How to use list_metrics method in localstack

Best Python code snippet using localstack_python

Metrics.py

Source:Metrics.py

1import numpy as np2import pandas as pd3import torch4import sklearn5from sklearn.neighbors import NearestNeighbors6def get_metrics(config, list_metrics, n_neighbors=2):7    """8    Generates an adversarial examples x' from an original sample x. Expected to contain9            Dataset, MaxIters, Alpha, Lambda, TrainData, TestData, ValidData, Scaler,10            FeatureNames, Target, Weights, Bounds, Model11    :param config: dictionnary containing the configuration for the experiment12    :param list_metrics: dictionnary containing the metrics to be computed. Choose from13            from SuccessRate, iter_means, iter_std, normdelta_median, normdelta_mean,14            n_std, weighted_median, weighted_mean, w_std, mean_dists_at_org,15            median_dists_at_org, mean_dists_at_tgt, mean_dists_at_org_weighted, mdow_std,16            median_dists_at_org_weighted, mean_dists_at_tgt_weighted, mdtw_std, prop_same_class_arg_org,17            prop_same_class_arg_adv,18    :param n_neighbors: number of neighbors to compute the distance to n_neighbors closest neighbors19    """20    metrics_for_conf = []21    df_test = config['TestData']22    dfs_adv = config['AdvData']23    24    for method, df_adv in dfs_adv.items():25        metrics_for_method = [method]26        # Get success rate before removing samples from dataframe27        if list_metrics['SuccessRate']:28            sr = metric_success_rate_for(df_adv)29            metrics_for_method.append(sr)30        # Removes samples that did cross frontier31        df_adv = remove_non_converted(df_adv)        32        df_adv = add_normdelta_to(df_adv, config, df_test)33        # Adding proportion of neighbors from diff classes34        df_adv, df_adv_weighted = add_maj_neighbors(df_adv, df_test, config, n_neighbors=n_neighbors)            35        # Mean, std, number of iterations36        if list_metrics['iter_means']:37            means_iters, stds_iters = mean_norm_for_col(df_adv, col='iters')38            metrics_for_method.append(means_iters)39            if list_metrics['iter_std']:40                metrics_for_method.append(stds_iters)41        # Median, norm of perturbation42        if list_metrics['normdelta_median']:43            median = median_norm_for_col(df_adv, col='normdelta')44            metrics_for_method.append(median)45        # Mean, std, norm of perturbation46        if list_metrics['normdelta_mean']:47            means, stds = mean_norm_for_col(df_adv, col='normdelta')48            metrics_for_method.append(means)49            if list_metrics['n_std']:50                metrics_for_method.append(stds)51        # Median, norm of perturbation, weighted52        if list_metrics['weighted_median']:53            median_w = median_norm_for_col(df_adv, col='normdelta_weighted')54            metrics_for_method.append(median_w)55        # Mean, std, norm of perturbation, weighted56        if list_metrics['weighted_mean']:57            means_w, stds_w = mean_norm_for_col(df_adv, col='normdelta_weighted')58            metrics_for_method.append(means_w)59            if list_metrics['w_std']:60                metrics_for_method.append(stds_w) 61        # Mean, std, number of neighbors of a particular class at perturbed sample62        if list_metrics['mean_dists_at_org']:63            mean, std = mean_norm_for_col(df_adv, col='mean_dists_at_org')64            metrics_for_method.append(mean)65        # Mean, std, number of neighbors of a particular class at perturbed sample66        if list_metrics['median_dists_at_org']:67            med = median_norm_for_col(df_adv, col='mean_dists_at_org')68            