Constructing sampling schemes via coupling: Markov semigroups and optimal transport

DOI10.1137/18M119896XzbMATH Open1454.60118OpenAlexW2963521170WikidataQ128175677 ScholiaQ128175677MaRDI QIDQ5228360FDOQ5228360

Authors: Nikolas Nüsken, G. A. Pavliotis 

Publication date: 12 August 2019

Published in: SIAM/ASA Journal on Uncertainty Quantification (Search for Journal in Brave)

Abstract: In this paper we develop a general framework for constructing and analysing coupled Markov chain Monte Carlo samplers, allowing for both (possibly degenerate) diffusion and piecewise deterministic Markov processes. For many performance criteria of interest, including the asymptotic variance, the task of finding efficient couplings can be phrased in terms of problems related to optimal transport theory. We investigate general structural properties, proving a singularity theorem that has both geometric and probabilistic interpretations. Moreover, we show that those problems can often be solved approximately and support our findings with numerical experiments. For the particular objective of estimating the variance of a Bayesian posterior, our analysis suggests using novel techniques in the spirit of antithetic variates. Addressing the convergence to equilibrium of coupled processes we furthermore derive a modified Poincar'e inequality.

Full work available at URL: https://arxiv.org/abs/1806.11026

Recommendations

• Coupling control variates for Markov chain Monte Carlo
• Transport map accelerated Markov chain Monte Carlo
• Coupling and Ergodicity of Adaptive Markov Chain Monte Carlo Algorithms
• Unbiased Hamiltonian Monte Carlo with couplings
• Antithetic coupling of two Gibbs sampler chains.

zbMATH Keywords

samplingMCMCoptimal transportMarkov semigroupsparticle methods

Mathematics Subject Classification ID

Monte Carlo methods (65C05) Markov semigroups and applications to diffusion processes (47D07) Transition functions, generators and resolvents (60J35) Stochastic particle methods (65C35) PDEs in connection with statistics (35Q62)

Cites Work

• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• Title not available (Why is that?)
• A Coupling-Regeneration Scheme for Diagnosing Convergence in Markov Chain Monte Carlo Algorithms
• A Liapounov bound for solutions of the Poisson equation
• A consensus-based model for global optimization and its mean-field limit
• Accelerating diffusions
• Analysis and geometry of Markov diffusion operators
• Analytical methods for Markov semigroups.
• Antithetic coupling of two Gibbs sampler chains.
• Computation of sensitivities for the invariant measure of a parameter dependent diffusion
• Conservativeness and extensions of Feller semigroups
• Control Variates for Estimation Based on Reversible Markov Chain Monte Carlo Samplers
• Coupling and convergence for Hamiltonian Monte Carlo
• Coupling control variates for Markov chain Monte Carlo
• Density functional theory and optimal transportation with Coulomb cost
• Eigenvalues, Inequalities, and Ergodic Theory
• Ensemble preconditioning for Markov chain Monte Carlo simulation
• Ergodicity for Infinite Dimensional Systems
• Ergodicity of the zigzag process
• Exact estimation for Markov chain equilibrium expectations
• Feller semigroups and invariant measures
• Fluctuations in Markov processes. Time symmetry and martingale approximation.
• Free energy computations. A mathematical perspective
• Gibbs Flow for Approximate Transport with Applications to Bayesian Computation
• Handbook of Monte Carlo Methods
• Handbook of uncertainty quantification. In 2 volumes
• Improving the convergence of reversible samplers
• Introduction to uncertainty quantification
• Invariant measure of duplicated diffusions and application to Richardson-Romberg extrapolation
• Limit theorems for the zig-zag process
• Lévy matters III. Lévy-type processes: construction, approximation and sample path properties
• Lévy matters VI. Lévy-type processes: moments, construction and heat kernel estimates
• MCMC using Hamiltonian dynamics
• Markov chain Monte Carlo and irreversibility
• Mean field simulation for Monte Carlo integration
• Multiprocess parallel antithetic coupling for backward and forward Markov chain Monte Carlo
• On tensor products of Banach spaces
• On the Poisson equation for Metropolis-Hastings chains
• On the functional central limit theorem and the law of the iterated logarithm for Markov processes
• One-Parameter Semigroups for Linear Evolution Equations
• One-parameter semigroups of positive operators
• Optimal Transport
• Parallel hierarchical sampling: a general-purpose interacting Markov chains Monte Carlo algorithm
• Partial differential equations and stochastic methods in molecular dynamics
• Particle Markov chain Monte Carlo for efficient numerical simulation
• Piecewise deterministic Markov processes and their invariant measures
• Piecewise deterministic Markov processes for continuous-time Monte Carlo
• Probabilistic Forecasting and Bayesian Data Assimilation
• Pseudo differential operators and Markov processes. In 3 vol. Vol. 1: Fourier analysis and semigroups
• Randomized Hamiltonian Monte Carlo
• Rectifiability of optimal transportation plans
• Recurrence and invariant measures for degenerate diffusions
• Reflection couplings and contraction rates for diffusions
• Selected papers on probability and statistics
• Spectral methods for Langevin dynamics and associated error estimates
• Stochastic differential equations. An introduction with applications.
• Stochastic processes and applications. Diffusion processes, the Fokker-Planck and Langevin equations
• Studying Convergence of Markov Chain Monte Carlo Algorithms Using Coupled Sample Paths
• The Bouncy Particle Sampler: A Non-Reversible Rejection-Free Markov Chain Monte Carlo Method
• The computation of averages from equilibrium and nonequilibrium Langevin molecular dynamics
• The zig-zag process and super-efficient sampling for Bayesian analysis of big data
• Unbiased Hamiltonian Monte Carlo with couplings
• Using perturbed underdamped Langevin dynamics to efficiently sample from probability distributions
• Variance reduction using nonreversible Langevin samplers
• Yet Another Look at Harris’ Ergodic Theorem for Markov Chains

Cited In (5)

• Metropolis-Hastings transition kernel couplings
• Coupling control variates for Markov chain Monte Carlo
• Nonasymptotic bounds for sampling algorithms without log-concavity
• A New Method for Coupling Random Fields
• Transport map accelerated Markov chain Monte Carlo

Uses Software

• EnsembleQN

This page was built for publication: Constructing sampling schemes via coupling: Markov semigroups and optimal transport

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q5228360)