Markov chain Monte Carlo (MCMC) is a powerful methodology for the approximation of posterior distributions. However, the iterative nature of MCMC does not naturally facilitate its use with modern highly parallel compu...
详细信息
Markov chain Monte Carlo (MCMC) is a powerful methodology for the approximation of posterior distributions. However, the iterative nature of MCMC does not naturally facilitate its use with modern highly parallel computation on HPC and cloud environments. Another concern is the identification of the bias and Monte Carlo error of produced averages. The above have prompted the recent development of fully ('embarrassingly') parallel unbiased Monte Carlo methodology based on coupling of MCMC algorithms. A caveat is that formulation of effective coupling is typically not trivial and requires model-specific technical effort. We propose coupling of MCMC chains deriving from sequential Monte Carlo (SMC) by considering adaptive SMC methods in combination with recent advances in unbiased estimation for state-space models. Coupling is then achieved at the SMC level and is, in principle, not problem-specific. The resulting methodology enjoys desirable theoretical properties. A central motivation is to extend unbiased MCMC to more challenging targets compared to the ones typically considered in the relevant literature. We illustrate the effectiveness of the algorithm via application to two complex statistical models: (i) horseshoe regression;(ii) Gaussian graphical models.
We propose both the first embarrassinglyparallel consensus variational inference algorithm and a new consensus Monte Carlo algorithm for efficient implementation of Bayesian nonparametric mixture models. The proposed...
详细信息
ISBN:
(纸本)9781728162515
We propose both the first embarrassinglyparallel consensus variational inference algorithm and a new consensus Monte Carlo algorithm for efficient implementation of Bayesian nonparametric mixture models. The proposed algorithms are based on a group clustering approach, and they substantially accelerate inference and reduce memory costs compared with standard Markov chain Monte Carlo and variational inference algorithms for clustering. We demonstrate that our proposed algorithms are significantly faster than competing methods while maintaining the same clustering accuracy. Due to their simplicity and embarrassinglyparallel nature, our proposed algorithms are straightforward to implement and widely applicable beyond the models and applications considered in this paper.
暂无评论