On the infinite swapping limit for parallel tempering

DOI10.1137/110853145MaRDI QIDQ4903663zbMATH OpenOpenAlexFDO

Authors Paul Dupuis, Yufei Liu, Nuria Plattner, J. Doll

Publication date 24 January 2013

Published in Multiscale Modeling & Simulation (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1110.4984

zbMATH Keywords

empirical measureGibbs measureslarge deviationsMonte Carloparallel temperingrate of convergencerelative entropystochastic differential equations

Mathematics Subject Classification ID

Ordinary differential equations and systems with randomness (34F05) Large deviations (60F10) Random measures (60G57) Stochastic ordinary differential equations (aspects of stochastic analysis) (60H10) Monte Carlo methods (65C05) Numerical solutions to stochastic differential and integral equations (65C30)

Abstract: Parallel tempering, also known as replica exchange sampling, is an important method for simulating complex systems. In this algorithm simulations are conducted in parallel at a series of temperatures, and the key feature of the algorithm is a swap mechanism that exchanges configurations between the parallel simulations at a given rate. The mechanism is designed to allow the low temperature system of interest to escape from deep local energy minima where it might otherwise be trapped, via those swaps with the higher temperature components. In this paper we introduce a performance criteria for such schemes based on large deviation theory, and argue that the rate of convergence is a monotone increasing function of the swap rate. This motivates the study of the limit process as the swap rate goes to infinity. We construct a scheme which is equivalent to this limit in a distributional sense, but which involves no swapping at all. Instead, the effect of the swapping is captured by a collection of weights that influence both the dynamics and the empirical measure. While theoretically optimal, this limit is not computationally feasible when the number of temperatures is large, and so variations that are easy to implement and nearly optimal are also developed.

Recommendations

• Methodological and computational aspects of parallel tempering methods in the infinite swapping limit
• A large deviations analysis of certain qualitative properties of parallel tempering and infinite swapping algorithms
• Analysis and optimization of certain parallel Monte Carlo methods in the low temperature limit
• Ergodicity of the infinite swapping algorithm at low temperature
• State-dependent swap strategies and automatic reduction of number of temperatures in adaptive parallel tempering algorithm

Cited in

(28)

• State-dependent swap strategies and automatic reduction of number of temperatures in adaptive parallel tempering algorithm
• A large deviations analysis of certain qualitative properties of parallel tempering and infinite swapping algorithms
• Methodological and computational aspects of parallel tempering methods in the infinite swapping limit
• Large deviations of empirical measures of diffusions in weighted topologies
• Analysis of stochastic gradient descent in continuous time
• Generalized parallel tempering on Bayesian inverse problems
• Anytime parallel tempering
• Large deviations for the empirical measure of the zig-zag process
• Large deviation properties of the empirical measure of a metastable small noise diffusion
• Spectral gap of replica exchange Langevin diffusion on mixture distributions
• Ergodicity of the infinite swapping algorithm at low temperature
• Improving the convergence of reversible samplers
• Optimal variance reduction for Markov chain Monte Carlo
• Stationary Averaging for Multiscale Continuous Time Markov Chains Using Parallel Replica Dynamics
• The simulated tempering method in the infinite switch limit with adaptive weight learning
• Gradient flows and randomised thresholding: sparse inversion and classification
• Analysis and optimization of certain parallel Monte Carlo methods in the low temperature limit
• Infinite Swapping Algorithm for Training Restricted Boltzmann Machines
• Replica exchange for non-convex optimization
• Overcoming the timescale barrier in molecular dynamics: Transfer operators, variational principles and machine learning
• Massively parallel simulations for disordered systems
• A large deviation principle for the empirical measures of Metropolis-Hastings chains
• The enhanced sampling in parallel finite-time dynamics method with replica exchange
• Monte Carlo methods for rough free energy landscapes: population annealing and parallel tempering
• Infinite swapping using IID samples
• Convergence speed and approximation accuracy of numerical MCMC
• Large deviations for independent Metropolis Hastings and Metropolis-adjusted Langevin algorithm
• Concentration inequalities and UQ bounds for hypocoercive MCMC samplers

This page was built for publication: On the infinite swapping limit for parallel tempering

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4903663)