Spatiotemporal blocking of the bouncy particle sampler for efficient inference in state-space models

Abstract

We propose a novel blocked version of the continuous-time bouncy particle sampler of Bouchard-Côté et al. (J Am Stat Assoc 113(522):855–867, 2018) which is applicable to any differentiable probability density. This alternative implementation is motivated by blocked Gibbs sampling for state-space models (Singh et al. in Biometrika 104(4):953–969, 2017) and leads to significant improvement in terms of effective sample size per second, and furthermore, allows for significant parallelization of the resulting algorithm. The new algorithms are particularly efficient for latent state inference in high-dimensional state-space models, where blocking in both space and time is necessary to avoid degeneracy of MCMC. The efficiency of our blocked bouncy particle sampler, in comparison with both the standard implementation of the bouncy particle sampler and the particle Gibbs algorithm of Andrieu et al. (J R Stat Soc Ser B Stat Methodol 72(3):269–342, 2010), is illustrated numerically for both simulated data and a challenging real-world financial dataset.

Can we persuade company car drivers to opt out?

“…The probability of getting at least one significant result due to chance rises exponentially with the number of hypotheses. Therefore, the chance of getting a type I error (i.e., finding a significant result that in reality is not there) increases with each hypothesis test (Goldman, 2008). Two other limitations of this study include the potential biases associated with self-reported data obtained through online surveys, as well as the relevance of using data from a survey conducted in 2018.…”

Concave-Convex PDMP-based sampling

“…Monte Carlo methods based on continuous-time Markov processes have shown potential for efficient sampling in Bayesian inference challenges (Goldman and Singh, 2021;Fearnhead et al, 2018). These new sampling algorithms are based on simulating piecewise-deterministic Markov processes (PDMPs).…”

“…Among these approaches, Cotter et al (2020) propose numerical integration and root-finding algorithms to facilitate simulation. Others have considered using approximate local bounds which can be simulated exactly (Goldman and Singh, 2021;Goan and Fookes, 2021;Pakman et al, 2017). Both of these approaches sacrifice exact sampling of the posterior and involve a trade-off between the computational cost and the approximation of the sampling distribution.…”