Multiscale Community Detection in Networks with Heterogeneous Degree Distributions

Abstract

AbstractGraph clustering has been widely applied in exploring regularities emerging in relational data, e.g., community detection. Most existing methods investigate the community structure at a single topological scale. However, community structure of real world networks often exhibits multiple topological descriptions. Furthermore, the detection of multiscale community structure is heavily affected by the heterogeneous distribution of node degree. In this chapter, we propose a unified framework for detecting community structure from the perspective of dimensionality reduction. We first prove that Laplacian matrix and modularity matrix are two kinds of covariance matrices used in dimensionality reduction. We further develop a novel method to handle heterogeneity of networks by introducing a rescaling transformation into the covariance matrices in our framework. Extensive tests on real world and artificial networks demonstrate that the proposed method possesses high performance at identifying multiscale community structure in heterogeneous networks.KeywordsCovariance MatriceCorrelation MatriceCommunity DetectionNormalize Mutual InformationLaplacian MatrixThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.