added class to compute graph metrics

Author: nimRobotics

legacy.py

@@ -0,0 +1,69 @@
+import numpy as np
+import bct as bct
+import networkx as nx
+import matplotlib.pyplot as plt
+import pandas as pd
+import utils
+
+# set seed
+np.random.seed(123)
+
+if __name__ == '__main__':
+    nnodes = 10
+    ngraphs = 1000
+
+    # plot the variation of global efficiency with sparsity from 0 to 1
+    sparsity = np.linspace(0, 1, 100)
+
+    global_efficiency_mean = []
+    global_efficiency_std = []
+    global_efficiency_fnirs = []
+
+    cluster_coef_mean = []
+    cluster_coef_std = []
+    cluster_coef_fnirs = []
+
+    fnirs_adj_matrix_stim_0 = np.loadtxt('output/stim_0_fc_mean.csv', delimiter=',')
+    # get min and max weights ignoring nan values
+    min_weight = np.nanmin(fnirs_adj_matrix_stim_0)
+    max_weight = np.nanmax(fnirs_adj_matrix_stim_0)
+    print('min weight: ', min_weight)
+    print('max weight: ', max_weight)
+
+    for s in sparsity:
+        ge_list = []
+        cc_list = []
+        for i in range(ngraphs):
+            adj_matrix = utils.RandomBinGraph(nnodes=nnodes, binarize_type='threshold', binarize_param=s).generate(min_weight=min_weight, max_weight=max_weight)
+            ge = bct.efficiency_bin(adj_matrix)
+            cc = np.mean(bct.clustering_coef_bu(adj_matrix))
+            ge_list.append(ge)
+            cc_list.append(cc)
+
+        ge_mean = np.mean(ge_list)
+        cc_mean = np.mean(cc_list)
+        ge_std = np.std(ge_list)
+        cc_std = np.std(cc_list)
+
+        global_efficiency_mean.append(ge_mean)
+        cluster_coef_mean.append(cc_mean)
+        global_efficiency_std.append(ge_std)
+        cluster_coef_std.append(cc_std)
+        global_efficiency_fnirs.append(bct.efficiency_bin(utils.BinarizeMatrix(fnirs_adj_matrix_stim_0, 'threshold', s).binarize()))
+        cluster_coef_fnirs.append(np.mean(bct.clustering_coef_bu(utils.BinarizeMatrix(fnirs_adj_matrix_stim_0, 'threshold', s).binarize())))
+
+    plt.errorbar(sparsity, global_efficiency_mean, yerr=global_efficiency_std, fmt='-o', capsize=3)
+    plt.plot(sparsity, global_efficiency_fnirs, '-o')
+    plt.xlabel('Sparsity')
+    plt.ylabel('Global Efficiency - Stim 0')
+    plt.legend(['fnirs data', 'random graph'])
+    plt.savefig('output/global_efficiency_stim_0.png')
+
+    plt.clf()
+
+    plt.errorbar(sparsity, cluster_coef_mean, yerr=cluster_coef_std, fmt='-o', capsize=3)
+    plt.plot(sparsity, cluster_coef_fnirs, '-o')
+    plt.xlabel('Sparsity')
+    plt.ylabel('Clustering Coefficient - Stim 0')
+    plt.legend(['fnirs data', 'random graph'])
+    plt.savefig('output/clustering_coef_stim_0.png')

main.py

@@ -63,85 +63,35 @@ def plot_features(self):
         Plot the features of the random graphs.
         '''
         for i, feature in enumerate(self.features):
-            plt.errorbar(self.search_space, [np.mean(self.random_graphs_features['{}_{}'.format(feature, sparsity)]) for sparsity in self.search_space], 
-                         yerr=[np.std(self.random_graphs_features['{}_{}'.format(feature, sparsity)]) for sparsity in self.search_space], label='Random Graphs')
-            plt.plot(self.search_space, [self.fc_graph_features['{}_{}'.format(feature, sparsity)] for sparsity in self.search_space], label='Real Graph', marker='o')
-            plt.xlabel('Sparsity')
+            plt.errorbar(self.search_space, 
+                        [np.mean(self.random_graphs_features['{}_{}'.format(feature, sparsity)]) for sparsity in self.search_space], 
+                        yerr=[np.std(self.random_graphs_features['{}_{}'.format(feature, sparsity)]) for sparsity in self.search_space], 
+                        label='Random Graphs', 
+                        fmt='-o',
+                        capsize=3)
+            plt.plot(self.search_space, 
+                    [self.fc_graph_features['{}_{}'.format(feature, sparsity)] for sparsity in self.search_space], 
+                    label='Real Graph', 
+                    marker='o')
+            plt.xlabel(self.analysis_type)
             plt.ylabel(feature)
             plt.legend()
-            plt.savefig('output/{}_sparsity.png'.format(feature))
+            plt.savefig('output/{}_{}.png'.format(feature, self.analysis_type), dpi=600)
             plt.clf()
 
