References & Citations
Statistics > Computation
Title: Unbiased Hamiltonian Monte Carlo with couplings
(Submitted on 1 Sep 2017 (v1), revised 13 Feb 2018 (this version, v2), latest version 27 Aug 2018 (v3))
Abstract: We propose a methodology to parallelize Hamiltonian Monte Carlo estimators. Our approach constructs a pair of Hamiltonian Monte Carlo chains that are coupled in such a way that they meet exactly after some random number of iterations. These chains can then be combined so that resulting estimators are unbiased. This allows us to produce independent replicates in parallel and average them to obtain estimators that are consistent in the limit of the number of replicates, instead of the usual limit of the number of Markov chain iterations. The choice of algorithmic parameters and the efficiency of our proposed methodology are illustrated on a truncated Gaussian distribution, a logistic regression with 300 covariates, and a log-Gaussian Cox point processes model with a finely grained discretization.
Submission history
From: Jeremy Heng [view email][v1] Fri, 1 Sep 2017 17:54:56 GMT (1593kb,D)
[v2] Tue, 13 Feb 2018 22:53:33 GMT (71kb)
[v3] Mon, 27 Aug 2018 16:45:13 GMT (539kb)
Link back to: arXiv, form interface, contact.