Spectral Algorithm for Community Detection Under Heterogeneous Networks

Compared with stochastic networks, the distribution of node degrees in real-world networks is characterized by heterogeneity. Spectral strategies based on the matrix eigenvector have been widely applied in the analysis of this type of network data, particularly for spatial division and community detection. Considering the performance of normalized Laplace matrix in heterogeneous networks outperforms some other common matrices, herein, a spectral strategy based on the eigenvector of matrix is proposed to tackle the modularity maximization problem in community detection.