = torch.utils.data.DataLoader(\n","      torchvision.datasets.MNIST(\n","          root=datasets_path,\n","          train=False,\n","          transform=transform,\n","      ),\n","      batch_size=batch_size\n","  )\n","  return test_loader_MNIST_square"]},{"cell_type":"markdown","metadata":{"id":"0RaOHfoHhRgY"},"source":["### Cifar10-BW"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"PwsGxh1yhRAI"},"outputs":[],"source":["def load_CIFAR10_BW(batch_size, test_only=True):\n","  transform = torchvision.transforms.Compose(\n","      [\n","          torchvision.transforms.ToTensor(),\n","          torchvision.transforms.Grayscale(num_output_channels=1),\n","          torchvision.transforms.Resize((28,28))\n","      ]\n","  )\n","\n","  test_data_CIFAR10 = torchvision.datasets.CIFAR10(\n","      root=datasets_path,\n","      train=False,\n","      download=True,\n","      transform=transform,\n","  )\n","  test_loader_CIFAR10 = torch.utils.data.DataLoader(\n","      test_data_CIFAR10,\n","      batch_size=batch_size,\n","      shuffle=False\n","  )\n","  return test_loader_CIFAR10"]},{"cell_type":"markdown","metadata":{"id":"YTw_xgswvcGh"},"source":["### Cifar10 colored"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"_MRYal83vcGj"},"outputs":[],"source":["def load_CIFAR10(batch_size, test_only=False):\n","  transform = torchvision.transforms.Compose(\n","      [\n","          torchvision.transforms.ToTensor()\n","      ]\n","  )\n","  test_data_CIFAR10 = torchvision.datasets.CIFAR10(\n","      root=datasets_path,\n","      train=False,\n","      download=True,\n","      transform=transform,\n","  )\n","  test_loader_CIFAR10 = torch.utils.data.DataLoader(\n","      test_data_CIFAR10,\n","      batch_size=batch_size,\n","      shuffle=False\n","  )\n","\n","  if test_only is False:\n","    train_data_CIFAR10 = torchvision.datasets.CIFAR10(\n","        root=datasets_path,\n","        train=True,\n","        download=True,\n","        transform=transform,\n","    )\n","\n","    train_loader_CIFAR10 = torch.utils.data.DataLoader(\n","        train_data_CIFAR10,\n","        batch_size=batch_size,\n","        shuffle=True\n","    )\n","    return train_data_CIFAR10, train_loader_CIFAR10, test_loader_CIFAR10\n","\n","  else:\n","    return test_loader_CIFAR10"]},{"cell_type":"markdown","metadata":{"id":"Z4ZKCp_0nw46"},"source":["### Letters"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"Ip68Ab1Royfd"},"outputs":[],"source":["from torchvision.transforms import Lambda\n","def load_MNIST_Letters(batch_size, test_only=False):\n","  # The rotation and horizontal flip are for getting the images in a similar \n","  # way to the MNIST dataset i.e. vertical and readable from left to right\n","  transform = torchvision.transforms.Compose(\n","      [\n","          lambda img: torchvision.transforms.functional.rotate(img, -90),\n","          lambda img: torchvision.transforms.functional.hflip(img),\n","          torchvision.transforms.ToTensor(),\n","      ]\n","  )\n","  test_data_letters = torchvision.datasets.EMNIST(\n","      root=datasets_path,\n","      split= \"letters\",\n","      train=False,\n","      download=True,\n","      transform=transform,\n","      target_transform=Lambda(\n","          lambda y: y-1\n","      ),\n","  )\n","  # Eliminate the first class, that is non-existant for our case\n","  test_data_letters.classes = test_data_letters.classes[1:] \n","  test_loader_letters = torch.utils.data.DataLoader(\n","      test_data_letters,\n","      batch_size=batch_size,\n","      shuffle=False\n","  )\n","  # To obtain 10.000 test samples, pass trought the function\n","  #test_loader_letters = parse_size_of_dataloader(test_loader_letters, batch_size)\n","  if test_only is False:\n","    train_data_letters = torchvision.datasets.EMNIST(\n","        root=datasets_path,\n","        split= \"letters\",\n","        train=True,\n","        download=True,\n","        transform=transform,\n","        target_transform=Lambda(\n","          lambda y: y-1\n","      ),\n","    )\n","    train_data_letters.classes = train_data_letters.classes[1:] \n","    train_loader_letters = torch.utils.data.DataLoader(\n","        train_data_letters,\n","        batch_size=batch_size,\n","        shuffle=True\n","    )\n","\n","    return train_data_letters, train_loader_letters, test_loader_letters\n","  else:\n","    return test_loader_letters"]},{"cell_type":"markdown","metadata":{"id":"sfZyNhW3a9nG"},"source":["### Omniglot"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"w3PqkZypa9u8"},"outputs":[],"source":["def load_omniglot(batch_size, test_only=True):\n","  transform = torchvision.transforms.Compose(\n","      [\n","          torchvision.transforms.ToTensor(),\n","          torchvision.transforms.Resize((28,28)),\n","          Lambda( lambda img: torchvision.transforms.functional.invert(img)),\n","      ]\n","  )\n","  test_data = torchvision.datasets.Omniglot(\n","      root=datasets_path,\n","      background=False,\n","      download=True,\n","      transform=transform,\n","  )\n","  test_loader = torch.utils.data.DataLoader(\n","      test_data,\n","      batch_size=batch_size,\n","      shuffle=False\n","  )\n","  test_loader = parse_size_of_dataloader(test_loader, batch_size)\n","  return test_loader"]},{"cell_type":"markdown","metadata":{"id":"cViJtIKdkKEW"},"source":["### notMNIST"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"8VgQPTsywuKy"},"outputs":[],"source":["class notMNIST(VisionDataset):\n","    def __init__(self, root: Path, transform=None, samples_per_class=None):\n","        \"\"\"\n","        Args:\n","            root (string): Directory with the images of the selected option.\n","            transform (callable, optional): Optional transform to be applied\n","                on a sample.\n","        \"\"\"\n","        self.root = root\n","        classes = []\n","        images = []\n","        targets = []\n","        # Every directory is a class in the structure of notMNINST\n","        for cl_index, class_dir_name in enumerate(sorted(root.iterdir())): \n","          # Extract the class name from the path\n","          classes.append(class_dir_name.as_posix().split('/')[-1])\n","          class_dir_path = root / class_dir_name\n","          if samples_per_class is None:\n","            # We put the limit way above the length of the dataset to not limit\n","            # the data collected at all, as None indicates that we want all the \n","            # data available\n","            limit = 1000000\n","          else:\n","            limit = samples_per_class\n","          for index, png_im in enumerate(sorted(class_dir_path.iterdir())):\n","            if index == limit:\n","              break\n","            else:\n","              # Some images are corrupted, so we skip those images\n","              try:\n","                # Get images in range 0-1\n","                images.append(torchvision.io.read_image(png_im.as_posix())/255)\n","                targets.append(cl_index)\n","              except RuntimeError:\n","                # we have to update the limit to obtain the desired\n","                # number of imgs per class\n","                limit += 1\n","                continue\n","\n","        # Transform list to tensors\n","        images = torch.stack(images, dim=0)\n","        targets = torch.tensor(targets, dtype=torch.uint8)\n","  \n","        '''\n","        images = torch.zeros(10*samples_per_class, 1, 28, 28, \n","                             dtype=torch.float32)\n","        labels = torch.zeros(10*samples_per_class, dtype=torch.uint8)\n","        # Every directory is a class in the structure of notMNINST\n","        for cl_index, class_dir_name in enumerate(sorted(root.iterdir())): \n","          # Extract the class name from the path\n","          classes.append(class_dir_name.as_posix().split('/')[-1])\n","          class_dir_path = root / class_dir_name\n","          labels[cl_index*samples_per_class:(cl_index+1)*samples_per_class] = torch.ones(samples_per_class, dtype=torch.uint8)*cl_index\n","          limit = samples_per_class\n","          for index, png_im in enumerate(sorted(class_dir_path.iterdir())):\n","            if index == limit:\n","              break\n","            else:\n","              # Some images are corrupted, so we skip those images\n","              try:\n","                # Get images in range 0-1\n","                images[samples_per_class*cl_index + index] = torchvision.io.read_image(png_im.as_posix())/255\n","              except RuntimeError:\n","                # we have to update the limit to obtain the desired\n","                # number of imgs per class\n","                limit += 1\n","                continue\n","        '''\n","        # Shuffle fixed for reproducibility reasons\n","        self.images, self.targets = skl_shuffle(images, targets,\n","                                                         random_state=7)\n","        self.classes = classes\n","        self.transform = transform\n","\n","    def __len__(self):\n","        return len(self.targets)\n","\n","    def __getitem__(self, idx):\n","        if torch.is_tensor(idx):\n","            idx = idx.tolist()\n","        sample = [self.images[idx], self.targets[idx]]\n","        if self.transform:\n","            sample = self.transform(sample)\n","        return sample\n","\n","def load_notMNIST(batch_size, test_only=True, samples_per_class=1000):\n","  # Test only for compatibility issues\n","  notmnist_path = datasets_path / 'notMNIST_small'\n","  if not notmnist_path.exists():\n","    !wget -O notMNIST.tar.gz --no-check-certificate \"https://tecnalia365-my.sharepoint.com/:u:/g/personal/aitor_martinez_tecnalia_com/EZTrDELAyABNqHplnGnNaqgBTCiHIPBAW5D-DBsAmBIGcw?download=1\"\n","    !tar -zxf notMNIST.tar.gz -C $datasets_path\n","    !rm notMNIST.tar.gz\n","\n","  loader = torch.utils.data.DataLoader(\n","      notMNIST(notmnist_path, samples_per_class=samples_per_class), \n","      batch_size=256,\n","      shuffle=False\n","      )\n","  return loader"]},{"cell_type":"markdown","metadata":{"id":"p2ZjT9lwkNkh"},"source":["### MNIST-C"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"4tnR37w4kNqg"},"outputs":[],"source":["class MNIST_C(VisionDataset):\n","    def __init__(self, root_dir, transform=None):\n","        \"\"\"\n","        Args:\n","            root_dir (string): Directory with the images of the selected option.\n","            transform (callable, optional): Optional transform to be applied\n","                on a sample.\n","        \"\"\"\n","        self.root = Path(root_dir)\n","        self.images = np.load(self.root / 'test_images.npy')\n","        self.targets = np.load(self.root / 'test_labels.npy').astype('uint8')\n","        self.transform = transform\n","        self.classes = [str(x) for x in range(10)]\n","\n","    def __len__(self):\n","        return len(self.targets)\n","\n","    def __getitem__(self, idx):\n","        if torch.is_tensor(idx):\n","            idx = idx.tolist()\n","        sample = [self.images[idx], self.targets[idx]]\n","        if self.transform:\n","            sample = self.transform(sample)\n","        return sample\n","\n","\n","class ToTensor(object):\n","    \"\"\"Convert ndarrays in sample to Tensors.\"\"\"\n","\n","    def __call__(self, sample):\n","        image, label = sample\n","        # swap color axis because\n","        # numpy image: H x W x C\n","        # torch image: C x H x W\n","        image = image.transpose((2, 0, 1))/255\n","        return [torch.from_numpy(image), torch.tensor(label)]\n","\n","\n","def load_MNIST_C(batch_size, test_only=True, option='zigzag'):\n","  if not (datasets_path / 'mnist_c').exists():\n","    !wget -O mnist_c.gz --no-check-certificate \"https://tecnalia365-my.sharepoint.com/:u:/g/personal/aitor_martinez_tecnalia_com/ERi3c4DxluJFqpv4wtlTkKEBvhdrY4WwqNRJWKyyVoTQqg?download=1\"\n","    !unzip mnist_c.gz -d $datasets_path\n","    !rm mnist_c.gz\n"," \n","  mnist_c_loader = torch.utils.data.DataLoader(\n","      MNIST_C(datasets_path / 'mnist_c' / option, ToTensor()), \n","      batch_size=256,\n","      shuffle=False\n","      )\n","  return mnist_c_loader"]},{"cell_type":"markdown","metadata":{"id":"hlQlo2HI-ZC3"},"source":["## Model definition (encoder, hidden layers and decoder)"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"gvGnC4XT-O3X"},"outputs":[],"source":["#@markdown ##Model class\n","class Model(torch.nn.Module):\n","    def __init__(self, encoder, snn, decoder):\n","        super(Model, self).__init__()\n","        self.encoder = encoder\n","        self.snn = snn\n","        self.decoder = decoder\n","\n","    def forward(self, x, flag=None):\n","        x = self.encoder(x)\n","        if flag is None:\n","          x = self.snn(x)\n","          x = self.decoder(x)\n","          return x\n","\n","        elif flag == \"hidden_spikes_and_logits\":\n","          x, hdd_spks = self.snn(x, flag)\n","          x, _ = torch.max(x, 0) \n","          return x, hdd_spks\n","\n","        elif flag == \"hidden_and_conv_spikes_and_logits\":\n","          x, hdd_spks, conv_spks = self.snn(x, flag)\n","          x, _ = torch.max(x, 0) \n","          return x, hdd_spks, conv_spks"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"IusxPaOl-jkX"},"outputs":[],"source":["#@markdown ##Decoding function\n","def decode(x):\n","  # First take the max across all time steps, the first dimension\n","  # [time_step, batch_size, output_neurons ]\n","  x, _ = torch.max(x, 0)\n","  # Then compute the logsoftmax across the \n","  log_p_y = torch.nn.functional.log_softmax(x, dim=1)\n","  return log_p_y"]},{"cell_type":"markdown","metadata":{"id":"g8y4CMv0Ltpr"},"source":["### Fully Connected models"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"ID__wSUzS7rz"},"outputs":[],"source":["#@markdown ## Fully connected model class\n","from norse.torch import LIFParameters\n","from norse.torch.module.lif import LIFCell\n","from norse.torch import LICell\n","\n","MODEL_NAME = 'Fully connected'\n","class FC_SNN_1(torch.nn.Module):\n","    def __init__(self, input_features, hidden_features, output_features, dt=0.001):\n","        #super(FC_SNN_1, self).__init__()\n","        super().__init__()\n","\n","        # Hidden layer\n","        self.fc1 = torch.nn.Linear(input_features, hidden_features, bias=False)\n","        self.lif1 = LIFCell(\n","            p=LIFParameters(alpha=100, v_th=torch.tensor(0.5)),\n","            dt=dt            \n","        )\n","\n","        # Output layer\n","        self.fc_out = torch.nn.Linear(hidden_features, output_features, bias=False)\n","        self.out = LICell(dt=dt)\n","\n","        self.input_features = input_features\n","        self.hidden_features = hidden_features\n","        self.output_neurons = output_features\n","      \n","    def forward(self, x, flag=None):\n","        # Input shape = [time_step,batch_size, ... , ... , ... ]\n","        seq_length, batch_size, _, _, _ = x.shape\n","        s1 = so = None\n","        #voltages = []\n","\n","        voltages = torch.zeros(\n","            seq_length, batch_size, self.output_neurons,\n","            device=x.device, dtype=x.dtype\n","        )\n","        if flag is None:\n","          for ts in range(seq_length):\n","              # Flatten the input to [batch_size, input_features]\n","              z = x[ts, :, :, :].view(-1, self.input_features)\n","              \n","              # Hidden layer\n","              z = self.fc1(z)\n","              z, s1 = self.lif1(z, s1)\n","\n","              # Output layer\n","              z = self.fc_out(z)\n","              vo, so = self.out(z, so)\n","              voltages[ts] = vo\n","\n","          return voltages\n","\n","        elif flag == \"hidden_spikes_and_logits\":\n","          hdn_spikes = torch.zeros(\n","              seq_length, batch_size, self.hidden_features,\n","              device=x.device, dtype=x.dtype\n","          )\n","          for ts in range(seq_length):\n","              # Flatten the input to [batch_size, input_features]\n","              z = x[ts, :, :, :].view(-1, self.input_features)\n","              \n","              # Hidden layer\n","              z = self.fc1(z)\n","              z, s1 = self.lif1(z, s1)\n","              hdn_spikes[ts] = z\n","\n","              # Output layer\n","              z = self.fc_out(z)\n","              vo, so = self.out(z, so)\n","              voltages[ts] = vo\n","        else:\n","          raise NameError('Wrong flag')\n","\n","        return voltages, hdn_spikes\n","        "]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"Eg-plICn463G"},"outputs":[],"source":["#@markdown ## Fully connected model class 2 hidden layers\n","from norse.torch import LIFParameters\n","from norse.torch.module.lif import LIFCell\n","from norse.torch import LICell\n","\n","MODEL_NAME = 'Fully connected'\n","class FC_SNN_2(torch.nn.Module):\n","    def __init__(self, input_features, hidden_features, output_features, dt=0.001):\n","        #super(FC_SNN_2, self).__init__()\n","        super().__init__()\n","\n","        # Hidden layer\n","        self.fc1 = torch.nn.Linear(input_features, 300, bias=False)\n","        self.fc2 = torch.nn.Linear(300, hidden_features, bias=False)\n","        self.lif1 = LIFCell(\n","            p=LIFParameters(alpha=100, v_th=torch.tensor(0.25)),\n","            dt=dt            \n","        )\n","        self.lif2 = LIFCell(\n","            p=LIFParameters(alpha=100, v_th=torch.tensor(0.25)),\n","            dt=dt            \n","        )\n","\n","        # Output layer\n","        self.fc_out = torch.nn.Linear(hidden_features, output_features, bias=False)\n","        self.out = LICell(dt=dt)\n","\n","        self.input_features = input_features\n","        self.hidden_features = hidden_features\n","        self.output_neurons = output_features\n","      \n","    def forward(self, x, flag=None):\n","        # Input shape = [time_step,batch_size, ... , ... , ... ]\n","        seq_length, batch_size, _, _, _ = x.shape\n","        s1 = s2 = so = None\n","        #voltages = []\n","\n","        voltages = torch.zeros(\n","            seq_length, batch_size, self.output_neurons,\n","            device=x.device, dtype=x.dtype\n","        )\n","        if flag is None:\n","          for ts in range(seq_length):\n","              # Flatten the input to [batch_size, input_features]\n","              z = x[ts, :, :, :].view(-1, self.input_features)\n","              \n","              # Hidden layer 1\n","              z = self.fc1(z)\n","              z, s1 = self.lif1(z, s1)\n","\n","              # Hidden layer 2\n","              z = self.fc2(z)\n","              z, s2 = self.lif2(z, s2)\n","\n","              # Output layer\n","              z = self.fc_out(z)\n","              vo, so = self.out(z, so)\n","              voltages[ts] = vo\n","\n","          return voltages\n","\n","        elif flag == \"hidden_spikes_and_logits\":\n","          hdn_spikes = torch.zeros(\n","              seq_length, batch_size, self.hidden_features,\n","              device=x.device, dtype=x.dtype\n","          )\n","          for ts in range(seq_length):\n","              # Flatten the input to [batch_size, input_features]\n","              z = x[ts, :, :, :].view(-1, self.input_features)\n","              \n","              # Hidden layer 1\n","              z = self.fc1(z)\n","              z, s1 = self.lif1(z, s1)\n","              #hdn_spikes[ts] = z\n","\n","              # Hidden layer 2\n","              z = self.fc2(z)\n","              z, s2 = self.lif2(z, s2)\n","              hdn_spikes[ts] = z\n","\n","              # Output layer\n","              z = self.fc_out(z)\n","              vo, so = self.out(z, so)\n","              voltages[ts] = vo\n","        else:\n","          raise NameError('Wrong flag')\n","\n","        return voltages, hdn_spikes"]},{"cell_type":"markdown","metadata":{"id":"tHl4KfmaL6ex"},"source":["### Convolutional Networks"]},{"cell_type":"code","source":["#@markdown SNN - ConvNet - (20, 3x3) -> (50, 3x3) - AVG POOL\n","from norse.torch import LIFParameters, LIFState\n","from norse.torch.module.lif import LIFCell, LIFRecurrentCell\n","# Notice the difference between \"LIF\" (leaky integrate-and-fire) and \"LI\" (leaky integrator)\n","from norse.torch import LICell, LIState\n","\n","class ConvNet_1(torch.nn.Module):\n","    def __init__(self, hidden_neurons, output_neurons, num_channels=1, \n","        feature_size=28, alpha=100):\n","        #super(ConvNet, self).__init__()\n","        super().__init__()\n","\n","        self.features = int(((feature_size - 2) / 2) - 2)\n","\n","        self.conv1 = torch.nn.Conv2d(num_channels, 20, 3, 1, bias=False)\n","        self.conv2 = torch.nn.Conv2d(20, 50, 3, 1, bias=False)\n","        self.fc1 = torch.nn.Linear(self.features * self.features * 50, hidden_neurons, bias=False)\n","        self.fc2 = torch.nn.Linear(hidden_neurons, output_neurons, bias=False) # Out fc\n","        self.lif0 = LIFCell(p=LIFParameters(v_th=torch.tensor(0.2), alpha=alpha))\n","        self.lif1 = LIFCell(p=LIFParameters(v_th=torch.tensor(0.2), alpha=alpha))\n","        self.lif2 = LIFCell(p=LIFParameters(v_th=torch.tensor(0.1), alpha=alpha))\n","        self.out = LICell()\n","\n","        self.hidden_neurons = hidden_neurons\n","        self.output_neurons = output_neurons\n","\n","    def forward(self, x, flag=None):\n","        seq_length = x.shape[0]\n","        batch_size = x.shape[1]\n","        \n","        # specify the initial states\n","        s0 = s1 = s2 = so = None\n","        voltages = torch.zeros(\n","            seq_length, batch_size, self.output_neurons, device=x.device, dtype=x.dtype\n","        )\n","        if flag is None:\n","          for ts in range(seq_length): # A la derecha pongo la salida del modelo\n","              # First convolution\n","              z = self.conv1(x[ts, :]) \n","              z, s0 = self.lif0(z, s0) \n","              z = torch.nn.functional.avg_pool2d(z, 2) # (batch_size, filters (20), (H-4)/2, (W-4)/2)\n","\n","              # Second convolution\n","              z = self.conv2(z)\n","              z, s1 = self.lif1(z, s1)\n","              #z = torch.nn.functional.avg_pool2d(z, 2)\n","              \n","              # Fully connected part\n","              z = z.view(-1, self.features ** 2 * 50) # Flatten -Z (batch_size, 800)\n","\n","              # First linear connection\n","              z = self.fc1(z) # (batch_size, 500)\n","              z, s2 = self.lif2(z, s2) # The neuron is the activation function\n","\n","              # Second linear connection\n","              z = self.fc2(z) # (batch_size, 10)\n","              v, so = self.out(z, so)\n","              voltages[ts, :, :] = v\n","\n","          return voltages\n","\n","        elif flag == \"hidden_spikes_and_logits\":\n","          hdn_spk_last_layer = torch.zeros(\n","              seq_length, batch_size, self.hidden_neurons, device=x.device, dtype=torch.int8\n","          )\n","          for ts in range(seq_length): # A la derecha pongo la salida del modelo\n","              # First convolution\n","              z = self.conv1(x[ts, :]) # (batch_size, filters (20), H-2, W-2)\n","              z, s0 = self.lif0(z, s0) # (batch_size, filters (20), H-2, W-2)\n","              z = torch.nn.functional.avg_pool2d(z, 2) # (batch_size, filters (20), (H-4)/2, (W-4)/2)\n","\n","              # Second convolution\n","              z = self.conv2(z) # Multiplica por 10 para que el valor de \n","              z, s1 = self.lif1(z, s1)              \n","              \n","              # Fully connected part\n","              z = z.view(-1, self.features ** 2 * 50) # Flatten -Z (batch_size, 800)\n","\n","              # First linear connection\n","              z = self.fc1(z) # (batch_size, 500)\n","              z, s2 = self.lif2(z, s2) # The neuron is the activation function\n","              hdn_spk_last_layer[ts, :, :] = z # To save the spikes (ts, batch_size, 500)\n","\n","              # Second linear connection\n","              z = self.fc2(z) # (batch_size, 10)\n","              v, so = self.out(z, so)\n","              voltages[ts, :, :] = v\n","\n","          return voltages, hdn_spk_last_layer\n","\n","        \n","        elif flag == \"hidden_and_conv_spikes_and_logits\":\n","          hdn_spk_last_layer = torch.zeros(\n","              seq_length, batch_size, self.hidden_neurons, device=x.device, dtype=torch.int8\n","          )\n","          conv_spikes = torch.zeros(seq_length, batch_size, 50, 11, 11, \n","                                    device=x.device, dtype=torch.bool)\n","          for ts in range(seq_length): # A la derecha pongo la salida del modelo\n","              # First convolution\n","              z = self.conv1(x[ts, :]) # (batch_size, filters (20), H-2, W-2)\n","              z, s0 = self.lif0(z, s0) # (batch_size, filters (20), H-2, W-2)\n","              z = torch.nn.functional.avg_pool2d(z, 2) # (batch_size, filters (20), (H-4)/2, (W-4)/2)\n","\n","              # Second convolution\n","              z = self.conv2(z) # Multiplica por 10 para que el valor de \n","              z, s1 = self.lif1(z, s1)\n","              conv_spikes[ts] = z\n","                       \n","              \n","              # Fully connected part\n","              z = z.view(-1, self.features ** 2 * 50) # Flatten -Z (batch_size, 800)\n","\n","              # First linear connection\n","              z = self.fc1(z) # (batch_size, 500)\n","              z, s2 = self.lif2(z, s2) # The neuron is the activation function\n","              hdn_spk_last_layer[ts, :, :] = z # To save the spikes (ts, batch_size, 500)\n","\n","              # Second linear connection\n","              z = self.fc2(z) # (batch_size, 10)\n","              v, so = self.out(z, so)\n","              voltages[ts, :, :] = v\n","\n","          return voltages, hdn_spk_last_layer, conv_spikes"],"metadata":{"id":"omXbAnY45aGj","cellView":"form"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["#@markdown SNN - ConvNet - (20, 3x3) -> (50, 3x3) - AVG POOL, 2 FC\n","from norse.torch import LIFParameters, LIFState\n","from norse.torch.module.lif import LIFCell, LIFRecurrentCell\n","# Notice the difference between \"LIF\" (leaky integrate-and-fire) and \"LI\" (leaky integrator)\n","from norse.torch import LICell, LIState\n","\n","\n","MODEL_NAME = 'ConvNet'\n","class ConvNet_2(torch.nn.Module):\n","    def __init__(self, hidden_neurons, output_neurons, num_channels=1, \n","        feature_size=28, alpha=100):\n","        #super(ConvNet, self).__init__()\n","        super().