Translator Disclaimer
December 1996 Exponential convergence of Langevin distributions and their discrete approximations
Gareth O. Roberts, Richard L. Tweedie
Bernoulli 2(4): 341-363 (December 1996).

Abstract

In this paper we consider a continuous-time method of approximating a given distribution π using the Langevin diffusion d L t=dW t+1 2 logπ(L t)dt . We find conditions under which this diffusion converges exponentially quickly to π or does not: in one dimension, these are essentially that for distributions with exponential tails of the form π (x)exp(-γ|x| β ) , 0 <β< , exponential convergence occurs if and only if β 1 . We then consider conditions under which the discrete approximations to the diffusion converge. We first show that even when the diffusion itself converges, naive discretizations need not do so. We then consider a 'Metropolis-adjusted' version of the algorithm, and find conditions under which this also converges at an exponential rate: perhaps surprisingly, even the Metropolized version need not converge exponentially fast even if the diffusion does. We briefly discuss a truncated form of the algorithm which, in practice, should avoid the difficulties of the other forms.

Citation

Download Citation

Gareth O. Roberts. Richard L. Tweedie. "Exponential convergence of Langevin distributions and their discrete approximations." Bernoulli 2 (4) 341 - 363, December 1996.

Information

Published: December 1996
First available in Project Euclid: 4 May 2007

zbMATH: 0870.60027
MathSciNet: MR1440273

Rights: Copyright © 1996 Bernoulli Society for Mathematical Statistics and Probability

JOURNAL ARTICLE
23 PAGES


SHARE
Vol.2 • No. 4 • December 1996
Back to Top