Abstract:Stochastic Block Models (SBMs) have emerged as a powerful framework for modelling community structures in networks. However, accurately determining the number of communities in SBMs remains challenging. In this paper, we propose a novel approach for community detection in stochastic block models through penalized variational estimation. Our method introduces two penalties on the variational objective function to ensure elimination of redundant blocks. We present the variational computing procedure, and demonstrate through simulation and empirical analysis that our approach can effectively identifies community structures and achieves accurate parameter estimation simultaneously.

Community detection in stochastic block models via penalized variational estimation