GPU Accelerated PMCMC Algorithm with System Dynamics Modelling

Abstract

Recent work demonstrates that coupling Bayesian computational statistics methods with dynamic models can facilitate analysis and understanding of complex systems associated with diverse time series, those involving social and behavioral dynamics. Particle Markov Chain Monte Carlo (PMCMC) is a particularly powerful class of Bayesian methods combining aspects of batch Markov Chain Monte Carlo (MCMC) and the sequential Monte Carlo method of Particle Filtering (PF). PMCMC can flexibly combine theory-capturing dynamic models with diverse empirical data streams. PMCMC has demonstrated great potential for broad applicability across social and behavioral domains. While PMCMC offers high analytic power, such power imposes a high computational burden. In this work, we investigated the effectiveness of using Graphical Processing Units (GPUs) in reducing run times. Specifically, we designed and implemented a GPU-enabled parallel PMCMC version with compartmental simulation models. Evaluating this work’s impact with a realistic PMCMC health application showed that GPU-based acceleration achieves up to $160\times$ speedup compared to a corresponding sequential CPU-based version. Use of the GPU accelerated PMCMC algorithm with dynamic models offers researchers a powerful toolset to greatly accelerate learning and secure additional insight from the high-velocity data increasingly prevalent within social and behavioral spheres.