+    def compute_auc(self):
+        '''
+        Compute the AUC for a given feature.
+        '''
+        pass
+
 
 if __name__ == '__main__':
     ngraphs = 1000
     fc_matrix = np.loadtxt('output/stim_0_fc_mean.csv', delimiter=',')
-    x = RandomGraphAnalyzer(fc_matrix, n_rnd_graphs=ngraphs, analysis_type='threshold', features=['global_efficiency', 'local_efficiency', 'clustering_coefficient'])
+    x = RandomGraphAnalyzer(fc_matrix, n_rnd_graphs=ngraphs, analysis_type='sparsity', features=['global_efficiency', 'local_efficiency', 'clustering_coefficient'])
     rgraphs = x.gen_random_graphs()
     print(len(rgraphs))
     print(len(rgraphs[0]))
     x.compute_features()
-    x.plot_features()
-
-
-
-
-# if __name__ == '__main__':
-#     nnodes = 10
-#     ngraphs = 1000
-
-#     # plot the variation of global efficiency with sparsity from 0 to 1
-#     sparsity = np.linspace(0, 1, 100)
-
-#     global_efficiency_mean = []
-#     global_efficiency_std = []
-#     global_efficiency_fnirs = []
-
-#     cluster_coef_mean = []
-#     cluster_coef_std = []
-#     cluster_coef_fnirs = []
-
-#     fnirs_adj_matrix_stim_0 = np.loadtxt('output/stim_0_fc_mean.csv', delimiter=',')
-#     # get min and max weights ignoring nan values
-#     min_weight = np.nanmin(fnirs_adj_matrix_stim_0)
-#     max_weight = np.nanmax(fnirs_adj_matrix_stim_0)
-#     print('min weight: ', min_weight)
-#     print('max weight: ', max_weight)
-
-#     for s in sparsity:
-#         ge_list = []
-#         cc_list = []
-#         for i in range(ngraphs):
-#             adj_matrix = utils.RandomBinGraph(nnodes=nnodes, binarize_type='threshold', binarize_param=s).generate(min_weight=min_weight, max_weight=max_weight)
-#             ge = bct.efficiency_bin(adj_matrix)
-#             cc = np.mean(bct.clustering_coef_bu(adj_matrix))
-#             ge_list.append(ge)
-#             cc_list.append(cc)
-
-#         ge_mean = np.mean(ge_list)
-#         cc_mean = np.mean(cc_list)
-#         ge_std = np.std(ge_list)
-#         cc_std = np.std(cc_list)
-
-#         global_efficiency_mean.append(ge_mean)
-#         cluster_coef_mean.append(cc_mean)
-#         global_efficiency_std.append(ge_std)
-#         cluster_coef_std.append(cc_std)
-#         global_efficiency_fnirs.append(bct.efficiency_bin(utils.BinarizeMatrix(fnirs_adj_matrix_stim_0, 'threshold', s).binarize()))
-#         cluster_coef_fnirs.append(np.mean(bct.clustering_coef_bu(utils.BinarizeMatrix(fnirs_adj_matrix_stim_0, 'threshold', s).binarize())))
-
-#     plt.errorbar(sparsity, global_efficiency_mean, yerr=global_efficiency_std, fmt='-o', capsize=3)
-#     plt.plot(sparsity, global_efficiency_fnirs, '-o')
-#     plt.xlabel('Sparsity')
-#     plt.ylabel('Global Efficiency - Stim 0')
-#     plt.legend(['fnirs data', 'random graph'])
-#     plt.savefig('output/global_efficiency_stim_0.png')
-
-#     plt.clf()
-
-#     plt.errorbar(sparsity, cluster_coef_mean, yerr=cluster_coef_std, fmt='-o', capsize=3)
-#     plt.plot(sparsity, cluster_coef_fnirs, '-o')
-#     plt.xlabel('Sparsity')
-#     plt.ylabel('Clustering Coefficient - Stim 0')
-#     plt.legend(['fnirs data', 'random graph'])
-#     plt.savefig('output/clustering_coef_stim_0.png')
+    x.plot_features()