metrics_for_method.append(med)69        # Mean, std, number of neighbors of a particular class at perturbed sample70        if list_metrics['mean_dists_at_tgt']:71            mean, std = mean_norm_for_col(df_adv, col='mean_dists_at_tgt')72            metrics_for_method.append(mean)73        # Mean, std, number of neighbors of a particular class at perturbed sample74        if list_metrics['mean_dists_at_org_weighted']:75            mean, std = mean_norm_for_col(df_adv_weighted, col='mean_dists_at_org')76            metrics_for_method.append(mean)77            if list_metrics['mdow_std']:78                metrics_for_method.append(std)79        # Mean, std, number of neighbors of a particular class at perturbed sample80        if list_metrics['median_dists_at_org_weighted']:81            median = median_norm_for_col(df_adv_weighted, col='mean_dists_at_org')82            metrics_for_method.append(median)83        # Mean, std, number of neighbors of a particular class at perturbed sample84        if list_metrics['mean_dists_at_tgt_weighted']:85            mean, std = mean_norm_for_col(df_adv_weighted, col='mean_dists_at_tgt')86            metrics_for_method.append(mean)87            if list_metrics['mdtw_std']:88                metrics_for_method.append(std)89        # Mean, std, number of neighbors of a particular class at perturbed sample90        if list_metrics['prop_same_class_arg_org']:91            mean, std = mean_norm_for_col(df_adv, col='prop_same_class_arg_org')92            metrics_for_method.append(mean)93        # Mean, std, number of neighbors of a particular class at perturbed sample94        if list_metrics['prop_same_class_arg_adv']:95            mean, std = mean_norm_for_col(df_adv, col='prop_same_class_arg_adv')96            metrics_for_method.append(mean)97            98        metrics_for_conf.append(metrics_for_method)99    return metrics_for_conf100  101def metric_success_rate_for(df):102    return len(df[df['orig_pred'] != df['adv_pred']]) / df.shape[0]103def remove_non_converted(df):104    df_return = df.copy()105    return df[df['orig_pred'] != df['adv_pred']]106def mean_norm_for_col(df, col):107    tmp = df[col]    108    mean, std = np.mean(tmp), np.std(tmp)109    return (mean, std)110def median_norm_for_col(df, col):111    tmp = df[col]    112    median = np.median(tmp)113    return median114def add_normdelta_to(df_adv, conf, df):115    # Drop columns if already there116    df_return = df_adv.copy()117    if 'normdelta' in df_return.columns:118        df_return = df_return.drop(columns='normdelta')119    if 'normdelta_weighted' in df_return.columns:120        df_return = df_return.drop(columns='normdelta_weighted')121        122    feature_names = conf['FeatureNames']123    weights = conf['Weights']124    norms = []125    norms_weighted = []126    127    # Iterate over all rows128    for index, row in df_return.iterrows():129        orig = df.loc[index][feature_names].values130        adv = row[feature_names].values 131        132        # Compute deltas133        delta = np.abs(orig-adv)134        assert(len(delta) == len(weights))135        136        # Norms delta137        norms.append(np.linalg.norm(delta))138        139        # Norms delta weighted140        norms_weighted.append(np.linalg.norm(delta * weights))141    df_return.insert(0, 'normdelta', norms)142    df_return.insert(0, 'normdelta_weighted', norms_weighted)143    144    return df_return145def get_majority_neighbors(df_adv, df_orig, conf, knn, n_neighbors):146    147    # orig, adv148    mean_dists = [[], []]149    prop_same_class = [[], []]150    151    feature_names = conf['FeatureNames']152    target = conf['Target']    153    154    # For each sample155    for index, row in df_adv.iterrows():156        