Conditions for rapid mixing of parallel and simulated tempering on multimodal distributions

Abstract

We give conditions under which a Markov chain constructed via parallel or simulated tempering is guaranteed to be rapidly mixing, which are applicable to a wide range of multimodal distributions arising in Bayesian statistical inference and statistical mechanics. We provide lower bounds on the spectral gaps of parallel and simulated tempering. These bounds imply a single set of sufficient conditions for rapid mixing of both techniques. A direct consequence of our results is rapid mixing of parallel and simulated tempering for several normal mixture models, and for the mean-field Ising model.

arXiv commentPublished in at http://dx.doi.org/10.1214/08-AAP555 the Annals of Applied Probability (http://www.imstat.org/aap/) by the Institute of Mathematical Statistics (http://www.imstat.org)

References in corpus (5)

• Simulated Tempering: A New Monte Carlo Scheme
• General state space Markov chains and MCMC algorithms
• Equi-energy sampler with applications in statistical inference and statistical mechanics
• Conditions for rapid mixing of parallel and simulated tempering on multimodal distributions
• The incomplete beta function law for parallel tempering sampling of classical canonical systems

Cited by in corpus (32)

• Enhanced sampling methods for molecular dynamics simulations
• Replica exchange and expanded ensemble simulations as Gibbs sampling: Simple improvements for enhanced mixing
• Conditions for rapid mixing of parallel and simulated tempering on multimodal distributions
• Non-Reversible Parallel Tempering: a Scalable Highly Parallel MCMC Scheme
• Complexity Results for MCMC derived from Quantitative Bounds
• Convergence rate of Markov chain methods for genomic motif discovery
• Generalized Parallel Tempering on Bayesian Inverse Problems
• Accelerating Parallel Tempering: Quantile Tempering Algorithm (QuanTA)
• Simulated Tempering Langevin Monte Carlo II: An Improved Proof using Soft Markov Chain Decomposition
• Estimating Convergence of Markov chains with L-Lag Couplings
• Non-asymptotic Error Bounds for Sequential MCMC Methods in Multimodal Settings
• Adaptive parallel tempering algorithm
• Beyond Log-concavity: Provable Guarantees for Sampling Multi-modal Distributions using Simulated Tempering Langevin Monte Carlo
• Simulated Tempering and Swapping on Mean-Field Models
• Complexity Bounds for MCMC via Diffusion Limits
• An approach to large-scale Quasi-Bayesian inference with spike-and-slab priors
• Non-asymptotic confidence intervals for MCMC in practice
• Simple Conditions for Metastability of Continuous Markov Chains
• Optimal Scaling of Random-Walk Metropolis Algorithms on General Target Distributions
• Adaptive Path Sampling in Metastable Posterior Distributions
• Parallel Tempering on Optimized Paths
• Convergence Speed and Approximation Accuracy of Numerical MCMC
• Weight-Preserving Simulated Tempering
• Efficient Multimodal Sampling via Tempered Distribution Flow
• Approximate spectral gaps for Markov chains mixing times in high dimensions
• Optimal Temperature Spacing for Regionally Weight-preserving Tempering
• Parallel Markov Chain Monte Carlo via Spectral Clustering
• Mixing times for the Swapping Algorithm on the Blume-Emery-Griffiths Model
• Sampling metastable systems using collective variables and Jarzynski-Crooks paths
• Survival modelling of smartphone trigger data for earthquake parameter estimation in early warning. With applications to 2023 Turkish-Syrian and 2019 Ridgecrest events
• On the Infinite Swapping Limit for Parallel Tempering
• Improved annealing for sampling from multimodal distributions via landscape modification