python 实现spectral clustering

import numpy as np
import math
import sys
from scipy.cluster.vq import kmeans2

def SpectralClustering(simi_matrix,cluster_num):
	N,N = np.shape(simi_matrix);
	DN = np.diag(1/np.sqrt(np.sum(simi_matrix,axis=1)));
	LapN = np.eye(N) - np.dot(np.dot(DN,simi_matrix),DN);
	U,s,V = np.linalg.svd(LapN,full_matrices=True);
	kerN = U[:,N-cluster_num+1:N];
	for i in range(N):
		kerN[i,:] = kerN[i,:] / np.linalg.norm(kerN[i,:]);
	centroids,label = kmeans2(kerN,cluster_num,iter=20);
	del DN,U,s,V,kerN,centroids
	return label;