Bayesian Network Structure Learning: A Sequential Monte Carlo Approach

• Activity Number: 401
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Computing
• Author(s): Kaixian Yu*+ and Jinfeng Zhang
• Companies: Florida State University and Florida State University
• Keywords: Bayesian Networks ; Structure Learning ; Sequential Monte Carlo ; Parallel computing

Abstract

Bayesian networks (BNs) provide a probabilistic, graphical framework for modeling high-dimensional joint distributions with complex correlation structures. BNs have wide applications in many scientific disciplines, including biology, computer science, social science and medicine. Despite extensive studies in the past, network structure learning from data is still an open question in BN research. In this study, we propose a new approach based on sequential Monte Carlo (SMC). Compared to MCMC based methods, SMC generates independent, weighted samples which may be used to perform Bayesian model averaging to provide better predictive ability. SMC can also be implemented efficiently for sampling large numbers of networks with diverse structures. We compare our method with several state of art methods in terms of skeleton discovery and v-structure preservation.