Nonnegative Bayesian nonparametric factor models with completely random measures for community detection

Abstract

We present a Bayesian nonparametric Poisson factorization model for modeling network data with an unknown and potentially growing number of overlapping communities. The construction is based on completely random measures and allows the number of communities to either increase with the number of nodes at a specified logarithmic or polynomial rate, or be bounded. We develop asymptotics for the number of nodes and the degree distribution of the network and derive a Markov chain Monte Carlo algorithm for targeting the exact posterior distribution for this model. The usefulness of the approach is illustrated on various real networks.

References in corpus (6)

• An efficient and principled method for detecting communities in networks
• Beta-Negative Binomial Process and Poisson Factor Analysis
• Sparse graphs using exchangeable random measures
• MCMC for Normalized Random Measure Mixture Models
• Using Triangles to Improve Community Detection in Directed Networks
• Poisson Latent Feature Calculus for Generalized Indian Buffet Processes

Cited by in corpus (3)

• A unified construction for series representations and finite approximations of completely random measures
• Slice Sampling for General Completely Random Measures
• Posterior distributions for Hierarchical Spike and Slab Indian Buffet processes