Residual Belief Propagation for Topic Modeling

Abstract

Fast convergence speed is a desired property for training topic models such as latent Dirichlet allocation (LDA), especially in online and parallel topic modeling algorithms for big data sets. In this paper, we develop a novel and easy-to-implement residual belief propagation (RBP) algorithm to accelerate the convergence speed for training LDA. The proposed RBP uses an informed scheduling scheme for asynchronous message passing, which passes fast convergent messages with a higher priority to influence those slow convergent messages at each learning iteration. Extensive empirical studies confirm that RBP significantly reduces the training time until convergence while achieves a much lower predictive perplexity than several state-of-the-art training algorithms for LDA, including variational Bayes (VB), collapsed Gibbs sampling (GS), loopy belief propagation (BP), and residual VB (RVB).