157        orig = df_orig.loc[index][feature_names].values158        adv = row[feature_names].values159        160        preds = [row['orig_pred'], row['adv_pred']]161        samples = [orig, adv]162        163        for i in range(len(preds)):164            165            sample = samples[i]166            pred = preds[i]167            168            distance, neighbors_idxs = knn.kneighbors([sample], n_neighbors)169            neighbors_samples = df_orig.iloc[neighbors_idxs[0]]170            171           172            distance = [distance[0][1:]]173            neighbors_idxs = [neighbors_idxs[0][1:]]174            175            # Distance to closest neighbors176            if len(distance[0]) > 0 :177                dst_mean = np.mean(distance[0])178            else:179                print('Error, no neighbor found')180            mean_dists[i].append(dst_mean)181            182            neighbors_pts_target = np.array(neighbors_samples[target]).astype(int)183            prop = list(neighbors_pts_target).count(pred)184            prop_same_class[i].append(float(prop)/float(n_neighbors))185                        186    return mean_dists, prop_same_class187def add_maj_neighbors_to(df_adv, df_orig, conf, knn, n_neighbors):188    df_return = df_adv.copy()189    190    if 'mean_dists_at_org' in df_return.columns:191        df_return = df_return.drop(columns='mean_dists_at_org')192    if 'mean_dists_at_tgt' in df_return.columns:193        df_return = df_return.drop(columns='mean_dists_at_tgt')194    if 'prop_same_class_arg_org' in df_return.columns:195        df_return = df_return.drop(columns='prop_same_class_arg_org')196    if 'prop_same_class_arg_adv' in df_return.columns:197        df_return = df_return.drop(columns='prop_same_class_arg_adv')198        199    mean_dists, prop_same_class = get_majority_neighbors(df_adv, df_orig, conf, knn, n_neighbors)200    201    df_return.insert(0, 'mean_dists_at_org', mean_dists[0])202    df_return.insert(0, 'mean_dists_at_tgt', mean_dists[1])203    df_return.insert(0, 'prop_same_class_arg_org', prop_same_class[0])204    df_return.insert(0, 'prop_same_class_arg_adv', prop_same_class[1])205    206    return df_return207def scale_data(conf, df_orig):208    print('Before')209    print(df.describe(include='all'))210    print(weights)211    for col, weight in zip(list(df.columns), weights):212        df[col] = df[col].apply(lambda x: x * weight)213        214    bounds = [[bounds[i][x] * weight for x, weight in enumerate(weights)] for i in range(len(bounds))]215    print(df.describe(include='all'))216    return df, bounds217def weighted_distance(x, y, w):218    sum_ = 0219    assert(len(x) == len(y) == len(w))220    for i in range(len(x)):221        sum_ += (w[i] * (y[i] - x[i])) ** 2222    sum_ = np.sqrt(sum_)223    return sum_224def add_maj_neighbors(df_adv, df_orig, conf, n_neighbors):225    # Otherwise we have issues because the KNN returns indexes in len(df) and not based on the real indexes on the samples226    df_adv = df_adv.reset_index().drop(columns=['index'])227    df_orig = df_orig.reset_index().drop(columns=['index'])228    weights = conf['Weights']229    assert(weights[0] > 0)230    feature_names = conf['FeatureNames']231    target = conf['Target']232        233        234    knn = NearestNeighbors(n_neighbors, metric='l2')235    knn.fit(df_orig[feature_names], df_orig[target])236    237    knn_weighted = NearestNeighbors(n_neighbors, metric=weighted_distance, metric_params={'w' : weights})238    knn_weighted.fit(df_orig[feature_names], df_orig[target])239    240    df_adv_return = add_maj_neighbors_to(df_adv, df_orig, conf, knn, n_neighbors)241    df_adv_weighted = add_maj_neighbors_to(df_adv, df_orig, conf, knn_weighted, n_neighbors)242    ...

