Anchored Reversible Jump Sequential Monte Carlo

2026-10-07

AI Summary

This study addresses the challenges of constructing cross-model proposals in reversible jump Markov chain Monte Carlo (RJMCMC) and its incompatibility with GPU parallelization. We propose a novel framework that integrates sequential Monte Carlo (SMC) with RJMCMC. Without requiring domain-specific prior knowledge, our method constructs kernel density estimates from SMC particle distributions to automatically generate efficient trans-dimensional proposal distributions, thereby enabling parallel exploration of the model space. This approach overcomes the serial computation bottleneck inherent in traditional RJMCMC, substantially improving the computational efficiency of Bayesian model comparison. The proposed method achieves performance comparable to existing state-of-the-art techniques, offering an effective pathway for large-scale, scalable Bayesian inference.

Abstract

Reversible Jump Markov Chain Monte Carlo (RJMCMC) is a principled framework for Bayesian model comparison, but its practical use is often limited due to the difficulty of designing between-model proposals that reach regions of high posterior probability. In addition, its inherently sequential nature limits efficient use of modern GPU architectures. We address both challenges by introducing Anchored Reversible Jump Sequential Monte Carlo. Our approach combines RJMCMC with Sequential Monte Carlo (SMC) to explore the model space in parallel. Importantly, the particle population enables effective between-model proposals without problem-specific knowledge: particles are used to construct kernel density approximations of the target distributions within each model, which are then used to generate trans-dimensional proposals. Numerical experiments show that the proposed method is computationally efficient and competitive with state-of-the-art RJMCMC approaches, demonstrating its potential for scalable Bayesian model comparison.

Problem

• Reversible Jump MCMC
• Bayesian model comparison
• between-model proposals
• GPU parallelization
• Sequential Monte Carlo

Innovation

• Reversible Jump MCMC
• Sequential Monte Carlo
• Bayesian model comparison
• Kernel density approximation
• Trans-dimensional proposals