__init__()\n","\n","        self.features = int(((feature_size - 2) / 2) - 2)\n","\n","        self.conv1 = torch.nn.Conv2d(num_channels, 20, 3, 1, bias=False)\n","        self.conv2 = torch.nn.Conv2d(20, 50, 3, 1, bias=False)\n","        self.fc1 = torch.nn.Linear(self.features * self.features * 50, 500, bias=False)\n","        self.fc2 = torch.nn.Linear(500, hidden_neurons, bias=False)\n","        self.fcout = torch.nn.Linear(hidden_neurons, output_neurons, bias=False) # Out fc\n","        self.lif0 = LIFCell(p=LIFParameters(v_th=torch.tensor(0.2), alpha=alpha))\n","        self.lif1 = LIFCell(p=LIFParameters(v_th=torch.tensor(0.2), alpha=alpha))\n","        self.lif2 = LIFCell(p=LIFParameters(v_th=torch.tensor(0.1), alpha=alpha))\n","        self.lif3 = LIFCell(p=LIFParameters(v_th=torch.tensor(0.05), alpha=alpha))\n","        self.out = LICell()\n","\n","        self.hidden_neurons = hidden_neurons\n","        self.output_neurons = output_neurons\n","\n","    def forward(self, x, flag=None):\n","        seq_length = x.shape[0]\n","        batch_size = x.shape[1]\n","        \n","        # specify the initial states\n","        s0 = s1 = s2 = s3 = so = None\n","        voltages = torch.zeros(\n","            seq_length, batch_size, self.output_neurons, device=x.device, dtype=x.dtype\n","        )\n","        if flag is None:\n","          for ts in range(seq_length): # A la derecha pongo la salida del modelo\n","              # First convolution\n","              z = self.conv1(x[ts, :]) \n","              z, s0 = self.lif0(z, s0) \n","              z = torch.nn.functional.avg_pool2d(z, 2) # (batch_size, filters (20), (H-4)/2, (W-4)/2)\n","\n","              # Second convolution\n","              z = self.conv2(z)\n","              z, s1 = self.lif1(z, s1)\n","              #z = torch.nn.functional.avg_pool2d(z, 2)\n","              \n","              # Fully connected part\n","              z = z.view(-1, self.features ** 2 * 50) # Flatten -Z (batch_size, 800)\n","\n","              # First linear connection\n","              z = self.fc1(z) # (batch_size, 500)\n","              z, s2 = self.lif2(z, s2) # The neuron is the activation function\n","\n","              # Second linear connection\n","              z = self.fc2(z) \n","              z, s3 = self.lif3(z, s3) # The neuron is the activation function\n","\n","              # Final linear connection\n","              z = self.fcout(z) # (batch_size, 10)\n","              v, so = self.out(z, so)\n","              voltages[ts, :, :] = v\n","\n","          return voltages\n","\n","        elif flag == \"hidden_spikes_and_logits\":\n","          hdn_spk_last_layer = torch.zeros(\n","              seq_length, batch_size, self.hidden_neurons, device=x.device, dtype=torch.int8\n","          )\n","          for ts in range(seq_length): # A la derecha pongo la salida del modelo\n","              # First convolution\n","              z = self.conv1(x[ts, :]) # (batch_size, filters (20), H-2, W-2)\n","              z, s0 = self.lif0(z, s0) # (batch_size, filters (20), H-2, W-2)\n","              z = torch.nn.functional.avg_pool2d(z, 2) # (batch_size, filters (20), (H-4)/2, (W-4)/2)\n","\n","              # Second convolution\n","              z = self.conv2(z) # Multiplica por 10 para que el valor de \n","              z, s1 = self.lif1(z, s1)              \n","              \n","              # Fully connected part\n","              z = z.view(-1, self.features ** 2 * 50) # Flatten -Z (batch_size, 800)\n","\n","              # First linear connection\n","              z = self.fc1(z) # (batch_size, 500)\n","              z, s2 = self.lif2(z, s2) # The neuron is the activation function\n","              #hdn_spk_last_layer[ts, :, :] = z # To save the spikes (ts, batch_size, 500)\n","\n","              # Second linear connection\n","              z = self.fc2(z) \n","              z, s3 = self.lif3(z, s3) # The neuron is the activation function\n","              hdn_spk_last_layer[ts, :, :] = z # To save the spikes (ts, batch_size, 500)\n","\n","              # Second linear connection\n","              z = self.fcout(z) # (batch_size, 10)\n","              v, so = self.out(z, so)\n","              voltages[ts, :, :] = v\n","\n","          return voltages, hdn_spk_last_layer\n","\n","        \n","        elif flag == \"hidden_and_conv_spikes_and_logits\":\n","          hdn_spk_last_layer = torch.zeros(\n","              seq_length, batch_size, self.hidden_neurons, device=x.device, dtype=torch.int8\n","          )\n","          conv_spikes = torch.zeros(seq_length, batch_size, 50, 11, 11, \n","                                    device=x.device, dtype=torch.bool)\n","          for ts in range(seq_length): # A la derecha pongo la salida del modelo\n","              # First convolution\n","              z = self.conv1(x[ts, :]) # (batch_size, filters (20), H-2, W-2)\n","              z, s0 = self.lif0(z, s0) # (batch_size, filters (20), H-2, W-2)\n","              z = torch.nn.functional.avg_pool2d(z, 2) # (batch_size, filters (20), (H-4)/2, (W-4)/2)\n","\n","              # Second convolution\n","              z = self.conv2(z) # Multiplica por 10 para que el valor de \n","              z, s1 = self.lif1(z, s1)\n","              conv_spikes[ts] = z\n","                       \n","              \n","              # Fully connected part\n","              z = z.view(-1, self.features ** 2 * 50) # Flatten -Z (batch_size, 800)\n","\n","              # First linear connection\n","              z = self.fc1(z) # (batch_size, 500)\n","              z, s2 = self.lif2(z, s2) # The neuron is the activation function\n","              #hdn_spk_last_layer[ts, :, :] = z # To save the spikes (ts, batch_size, 500)\n","\n","              # Second linear connection\n","              z = self.fc2(z) \n","              z, s3 = self.lif3(z, s3) # The neuron is the activation function\n","              hdn_spk_last_layer[ts, :, :] = z # To save the spikes (ts, batch_size, 500)\n","\n","              # Second linear connection\n","              z = self.fcout(z) # (batch_size, 10)\n","              v, so = self.out(z, so)\n","              voltages[ts, :, :] = v\n","\n","          return voltages, (hdn_spk_last_layer, conv_spikes)"],"metadata":{"cellView":"form","id":"XgP4V44F5edx"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"L__CXwGL1vkt"},"source":["### Load model function"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"Z8WxEgu1saH_"},"outputs":[],"source":["def load_model(option, device, hidden_neurons=None, output_neurons=10, \n","               n_hidden_layers=1):\n","\n","  if option == 'Fully_connected':\n","    if hidden_neurons is None:\n","      hidden_neurons = 200\n","\n","    encoder=PoissonEncoder(seq_length=NUM_TIME_STEPS, f_max=100)\n","\n","    if n_hidden_layers == 1:\n","      \n","      model = Model(\n","          encoder=encoder,\n","          snn=FC_SNN_1(input_features=INPUT_FEATURES,\n","                    hidden_features=hidden_neurons,\n","                    output_features=output_neurons),\n","          decoder=decode\n","      ).to(device)\n","\n","    elif n_hidden_layers == 2:\n","      model = Model(\n","          encoder=encoder, \n","          snn=FC_SNN_2(input_features=INPUT_FEATURES,\n","                    hidden_features=hidden_neurons,\n","                    output_features=output_neurons),\n","          decoder=decode\n","      ).to(device)\n","\n","    else:\n","      raise NameError('Wrong number of layers')\n","\n","  elif option == 'ConvNet':\n","    \n","    if hidden_neurons is None:\n","      hidden_neurons = 300\n","\n","    if n_hidden_layers == 1:\n","      \n","      model = Model(\n","      encoder=PoissonEncoder(NUM_TIME_STEPS),\n","      snn=ConvNet_1(hidden_neurons=hidden_neurons,\n","                  output_neurons=output_neurons,\n","                  alpha=80),\n","      decoder=decode\n","      ).to(DEVICE)\n","\n","    elif n_hidden_layers == 2:\n","      model = Model(\n","      encoder=PoissonEncoder(NUM_TIME_STEPS),\n","      snn=ConvNet_2(hidden_neurons=hidden_neurons,\n","                  output_neurons=output_neurons,\n","                  alpha=80),\n","      decoder=decode\n","      ).to(DEVICE)\n","\n","    else:\n","      raise NameError('Wrong number of layers')\n","  else:\n","    raise ValueError('Wrong option introduced')\n","\n","  return model"]},{"cell_type":"markdown","metadata":{"id":"vUPCLvjIGMSB"},"source":["## Dictionary to load the datasets"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"Iq0iSCG2GMXA"},"outputs":[],"source":["# TODO: Aqui he dejado lo que tengo que definir para poder realizar el benchmark\n","#   automatico, pero como también sirve para lo otro tendre que definirlo en\n","#   algún punto común de ambos\n","BATCH_SIZE = 256\n","\n","IN_DISTRIBUTION_DATASETS = {\n","    'MNIST': load_MNIST,\n","    'Fashion_MNIST': load_Fashion_MNIST,\n","    'KMNIST': load_KMNIST,\n","    'Letters': load_MNIST_Letters,\n","    'MNIST_Square': load_MNIST_square,\n","    'CIFAR10': load_CIFAR10,\n","}\n","\n","OUT_DISTRIBUTION_DATASETS = {\n","    'MNIST': load_MNIST,\n","    'Fashion_MNIST': load_Fashion_MNIST,\n","    'KMNIST': load_KMNIST,\n","    'Letters': load_MNIST_Letters,\n","    'noMNIST': load_notMNIST,\n","    'omniglot': load_omniglot,\n","    'CIFAR10-BW': load_CIFAR10_BW,\n","    'MNIST-C': load_MNIST_C,\n","    'MNIST_Square': load_MNIST_square\n","}"]},{"cell_type":"markdown","metadata":{"id":"sxiAnkxODitR"},"source":["## Train and test functions\n"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"hyg9mGYL-O3Y"},"outputs":[],"source":["def train(model, device, train_loader, optimizer, epoch, max_epochs):\n","    model.train() # Defines que estas en train\n","    losses = []\n","\n","    for (data, target) in tqdm(train_loader, leave=False, desc='Progress of one epoch'):\n","        # Send data to the GPU if available\n","        data, target = data.to(device), target.to(device)\n","        # Converts gradients to zero for every minibatch (must be called\n","        # anywhere before loss.backward())\n","        optimizer.zero_grad()\n","        output = model(data)\n","        # Negative loglikelihoog loss, for classification problems. The input must contain log probabilities of each class\n","        loss = torch.nn.functional.nll_loss(output, target)\n","        # Computes the gradient of the current tensor (loss in this case) \n","        # w.r.t. the graph leaves\n","        loss.backward()\n","        # Once gradients computed, the .step() can be called to update the parameters\n","        optimizer.step()\n","        # Store the losses of every minibatch\n","        losses.append(loss.item())\n","\n","    mean_loss = np.mean(losses)\n","    return losses, mean_loss"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"3pEsj9fc-O3Z"},"outputs":[],"source":["def test(model, device, test_loader, return_logits=False, return_conv_spikes=False):\n","\n","    # In inference, don't calculate losses\n","\n","    # To accumulate all the spikes across different batches\n","    preds   = []\n","    all_logits = []\n","    hidden_spikes = []\n","    conv_spikes = []\n","    losses = []\n","    correct = 0\n","\n","    n = 0\n","    # Puts model in inference mode\n","    model.eval() # Turns model.training off\n","\n","    # torch.no_grad() -> Context-manager that disabled gradient calculation. \n","    # Disabling gradient calculation is useful for inference,\n","    # when you are sure that you will not call Tensor.backward().\n","    # It will reduce memory consumption for computations\n","    with torch.no_grad():\n","        for data, target in test_loader:\n","            # Send data to GPU if available and evaluate it\n","            n = n + 1 \n","            data, target = data.to(device), target.to(device)\n","            if return_conv_spikes is True:\n","              logits, hdd_spk, conv_spk = model(data, flag=\"hidden_and_conv_spikes_and_logits\")\n","              output = torch.nn.functional.log_softmax(logits, dim=1)\n","              all_logits.append(logits.cpu().numpy())\n","              hidden_spikes.append(hdd_spk.detach().cpu().numpy())\n","              conv_spikes.append(conv_spk.detach().cpu().numpy())\n","\n","            elif return_logits is True:\n","              logits, hdd_spk = model(data, flag=\"hidden_spikes_and_logits\")\n","              output = torch.nn.functional.log_softmax(logits, dim=1)\n","              all_logits.append(logits.cpu().numpy())\n","              hidden_spikes.append(hdd_spk.detach().cpu().numpy())\n","\n","            else:\n","              output = model(data)\n","            # Compute and sum the loss\n","              test_loss = torch.nn.functional.nll_loss(output, target)\n","              losses.append(test_loss.item())              \n","            # Get the index of the max log-probability\n","            pred = output.argmax(dim=1, keepdim=True)\n","            correct += pred.eq(target.view_as(pred)).sum().item()\n","            # Extract the labels, predictions and the hidden layer spikes\n","            preds.append(pred.cpu().numpy())    \n","\n","    #test_loss /= len(test_loader.dataset)\n","    accuracy = 100.0 * correct / len(test_loader.dataset)\n","    if return_logits is True:\n","      if conv_spikes == []:\n","        return accuracy, np.concatenate(preds).squeeze(), np.concatenate(all_logits), np.hstack(hidden_spikes)\n","      # If conv_spikes is not empty, we must return it. To be compliant with the\n","      # rest of the code, we pack the conv spikes with the hidden spikes\n","      else: \n","        return accuracy, np.concatenate(preds).squeeze(), np.concatenate(all_logits), np.hstack(hidden_spikes), np.hstack(conv_spikes)\n","    # Concatenate is used to attach each batch to the previous one \n","    # in the same dimensions, to obtain the full test split of spikes \n","    # in a single array. Squeeze() is to eliminate one extra dimension.\n","    return np.mean(losses), accuracy, np.concatenate(preds).squeeze()"]},{"cell_type":"markdown","metadata":{"id":"luLXrDuuspf6"},"source":["## Check CUDA availability"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"TWqW5t9CsnOT"},"outputs":[],"source":["if torch.cuda.is_available():\n","  DEVICE = torch.device(\"cuda\")\n","else:\n","  DEVICE = torch.device(\"cpu\")"]},{"cell_type":"markdown","metadata":{"id":"uUwSw1-LkLVR"},"source":["## Download pretrained weights"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":10656,"status":"ok","timestamp":1663926491916,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"},"user_tz":-120},"id":"7HBCl2UKkZz8","outputId":"f0ebbbea-6420-4f4c-b122-5ba891c804ba"},"outputs":[{"output_type":"stream","name":"stdout","text":["--2022-09-23 09:48:00--  https://tecnalia365-my.sharepoint.com/:u:/g/personal/aitor_martinez_tecnalia_com/Ea2uSuEbePRIklCHnUxhAB0BLUVrQ4IxcnLyu4BnU_i8ag?download=1\n","Resolving tecnalia365-my.sharepoint.com (tecnalia365-my.sharepoint.com)... 13.107.136.9, 13.107.138.9\n","Connecting to tecnalia365-my.sharepoint.com (tecnalia365-my.sharepoint.com)|13.107.136.9|:443... connected.\n","HTTP request sent, awaiting response... 302 Found\n","Location: /personal/aitor_martinez_tecnalia_com/Documents/SNN/Pretrained_weights.zip?ga=1 [following]\n","--2022-09-23 09:48:03--  https://tecnalia365-my.sharepoint.com/personal/aitor_martinez_tecnalia_com/Documents/SNN/Pretrained_weights.zip?ga=1\n","Reusing existing connection to tecnalia365-my.sharepoint.com:443.\n","HTTP request sent, awaiting response... 200 OK\n","Length: 81815613 (78M) [application/x-zip-compressed]\n","Saving to: ‘Pretrained_weights.zip’\n","\n","Pretrained_weights. 100%[===================>]  78.03M  14.4MB/s    in 6.1s    \n","\n","2022-09-23 09:48:10 (12.7 MB/s) - ‘Pretrained_weights.zip’ saved [81815613/81815613]\n","\n","Archive:  Pretrained_weights.zip\n","  inflating: /content/Pretrained_weights/state_dict_Fashion_MNIST_ConvNet_300_10_1_layers.pth  \n","  inflating: /content/Pretrained_weights/state_dict_Fashion_MNIST_ConvNet_300_10_2_layers.pth  \n","  inflating: /content/Pretrained_weights/state_dict_Fashion_MNIST_Fully_connected_200_10_1_layers.pth  \n","  inflating: /content/Pretrained_weights/state_dict_Fashion_MNIST_Fully_connected_200_10_2_layers.pth  \n","  inflating: /content/Pretrained_weights/state_dict_KMNIST_ConvNet_300_10_1_layers.pth  \n","  inflating: /content/Pretrained_weights/state_dict_KMNIST_ConvNet_300_10_2_layers.pth  \n","  inflating: /content/Pretrained_weights/state_dict_KMNIST_Fully_connected_200_10_1_layers.pth  \n","  inflating: /content/Pretrained_weights/state_dict_KMNIST_Fully_connected_200_10_2_layers.pth  \n","  inflating: /content/Pretrained_weights/state_dict_Letters_ConvNet_300_26_1_layers.pth  \n","  inflating: /content/Pretrained_weights/state_dict_Letters_ConvNet_300_26_2_layers.pth  \n","  inflating: /content/Pretrained_weights/state_dict_Letters_Fully_connected_200_26_1_layers.pth  \n","  inflating: /content/Pretrained_weights/state_dict_Letters_Fully_connected_200_26_2_layers.pth  \n","  inflating: /content/Pretrained_weights/state_dict_MNIST_ConvNet_300_10_1_layers.pth  \n","  inflating: /content/Pretrained_weights/state_dict_MNIST_ConvNet_300_10_2_layers.pth  \n","  inflating: /content/Pretrained_weights/state_dict_MNIST_Fully_connected_200_10_1_layers.pth  \n","  inflating: /content/Pretrained_weights/state_dict_MNIST_Fully_connected_200_10_2_layers.pth  \n"]}],"source":["!wget -O Pretrained_weights.zip --no-check-certificate \"https://tecnalia365-my.sharepoint.com/:u:/g/personal/aitor_martinez_tecnalia_com/Ea2uSuEbePRIklCHnUxhAB0BLUVrQ4IxcnLyu4BnU_i8ag?download=1\"\n","!unzip Pretrained_weights.zip -d $root_path\n","!rm Pretrained_weights.zip"]},{"cell_type":"markdown","metadata":{"id":"f7T45nyDUReS"},"source":["# **Automated benchmark** and Bayesian analysis and CD graph\n","---"]},{"cell_type":"markdown","source":["## Automated benchmark"],"metadata":{"id":"FRvwz4NgvYu7"}},{"cell_type":"markdown","metadata":{"id":"vtG4jQhKUlN8"},"source":["Run this cell with the In and Out Distributions you want to test (all combinations are going to be made)\n","\n","- Faltan las instrucciones sobre los pesos y los modelos y su accuracy"]},{"cell_type":"markdown","metadata":{"id":"KaVn5-8sZC1M"},"source":["\n","\n","*   in_distribution_datasets_benchmark = [ 'MNIST', 'Fashion_MNIST', 'KMNIST', 'Letters' ]\n","*   out_of_distribution_datasets_benchmark = [ 'MNIST', 'Fashion_MNIST',' KMNIST','Letters', 'noMNIST', 'omniglot', 'CIFAR10-BW', 'MNIST-C' ]\n","* mnist-c options = [ 'MNIST-C/brightness',\n"," 'MNIST-C/canny_edges',\n"," 'MNIST-C/dotted_line',\n"," 'MNIST-C/fog',\n"," 'MNIST-C/glass_blur',\n"," 'MNIST-C/identity',\n"," 'MNIST-C/impulse_noise',\n"," 'MNIST-C/motion_blur',\n"," 'MNIST-C/rotate',\n"," 'MNIST-C/scale',\n"," 'MNIST-C/shear',\n"," 'MNIST-C/shot_noise',\n"," 'MNIST-C/spatter',\n"," 'MNIST-C/stripe',\n"," 'MNIST-C/translate',\n"," 'MNIST-C/zigzag' ]\n","\n"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"uFyfLvRLURZ0"},"outputs":[],"source":["# Params\n","NUM_TIME_STEPS = 50\n","HEIGHT = 28\n","WIDTH = 28\n","INPUT_FEATURES = HEIGHT * WIDTH\n","\n","# Dataset for the benchmark\n","in_distribution_datasets_benchmark = ['MNIST', 'Fashion_MNIST', 'KMNIST', 'Letters']\n","\n","out_of_distribution_datasets_benchmark = ['MNIST', 'Fashion_MNIST', 'KMNIST', \n","                                          'Letters', 'noMNIST', 'omniglot',\n","                                          'CIFAR10-BW', 'MNIST-C/brightness',\n","                                          'MNIST-C/canny_edges',\n","                                          'MNIST-C/dotted_line', 'MNIST-C/fog',\n","                                          'MNIST-C/glass_blur', \n","                                          'MNIST-C/identity',\n","                                          'MNIST-C/impulse_noise', \n","                                          'MNIST-C/motion_blur',\n","                                          'MNIST-C/rotate', 'MNIST-C/scale',\n","                                          'MNIST-C/shear', 'MNIST-C/shot_noise',\n","                                          'MNIST-C/spatter', 'MNIST-C/stripe',\n","                                          'MNIST-C/translate', 'MNIST-C/zigzag',\n","                                          'MNIST_Square']\n","                     \n","out_of_distribution_datasets_benchmark = ['MNIST', 'Fashion_MNIST', 'KMNIST', \n","                                          'Letters', 'noMNIST', 'omniglot',\n","                                          'CIFAR10-BW']                                          \n","\n","# Models to test\n","models_for_benchmark = ['Fully_connected', 'ConvNet']\n","\n","# Hyperparams dict is a dict of a dict of lists. The list are of the shape of\n","# [ hidden_neurons , output_neurons ]\n","hyperparams_dict = {\n","    'Fully_connected': {\n","        'MNIST': [200, 10],\n","        'Fashion_MNIST': [200, 10], \n","        'KMNIST': [200, 10], \n","        'Letters': [200, 26],\n","    },\n","    'ConvNet': {\n","        'MNIST': [300, 10],\n","        'Fashion_MNIST': [300, 10], \n","        'KMNIST': [300, 10], \n","        'Letters': [300, 26],\n","    }\n","}\n","\n","# Number of hidden layers of the model (1 or 2)\n","number_of_hidden_layers = 1"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"s8VSw0YQF1Z0"},"outputs":[],"source":["def my_custom_logger(logger_name, level=logging.INFO):\n","    \"\"\"\n","    Method to return a custom logger with the given name and level\n","    \"\"\"\n","    logger = logging.getLogger(logger_name)\n","    logger.setLevel(level)\n","    format_string = (\"%(asctime)s — %(levelname)s — %(message)s\")\n","    log_format = logging.Formatter(format_string)\n","    # Creating and adding the console handler\n","    #console_handler = logging.StreamHandler(sys.stdout)\n","    #console_handler.setFormatter(log_format)\n","    #logger.addHandler(console_handler)\n","    # Creating and adding the file handler\n","    file_handler = logging.FileHandler(logger_name, mode='w')\n","    file_handler.setFormatter(log_format)\n","    logger.addHandler(file_handler)\n","    return logger"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"aG-h6cQcVHO1"},"outputs":[],"source":["# Create Data directory if it does not exist\n","results_path = Path(\"./Results\")\n","results_path.mkdir(parents=True, exist_ok=True)\n","\n","# Dataframe to store the results\n","COLUMNS = ['Timestamp', 'In-Distribution', 'Out-Distribution', 'Model', \n","           'Training accuracy', 'Accuracy in OOD test set', 'OoD Method',\n","           'AUROC', 'AUPR', 'FPR95', 'FPR80', 'Temperature']\n","df_results = pd.DataFrame(columns=COLUMNS)\n","\n","for in_dataset in tqdm(in_distribution_datasets_benchmark, desc=f'In-Distribution dataset loop'):\n","\n","  # For each in dataset we reset the results list, previously saving the results\n","  # in dataframe and in a checkpoint file\n","  #results_list = []\n","  \n","  # New logger for each In-Distribution Dataset\n","  logger = my_custom_logger(f\"Logger_{in_dataset}.log\")\n","\n","  # Load in-distribution data from the dictionary\n","  train_data, train_loader, test_loader = IN_DISTRIBUTION_DATASETS[in_dataset](BATCH_SIZE)\n","\n","  logger.info(f'Starting In-Distribution dataset {in_dataset}')\n","\n","  for model_name in tqdm(models_for_benchmark, desc='Model loop'):\n","    \n","    logger.info(f'Logs for benchmark with the model {model_name}')\n","\n","    # Load model and its weights\n","    hidden_neurons = hyperparams_dict[model_name][in_dataset][0]\n","    output_neurons = hyperparams_dict[model_name][in_dataset][1]\n","    model = load_model(model_name, DEVICE, hidden_neurons, output_neurons, number_of_hidden_layers)\n","    model.load_state_dict(torch.load(pretrained_weights_path / f'state_dict_{in_dataset}_{model_name}_{hidden_neurons}_{output_neurons}_{}_layers.pth'))\n","    # Show test accuracy and extract the test logits and spikes\n","    model.eval()\n","    test_accuracy, preds_test, logits_test, _spk_count_test = test(model, DEVICE, test_loader, return_logits=True, return_conv_spikes=False)\n","    logger.info(f\"The accuracy of the model with loaded weights of {in_dataset} is {test_accuracy} %\")\n","\n","    # Train subset for creating the clusters\n","    number_of_samples_per_class = 1200\n","    selected_indices_per_class = indices_of_every_class_for_subset(\n","          train_loader,\n","          n_samples_per_class=number_of_samples_per_class,\n","          dataset_name=in_dataset\n","    )\n","    training_subset_clusters = torch.utils.data.Subset(train_data, [x for x in selected_indices_per_class])\n","    subset_train_loader_clusters = torch.utils.data.DataLoader(training_subset_clusters, batch_size=256, shuffle=False)\n","    accuracy_subset_train_clusters, preds_train_clusters, _, _spk_count_train_clusters = test(model, DEVICE, subset_train_loader_clusters, return_logits=True, return_conv_spikes=False)\n","    logger.info(f'Accuracy for the train clusters subset is {accuracy_subset_train_clusters:.3f} %')\n","\n","    # Train subset to create the thresholds\n","    # Introduce a the init_pos parameters to not select the same indices that for\n","    # the subset for creating the clusters\n","    selected_indices_per_class = indices_of_every_class_for_subset(\n","          train_loader,\n","          n_samples_per_class=1000,\n","          dataset_name=in_dataset,\n","          init_pos=number_of_samples_per_class*len(train_loader.dataset.classes)      \n","    )\n","    training_subset = torch.utils.data.Subset(train_data, [x for x in selected_indices_per_class])\n","    subset_train_loader = torch.utils.data.DataLoader(training_subset, batch_size=BATCH_SIZE, shuffle=False)\n","    # Extract the logits and the hidden spikes\n","    accuracy_subset_train, preds_train, logits_train, _spk_count_train = test(model, DEVICE, subset_train_loader, return_logits=True, return_conv_spikes=False)\n","    logger.info(f'Accuracy for the train subset is {accuracy_subset_train_clusters:.3f} %')\n","\n","    # Convert spikes to counts\n","    if isinstance(_spk_count_train_clusters, tuple):\n","      _spk_count_train_clusters, _ = _spk_count_train_clusters\n","      _spk_count_train, _ = _spk_count_train \n","      _spk_count_test, _ = _spk_count_test\n","    # Train\n","    spk_count_train_clusters = np.sum(_spk_count_train_clusters, axis=0, dtype='uint16')\n","    logger.info(f'Train subset for clusters: {spk_count_train_clusters.shape}')\n","    spk_count_train = np.sum(_spk_count_train, axis=0, dtype='uint16')\n","    logger.info(f'Train subset for threshold: {spk_count_train.shape}')\n","    # Test\n","    spk_count_test = np.sum(_spk_count_test, axis=0, dtype='uint16')\n","    logger.info(f'Test set: {spk_count_test.shape}')\n","\n","    # Create clusters\n","    '''\n","    if hidden_neurons == 200:\n","      dist_clustering = (1000,2000)\n","    elif hidden_neurons == 300:\n","      dist_clustering = (1500,2500)\n","    '''\n","    dist_clustering = (500, 5000)\n","\n","    clusters_per_class, logging_info = create_clusters(\n","        subset_train_loader_clusters,\n","        preds_train_clusters,\n","        spk_count_train_clusters,\n","        distance_for_clustering=dist_clustering,\n","        size=1000,\n","        verbose=1)\n","    logger.info(logging_info)\n","\n","    # Initialize for every model, as we save the results for every model\n","    results_list = []\n","\n","    for ood_dataset in tqdm(out_of_distribution_datasets_benchmark, desc='Out-of-Distribution dataset loop'):\n","\n","      logger.info(f'Logs for benchmark with the OoD dataset {ood_dataset}')\n","      \n","      # Load OoD dataset from the dictionary. In case it is MNIST-C, load the \n","      # selected option\n","      if ood_dataset.split('/')[0] == 'MNIST-C':\n","        test_loader_ood = OUT_DISTRIBUTION_DATASETS[ood_dataset.split('/')[0]](\n","            BATCH_SIZE, \n","            test_only=True,\n","            option=ood_dataset.split('/')[1]\n","        )\n","      else:\n","        test_loader_ood = OUT_DISTRIBUTION_DATASETS[ood_dataset](BATCH_SIZE, \n","                                                                 test_only=True)\n","\n","      # Extract the spikes and logits for OoD\n","      accuracy_ood, preds_ood, logits_ood, _spk_count_ood = test(model, DEVICE, test_loader_ood, return_logits=True, return_conv_spikes=False)\n","      logger.info(f'Accuracy for the ood dataset {ood_dataset} is {accuracy_ood:.3f} %')\n","\n","      # Convert spikes to counts\n","      # OoD\n","      if isinstance(_spk_count_ood, tuple):\n","        _spk_count_ood, _ = _spk_count_ood\n","      spk_count_ood = np.sum(_spk_count_ood,axis=0, dtype='uint16')\n","      logger.info(f'OoD set: {spk_count_ood.shape}')\n","\n","      #Create the median aggregations for each cluster of each class\n","      agg_counts_per_class_cluster = average_per_class_and_cluster(spk_count_train_clusters,preds_train_clusters,clusters_per_class,n_samples=1000, option='median')\n","\n","      # Computation of the distances of train, test and ood\n","      distances_train_per_class, _ = distance_to_clusters_averages(spk_count_train, preds_train, agg_counts_per_class_cluster)\n","      distances_test_per_class, _ = distance_to_clusters_averages(spk_count_test, preds_test, agg_counts_per_class_cluster)\n","      distances_ood_per_class, _ = distance_to_clusters_averages(spk_count_ood, preds_ood, agg_counts_per_class_cluster)\n","\n","      ## Metrics ##\n","\n","      # Creation of the array with the thresholds for each TPR (class, dist_per_TPR)\n","      distance_thresholds_train = thresholds_per_class_for_each_TPR(distances_train_per_class)\n","      # Conmputing precision, tpr and fpr\n","      precision, tpr_values, fpr_values = compute_precision_tpr_fpr_for_test_and_ood(distances_test_per_class, distances_ood_per_class,distance_thresholds_train)\n","      # Appending that when FPR = 1 the TPR is also 1:\n","      tpr_values_auroc = np.append(tpr_values,1)\n","      fpr_values_auroc = np.append(fpr_values,1)\n","      # Metrics\n","      auroc = round(np.trapz(tpr_values_auroc, fpr_values_auroc)*100,2)\n","      aupr = round(np.trapz(precision, tpr_values)*100,2)\n","      fpr95 = round(fpr_values_auroc[95]*100,2)\n","      fpr80 = round(fpr_values_auroc[80]*100,2)\n","\n","      # Save results to list\n","      local_time = datetime.datetime.now(pytz.timezone('Europe/Madrid')).ctime()\n","      results_list.append([local_time, in_dataset, ood_dataset, model_name,\n","                           test_accuracy, accuracy_ood, 'Ours',auroc, aupr, fpr95, fpr80, 0.0])\n","\n","      ### Other approaches ### \n","      ## Baseline method ##\n","      softmax_train_winners = np.max(scysp.softmax(logits_train, axis=1),axis=1)\n","      softmax_test_winners = np.max(scysp.softmax(logits_test, axis=1),axis=1)\n","      softmax_ood_winners = np.max(scysp.softmax(logits_ood, axis=1),axis=1)\n","      # Creation of the array with the thresholds for each TPR (class, dist_per_TPR)\n","      likelihood_thresholds_train = thresholds_for_each_TPR_likelihood(softmax_train_winners)\n","      # Conmputing precision, tpr and fpr\n","      precision, tpr_values, fpr_values = likelihood_method_compute_precision_tpr_fpr_for_test_and_ood(softmax_test_winners, softmax_ood_winners, likelihood_thresholds_train)\n","      # Appending that when FPR = 1 the TPR is also 1:\n","      tpr_values_auroc = np.append(tpr_values,1)\n","      fpr_values_auroc = np.append(fpr_values,1)\n","      # Metrics\n","      auroc = round(np.trapz(tpr_values_auroc, fpr_values_auroc)*100,2)\n","      aupr = round(np.trapz(precision, tpr_values)*100,2)\n","      fpr95 = round(fpr_values_auroc[95]*100,2)\n","      fpr80 = round(fpr_values_auroc[80]*100,2)\n","\n","      # Save results to list\n","      local_time = datetime.datetime.now(pytz.timezone('Europe/Madrid')).ctime()\n","      results_list.append([local_time, in_dataset, ood_dataset, model_name,\n","                           test_accuracy,  accuracy_ood, 'Baseline', auroc, aupr, fpr95, fpr80, 0.0])\n","\n","\n","      ## ODIN ## \n","      prelim_results = []\n","      for temp in [1, 10, 100, 1000]:\n","        # Temperature scaling the softmax\n","        temp_softmax_train = scysp.softmax(logits_train/temp,axis=1)\n","        temp_softmax_test = scysp.softmax(logits_test/temp,axis=1)\n","        temp_softmax_ood = scysp.softmax(logits_ood/temp,axis=1)\n","        # Getting only the winners\n","        temp_softmax_train_winners = np.max(temp_softmax_train,axis=1)\n","        temp_softmax_test_winners = np.max(temp_softmax_test,axis=1)\n","        temp_softmax_ood_winners = np.max(temp_softmax_ood,axis=1)\n","        # Creation of the array with the thresholds for each TPR (class, dist_per_TPR)\n","        likelihood_thresholds_train = thresholds_for_each_TPR_likelihood(temp_softmax_train_winners)\n","        # Conmputing precision, tpr and fpr\n","        precision, tpr_values, fpr_values = likelihood_method_compute_precision_tpr_fpr_for_test_and_ood(temp_softmax_test_winners, temp_softmax_ood_winners, likelihood_thresholds_train)\n","        # Appending that when FPR = 1 the TPR is also 1:\n","        tpr_values_auroc = np.append(tpr_values,1)\n","        fpr_values_auroc = np.append(fpr_values,1)\n","        # Metrics\n","        auroc = round(np.trapz(tpr_values_auroc, fpr_values_auroc)*100,2)\n","        aupr = round(np.trapz(precision, tpr_values)*100,2)\n","        fpr95 = round(fpr_values_auroc[95]*100,2)\n","        fpr80 = round(fpr_values_auroc[80]*100,2)\n","        prelim_results.append([auroc, aupr, fpr95, fpr80, temp])\n","      \n","      # Extrac the best result for different temperatures\n","      prelim_results = np.array(prelim_results)\n","      index_max = np.argmax(prelim_results[:,0])\n","      auroc, aupr, fpr95, fpr80, temp = prelim_results[index_max]\n","\n","      # Save results to list\n","      local_time = datetime.datetime.now(pytz.timezone('Europe/Madrid')).ctime()\n","      results_list.append([local_time, in_dataset, ood_dataset, model_name,\n","                           test_accuracy,  accuracy_ood, 'ODIN', auroc, aupr, fpr95, fpr80, temp])\n","\n","\n","      ## Energy based OoD ##\n","      prelim_results = []\n","      for temp in [1,10,100,1000,10000, 100000]:\n","        # Compute the energies \n","        energy_train = -(-temp*torch.logsumexp(torch.Tensor(logits_train) / temp, dim=1)).numpy()\n","        energy_test = -(-temp*torch.logsumexp(torch.Tensor(logits_test) / temp, dim=1)).numpy()\n","        energy_ood = -(-temp*torch.logsumexp(torch.Tensor(logits_ood) / temp, dim=1)).numpy()\n","        # Creation of the array with the thresholds for each TPR (class, dist_per_TPR)\n","        distance_thresholds_train = thresholds_for_each_TPR_likelihood(energy_train)\n","        # Conmputing precision, tpr and fpr\n","        precision, tpr_values, fpr_values = likelihood_method_compute_precision_tpr_fpr_for_test_and_ood(energy_test, energy_ood, distance_thresholds_train)\n","        # Appending that when FPR = 1 the TPR is also 1:\n","        tpr_values_auroc = np.append(tpr_values,1)\n","        fpr_values_auroc = np.append(fpr_values,1)\n","        # Metrics\n","        auroc = round(np.trapz(tpr_values_auroc, fpr_values_auroc)*100, 2)\n","        aupr = round(np.trapz(precision, tpr_values)*100, 2)\n","        fpr95 = round(fpr_values_auroc[95]*100, 2)\n","        fpr80 = round(fpr_values_auroc[80]*100, 2)\n","        prelim_results.append([auroc, aupr, fpr95, fpr80, temp])\n","      \n","      # Extrac the best result for different temperatures\n","      prelim_results = np.array(prelim_results)\n","      index_max = np.argmax(prelim_results[:,0])\n","      auroc, aupr, fpr95, fpr80, temp = prelim_results[index_max]\n","      # Save results to list\n","      local_time = datetime.datetime.now(pytz.timezone('Europe/Madrid')).ctime()\n","      results_list.append([local_time, in_dataset, ood_dataset, model_name,\n","                           test_accuracy,  accuracy_ood, 'Free energy', auroc, aupr, fpr95, fpr80, temp])\n","                \n","\n","    # Save the results in the results list to a dataframe and the save it to \n","    # a file\n","    df_results_one_run = pd.DataFrame(results_list, columns=COLUMNS)\n","    # Save into .csv format\n","    #df_results_one_run.to_csv(results_path / f'checkpoint_{in_dataset}.csv')\n","    df_results = pd.concat([df_results, df_results_one_run])\n","    # Save into .csv\n","    #df_results.to_csv(results_path / f'Benchmark_results.csv')\n","\n","# Save all the results to excel\n","df_results.to_excel('benchmark_results_full.xlsx')"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"cAMkqVtA8ESI"},"outputs":[],"source":["#@markdown ###Process the results to delete redundancies and to separate MNIST-C\n","df_results = pd.read_excel('benchmark_results_full.xlsx', index_col=0)\n","\n","# Eliminate redundancies\n","df_results_no_redundancies = df_results.loc[\n","  df_results['In-Distribution'] != df_results['Out-Distribution']\n","]\n","df_results_no_redundancies.to_excel('benchmark_results.xlsx')"]},{"cell_type":"markdown","metadata":{"id":"L1qSOfqJS9iW"},"source":["## Bayesian Analysis / CD GRAPH"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"NyztyQ7d3b62"},"outputs":[],"source":["#@markdown  ## Set latex fonts for matplotlib\n","!apt-get update\n","!apt install texlive-fonts-recommended texlive-fonts-extra cm-super dvipng\n","# Set LaTex font\n","plt.rcParams.update({\n","    \"text.usetex\": True,\n","    \"font.family\": \"Latin Modern Sans\"})"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"0PFW0YcfbsEf"},"outputs":[],"source":["#@markdown Select the excel to load\n","benchmark_choice = 'No MNIST-C' #@param ['Full','No MNIST-C','Only MNIST-C']\n","model_choice = 'Fully_connected' #@param ['All', 'Fully_connected', 'ConvNet']\n","\n","df_results = pd.read_excel('benchmark_results.xlsx')\n","\n","df_results = df_results.loc[df_results['Out-Distribution'] != 'MNIST_Square']\n","\n","if benchmark_choice == 'Full':\n","  pass\n","elif benchmark_choice == 'No MNIST-C':\n","  df_results = isolate_or_remove_mnistc(df_results, 'Remove')\n","elif benchmark_choice == 'Only MNIST-C':\n","  df_results = isolate_or_remove_mnistc(df_results, 'Isolate')\n","else:\n","  raise ValueError('Wrong option selected')\n","\n","if model_choice == 'All':\n","  pass\n","else:\n","  df_results = isolate_model(df_results, model_choice)"]},{"cell_type":"code","source":["import tutorial.Python.bayesiantests as bt"],"metadata":{"id":"uXn3H8CwxeCa"},"execution_count":null,"outputs":[]},{"cell_type":"code","execution_count":null,"metadata":{"id":"jpeKGc6BZvjm"},"outputs":[],"source":["df_results = df_results.loc[df_results['OoD Method'] != 'JointEnergy']"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"wFqUy1MfiLld","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1663931292966,"user_tz":-120,"elapsed":3,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"}},"outputId":"2c0dd405-0d37-4c92-8cb2-777cfceb4056"},"outputs":[{"output_type":"stream","name":"stdout","text":["['Fashion_MNIST' 'KMNIST' 'Letters' 'noMNIST' 'omniglot' 'CIFAR10-BW'\n"," 'MNIST']\n","['Fully_connected']\n","['Ours' 'Baseline' 'ODIN' 'Free energy']\n"]}],"source":["print(df_results['Out-Distribution'].unique())\n","print(df_results['Model'].unique())\n","print(df_results['OoD Method'].unique())"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"wN31mX8oXCnF","colab":{"base_uri":"https://localhost:8080/","height":150},"executionInfo":{"status":"ok","timestamp":1663930411766,"user_tz":-120,"elapsed":4,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"}},"outputId":"4929a962-8178-4764-87e0-fad4959aa24a"},"outputs":[{"output_type":"display_data","data":{"text/plain":["<Figure size 360x118.8 with 1 Axes>"],"image/png":"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\n"},"metadata":{}}],"source":["#@markdown ###CD Graph\n","from matplotlib.transforms import Bbox\n","from Orange.evaluation import compute_CD, graph_ranks\n","# Create an array of shape [auroc_scores, methods]\n","list_df = []\n","methods_to_test = list(df_results.groupby(df_results['OoD Method']).groups.keys())\n","#methods_to_test[3] = 'SCP'\n","for method in methods_to_test:\n","  list_df.append(df_results.loc[df_results['OoD Method'] == method][['AUROC']].to_numpy()/100)\n","auroc_results_per_method = np.concatenate(list_df, axis=1)\n","methods_to_test[3] = 'SCP' # To change the name in the figure\n","\n","# Compute N and M\n","N = len(methods_to_test) # Models\n","M = np.shape(auroc_results_per_method)[0] # Datasets\n","\n","order = np.argsort(-auroc_results_per_method, axis=1)\n","ranks = np.argsort(order, axis=1)+1\n","avgranks = np.mean(ranks, axis=0)\n","\n","CD = compute_CD(avgranks, M)\n","graph_ranks(avgranks, methods_to_test, cd=CD, width=5, textspace=0.8)\n","plt.savefig(figures_path / f'CD_Graph_{model_choice}.pdf', bbox_inches='tight')\n","#plt.savefig(f'CD_Graph_{model_choice}.pdf', bbox_inches=Bbox([[0.2, 0], [0.8, 1]]))\n","plt.show()"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"5f9V73jF8psA"},"outputs":[],"source":["#@markdown Automated bayesian analysis (Signrank test)\n","scores_dict = OrderedDict()\n","methods_to_test = list(df_results.groupby(df_results['OoD Method']).groups.keys())\n","for method in methods_to_test:\n","  scores_dict[method] = df_results.loc[df_results['OoD Method'] == method][['AUROC']].to_numpy().squeeze()\n","\n","# Select the rope (Region of Practical Equivalence)\n","ROPE = 0.025\n","methods_to_test.remove('Ours')\n","for method in methods_to_test:\n","  scores =  np.array(\n","      [[s1,s2] for (s1,s2) in zip(scores_dict['Ours'], scores_dict[method])]\n","  )\n","\n","  modelLEFT = \"Ours\"\n","  modelRIGHT = method\n","  \n","  names = (modelLEFT, modelRIGHT)\n","\n","  import bayesiantests as bt\n","  left, within, right = bt.signrank(scores/100, rope=ROPE, verbose=True, names=names)\n","  samples = bt.signrank_MC(scores/100, rope=ROPE)\n","  fig = bt.plot_posterior(samples, names)\n","  \"\"\"\n","  if left > right:\n","    fig.set_title(f'Our method is better than {method} the {left*100:.2f}% of the times', y=0.92)\n","  elif left < right:\n","    fig.set_title(f'The {method} method is better than our method the {left*100:.2f}% of the times', y=0.92)\n","  else:\n","    fig.set_title(f'Our method and {method} method are practically equivalent the {left*100:.2f}% of the times', y=0.92)\n","  \"\"\"\n","  plt.savefig(figures_path / f'signrank_test_{method}_{model_choice}_rope{ROPE}_better{left*100:.2f}.pdf', bbox_inches='tight')\n","  plt.show()"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"QT6aIfprFru5","colab":{"base_uri":"https://localhost:8080/","height":1000},"executionInfo":{"status":"ok","timestamp":1663931303268,"user_tz":-120,"elapsed":2492,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"}},"outputId":"e63df29a-301c-4e26-8770-851dda55cf1a","cellView":"form"},"outputs":[{"output_type":"stream","name":"stdout","text":["P(SCP > Baseline) = 0.95336, P(rope) = 0.0, P(Baseline > SCP) = 0.04664\n"]},{"output_type":"display_data","data":{"text/plain":["<Figure size 576x504 with 1 Axes>"],"image/png":"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\n"},"metadata":{"needs_background":"light"}},{"output_type":"stream","name":"stdout","text":["P(SCP > Free energy) = 0.85904, P(rope) = 0.00012, P(Free energy > SCP) = 0.14084\n"]},{"output_type":"display_data","data":{"text/plain":["<Figure size 576x504 with 1 Axes>"],"image/png":"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\n"},"metadata":{"needs_background":"light"}},{"output_type":"stream","name":"stdout","text":["P(SCP > ODIN) = 0.79732, P(rope) = 4e-05, P(ODIN > SCP) = 0.20264\n"]},{"output_type":"display_data","data":{"text/plain":["<Figure size 576x504 with 1 Axes>"],"image/png":"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\n"},"metadata":{"needs_background":"light"}}],"source":["#@markdown Automated bayesian analysis (Signed test)\n","scores_dict = OrderedDict()\n","methods_to_test = list(df_results.groupby(df_results['OoD Method']).groups.keys())\n","for method in methods_to_test:\n","  scores_dict[method] = df_results.loc[df_results['OoD Method'] == method][['AUROC']].to_numpy().squeeze()\n","\n","# Select the rope (Region of Practical Equivalence)\n","ROPE = 0.01\n","methods_to_test.remove('Ours')\n","for method in methods_to_test:\n","  scores =  np.array(\n","      [[s1,s2] for (s1,s2) in zip(scores_dict['Ours'], scores_dict[method])]\n","  )\n","\n","  modelLEFT = \"SCP\"\n","  modelRIGHT = method\n","  \n","  names = (modelLEFT, modelRIGHT)\n","\n","  #import bayesiantests as bt\n","  left, within, right = bt.signtest(scores/100, rope=ROPE, verbose=True, names=names)\n","  samples = bt.signtest_MC(scores/100, rope=ROPE)\n","  fig = bt.plot_posterior(samples,names)\n","  \"\"\"\n","  if left > right:\n","    fig.set_title(f'Our method is better than {method} the {left*100:.2f}% of the times', y=0.92)\n","  elif left < right:\n","    fig.set_title(f'The {method} method is better than our method the {left*100:.2f}% of the times', y=0.92)\n","  else:\n","    fig.set_title(f'Our method and {method} method are practically equivalent the {left*100:.2f}% of the times', y=0.92)\n","  \"\"\"\n","  plt.savefig(figures_path / f'signed_test_{method}_{model_choice}_rope{ROPE}_better{left*100:.2f}.pdf', bbox_inches='tight')\n","  plt.show()"]},{"cell_type":"markdown","metadata":{"id":"a70b3IFg1RHb"},"source":["# Dataset and model selection\n","---"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"8SwDcirNel-P"},"outputs":[],"source":["#@title #Dataset selection\n","#@markdown Click play once decided the in and out distributions to confirm the selection\n","IN_DATASET_NAME = 'MNIST' #@param ['MNIST','Fashion_MNIST','KMNIST', \"Letters\", 'MNIST_Square', 'CIFAR10']\n","OD_DATASET_NAME = 'MNIST_Square' #@param ['MNIST','Fashion_MNIST','KMNIST', 'Letters', 'MNIST_Square','CIFAR10-BW', 'omniglot', 'noMNIST', 'MNIST-C']\n"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"d8XkQY_qfWe3"},"outputs":[],"source":["#@markdown MNIST-C option selection\n","mnist_c_option = 'zigzag' #@param ['zigzag', 'scale', 'identity', 'motion_blur', 'canny_edges', 'glass_blur', 'translate', 'rotate', 'brightness', 'spatter', 'dotted_line', 'impulse_noise', 'fog', 'shear', 'stripe', 'shot_noise']"]},{"cell_type":"markdown","source":["## Model selection\n","\n","Available pretrained weights only for the indicated number of hidden neurons\n","\n","Model type | Number of hidden neurons\n","--- | ---\n","Fully connected | 200\n","ConvNet | 300\n"],"metadata":{"id":"ZYjXMdCmf6OV"}},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"iE1B07pzlzXl","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1663746119801,"user_tz":-120,"elapsed":4471,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"}},"outputId":"3bb0cf73-5630-494d-aec4-88eaa72ada00"},"outputs":[{"output_type":"stream","name":"stdout","text":["Model(\n","  (encoder): PoissonEncoder()\n","  (snn): FC_SNN_1(\n","    (fc1): Linear(in_features=784, out_features=200, bias=False)\n","    (lif1): LIFCell(p=LIFParameters(tau_syn_inv=tensor(200.), tau_mem_inv=tensor(100.), v_leak=tensor(0.), v_th=tensor(0.5000), v_reset=tensor(0.), method='super', alpha=tensor(100)), dt=0.001)\n","    (fc_out): Linear(in_features=200, out_features=10, bias=False)\n","    (out): LICell(p=LIParameters(tau_syn_inv=tensor(200.), tau_mem_inv=tensor(100.), v_leak=tensor(0.)), dt=0.001)\n","  )\n",")\n"]}],"source":["#@markdown Click play once decided to confirm the selection \n","NUM_TIME_STEPS = 50\n","LR = 0.002\n","HEIGHT = 28\n","WIDTH = 28\n","INPUT_FEATURES = HEIGHT * WIDTH\n","selected_model = 'Fully_connected' #@param ['Fully_connected', 'ConvNet']\n","hidden_neurons = 200 #@param {type:\"slider\", min:100, max:500, step:25}\n","number_of_hidden_layers = \"1\" #@param [1, 2]\n","number_of_hidden_layers = int(number_of_hidden_layers)\n","output_neurons = 10\n","if IN_DATASET_NAME == 'Letters':\n","  output_neurons = 26\n","  model = load_model(selected_model, DEVICE, hidden_neurons, output_neurons,\n","                     number_of_hidden_layers)\n","else:\n","  model = load_model(selected_model, DEVICE, hidden_neurons, output_neurons,\n","                     number_of_hidden_layers)\n","optimizer = torch.optim.Adam(model.parameters(), lr=LR)\n","print(model)"]},{"cell_type":"markdown","source":["## Select to Train the model or to Load the pretrained weights (if available)"],"metadata":{"id":"1sDE3sDqhBcb"}},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"T8m6NPLa5ewR","colab":{"base_uri":"https://localhost:8080/","height":457,"referenced_widgets":["2cc4b7e58ff0464d8f39c9f81851d508","449be4d4154d4b9bb7e22e7de8220e80","1a43f874e5aa493aa2a8930ce9064caf","4d8ed3cf1c304ae1b29c67d4bbc18a0f","832d1e6476f749078c9cc95123aba684","ec5f7def061e4411b15334e960ea0b5d","9e6308fa6c7542bdab60efe8155ae443","0f5593c8c2214fdba680ab8739696f1c","52ed85d9daf445098e3550e28dc52e9a","6d7b6701dec343ec82ee0d0a7c46964c","13b00de3c87943c6b3e2abdbe2e1e681","ec72ba0358914a31b9f697448d9854da","b3f95b6a61ea4187a6bc3e83ac446ed5","523cd76739e641ac891646c7530d284a","30f69af25d224aa18508bafb665ea55c","35f2223cf93c4c108ccd054c422abbe7","1bb4959231b44f5c9ad8aa9ab8e1e50e","a051eb15d39940568639ab1ff830618f","3acff4df63894f8fae92e250f26eb6af","5d098b38c2eb48a595ace4d61156cb0b","8efba5efea5a41098a64cb0638522718","1b9659c792fb4c77911bdeacfc2dc516","cbb560dd6e754ff181f36e55614e8983","483bd8b769cb4360954eb615ebcb3852","9ca65c8f48bd43748a406b8156467ee0","cb084fe409804ae0b62e7e6638a0b343","2a59052ee73546f9bb547e4462be7733","4d6e86bf76b64e2f9e9eb479947568c6","332cc880de8e4630bbb08f82c7cd992e","7de526d3f46844f294aed96443dc042d","eb610b88cc904468a71e3602412215d8","4111f0c5c7f84a6ca9a3df82f1214813","1929ad137d334cd6889b9f91ea97435e","c4f3070e663a4e6694466403518b8c17","9e3a1190de6543b0a808e5ef3026e1a3","704fe4fb45804d60adaa974faf78467b","cb4a641d46474738ac3e0fd965e426c5","ba8561b43aaf4e66a12bd4d1987f1039","f0c0ea575e5c4345ba0245dae4d3616c","a40d0197c3e64addb2d059848819c07d","63474d61b6294bc595f04ee32f73787f","4919ac3ddeea456da720acb6836db00c","4102125a74fe4d98a0dae4207554b558","5427913e518846bb82eb0b7923d687dd"]},"executionInfo":{"status":"ok","timestamp":1663746125560,"user_tz":-120,"elapsed":5767,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"}},"outputId":"17eb7bc7-41c7-4df5-9613-9939c8391d38"},"outputs":[{"output_type":"stream","name":"stdout","text":["Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz\n","Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz to /content/Datasets/MNIST/raw/train-images-idx3-ubyte.gz\n"]},{"output_type":"display_data","data":{"text/plain":["  0%|          | 0/9912422 [00:00<?, ?it/s]"],"application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"2cc4b7e58ff0464d8f39c9f81851d508"}},"metadata":{}},{"output_type":"stream","name":"stdout","text":["Extracting /content/Datasets/MNIST/raw/train-images-idx3-ubyte.gz to /content/Datasets/MNIST/raw\n","\n","Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz\n","Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz to /content/Datasets/MNIST/raw/train-labels-idx1-ubyte.gz\n"]},{"output_type":"display_data","data":{"text/plain":["  0%|          | 0/28881 [00:00<?, ?it/s]"],"application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"ec72ba0358914a31b9f697448d9854da"}},"metadata":{}},{"output_type":"stream","name":"stdout","text":["Extracting /content/Datasets/MNIST/raw/train-labels-idx1-ubyte.gz to /content/Datasets/MNIST/raw\n","\n","Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz\n","Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz to /content/Datasets/MNIST/raw/t10k-images-idx3-ubyte.gz\n"]},{"output_type":"display_data","data":{"text/plain":["  0%|          | 0/1648877 [00:00<?, ?it/s]"],"application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"cbb560dd6e754ff181f36e55614e8983"}},"metadata":{}},{"output_type":"stream","name":"stdout","text":["Extracting /content/Datasets/MNIST/raw/t10k-images-idx3-ubyte.gz to /content/Datasets/MNIST/raw\n","\n","Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz\n","Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz to /content/Datasets/MNIST/raw/t10k-labels-idx1-ubyte.gz\n"]},{"output_type":"display_data","data":{"text/plain":["  0%|          | 0/4542 [00:00<?, ?it/s]"],"application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"c4f3070e663a4e6694466403518b8c17"}},"metadata":{}},{"output_type":"stream","name":"stdout","text":["Extracting /content/Datasets/MNIST/raw/t10k-labels-idx1-ubyte.gz to /content/Datasets/MNIST/raw\n","\n","\n","The accuracy of the model with loaded weights of MNIST is 96.15%\n"]}],"source":["load_or_train = 'Load' #@param [\"Load\", \"Train\"]\n","BATCH_SIZE = 256\n","\n","# Load from the dictionary\n","train_data, train_loader, test_loader = IN_DISTRIBUTION_DATASETS[IN_DATASET_NAME](BATCH_SIZE)\n","\n","# Load or create the DF to store the results. Also create the auxiliary list\n","hyperparams = []\n","hyperparams_path = Path(\"./model_hyperparameters.csv\")\n","HYPERPARAMS_COLUMNS = ['In-Dataset', 'Model architecture', 'Number of steps', \n","                       'Hidden neurons', 'Output neurons',\n","                       'Epochs', 'LR', 'Test accuracy']\n","if hyperparams_path.exists():\n","  df_model_hyperparams = pd.read_csv(hyperparams_path,\n","                                     sep=CSV_SEPARATOR, decimal=CSV_DECIMAL)\n","else:\n","  df_model_hyperparams = pd.DataFrame(columns=HYPERPARAMS_COLUMNS)\n","\n","if load_or_train == 'Train':\n","  training_losses   = []\n","  mean_losses       = []\n","  test_losses       = []\n","  accuracies        = []\n","\n","  EPOCHS = 10\n","  # Training loop\n","  for epoch in trange(EPOCHS, desc='Number of epochs', leave=False):\n","      _, mean_training_loss = train(model, DEVICE, train_loader, optimizer, epoch, max_epochs=EPOCHS)\n","      mean_test_loss, accuracy, _ = test(model, DEVICE, test_loader)\n","      training_losses.append(mean_training_loss)\n","      test_losses.append(mean_test_loss)\n","      accuracies.append(accuracy)\n","      print(f\"\\nThe accuracy of the model with {IN_DATASET_NAME} for epoch {epoch+1} is {accuracies[-1]}%\")\n","\n","  #print(f\"\\nThe accuracy of the model with {IN_DATASET_NAME} is {accuracies[-1]}%\")\n","  torch.save(model.state_dict(),f'state_dict_{IN_DATASET_NAME}_{selected_model}_{hidden_neurons}_{output_neurons}_{number_of_hidden_layers}_layers.pth')\n","  plot_loss_history(training_losses, test_losses)\n","\n","  hyperparams = [[IN_DATASET_NAME, MODEL_NAME, NUM_TIME_STEPS, hidden_neurons,\n","                  output_neurons, EPOCHS, LR, accuracy]]\n","  df_one_model_hyperparams = pd.DataFrame(data=hyperparams, columns=HYPERPARAMS_COLUMNS)\n","  df_model_hyperparams = pd.concat([df_model_hyperparams, df_one_model_hyperparams])\n","  df_model_hyperparams.to_csv(hyperparams_path, index=False, sep=CSV_SEPARATOR, \n","                              decimal=CSV_DECIMAL)\n","\n","\n","elif load_or_train == 'Load':\n","  model.load_state_dict(torch.load(pretrained_weights_path / f'state_dict_{IN_DATASET_NAME}_{selected_model}_{hidden_neurons}_{output_neurons}_{number_of_hidden_layers}_layers.pth'))\n","  model.eval()\n","  test_loss, accuracy, _ = test(model, DEVICE, test_loader, 0)\n","  print(f\"\\nThe accuracy of the model with loaded weights of {IN_DATASET_NAME} is {accuracy}%\")\n","\n","else:\n","  raise NameError('Wrong option introduced')"]},{"cell_type":"markdown","metadata":{"id":"0Gjy0XR1KVe5"},"source":["# Spike Count Pattern based detector (SCP)\n","---"]},{"cell_type":"markdown","metadata":{"id":"gq7AqLa_Ve28"},"source":["## Extraction of hidden spikes and logits"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"-pS5d5K7Upgy","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1663746131873,"user_tz":-120,"elapsed":6317,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"}},"outputId":"0b332ffc-8a23-47e3-bc74-5a5c201e33dc"},"outputs":[{"output_type":"stream","name":"stdout","text":["Accuracy for the train subset is 96.533 %\n"]}],"source":["#@markdown ##Training subset to create the clusters\n","number_of_samples_per_class = 1200\n","selected_indices_per_class = indices_of_every_class_for_subset(\n","      train_loader,\n","      n_samples_per_class=number_of_samples_per_class,\n","      dataset_name=IN_DATASET_NAME\n",")\n","training_subset_clusters = torch.utils.data.Subset(train_data, [x for x in selected_indices_per_class])\n","subset_train_loader_clusters = torch.utils.data.DataLoader(training_subset_clusters, batch_size=256, shuffle=False)\n","if selected_model == 'ConvNet':\n","  accuracy_subset_train_clusters, preds_train_clusters, _, _spk_count_train_clusters = test(model, DEVICE, subset_train_loader_clusters, return_logits=True, return_conv_spikes=False)\n","else:\n","  accuracy_subset_train_clusters, preds_train_clusters, _, _spk_count_train_clusters = test(model, DEVICE, subset_train_loader_clusters, return_logits=True, return_conv_spikes=False)\n","\n","print(f'Accuracy for the train subset is {accuracy_subset_train_clusters:.3f} %')"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"nM6V8sMcx4ZR","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1663746134300,"user_tz":-120,"elapsed":2440,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"}},"outputId":"5751fc3b-0ae8-409a-920e-2134c8af8050"},"outputs":[{"output_type":"stream","name":"stdout","text":["Accuracy for the train subset is 96.550 %\n"]}],"source":["#@markdown ##Training subset to create the thresholds to classify ood data\n","# If we have more than 10000 test samples, we limit it to 10000 mantaining a \n","# balance in the number of samples predicted per class\n","\n","# Introduce a the init_pos parameters to not select the same indices that for\n","# the subset for creating the clusters\n","selected_indices_per_class = indices_of_every_class_for_subset(\n","      train_loader,\n","      n_samples_per_class=1000,\n","      dataset_name=IN_DATASET_NAME,\n","      init_pos=number_of_samples_per_class*len(train_loader.dataset.classes)      \n",")\n","training_subset = torch.utils.data.Subset(train_data, [x for x in selected_indices_per_class])\n","subset_train_loader = torch.utils.data.DataLoader(training_subset, batch_size=BATCH_SIZE, shuffle=False)\n","# Extract the logits and the hidden spikes\n","if selected_model == 'ConvNet':\n","  accuracy_subset_train, preds_train, logits_train, _spk_count_train = test(model, DEVICE, subset_train_loader, return_logits=True, return_conv_spikes=False)\n","else:\n","  accuracy_subset_train, preds_train, logits_train, _spk_count_train = test(model, DEVICE, subset_train_loader, return_logits=True, return_conv_spikes=False)\n","print(f'Accuracy for the train subset is {accuracy_subset_train:.3f} %')"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":2318,"status":"ok","timestamp":1663746136614,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"},"user_tz":-120},"id":"mzDfnO524-nv","outputId":"e5196726-95bd-478d-df1a-5939e319f87c"},"outputs":[{"output_type":"stream","name":"stdout","text":["Accuracy for the test set of In-Distribution dataset is 96.160 %\n"]}],"source":["#@markdown ##Test set to obtain the metrics\n","\n","# Extract logits and hidden spikes\n","if selected_model == 'ConvNet':\n","  accuracy_test, preds_test, logits_test, _spk_count_test = test(model, DEVICE, test_loader, return_logits=True, return_conv_spikes=False)\n","else:\n","  accuracy_test, preds_test, logits_test, _spk_count_test = test(model, DEVICE, test_loader, return_logits=True, return_conv_spikes=False)\n","print(f'Accuracy for the test set of In-Distribution dataset is {accuracy_test:.3f} %')"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":2782,"status":"ok","timestamp":1663746139384,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"},"user_tz":-120},"id":"e6IJrCqr70X6","outputId":"ee0d103f-bd69-4685-a48a-9a57cafe3303"},"outputs":[{"output_type":"stream","name":"stdout","text":["Accuracy for the ood dataset MNIST_Square is 38.040 %\n"]}],"source":["#@markdown ##OoD test set\n","# Load from the dictionary\n","if OD_DATASET_NAME == 'MNIST-C':\n","  test_loader_ood = OUT_DISTRIBUTION_DATASETS[OD_DATASET_NAME](\n","      BATCH_SIZE,\n","      test_only=True,\n","      option=mnist_c_option\n","      )\n","else:\n","  test_loader_ood = OUT_DISTRIBUTION_DATASETS[OD_DATASET_NAME](BATCH_SIZE,\n","                                                               test_only=True)\n","\n","# Extract logits and hidden spikes\n","if selected_model == 'ConvNet':\n","  accuracy_ood, preds_ood, logits_ood, _spk_count_ood = test(model, DEVICE, test_loader_ood, return_logits=True, return_conv_spikes=False)\n","else:\n","  accuracy_ood, preds_ood, logits_ood, _spk_count_ood = test(model, DEVICE, test_loader_ood, return_logits=True, return_conv_spikes=False)\n","print(f'Accuracy for the ood dataset {OD_DATASET_NAME} is {accuracy_ood:.3f} %')"]},{"cell_type":"markdown","metadata":{"id":"eyK1Kbn3gVBe"},"source":["## Compute spike counts"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":1069,"status":"ok","timestamp":1663746140449,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"},"user_tz":-120},"id":"wDLav-LMKVky","outputId":"f25e6579-895a-4cbf-9ee4-f022d586ed48"},"outputs":[{"output_type":"stream","name":"stdout","text":["Shape of the arrays\n","--------------------------------------------------\n","Train subset for clusters:\t (12000, 200)\n","Train subset for threshold:\t (10000, 200)\n","Test set:\t\t\t (10000, 200)\n","ood set:\t\t\t (10000, 200)\n"]}],"source":["#@markdown Spike counts of hidden spikes\n","# In case we have returned the convolutional spikes, we will have a tuple \n","# instead of an array, so we unpack the tuple first\n","print('Shape of the arrays')\n","print('-'*50)\n","spk_count_train_clusters = np.sum(_spk_count_train_clusters,axis=0, dtype='uint16')\n","print('Train subset for clusters:\\t',spk_count_train_clusters.shape)\n","spk_count_train = np.sum(_spk_count_train,axis=0, dtype='uint16')\n","print('Train subset for threshold:\\t', spk_count_train.shape)\n","# Test\n","spk_count_test = np.sum(_spk_count_test,axis=0, dtype='uint16')\n","print('Test set:\\t\\t\\t', spk_count_test.shape)\n","# ood\n","spk_count_ood = np.sum(_spk_count_ood,axis=0, dtype='uint16')\n","print('ood set:\\t\\t\\t', spk_count_ood.shape)"]},{"cell_type":"markdown","metadata":{"id":"nKkyGQmew0GW"},"source":["## Training phase of the detector: Creating the clusters"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/","referenced_widgets":["d2e15f0fd13340fa8423867719e4ac4e","8287503040974fcd8b17ff68962f3618","73fe7b1bc0c34976941af9c6416f084b","ca4b077f3333452ba42ffa939e2912b5","295e621961d6484488cbfc149f4b104f","bed292f0a89946768aabb3ff41260b01","47f528c8faaf45ab860056120680142c","2c5d4dac0c4c4522a1b3c36e6c8974f2","c612f1812b2f4ab6ac55bbb950ecb501","b3af29d905cb48f5811bf39c56660bd3","59bd24ba9995475797d7d4fbd4156eac","d059d2b1271141a3815097af381734a1","189824963eaa483faca91024e6d66607","18ae8631dac54d4c8d4cde50d7f6d25f","5752f91c41004a0b9445070508f12f68","c2c5d2fbf46046d896d8d650a40fce2b","019380ad0fbb4dab920c4462522402b6","90d48dc4f10a4624b95d64d893a8609c","5e9634959c2b4c508fbfdd0ea502625f","5e14cbd1a3ce4550ac5c08376bc04dff","cc35cf544e764c78be05f18827f49662","2b85fc5fafc74dd59a0c40d14a6628d4"]},"executionInfo":{"elapsed":64712,"status":"ok","timestamp":1663746205154,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"},"user_tz":-120},"id":"HYLg4Y4DxObV","outputId":"bfaa9640-a65a-4be8-fb57-e7cf4a059b53","cellView":"form"},"outputs":[{"output_type":"display_data","data":{"text/plain":["Computing silhuette score for various distance thresholds:   0%|          | 0/10 [00:00<?, ?it/s]"],"application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"d2e15f0fd13340fa8423867719e4ac4e"}},"metadata":{}},{"output_type":"stream","name":"stdout","text":["Selected distance thresholds:\n"," [1326.5306122448978, 959.1836734693877, 1418.3673469387754, 1510.204081632653, 1142.8571428571427, 1510.204081632653, 1142.8571428571427, 1326.5306122448978, 1602.0408163265306, 1326.5306122448978]\n"]},{"output_type":"display_data","data":{"text/plain":["Create the clusters with the selected distance thresholds: 0it [00:00, ?it/s]"],"application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"d059d2b1271141a3815097af381734a1"}},"metadata":{}},{"output_type":"stream","name":"stdout","text":["\n","Created clusters:\n","---------------------------------------------------------------------------\n","Clase 0 - zero        \t {0: 80, 1: 88, 2: 223, 3: 19, 4: 49, 5: 29, 6: 53, 7: 68, 8: 81, 9: 39, 10: 92, 11: 3, 12: 37, 13: 50, 14: 22, 15: 24, 16: 19, 17: 24}\n","---------------------------------------------------------------------------\n","Clase 1 - one         \t {0: 8, 1: 67, 2: 135, 3: 32, 4: 256, 5: 9, 6: 101, 7: 1, 8: 20, 9: 25, 10: 19, 11: 241, 12: 82, 13: 4}\n","---------------------------------------------------------------------------\n","Clase 2 - two         \t {0: 30, 1: 12, 2: 42, 3: 25, 4: 36, 5: 6, 6: 25, 7: 69, 8: 69, 9: 8, 10: 31, 11: 59, 12: 53, 13: 24, 14: 34, 15: 99, 16: 86, 17: 9, 18: 7, 19: 39, 20: 35, 21: 5, 22: 29, 23: 16, 24: 5, 25: 95, 26: 45, 27: 3, 28: 4}\n","---------------------------------------------------------------------------\n","Clase 3 - three       \t {0: 70, 1: 130, 2: 45, 3: 100, 4: 116, 5: 95, 6: 55, 7: 144, 8: 7, 9: 29, 10: 26, 11: 5, 12: 45, 13: 40, 14: 9, 15: 12, 16: 17, 17: 13, 18: 3, 19: 1, 20: 38}\n","---------------------------------------------------------------------------\n","Clase 4 - four        \t {0: 71, 1: 197, 2: 6, 3: 25, 4: 249, 5: 51, 6: 101, 7: 10, 8: 7, 9: 1, 10: 99, 11: 35, 12: 11, 13: 89, 14: 48}\n","---------------------------------------------------------------------------\n","Clase 5 - five        \t {0: 65, 1: 35, 2: 93, 3: 152, 4: 214, 5: 7, 6: 64, 7: 15, 8: 23, 9: 51, 10: 44, 11: 167, 12: 70}\n","---------------------------------------------------------------------------\n","Clase 6 - six         \t {0: 129, 1: 64, 2: 30, 3: 152, 4: 16, 5: 28, 6: 17, 7: 52, 8: 15, 9: 24, 10: 97, 11: 3, 12: 34, 13: 65, 14: 25, 15: 6, 16: 3, 17: 12, 18: 4, 19: 53, 20: 135, 21: 36}\n","---------------------------------------------------------------------------\n","Clase 7 - seven       \t {0: 78, 1: 79, 2: 336, 3: 101, 4: 144, 5: 18, 6: 25, 7: 23, 8: 48, 9: 1, 10: 26, 11: 18, 12: 42, 13: 14, 14: 45, 15: 2}\n","---------------------------------------------------------------------------\n","Clase 8 - eight       \t {0: 67, 1: 310, 2: 210, 3: 11, 4: 50, 5: 3, 6: 151, 7: 18, 8: 60, 9: 67, 10: 6, 11: 19, 12: 20, 13: 8}\n","---------------------------------------------------------------------------\n","Clase 9 - nine        \t {0: 380, 1: 31, 2: 41, 3: 91, 4: 2, 5: 5, 6: 13, 7: 27, 8: 10, 9: 3, 10: 10, 11: 194, 12: 28, 13: 85, 14: 80}\n","---------------------------------------------------------------------------\n"]},{"output_type":"display_data","data":{"text/plain":["<Figure size 2160x864 with 10 Axes>"],"image/png":"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\n"},"metadata":{"needs_background":"light"}},{"output_type":"display_data","data":{"text/plain":["<Figure size 2160x864 with 10 Axes>"],"image/png":"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\n"},"metadata":{"needs_background":"light"}}],"source":["#@markdown ##Compute the optimal clusters\n","dist_clustering = (500, 5000)\n","\n","clusters_per_class = create_clusters(\n","    subset_train_loader_clusters,\n","    preds_train_clusters, \n","    spk_count_train_clusters, \n","    distance_for_clustering=dist_clustering)"]},{"cell_type":"code","source":["#@markdown ##Create the median aggregations for each cluster of each class\n","agg_counts_per_class_cluster = average_per_class_and_cluster(\n","    spk_count_train_clusters, \n","    preds_train_clusters, \n","    clusters_per_class,\n","    n_samples=1000,\n","    option='median'\n",")\n","print(\"Median aggregations computed succesfully!\")"],"metadata":{"cellView":"form","id":"MnsOV9tPhZ2a","executionInfo":{"status":"ok","timestamp":1663746205154,"user_tz":-120,"elapsed":26,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"}},"colab":{"base_uri":"https://localhost:8080/"},"outputId":"125fe39a-5a7a-4613-e8e4-2760e1e247b4"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["Median aggregations computed succesfully!\n"]}]},{"cell_type":"markdown","metadata":{"id":"hKJZIvKo987S"},"source":["## Testing the detector"]},{"cell_type":"code","source":["#@title Computation of the distances of train, test and ood\n","distances_train_per_class, closest_clusters_per_cls_train = distance_to_clusters_averages(spk_count_train,preds_train,agg_counts_per_class_cluster)\n","distances_test_per_class, closest_clusters_per_cls_test = distance_to_clusters_averages(spk_count_test,preds_test,agg_counts_per_class_cluster)\n","distances_ood_per_class, closest_clusters_per_cls_ood = distance_to_clusters_averages(spk_count_ood,preds_ood,agg_counts_per_class_cluster)"],"metadata":{"cellView":"form","id":"Nd10cU7ahgoO"},"execution_count":null,"outputs":[]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":3576,"status":"ok","timestamp":1663746209225,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"},"user_tz":-120},"id":"wRPCz8qf-MJq","outputId":"e3a39d8f-0cb1-4847-f9d3-85f264a51c64"},"outputs":[{"output_type":"display_data","data":{"text/plain":["<Figure size 2160x576 with 10 Axes>"],"image/png":"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\n"},"metadata":{"needs_background":"light"}}],"source":["#@title Classwise distance distribution\n","# Fuction to plot classwise distances\n","def plot_classwise_distances_to_cluster_averages(dist_train, dist_test, dist_ood, preds_ood, n_classes):\n","  if n_classes == 10:\n","    rows = 2\n","    cols = 5\n","  elif n_classes == 26:\n","    rows = 2\n","    cols = 13\n","  else:\n","    raise ValueError('Plot not implemented for that number of classes')\n","\n","  fig, axes = plt.subplots(rows, cols, figsize=(6*cols, 4*rows),\n","                             tight_layout=True)\n","  for class_index, ax in enumerate(axes.flat):\n","    n_samples = len(dist_ood[class_index])\n","    ood_bins = n_samples+1 if n_samples < 25 else 25\n","    ax.hist(dist_train[class_index],bins=25,color='dodgerblue',density=True, label='Train')\n","    ax.hist(dist_test[class_index],bins=25,color='green',alpha=0.6,density=True, label='Test')\n","    ax.hist(dist_ood[class_index],bins=ood_bins,color='darkorange',alpha=0.6,density=True, label='ood')\n","    ax.set_title(f'{train_data.classes[class_index]} // Number of samples ood = {len(preds_ood[searchIndicesOfClass(class_index,preds_ood)])}')\n","    ax.legend()\n","  return fig\n","\n","fig = plot_classwise_distances_to_cluster_averages(\n","    distances_train_per_class, distances_test_per_class,\n","    distances_ood_per_class, preds_ood, len(train_loader.dataset.classes)\n","    )\n","#my_suptitle = fig.suptitle(f'{IN_DATASET_NAME} vs {OD_DATASET_NAME}', y=1.03, fontsize=15)\n","#fig.savefig(f'{IN_DATASET_NAME}_vs_{OD_DATASET_NAME}_Classwise_Distribution.pdf',\n","            #dpi=250, bbox_inches='tight', bbox_extra_artists=[my_suptitle])"]},{"cell_type":"markdown","metadata":{"id":"gWvYV1QHpP5-"},"source":["## Metrics"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","executionInfo":{"elapsed":959,"status":"ok","timestamp":1663746210179,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"},"user_tz":-120},"id":"rs2czO_44vqZ","colab":{"base_uri":"https://localhost:8080/"},"outputId":"95f39693-aef0-457f-ee8c-d278cc506abd"},"outputs":[{"output_type":"stream","name":"stderr","text":["/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:60: RuntimeWarning: invalid value encountered in true_divide\n"]}],"source":["#@markdown ##Threstholds calculation\n","# Creation of the array with the thresholds for each TPR (class, dist_per_TPR)\n","distance_thresholds_train = thresholds_per_class_for_each_TPR(distances_train_per_class)\n","# Conmputing precision, tpr and fpr\n","precision, tpr_values, fpr_values = compute_precision_tpr_fpr_for_test_and_ood(distances_test_per_class, distances_ood_per_class,distance_thresholds_train)\n","# Appending that when FPR = 1 the TPR is also 1:\n","tpr_values_auroc = np.append(tpr_values,1)\n","fpr_values_auroc = np.append(fpr_values,1)"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":8,"status":"ok","timestamp":1663746210180,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"},"user_tz":-120},"id":"5pG261OLpWPs","outputId":"f73b59e0-bb13-4591-bea4-165d9d75d5da"},"outputs":[{"output_type":"stream","name":"stdout","text":["------------------------------\n","Results\n","------------------------------\n","AUROC = 95.825 %\n","AUPR  = 96.477 %\n","------------------------------ \n","\n"]},{"output_type":"display_data","data":{"text/plain":["<Figure size 1080x864 with 1 Axes>"],"image/png":"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\n"},"metadata":{"needs_background":"light"}},{"output_type":"display_data","data":{"text/plain":["<Figure size 1080x864 with 1 Axes>"],"image/png":"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\n"},"metadata":{"needs_background":"light"}}],"source":["#@title AUROC and AUPR\n","#AUROC\n","auroc = np.trapz(tpr_values_auroc, fpr_values_auroc)\n","\n","# Plot\n","plt.figure(figsize=(15,12))\n","roc = plt.plot(fpr_values_auroc,tpr_values_auroc,label='ROC curve', lw=3)\n","rnd_roc = plt.plot(np.linspace(0,1,100),np.linspace(0,1,100),'k--', label='Random ROC curve')\n","plt.xticks(fontsize=15)\n","plt.yticks(fontsize=15)\n","plt.xlabel('FPR',fontsize=20)\n","plt.ylabel('TPR',fontsize=20)\n","plt.title(f'ROC curve, AUC = {auroc*100:.3f}%', fontsize=25, pad=10)\n","plt.fill_between(fpr_values_auroc, tpr_values_auroc, alpha=0.3)\n","# Create empty plot with blank marker containing the extra label\n","plt.plot([], [], ' ', label='FPR at 95% TPR = {}%'.format(round(fpr_values_auroc[95]*100,2)))\n","plt.plot([], [], ' ', label='FPR at 80% TPR = {}%'.format(round(fpr_values_auroc[80]*100,2)))\n","#plt.text(0.60,0.975,'FPR at 95% TPR = {}%'.format(round(array_TPR_FPR_x_threshold[95,1]*100,2)),fontsize=20,bbox=dict(boxstyle=\"round\",facecolor='white', alpha=0.5))\n","plt.legend(fontsize=20,loc='upper left')\n","#plt.axhline(y=0.95,linestyle='-',zorder=0)\n","#plt.axvline(x=array_TPR_FPR_x_threshold[95,1],linestyle='dashed',zorder=0)\n","#plt.savefig('/content/AUROC')\n","plt.savefig(f'/content/AUROC_{IN_DATASET_NAME}_vs_{OD_DATASET_NAME}.pdf')\n","\n","# AUPR\n","aupr = np.trapz(precision, tpr_values)\n","# Plot\n","plt.figure(figsize=(15,12))\n","roc = plt.plot(tpr_values,precision,label='ROC curve',lw=3)\n","plt.xticks(fontsize=15)\n","plt.yticks(fontsize=15)\n","plt.xlabel('FPR',fontsize=20)\n","plt.ylabel('Precision',fontsize=20)\n","plt.title(f'PR curve, AUC = {aupr*100:.3f} %', fontsize=25, pad=10)\n","plt.fill_between(tpr_values,precision,alpha=0.3)\n","plt.legend(fontsize=20,loc='upper left')\n","plt.savefig(f'/content/AUPR_{IN_DATASET_NAME}_vs_{OD_DATASET_NAME}.pdf')\n","\n","print('-'*30)\n","print(f'Results')\n","print('-'*30)\n","print(f'AUROC = {auroc*100:.3f} %')\n","print(f'AUPR  = {aupr*100:.3f} %')\n","print('-'*30,'\\n')"]},{"cell_type":"markdown","metadata":{"id":"ReezkOBfj1-1"},"source":["# OoD detection benchmark"]},{"cell_type":"markdown","metadata":{"id":"XH0lK8PotW7p"},"source":["To change between different datasets:\n","\n","1. You must go to the training stage, select the In and Out datasets, and retrain the model or load it with the corresponding weights\n","2. Recompute the Spike and logits extraction\n","3. Compute all the cells of the OoD method you want to test again"]},{"cell_type":"markdown","metadata":{"id":"sBO1GtqdQxwO"},"source":["## Baseline method"]},{"cell_type":"markdown","metadata":{"id":"sypujNlqJneO"},"source":["DISCLAIMER: \n","\n","---\n","\n","This method is inspired by the baseline method (http://arxiv.org/abs/1610.02136) but does not follow the exact implementation of the paper, as we want the method to be comparable with ours. Therefore, as we don't make usage of prior knowledge of the predicted class, in our implementation of the baseline method we don't make the distinction between correctly and incorrectly classified samples."]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"dZz_7_efYZ48"},"outputs":[],"source":["#@markdown Max softmax probailities\n","# Winners\n","softmax_train_winners = np.max(scysp.softmax(logits_train, axis=1),axis=1)\n","softmax_test_winners = np.max(scysp.softmax(logits_test, axis=1),axis=1)\n","softmax_ood_winners = np.max(scysp.softmax(logits_ood, axis=1),axis=1)"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","colab":{"base_uri":"https://localhost:8080/","height":369},"executionInfo":{"elapsed":4,"status":"ok","timestamp":1660299614993,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"},"user_tz":-120},"id":"nVfo8DR2kEbc","outputId":"3474229e-717d-4151-a34e-f67c71f65561"},"outputs":[{"output_type":"display_data","data":{"text/plain":["<Figure size 720x360 with 1 Axes>"],"image/png":"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\n"},"metadata":{"needs_background":"light"}}],"source":["#@markdown Histogram of softmax probabilities of train, test and ood\n","fig = plt.figure(figsize=(10,5),tight_layout=True)\n","plt.hist(softmax_train_winners,bins=50,color='blue',alpha=0.6,density=True, label='Train')\n","plt.hist(softmax_test_winners,bins=50,color='green',alpha=0.6,density=True, label='Test')\n","plt.hist(softmax_ood_winners,bins=50,color='darkorange',alpha=0.6,density=True, label='ood')\n","plt.ylim([0,10])\n","plt.legend(fontsize=18)\n","plt.show()"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","colab":{"base_uri":"https://localhost:8080/","height":1000},"executionInfo":{"elapsed":6,"status":"ok","timestamp":1660299620075,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"},"user_tz":-120},"id":"iki3dKP30q6a","outputId":"0ee4460b-759d-4d3f-ccc6-950c8b3eceea"},"outputs":[{"output_type":"stream","name":"stderr","text":["/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:160: RuntimeWarning: invalid value encountered in true_divide\n"]},{"output_type":"display_data","data":{"text/plain":["<Figure size 1080x864 with 1 Axes>"],"image/png":"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\n"},"metadata":{"needs_background":"light"}},{"output_type":"display_data","data":{"text/plain":["<Figure size 1080x864 with 1 Axes>"],"image/png":"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\n"},"metadata":{"needs_background":"light"}}],"source":["#@markdown ##Metrics\n","# Creation of the array with the thresholds for each TPR (class, dist_per_TPR)\n","likelihood_thresholds_train = thresholds_for_each_TPR_likelihood(softmax_train_winners)\n","# Conmputing precision, tpr and fpr\n","precision, tpr_values, fpr_values = likelihood_method_compute_precision_tpr_fpr_for_test_and_ood(softmax_test_winners, softmax_ood_winners, likelihood_thresholds_train)\n","# Appending that when FPR = 1 the TPR is also 1:\n","tpr_values_auroc = np.append(tpr_values,1)\n","fpr_values_auroc = np.append(fpr_values,1)\n","\n","# Plot AUROC\n","plot_AUROC(fpr_values_auroc, tpr_values_auroc)\n","\n","# Plot AUPR\n","plot_AUPR(precision, tpr_values)"]},{"cell_type":"markdown","metadata":{"id":"W61F95XALkny"},"source":["## ODIN"]},{"cell_type":"markdown","source":["### Select the temperature manually"],"metadata":{"id":"QlzmJgYjXx-J"}},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"P59tdVY7MtLd"},"outputs":[],"source":["#@markdown Temperature scaling the softmax\n","temp= 1 #@param {type:\"number\"}\n","temp_softmax_train = scysp.softmax(logits_train/temp,axis=1)\n","temp_softmax_test = scysp.softmax(logits_test/temp,axis=1)\n","temp_softmax_ood = scysp.softmax(logits_ood/temp,axis=1)\n","# Getting only the winners\n","temp_softmax_train_winners = np.max(temp_softmax_train,axis=1)\n","temp_softmax_test_winners = np.max(temp_softmax_test,axis=1)\n","temp_softmax_ood_winners = np.max(temp_softmax_ood,axis=1)"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"vDnjCxpQN30R"},"outputs":[],"source":["#@markdown Histogram of softmax probabilities of train, test and ood\n","fig = plt.figure(figsize=(10,5),tight_layout=True)\n","plt.hist(temp_softmax_train_winners,bins=50,color='blue',alpha=0.6,density=True, label='Train')\n","plt.hist(temp_softmax_test_winners,bins=50,color='green',alpha=0.6,density=True, label='Test')\n","plt.hist(temp_softmax_ood_winners,bins=50,color='darkorange',alpha=0.6,density=True, label='ood')\n","#plt.ylim([0,10])\n","plt.legend(fontsize=18)\n","plt.show()"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"a_LIaZUBOSCZ"},"outputs":[],"source":["#@markdown Metrics\n","# Execute the functions\n","# Creation of the array with the thresholds for each TPR (class, dist_per_TPR)\n","likelihood_thresholds_train = thresholds_for_each_TPR_likelihood(temp_softmax_train_winners)\n","# Conmputing precision, tpr and fpr\n","precision, tpr_values, fpr_values = likelihood_method_compute_precision_tpr_fpr_for_test_and_ood(temp_softmax_test_winners, temp_softmax_ood_winners, likelihood_thresholds_train)\n","# Appending that when FPR = 1 the TPR is also 1:\n","tpr_values_auroc = np.append(tpr_values,1)\n","fpr_values_auroc = np.append(fpr_values,1)\n","\n","# Plot AUROC\n","plot_AUROC(fpr_values_auroc, tpr_values_auroc)\n","\n","# Plot AUPR\n","plot_AUPR(precision, tpr_values)"]},{"cell_type":"markdown","source":["### Testing metrics for several temperatures"],"metadata":{"id":"oYU2dqkNX4Qp"}},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":1394,"status":"ok","timestamp":1660299633248,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"},"user_tz":-120},"id":"ubrt7qDoTTTV","outputId":"df9035c3-e861-4fc2-9140-3c1567c3946f"},"outputs":[{"output_type":"stream","name":"stdout","text":["------------------------------\n","Results for temperature = 1\n","------------------------------\n","AUROC = 63.288 %\n","AUPR  = 59.947 %\n","------------------------------ \n","\n","------------------------------\n","Results for temperature = 10\n","------------------------------\n","AUROC = 61.651 %\n","AUPR  = 57.896 %\n","------------------------------ \n","\n","------------------------------\n","Results for temperature = 100\n","------------------------------\n","AUROC = 61.169 %\n","AUPR  = 57.550 %\n","------------------------------ \n","\n","------------------------------\n","Results for temperature = 1000\n","------------------------------\n","AUROC = 61.124 %\n","AUPR  = 57.509 %\n","------------------------------ \n","\n","------------------------------\n","Results for temperature = 10000\n","------------------------------\n","AUROC = 61.119 %\n","AUPR  = 57.505 %\n","------------------------------ \n","\n"]},{"output_type":"stream","name":"stderr","text":["/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:160: RuntimeWarning: invalid value encountered in true_divide\n"]}],"source":["#@markdown AUROC and AUPR for different temperatures\n","for temp in [1, 10, 100, 1000, 10000]:\n","  # Temperature scaling the softmax\n","  temp_softmax_train = scysp.softmax(logits_train/temp,axis=1)\n","  temp_softmax_test = scysp.softmax(logits_test/temp,axis=1)\n","  temp_softmax_ood = scysp.softmax(logits_ood/temp,axis=1)\n","  # Getting only the winners\n","  temp_softmax_train_winners = np.max(temp_softmax_train,axis=1)\n","  temp_softmax_test_winners = np.max(temp_softmax_test,axis=1)\n","  temp_softmax_ood_winners = np.max(temp_softmax_ood,axis=1)\n","  #fig = plt.figure(figsize=(10,5),tight_layout=True)\n","  #plt.hist(temp_softmax_train_winners,bins=50,color='blue',alpha=0.6,density=True, label='Train')\n","  #plt.hist(temp_softmax_test_winners,bins=50,color='green',alpha=0.6,density=True, label='Test')\n","  #plt.hist(temp_softmax_ood_winners,bins=50,color='darkorange',alpha=0.6,density=True, label='ood')\n","  #plt.ylim([0,10])\n","  #plt.legend(fontsize=18)\n","  #plt.show()\n","  # Execute the functions\n","  # Creation of the array with the thresholds for each TPR (class, dist_per_TPR)\n","  likelihood_thresholds_train = thresholds_for_each_TPR_likelihood(temp_softmax_train_winners)\n","  # Conmputing precision, tpr and fpr\n","  precision, tpr_values, fpr_values = likelihood_method_compute_precision_tpr_fpr_for_test_and_ood(temp_softmax_test_winners, temp_softmax_ood_winners, likelihood_thresholds_train)\n","  # Appending that when FPR = 1 the TPR is also 1:\n","  tpr_values_auroc = np.append(tpr_values,1)\n","  fpr_values_auroc = np.append(fpr_values,1)\n","\n","  # Plot AUROC\n","  #plot_AUROC(fpr_values_auroc, tpr_values_auroc)\n","\n","  # Plot AUPR\n","  #plot_AUPR(precision, tpr_values)\n","\n","  # AUROC\n","  auroc = np.trapz(tpr_values_auroc, fpr_values_auroc)\n","\n","  # AUPR\n","  aupr = np.trapz(precision, tpr_values)\n","\n","  #results.append([temp, auroc, aupr])\n","\n","  print('-'*30)\n","  print(f'Results for temperature = {temp}')\n","  print('-'*30)\n","  print(f'AUROC = {auroc*100:.3f} %')\n","  print(f'AUPR  = {aupr*100:.3f} %')\n","  print('-'*30,'\\n')\n"]},{"cell_type":"markdown","metadata":{"id":"M3HxHd4WRceW"},"source":["## Energy-based ood detection"]},{"cell_type":"markdown","source":["### Select the temperature manually"],"metadata":{"id":"otAKfNoLYkSx"}},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"j95lqJPoRcYN"},"outputs":[],"source":["#@markdown Compute energy\n","temp = 1 #@param {type:\"number\"}\n","energy_train = -(-temp*torch.logsumexp(torch.Tensor(logits_train) / temp, dim=1)).numpy()\n","energy_test = -(-temp*torch.logsumexp(torch.Tensor(logits_test) / temp, dim=1)).numpy()\n","energy_ood = -(-temp*torch.logsumexp(torch.Tensor(logits_ood) / temp, dim=1)).numpy()"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"hadFVJi3Vra0"},"outputs":[],"source":["#@markdown Plot histogram of the free energy\n","plot_histogram(energy_train, energy_test, energy_ood)"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"sZOuf18XWYnD"},"outputs":[],"source":["#@markdown ##Metrics\n","# Execute the functions\n","# Creation of the array with the thresholds