Abstract:Many Markov chain Monte Carlo techniques currently available rely on discrete-time reversible Markov processes whose transition kernels are variations of the Metropolis–Hastings algorithm. We explore and generalize an alternative scheme recently introduced in the physics literature (Peters and de With 2012) where the target distribution is explored using a continuous-time nonreversible piecewise-deterministic Markov process. In the Metropolis–Hastings algorithm, a trial move to a region of lower target density, equivalently of higher “energy,” than the current state can be rejected with positive probability. In this alternative approach, a particle moves along straight lines around the space and, when facing a high energy barrier, it is not rejected but its path is modified by bouncing against this barrier. By reformulating this algorithm using inhomogeneous Poisson processes, we exploit standard sampling techniques to simulate exactly this Markov process in a wide range of scenarios of interest. Additionally, when the target distribution is given by a product of factors dependent only on subsets of the state variables, such as the posterior distribution associated with a probabilistic graphical model, this method can be modified to take advantage of this structure by allowing computationally cheaper “local” bounces, which only involve the state variables associated with a factor, while the other state variables keep on evolving. In this context, by leveraging techniques from chemical kinetics, we propose several computationally efficient implementations. Experimentally, this new class of Markov chain Monte Carlo schemes compares favorably to state-of-the-art methods on various Bayesian inference tasks, including for high-dimensional models and large datasets. Supplementary materials for this article are available online.

Irreversible Samplers from Jump and Continuous Markov Processes.

Efficient Irreversible Monte Carlo Samplers

Nonreversible Langevin Samplers: Splitting Schemes, Analysis and Implementation

The Bouncy Particle Sampler: A Nonreversible Rejection-Free Markov Chain Monte Carlo Method

Reversible Jump Metropolis Light Transport Using Inverse Mappings

Informed reversible jump algorithms

A generalized Multiple-try Metropolis version of the Reversible Jump algorithm

A note on the Metropolis-Hastings acceptance probabilities for mixture spaces

Reversible jump Markov chain Monte Carlo and multi-model samplers

The transcoding sampler for stick-breaking inferences on Dirichlet process mixtures

Importance sampling for jump processes and applications to finance

Stochastic Gradient Piecewise Deterministic Monte Carlo Samplers

Infinite dimensional Piecewise Deterministic Markov Processes

Piecewise deterministic sampling with splitting schemes

Bouncy Hybrid Sampler as a Unifying Device

The Coordinate Sampler: A Non-Reversible Gibbs-like MCMC Sampler

Schr{ö}dinger-F{ö}llmer Sampler: Sampling without Ergodicity

Metropolis-adjusted interacting particle sampling

Metropolis-Hastings algorithms with autoregressive proposals, and a few examples

Exact Langevin Dynamics with Stochastic Gradients

General Construction of Irreversible Kernel in Markov Chain Monte Carlo