graph.py

Source:graph.py

1import os, copy2import modules.node3import modules.extra4class Graph(object):5	def __init__(self, graphfile):6		self.graphfile = graphfile7		self.nodes = {}8		9		self.n = 010		self.m = 011		self.density = 0.012		13		self.list_local_degree = {}14		self.list_local_cc = {}15		self.list_local_rc = {}16		self.list_metrics = {}17		self.list_correlations = {}18	def compute_metrics(self):19		self.__compute_degree_informations()				20		self.__compute_cc_informations()21		self.__compute_rc_informations()22	def create_metrics(self):23		self.list_metrics["degree"] = metric.Metric(self, "degree", self.list_local_degree)24		self.list_metrics["degree"].compile()25		#self.list_metrics["degree_norm"] = self.list_metrics["degree"].normalization()26		self.list_metrics["cc"] = metric.Metric(self, "cc", self.list_local_cc)27		self.list_metrics["cc"].compile()28		self.list_metrics["rc"] = metric.Metric(self, "rc", self.list_local_rc)29		self.list_metrics["rc"].compile()30	def treat_correlations(self):31		for metric1 in self.list_metrics:32			for metric2 in self.list_metrics:33				if metric1 != metric2:			34					name_correlation = "%s-%s" % (metric1, metric2)35					self.list_correlations[name_correlation] = metric.Correlation(self, self.list_metrics[metric1], self.list_metrics[metric2])36					self.list_correlations[name_correlation].compile()37		38	def __compute_degree_informations(self):39		for id_node in self.nodes:40			degree = self.nodes[id_node].degree41			self.list_local_degree[id_node] = degree42			self.m += degree43		self.m /= 244		self.average_degree = (2 * self.m) / float(self.n)45		self.density = (2 * self.m) / float(self.n * (self.n - 1))46	47	def __compute_cc_informations(self):48		for id_node in self.nodes:49			self.list_local_cc[id_node] = -1.050			k = self.nodes[id_node].degree * (self.nodes[id_node].degree - 1)51			# ALTERNATIVE METHOD (deprecated degree 1 nodes)52			'''k = 053			for id_neighbour in self.nodes[id_node].list_neighbours:54				if self.nodes[id_neighbour].degree > 1:55					k += 156			k = k * (k - 1)	'''57			v = 058			for id_neighbour1 in self.nodes[id_node].list_neighbours:59				for id_neighbour2 in self.nodes[id_node].list_neighbours:60					if id_neighbour1 != id_neighbour2:						61						if (id_neighbour2 in self.nodes[id_neighbour1].list_neighbours):62							v += 163			if k != 0 and v != 0:64				self.list_local_cc[id_node] = v / float(k)65		66	def __compute_rc_informations(self):	67		for id_node in self.nodes:68			self.list_local_rc[id_node] = -1.069			k = self.nodes[id_node].degree * (self.nodes[id_node].degree - 1)70			# ALTERNATIVE METHOD (deprecated degree 1 nodes)71			'''k = 072			for id_neighbour in self.nodes[id_node].list_neighbours:73				if self.nodes[id_neighbour].degree > 1:74					k += 175			k = k * (k - 1)	'''76			v = 077			78			for id_neighbour1 in self.nodes[id_node].list_neighbours:79				for id_neighbour2 in self.nodes[id_node].list_neighbours:80					if id_neighbour2 != id_neighbour1:81						for id_neighbour_id1 in self.nodes[id_neighbour1].list_neighbours:82							for id_neighbour_id2 in self.nodes[id_neighbour2].list_neighbours:83								if id_neighbour_id1 != id_node and id_neighbour_id1 == id_neighbour_id2:84									v += 0.585									break86							else:87								continue88							break89			90			if k != 0 and v != 0:91				self.list_local_rc[id_node] = (2*v) / float(k)92	93	def load(self):94		file = open(self.graphfile, 'r')95		# A AMERLIORER96		for line in file.read().splitlines():97			nodes_id = line.split()98			if nodes_id[0] not in self.nodes:99				self.nodes[nodes_id[0]] = modules.node.Node(nodes_id[0])100			for i in range(1, len(nodes_id)):			101				self.nodes[nodes_id[0]].add_neighbour(nodes_id[i])102		file.close()103		self.n = len(self.nodes)104	105	def informations(self):106		info = "\t\t#### [Statistics of graph (%s)] ####\n\n" % self.name107		info += "- GLOBAL STATS :\n\n"108		info += "\t# Size (n) = %d\n" % self.n109		info += "\t# Number of links (m) = %d\n" % self.m				110		info += "\t# Density = %0.6f\n" % self.density111		info += "\n- LOCAL METRICS :\n\n"112		for metric_name in self.list_metrics:113			info += str(self.list_metrics[metric_name])114		return info115	def informations_correlations(self):116		info = "\n\t ---- Correlations ---- \n"117		for correlation_name in self.list_correlations:118			info += str(self.list_correlations[correlation_name])119		return info120	def save_local_informations(self, filename):121		f = open(filename, 'w')122		com = "#ID\tDegree\tClustering coefficient\tRedundancy coefficient"123		if isinstance(self, graph_bipartite.Bipartite):124			com += "\tBipartite TOP(0)/BOT(1) part"125		com +="\n"126		f.write(com)127		for id_node in self.nodes:128			line = "%s\t%d\t%0.6f\t%0.6f" % (id_node, self.list_local_degree[id_node], self.list_local_cc[id_node], self.list_local_rc[id_node])129			if isinstance(self, graph_bipartite.Bipartite):130				if id_node in self.list_top_nodes:131					line += "\t0"132				else:133					line += "\t1"134			line +="\n"135			f.write(line)136		f.close()137	def save_metrics(self, directory_data):138		if not os.path.isdir(directory_data):139			os.mkdir(directory_data)140		for metric_name in self.list_metrics:141			self.list_metrics[metric_name].save(directory_data)142	def save_correlations(self, directory_data):143		directory_correlation = directory_data + "/correlations"144		if not os.path.isdir(directory_correlation):145			os.mkdir(directory_correlation)146		for correlation_name in self.list_correlations:147			self.list_correlations[correlation_name].save(directory_correlation)148	def __str__(self):		...

Automation Testing Tutorials

Learn to execute automation testing from scratch with LambdaTest Learning Hub. Right from setting up the prerequisites to run your first automation test, to following best practices and diving deeper into advanced test scenarios. LambdaTest Learning Hubs compile a list of step-by-step guides to help you be proficient with different test automation frameworks i.e. Selenium, Cypress, TestNG etc.