for each TPR (class, dist_per_TPR)\n","distance_thresholds_train = thresholds_for_each_TPR_likelihood(energy_train)\n","# Conmputing precision, tpr and fpr\n","precision, tpr_values, fpr_values = likelihood_method_compute_precision_tpr_fpr_for_test_and_ood(energy_test, energy_ood, distance_thresholds_train)\n","# Appending that when FPR = 1 the TPR is also 1:\n","tpr_values_auroc = np.append(tpr_values,1)\n","fpr_values_auroc = np.append(fpr_values,1)\n","\n","# Plot AUROC\n","plot_AUROC(fpr_values_auroc, tpr_values_auroc)\n","\n","# Plot AUPR\n","plot_AUPR(precision, tpr_values)"]},{"cell_type":"markdown","metadata":{"id":"sL0zr9BJdvVE"},"source":["### Testing metrics for several temperatures"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":6,"status":"ok","timestamp":1660299639802,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"},"user_tz":-120},"id":"hxC6fr8lcF7S","outputId":"f4230967-cf91-4e74-f995-45ea91d421b9","cellView":"form"},"outputs":[{"output_type":"stream","name":"stdout","text":["------------------------------\n","Results for temperature = 1\n","------------------------------\n","AUROC = 60.103 %\n","AUPR  = 56.781 %\n","------------------------------ \n","\n","------------------------------\n","Results for temperature = 10\n","------------------------------\n","AUROC = 53.111 %\n","AUPR  = 51.763 %\n","------------------------------ \n","\n","------------------------------\n","Results for temperature = 100\n","------------------------------\n","AUROC = 45.346 %\n","AUPR  = 47.940 %\n","------------------------------ \n","\n","------------------------------\n","Results for temperature = 1000\n","------------------------------\n","AUROC = 44.703 %\n","AUPR  = 47.513 %\n","------------------------------ \n","\n","------------------------------\n","Results for temperature = 10000\n","------------------------------\n","AUROC = 44.637 %\n","AUPR  = 47.444 %\n","------------------------------ \n","\n","------------------------------\n","Results for temperature = 100000\n","------------------------------\n","AUROC = 44.611 %\n","AUPR  = 47.425 %\n","------------------------------ \n","\n"]},{"output_type":"stream","name":"stderr","text":["/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:160: RuntimeWarning: invalid value encountered in true_divide\n"]}],"source":["#@markdown AUROC and AUPR for different temperatures\n","results = []\n","temperatures = [1,10,100,1000,10000, 100000]\n","for temp in temperatures:\n","  # Compute the energies \n","  energy_train = -(-temp*torch.logsumexp(torch.Tensor(logits_train) / temp, dim=1)).numpy()\n","  energy_test = -(-temp*torch.logsumexp(torch.Tensor(logits_test) / temp, dim=1)).numpy()\n","  energy_ood = -(-temp*torch.logsumexp(torch.Tensor(logits_ood) / temp, dim=1)).numpy()\n","\n","  # Creation of the array with the thresholds for each TPR (class, dist_per_TPR)\n","  distance_thresholds_train = thresholds_for_each_TPR_likelihood(energy_train)\n","  # Conmputing precision, tpr and fpr\n","  precision, tpr_values, fpr_values = likelihood_method_compute_precision_tpr_fpr_for_test_and_ood(energy_test, energy_ood, distance_thresholds_train)\n","  # Appending that when FPR = 1 the TPR is also 1:\n","  tpr_values_auroc = np.append(tpr_values,1)\n","  fpr_values_auroc = np.append(fpr_values,1)\n","\n","  # AUROC\n","  auroc = np.trapz(tpr_values_auroc, fpr_values_auroc)\n","\n","  # AUPR\n","  aupr = np.trapz(precision, tpr_values)\n","\n","  results.append([temp, auroc, aupr])\n","# Print results\n","for res in results:\n","  print('-'*30)\n","  print(f'Results for temperature = {res[0]}')\n","  print('-'*30)\n","  print(f'AUROC = {res[1]*100:.3f} %')\n","  print(f'AUPR  = {res[2]*100:.3f} %')\n","  print('-'*30,'\\n')"]},{"cell_type":"markdown","metadata":{"id":"IPet7FyD9ryM"},"source":["# Explainability"]},{"cell_type":"markdown","source":["Instructions:\n","- Set the LaTeX fonts for the plots if you want using the cell below\n","- Compute the prerequisites to enable the plots to work\n","- The figure of the paper has its own prerequisites in its section befor the plot\n"],"metadata":{"id":"oGXNbgYrklg7"}},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"RHziFVYoDJVr"},"outputs":[],"source":["#@markdown  ## Set LaTeX fonts for matplotlib\n","!apt-get update\n","!apt install texlive-fonts-recommended texlive-fonts-extra cm-super dvipng\n","# Set LaTex font\n","plt.rcParams.update({\n","    \"text.usetex\": True,\n","    \"font.family\": \"Latin Modern Sans\"})"]},{"cell_type":"markdown","source":["## Prerequisites"],"metadata":{"id":"64ptWxaSwslF"}},{"cell_type":"code","source":["#@markdown ##Retrieve weigths from the model\n","if number_of_hidden_layers == 1:\n","  if selected_model == 'ConvNet':\n","    weights = model.snn.state_dict()['fc1.weight'].cpu().numpy()\n","  elif selected_model == 'Fully_connected':\n","    weights = model.state_dict()['snn.fc1.weight'].cpu().numpy()\n","  else:\n","    raise NameError\n","\n","\n","elif number_of_hidden_layers > 1:\n","  if selected_model == 'ConvNet':\n","    weights = [\n","              model.snn.state_dict()['fc2.weight'].cpu().numpy(),\n","              model.snn.state_dict()['fc1.weight'].cpu().numpy(),\n","              ]\n","  elif selected_model == 'Fully_connected':\n","    weights = [\n","              model.snn.state_dict()['fc2.weight'].cpu().numpy(),\n","              model.snn.state_dict()['fc1.weight'].cpu().numpy(),\n","              ]\n","  else:\n","    raise NameError\n","\n","else:\n","  raise ValueError('Wrong number of layers')"],"metadata":{"cellView":"form","id":"qMnkYhTUzAnX"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["#@markdown ##Computation of the distances of train, test and ood \n","distances_train_per_class, closest_clusters_per_cls_train = distance_to_clusters_averages(spk_count_train,preds_train,agg_counts_per_class_cluster)\n","distances_test_per_class, closest_clusters_per_cls_test = distance_to_clusters_averages(spk_count_test,preds_test,agg_counts_per_class_cluster)\n","distances_ood_per_class, closest_clusters_per_cls_ood = distance_to_clusters_averages(spk_count_ood,preds_ood,agg_counts_per_class_cluster)"],"metadata":{"cellView":"form","id":"0rSR027XZF1t"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["#@markdown ##Create the reconstruction\n","# Compute reconst\n","reconst_test = compute_reconstruction(spk_count_test, weights)\n","reconst_ood = compute_reconstruction(spk_count_ood, weights)\n","# Extract positive part\n","reconst_test = np.where(reconst_test >= 0, reconst_test, 0)\n","reconst_ood = np.where(reconst_ood >= 0, reconst_ood, 0)\n","\n","### Backprop the hidden counts of the median ###\n","agg_counts_reconstructed_per_class_and_cluster = compute_reconstruction_per_class(agg_counts_per_class_cluster, weights)\n","# Extract the positive part\n","avg_frecs_reconst_per_cl_clu_positive = extract_positive_part_per_class(agg_counts_reconstructed_per_class_and_cluster)\n","\n","if selected_model == 'ConvNet':\n","  # Rearrange\n","  reconst_test = rearrange_to_ftmaps(reconst_test, ftmaps_shape=(50, 11, 11))\n","  reconst_ood = rearrange_to_ftmaps(reconst_ood, ftmaps_shape=(50, 11, 11))\n","  avg_frecs_reconst_per_cl_clu_positive = rearrange_to_ftmaps_per_class(avg_frecs_reconst_per_cl_clu_positive, (50,11,11))\n","\n","# Reorder to be a list of reconstructions per class\n","reconst_per_class_test = order_array_of_samples_to_per_class_list(reconst_test, preds_test)\n","reconst_per_class_ood = order_array_of_samples_to_per_class_list(reconst_ood, preds_ood)"],"metadata":{"cellView":"form","id":"SEgYOoTyZfkH"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["#@markdown ##Show AUROC and AUPR and extract distance thresholds\n","thresholds = auroc_aupr(distances_train_per_class, distances_test_per_class, distances_ood_per_class)"],"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"id":"JBPV8bL_hdyt","executionInfo":{"status":"ok","timestamp":1663757807740,"user_tz":-120,"elapsed":477,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"}},"outputId":"876cb01e-e075-4fd2-e36a-83801198c3fd","cellView":"form"},"execution_count":null,"outputs":[{"output_type":"stream","name":"stdout","text":["------------------------------\n","Results\n","------------------------------\n","AUROC = 95.825 %\n","AUPR  = 96.477 %\n","------------------------------\n","FPR at 95% TPR: 23.51%\n","Threshold mean and std for all classes at 95% TPR: 701.90, 113.17\n","------------------------------ \n","\n"]},{"output_type":"stream","name":"stderr","text":["/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:60: RuntimeWarning: invalid value encountered in true_divide\n"]}]},{"cell_type":"code","source":["#@markdown ##Define the plot range\n","# To define the plot range we take 100 examples of test and process them to\n","# obtain the attribution. Then, we take the quantile 80 of all the pixel values\n","# in the 100 examples taken. \n","array_histrogram = np.zeros((100, 28, 28), dtype='float32')\n","for ind, i in tqdm(enumerate(np.random.randint(0, 10000, 100))):\n","  prediction = preds_test[i]\n","  # Obtain the index of the represented number in the per class order to extract the closest cluster (as it is in per class order)\n","  i_in_per_cls_order = np.where(searchIndicesOfClass(prediction, preds_test) == np.array(i))[0][0]\n","  if selected_model == 'ConvNet':\n","    array_histrogram[ind] = resize(np.sum(np.abs(reconst_test[i] - avg_frecs_reconst_per_cl_clu_positive[prediction][closest_clusters_per_cls_test[prediction][i_in_per_cls_order]]), axis=0), (28, 28))\n","  elif selected_model == 'Fully_connected':\n","    array_histrogram[ind] = np.abs(reconst_test[i] - avg_frecs_reconst_per_cl_clu_positive[prediction][closest_clusters_per_cls_test[prediction][i_in_per_cls_order]]).reshape(28,28)\n","  else:\n","    raise NameError\n","plot_range = [int(np.quantile(array_histrogram.flatten(), 0.90)), int(np.max(array_histrogram.flatten()))]\n","print(\"Plot range for heatmaps:\", plot_range)\n","\n","array_histrogram = np.zeros((100, 28, 28), dtype='float32')\n","for ind, i in tqdm(enumerate(np.random.randint(0, 10000, 100))):\n","  prediction = preds_test[i]\n","  # Obtain the index of the represented number in the per class order to extract the closest cluster (as it is in per class order)\n","  i_in_per_cls_order = np.where(searchIndicesOfClass(prediction, preds_test) == np.array(i))[0][0]\n","  if selected_model == 'ConvNet':\n","    array_histrogram[ind] = resize(np.sum(reconst_test[i], axis=0), (28, 28))\n","  elif selected_model == 'Fully_connected':\n","    array_histrogram[ind] = reconst_test[i].reshape(28,28)\n","  else:\n","    raise NameError\n","plot_range_inputs = [int(np.quantile(array_histrogram.flatten(), 0.80)), int(np.max(array_histrogram.flatten()))]\n","max_plot_value_inputs = plot_range_inputs[1]\n","print(\"Plot max value for inputs:\", max_plot_value_inputs)"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":116,"referenced_widgets":["d40a74633a4e4d02b367c7864760a0f0","9cb5c0c528f140caa7b01b2ea307aeef","e86f42f615914675a69c03595a2e0965","99d2db05ed3a41d5bee7913abc0d8663","c7bdd99223ce4d6c88de0300be8e95ae","f4665222fa824d76a7a383f5d83e59f3","685274e6e591482298dbb9ffee146851","41dfb5e193bb4b84be9ebd33835ecee2","ca91d0e56d624c91843662a3b617bce4","2f18ab1c4cd347cba56f0822a8dfab0f","d63a2c90243a41ee8795b0fc279c7e24","895daecc521046859758fcb876b80c13","257aa47024bd4bfb81de1b82809bfd13","ca13eee58102488199021886c4913fdb","e53a1ef479794ab5a99dcf77cc9e6969","860547aa48144bfa8870f5c9fdb0281c","53a3e7bda0ff44f7a86594610a4c680a","d08e77d3b5b645a9b93a1c62267dfc1e","61033ba9035845749b89c8cd193eb81a","56ab2d070f5746bfb7df0aa6d48675d8","392d8a826a1a4c779c0b2b8ba319586b","7550ae4115574784b50446f02558cef6"]},"id":"K5vLAzir5dwJ","executionInfo":{"status":"ok","timestamp":1663757866854,"user_tz":-120,"elapsed":429,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"}},"outputId":"9d5b35fc-a984-4240-cc6e-4de9b6f82e0d","cellView":"form"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"text/plain":["0it [00:00, ?it/s]"],"application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"d40a74633a4e4d02b367c7864760a0f0"}},"metadata":{}},{"output_type":"stream","name":"stdout","text":["Plot range for heatmaps: [20, 102]\n"]},{"output_type":"display_data","data":{"text/plain":["0it [00:00, ?it/s]"],"application/vnd.jupyter.widget-view+json":{"version_major":2,"version_minor":0,"model_id":"895daecc521046859758fcb876b80c13"}},"metadata":{}},{"output_type":"stream","name":"stdout","text":["Plot max value for inputs: 288\n"]}]},{"cell_type":"markdown","source":["## Plot type #1"],"metadata":{"id":"8TflxieoCqDr"}},{"cell_type":"code","source":["#@markdown ##Plot attribution test\n","initial_pos = 0\n","rows = 10\n","cols = 20\n","fig, axes = plt.subplots(rows, cols, figsize=(2*cols, 2*rows))\n","fig.suptitle(\"In every two columns, the first representes the image with the heatmap superimposed and the second is the heatmap alone\\n\" \\\n","             \"The number above the images correspond to the difference between the OOD score of the image to the closest centroid compared with the threshold if set to obtain a 80% FPR\", fontsize=20, y=0.94)\n","for cl_ind, ax_row in enumerate(axes):\n","  # Extract the indices of the correspondent examples\n","  ind = searchIndicesOfClass(cl_ind, preds_test)\n","  ind = ind[initial_pos:] # To start from the selected initial position\n","  preds_test = preds_test[initial_pos:] # To start from the selected initial position\n","\n","  for col, ax in enumerate(ax_row):\n","    ax.set_xticks([])\n","    ax.set_yticks([])\n","\n","    # For distinguising even and odd columns\n","    if col % 2 == 0:\n","      # Compute the distance compared to the threshold at the FPR seleted\n","      dist = distances_test_per_class[cl_ind][col//2 + initial_pos]\n","      which_fpr = 80 # To select the FPR (80 or 95)\n","      if dist - thresholds[preds_test[col//2], which_fpr] >= 0:\n","        distance_to_centroid = '+ ' + str(dist - thresholds[preds_test[col//2], which_fpr])\n","      else:\n","        distance_to_centroid = '\\N{MINUS SIGN} ' + str(abs(dist - thresholds[preds_test[col//2], which_fpr]))\n","      # Plot image of the number and title \n","      ax.imshow(test_loader.dataset[ind[col//2]][0].squeeze(), cmap=\"binary\")\n","      alpha=0.65\n","      ax.set_title(f'{distance_to_centroid}', fontsize=12)\n","    else:\n","      alpha=1 # No transparency if no image is superimposed\n","\n","    # The image to be plotted\n","    img = np.abs(reconst_per_class_test[cl_ind][col//2 + initial_pos] - avg_frecs_reconst_per_cl_clu_positive[cl_ind][closest_clusters_per_cls_test[cl_ind][col//2 + initial_pos]])\n","    if selected_model == \"ConvNet\":\n","      img = resize(np.sum(img, axis=0), (28,28))\n","\n","    # Plot the image\n","    im = ax.imshow(np.reshape(img, (28,28)),\n","                   alpha=alpha, vmin=plot_range[0], vmax=plot_range[1], cmap='Reds')\n","    # The first column includes the predicted label for all the row\n","    if col == 0:\n","      ax.set_ylabel(train_data.classes[cl_ind], fontsize=18)\n","    if col % 2 == 1:\n","      #plt.colorbar(im, ax=ax)\n","      pass\n","plt.show()"],"metadata":{"id":"TPQpCZk53NkS","cellView":"form"},"execution_count":null,"outputs":[]},{"cell_type":"code","source":["#@markdown ##Plot attribution OOD\n","rows = 10\n","cols = 20\n","fig, axes = plt.subplots(rows, cols, figsize=(2*cols, 2*rows))\n","fig.suptitle(\"In every two columns, the first representes the image with the heatmap superimposed and the second is the heatmap alone\\n\" \\\n","             \"The number above the images correspond to the difference between the OOD score of the image to the closest centroid compared with the threshold if set to obtain a 80% FPR\", fontsize=20, y=0.94)\n","for cl_ind, ax_row in enumerate(axes):\n","  ind = searchIndicesOfClass(cl_ind, preds_ood)\n","  ind = ind[initial_pos:]\n","  preds_ood = preds_ood[initial_pos:]\n","  for col, ax in enumerate(ax_row):\n","    ax.set_xticks([])\n","    ax.set_yticks([])\n","\n","    # For distinguising even and odd columns\n","    if col % 2 == 0:\n","      # Compute the distance compared to the threshold at the FPR seleted\n","      dist = distances_ood_per_class[cl_ind][col//2 + initial_pos]\n","      which_fpr = 80 # To select the FPR (80 or 95)\n","      if dist - thresholds[preds_ood[col//2], which_fpr] >= 0:\n","        distance_to_centroid = '+ ' + str(dist - thresholds[preds_ood[col//2], which_fpr])\n","      else:\n","        distance_to_centroid = '\\N{MINUS SIGN} ' + str(abs(dist - thresholds[preds_ood[col//2], which_fpr]))\n","\n","      ax.imshow(test_loader_ood.dataset[ind[col//2]][0].squeeze(), cmap=\"binary\")\n","      alpha=0.65\n","      ax.set_title(f'{distance_to_centroid}')\n","    else:\n","      alpha=1 # No transparency if no image is superimposed\n","\n","    # Compute the image to be plotted\n","    img = np.abs(reconst_per_class_ood[cl_ind][col//2 + initial_pos] - avg_frecs_reconst_per_cl_clu_positive[cl_ind][closest_clusters_per_cls_ood[cl_ind][col//2 + initial_pos]])\n","    if selected_model == \"ConvNet\":\n","      img = resize(np.sum(img, axis=0), (28,28))\n","      \n","    # Plot the image\n","    im = ax.imshow(np.reshape(img, (28,28)),  \n","                   alpha=alpha,  vmin=plot_range[0], vmax=plot_range[1], cmap='Reds')\n","    \n","    # The first column includes the predicted label for all the row\n","    if col == 0:\n","      ax.set_ylabel(train_data.classes[cl_ind], fontsize=18)\n","    if col % 2 == 1:\n","      #plt.colorbar(im, ax=ax)\n","      pass\n","plt.show()"],"metadata":{"id":"MgbuEYsY9NeP","cellView":"form"},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":["## Plot type #2"],"metadata":{"id":"vh4uIGuYCj8R"}},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"4srxUzluCioW"},"outputs":[],"source":["#@markdown ##Reconstruction of the input\n","initial_pos = 0\n","transp = 0.65\n","clr_map = 'Reds'\n","fig, axes = plt.subplots(10,15, figsize=(35,18))\n","fig.suptitle(f'First part of the title is the actual label of the image, on the left of the first column is the predicted class for that entire row\\n' \\\n","             \"The second part of the title is the difference between the OOD score of the image to the closest centroid compared with the threshold if set to obtain a 80% FPR\", fontsize=22, y=0.93)\n","for cl_index, ax_row in enumerate(axes):\n","  indices_test = np.array(searchIndicesOfClass(cl_index, preds_test, 7, initial_pos), dtype='uint16')\n","  indices_ood = np.array(searchIndicesOfClass(cl_index, preds_ood, 8, initial_pos), dtype='uint16')\n","  indices = np.concatenate((indices_test, indices_ood), dtype='uint16')\n","  for index, ax in enumerate(ax_row):\n","    ax.set_xticks([])\n","    ax.set_yticks([])\n","\n","    # First part of the image for test images\n","    if index < 7:\n","      reconst = reconst_per_class_test[cl_index][index+initial_pos]\n","      ax.imshow(test_loader.dataset[indices[index]][0].squeeze(), cmap='binary')\n","      if selected_model == \"ConvNet\":\n","        reconst = resize(np.sum(reconst, axis=0), (28,28))\n","      im = ax.imshow(reconst.reshape(28,28), vmin=0, vmax=max_plot_value_inputs, cmap=clr_map, alpha=transp)\n","\n","      # Compute distance to centroid for title\n","      dist = distances_test_per_class[cl_index][index+initial_pos]\n","      if dist - thresholds[cl_index, WHICH_FPR] >= 0:\n","        distance_to_centroid = '+ ' + str(int(dist - thresholds[cl_index, WHICH_FPR]))\n","      else:\n","        distance_to_centroid = '\\N{MINUS SIGN} ' + str(int(abs(dist - thresholds[cl_index, WHICH_FPR])))\n","      ax.set_title(test_loader.dataset.classes[test_loader.dataset.targets[indices[index]]] + f'/ {distance_to_centroid}', fontsize=11, pad=3)\n","\n","    # Second part of the image for OoD images\n","    else: \n","      try: # In case there are no sufficient OoD images for a row, skip the space left for that row\n","        # Extract the reconstruction\n","        reconst = reconst_per_class_ood[cl_index][index+initial_pos-7]\n","        # Plot the image\n","        ax.imshow(test_loader_ood.dataset[indices[index]][0].squeeze(), cmap='binary')\n","        # In case the model is a ConvNet, we must sum over channels and resize\n","        if selected_model == \"ConvNet\":\n","          reconst = resize(np.sum(reconst, axis=0), (28,28))\n","\n","        # Compute distance to centroid for title\n","        dist = distances_ood_per_class[cl_index][index+initial_pos-7]\n","        if dist - thresholds[cl_index, WHICH_FPR] >= 0:\n","          distance_to_centroid = '+ ' + str(int(dist - thresholds[cl_index, WHICH_FPR]))\n","        else:\n","          distance_to_centroid = '\\N{MINUS SIGN} ' + str(int(abs(dist - thresholds[cl_index, WHICH_FPR])))\n","        \n","        im = ax.imshow(reconst.reshape(28,28), vmin=-0, vmax=max_plot_value_inputs, cmap=clr_map, alpha=transp)\n","        try:\n","          ax.set_title(test_loader_ood.dataset.classes[test_loader_ood.dataset.targets[indices[index]]] + f'/ {distance_to_centroid}', fontsize=11, pad=3)\n","        except AttributeError:\n","          ax.set_title(f'{int(distances_ood_per_class[cl_index][index+initial_pos-7])}', fontsize=11, pad=3)\n","      except IndexError:\n","        ax.axis('off')\n","\n","    if index == 0:\n","      ax.set_ylabel(train_data.classes[cl_index], fontsize=18)\n","plt.colorbar(im, ax=axes.ravel().tolist(), pad=0.02, aspect=30)\n","plt.show()"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"8np0huuyCioW"},"outputs":[],"source":["#@markdown ##Reconstruction of the median closest to the correspondent image\n","initial_pos = 0\n","transp = 1\n","rep_range = plot_range_inputs[1]\n","clr_map = 'Reds'\n","fig, axes = plt.subplots(10,15, figsize=(35,18))\n","fig.suptitle(f'First part of the title is the actual label of the image, on the left of the first column is the predicted class for that entire row\\n' \\\n","             \"The second part of the title is the difference between the OOD score of the image to the closest centroid compared with the threshold if set to obtain a 80% FPR\", fontsize=22, y=0.93)\n","for cl_index, ax_row in enumerate(axes):\n","  indices_test = np.array(searchIndicesOfClass(cl_index, preds_test, 7, initial_pos), dtype='uint16')\n","  indices_ood = np.array(searchIndicesOfClass(cl_index, preds_ood, 8, initial_pos), dtype='uint16')\n","  indices = np.concatenate((indices_test, indices_ood), dtype='uint16')\n","  for index, ax in enumerate(ax_row):\n","    ax.set_xticks([])\n","    ax.set_yticks([])\n","\n","    if index < 7:\n","      aggretated_reconst = avg_frecs_reconst_per_cl_clu_positive[cl_index][closest_clusters_per_cls_test[cl_index][index]]\n","      ax.imshow(test_loader.dataset[indices[index]][0].squeeze(), cmap='binary')\n","      if selected_model == \"ConvNet\":\n","        aggretated_reconst = resize(np.sum(aggretated_reconst, axis=0), (28,28))\n","\n","      im = ax.imshow(aggretated_reconst.reshape(28,28), vmin=0, vmax=max_plot_value_inputs, cmap=clr_map, alpha=transp)\n","\n","      # Compute distance to centroid for title\n","      dist = distances_test_per_class[cl_index][index+initial_pos]\n","      if dist - thresholds[cl_index, WHICH_FPR] >= 0:\n","        distance_to_centroid = '+ ' + str(int(dist - thresholds[cl_index, WHICH_FPR]))\n","      else:\n","        distance_to_centroid = '\\N{MINUS SIGN} ' + str(int(abs(dist - thresholds[cl_index, WHICH_FPR])))\n","      ax.set_title(test_loader.dataset.classes[test_loader.dataset.targets[indices[index]]] + f'/ {distance_to_centroid}', fontsize=12, pad=3)\n","    \n","    else:\n","      try:\n","        aggretated_reconst = avg_frecs_reconst_per_cl_clu_positive[cl_index][closest_clusters_per_cls_ood[cl_index][index-7]]\n","        ax.imshow(test_loader_ood.dataset[indices[index]][0].squeeze(), cmap='binary')\n","        if selected_model == \"ConvNet\":\n","          aggretated_reconst = resize(np.sum(aggretated_reconst, axis=0), (28,28))\n","\n","        im = ax.imshow(aggretated_reconst.reshape(28,28), vmin=0, vmax=max_plot_value_inputs, cmap=clr_map, alpha=transp)\n","\n","        # Compute distance to centroid for title\n","        dist = distances_ood_per_class[cl_index][index+initial_pos-7]\n","        if dist - thresholds[cl_index, WHICH_FPR] >= 0:\n","          distance_to_centroid = '+ ' + str(int(dist - thresholds[cl_index, WHICH_FPR]))\n","        else:\n","          distance_to_centroid = '\\N{MINUS SIGN} ' + str(int(abs(dist - thresholds[cl_index, WHICH_FPR])))\n","        ax.set_title(test_loader_ood.dataset.classes[test_loader_ood.dataset.targets[indices[index]]] + f'/ {distance_to_centroid}', fontsize=12, pad=3)\n","      \n","      except IndexError:\n","        ax.axis('off')\n","    if index == 0:\n","      ax.set_ylabel(train_data.classes[cl_index], fontsize=18)\n","plt.colorbar(im, ax=axes.ravel().tolist(), pad=0.02, aspect=30)\n","plt.show()"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"WoLjprXFCioX"},"outputs":[],"source":["#@markdown ##Heatmap or attribution map\n","initial_pos = 0\n","transp = 0.65\n","clr_map = 'Reds'\n","fig, axes = plt.subplots(10,15, figsize=(35,18))\n","fig.suptitle(\"First part of the title is the actual label of the image, on the left of the first column is the predicted class for that entire row\\n\" \\\n","             \"The second part of the title is the difference between the OOD score of the image to the closest centroid compared with the threshold if set to obtain a 80% FPR\", fontsize=22, y=0.93)\n","for cl_index, ax_row in enumerate(axes):\n","  indices_test = np.array(searchIndicesOfClass(cl_index, preds_test, 7, initial_pos), dtype='uint16')\n","  indices_ood = np.array(searchIndicesOfClass(cl_index, preds_ood, 8, initial_pos), dtype='uint16')\n","  indices = np.concatenate((indices_test, indices_ood), dtype='uint16')\n","  for index, ax in enumerate(ax_row):\n","    ax.set_xticks([])\n","    ax.set_yticks([])\n","\n","    if index < 7: \n","      reconst = reconst_per_class_test[cl_index][index+initial_pos]\n","      aggretated_reconst = avg_frecs_reconst_per_cl_clu_positive[cl_index][closest_clusters_per_cls_test[cl_index][index]]\n","      # Plots\n","      ax.imshow(test_loader.dataset[indices[index]][0].squeeze(), cmap='binary')\n","\n","      img = np.abs(reconst-aggretated_reconst)\n","      if selected_model == \"ConvNet\":\n","        img = resize(np.sum(img, axis=0), (28,28))\n","\n","      # Compute distance to centroid for title\n","      WHICH_FPR = 80\n","      dist = distances_test_per_class[cl_index][index+initial_pos]\n","      if dist - thresholds[cl_index, WHICH_FPR] >= 0:\n","        distance_to_centroid = '+ ' + str(int(dist - thresholds[cl_index, WHICH_FPR]))\n","      else:\n","        distance_to_centroid = '\\N{MINUS SIGN} ' + str(int(abs(dist - thresholds[cl_index, WHICH_FPR])))\n","\n","      im = ax.imshow(img.reshape(28,28), vmin=plot_range[0], vmax=plot_range[1], cmap=clr_map, alpha=transp)\n","      ax.set_title(test_loader.dataset.classes[test_loader.dataset.targets[indices[index]]] + f'/ {distance_to_centroid}', fontsize=12, pad=3)\n","    \n","    else:\n","      try:\n","        reconst = reconst_per_class_ood[cl_index][index+initial_pos-7]\n","        aggretated_reconst = avg_frecs_reconst_per_cl_clu_positive[cl_index][closest_clusters_per_cls_ood[cl_index][index-7]]\n","        # Plots\n","        ax.imshow(test_loader_ood.dataset[indices[index]][0].squeeze(), cmap='binary')\n","        \n","        img = np.abs(reconst-aggretated_reconst)\n","        if selected_model == \"ConvNet\":\n","          img = resize(np.sum(img, axis=0), (28,28))\n","\n","        im = ax.imshow(img.reshape(28,28), vmin=plot_range[0], vmax=plot_range[1], cmap=clr_map, alpha=transp)\n","\n","        # Set title\n","        try:\n","          dist = distances_ood_per_class[cl_index][index+initial_pos-7]\n","          if dist - thresholds[cl_index, WHICH_FPR] >= 0:\n","            distance_to_centroid = '+ ' + str(int(dist - thresholds[cl_index, WHICH_FPR]))\n","          else:\n","            distance_to_centroid = '\\N{MINUS SIGN} ' + str(int(abs(dist - thresholds[cl_index, WHICH_FPR])))\n","          