Variance reduction for additive functionals of Markov chains via martingale representations

DOI10.1007/S11222-021-10073-ZzbMATH Open1482.62003arXiv1903.07373OpenAlexW4210453183MaRDI QIDQ2114045FDOQ2114045

D. Belomestny, Eric Moulines, S. P. Samsonov

Publication date: 14 March 2022

Published in: Statistics and Computing (Search for Journal in Brave)

Abstract: In this paper we propose an efficient variance reduction approach for additive functionals of Markov chains relying on a novel discrete time martingale representation. Our approach is fully non-asymptotic and does not require the knowledge of the stationary distribution (and even any type of ergodicity) or specific structure of the underlying density. By rigorously analyzing the convergence properties of the proposed algorithm, we show that its cost-to-variance product is indeed smaller than one of the naive algorithm. The numerical performance of the new method is illustrated for the Langevin-type Markov Chain Monte Carlo (MCMC) methods.

Full work available at URL: https://arxiv.org/abs/1903.07373

Recommendations

• Variance reduction for Markov chains with application to MCMC
• Variance reduction for dependent sequences with applications to stochastic gradient MCMC
• Variance reduction for Metropolis-Hastings samplers
• Variance reduction for discretised diffusions via regression
• Variance reduction for diffusions

zbMATH Keywords

MCMCvariance reductionMarkov chainsmartingale representations

Mathematics Subject Classification ID

Computational methods for problems pertaining to statistics (62-08)

Cites Work

• Zero variance Markov chain Monte Carlo for Bayesian estimators
• A Multivariate Faa di Bruno Formula with Applications
• A distribution-free theory of nonparametric regression
• Rates of convergence of the Hastings and Metropolis algorithms
• Concentration inequalities. A nonasymptotic theory of independence
• Title not available (Why is that?)
• Monte Carlo complexity of parametric integration
• Asymptotic behaviour of the stochastic Lotka-Volterra model.
• Control Functionals for Monte Carlo Integration
• Ergodicity for SDEs and approximations: locally Lipschitz vector fields and degenerate noise.
• Control Variates for Estimation Based on Reversible Markov Chain Monte Carlo Samplers
• Monte Carlo Methods for Applied Scientists
• Markov Chains
• An adaptive scheme for the approximation of dissipative systems
• Recursive computation of the invariant distribution of a diffusion
• Preliminary control variates to improve empirical regression methods
• Nonasymptotic convergence analysis for the unadjusted Langevin algorithm
• Theoretical Guarantees for Approximate Sampling from Smooth and Log-Concave Densities
• Title not available (Why is that?)
• Dynamic analysis of stochastic Lotka-Volterra predator-prey model with discrete delays and feedback control
• Monte-Carlo methods and stochastic processes. From linear to non-linear
• Variance reduction for Markov chains with application to MCMC
• Weighted multilevel Langevin simulation of invariant measures
• Simulation and the Monte Carlo Method
• Variance reduction for discretised diffusions via regression
• Adaptive simulation using perfect control variates

Cited In (1)

• Variational principles for asymptotic variance of general Markov processes

This page was built for publication: Variance reduction for additive functionals of Markov chains via martingale representations

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2114045)