Concave-Convex PDMP-based samplers
Date: 16 September 2022, FridayTime: 4pm AEST
Speaker: Matthew Sutton (QUT)
Abstract: Recently non-reversible samplers based on simulating piecewise deterministic Markov processes (PDMPs) have shown potential for efficient sampling in Bayesian inference problems. In this talk, I will show how these methods may be implemented efficiently when the rate function admits a concave-convex decomposition [1]. This approach facilitates simple implementation and computationally efficient thinning for a wide range of problems. I will show the merits of this approach with empirical scaling analysis and applications to variable selection problems using reversible-jump PDMP-based samplers [2].
[1] Sutton, M., & Fearnhead, P. (2021). Concave-Convex PDMP-based sampling. In arXiv [stat.ME]. arXiv. http://arxiv.org/abs/2112.12897
[2] Chevallier, A., Fearnhead, P., & Sutton, M. (2022) Reversible Jump PDMP Samplers for Variable Selection, Journal of the American Statistical Association, DOI: 10.1080/01621459.2022.2099402
Password: 017349