Coupling control variates for Markov chain Monte Carlo

JOURNAL OF COMPUTATIONAL PHYSICS

卷:228

Article

Goodman, Jonathan B.¹  Lin, Kevin K.² 

[1] NYU, Courant Inst Math Sci, New York, NY 10012 USA

[2] Univ Arizona, Dept Math, Tucson, AZ 85721 USA

关键词: Monte Carlo;    Control variates;    Variance reduction;    Coupling;    Markov chain Monte Carlo;    Nonequilibrium statistical mechanics;   

DOI  :  10.1016/j.jcp.2009.03.043

来源: Elsevier

【 摘 要 】

We show that Markov couplings can be used to improve the accuracy of Markov chain Monte Carlo calculations in some situations where the steady-state probability distribution is not explicitly known. The technique generalizes the notion of control variates from classical Monte Carlo integration. We illustrate it using two models of nonequilibrium transport. (C) 2009 Elsevier Inc. All rights reserved

【 授权许可】

Free