Fitting large Bayesian network models quickly become computationally infeasible when the number of nodes grows into the hundreds of thousands and millions. In particular, the mixed membership stochastic blockmodel (MMSB) is a popular Bayesian network model used for community detection. In this paper, we introduce a scalable inference method that leverages nodal information that often accompanies real-world networks. Conditioning on this extra information leads to a model that admits a parallel variational inference algorithm. We apply our method to a citation network with over two million nodes and 25 million edges. Our method recovers parameters and achieves convergence better on simulated networks generated according to the MMSB.
تحميل البحث