Improving the convergence of reversible samplers

Abstract

In Monte-Carlo methods the Markov processes used to sample a given target distribution usually satisfy detailed balance, i.e. they are time-reversible. However, relatively recent results have demonstrated that appropriate reversible and irreversible perturbations can accelerate convergence to equilibrium. In this paper we present some general design principles which apply to general Markov processes. Working with the generator of Markov processes, we prove that for some of the most commonly used performance criteria, i.e., spectral gap, asymptotic variance and large deviation functionals, sampling is improved for appropriate reversible and irreversible perturbations of some initially given reversible sampler. Moreover we provide specific constructions for such reversible and irreversible perturbations for various commonly used Markov processes, such as Markov chains and diffusions. In the case of diffusions, we make the discussion more specific using the large deviations rate function as a measure of performance.

Final version will appear in the Journal of Statistical Physics

References in corpus (4)

• Markov Chain Monte Carlo Method without Detailed Balance
• Accelerating diffusions
• Irreversible Langevin samplers and variance reduction: a large deviation approach
• Improving the Asymptotic Performance of Markov Chain Monte-Carlo by Inserting Vortices

Cited by in corpus (17)

• Acceleration of convergence to equilibrium in Markov chains by breaking detailed balance
• Limit theorems for the Zig-Zag process
• Role of current fluctuations in nonreversible samplers
• Large deviations for the Skew-Detailed-Balance Lifted-Markov processes to sample the equilibrium distribution of the Curie-Weiss model
• Breaking Reversibility Accelerates Langevin Dynamics for Global Non-Convex Optimization
• Irreversible Samplers from Jump and Continuous Markov Processes
• Peskun-Tierney ordering for Markov chain and process Monte Carlo: beyond the reversible scenario
• Gibbsian stationary non equilibrium states
• Concentration Inequalities and UQ Bounds for Hypocoercive MCMC Samplers
• Lifting Markov Chains To Mix Faster: Limits and Opportunities
• Symmetries and Geometrical Properties of Dynamical Fluctuations in Molecular Dynamics
• Non-Convex Optimization via Non-Reversible Stochastic Gradient Langevin Dynamics
• Constructing sampling schemes via coupling: Markov semigroups and optimal transport
• On decompositions of non-reversible processes
• Non-reversible sampling schemes on submanifolds
• Markov chain entropy games and the geometry of their Nash equilibria
• Improving the convergence of Markov chains via permutations and projections