ax.set_title(test_loader_ood.dataset.classes[test_loader_ood.dataset.targets[indices[index]]] + f'/ {distance_to_centroid}', fontsize=12, pad=3)\n","        except AttributeError:\n","          ax.set_title(f'{int(distances_ood_per_class[cl_index][index+initial_pos-7])}', fontsize=12, pad=3)\n","\n","      except IndexError:\n","        ax.axis('off')\n","    if index == 0:\n","      ax.set_ylabel(train_data.classes[cl_index], fontsize=18)\n","plt.colorbar(im, ax=axes.ravel().tolist(), pad=0.02, aspect=30)\n","plt.show()"]},{"cell_type":"markdown","metadata":{"id":"Z4e7K_m6R-nd"},"source":["## Figure of the paper"]},{"cell_type":"code","execution_count":null,"metadata":{"id":"kf-oDgCMDk2X","cellView":"form","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"status":"ok","timestamp":1663157300705,"user_tz":-120,"elapsed":10436,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"}},"outputId":"3bb31f9d-954e-458b-ec7b-b9570aab6edb"},"outputs":[{"output_type":"stream","name":"stdout","text":["Accuracy for the ood dataset MNIST_Square is 93.040 %\n","Accuracy for the ood dataset MNIST-C is 85.200 %\n"]}],"source":["#@markdown Compute the loaders, preds and counts of mnist sq and c\n","mnist_c_opt = 'zigzag' #@param ['zigzag', 'scale', 'identity', 'motion_blur', 'canny_edges', 'glass_blur', 'translate', 'rotate', 'brightness', 'spatter', 'dotted_line', 'impulse_noise', 'fog', 'shear', 'stripe', 'shot_noise']\n","loader_mnist_sq, preds_mnist_sq, spk_count_mnist_sq = create_spk_count(\n","      model, DEVICE, OUT_DISTRIBUTION_DATASETS, 'MNIST_Square')\n","loader_mnist_c, preds_mnist_c, spk_count_mnist_c = create_spk_count(\n","      model, DEVICE, OUT_DISTRIBUTION_DATASETS, 'MNIST-C', mnist_c_opt=mnist_c_opt)"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","colab":{"base_uri":"https://localhost:8080/","height":186},"executionInfo":{"elapsed":391,"status":"ok","timestamp":1663157022098,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"},"user_tz":-120},"id":"9kpItlzHgayw","outputId":"1f31cc2a-bd8f-4026-8ce1-7eafdbd7050c"},"outputs":[{"output_type":"display_data","data":{"text/plain":["<Figure size 648x180 with 4 Axes>"],"image/png":"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\n"},"metadata":{"needs_background":"light"}}],"source":["#@markdown Check datasets\n","n = 456\n","title=15\n","subtitle=13\n","fig, axes = plt.subplots(1,3, figsize=(9,2.5))\n","cmap='binary'\n","axes[0].imshow(test_loader.dataset[n][0].squeeze(), cmap=cmap)\n","axes[0].set_title(f'MNIST', fontsize=title)\n","axes[0].set_xlabel('InD', fontsize=subtitle)\n","\n","axes[1].imshow(loader_mnist_sq.dataset[n][0].squeeze(), cmap=cmap)\n","axes[1].set_title(f'MNIST Square', fontsize=title)\n","axes[1].set_xlabel('OOD', fontsize=subtitle)\n","\n","im = axes[2].imshow(loader_mnist_c.dataset[n][0].squeeze(), cmap=cmap)\n","axes[2].set_title(f'MNIST-C', fontsize=title)\n","axes[2].set_xlabel('OOD', fontsize=subtitle)\n","\n","for ax in axes.flatten():\n","  ax.set_xticks([])\n","  ax.set_yticks([])\n","cbar = plt.colorbar(im, ax=axes.ravel().tolist(), pad=0.04)\n","cbar.ax.tick_params(labelsize=12)\n","fig.savefig(figures_path / f'attribution_datasets.pdf', bbox_inches='tight')\n","plt.show()"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","id":"UWjccjoOupeS"},"outputs":[],"source":["#@markdown Distances, reconst and agg counts\n","# Compute distances per class an closest clusters per class\n","distances_test_per_class, closest_clusters_per_cls_test = distance_to_clusters_averages(spk_count_test, preds_test, agg_counts_per_class_cluster)\n","distances_mnist_sq_per_class, closest_clusters_per_cls_mnist_sq = distance_to_clusters_averages(spk_count_mnist_sq, preds_mnist_sq, agg_counts_per_class_cluster)\n","distances_mnist_c_per_class, closest_clusters_per_cls_mnist_c = distance_to_clusters_averages(spk_count_mnist_c, preds_mnist_c, agg_counts_per_class_cluster)\n","\n","# Create the reconstruction\n","if selected_model == 'ConvNet':\n","  weights = model.snn.state_dict()['fc1.weight'].cpu().numpy()\n","elif selected_model == 'Fully_connected':\n","  weights = model.state_dict()['snn.fc1.weight'].cpu().numpy()\n","else:\n","  raise NameError\n","\n","reconst_test = np.dot(spk_count_test, weights)\n","reconst_mnist_sq = np.dot(spk_count_mnist_sq, weights)\n","reconst_mnist_c = np.dot(spk_count_mnist_c, weights)\n","\n","# Extract negative part\n","reconst_test_neg      = np.where(reconst_test <= 0, -reconst_test, 0)\n","reconst_mnist_sq_neg  = np.where(reconst_mnist_sq <= 0, -reconst_mnist_sq, 0)\n","reconst_mnist_c_neg   = np.where(reconst_mnist_c <= 0, -reconst_mnist_c, 0)\n","\n","# Extract positive part\n","reconst_test = np.where(reconst_test >= 0, reconst_test, 0)\n","reconst_mnist_sq = np.where(reconst_mnist_sq >= 0, reconst_mnist_sq, 0)\n","reconst_mnist_c = np.where(reconst_mnist_c >= 0, reconst_mnist_c, 0)\n","\n","\n","### Backprop the hidden counts of the median ###\n","agg_counts_reconstructed_per_class_and_cluster = compute_reconstruction_per_class(agg_counts_per_class_cluster, weights)\n","# Extract the positive part\n","avg_frecs_reconst_per_cl_clu_positive = extract_positive_part_per_class(agg_counts_reconstructed_per_class_and_cluster)\n","#avg_frecs_reconst_per_cl_clu_negative = extract_negative_part_per_class(agg_counts_reconstructed_per_class_and_cluster)\n","\n","if selected_model == 'ConvNet':\n","  # Rearrange\n","  reconst_test = rearrange_to_ftmaps(reconst_test, ftmaps_shape=(50, 11, 11))\n","  reconst_mnist_sq = rearrange_to_ftmaps(reconst_mnist_sq, ftmaps_shape=(50, 11, 11))\n","  reconst_mnist_c = rearrange_to_ftmaps(reconst_mnist_c, ftmaps_shape=(50, 11, 11))\n","  avg_frecs_reconst_per_cl_clu_positive = rearrange_to_ftmaps_per_class(avg_frecs_reconst_per_cl_clu_positive, (50,11,11))"]},{"cell_type":"code","execution_count":null,"metadata":{"cellView":"form","colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":357,"status":"ok","timestamp":1663157091908,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"},"user_tz":-120},"id":"TdwX5RLkblAQ","outputId":"87401b35-fd1f-4b1e-e767-1bd6bbbaa9ca"},"outputs":[{"output_type":"stream","name":"stdout","text":["------------------------------\n","Results\n","------------------------------\n","AUROC = 72.048 %\n","AUPR  = 71.099 %\n","------------------------------\n","FPR at 95% TPR: 81.52%\n","Threshold mean and std for all classes at 95% TPR: 1950.40, 140.82\n","------------------------------ \n","\n","------------------------------\n","Results\n","------------------------------\n","AUROC = 70.926 %\n","AUPR  = 69.110 %\n","------------------------------\n","FPR at 95% TPR: 80.19%\n","Threshold mean and std for all classes at 95% TPR: 1950.40, 140.82\n","------------------------------ \n","\n"]},{"output_type":"stream","name":"stderr","text":["/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:60: RuntimeWarning: invalid value encountered in true_divide\n"]}],"source":["#@markdown AUROC and AUPR and extrac distance threshodls\n","thresholds = auroc_aupr(distances_train_per_class, distances_test_per_class, distances_mnist_sq_per_class)\n","\n","# For both OOD datasets, the thresholds are the same as they are based on the \n","# In-Distribution training set \n","_ = auroc_aupr(distances_train_per_class, distances_test_per_class, distances_mnist_c_per_class)"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/"},"executionInfo":{"elapsed":356,"status":"ok","timestamp":1663157132730,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"},"user_tz":-120},"id":"ug727RFxM7NN","outputId":"f559a71b-2017-4ed7-c529-1dff90f07c7e","cellView":"form"},"outputs":[{"output_type":"stream","name":"stdout","text":["[3714 8210 6522]\n"]}],"source":["#@markdown ## Indices selection\n","indices = np.random.randint(0, 10000, 3)\n","print(indices)"]},{"cell_type":"code","execution_count":null,"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":591},"executionInfo":{"elapsed":1158,"status":"ok","timestamp":1663157740448,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"},"user_tz":-120},"id":"tnUIctTSyo9w","outputId":"69176dad-f00e-4b30-fd2f-04b902e4dc36","cellView":"form"},"outputs":[{"output_type":"display_data","data":{"text/plain":["<Figure size 648x648 with 9 Axes>"],"image/png":"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\n"},"metadata":{}}],"source":["#@markdown Plot with xlabel\n","nrows = len(indices)\n","ncols = 3\n","fig, axes = plt.subplots(nrows, ncols, figsize=(ncols*3, nrows*3))\n","for i, ax_row in enumerate(axes):\n","    \n","  for j, ax in enumerate(ax_row):\n","\n","    if j == 0:\n","      image = test_loader.dataset[indices[i]][0].squeeze()\n","      prediction = preds_test[indices[i]]\n","      # Obtain the index of the represented number in the per class order to extract the closest cluster (as it is in per class order)\n","      i_in_per_cls_order = np.where(searchIndicesOfClass(prediction, preds_test) == np.array(indices[i]))[0][0]\n","      diff = np.abs(reconst_test[indices[i]] - avg_frecs_reconst_per_cl_clu_positive[prediction][closest_clusters_per_cls_test[prediction][i_in_per_cls_order]])\n","\n","      dist = distances_test_per_class[prediction][i_in_per_cls_order]\n","\n","    elif j == 1:\n","      image = loader_mnist_sq.dataset[indices[i]][0].squeeze()\n","      prediction = preds_mnist_sq[indices[i]]\n","      i_in_per_cls_order = np.where(searchIndicesOfClass(prediction, preds_mnist_sq) == np.array(indices[i]))[0][0]\n","      diff = np.abs(reconst_mnist_sq[indices[i]] - avg_frecs_reconst_per_cl_clu_positive[prediction][closest_clusters_per_cls_mnist_sq[prediction][i_in_per_cls_order]])\n","      \n","      dist = distances_mnist_sq_per_class[prediction][i_in_per_cls_order]\n","\n","    elif j == 2:\n","      image = loader_mnist_c.dataset[indices[i]][0].squeeze()\n","      prediction = preds_mnist_c[indices[i]]\n","      i_in_per_cls_order = np.where(searchIndicesOfClass(prediction, preds_mnist_c) == np.array(indices[i]))[0][0]\n","      diff = np.abs(reconst_mnist_c[indices[i]] - avg_frecs_reconst_per_cl_clu_positive[prediction][closest_clusters_per_cls_mnist_c[prediction][i_in_per_cls_order]])\n","      dist = distances_mnist_c_per_class[prediction][i_in_per_cls_order]\n","\n","    else:\n","      raise IndexError\n","\n","    if selected_model == 'ConvNet':\n","      diff = resize(np.sum(diff, axis=0), (28, 28))\n","    plot_ax(ax, img=image, plt_range=[0,1], cmap='binary')\n","\n","    which_fpr = 80 # To select the FPR (80 or 95)\n","    if dist - thresholds[prediction, which_fpr] >= 0:\n","      distance_to_centroid = '+ ' + str(dist - thresholds[prediction, which_fpr])\n","    else:\n","      distance_to_centroid = '\\N{MINUS SIGN} ' + str(abs(dist - thresholds[prediction, which_fpr]))\n","\n","    if selected_model == 'ConvNet':\n","      plot_ax(ax, img=diff.reshape(28,28), plt_range=plot_range, cmap='Reds', alpha=0.65, title=f'{distance_to_centroid}',xlabel=f'Prediction = {prediction}', fontsize=18)\n","    elif selected_model == 'Fully_connected':\n","      plot_ax(ax, img=diff.reshape(28,28), plt_range=plot_range, cmap='Reds', alpha=0.65, title=f'{distance_to_centroid}',xlabel=f'Prediction = {prediction}', fontsize=18)\n","    else:\n","      raise NameError \n","      \n","    #plot_ax(ax, img=diff.reshape(28,28), plt_range=[200,1000], cmap='Reds', alpha=0.65, title=f'{distance_to_centroid}',xlabel=f'Prediction = {prediction}', fontsize=18)\n","    #plot_ax(ax, img=diff.reshape(28,28), plt_range=[200,1000], cmap='Reds', alpha=0.65, title=f'{prediction} | {dist}', fontsize=18)\n","plt.subplots_adjust(left=0.1,\n","                    bottom=0.1, \n","                    right=0.9, \n","                    top=0.9, \n","                    wspace=0.2, \n","                    hspace=0.45)\n","fig.savefig(figures_path / f'{selected_model}_reconstruction.pdf', bbox_inches='tight')\n","#fig.savefig(figures_path / f'{selected_model}_reconstruction.png', dpi=300, bbox_inches='tight')"]},{"cell_type":"code","source":["#@markdown Plot with closest cluster also\n","nrows = len(indices)\n","ncols = 3\n","fig, axes = plt.subplots(nrows, ncols, figsize=(ncols*3, nrows*3))\n","for i, ax_row in enumerate(axes):\n","    \n","  for j, ax in enumerate(ax_row):\n","\n","    if j == 0:\n","      image = test_loader.dataset[indices[i]][0].squeeze()\n","      prediction = preds_test[indices[i]]\n","      # Obtain the index of the represented number in the per class order to extract the closest cluster (as it is in per class order)\n","      i_in_per_cls_order = np.where(searchIndicesOfClass(prediction, preds_test) == np.array(indices[i]))[0][0]\n","      diff = np.abs(reconst_test[indices[i]] - avg_frecs_reconst_per_cl_clu_positive[prediction][closest_clusters_per_cls_test[prediction][i_in_per_cls_order]])\n","      dist = distances_test_per_class[prediction][i_in_per_cls_order]\n","      clu = [closest_clusters_per_cls_test[prediction][i_in_per_cls_order]]\n","\n","    elif j == 1:\n","      image = loader_mnist_sq.dataset[indices[i]][0].squeeze()\n","      prediction = preds_mnist_sq[indices[i]]\n","      i_in_per_cls_order = np.where(searchIndicesOfClass(prediction, preds_mnist_sq) == np.array(indices[i]))[0][0]\n","      diff = np.abs(reconst_mnist_sq[indices[i]] - avg_frecs_reconst_per_cl_clu_positive[prediction][closest_clusters_per_cls_mnist_sq[prediction][i_in_per_cls_order]])\n","      dist = distances_mnist_sq_per_class[prediction][i_in_per_cls_order]\n","      clu = [closest_clusters_per_cls_mnist_sq[prediction][i_in_per_cls_order]]\n","\n","    elif j == 2:\n","      image = loader_mnist_c.dataset[indices[i]][0].squeeze()\n","      prediction = preds_mnist_c[indices[i]]\n","      i_in_per_cls_order = np.where(searchIndicesOfClass(prediction, preds_mnist_c) == np.array(indices[i]))[0][0]\n","      diff = np.abs(reconst_mnist_c[indices[i]] - avg_frecs_reconst_per_cl_clu_positive[prediction][closest_clusters_per_cls_mnist_c[prediction][i_in_per_cls_order]])\n","      dist = distances_mnist_c_per_class[prediction][i_in_per_cls_order]\n","      clu = [closest_clusters_per_cls_mnist_c[prediction][i_in_per_cls_order]]\n","\n","    else:\n","      raise IndexError\n","\n","    if selected_model == 'ConvNet':\n","      diff = resize(np.sum(diff, axis=0), (28, 28))\n","    plot_ax(ax, img=image, plt_range=[0,1], cmap='binary')\n","\n","    which_fpr = 80 # To select the FPR (80 or 95)\n","    if dist - thresholds[prediction, which_fpr] >= 0:\n","      distance_to_centroid = '+ ' + str(dist - thresholds[prediction, which_fpr])\n","    else:\n","      distance_to_centroid = '\\N{MINUS SIGN} ' + str(abs(dist - thresholds[prediction, which_fpr]))\n","    plot_ax(ax, img=diff.reshape(28,28), plt_range=plot_range, cmap='Reds', alpha=0.65, title=f'{distance_to_centroid}',xlabel=f'Prediction = {prediction}\\nClosest cluster = {clu[0]}', fontsize=18)\n","    #plot_ax(ax, img=diff.reshape(28,28), plt_range=[200,1000], cmap='Reds', alpha=0.65, title=f'{prediction} | {dist}', fontsize=18)\n","plt.subplots_adjust(left=0.1,\n","                    bottom=0.1, \n","                    right=0.9, \n","                    top=0.9, \n","                    wspace=0.2, \n","                    hspace=0.52)\n","fig.savefig(figures_path / f'{selected_model}_reconstruction.pdf', bbox_inches='tight')\n","#fig.savefig(figures_path / f'{selected_model}_reconstruction.png', dpi=300, bbox_inches='tight')"],"metadata":{"colab":{"base_uri":"https://localhost:8080/","height":612},"cellView":"form","id":"u08lK9VZ-gRc","executionInfo":{"status":"ok","timestamp":1663157811290,"user_tz":-120,"elapsed":1616,"user":{"displayName":"Aitor Martinez","userId":"09566376405051483384"}},"outputId":"36c6b8d7-aa60-4f6a-eed1-9b4b7db650a8"},"execution_count":null,"outputs":[{"output_type":"display_data","data":{"text/plain":["<Figure size 648x648 with 9 Axes>"],"image/png":"iVBORw0KGgoAAAANSUhEUgAAAjIAAAJTCAYAAAASZPymAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOydd5xdV3Xvf2u6epcs2ZZkWdVd7gVjGUxvAUIAG4IBJy+PkIQEkjxCAJM8SAj1kUaopth04oRQDVjGXbIty2q2ZVUjy5JVp2n6fn/sM/bda62Ze6ZIc8/M7/v53M/VXrPOPvueu87Rvmf/zloSQgAhhBBCSBGpGukBEEIIIYQMFk5kCCGEEFJYOJEhhBBCSGHhRIYQQgghhYUTGUIIIYQUFk5kCCGEEFJYOJEhhBBCSGHhROY4ISKniMgXRGS3iLSLyNMi8lMROaPEp1ZEPi8iD4rIgcxvh4h8R0RWOn2+QkT+W0R2ikiriBwWkYdE5D0i0jCAsYU+Xs3D9fnJ6ENEJorIh7MY/G0WM6vLbPNyEblHRFpE5JCIfE9ETlM+IiJvEZFvi8gTWWzvzvZzyQDHOEVE/llE9ohIm4hsEpH/LSIyiI9MxjgDveaKyOp+rq8XDmC/y0Tk1mx/LSJyp4i8YHg/3eihZqQHMBrJJiG/BNAE4CsAdgOYDuBCALNKXOsy290AvpH5zwfwdgD3i8hLQwi/LvE/G0A3gC8D2AtgHIArAXwGwCtE5MUhf4bDOwF8Qdk6835GMiaZCeBGAPsAPAhgTn/OIvI6AN8HsB7AXwKYAuA9AO4WkQtDCE9lrvWI8f8wgG8D2AFgLoA/AnCviPx+COGb5QYnInUAbgOwEsA/A9gC4GUA/i0b6435PyohAAZ3zT0A4M+dvrbn2aGInA7gHgBdAP4JwFEAfwDg5yLyshDCLwfzQUY1IQS+hvEFoAHAVgDrAEweZB9zEScVP8np/68AAoCLc/oHADeN9LHiq1gvxAnHKSXtZgCr+/CtBbAHwC4AE0vs5yH+x/CFElsNgKucPuYg/qewD0BVjvG9K4vtP1H2HwDoALBgpI8hXyf2BWBVFhOrhrlf95oLYDWAnUPs+7vZOXJeiW1idi49BkBG+rhW2otLS8PP7wFYDOBDIYRGEakXkfoB9rEfQBuAaTn9d2Xvef0BxF+wIjJxINuQsUsIoT2E8Nuc7lcBmAfgSyGEZ5csQwgPI17s3ygitZmtK4Rwh7O/fQDuADA7e5XjWgCtAL6o7J9FnFi9MefYCSlHv9dcEakSkckDXdIUkQkAXo34A+HhXnt2Dn0JwFIAFw1uyKMXTmSGn5dn70dE5DcAjgFoE5F1IvISbwMRqRaRmSJykohcBOAWxBn4T/rwn5T5LxKRtwL4awAHAdw/gHH+LuJFv0lE9me6gikD2J6Q/ui92N7r/O0+AJMRL8rlOAXxbsqR/pxEpArA+QDWhRDa1J/XIP565n8AZFAM8Jp7MuLdyqMAmkXkhyKyPOeuzkG889nXeQMwjg3UyAw/y7L3HyAG+ZsQ9TEfAPATEXlJsGucKwBsKGkfBfAP2cvjqwBeX9K+H8AfhxD6vdiXsAbA9wA8gfgfyssBvBvAVSJyeekvaEIGybzsfY/zt17byQA29dWBiLwcwMUAvuFMTjTTEPULZn8hhHYROZDtj5DBkPeauwNR8/gI4vLQJYjX1heKyPNCCBvQP3nPG1LCmJ7IiMhURPFhXj4XQjhUxmdS9v4ogFeHbIFTRH4FYDOAjyIKgUvZAeBFiOLfxQDegiiMrEcUfGk+AuDziMLhqxFn8TPyfogQgn4S5Osi8kg2tj/L3kmBOU6xPRDGZ+/tzt/alI9BRJYgCoD3AHjvEPfXu88+90eKT3ZXbroy995lniIiM9XfDoUQenJ2n+uaG0J4uzJ9X0T+G3E59dOI1/n+GNJ5M1YZ0xMZAFMBfHgA/t8EUO5ifyx7/3rvJAYAQghbReQeAFeKyIQQQkvJ31pQMrkRka8AeAjxrs5L9Q6yWX3vzP5bIvK/APxURJ4fQrh7AJ+nlE8gHotXgBOZ0cDxiO2B0Jq9e/qwBuWTkD2e/SvE5aCXhRCeGeL+evfp7o+MGuYj/ij0uNWxnQZgZ56Oh3LNDSHcmckMrhaRcSGEY/24D/q8GcuMaY1MCGFnCEEG8HoiR7e9Ysinnb/tBSB47ldCX+NqBvBDAC/JHsUrR++jqX+Uw7evfXYCeArxEVtScI5TbA+E3kervdvgvTZz+1xEFgK4HVEj9qIct+J7OYz4I8LsLxPbz/T2R0YVTyPe8Sh9vS/72/ucv3nX6LwM9Jq7E0A1yj+QMajzZqwz1u/IHA/WIN5FOcX52ymIS0V5fvmOy96nA9hWxrcOcVKqb6vmJkvudAqeE5QRMhTWZu+XwS6lXgqgEcDjpcZsErMacaJ/TQhhXd6dhRB6ROQhACtFpD6EUHpr/mLEHxAPDGD8pGBkOqok1kSkd2n+wRDC6mHc3UCvuUuQ79q/AXFZ6TLnb5dm74xjxZi+I3OcuAVR5HWDiDw7URSRcxGD8/Ze4aKIzMrWdRNE5CQAb0BUvm9Sdo8/zd6TSYiInK7V8iLSl5bm7xEntj/q+6MRkps7EO9A3lD6iH92HqwC8L3sLmCvfQHinZipAF4cQniwr44lZsReLiLz1Z++hagf+ENlfw/ifyLfGfzHIWORgVxzs6zS1U4frwBwBYDbSkXrmf/yUu1Odjf+RwBWZedKr+9EADcg5ihbM4SPNCqREhkHGSZE5GMA3o+YnfHbiLP2P0WcxV8RQngk83sP4kX2PxHXdjsQH0l9G+ItyBtCCF8p6fcAgLsQ9TN7EG+XvwjACxFn8leEEJpK/HciJgGTEttnEGf2tyNmHJ6I+NTS1YhK/KvLrOGSMYyIvBtxsgEAH0S8Ff7lrL0+hPCjEt83IE4e1iPmdpmMmPE0ALgghLAn85uU+ZyGmJHXu1DfluWV6b1zswPAHSGEVSX7q0M8584F8DnEzL4vB/BaAP83hPDBIX14UjhEZBXite7qwdyRGcg1V0R+B1HQ+yPELL5diHcD34J4J+aKEMLjJX1fj/g01EdCCDeW2BcjngOdiBmEGxEz+54N4BUhhJ8P9HOMekY6I99ofSH+KlyPqDQ/jKh5OVP5XIC41roV8e5LB4AnES/+lzt9fhCxtMA+xCBvRLyF/34AExz/nfErTmyvAfBzxJOyDUALYmr4vwHQMNLHja/KfvXGVB+vmxz/VyL+am3NzoPvAzhd+Szsp8/e1yrHf7Wzv6kA/gVxgtWO+KTgu8FsqGPyhSFm9h3INRcxjcZ3EaUAzVn8bUPMAnyy0/f12dhudP62AsB/IeZPakWcTF0z0sezUl+8I0MIIYSQwkKNDCGEEEIKCycyhBBCCCksnMgQQgghpLBwIkMIIYSQwsKJDCGEEEIKS8Vk9p05c2ZYuEDntyJjhZ27duPAgQNS3rN4jHhse3XxevTTinmfXpR+m7l93N1po7Oh5LRVEKM6tqdNDQtPnpsaq9KccAf320oA0ydOTNriPD3b3WPjtrU1LTPUecymvKpWOUa7OjqNz9RptlKAHkKV940po1TZewFdzvnW2NqStGdMcxICe08Q60HU1Fofbct7PuhcrFUmlx+gP58a487du3HgwMERj+2KmcgsXDAfD9xz10gPg4wQF17+vJEewnHjRMe2SanQ2WadOlRx3dBtfdwLorJVO5cQffHz+nH+kzI278LqXcirU5tU2MRmVMf2yXOx5ntfTY2TpibNmz/zcbPdG5+fHpPqLjvZaGlpMbaHHkgTPv92yxbjM7UmLQ594Ld2IvW6177e2Lq60virq7PxV1WfxnvNxAbjc6jLnm+3rU1zPF77xjcZHzjHAPVp7UiZOcf6zFITSe+c9M6JWtX3BFsCUBrSCWfo6kjaF115te13BODSEiGEEEIKCycyhBBCCCksFbO0RAgZLnLoXfTyj9HM9NGPXrYabGbwXBoZUvGIAHXp8oo0TEjaF15qCznfdtsvkvaSufOMz5ZNG+3uutMl0IvOO9/4LDjplKS97j5bLPpXv/qVsV1xxaqk3d3dZXykPbVtX7fe+Ow8us/YMGFc0rzrV7ogPPC8Sy81NoMzJrtM6ywTO+fWLbd8O2lf90fvtluZ813vqzLOWd6RIYQQQkhh4USGEEIIIYWFExlCCCGEFBZqZCqYQ4cPGdtPf/rTpF2lcwEAeOE11xjb7Fmzhm9gpKIRFRPuKnZPutYenMehvRwZqK7WTt4AyowQOHT4sLH99Mf/k7Srau2j1te85GXGNuskq68gJ4ieHqC1MTEFFXHLnBxKD6jvf+2BA8bnlAULjO38lakmZnx7u/HpPHA0aZ9z9tnGZ++ePca2bdtjSXvJGYuNz/qt6ePeTzfbOJ52mv287W3pI9m7nfhfs9k+Sn7xlerR/Xr7uLdOU7D/8BHjsmHjJrtdQ6rbQY2dDuhUBsGkRKiMVAe8I0MIIYSQwsKJDCGEEEIKCycyhBBCCCksnMgQQgghpLBQ7FshbN6y2dg2bXIEWkpc1eMkJHt6715jo9h3DOMUsQs6sVaPk0Srqt7aRIt9y9dj2rzZie2NTrKzqvIi5b1ebM892fEkJ4SebqA5FfuiQ9UaaksLPQLA+EmTkvbSpcuMz5nn22R3Wtga9lrRrsG5Rr7kpS81tlu+/a2kva/nqPGpmp1eRyfNs7WPmrrsuSSqrlHosontnthnE+ldMj6tdaSTDwLAxkcfTdpbd+w2PhOnzzC2s8+/KGnfcdc9xmfVC9SDI1VqylAhdc14R4YQQgghhYUTGUIIIYQUFk5kCCGEEFJYqJEZIVpb03XjrVu3Gp+Ojk5jq1YJyebNnWt8li5bOsTRkSITtCbGEZuYpHnVzm8avR4O2IR4TtK81raOpL11p12z73TGVF2fJuiae8qpxmfZmWfZDcnIEQLQmX7fpu1811de86KkPXPmTKdvZ38dKgGec43s6Ug1Kl5d06pandgNqB5Xl7SnnWy1V2dcfXXSvmv9BuPT2Wo1QVW1ad/zZs82PuMb6oztP39xW9KeceppxmdvY1PSXrTCJgC8+PIrjA016f5++qvbrY9CJ8irFHhHhhBCCCGFhRMZQgghhBQWTmQIIYQQUlg4kSGEEEJIYaHY9wRwrO2YsW3bti1pNze3GJ8apxrp+eevTNpLFi8Z4ujIqKPLCiANNWllafGq2DrxpwXAbU5ir+07dibtllYb/zVK2AswtgtJCEC3Epdrse+48WazmQsXpQYt4gUQmmxCOjSmyfe6m6ywtqc9jcmeHqv2ra61v+EvuPySpH365Rcbn22H0qrVLfqzA6ieONnYzr/ggqS95PTTjY+XlPKxb38nade1dxifiy+9PGkvWnGm7dsT7iux7wtf+ELrUxB4R4YQQgghhYUTGUIIIYQUFk5kCCGEEFJYOJEhhBBCSGGh2PcE4GXt3bAhrf5bU2MzTWrxI0ABJMlBUKJBLxlnta5i6/ym0T6AEQ1u3bLFuDC2xxAhAKqSelDCXRk/wW6nbT1WNIt2RwB8JBUAdzdZIXmHyuxbXWVPAF1pHQDOuuDcpL3hyd8an40HDqV9T7FVpc+/4EJjW6KzrXuf10lBvGDJ8qR94WWXG58GIy72zmV7DoqqJF5fZ32KAu/IEEIIIaSwcCJDCCGEkMLCiQwhhBBCCgs1MkMkOCVad+9+Mmlv2rjJ+FSrNdoLVMIkAFh8+uIhjo6MRaTOJiDTBK8ksOnImhjbpCxab+XpQXTVaichnlvZuj3Vv/Qcsz5tbalPrZP8rt7uDbuf3pe0Nznaxqq5aUX2Cy6+xPgsXuwlu9Pnm3NyOZWln/c8p2q1pjVNEuj9n+QVrTZeNfaoVGq1aw3vyBBCCCGksHAiQwghhJDCwokMIYQQQgoLJzKEEEIIKSwU+w6Rrk5b/feuu+4qu93CBQuSNsWP5ESSR8TX2WmFlIxtUhYdWp6wvEtVcXbFvrbSc097GpPtHbZitBb7enR2O9ftNWuStsyeY3zmL0qFvIuXLrOdB0fcDDUmJ0mfixYJe8dJVxt3jndwEgCKdtP9AAgmUabayKnYPRLwjgwhhBBCCgsnMoQQQggpLJzIEEIIIaSwUCMzRB5e/3BZn2nTphnbhRfaJGGEVBKMbVKWnh6E5qbU1pXqJoKTEE+UHiM0NRmfnsZmY+tuakv7caQmEyak/601NNhiiBu22ESOmmlTpxjbhSvPSw1dVkfmamS0rdvRDXlaIq3laW+zPm22cKahxv5XH7qVvsX7LD1q/1oz42iNRgLekSGEEEJIYeFEhhBCCCGFhRMZQgghhBQWTmQIIYQQUlgo9h0ALS0txrZ9+46y282ZPdvY6uu8+quEjAyMbTIYQlcXwsFDypaKSKvqa+12dXVJu6fJxl/HvqPGdrQpFQlPnlRnfMbNnJi028UKUnet22tsqE5FwbMdsW99txLEtllBsqtA1nQ5ieR0kkAA6FJj94S9XpI8TY0VPJu+W6zgGi1pZW3oc9tJmjkS8I4MIYQQQgoLJzKEEEIIKSycyBBCCCGksHAiQwghhJDCQrHvANiyZYuxdWnBlMPy5cuPx3AIGTYY22QwhK5udOxXgtDuNIut1FqhqdSl//V0N1oR6+EjVsTa0pbG5NTp44xP7SmpAH395s3Gp8sTpCvR7LJFi6zPsVZr05iK0QB09WlPoOtl7VV+wRPXqu1cqXGtFVyjRo3h0AHjEg4dTA3jxqdtLxvwCMA7MoQQQggpLJzIEEIIIaSwcCJDCCGEkMJCjUw/HG1MEzLt3r17UP0E2KqmO3bs6LedF6myK6LnnnuusU2dmlYprsqTtImMWhjbZDjo6Q5oa0q1Fl1daUzU19nfy7XjUh1NZ6etGN3dZWPLWJwYOdqS6liefPpp44NxVltjdCTVVtuzY+vWtL3TiW1HW6a1LeJUBD/vrDONberENLmfdNikeUHbmpwkfQet/sXodtqsbie0Kf2N7psJ8QghhBBChgYnMoQQQggpLJzIEEIIIaSwcCJDCCGEkMJCsW8/dHakQqZjbU7CohzcfvtqY2ttTau9dnnVUAfJU0/Zyq7nn78yaS9btsz4VHmJnMiohLFNhoOenoDm5jSWOrtSIWvXOPvfzATVdsW+3Y7Y15oMnUpI29ZjN5LxegQAulOR7u2/udO4tLakYteuY041aqeSPFpTAXJwPshTj6w3tpVnp+L2JQtskr7Qlh7/4HxenaQwjiFte+J6T0xdSs8wnttDgWc3IYQQQgoLJzKEEEIIKSycyBBCCCGksFAjcwJobGw0Np20q1onJ4K/jtqTZ5HY4aGH1iVtcUqLsQAgGSiM7bFNCEC30mR0q+/RS2zXqTQb3ldf5yTSq+lJv9uqBqcYok6b53XuJLtDXV3SbHISxInSWlVV2/9Cg5OQsUd/3i5Hs+IkyXvgrvuSdse+o8Zn/qmnJ+2uTvt5u52+NfX19pg0KJtOdhgcHdNIwDsyhBBCCCksnMgQQgghpLBwIkMIIYSQwsKJDCGEEEIKC8W+J4CZM2YY24oVK5L2/Pnzjc+27duNraU5TcjkiSY3btpUdkzbtm0zNj2G8ePHl+2HjG0Y22MbEaC2Nv09XK2SqFXXWPFrUNpT3QcAjBtnxafVdel/WdUT6u2gVDXo0GaT1sk45/uvaUiaM06aa1xWnHFG0p4/9yTjs339OmNr2re/7Jg2PvigsXWpZHebNzxmfKpr03NwvFPZu9oRINfUpMdcfyeA1Unrc7IypL68I0MIIYSQAsOJDCGEEEIKCycyhBBCCCksnMgQQgghpLBQ7HsCeMELXmBstbVeRsqU0xfZSqea4MmtHGHXxo0bk/aRozZDZKuq5EpBJCkHY3tsUyWCulolyi3/9RsRqSf2rZ1mK1TXaJtX1fzwobR9zKnsXm8FsVBx+4JXvNK6aJFwj63+vOiCi23fqmp2aLIxKo64fd2v7krae/cdND7TjqbZtU+qswLo8U527SqVkbnHqZqtK2kPNvv28YZ3ZAghhBBSWDiRIYQQQkhh4USGEEIIIYWFGpl+mDZtWtKeN88mSHrqqb1l+2lsajK2GdOnD2pMXd3pmuyaNWuMz549ewbVNxk7MLbJcFDdUIPpy2YnNqlWidY6u8x2PR3pd11VZ5PfyQRHx6Q1Ko72Y9qkiUl73rJlxmdvq6ObmTglaTZJnXGZPkOdJ15V6QZbEb6r7nDSXnPvvcZnz65dxlal9EdeheqJNanN08PUV1ltma4u7vVdOyE9BjXt6XfpdDsi8I4MIYQQQgoLJzKEEEIIKSycyBBCCCGksHAiQwghhJDCQrFvP1RXV/fbzsvmzZuNbenSJUn7sUdtVdNjx2yF1G4lLjt8+LDxycOcOXOMbcqUKY4nGY0wtslwIA0NqFqqxLQqIZu0ttjtjM1JtFZj/3uSamXTbQA1SuxaPc1WaEdVs7VNSGNky5NPG5elU9PYetSL7aM2aV13c5oA7/AB6+PlmhOVKHDB/HnG55QZqbi+VicohF+BXCcyrB/nHMtJaeLAnpq0srg4wuKRoDJGQQghhBAyCDiRIYQQQkhh4USGEEIIIYWFExlCCCGEFBaKfQfAksVLjO2Z/c+U3W737t25bMeTyZMnJ+0zzzzD+NQ64joyNhjtsV2jqmYHR1kpTmVtUoa6BuC0FdZWSqvNdCva1tFp++6xGYERVCbdKueapSpiL1l5oXF55uFH7Gbj0zjavdeKfXcftFWrDZ3t1tauPl+DU327xop0J0+ZmrTPWWizFE9oSI+3F8c1jthXZ/KtmT7R+FRPTwXQVcdaVceVcS+kMkZBCCGEEDIIOJEhhBBCSGHhRIYQQgghhYWiiAEwd66tELxq1Spj26WqmD7xxBPGp7PLWf8dBLW19is8ed7JxnbZZZcl7aoKSWREKoPRFNu33HKz7UzrFrzsY15SNq3JcD7btW98Q9KWKqV16B6e41GR1NRCpp+U2rRuRWlPAADdSjPS2WFcQodNmogOp2p1GeZOm2Vsq2bMNrZdT6aV1bfttRqxfN9k+TiqcRJQzjv1VGO7ZOGipN19uNX4HGtMY7um2uphasfVGlvN1LSSePVce0xk3ilpuzFNUil1tkL4SMD/zQghhBBSWDiRIYQQQkhh4USGEEIIIYWFExlCCCGEFBaKfYeITsYFAFsefTRpr1y50vjs2rkzaS9YuHBQ+z/JqfQ7ffp0x5OQgVHY2O7ptrYuJSbtchKwOYnMghaXHrOVnLFnW7pNrRJAOkLWUYMIpNoKSRNq68vbGqyL+x1pAbAWYwNWyO3Ew2TnN/yjP/tZ0j7/yhcYn53706rVCxeeZvffbgW52ja7e7FxmdZkq72HffuTdo+TgE6Lez1hb/Uke4Crp6Xnt5xkK2tjwdK0fTgdj650PlLwjgwhhBBCCgsnMoQQQggpLJzIEEIIIaSwUCMzRGpr7XrkdddeW3a7M1asKOtDyEhS2NjucXQTWm/hFfbzEvl5fWl0ckkWnxwWpMbGX9CH1ks22K01Mfb7qBs/wdiuvf4d6VZ1trDjinO0JsRJftduE/mFY02p4YAtSIluR0s1bVrSrKu3CeiqJ6Q6rqp6+9+61DtalsmT0vYkq4nTxyCMV9tUSGLVyhgFIYQQQsgg4ESGEEIIIYWFExlCCCGEFBZOZAghhBBSWCj2JYSMLrwkaVrcqxPkAa5u04gZa5xLZr0Sherq1xT/Dhs6+V7QxxoAoJIYBuf3epUVEkuP/m6dgNBx1GPFxuFYs92uVYl9vVsIk6bYMal4C+0TjU/1FDUmL9yqHQHwRCXcbbACaOhkh/VpxWxIZdwLqYxREEIIIYQMAk5kCCGEEFJYOJEhhBBCSGHhRIYQQgghhYViX0LI6McIgJ3fcI5IUrSYsd7ZTmc71R1VSPbT0Yj5fuAIgMWrkO1USFcxEjqdrME6Q7Sujg74FdLbVEVsT5CrReMAoDLySocjUtfV1XX1b8AXnDco4a6JYwCqkrt0qyraFPsSQgghhAwNTmQIIYQQUlg4kSGEEEJIYaFGhhBSaIKXAE8zXlX27XE0EnkqXXuJxabMUgNS/TjbkOOIqOMdbKXz0NJot2s6nLa7O61Pp7JpzQxgNSueX12D9WlwbDp2bPFrq8HyNFlVTgyqCuDiJMQzCQhN8sfKuBdSGaMghBBCCBkEnMgQQgghpLBwIkMIIYSQwsKJDCGEEEIKC1VohJCKJHiCXK9qtRLpXnfdddZHixLdvp0EaBpHpCk1qQLTiI/dCs3keCHV6fEOnoi7+Yi17ftt2nZEu6FLxY0XRx4qSZ04uedQn0Ps64lrtU+NreyNGkclrBPgeQn5FFJTrwyVUdmdd2QIIYQQUlg4kSGEEEJIYeFEhhBCCCGFhRoZQkhFIo62JHhr/VqnkEf/4ibEy6GRcTQ6AapIn+47T8I+MmxYjZKj4/D0IJOmpm1HIyM6sV23E0fdThxpW43zX68nN9GfpduJJd23UzPSjcEOpaXxNGJKE2N0a96+RgDekSGEEEJIYeFEhhBCCCGFhRMZQgghhBQWTmQIIYQQUlgkhMpQ64jIMwB2jfQ4yIixIIQwq7xb8WBsj3kY22S0UhGxXTETGUIIIYSQgcKlJUIIIYQUFk5kCCGEEFJYOJEhhBBCSGHhRKYPRGShiAQRubE/2/HaFyHHC8Y2Ga0wtscmFTWREZFVWWCUvppF5EER+TMRsTnLC0IW9DeKyHkjPZYTjYiMF5EPicgmETkmIodE5F4Ree1Ij+1EwdgenTC2GdujlSLFdqXWWvoWgJ8gVp+YB+B6AJ8FcCaAPxy5YWEXgHEAchRlMSwE8GEAOwE8PIz9VjQiMg3ArwAsAfBVAJ8GMAHACgALRnBoIwVje5TA2DYwtkcJRYvtSp3IPBRC+GZvQ0T+HcAWADeIyAdDCPu8jURkUgih6XgNKsRn1duK0m+F8DkAiwFcEkLYPNKDqQAY26MHxnYKY3v0UKjYrqilpb4IITQCuBdxpr8IAERkp4isFpGVIvJzETkK4JHebURkiYh8Q0T2ikhH5v8JEZmg+xeR54nI3dnts30i8i8AJjp+fa6JisjrsxxLAxIAACAASURBVPEcEZFWEXlMRD4nInUicj2A2zPXr5bcfl3dX78iUiMify0im0WkTUQOish/isjZfY1LRF4pImsz/73ZZx6RCauILARwLYAvhhA2i0i1iJjjOpZhbDO2RyuMbcb2iaJS78gkiIggzg4B4EDJn+YD+DWA7wH4AbIgFpELMvsRAP8BYA+AcwH8KYArROSqEEJn5nsJgF8CaALw8WybNwH4+gDG91EAfwNgM4DPANgL4HQArwfwIQC/AfCxzOcLAO7MNnV/oZRwM4DfA3AbgH8HcBKAPwZwr4hcGUJYp/xfDuBdAD4P4CsAXgPgfQAOZ/sv9zmmAKgt55fRGkJoLePzUsTJ8mYR+Ub2WepEZA+AT4UQPpNzX6MWxjZje7TC2GZsnzBCCBXzArAKQEAMopkAZgE4B8AXM/u9Jb47M9sNTj/rATwKYJKyvzbb5voS2z0AOgAsLbHVAViT+d5YYl/o2C7ObL8G0KD2J3gue/Iqve8y/b4os32nt4/Mfi7imuydzvYtABaq/W8EsDfn8V+d9ZPndWOO/j6T+e5HvFi8HcB1iBeEAOAjIx1zjG3GNmObsc3YLnZsV+odmY9kr156APw3rGDsEKIQ6Vmy23fnIAq06kWkvuTPdyEGzYsB3CQiswFcBuD7IYTHe51CCB0i8hkAt+QY63XZ+/tDCMl6aciiYpD0KsM/WtpPCGG9iPwIwO+IyKwQwjMl29waQthZun8RuR3Au0VkYgihucw+3wtgWs7xbc/hMyl7rwNwZQjhIACIyHcRT5C/EpHPhhAO59znaICxzdgerTC2GdsjQqVOZL6AeNuxd7b6eAjhkOO3LYTQrWwrsnd9UpUyJ3tflL0/6vjkFTgtyca5Pqd/Xk5DvBBscf62CcDvZD6lJ4QXpAez9xkA+j0hQggPDnyY/XIse/+f3pMh20+niNyC+AvuUgA/Heb9VjKMbcb2aIWxzdgeESp1IrM1hPDLHH7eWp9k758C8LM+thvumWTvbbuRRl8cSpF+/hYdRKYjzsLz0Jzjl8Jvs/ennb/tzd7z/pIYLTC2Bwdju/JhbA8OxvYQqdSJzFDYmr135zipdmTvy52/nZFzf48DeBniGuiafvwGesJsRxRcrUCJql+NbQeGlx8CuCqn70cA3FjGp/d4nOL8rde2P+f+CGN7KDC2KxvG9uAZ87E9Gicy6xCFUn8kIv8RQkhu22WPtE0OIRwKIewTkfsAvEZElvaut4pIHYA/z7m/WwD8GYCPicjLQwgdan+SrZX2zoKn5+z3VkQl+/tF5Nre9VYROQvAqwHcpdZZh4PhXmv9DWLSqFeJyMkhhD0AkD1K+fuITxrcO5iBjlEY24OHsV3ZMLYHz5iP7VE3kcmEUm9FVKM/IiJfQVybHI/4KODrALwfwE3ZJn+BqPq+W0T+Fc89xpfr2IQQ1ojIxwH8NYCHROQ7iLfkTgPwu4jq+COIa7dNAN4lIq2ZbX8I4dd99HtbJq56E4BpIvI/eO4xvjbERxKHleFeaw0hdIvIuxAFf/eKyL8hPmnwdgCnAnhnCKFlOPc5mmFsDx7GdmXD2B48jG1U7OPX78vhuxPA6n7+vgDxufydiF/CQQAPAvgHAKcq3+cjPs7Xhpgj4F8BnIUcj/GV/O3NAO5GDPoWRCHaZwHUlfi8HMBD2X5C7/j76hfxpPxrROFYO6La/1YAZyu//sZ1Y/a3hSP4vV6FeIFqQlwfvxvAq0Y63hjbjG3GNmObsV382O59Vp4QQgghpHAUokQBIYQQQogHJzKEEEIIKSycyBBCCCGksHAiQwghhJDCwokMIYQQQgrLCZ3IiMhqEdl5IvdZyYjIQhEJInLjSI+F5IMxnMIYHj0wtlMY28VhyBMZERkvIu8RkTtF5JCIdIrIPhH5iYhcn2VkLDTZ57t+pMcxWETkPBG5UUQWjvRYhoMsrkKZ18kD6I8xXOGMwhg+WUTeLyJ3iMheEWkRkU0i8gkRmdHHNvNE5Osi8oyIHBORB0TkDWX2w9iucEZbbAOAiPwvEblZRB4VkW4R6TPPi4i8N5tE7xWR9uz9dhF5bV/baIYUxCKyGMCPASwF8EvEpEUHAMwGcA1iqfYzAPzVUPZTAbwHMUHTTSM7jEFzHoAPI2bC3DmiIxkefgPgrY59LoB/ArAuZGm1y8EYLgyjLYZfhZj07McAPoGYdOxixO/pTSJyUQjh2aJ9EgsD3oUYl59GLOx3LYDvisg7Qghf1TtgbBeG0RbbQMzCPAOx9MQE+HWberkY8XP/BDE+pwN4A4AfisiHQgh/X25ng57IiMg4AP+DWFL99SGEHyqXj4vIRQAuGuw+SOUjIgJgQihfUXXYCLEOi6kZIiLvz/755Tz9MIYJMDIxDOBOAAtKJysAvigi9wP4IoD3Za9e/g9i+vxXhxB+BAAi8mXEmjefFJHvlY6fsU2AEYttIGZ73h1C6MnKNPQ5kQkhvFHbROSziBmd/0pEPhZC6K9C+OBLFAD4E8QUyv84gG1WA9jp2J8P4DYARwEcQ0wH/U7H70wA3wOwBzH189MAbgfwCuVXD+BvEGt1tCHWx/gRgJXKrwpxNv8I4i+iRgCPIf5HWJv5hD5eZVNHA3h99pmPIKZ4fgzA55Clv4aTohrPpfu+3unvpviV5T8meC7VtX7dNIjj9ezYEGuHbM72aVJsn+gXYrn7rdlxnsoYZgwXLYaz8U3KxvczZf8tgCcc/7dm/r/H2GZsV2JsI06owyC2+wmAHgAN5XyHsrT0u9n7F4bQB0TkVQD+E/EL/hRiwL4JwJdEZFEI4QOZ3wzEug9ArMWxC8BMABcCuATxFipEpBbAzwBcDuAbAP4FwBQAf4BYYOz5IYQHsn4+AODvEIPi8wC6kf3qQQyeTsQLxWcQb3l9tGTo/VYwFZGPIgbf5mz7vQBORzx5PoRYR2RI5DwmP0RccvlDAB9DrP8BANuyPgZyvHp5D+Jtwy8ifm9PlhnnFAC1OT9WawihNadvKVchFpf7ZgjhSM5tGMP9fy7G8HPjPBExDDz3y3Vfyb7nAjgZwM2O/33Z+0UAvltiZ2z3/7kY28+N80TFdr9ky6fViMfoDQBeCuD2EEJb2Y2HMMs6CODoALdZjZIZfzboXYgzzHkl9jrEAlXdAJZktlfD+eXh7OPPM7+XKPtkALtRUrAM8ZfF5hzj3ol+Cp05/hdnY/g11GwS8c5Bb42rhRjCjH8Ax+T6zG/VEI9X79gOAZg9wO+9r19O+nVj3n7VPr6RbX8VY5gxXMQYzvbz3ayPF5TYLshsH3f8x2d/u4WxzdiuxNhGzjsyiJPR3v10It7JmpVnH0O5IzMZJb8aBskFAOYD+EwI4aleYwihQ0T+CbFi6GsAfBLxtiYAvExEfhZCaOyjz7cgVjB9UERmqr/dBuBtIjIuhHAs6/N0EXleCOGuIX6WUq7L3t8f1GwyZN/YMJH3mPTHQI5XL18PIewfwD7eC2BaTl+jfSmHiExF/CX1RAjhjgFsyhjuG8ZwynGNYSA+vYH4S/QLIYRfl/xpfPbe7mzWpnx6YWz3DWM75bjHdk5eB6AB8e7jGwCMQ1xq7ffuGjC0p5Yas50MhdOy903O33ptiwAghHCHiHwdcfZ6nYisRVTifyeEsLlkuxWIB6C/Dz8T8Xbb3yCejHeKyFOIM9MfA/h+CGEotxaXIM4q1w+hj7IM4Jj0x0COVy+PD3CcDw7EfxBci/gZcol8S2AM9w1jOB3ncY1hEbkB8emlHwN4t/pz7638emfTBuXTC2O7bxjb6TiP9/U5FyGE35Q0vyoi30JcPjsjhHC4v22HMpHZCOD52Trp8ZylPUsI4W0i8gkALwNwJeJM8gMi8p4Qwr9kbgJgA4C/6KerZ7L+7hWR0wG8BMDV2etaAH+b/Qo4NJThZq/BbNcX5vvKeUz6I/fxKmFAa6TZ2mddTvfmMHCF/TsBdGHgj18yhssMF4zhuIPjGMMi8g5ELcsvEJ8w6lQuvXdDvNxIvTadboCxXWa4YGzHHRz/6/Ng+RqiHut1KPcjdaDrXSXrWX+K+KV+bADbrEa6BntR1senHd/e9cX39dPfVESleTOeW9d8BFG4VTXIz/WubL9/WWLbgYGtwX426+PiMn4LYddgz85sf+H434Mya419HJO3oe812NzHC/2sD+f43o/LGixiDoYA4FbGMGO4aDEM4B2IT2b8Av08nYHyTy29kbHN2K6k2C7pf7BPLb1Wf9d9vYaS2fdL2RfyPhF5jecgIheIyLv66eMhRMHS20XkpJLtagH8ZfYh/iuzTReRZLwhPp2yA3F9uPcW69cBnIQ+ZrAiMqfk33rNsXdMQEzK00uzapfjluz9YyJiZrrZs/19sQPx7sI1apvLAVyqbHmPSe8M2vsMuY/XEHgvgBflfH19gH3fkL0PdFkJYAz3B2M4ZdhjWGI22i8iik5fE/p/OuNbiHqRV5VsX434mPURxEdVS2Fs9w1jO+V4Xp/7RUQmiMhEx16N+Ag58NyTeX0y6KWlEEKriLwScc3yVhH5BaL46CCAWYi3AV+CmGm1rz66ReTdiI/3rRWRLyA+3vdGxKD4WAhha+b++wD+XET+E8ATiKrmq7J9fDc8J3b6f4gH/BMi8gLEi0QjomjthYjiuKsz3y0ich+A+xFv3/Y+BtcB4NslQ70PwDtF5O8RH4/rAfCjEEJLH59rjYh8HMBfA3hIRL6D+BjcaYiPRV6MePHxtm0WkZsA3JCtEa5GXNN9O+Ls/NwS97zHZG025g+IyDQALQB2hBDuH+DxGhThOK3BikgDonDvKdgLeVkYw4zhvAx3DIvIqxEn340AvgPg9er/z+YQwq0l7X9EFEDeIiKfRlxKejPiXZMbQghNaryMbcZ2Lo7H9TmbcPcei8WZ7W+z9pHw3LLaEgB3iMj3ESfehxCXS98MYBmAr4UQ7iy7w4He7nFu/4xHfETsLgCHEb+sfYgn0FsBVKtbWDudPq5CPMkaEb+YdVAJlxCXEL6GGBAtme96xNlkvfKtQby1ujbzbUFMlnYzgBeX+P0fxHT3+xGfCHgS8ZGv81V/swH8IDvIPYi/RBbmODZvRnxMsSkbw6OItzX7TLiU2Sci/qI6iLjeeSdiHoGbUHKLboDH5G2IORM6sn3eNIjjtQqDuHV5vF6I6+UBwEcZw4zhIsUw+k6E1vvyYuxkxDQDB7IYewhqSYmxzdge6djOxnJTnthGFCr/S3ZcDmXxeSCLt+uQLb+Ve/Wu0RFCCCGEFI4hV78mhBBCCBkpOJEhhBBCSGHhRIYQQgghhYUTGUIIIYQUFk5kCCGEEFJYOJEhhBBCSGHhRIYQQgghhYUTGUIIIYQUFk5kTiAicrqI3Cwi+0SkXUSeEJGPZKn2te+LROTzIrJWRNpEJIjIqgHu78ZsO+/1vmH7YGRMIyLni8gnReQhETmcvdaKyLuyujzltv94FpMDqqorIvNE5Osi8oyIHBORB0TkDYP/JGSsIiJXici/isgGEWnMYupuEXmzrr0kIqv6ua72vq5Q28wXkf/IrvnHRGSPiPxIRJ4/gDHWi8jficiO7P+PbSLyt3nOsdHOoGstkYEhIssB3It4zP8VsXDYZQA+COASEXlZSNMsX4eYgn8jYv2Q84aw+z9HTPtcynGpf0TGJH+FWETvVsQiiNUAXokY568RkZeGPlKIi8h5iAXxmgH0V6xPbzcdMe3+bACfRqwOfS2A74rIO0IIXx38xyFjkI8DOAWxrtQGABMQa0rdAuAFAP6gxHcLYnkHTT2ALyBea9f0GkVkHuL1tgbAfyCWF5iX9Xm7iLw6hPDjHGP8DoDXAPgK4v8llwH4e8RaRtfn+5ijlJGuyTBWXogX+R4Alyv7+xHrT7xF2U9GVo8DwPvQR5n3Mvu8ETnrjvDF12BfAK4A0ODYv5nF3yv72K4asX7MfyPW+WkewD7/Kev7Vaq/NYg1cCaO9HHh68S98FydoVWD3P4qlNSdymxVAO7I+j0rRx9vznw/oey91/jXKPvizH5rjr5fnvl+Stk/ldkvL9fHaH5xaenEcTWAx0MI9yj7Tdn720uNIYQ9IYT24dq5iEwWEd6BI8NOCOHuEEKb86fvZO9n9bHpnwI4A8CfDGK31wLYFkL4Uck4ugH8M4DpiBd+QnIRQrgji59SWw+A72fNvmK4lBuy9y8p++Ts/Sllfxrxx61bpVtxbfb+WWXvbb8lRx+jFk5kThz1iJVSNb22i/Va7DDyCICjANpE5B4Redlx2g8hpZySve/TfxCRBYi3xT8SQtg1kE5FZC7iHcv7nD/32i4aSJ+E9EGfMVyKiJyG+GP1rhDCY+rPv8je/y3T15wsIhcB+BbikuqncozjIgB7QghPlhqz9lMY4/HOicyJYxOAFSJykrJfnb1PBDBtmPd5BHHN9k8Q11bfD2ABgB+LyPXDvC9CnkVEJgL4S8QJ9H85Lv8OYDuivmWgzMve9zh/67WdPIh+CXmWTNvyh4hxelcZ93cgarz03RiEEG4H8McATgNwO6Keaw2AZQAuDSE8lGM48+DHOzL7mI53LjU4iMhUAO8ZwCafCyEcKuPzKQA3A/gvEfkrADsBXALg/wHoBFALYDyAcv3kJoSgb0NCRL6CKCD+jIh8P4QwoCdFSLE5TrGt91GNqI85DcC1ensReTOAlwJ4XgihayB9Z4zP3r2l1zblQ0YZIlKFuHxYypTedxGZqf52KFsmGsg+xiMKfycCeHUIobMf32pEsW0jgO/14fYMgAcA/BLA4wCWIk70fywiV+k7LQ7j4cc7EGN+TMc7JzI+UwF8eAD+30SZCUgI4RYRmYF4O311Zu4A8DEAr0C8Ndg44JEOkBDCQRH5PKIQ+HI8d9uTjA2GPbZLyf6T+QriHcAPhBC+pf4+HXFd/8uOXiwvvcux9c7fGpQPGX3MR3zq0+NWx3Ya4g/HXGTpMG4FcCGAt4UQ7iyzyUsQl6D+I4Rg4k5E/gDAvwFYGULYWGL/OYCHAPwDymtcWuHHOxBjfkzHOycyDiGEnRjAo6AD6PefReQLAM5GDMpNIYQjIvLHAPaGEI77RCZjZ/auf7mQUc7xim3g2UnMlwD8PqL25WOO24cRH239oogsLrGPi13IYgDtZX6h9oomvdvpvba+bsOT4vM0gBcp27kAPon4hOd6xz8XJZOYawC8M4TwzRybvTN7N8tKGe8H8GjpJAYAQggbRORRxCemyvEU+l4+OhljPN45kTnBZE8iPdDbFpELAcwC8OUTOIwl2Xu/AjZC8lIyiXk7gP8bQrixD9cFiBOZ+/v4+1ZEPVmfT4mEEPaKyB4Alzp/7rU94PyNjAKyJ+R+WWoTkd4lygdDCKsH02/JJObFAP4w5MhFJCKzAbwKwPoQQl8xdzKAbX38rQb5/h9eC+A6ETm1dJIvIqci6mf+O0cfoxaKfUeQ7MT5LOLa5yeH0M8UEVleujYsIjUiMsXxPRXA/0bMtTHYW/uEPEv2tN0XEScxHwshfLAf948DeIPz2oy41v8GxASOvX2Pz2J7rurnWwBOF5FXlfhWIwrbjwD4yVA/Fxk7iEg9oibmxQD+KITQ190Vze8j6hv7+yG6GcAyEUkm3iJyGaJWZq2yLxeR01UfvUu0Wt/W274553hHJZIl1SHHGRE5EzFnzP8gqtbnAHgbgNMBvD2E8A3lfw6AV2fNKxDFkV/Bc2vD/xxCOJr5Xg/gq4i382/MbFMz31sRM1EeRlTJ34AoYHtzCKEvYRohuRGRTwJ4L+ItfW9Cvi2EcG+ZPlYDuDCEMFHZVyE+6fG1EML1JfYZiNlSZyA++bQHMSHZKgA3hBBO5B1OMsKUxMnVg7kjIyLfB/B6xDs9X3NcHgkhPOJstwXAQgDzQgiH++j7NQB+iKhj+TziXccliD8oqxFF7w+W+AcAu0IIC1U/P0LMmP1lPJfZ950AvhlC8DINjxm4tHTiOIA4gfkDxLTqRwHcCeCtIYQ1jv/5iMLgUt5R8u9vZn30xTEAP0B8Mup3ECcvBxBP1H/qY5+EDIYLs/dzAXzD+fvXEC+8w0YmWr8CwD8iPto6EfGX75tCCN/pd2NCLL0xfE320nwEMR/Xs4jI5QCWA7ilr0kMAIQQ/ktEXoT4lNI7EJ+wOgzg5wD+PoTwcM4xvgHA3yIKg9+KOHn/EOI5MKbhHRlCCCGEFBZqZAghhBBSWDiRIYQQQkhh4USGEEIIIYWFExlCCCGEFBZOZAghhBBSWCrm8euZM2eGhQvmj/QwhgfzIJj3ZFiep8WGM5O89NscaXbu2o0DBw5U2KiGh1EV23kwT0I6X+uo/KZ9GNsnGBN/zrXWvfzmiFuN6+LF+6j8+ismtitmIrNwwXw8cE+5SunFIPR0pwbddm3OmSXV1lY12JtoajunHxnBk+3Cy583Yvs+3oym2M5D6NYFrW1cSbUT28cRXfw4VlQ4MTC2TyyhWxWqNvEIoMcphq2vye61Vsr7VDv/rValtpG81g4nlRLbXFoihBBCSGHhRIYQQgghhaVilpZGE1KV3jYP3m1svY4bnFud7t3Hyl4iIsQsk3qxjRO7tDSmRDljHKmuTQ26DcDNaK+Xm9y41Tbn2l5l/1vlNfn4wjsyhBBCCCksnMgQQgghpLBwIkMIIYSQwkKNzABw11XdR/u0zctjoGzeGqpbmVw9Ithl968fNY39a22Ns457gh+JJSNH6FKPqJq1f+R7/N+NUeVTXQG/l/Sjtd4jsqRwuNe6jmNp28Q63JQYOm2A0doAgL5GOtdRBKdv7Vdlzy3qaAZPBVxhCCGEEEIGBycyhBBCCCksnMgQQgghpLBwIkMIIYSQwkLFW4YrGtMiMU801tVh++poS9pSV2+300IyR/yFKk9IqcbpiY29ceqEYPUNzpgo9h2NWGEvgLaWtG0E6rACcc/mxXZtGlsnWsQYusvXNgsUW44OvOtva2NqaGk0PlDXaABAZ3qehAmTrM+41CbedbTHq7WkxeY2/oInrh9ETbDBPrRha6TBCqe1SN5NGnji4R0ZQgghhBQWTmQIIYQQUlg4kSGEEEJIYRmzGhmT3K6z3Topm6s18LZT67ahyq692/V4r4iZNekEZMFZI0Zba9ntpMeu/xptQW2d8XGTRJHKxtMRtLeW9XGTxtWk3784sa2TO5pkYNHJmnRfTtIyU9gPsAnIPI1Mtzp3HT1AqLV6ByaJrHCchKDm+qdjHQA6cugdPf1XvYotLx7Fu3CrewadzvnmxbtJnOoVqUxtnv7LKy5syKP31Nd/7/OPALwjQwghhJDCwokMIYQQQgoLJzKEEEIIKSycyBBCCCGksIwJsW+uqtWOkDfoZEAdjrDXSwik9+eID0NVV1kfN0mZFrfpxGYAcKzZGVM6TiP2BICGCUlTxk+23dSPs9uNcAI0kmISW3lCXv39e8mwah2xoxIbBke0KFoA6AkN3WR7WuzrnFvOOWE+r5fsrF2dy/Xj7e7HTTC2UKcEwLU2/hnvJw5zLffiVtvyClJ1nOYRyHrfvZvETgngvYdEmg5Zm/4/x01AqWy5xb55HjhR/2/o/yeZEI8QQgghZGhwIkMIIYSQwsKJDCGEEEIKCycyhBBCCCksY0Lsm0s06AkitWjWy+zrZS3V2UC9jI2qr7X3329ctm7eaLfTIrHWJuNy6TlnG9ui+aemhjpH7Ft3NGmGKVaQJlNm2u0UocbJCOwJ0MjxQWWxdbM/6zjyxOZeZl9Njmygax94wPhs3b7d9qWr/+psvAAuvfACY1t0yrzU4AnZjynbOCv2dQWYE6emQ/SEnErsHipEADkq6VZZ050YsWJfR8TqkUfcq1i7dq2xbd2+0zqqjNjh6AHjcumik41t0ZzZqaHBediiPofY1xMlaz9P3K9t+v/NvMf2OMM7MoQQQggpLJzIEEIIIaSwcCJDCCGEkMIy6jQybvK7Hq/SqdIReImVjLbGWWf0quMqjUhHp93/+gfXJe2tW7caHy/RVtA2Z83+vvus3gbtqW5h0YIF1kev7TtrxmGcrZotOkmeqxGoVi5Kk1EZS62Fw1QsB6wmy1szdxLCGRqcpHFqzby91SZkXL92TdLe+uijtp/ye3fPyftW/9o6nn9e0lw0Z5b1aVFaMm8ADTYhnnHzri8afd3Isw0xuNdynQDU0zXpa6JXod2rmt2R6m86uu117GF13X5iz9N2905iRX0OmmsmgPse2Wy3OzM9lxfNn2992lSyR6XHAQDUeTad7NHRyOi+OtUxqZBckLwjQwghhJDCwokMIYQQQgoLJzKEEEIIKSycyBBCCCGksIw6sW+uaqiATbblJQ3LUzXVEVIeakyFhevWPWR89j9tRWKaKVOnGtvS0xcl7V2bHzE++7bZ6tdr7rk7aR/YscP4XHTWmUnbE6RhmiOu00mRxB63AC3u1T4URA4KT8iu4t0VjSshr5uw0BH/HTramLTX3X+v8dm3a1tqcBJJTpliReNLTz89ae/a/oTt+6mnjG3NHbcn7QOzbdLGi5YvT9riVRGe7AinzbGzcWpEqUyINzx02MSGoTlN2ukmG9XfmZfozhHJH9ybxtbDj9sHMPY1KbHxhCnGZ+pJNrHd0nPPT9q7Httk+3aSm655ZEPSPrDnt8bnomVLk7Z4wl5XgJzjv3+dFNNLClsB8I4MIYQQQgoLJzKEEEIIKSycyBBCCCGksIw+jYyXoMfRugS91ufpYczaqu38mYOHjG3t+nRd82hjo/FBXao/ufTii4zLKc5af50q2jV/mtUatJ5mk909/US63tt2yI47HEsTK0n7MePjaolM4ipH71JeajCm0LoKT8dit/GKnzoaAe3nJRbTScOc/T+zf7+xrX0o1Xsd2e9ovVSysUsuWGlcTp0319jqalKdzvyZ041Pa+NRY9u3e3fSbjt62I5J9e0mDfNsupBlpWQAGwvkKUjoJn9rKOuzPg9VVAAAIABJREFU/8BBY1u7KU1Id7TNudap6/Ylz3u+cZm/dIXdbPK01GeOvba3rFhqbPu2bknabfv22DHp/6e8Yq/ucVLFfb0EmCpJZNB9V0gxYN6RIYQQQkhh4USGEEIIIYWFExlCCCGEFBZOZAghhBBSWEaf2Dev+ChPYh8lbGprt+Kv2++4w9i6daKhGpt869IrU5HYooVONep2LyHUkaRdV23nojt27zK2meNTIdeMOXNs30rc5lVWFi+5oBGhWkGkTrhm9acUUZbFEz+61d57lEt5kfCx1jbj8utf/MzYurUoXlcjBnDJ+ecm7UWn2gRhXkIydKW2OickdnqxPSVNSjZjyenGx4ibJ022PvVO0jAjbszx28+pSD+WGIyQ3cW5tmHcxLTvaud6r66/bbo6NIDVGx43tu7xqbhcJtr9X3r5FUl70RIr0HXFtuocrKu14tudzxwwtpnqOj1j4am272aVSK++wbjcqxLrAcBlL3xJ0pYJNgGrjFMPk2jRcJ7z4QRQGaMghBBCCBkEnMgQQgghpLBwIkMIIYSQwsKJDCGEEEIKy6gT+4ojtHPzQ3oZURVdasM77rq7rA9gZatLVpxhfE5futzYNJ7YVgsJ9z+117g8fK+tSDy7PhUcX3WlzUhpRHpetmNP7KuPZR5tnxaJjXKtr6mQPMC/RycnHrys1Vrc22W/s66OtIr5Hbf9wvroSsOx86S5eIEVHxpxryes92JL2fY7lX7X3XePsc2ZPTtpr3rFq2zfE1WVYqdqsTgiSWgxqcn0a8WswcT2KA7uEBD0NUFdD0KNyiCLnAJg51ouKmvvhi2PGp8VZ52dtFfff5fx6W6wGdFRn9qWOKLxRWeea2ya21ffbmxXr7o6ae93smave3idsc2elArQV11+ibNHdZzq7cMl+5qs4HlfcyrwP2m2zeyrkVp1jlSIsL0yRkEIIYQQMgg4kSGEEEJIYeFEhhBCCCGFZdRpZFy8dTxn3VazcUO6ZnnwqK1iLU5V0TPOPidtr7TVf3NRZdeRjzSnCch+4+hhupwkUQ1T02RHMmmi8RG1ti+TrI7Aq5CKanUMqssf20EnySoioQfoVGvU+pg5lcaDTojo6ZPyHEdHo7LhoQeS9sH9+2zXtfZ7POPc85L2iiWLnTGpdq4kisARVZH9N2seMD7dTsLLhnHl1/bR44xB4yXTHES1X0+nN2rp6QGONac2reXSsQ4g1Kjj6mnEPI2g0lFt3rjRuDyycVPS1sk4AbiJ3M4888ykvcLRNkqO/zeONttz+Xs/+GHS7mppMj7BSXjZUKeOnZPcD01pklRMsMke586eZWx333Vn0n79qTYpa5C02je6VFJYL9nmCDCGzjhCCCGEjDY4kSGEEEJIYeFEhhBCCCGFhRMZQgghhBSWsSH2dUSzWsjYesyKqJ7YlVbaFaeq6Ulz5xnbijPOStp1DeOMTy4ckdqBI2mSsg4vIV+dTYg0RVe7nuiIffV2U2w1VL9CcDrOMSXkzUNPD9CuqkvXqyRqLUqwBwD7VEI477A6wj5dIbj1qO172yZVDdcRGs6Zv9DYVpyTin3rtWgTQNAJEj3RpiMSPPBUKjju8KoIO+L6ydOmWT+NTsDnxaiX7M4RqpISeroQmlORtknA6FyPUKtsniDX+Y70taWrs9346IcNgiO+nnvSXGNbsXxF0q6vKy/s9ahxHrZoVeeAn/7Sft7Jk1XivrZW4xMa0/8Tdj9pE0lue3y7sS298LK0H+eBA+ksI+71EluOALwjQwghhJDCwokMIYQQQgoLJzKEEEIIKSxjQyPjJWhTpvvutAUhO5GudVbX2cN1tkoQBgD1OomRl+wph4xk/SM22dPj2lbvJAPzEoRNSLUtMtHRVuiieeOtj3j704XESEpPD0JbmjRM9Dp6i022iEalbanOmdhRFda79+47jUuH0uxUOTqGc846y9jqdZI8T8egdSxO/K9fZwvkPbZlc2qoc7Rlnt5M2/L45NS+aL2Pm1xtLNPTA7T2nxDvsc1pgjoAWKYKO8JJvujF9kOPbLB+hvIFWBsb7flW7xRbLMfPf2GLrTY22b5rlZaswysA220TVx7Y+1RqmG6vyfueSn3uX7fe+Kx8wTXGtuKsVBOEY06SPvMflTq2OYovnwh4R4YQQgghhYUTGUIIIYQUFk5kCCGEEFJYOJEhhBBCSGEZE2Jfr0K1TqzVJo5PQyqQra21Qr9Zs2fb7XSSph4rhj14OBVybthoRWxPKxEXAHSrxGVSb4WNwRPyTlLJ7SZ4ie2UuG6cTZrnCSmZAK8MoccksgpaNNqhEua5eMfZSUilBHjt7U7SMJUkrK7Wxv/MqZOMLbSl1ddlnI0jXSV+w0Yr9tz32yeNrVslDfOqbwdnfya2J0+3PjqW6xyBulchW30vwUl2hiotblYCSE/sP1oIwR637jQmH7j/PrNZh0qad/YFFxmfjZs3G9uO3TpJpJPEUMW2d/RrnESOt6++PWlfvepq4/O9738vaXd0dBifaieRY2drKqSVDntO9hxrMbYnn9mTtDc64fe4Es4vO+NM47Ni/ql2wyaVyNB7cEAnLtSV7JkQjxBCCCFkaHAiQwghhJDCwokMIYQQQgoLJzKEEEIIKSxjQuybB1euqkRjXc68b926h40ttKbVSHc8Y6sPtwfdl5WkTZpoxZaLli5O2o8+/KDxafMqwk6akrY9Ia/6vF4WX6kZePbLMU9PD6BEsqaKrCMaNHiiak9rNwhBXmenzSq6bs39zhjSGNlxxAoU26vSGDHCZgCTJzmxvXxp0n50wyPGp82pmi2TVPXryTOsjxbFexlJu+wxCMrmPjhQrfrSfTtjHj0EG29K7HvhSpv9fN0jafbZTY8/YXyqJk4xtudf8+K0n4dtjBw5ml5/a2ps/E2dbB+I2K0E6DffcrPxycP4Wvv/xKKFy5P2ow+uMT5t7fZcCkoUvP4R+3mXLTsjaZ977rl2UI6QGLra9Xjn/wT9X4C+TlHsSwghhBAyNDiRIYQQQkhh4USGEEIIIYWFGpmMhgabIEsrErq67Lr6lm07bGd6rbPBrj1OnZ6u/56x4gzjM3OGXetvOvB0uitnTDJxqrVNmZkadBuwGgydDIkMjqpqYIJakx+nNCKOPsOsa3saGU9/oRLLebENlRCs20ks9uj2Xc52SiOi9SkAps5K49aN7clWf9X4dJrsrL3ZqQjeZZPWhU6VTLDpsPU5pio0e8nvnL5NPw1O9XetJeuqTB3BcUGq7HWiKj2OptI1gLppaYzcu84mBL30siuMbe7JaWK3LY9vs2NqTpNP6iSiALD7SZuQMajEhVJlz7cpk/V1e4XxmeXF9r40uWlb01Hjo89bAFi5cmXSXrd2rfG58MILUoN3nXCSuUJXvNc6SgCYqM7vDqWrcRILjgS8I0MIIYSQwsKJDCGEEEIKCycyhBBCCCksnMgQQgghpLBUhlJnBNDCrssuu8z4bN++PWnv27/f+HiJ9EJ3KtxctHip8Tl1/vykXV1l55S6HwBo2q8EYV413hpbIXjinFOSticINtQ6IlEycKqrgQnp8RYlGtVVpQEANUq06iVx88S+qvqyG9u7U7Hj/iOOsNaprC6qQu5pK6yQc/5pi5K2G9ttzcbWqCtNaxEv4IqiJ+qEZ42HjI/ZzqtI7dn0+eWNSVcy1mJfnTBuNCFixb4qaaJXafy05WmF5l2HmqzP4sXGBpVc8bLnXWlcduxIH8Dwrtv+95+eS3PmzDEuy5enie2qHNG4dy43HlOfT4tmAVfsO3FCei2/7rpr7XZapO5UjUe98+CGTorqJZLUDyl0KCFzVWVMIXhHhhBCCCGFhRMZQgghhBQWTmQIIYQQUlg4kSGEEEJIYakMpc6IkIq9xo2zwsYzzzyz3/bxx0qJd+zcmRqqna/QEWDNX7xk4Hv3hMRk4EiVEfdKvRJkT7BZNcMEJfbtcISmjpBWiwbHjbeZRs88J62Qe6YjyBQvs7MS+6LOEQSrMWlhPQD3s2zfoCrJexV7nb7mz1FZqptstfnQovryhNNeBl4lLhXnOoFx6rvsVGJfL2vzaKGqGtAPDmhxsxNHUp1miL76RS/x+7ZbJq3xXmwP8rodtHDeiwcdf45A11SVBrB9y8bUoGOkj/3NVw+FuMJxncnXq9Cuhb2A+d7E8ZE69VCC/r/Fu/6MAJUxCkIIIYSQQcCJDCGEEEIKCycyhBBCCCksY1YjIzpp0zCi11oHu68jTTZp2O6n0urXnh7m0stt1Vg7Z/XWf/OPjQyAqiqridF4a9jTZqftFq8atLPW3qnW0T2Niqs/UJs5CRm1tkQcrUnQOpquduNzePujxvbkugdSg1PF99IrbQI0tKf93/zFL1ifo6racLsdU0+bPZZBVU6ucqoIX/f261ODrkjf5mibRgvVNZBJ05VRx5uTNlR/t+61x0tap+Nt+HR8+jodnOrXpmq6k5Dy6AGbgO/JrY+nBkdbc+mll3qD0gbro6lzEuLpxHYAZLyy5UiAqhNiupW2RwDekSGEEEJIYeFEhhBCCCGFhRMZQgghhBQWTmQIIYQQUljGrNj3uKIFkNWDmy9u2bLZdp2j2ui8U06xRq3JCm7d7nwDI8OOeEI7JaIMnkDXq/TcnkNcmieRla7qDABKAOyOSSdpbLWVjTf/5te26317U4OTfG7uNKdquxbu6uR3AHqOpGPoOtpqfDqa7edt70jP5SpH3Hh0y1NJW2uru9pHd/VrIwAdJlyx+QmsJC7Odx30hdQR+256ZL2x9bSoBzec82bu3Ll2EFpw7J23PcrHS0DoPUygHkAocgJU3pEhhBBCSGHhRIYQQgghhYUTGUIIIYQUFmpkjgeDkJp0dNr14Gf228RKWqMwa9Ys41LrDcAkTvOSTaW24KwRi1ekkhwf1PGXeqsZCZNn2O1UgUp4CRl1gTqv+J1O/uX5dTk+IS2a13HkoHF5Zud2u11rqluZNWu2calzxhl0kb5Wq3/palRjcvQwLS32HOzqSs+JhgarI6ipSY9vV5eTbJIMGF1YEgDg2U4k6hrZ2WGTKB54xrluq8vtrJPmGJe6Bqcgqf68IV+xSbN757wRfX5TI0MIIYQQcuLhRIYQQgghhYUTGUIIIYQUFk5kCCGEEFJYqNw8Aehq2ICttLpjuxU/Nh+xyc7qlf72nOXLjE91j1e1uMwg46DStqsHdkSh1WlCLC+R1FinXEV0N/mXpsZJdOXYdKVbI4YFgFZVSbvTERG66GxvzriVcH3Hxg3GpdkRstcqYe05S5cYnypvf0rs2NNqhbxdLamttdWKlNvabGyr4tcYP95eMusmpckMq9S+vCLKpPIJXtV4dR7v2L7NuDQ3NRtbvUrueM55K41PdZ1zLutkqs7/JUbs6/l0Oskt1QMgpmo9inMt5x0ZQgghhBQWTmQIIYQQUlg4kSGEEEJIYeFEhhBCCCGFhWLfIRJ0pWsARlnraK+CpEKyDevXWScna+q0mdOS9uzpU2zfHY64U+Nle1U2qfYG7gngVNupvjqm6ekB2tNss7mSP2snr4q6l+lUf4+OQNZk+jSZn3NyzFaaRlsafxvuu8fZzsbo1EmpSHn2tGnGB21OZW+lyO1pt5+3s7O82r3KOb61yjRugj3eNdMnKIM+j/h78URirsneNVqcLLa6snS3c04oUf6G9Q87fVvT9Nlpluo5XqVrT1frXW/L+XiC+A7nvKlLReqobbA+NSOcSTknPMMIIYQQUlg4kSGEEEJIYeFEhhBCCCGFhRqZoeKtYZpqpHZ9/uDBNNlde6tNouQpKZYuW65cnLX/Nlv916ybelWslS00TDAubkVaT29DnqOnG6HlqLEleGvR6lhLt/edOd+/SmIVPP2L3n+3U8Xaqchu4qi5ybgc2L0rabd7Vdy7rG5h2coVaozOuXXMrvX7OrWU+nHpsWyYZHVcVXX2+FY1pNtVT51ofGR2WoG+dkqafEzqiqEzGBQ9PQhtSielqyo72gs5njo6HQ9uEkVPa5WeS8Gp/n5w/76k3dZ4xPbj6G+WrLwgNTjJ59yklFon6X0Wc2139D81VssWalUiU62ZAcx1yVxL8mh4TgD8H4gQQgghhYUTGUIIIYQUFk5kCCGEEFJYOJEhhBBCSGGh2Heo5KpGagVRM6amyb+u/b3fG9zudRVjwE9S1qnEbZ64VAvQPEFa3Thr04mkSEp3F3D0QGrTQtoG57gqsXXwquP22O/RVKz1kmHlEUS2OYkVVWyFo0eNywzVftPLX2nH6CWJ02WiHUFwt1PZWvtde+11xiWoc7J2mhWyY+pUO87JKuFkrXdOqO9FC6e9bUYL3Z0IR1IBrHnYYIqOCADjJqVt73qUK2mdI1LXIl0vIWS7c41sVcJ177qthnnt637HGaMntlX/1WrxPwC0OmJffSzNgySwgmDv4QvvGq2v787DHUGLsrUgmWJfQgghhJChwYkMIYQQQgoLJzKEEEIIKSzUyAwVrUcAbPEvzycom1cwzEv0pTU53nponjV5vWaL/9/emcfZUVZ5//frTnens5MFkCB0SMKmyL6KJOIKLqg4jiMqcdRxBnVc0NcBt6gDjgviODoq8GpEAXEBFPVVQYgSCIsEEiFhC3SArJA9vaS38/7x1DW3nufpvnVvd+fe6v59P5/63K5zTz31VNWp6nOf59Q5COepY4mVYtTIPGnNwrpw/nmMN28fS0blxV5EkxHGEhv6NhHbzp8zr4u00xDZrs/rZ3MsZiV9bHV+7AsQjX+xQBZuV9cY2rvVewkAOyIxEV3pOILenWHcUJ1tDfvZ6R3f+HGBCpvTMvPjn2JxHCOFunpwfDqOyLwEeGyKxCON8Wy7wuKEFrNt/xkZeY7Fkr+Zb2+x569/LWM6MZkfpxaLdYk9k/1neRZbip3L2P8EXxZ7lvjPhXo/Hqd0d/YGGpERQgghRG6RIyOEEEKI3CJHRgghhBC5RY6MEEIIIXILrUYCNUk+B2BNSUUxUjnYzGaUVssfsu1Rj2xbjFRqwrZrxpERQgghhCgXTS0JIYQQIrfIkRFCCCFEbpEjI4QQQojcIkemH0i2kDSSCweSDde+hBguZNtipCLbHp3UlCNDcn5iGMXLLpL3k/wIGavrng8So19I8phq92VvQnICya+SXE1yN8kNJH9Icma1+7Y3kW2PPGTbDtn2yCNvtl2rtZauA/A7uEoOBwBYAOCbAF4E4F+q1y2sAdAMoJLiKS0APg+gFcCDQ9huzUKyGcCfARwL4GoASwHMAvBBAK8geZKZbahiF6uBbHsEINuOItseAeTRtmvVkVlmZj8prJD8LoBVAN5H8rNmtjG2EcmJZrZzuDpl7l31sNpcjbZbA3wAwHEALjazLxeEJH8NYAmA/wTwvir1rVrItkcGsu0Q2fbIIHe2XVNTS/1hZjvgvEICOAQASLaSXEzyWJJ/ILkdwIrCNiTnkvwxyfUkuxL9r5EMSrGSPJ3knSQ7SG4k+W0AEyJ6/c6Jkjw36c82ku0kHyX5LZKNJBcAuD1R/WHR8OvigdolOYbkp0iuJNlJcjPJG0ke1V+/SL6e5H2J/vrkmKvlsL48+fxhsdDM7gLwOIC3k4yUfR49yLZl2yMV2bZse29RqyMyKUgSwJxk9fmirw4CcBuAnwP4JRIjJnl8It8G4PsA1gI4GsC/A3gpyXlm1p3ongzgVgA7AXwl2ebtcENqWft3CYCLAawEcDmA9QBmAzgXwOcA/AXApYnOFQDuSDaN/kIp4hoAbwNwC4DvAtgfbnhvKcmXmdkDnv7ZAC4A8D0APwBwDoBPANia7L/UcUwGEKkBH6XdzNpL6DQVdGPbAxgP4CgA92Xc54hDti3bHqnItmXbew0zq5kFwHwABmdE0wHMAPASAFcm8qVFuq2J7H2RdpYDeATARE/+5mSbBUWyuwB0ATi0SNYI4N5Ed2GRvCUiOymR3QZgrLc/Yk/25Pn+vku0+6pEdn2hjUR+NNyc7B2R7dsAtHj7fwjA+oznf3HSTpZlYYb2Lkt03+TJXwCgI/nuLdW2O9m2bFu2LduWbefXtmt1ROYLyVKgD8CvEQaMbYE3/JUM370ELkCriWRT0ddL4Izm1QAWkdwXwKkAfmFmjxWUzKyL5OUArs3Q1/OSz4vMLDVfasnVr5A3J5+XFLdjZstJ3gzgTSRnmNlzRdvcZGatxfsneTuAD5GcYGa7SuzzQgD7ZOzfkxl0vgvgXwF8N7kOdwM4GMDXABTeZBiXcX8jBdm2bHukItuWbVeFWnVkroAbdix4q4+Z2ZaI3moz6/VkRySf/k1VzH7J5yHJ5yMRnZUZ+zo36efyjPpZmQX3IFgV+e5hAG9KdIpviJiRbk4+pwEY8IYws/vL7+aA7T1B8nUArgLw06KvbgBwP4B/A7BjKPeZA2Tbsu2Rimxbtl0VatWRedzMbs2gF5vDY/J5GYDf97Pd1op61T+FYbtq4z8ciuEA3zkFcirc8GwWdmX4pQAzW0xyLtyDajqAp8zsGZI/S1RiD6ORjGy7MmTbtY9suzJk24OkVh2ZwfB48tmb4aZ6Kvk8PPLdkRn39xiAs+DmQO8dQK/cG+ZJuLfKjkBRVL/Xt6cwtNwAYF5G3S8AWJhFMRli/fsvpWS48kwATxQPDYuSyLYrR7Zd28i2K2fU2/ZIdGQegAuU+leS3zez1LBd8krbJDPbYmYbSd4N4ByShxYuDslGAB/LuL9rAXwEwKUkzzazLm9/TAyi4AVPzdjuTXCR7BeRfEdhvpXkiwG8EcASb551KBjqudb+uBRuyPQTg2hjNCLbrhzZdm0j266cUW/bI86RSQKl3gUXjb6C5A/g5ibHwb0K+BYAFwFYlGzycbio7ztJfgd7XuPLdG7M7F6SXwHwKQDLSF4PYAPcPOhb4aLjt8F5tjsBXECyPZFtMrPb+mn3lmQY7+0A9iH5G+x5ja8T7pXEIWWo51oBgOT9cLkYHod7re9NcHkKrjCzRUO9v5GMbLtyZNu1jWy7cmTbI9CRAQAze5DksXCG/0a4COydcK/+LQLwpyLdpSRfBeC/APwHgO0AfgEXuf23jPv7D5LLAXwIwP+BG1p8Bi5dd3ui00Hy7XBZEb8JZxx/hrtx++M8AMvgUn1fBhdA92cAnzWzTH2rAZbCXYMD4V4/fBDAO8zsuqr2KqfItmsK2fYQItuuKXJl24V35YUQQgghckcuShQIIYQQQsSQIyOEEEKI3CJHRgghhBC5RY6MEEIIIXKLHBkhhBBC5JZhcWRILibZOhxt5xGSLSSN5MJq90UMDtl2Gtn2yEc2n0Y2X3tkdmRIjiP5UZJ3kNxCspvkRpK/I7kgybyYa5LjW1DtflQKyWNILiTZUu2+DBUkP0DyGpKPkOwlGc0XQMc7Sf6U5BMk20k+TfLXJE8usQ/Zdo0zmm3b2+Z1JG8luTWx8cdIfruCfcvma5yRZvMkZ5K8iOSfSa4n2UbyYZJfIzltMG1ncmRIzoFLIX05XHbCL8OVZv8GgAa4kuyXDqYjNcJH4ZIY5ZVjAHweQEuV+zGUFJJjbQKwbgC9JgA/BnAYXMXWD8NV4z0OwFKS74xtJNvODaPZtgEAJD8P4DdwCco+D5cl9qdwScsyI5vPDSPN5t8AV+dpM4CvwV2fu5LPB0nuX2nDJb1uks1wN88hAM41sxs8la+QPBHAiZV2QtQ+JAlgfJbKqUPMfABPm1lfku67v4d2D4D5ZvbnYiHJK+FSnV9G8loz6yv6TrYt8mDbIPlKuH8CnzOzL1W6Q9m8AKpm83cAONjMNhTJriR5D4Ar4Wo4VVbHycwGXOB+2RqA/yqlW7TNYgCtEfkZAG6BSyfdAZfG+b0RvRcB+DmAtQB2w9XAuB3A6zy9JgAXw/2j6oSrg3EzgGM9vTo4r28FXMrrHQAeBfB/ATQkOtbP0pLheM9NjnkbXGrrRwF8C0Bj8n1L0tbCom3mJ7IFkfYWISk+mvWcwD3kYv1fVMH5+nvf4GqErEz2ubDUuRjOBe4BbBVs98vkePaXbcu282jbcCnxNwIYk6xPAFBXwX5k87L5mrD5ov5NTPr3+0rbyDIP+tbk84oMuv1C8g0AboS7YJfBGeDbAVxF8hAz+3SiNw176lh8D8AaANMBnADgZAC/TfQaAPwewGlwUwrfBjAZwPvhComdYWZ/Tdr5NIAvwl3k7wHohSsO9kY4Y+gG8C64odbnAVxS1PUBK5WSvATOmFYm268HMBvuZvgcgK7+t85GxnNyA4AXwA0RXwpgVaK/OmmjnPNV4KNw1U6vhLtuz5To52S4oekstJtZe0bdwXIg3HXY5sll2wMfl2x7Tz+rZtskx8M5Db8D8F6SnwNwAIAOkr8G8BEz25ixOdn8wMclm9/Tz71l84WRyKw2HJLBW9oMYHuZHtZiFHnwAOrhLtY2AAcUyRsB3AlniHMT2RvhvLO3ldjHxxK913jySQCeBrC4SLYMwMoM/W4t3i6D/klJH24DMNb7jthTy6oFg/DgyzgnCxK9+YM8X4W+bQGwb5nXvb9fQv5S9q8BVDAiA+DsZH9Xy7Zl23m0bQBHJ+1ugvvlvRDAm+FiWnrh/umOy7gf2bxsviZsvmg/P0vaOLPSNrKMyEzCYDwlx/EADgJwuZn9PajNzLpIfhXATQDOAfB1uGFKADiL5O/NbEc/bb4TwCMA7ic53fvuFgDnk2w2s46kzdkkTzezJYM8lmLOSz4vMrPO4i8suUJDRNZzMhDlnK8CV5vZpjL2cSGAfTLqPllGuxVBci7cr5W1cH3zkW33j2w7TTVte2LyOQPA+83sqmT9RpI74AJCz4er/FwK2Xz/yObTDLvNk7wQwD8AuMLMBqooPiBZHJkd2HMjVcqs5PPhyHcF2SEAYGZ/Jnk1nDd6Hsn7ANwK4HozW1m03REAmjHwUOF0uOGzi+FurjtIroPzNH8L4Bcx5y+hAAAgAElEQVRmNpihwrlwnuTyQbRRkjLOyUCUc74KPFZmP+8vR384ITkLwJ/grs9ZZhY7btl2/8i20/2spm0X/hn1wTnmxfwIzpGZj2yOjGy+f2Tz6X4Oq82TfB/c20u/BfChwbSVxZF5CMAZybznsP+KBgAzO5/k1wCcBeBlcJ7hp0l+1MwKORMI4G8APj5AU88l7S0lORvAawC8PFneAeAziVe/ZTDdTZZKtuuP4LpkPCcDkfl8FVHWnCfJqXDDy1nYZcMUMZ/kXbgdLiDyFWb2t35UZdslugvZtttBdW372eRzq5nt9r5bn3xm/eUsmy/RXcjm3Q6G0eZJ/jNcnNYf4d6e6y6nbwEZ5q/+He4iXVrm3Fpr0fqJSRvfiOgW5gs/MUB7U+Aix3dhzzzlCribuOzI/WT7C5L9frJI9hTKm1P9ZtLGSSX0WhDOqR6VyD4e0b8LJWJB+jkn56P/OdXM5wsDzPdmuO5ViSPwznUrgK0Ajpdty7ZHgm3DxaT0wYuFATAn2edPMu5HNi+br6rNA/jnxJb/CC8WqdIlS0K8q5IT/AmS58QUSB5P8oIB2lgGF4D0HhYlvUkirz+ZnIhfJbKpJFP9MrNtcEY5DsDYRHw1gP3Rj0dKcr+iv/05xEKfAGBqkWyXt16Ka5PPS0kGnmvyrn5/PAWX++SV3janATjFk2U9JwWPOHYMmc/XILgQwKsyLlcPwf5SkDwYbiRmCoBXW+mhUdl2/8i201TVtuGmlAjgA57835LP32VsRzbfP7L5NENu83SZlq+EC6g+x7xYpEopObVkZu0kXw83j3UTyT/CBRNthgs+ezncEN9XB2ijl+SH4F7Xu4/kFXCv6/0j3EW+1MweT9TfDeBjJG8E8ATcq3Tzkn38zPYEL/033An8Gskz4U7MDrggtFfARfe/PNFdRfJuAPfAZdAsvNbWBZcZs8DdcK83fgnudbc+ADebWVs/x3Uvya8A+BSAZSSvh3utbRbca44nIXzlt7DtLpKLALyP5HVw3u9cAO+B87aPLlLPek7uS/r8aZL7AGgD8JSZ3VPm+aqIDI5D2dC95lk4F3MS2WeS9W2WDMOSnAjnxLQA+B8Ah5E8zGvuFit6TVW2LdvOSjVtO+GrcK8Af53koXBxHKfDBajeBuD6LPuUzcvmszLUNk/yjXC5fnbA2eu5nm+4y8xuqqjxMoaDxsG98rUEbti+Gy76/bdw7+zXe0NSrZE25sHdNDvgTvQD8BIowaVl/hHcBW5LdJfDeYdNnu4YuKHS+xLdNgCPA7gG7hd5Qe8/APwF7vXF3XABUD8HcJzX3r5wydO2wBmQIVsCpX+Ce+1wZ9KHR+CGKftNoJTIJ8D9QtoMN395B1xegEUoGoos85ycD/c6Zleyz0UVnK/5qGAocjiWwrnoZ2kt0msZQK+wzJdty7bzZttF+tPhAnrXJefgSbgcKWUPz8vmZfNVsPeFA9h71OazLoW5OCGEEEKI3JG5+rUQQgghRK0hR0YIIYQQuUWOjBBCCCFyixwZIYQQQuQWOTJCCCGEyC1yZIQQQgiRW+TICCGEECK3yJERQgghRG6RIzPEkDyF5C9IPkFyZ7I8RPLzJCdH9BeRtH6Wt3q6LQPoPlRmP99N8gGSHSQ3kryK5IzBHr8QBUhOJ/lVko+QbCe5geRt/dX46aeN/ux9WCqnC1EMyeNIfp3kMpJbk+U+khcktaWKdceSfD/JX5FsTZ6tT5K8juQRZexzwQB2n6Uy9qijZK0lUTaHwqX/vgYulXgdXLXYTwN4K8mTbE8tjWLeFZHd288+bgRwgyeL1gCJQfJjAL4B4M8APgLgQLjiY6cm/YvWIhEiKyTHwVX9fSFckbgVcMXvFsDV+LnAzL6bsbk7AFzhybqHqKtCDMT/gSsEeROcHdcDeD2A7wA4h+RrbU96/BY4O10CV1NoHYBD4Ap7viXRvb2MfV8KVyOqmEcrPI6RTTVrL4ymBXuqwr7Nky9CiRLvRbotKLNkeqSN6XA1Oe5Fup7KG5K2L672udKS/wWuXo0B+IgnnwJX1ffBjO2k6sto0bI3FwAvRaSWFYCfJLb5+iLZNADHRHSPhKsJ9deM+1yAAWrDaQkXTS3tPdYkn/vEvqRjEr3y7v2RDGOOq6Afb4IbMfofM+stCM3sZrgidO+soE0hfCYln+s8+XbsKXCXGZKNJCcMRceEyIqZ3WlmnZGvCtXGX1yku9nMHoy0sRLAQ8W6WSE5kWRjuduNNuTIDBMkxyUxAgeRfDOAr8BVML21n022J0sHyVtInjxA8xfCVVdtI/kMyS+SbMrYtROTz6WR7+4GcLj+YYgh4DYAPQC+TPJskgeSPApuyH0KXNXmrLwVzt53ktxE8n9i8WZC7EUOTD43llJMfpy+IIuux6+RVBYnuZykfmT2g2Jkho8vwjkcBR4G8AYzW+3pbQBwOYD74X6lHg3gowDuIHm2mRU7Pn1w/yBughvhmQHgbQA+Cxff8triUZZ+OCD5XBv5bi0AJjqPlWhHiH4xs8dJ/iOA/wbw26KvNgI408zuzNjUvQB+DuAJuFGeswF8CMA8kqeZmYJ+xV4l+aH3Sbgfnr/KsMm/wjkyX8q4i3YA18I96zcBmAXggwB+THK2mX2h7E6PcJjMyY1KSE6Bcxqy8i0z25Kx7bkADob79XkqgPkALjOzazNu+yCAdWY2N4P+FQDeD+CdZnZNCd0/ATgTLj6mz/vui3BO0bGxIVIxMhmu+4DkmXDO/L1w9jwzWZ8M4BVmtryC7oLkxXAjOp8xs3JGdsQIZjif50X7qAfwSwDnAHiHmV1XQv80OIfkEQCn9DNNlWW/TQD+CuBwAHPNrLWSdkYs1Q7SqeaCPcGzWZc5g9jXa5I2/imj/g8T/UMz6B6c6F6bQffmRLc58t1Xs+5Ty8hZhuM+SOy9F8BrPfn+cG/YLRlEfxvggifvqva501I7y3A/z+FCMX6EjC9FADgewFa42MOZQ3B85yf7/pdqn+taW0b11JI5r5Z7aV9/ILkRwAUABvTiE1qTz+koPc3zDNw/jekZ2i0EX86EG64vZibcjeIHaIoRzDDdB58C0GZmv/f2tYHkHQDOItloZl3lNmxm3STXIZu9i1HCcD7PkziXqwC8G8AXzOzSEvrHAbgFbvrp5WYWm8ovl9bkU3bvoWDfvctYuFwaWShMKWUJEDsELr9BFt37ks9TI9+dAuBRU9yBGDwzAdSRjP1jGQNnrxU9f0iOhQu2LDd4UoiyKXJi3gPgP81sYQn94+Be6tgJ58SsGUi/DMr5nzCqkCMzxJDcvx/5+XCxAXcXycYnD2Vf91gA/wBglRUFB5OcFtGtA/CfyerN3ncHkTzcy0D5KwAdAD6UzPcWdN8A5xANGGMjREZWAhgPZ8d/h+QsAGcA+JsVxQuQnE3ycE83sPeEL8E5Qzf3870QQ0LiiF8J58RcamafLaF/LNxIzC44J+apAXTHJc/nF3jy2HN+MtwoZxeAP5R9ICOcUR3sOxyQXAZgM9zrzU/DOS+nwwWHrQVwmpk9k+geA+D/wb2F9Dj2vLX0z3BvKL3azJYUtX0D3Jsbd8FNJ00HcC7cXOyvALzFigJ4SS4GMA/ALCsKDiN5IYCvA1gMN81VCMJ8BsCJGpERg4XkiQD+AjfychVcsO+BcFlOpwI4x8x+U6TfCuBgM2OR7HK4UcLb4e6lCXBvLb0cwD1w/yhiWbKFGBJIfh3u2bgc7pnps9rMlia6B8O9fToVwBcA+G+oAsCNlmROJzkfzrZ/ZGYLiva5Di7r+t/g3lpqgfuf8AIAF5rZN4bg0EYW1Q7SGWkL3IP6Vrg4ky4452QFgC8DmObp7g/gx3AR7Tvg0q4/DRdQdnik7ffCOR8bkrZ3wo3wXACgLqK/GC7mpSXy3QK4m7MT7mb5AYB9q33+tIycBcCxAH4BNxTeAxf4+HtEMpbCzf+bJzsH7tfn2sRO2+AcoosRybaqRctQL0XP0P6WRUW680vopp7FRfqLvH1eBucQbU7+JzwP4HcAXlPt81Gri0ZkhBBCCJFbFCMjhBBCiNwiR0YIIYQQuUWOjBBCCCFyixwZIYQQQuQWOTJCCCGEyC01U6Jg+vTp1nLwQdXuhqgSrWuexvPPP79XykXsbWTbw0zw5mXkTUxfxMhvuGGyPtl2Bnp7Qll3pHqFL+vuDlT6ur22IubAusjl8GV94YbBW76VvvSb0RpYn7ZTjon8y25oSK/XRWw7JvM70dcXqvT1epuk22ldvxHPb9tedduuGUem5eCD8Ne7lpRWFCOSE047vdpdGDZk28OL9Xr/zPyHb0w2JkioDdbXB7KhQLYdUpS3063v2BwqbQgz+9v6p9OCTZsCna516YLW1h3aQ11zY0lZ3+7QSbKutJNkPZF//llSmtSHjkXMuaqf2Jxe3y9Mds39XpAWjBsf7m9scyijZ++d7aFOu5cbtbEptXrSuz8cblMFNLUkhBBCiNwiR0YIIYQQuaVmppaEEGJoiEzZ+0W4qz6rP7qhH6NU3xDoWCQehPVpmdVHdBoz/FuLxch4MTH+NBIA9LSnY3S6u8OppdjMUp8nbBgTjiGMicjqmvz4l8j0Z7M3bTR+UqgzbmIo8895485Qx4+/aUhPLWGYpmPLRSMyQgghhMgtcmSEEEIIkVvkyAghhBAityhGRggxwogGKXgqlSYAEUOB9exOr+9uC5V2d4bbdfnbhTp9bZ5O5BXp+lhelbFpWf348BV9P2alIWJH9OOxYmTRAVDX7MWoxPLtbE2/bo6OjlCncUso8/vQEbkGHd4r2Q3ea+uxXD9VQCMyQgghhMgtcmSEEEIIkVvkyAghhBAit8iREUIIIURuUbDvMGB+XZdoXGGW4MNY8TtPFksIFUuaJMQA+LVvEmFE5q1HgiYzBTtWSFC0L65UWjaEwb7+uQuSvYmQLi9It2NXqBOr/bM7HVza1xHWQ9rd4dVDilzr5ubQ3lmftluODesxBUUbYwnhshRt7InUA4sF8vr0hDq2bZu3rx3hdrF70j8vnWGQsO1OB04Hx9+lYF8hhBBCiEEhR0YIIYQQuUWOjBBCCCFyi2JkyiA6P58ltsCPmQGAPm+uMzZlH5tr9effI32K9XM44xZE/rDu9Nx3pvn5GJEYAavzC91VFkdT8f0WD0qrQCe2q8h23rkMChmO8uR7QcwgENpbTCcWa9SYtq365jCOpWmCJ4uc/vqJzYGsbpwnixWW9G05FiPjF1oEgEavT5FYF3SF8T7BeYnZkt9Wd9iORWJZ+nb3eOvhdn2daRm9wpZ93ZHrVgU0IiOEEEKI3CJHRgghhBC5RY6MEEIIIXKLHBkhhBBC5BYF+w5AEKTWFwk0zBJD6wf2AkBPJLDLpy4MZEOdf8kiwV+RwE3z+xm0o4DgkYL1eIF9MVvzZZY1aM+3kYiNBgHpoUosZrfyJJHRxrw++dtUGCTfHVZbhleRGWMyBGiOJiLPzeBcxwJ7/UrLANg8Lt3OlMmBTlNwrSN9Gj8ulDV7wb6xZG9+IG3sf0IkAJhj021brGp07OWO7gxB0b1pWSywt2dLWNm6e1falnsiVcK7u9KyujHpk9vXm+He2wtoREYIIYQQuUWOjBBCCCFyixwZIYQQQuQWOTJCCCGEyC0K9h2IIKtiLLApln3XizaLBfv5AbnRirlh8C29QLJo1syozO97JLDXy0gaC37cunVrIHt45Upv9+H+582bF+5PDBrrCivWwpNZLGuvH6Aas+1o8LdXIThmt/VeZtNozG5Gux0y0v2OVfveuvm5QLby4YdT670dYdDkGaee5O1p+Cpt1yL+cyIIkI6c6yAe18+GDABNTaGsefyA7QCAxTLr+vuP6XjBxWY7Qx2vGnS0B7Gg3THp/UVfq4jcS+bLYgHI3rPdz8YLABueXhfIli59ILXe3RXefy89/WWp9Tov23Fvb23YtkZkhBBCCJFb5MgIIYQQIrfIkRFCCCFEblGMzED4c/2x5EexCqn+DOiYWGyNpxOrENwYVmgNdOoi1VcjsnXr0nOkvZFjefLJ1an1vojOunXrS/ZJDA/Wm6E6LhDaaWd7qNPuzf/H2hlTOo7Aj1kAENiyH9fliFTNzlLFui+UrVv3bFolkqRrdWDbYTvr12+I9NPvUnie2DzRE3j3cixmYgRRMpHmmEjSxAYv/qUpkqAuFtuVIY4qiL+JJJ+zSBwfur34p/Yw/sw6vISIkec/d48N227OkEgvgv/c7msLY7SeXJmO4+p6bluos/yxQNbWnj6/sZjIpqb0fepf6ui/vyowsu8wIYQQQoxo5MgIIYQQIrfIkRFCCCFEbpEjI4QQQojcomDfAQiC2KJBixkYEyZ2MqZPfSwgsi0S2OXLnn7mmUBnTWtrIOvyEin1ZaosnC2Sa9LEdLDjpMmTMm0nBiZI2pYx+DFIkrg7Euy7a7vXdiT4sSkSbN7k6U2M2FGksrrPUNr2bi9JWSxoMbTsWL9De5/o2fbkKVMCnZJB+aO8qnwsGNi8AGBGbC2a7DNLpXN//xGZ+YG9ANCevk+sI7xvetvStsYx4XO7s29LIGtr25VafzZi262tawLZ7vZ0P21nmKTPdqbb7tke9jv2tJ8yJf2cnjw5UkncC/b1GyoZ6L2X0IiMEEIIIXKLHBkhhBBC5BY5MkIIIYTILYqRGQb8OfpYYrlWb65/85ZwXjUa6xIkdxq6Oco5s+eU1Nlv//0D2UEvfGFqvW6EJwDba/gxMbEYmZ5IkrwuL2lXt1/oDmHRyFiMTGz+27u0PX47ANY88Xhqfcu27aHOcNp2pN9zZ88uuZlsey/in8dorEWs2K5np9HEct52PZH7pjO0W/NitHp7wnuidX06tmXrrjBmpfX5zYGsu8mLCcpo20G8V+SUzPZsu68zPN5995kWyA48YGZqvT4WAxrEjaU7UN9QG/dDbfRCCCGEEKIC5MgIIYQQIrfIkRFCCCFEbpEjI4QQQojcomDfMogl2upsC4O9Vv5tRWp91RNPBjrxisClmTFjRmp9ypR9Ap1pU6cGspaWltR6NLBL1BZ+QrBIYK/tDiv0BtWuuyIBwdGEiGk6vCReALDywQdS649ui+x/n3TQLGNVtCPItkcxsaDdmG23eYHjuyOB7J3p7awzbKdz8/OBbMWd96XWn9j4bKCDKV7SuOZIMsRxYSXvGV4g+RQv0SIATPPbBnDwAQek1us7OwMdtHn3e0PkfmuKVCAP7stYtXnvuniV3dmU7d4ebjQiI4QQQojcIkdGCCGEELlFjowQQgghcoscGSGEEELkFgX7lkFPkHkUuOuOvwSy9c+kq5jWNY8PdJqa0rK5h84NdKZNDbMxzpw5M5CJEYoX7Gs9of1FAyL9ate9ke28TKo9kazBS5csCWTrN25KrdcdOCvQaWpMBwAe+qKjAh3Z9ijHT1jbF8m+G6va7gegd4W2bV626Z5Ixeglf7w9kD39cDojdfOcFwQ6TTPTmZ7nHnt8oDNt4oRANnPG9LQg0m9E/r/4xxLNgFzvjUdMCgOJOW1GIMPEdPXraAZkP5Oyn+m3sSncpgpoREYIIYQQuUWOjBBCCCFyixwZIYQQQuQWxcgMwJqnn06tr1i+PNDZsSOs7Itx6bnHk089LVDxK5YKEeAlrGLD2EDFxkeSWPmVhRvCeew1G9LxACseXBbo7GyLxN9MTCftOun0MwKVOUefkFpnYyRpmBg1WKSyum1dnxZs2RToYOe2cLsOL26mO0z2+PTq1an15XfeFehsfXZd2LaX/O2EQ48MdOaedFxaMDa8J8HwnrTnn0sLYgkAY0kq/SSYuyNxNN3p2BZGqnZH91fnJY4cF7lP/WeHH8tUXxsuhEZkhBBCCJFb5MgIIYQQIrfIkRFCCCFEbpEjI4QQQojcUhuROlXATwB29933BDrPPvNMar03EjA1fkKYfOis1742td7QGKk8KkQp6j27aQgTVtFPUAWgx9Kyu++7P9B59sknUuu9kUfBuP0PDGRnnXVWar1p6r6Bjk+sajxjib1ELvGvb3BtO8OEdFj7VLqNbZtDnUiCth6v+vM9kaSNz7S2ptZ7d4T7H9cUBum+8i2vSa03RYJfdz+1IbVeNzas/lwXqQhd1+wF7sfs3w/SB8LEgZFK9n3d6eDe+khyS1iGYN/xkwIVjksH91u3V33bb6NKaERGCCGEELlFjowQQgghcoscGSGEEELkllEbI/PQQw+l1tesWdOP5h4aG8O5z8MPPyyQPfPss6n1+sjc56xZYbE9IYrx59EtlljOT5gF4KG/pW376ee2hNt5SRsb/cJzAI44Kiz2+OzzW1Pr9W1hgq5Zc/17IpLoC4qRGRH09YXFHcemC+JaeyRGZqeXSLRzd6jTEP57euiJdGzXmvXrAx2fpuYwHuaIQ8OEpGvXrU2t148J4z9aZh2SWu/rDO+/WPyL1Xv3cuSWiMnQm45t6esO41+sKy2rHxfGZO7YuCGQTZrgxcREihsH925fJNleDaARGSGEEELkFjkyQgghhMgtcmSEEEIIkVvkyAghhBAit4zaYN/ODi+xTzQgMU1XVxjYeP/9YbKxLNy1NF2RtbEhDNA67vjjA9nsQw4JZGJ0QEZ+d0QqW3f2ZggS9pJmdUWSPd6/4uFwO7/67ZgwAP6u+9KVtBvHTQh0jj/hhEAm284hfb2wDi+Y14917dgVbue9AMHxkUBTPxgVwG4/AePUGeF2XiBx144dgcoDK1cEsp6tbekuNofP5L+uTt8TTePDeytm24fMnZta7+sIg5v72iKyjnQwcXdH+D+oqyt9706MvFxy8y9vCGQnv/71qfU5J5wY6JiflNAP7I68bFANNCIjhBBCiNwiR0YIIYQQuUWOjBBCCCFyixwZIYQQQuSWURvsu8/Uqan1sWvX9qO5d4gFEi+LBBIrIFKUYur06an1dRmyn1ok2Bc9kWyrfnBffeQR4gUbdveG2UBl2yOEvl5gZzpztPk20pEOoo0yPgwIx7SwsvrUw49Ora/d9Fy43ZZ0AK6NidhoJLF0j6XtdszkSADyxImp1a5IVem7V64KZIe85CXp3UcyGceyBHfsTOt1dISZfTt2p++vp55cHej0btkeyO794x9S63NaDg504Ffb3t2RXlewrxBCCCHE4JAjI4QQQojcIkdGCCGEELll1MbIHHbooQOu722eWB3Oa95zzz2B7MEHH0ytH3PMMcPWJ5FPsti2+aV2YzEyFql061e/rQ8T4tGLm5Ftj2QIjPESx/lJEiPJPtHkVaSOJXuMxF8cetDM9Pq5bw236/RicnZsC1Rs69ZAhu2eXnMkkeTYdL9XP9UaqNxz55JAdu2116XWzzz55EBnxsTJgazZu03r6sLgHl922/Iw/mzBB98fyB7b9GxqPUh+B4Be4kz4OtGS3XsfjcgIIYQQIrfIkRFCCCFEbpEjI4QQQojcIkdGCCGEELll1Ab71hr1kYqlMZ5+5pnU+pEvelGg09gQBmCK/GOR5FuxzF6kV/06tl2fH6QXCdpjfSgbk5axLqLjIdsewZBBBXZ6wb/WGFZohy+LBY12RxIytqdt+5Zbbw1UDjviiNT6QdOnBTqM7M78W6k+YtuebHXrU6FOLHDZuwcWL70rUDmspSWQveSw9D3A+vB+X7LkL6n14084PtCJJbw84pDZaYGf/A4IXgIwP/i3RtCIjBBCCCFyixwZIYQQQuQWOTJCCCGEyC1yZIQQQgiRW0ZtsO+mTZtS621tYYXWWbNm7a3uYNv2sDqpGN3Ybs8mYwGRkWDbjZvT1Yijtt3i2TYj5YAjQbp+IHEWZNsjGDLM5OtXRI9VSPcDabu6Qp3OzkC0ae3a9Pq6ZwOdbTt2pNY3zDwg0DluTlhpfcy4SLVrH++eeH5baNtsCjMZj5+6T2p93yn7BDqr17QGsud33p1aH9cbPgO669PZto88KgySj1XW9sN2G8aH5zvIruxX7Y5lBK8CGpERQgghRG6RIyOEEEKI3CJHRgghhBC5ZdTGyPjVd3fs3BHoLF+xIrU+K5Kw6Oijj65o/7t3p+caN27cGNEK50MPeuELU+tKEJYDzIJEUn6F6J/88KpwO7+KbyyxnV95GAhjEiJxNCu8StMts+cEOscce1zYdgZk26MIMrQvfz2WIM6PreiOxMhE4i9uueEXacHYsEJ1lxfX8fiW5wKdLasfC2RzDz4otT47UjX+D3+6PS0YPyHQidWDPvjIdNzKMS85KtBZ+/SaQLZqWbqS9ZpVqwKdt7zhdWlBZ5i0rrstcn7b07L68WHiwroJ3vG17UqvK0ZGCCGEEGJwyJERQgghRG6RIyOEEEKI3CJHRgghhBC5ZdQG+x5x5JGp9W3btgU6jz76aGp9VSTQKrbdgV7QYnckkG3VqkdS6+3t7YFOfSSR1OTJkwOZqHH6+oCOdJCc+UnE/MBeALZza1oQSUbHaGVhLwCyIdRp84L2Vj3ySKCzfcfOQCbbFj70Eyd664899niwzaGTx6YFETuy3ZHq1zvTNnn47DCx3aNr0lXULRJsvNkPWgWweeOG1PrdD64IdNDkVfoePzFQqYvck5NnecH0+0wPdGZODpPkzZw9N7V+zTcvC7s0aVxqvasjfHGlvaM3kPX0pAN1G7eH92njlLSsb1d63XoV7CuEEEIIMSjkyAghhBAit8iREUIIIURuGbUxMnNmz06t90YS+0ycmJ7/3LhhQ6Cz6blNgezZtX4hs/IL7QHAKaecEshaDj64orZENTFYn5cQr8+ziVhCsF6vYFtfmGrLInZLL/leLGnVCccfn1rfsHlzoPPchnWBLA+2bbHimn1+jECk3zFRJJmgKIJ1QQwWvRitvz70cLDZxjHp6/Hi2WGB3gfv/2sgO/7MV6TW5x5+eKAzYf8n0/tavz7Q2RSR7e7sCGQhnpEEdgWccnSY7K5lsleQcvuWQKVWKz0AABSASURBVCdaFNbr0zvOOSdQ6X0uHadpPWGf6iNDFvWNaWH9+LGhkpfwr647/UxifWX3/1CjERkhhBBC5BY5MkIIIYTILXJkhBBCCJFb5MgIIYQQIreM2mBfn3o/qROAw7zqp/46AHR5wU8AsPSuu1Lrbe1hsrMWr5J2S0sY7Da2KZLsTOQPM8APwPXtrS+0I/R6QXt+GwDQF8qs20skFgn2nXvQzNT6oXNmBzrdkdjDu+65L7Xe1h0GFs6a1ZJaH27bNj8oOnae/HMZSS6IulBm9enEhaxXRe4UJNAQCRIt4oTTXhbIHr/jltT6727/c6Azbb/9A9nhp5+RFnSFSfMO8ypNR5/bkSSNS5fckVp/9vkwIPfYY49Nrbe88MBAZ2xXZyCzDV5wcSxGtj4SWL47/RKARZJU9uxMBwT3RZLUjWkI/781Nqb3Vz9pfKDDyVPS+/eDm2N9rgIakRFCCCFEbpEjI4QQQojcIkdGCCGEELlFjowQQgghcouCfQdJY0MY/Ddv3rwq9ETUNH7AbSQAN8APpItl/qxk30CQNdh6w6C9Br9CN4B5p780LWgMAz0ZqWw9VATBhkAY3BsL9u3xAoIjFZER6ze9c1cbsY21A1nyeh/2ojDT7QsnpbP/3vCznwY6r37LP4SNNTSm1zvCKtZA+HKFT2NzcyCbd+aZaUHsuPx7YncksDdSWdu2pSvZRzNGj4nIvCD1vs5IcL9XxbouksZ3bEPYdv1E7xxMiVSf32daapX+s2QY7/Vy0IiMEEIIIXKLHBkhhBBC5BY5MkIIIYTILbUxwSXESMeLdwkqNI8JE8RxvH97RrJoxRJS+fPvkWSPHDfR04m0w4jMn3+3MP7GPzbGks9lIFrFOpoU0OtDbDu/D5FzEi0RrOrXg4YNoW2P2z+dSO6o084IdOrGTwob8xMbRmK0AnuIXdfGxlDW49lWrBr2zu2pVYskO0Uk2Z75iSPrQxtlJCGjb7d148J+N3jHF6tIzUj8DcenE+Bx+n7h/vfxZH5s2ZjacCE0IiOEEEKI3CJHRgghhBC5RY6MEEIIIXKLHBkhhBBC5JbaiNQRQ0o0SNKj0gBMUQFkGCTnJXY77/wF4XZ+oGks+VRM5m8XCcjNlJAvZiO+LGZrPemKxBYLUo7anyeLVPaO9jtIkpcl2DdDkDSGN7nfqKY5HWx+9CkvDXX8JIYArLvLE2Sw40hix9BmIvuLBPLajh1pwc6wGrV1dUVkfpByuHv6gcxAcH+zOQxurh/vB+BGjjeSuJUTvYD/6WG1cU7wql8Hz6TaqAavERkhhBBC5BY5MkIIIYTILXJkhBBCCJFbNAE8AlH8Sy3ixW14CdmixRC9eBBmKZgIVJbELUs8DBDG+sQSy/mJ9GIFGiMy324tcrhA5Hj9OJ1Y3ESgE4mjicjMa4uxYxHl0+0VW/RjX4B4HIsvy1IgNNp2ZLtuTxaz/yYvuV8kjif69PXixqLP6Cy25fcR4bOD9ZHjjSSuMy8BH9t3BDroTvcbHV5MUJZYu72A7kohhBBC5BY5MkIIIYTILXJkhBBCCJFb5MgIIYQQIrcwS/K0vQHJ5wCsqXY/RNU42MxmVLsTw4Fse9Qj2xYjlZqw7ZpxZIQQQgghykVTS0IIIYTILXJkhBBCCJFb5MgIIYQQIrfIkekHki0kjeTCgWTDtS8hhgvZthipyLZHJzXlyJCcnxhG8bKL5P0kP0L6uc/zQ2L0C0keU+2+7E1ItkauaWGZXu3+7S1k2yMP2bZDtj3yyJtt12qtpesA/A6ubMUBABYA+CaAFwH4l+p1C2sANCNa8KUkLQA+D6AVwIND2G4eeATAJRH5zohspCPbHlnItvcg2x5Z5Ma2a9WRWWZmPymskPwugFUA3kfys2a2MbYRyYlmNmwn2dy76p0lFWuk3RpiY/H1HOXItkcWsu09yLZHFrmx7ZqaWuoPM9sBYCmcp38I8Pehr8UkjyX5B5LbAawobENyLskfk1xPsivR/xrJ8X77JE8neSfJDpIbSX4bwISIXr9zoiTPTfqzjWQ7yUdJfotkI8kFAG5PVH9YNES3eKB2SY4h+SmSK0l2ktxM8kaSR/XXL5KvJ3lfor8+OeaqO6zJsUyqdj9qDdm2bHukItuWbe8tqn6iskCSAOYkq88XfXUQgNsA/BzAL5EYMcnjE/k2AN8HsBbA0QD+HcBLSc4zs+5E92QAt8INl30l2ebtAK4uo3+XALgYwEoAlwNYD2A2gHMBfA7AXwBcmuhcAeCOZNPoL5QirgHwNgC3APgugP0BfBDAUpIvM7MHPP2zAVwA4HsAfgDgHACfALA12X+p45gMoKGUXkK7mbVn1D0ZQDuAhuTB9SsAF5nZuozbj1hk27LtkYpsW7a91zCzmlkAzAdgcEY0HcAMAC8BcGUiX1qk25rI3hdpZznc/N5ET/7mZJsFRbK7AHQBOLRI1gjg3kR3YZG8JSI7KZHdBmCstz9iT/bk+f6+S7T7qkR2faGNRH403JzsHZHt2wC0ePt/CMD6jOd/cdJOlmVhxjZ/C+AzcA+Ht8PdrD0AngVwQLVtTrYt25Zty7Zl2/m27VodkflCshToA/BrhAFjWwD8sFiQDN+9BC5Aq4lkU9HXS+CM5tUAFpHcF8CpAH5hZo8VlMysi+TlAK7N0Nfzks+LzCw1X2qJRVTIm5PPS4rbMbPlJG8G8CaSM8zsuaJtbjKz1uL9k7wdwIdITjCzXSX2eSGAfTL278ksSmb2Ok/0U5J/gfvV8gUA78+4v5GCbFu2PVKRbcu2q0KtOjJXwA07FrzVx8xsS0RvtZn1erIjkk//pipmv+TzkOTzkYjOyox9nZv0c3lG/azMgnsQrIp89zCANyU6xTdEzEg3J5/TAAx4Q5jZ/eV3s3zM7NpkWNe/WUYDsm3Z9khFti3brgq16sg8bma3ZtCLzfUx+bwMwO/72W5rRb3qn8KwXbXxHw7FcIDvnAI5FW54Ngu7MvxSGIhWAC8dxPZ5RbZdGbLt2ke2XRmy7UFSq47MYHg8+ezNcFM9lXweHvnuyIz7ewzAWXBzoPcOoFfuDfMk3FtlR6Aoqt/r21MYWm4AMC+j7hcALBzEvuagdNCcSCPbrhzZdm0j266cUW/bI9GReQAuUOpfSX7fzFLDdskrbZPMbIuZbSR5N4BzSB5amG8l2QjgYxn3dy2AjwC4lOTZZtbl7Y/JXGnBC56asd2b4CLZLyL5jsJ8K8kXA3gjgCXePOtQMKRzrSSnxoaWSX4QwIFwEf0iO7LtypFt1zay7coZ9bY94hyZJFDqXXDR6CtI/gBubnIcnDf5FgAXAViUbPJxuKjvO0l+B3te48t0bszsXpJfAfApAMtIXg9gA9w86FvhouO3wc3d7gRwAcn2RLbJzG7rp91bSP4s6cs+JH+DPa/xdcK9kjikDMNc67tJvhduqLgV7pzOh5snXg0X2CcyItuuHNl2bSPbrhzZNmr29etPZNBtBbB4gO8PhntlrBXuNb3NAO4H8GUAL/R0z4B7na8TbtjsOwBejAyv8RV9908A7oQz+ja4QLRvAmgs0jkbwLJkP1bof3/twhnQp+ACx3bDRfvfBOAoT2+gfi1MvmupwvV8KdxbC08D6EiOexWA/wIwpdr2JtuWbcu2Zduy7fzbduFdeSGEEEKI3JGLEgVCCCGEEDHkyAghhBAit8iREUIIIURukSMjhBBCiNwiR0YIIYQQuWVYHBmSi0m2DkfbeYRkC0kjubDafRGDQ7adRrY9cpBtp5Ft54fMjgzJcSQ/SvIOkltIdpPcSPJ3JBckmRdzTXJ8C6rdj0oheQzJhSRbqt2XoYLkB0heQ/IRkr0kB8wXQPJUkr8m+SzJDpKrSV5J8pABtpFt1ziybYDkYSRvIrmVZFtir2eW2Ea2XePItoNt/y1xII3k9CzbZHJkSM6BSyF9OVxynC/DlWb/BoAGuJLsl2btaA3zUQALqt2JQXAMXNbFlir3Yyi5CC619yYA6wZSJPlaAEvgarB8G8CH4RI7vQPAX0nOjGwj284Ho922Z8MlfzsVwFcBfBLABAB/IPnKfraRbeeDUW3bxZA8AC7xXlmFLUt64ySbAfwGrnT6uWZ2g6fyFZInAjixnB2LfEGSAMbb4CqnVsJ8AE+bWV+S7vvAAXQ/BldJ9jQze74gJPkwgCsB/ANc1s6CXLYt8mLbXwYwBcDxZvYgAJC8Gi6N/3dIHm5F2U1l2wLIjW0X8x24MggPA3hn5r1lSFf8YbhUyf9VRorjxQBaI/IzANwCYDtc6uNlAN4b0XsRgJ8DWAuX4nkDgNsBvM7TawJwcXLQnXB1MG4GcKynVwfnta+AS0W9A8CjAP4vgIZEx/pZSqaIBnBucszb4ErUPwrgW0jSXCOSihp70noviLS3yF2a7OcEe1Ja+8uiCs7X3/sGVyNkZbLPIJX23lzgHsw2wPdLk2tb78nPTo7n/bJt2XbebBvA+OSY/hT57rPJ8Zwk25Zt5822Pb03w/0QPalwLgFMz7KPLPOjb00+r8ig2y8k3wDgRrgLeRmcYb4dwFUkDzGzTyd60+AKhwGu5sYaANMBnADgZAC/TfQa4IpanQbgx3BTCZMBvB+ukNgZZvbXpJ1PA/gi3MX/HtzJmgU39NUEoBvAu+CGYJ8HcElR1wesVEryEjgjW5lsvx7AbLib5HNw9UIGRcZzcgOAF8ANHV8KVxsDcN5tueerwEcBTIMbzdgA4JkS/ZwMN2SdhXYza8+om5U/ADgFwI9Ifg3uWr4Yzt5WAfippy/bHvi4ZNt7+llN234J3LVcGvnu7uTzRAD3Fsll2wMfl2x7Tz+r/dwGyUlwx/Z9cwU9LyirgQxe0mYA28v0wBajyLMHUA93EbcBOKBI3ghXsKsXwNxE9kY4T+xtJfbxsUTvNZ58Elyxq8VFsmUAVmbodysGKGgW0T8p6cNtAMZ63xH4ey2rFgzCsy/jnCxI9OYP8nwV+rYFwL5lXvf+fiH5S9m/ElB6RKYJwP9iT3G3wvJbAJNk27LtPNo23D9XA/Bvke+OTL67VLYt286bbRfpfBfOmZxcfC4xhCMyk+Aqiw6G4wEcBOByM/t74I+ZdZH8Klxl0HMAfB1u+BIAziL5ezPb0U+b74SrVHp/JLL5FgDnk2w2s46kzdkkTzezJYM8lmLOSz4vMrPO4i8suRpDRNZzMhDlnK8CV5vZpjL2cSGAfTLqPllGu1nphRvCvRXuV+QWuEquHwbwU5LnmFl3kb5su39k22mqadvjks/dke86PZ0Csu3+kW2nqepzm+RLAXwAwHlmtr2UfowsjswOABMrabyIWcnnw5HvCrJDAMDM/pwEsS0AcB7J++D+MV1vZiuLtjsCQDMGHkKcDjesdjHcTXcHyXVwHuhvAfzCzAYzhDgXzmtcPog2SlLGORmIcs5XgcfK7Of95egPA4vghmBfVHRj30jyCTiP/3wAVxXpy7b7R7ad7mc1bbswlN8U+W6sp1NAtt0/su10P6tm2yQb4aY/bzWz6yptJ4sj8xCAM5L50OH4FR1gZucnMQ5nAXgZnMf4aZIfNbNvJ2oE8DcAHx+gqeeS9pYmry++BsDLk+UdAD6TePtbBtPdZKlku/4IrkvGczIQmc9XEWXNhZKcCjfsnIVdNoSR9CQPgvul9W3v1wnggu2+C2Ae0o6MbLtEdyHbdjuoom1jz+urQfqAItlaTy7bLtFdyLbdDqpr2x+ES5dxYZIuoEDBCZ9FclJJG84wv/XviMzBlthmMdJzrScmbXwjoluYR/zEAO1NgYso34U985cr4ObU6rL2y2vzgmS/nyySPYXy5lq/mbRxUgm9FoRzrUclso9H9O9C6TnF2Dk5H/3PtWY+XxhgHjjDda9WHMGpSbv/G/luv+S762Tbsu0c2vYElH5r6WTZtmw7h7b9zQz73FVqH1kS4l2VnPhPkDwnpkDy+BJRxsvgApPeQ3L/ou0a4BI7GYBfJbKpJFP9MrNtcMY6DnuGUq8GsD/68VRJ7lf0dyw74LLkc2qRbJe3Xoprk89LkyEyvw8cYNunAPQASCWzInka3Js3xbKs56TgKceOIfP5GgQXAnhVxuXqIdhfMY/Cxci8ieQU77sFyed9nly23T+y7TRVs21zv4BvBjCf5NEFOckJAN4H4HGk31gCZNsDIdtOU83n9g/h8nv5y+Lk+39GhnwyJaeWzKyd5Ovh5iZvIvlHuCCjzQBmwA33vQYu22R/bfSS/BBcAOZ9JK+Ae43vH+Eu/qVm9nii/m4AHyN5I4An4F6xm5fs42e2Z9rgv+FO7Nfo0nTfBjcvfBCAV8D9gnl5oruK5N0A7oEbpi287taF9Cu5dwN4L8kvwb0G1wfgZjNr6+e47iX5FQCfArCM5PVwr7vNgnv98SS4iP/YtrtILgLwPpLXwV24uQDeA+eFH12knvWc3Jf0+dMk9wHQBuApM7unzPNVETYMc610r38WzsWcRPaZZH2bJcOzZraF5DfhbsoHSF6JPcG+58G9zlg8rSTblm1nppq2nXAR3LH8keTlcMf3frippddZ8vO2qL+ybdl2Jqr83F6OSKxSYruAu47P+98HlDE8NA7uVbAlALbCXZSNcDfKu1CUhAz9J1aaB3cz7YC7AA/AS6wEl675R3AXvi3RXQ73D6rJ0x0DN4R6X6LbBvfr5BoAry7S+w8Af4FLl7wbLjDq5wCO89rbF8Av4f4B9sH94mjJcG7+Ce51xJ1JHx6BGzLrN7FSIp8A9891M9y85h1wwaqLUDQUV+Y5OR8uN0JXss9FFZyv+ahgiHI4Fux5DS+2tHq6hHu43wP3K6cb7tXM7wCYIduWbefVthP9I+BGQAoJ3JYAeGWJfci2Zds1b9sDbJ/p9evCHJ0QQgghRO7IXP1aCCGEEKLWkCMjhBBCiNwiR0YIIYQQuUWOjBBCCCFyixwZIYQQQuQWOTJCCCGEyC1yZIQQQgiRW+TICCGEECK3yJERQgghRG6RIyOEEEKI3PL/AQ4Y0aV4+WMOAAAAAElFTkSuQmCC\n"},"metadata":{}}]}],"metadata":{"accelerator":"GPU","colab":{"collapsed_sections":["tnAt_S0yIOwB","bRstJrMldjZV","dOBdw3VlDdfR","tF9WS8JC7xZv","yXCmg_x5OdKV","x0vIsj9fuZwA","0RaOHfoHhRgY","YTw_xgswvcGh","Z4ZKCp_0nw46","sfZyNhW3a9nG","cViJtIKdkKEW","p2ZjT9lwkNkh","L__CXwGL1vkt","vUPCLvjIGMSB","sxiAnkxODitR","luLXrDuuspf6","uUwSw1-LkLVR","f7T45nyDUReS","FRvwz4NgvYu7","a70b3IFg1RHb","0Gjy0XR1KVe5","gq7AqLa_Ve28","eyK1Kbn3gVBe","nKkyGQmew0GW","hKJZIvKo987S","ReezkOBfj1-1","QlzmJgYjXx-J","otAKfNoLYkSx"],"machine_shape":"hm","provenance":[{"file_id":"https://github.com/norse/notebooks/blob/master/mnist_classifiers.ipynb","timestamp":1637582196016}],"toc_visible":true},"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":"3.8.12"},"widgets":{"application/vnd.jupyter.widget-state+json":{"2cc4b7e58ff0464d8f39c9f81851d508":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HBoxModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HBoxView","box_style":"","children":["IPY_MODEL_449be4d4154d4b9bb7e22e7de8220e80","IPY_MODEL_1a43f874e5aa493aa2a8930ce9064caf","IPY_MODEL_4d8ed3cf1c304ae1b29c67d4bbc18a0f"],"layout":"IPY_MODEL_832d1e6476f749078c9cc95123aba684"}},"449be4d4154d4b9bb7e22e7de8220e80":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_ec5f7def061e4411b15334e960ea0b5d","placeholder":"​","style":"IPY_MODEL_9e6308fa6c7542bdab60efe8155ae443","value":"100%"}},"1a43f874e5aa493aa2a8930ce9064caf":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"FloatProgressModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"ProgressView","bar_style":"success","description":"","description_tooltip":null,"layout":"IPY_MODEL_0f5593c8c2214fdba680ab8739696f1c","max":9912422,"min":0,"orientation":"horizontal","style":"IPY_MODEL_52ed85d9daf445098e3550e28dc52e9a","value":9912422}},"4d8ed3cf1c304ae1b29c67d4bbc18a0f":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_6d7b6701dec343ec82ee0d0a7c46964c","placeholder":"​","style":"IPY_MODEL_13b00de3c87943c6b3e2abdbe2e1e681","value":" 9912422/9912422 [00:00&lt;00:00, 29187488.30it/s]"}},"832d1e6476f749078c9cc95123aba684":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"ec5f7def061e4411b15334e960ea0b5d":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"9e6308fa6c7542bdab60efe8155ae443":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"0f5593c8c2214fdba680ab8739696f1c":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"52ed85d9daf445098e3550e28dc52e9a":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"ProgressStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","bar_color":null,"description_width":""}},"6d7b6701dec343ec82ee0d0a7c46964c":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"13b00de3c87943c6b3e2abdbe2e1e681":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"ec72ba0358914a31b9f697448d9854da":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HBoxModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HBoxView","box_style":"","children":["IPY_MODEL_b3f95b6a61ea4187a6bc3e83ac446ed5","IPY_MODEL_523cd76739e641ac891646c7530d284a","IPY_MODEL_30f69af25d224aa18508bafb665ea55c"],"layout":"IPY_MODEL_35f2223cf93c4c108ccd054c422abbe7"}},"b3f95b6a61ea4187a6bc3e83ac446ed5":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_1bb4959231b44f5c9ad8aa9ab8e1e50e","placeholder":"​","style":"IPY_MODEL_a051eb15d39940568639ab1ff830618f","value":"100%"}},"523cd76739e641ac891646c7530d284a":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"FloatProgressModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"ProgressView","bar_style":"success","description":"","description_tooltip":null,"layout":"IPY_MODEL_3acff4df63894f8fae92e250f26eb6af","max":28881,"min":0,"orientation":"horizontal","style":"IPY_MODEL_5d098b38c2eb48a595ace4d61156cb0b","value":28881}},"30f69af25d224aa18508bafb665ea55c":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_8efba5efea5a41098a64cb0638522718","placeholder":"​","style":"IPY_MODEL_1b9659c792fb4c77911bdeacfc2dc516","value":" 28881/28881 [00:00&lt;00:00, 1270685.23it/s]"}},"35f2223cf93c4c108ccd054c422abbe7":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"1bb4959231b44f5c9ad8aa9ab8e1e50e":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"a051eb15d39940568639ab1ff830618f":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"3acff4df63894f8fae92e250f26eb6af":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"5d098b38c2eb48a595ace4d61156cb0b":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"ProgressStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","bar_color":null,"description_width":""}},"8efba5efea5a41098a64cb0638522718":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"1b9659c792fb4c77911bdeacfc2dc516":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"cbb560dd6e754ff181f36e55614e8983":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HBoxModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HBoxView","box_style":"","children":["IPY_MODEL_483bd8b769cb4360954eb615ebcb3852","IPY_MODEL_9ca65c8f48bd43748a406b8156467ee0","IPY_MODEL_cb084fe409804ae0b62e7e6638a0b343"],"layout":"IPY_MODEL_2a59052ee73546f9bb547e4462be7733"}},"483bd8b769cb4360954eb615ebcb3852":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_4d6e86bf76b64e2f9e9eb479947568c6","placeholder":"​","style":"IPY_MODEL_332cc880de8e4630bbb08f82c7cd992e","value":"100%"}},"9ca65c8f48bd43748a406b8156467ee0":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"FloatProgressModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"ProgressView","bar_style":"success","description":"","description_tooltip":null,"layout":"IPY_MODEL_7de526d3f46844f294aed96443dc042d","max":1648877,"min":0,"orientation":"horizontal","style":"IPY_MODEL_eb610b88cc904468a71e3602412215d8","value":1648877}},"cb084fe409804ae0b62e7e6638a0b343":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_4111f0c5c7f84a6ca9a3df82f1214813","placeholder":"​","style":"IPY_MODEL_1929ad137d334cd6889b9f91ea97435e","value":" 1648877/1648877 [00:00&lt;00:00, 3510475.71it/s]"}},"2a59052ee73546f9bb547e4462be7733":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"4d6e86bf76b64e2f9e9eb479947568c6":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"332cc880de8e4630bbb08f82c7cd992e":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"7de526d3f46844f294aed96443dc042d":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"eb610b88cc904468a71e3602412215d8":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"ProgressStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","bar_color":null,"description_width":""}},"4111f0c5c7f84a6ca9a3df82f1214813":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"1929ad137d334cd6889b9f91ea97435e":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"c4f3070e663a4e6694466403518b8c17":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HBoxModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HBoxView","box_style":"","children":["IPY_MODEL_9e3a1190de6543b0a808e5ef3026e1a3","IPY_MODEL_704fe4fb45804d60adaa974faf78467b","IPY_MODEL_cb4a641d46474738ac3e0fd965e426c5"],"layout":"IPY_MODEL_ba8561b43aaf4e66a12bd4d1987f1039"}},"9e3a1190de6543b0a808e5ef3026e1a3":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_f0c0ea575e5c4345ba0245dae4d3616c","placeholder":"​","style":"IPY_MODEL_a40d0197c3e64addb2d059848819c07d","value":"100%"}},"704fe4fb45804d60adaa974faf78467b":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"FloatProgressModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"ProgressView","bar_style":"success","description":"","description_tooltip":null,"layout":"IPY_MODEL_63474d61b6294bc595f04ee32f73787f","max":4542,"min":0,"orientation":"horizontal","style":"IPY_MODEL_4919ac3ddeea456da720acb6836db00c","value":4542}},"cb4a641d46474738ac3e0fd965e426c5":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_4102125a74fe4d98a0dae4207554b558","placeholder":"​","style":"IPY_MODEL_5427913e518846bb82eb0b7923d687dd","value":" 4542/4542 [00:00&lt;00:00, 173082.78it/s]"}},"ba8561b43aaf4e66a12bd4d1987f1039":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"f0c0ea575e5c4345ba0245dae4d3616c":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"a40d0197c3e64addb2d059848819c07d":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"63474d61b6294bc595f04ee32f73787f":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"4919ac3ddeea456da720acb6836db00c":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"ProgressStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","bar_color":null,"description_width":""}},"4102125a74fe4d98a0dae4207554b558":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"5427913e518846bb82eb0b7923d687dd":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"d2e15f0fd13340fa8423867719e4ac4e":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HBoxModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HBoxView","box_style":"","children":["IPY_MODEL_8287503040974fcd8b17ff68962f3618","IPY_MODEL_73fe7b1bc0c34976941af9c6416f084b","IPY_MODEL_ca4b077f3333452ba42ffa939e2912b5"],"layout":"IPY_MODEL_295e621961d6484488cbfc149f4b104f"}},"8287503040974fcd8b17ff68962f3618":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_bed292f0a89946768aabb3ff41260b01","placeholder":"​","style":"IPY_MODEL_47f528c8faaf45ab860056120680142c","value":"Computing silhuette score for various distance thresholds: 100%"}},"73fe7b1bc0c34976941af9c6416f084b":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"FloatProgressModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"ProgressView","bar_style":"success","description":"","description_tooltip":null,"layout":"IPY_MODEL_2c5d4dac0c4c4522a1b3c36e6c8974f2","max":10,"min":0,"orientation":"horizontal","style":"IPY_MODEL_c612f1812b2f4ab6ac55bbb950ecb501","value":10}},"ca4b077f3333452ba42ffa939e2912b5":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_b3af29d905cb48f5811bf39c56660bd3","placeholder":"​","style":"IPY_MODEL_59bd24ba9995475797d7d4fbd4156eac","value":" 10/10 [01:01&lt;00:00,  6.17s/it]"}},"295e621961d6484488cbfc149f4b104f":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"bed292f0a89946768aabb3ff41260b01":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"47f528c8faaf45ab860056120680142c":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"2c5d4dac0c4c4522a1b3c36e6c8974f2":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"c612f1812b2f4ab6ac55bbb950ecb501":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"ProgressStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","bar_color":null,"description_width":""}},"b3af29d905cb48f5811bf39c56660bd3":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"59bd24ba9995475797d7d4fbd4156eac":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"d059d2b1271141a3815097af381734a1":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HBoxModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HBoxView","box_style":"","children":["IPY_MODEL_189824963eaa483faca91024e6d66607","IPY_MODEL_18ae8631dac54d4c8d4cde50d7f6d25f","IPY_MODEL_5752f91c41004a0b9445070508f12f68"],"layout":"IPY_MODEL_c2c5d2fbf46046d896d8d650a40fce2b"}},"189824963eaa483faca91024e6d66607":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_019380ad0fbb4dab920c4462522402b6","placeholder":"​","style":"IPY_MODEL_90d48dc4f10a4624b95d64d893a8609c","value":"Create the clusters with the selected distance thresholds: "}},"18ae8631dac54d4c8d4cde50d7f6d25f":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"FloatProgressModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"ProgressView","bar_style":"success","description":"","description_tooltip":null,"layout":"IPY_MODEL_5e9634959c2b4c508fbfdd0ea502625f","max":1,"min":0,"orientation":"horizontal","style":"IPY_MODEL_5e14cbd1a3ce4550ac5c08376bc04dff","value":1}},"5752f91c41004a0b9445070508f12f68":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_cc35cf544e764c78be05f18827f49662","placeholder":"​","style":"IPY_MODEL_2b85fc5fafc74dd59a0c40d14a6628d4","value":" 10/? [00:00&lt;00:00, 31.56it/s]"}},"c2c5d2fbf46046d896d8d650a40fce2b":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"019380ad0fbb4dab920c4462522402b6":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"90d48dc4f10a4624b95d64d893a8609c":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"5e9634959c2b4c508fbfdd0ea502625f":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":"20px"}},"5e14cbd1a3ce4550ac5c08376bc04dff":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"ProgressStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","bar_color":null,"description_width":""}},"cc35cf544e764c78be05f18827f49662":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"2b85fc5fafc74dd59a0c40d14a6628d4":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"d40a74633a4e4d02b367c7864760a0f0":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HBoxModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HBoxView","box_style":"","children":["IPY_MODEL_9cb5c0c528f140caa7b01b2ea307aeef","IPY_MODEL_e86f42f615914675a69c03595a2e0965","IPY_MODEL_99d2db05ed3a41d5bee7913abc0d8663"],"layout":"IPY_MODEL_c7bdd99223ce4d6c88de0300be8e95ae"}},"9cb5c0c528f140caa7b01b2ea307aeef":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_f4665222fa824d76a7a383f5d83e59f3","placeholder":"​","style":"IPY_MODEL_685274e6e591482298dbb9ffee146851","value":""}},"e86f42f615914675a69c03595a2e0965":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"FloatProgressModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"ProgressView","bar_style":"success","description":"","description_tooltip":null,"layout":"IPY_MODEL_41dfb5e193bb4b84be9ebd33835ecee2","max":1,"min":0,"orientation":"horizontal","style":"IPY_MODEL_ca91d0e56d624c91843662a3b617bce4","value":1}},"99d2db05ed3a41d5bee7913abc0d8663":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_2f18ab1c4cd347cba56f0822a8dfab0f","placeholder":"​","style":"IPY_MODEL_d63a2c90243a41ee8795b0fc279c7e24","value":" 100/? [00:00&lt;00:00, 459.92it/s]"}},"c7bdd99223ce4d6c88de0300be8e95ae":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"f4665222fa824d76a7a383f5d83e59f3":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"685274e6e591482298dbb9ffee146851":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"41dfb5e193bb4b84be9ebd33835ecee2":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":"20px"}},"ca91d0e56d624c91843662a3b617bce4":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"ProgressStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","bar_color":null,"description_width":""}},"2f18ab1c4cd347cba56f0822a8dfab0f":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"d63a2c90243a41ee8795b0fc279c7e24":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"895daecc521046859758fcb876b80c13":{"model_module":"@jupyter-widgets/controls","model_name":"HBoxModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HBoxModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HBoxView","box_style":"","children":["IPY_MODEL_257aa47024bd4bfb81de1b82809bfd13","IPY_MODEL_ca13eee58102488199021886c4913fdb","IPY_MODEL_e53a1ef479794ab5a99dcf77cc9e6969"],"layout":"IPY_MODEL_860547aa48144bfa8870f5c9fdb0281c"}},"257aa47024bd4bfb81de1b82809bfd13":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_53a3e7bda0ff44f7a86594610a4c680a","placeholder":"​","style":"IPY_MODEL_d08e77d3b5b645a9b93a1c62267dfc1e","value":""}},"ca13eee58102488199021886c4913fdb":{"model_module":"@jupyter-widgets/controls","model_name":"FloatProgressModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"FloatProgressModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"ProgressView","bar_style":"success","description":"","description_tooltip":null,"layout":"IPY_MODEL_61033ba9035845749b89c8cd193eb81a","max":1,"min":0,"orientation":"horizontal","style":"IPY_MODEL_56ab2d070f5746bfb7df0aa6d48675d8","value":1}},"e53a1ef479794ab5a99dcf77cc9e6969":{"model_module":"@jupyter-widgets/controls","model_name":"HTMLModel","model_module_version":"1.5.0","state":{"_dom_classes":[],"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"HTMLModel","_view_count":null,"_view_module":"@jupyter-widgets/controls","_view_module_version":"1.5.0","_view_name":"HTMLView","description":"","description_tooltip":null,"layout":"IPY_MODEL_392d8a826a1a4c779c0b2b8ba319586b","placeholder":"​","style":"IPY_MODEL_7550ae4115574784b50446f02558cef6","value":" 100/? [00:00&lt;00:00, 318.22it/s]"}},"860547aa48144bfa8870f5c9fdb0281c":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"53a3e7bda0ff44f7a86594610a4c680a":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"d08e77d3b5b645a9b93a1c62267dfc1e":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}},"61033ba9035845749b89c8cd193eb81a":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":"20px"}},"56ab2d070f5746bfb7df0aa6d48675d8":{"model_module":"@jupyter-widgets/controls","model_name":"ProgressStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"ProgressStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","bar_color":null,"description_width":""}},"392d8a826a1a4c779c0b2b8ba319586b":{"model_module":"@jupyter-widgets/base","model_name":"LayoutModel","model_module_version":"1.2.0","state":{"_model_module":"@jupyter-widgets/base","_model_module_version":"1.2.0","_model_name":"LayoutModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"LayoutView","align_content":null,"align_items":null,"align_self":null,"border":null,"bottom":null,"display":null,"flex":null,"flex_flow":null,"grid_area":null,"grid_auto_columns":null,"grid_auto_flow":null,"grid_auto_rows":null,"grid_column":null,"grid_gap":null,"grid_row":null,"grid_template_areas":null,"grid_template_columns":null,"grid_template_rows":null,"height":null,"justify_content":null,"justify_items":null,"left":null,"margin":null,"max_height":null,"max_width":null,"min_height":null,"min_width":null,"object_fit":null,"object_position":null,"order":null,"overflow":null,"overflow_x":null,"overflow_y":null,"padding":null,"right":null,"top":null,"visibility":null,"width":null}},"7550ae4115574784b50446f02558cef6":{"model_module":"@jupyter-widgets/controls","model_name":"DescriptionStyleModel","model_module_version":"1.5.0","state":{"_model_module":"@jupyter-widgets/controls","_model_module_version":"1.5.0","_model_name":"DescriptionStyleModel","_view_count":null,"_view_module":"@jupyter-widgets/base","_view_module_version":"1.2.0","_view_name":"StyleView","description_width":""}}}}},"nbformat":4,"nbformat_minor":0}