add screenshots

Author: Jonas1312

Screenshots/BH_eigenvalues.png

@@ -12,7 +12,7 @@
 from stochasticBlockModel import SBM
 from sklearn.metrics.cluster import normalized_mutual_info_score
 
-n_vertices = 1000 # number of vertices
+n_vertices = 1500 # number of vertices
 n_communities = 2 # number of communities
 
 # cin > cout is referred to as the assortative case
@@ -25,15 +25,15 @@
 best_r = np.sqrt(np.mean(sbm.average_degree))
 print("r = sqrt(average degree) = " + str(best_r))
 
-x_axis = sorted([1, 2, 3, 3.5, 4, 6, 10, best_r])
+x_axis = sorted([1, 2, 3, 3.5, 4, 6, 15, best_r])
 y_axis = []
 
 plt.figure(1)
 
 for i, r in enumerate(x_axis):
 spectral_labels, eigvals, eigvects, W = SpectralClustering(n_communities, BetheHessian(sbm.adjacency_matrix, r), "BetheHessian") # spectral clustering
 plt.subplot(2,4,i+1)
-plt.title("Bethe Hessian eigenvalues with r = " + str(r))
+plt.title("r = " + str(r))
 plt.hist(eigvals, bins=100) # plot histogram of the eigenvalues
 nmi = normalized_mutual_info_score(sbm.community_labels, spectral_labels)
 if r == best_r: best_nmi = nmi
@@ -42,7 +42,7 @@
 
 plt.figure(2)
 plt.ylim(0,1)
-plt.title("NMI = f(r)")
+plt.title("NMI = f(r), cin = {}, cout = {}, n_vertices = {}".format(cin, cout, n_vertices))
 plt.xlabel("r")
 plt.ylabel("NMI")
 plt.plot(x_axis, y_axis)

Screenshots/cin cout matrices comparison.png

@@ -11,7 +11,7 @@
 from sklearn.metrics.cluster import normalized_mutual_info_score
 from spectralClustering import SpectralClustering
 
-n_vertices = 700
+n_vertices = 2000
 x_step = 0.5
 x_axis = [(15,i) for i in np.arange(4,8,x_step)]
 yNMI_bethehessian = [0]*len(x_axis)
@@ -20,7 +20,7 @@
 yNMI_adjacency = [0]*len(x_axis)
 
 n_communities = 2
-n_iterations = 2
+n_iterations = 4
 
 for n in xrange(n_iterations):
 print("Iteration {}, {} left".format(n+1, n_iterations-n-1))
@@ -29,21 +29,25 @@
 sbm = SBM(n_vertices, n_communities, probability_matrix)
 
 # BetheHessian matrix
+print("BetheHessian")
 spectral_labels, _, _, _ = SpectralClustering(n_communities, BetheHessian(sbm.adjacency_matrix), "BetheHessian")
 nmi = normalized_mutual_info_score(sbm.community_labels, spectral_labels)
 yNMI_bethehessian[i] += nmi
 
 # Modularity matrix
+print("Modularity")
 spectral_labels, _, _, _ = SpectralClustering(n_communities, ModularityMatrix(sbm.adjacency_matrix), "ModularityMatrix")
 nmi = normalized_mutual_info_score(sbm.community_labels, spectral_labels)
 yNMI_modularity[i] += nmi
 
 # Laplacian matrix
+print("Laplacian")
 spectral_labels, _, _, _ = SpectralClustering(n_communities, LaplacianMatrix(sbm.adjacency_matrix), "LaplacianMatrix")
 nmi = normalized_mutual_info_score(sbm.community_labels, spectral_labels)
 yNMI_laplacian[i] += nmi
 
 # Adjacency matrix
+print("Adjacency")
 spectral_labels, _, _, _ = SpectralClustering(n_communities, sbm.adjacency_matrix, "AdjacencyMatrix")
 nmi = normalized_mutual_info_score(sbm.community_labels, spectral_labels)
 yNMI_adjacency[i] += nmi

Screenshots/figured_1.png renamed to Screenshots/dense vs sparse graphs.png

@@ -11,17 +11,16 @@
 from sklearn.metrics.cluster import normalized_mutual_info_score
 from spectralClustering import SpectralClustering
 
-n_vertices = 800
+n_vertices = 1800
 cin = 15
 cout = 5
-x_axis = range(1,6)
+x_axis = range(2,6)
 yNMI_bethehessian = [0]*len(x_axis)
 yNMI_modularity = [0]*len(x_axis)
 yNMI_laplacian = [0]*len(x_axis)
 yNMI_adjacency = [0]*len(x_axis)
 
-
-n_iterations = 1
+n_iterations = 3
 
 for n in xrange(n_iterations):
 print("Iteration {}, {} left".format(n+1, n_iterations-n-1))