A Bayesian Graph Clustering Approach Using the Prior Based on Degree Distribution

Newman et al. proposed a stochastic graph clustering approach using a mixture model with an assumption that a group of vertices is regarded as a class when the vertices have a similar connection pattern. Kuwata et al. recently adopted a nonparametric Bayesian approach and improved Newman's one in such a way that the number of classes can also be empirically estimated. In this paper, we propose a new approach that can incorporate the degree distribution of the network structure as priors for Bayesian estimation. We show the effectiveness of our method through experiments using both artificial and real data.