Beta-Negative Binomial Process and Exchangeable Random Partitions for Mixed-Membership Modeling

摘要: The beta-negative binomial process (BNBP), an integer-valued stochastic process, is employed to partition a count vector into latent random matrix. As the marginal probability distribution of BNBP that governs exchangeable partitions grouped data has not yet been developed, current inference for truncate number atoms beta process. This paper introduces function explicitly describe how clusters points each group partitions, which are shared across all groups. A fully collapsed Gibbs sampler developed BNBP, leading novel nonparametric Bayesian topic model distinct from existing ones, with simple implementation, fast convergence, good mixing, and state-of-the-art predictive performance.