Ensemble samplers with affine invariance

Jonathan Goodman, Jonathan Weare

Research output: Contribution to journalArticle

Abstract

We propose a family of Markov chain Monte Carlo methods whose performance is unaffected by affine tranformations of space. These algorithms are easy to construct and require little or no additional computational overhead. They should be particularly useful for sampling badly scaled distributions. Computational tests show that the affine invariant methods can be significantly faster than standard MCMC methods on highly skewed distributions.

Original languageEnglish (US)
Pages (from-to)65-80
Number of pages16
JournalCommunications in Applied Mathematics and Computational Science
Volume5
Issue number1
DOIs
StatePublished - 2010

Fingerprint

Affine Invariance
Invariance
Markov processes
Ensemble
Monte Carlo methods
MCMC Methods
Sampling
Skewed Distribution
Affine Invariant
Markov Chain Monte Carlo Methods

Keywords

  • Affine invariance
  • Ensemble samplers
  • Markov chain Monte Carlo

ASJC Scopus subject areas

  • Computer Science Applications
  • Computational Theory and Mathematics
  • Applied Mathematics

Cite this

Ensemble samplers with affine invariance. / Goodman, Jonathan; Weare, Jonathan.

In: Communications in Applied Mathematics and Computational Science, Vol. 5, No. 1, 2010, p. 65-80.

Research output: Contribution to journalArticle

Goodman, Jonathan ; Weare, Jonathan. / Ensemble samplers with affine invariance. In: Communications in Applied Mathematics and Computational Science. 2010 ; Vol. 5, No. 1. pp. 65-80.
@article{64ebe435829647f594fea6d945067389,
title = "Ensemble samplers with affine invariance",
abstract = "We propose a family of Markov chain Monte Carlo methods whose performance is unaffected by affine tranformations of space. These algorithms are easy to construct and require little or no additional computational overhead. They should be particularly useful for sampling badly scaled distributions. Computational tests show that the affine invariant methods can be significantly faster than standard MCMC methods on highly skewed distributions.",
keywords = "Affine invariance, Ensemble samplers, Markov chain Monte Carlo",
author = "Jonathan Goodman and Jonathan Weare",
year = "2010",
doi = "10.2140/camcos.2010.5.65",
language = "English (US)",
volume = "5",
pages = "65--80",
journal = "Communications in Applied Mathematics and Computational Science",
issn = "1559-3940",
publisher = "Mathematical Sciences Publishers",
number = "1",

}

TY - JOUR

T1 - Ensemble samplers with affine invariance

AU - Goodman, Jonathan

AU - Weare, Jonathan

PY - 2010

Y1 - 2010

N2 - We propose a family of Markov chain Monte Carlo methods whose performance is unaffected by affine tranformations of space. These algorithms are easy to construct and require little or no additional computational overhead. They should be particularly useful for sampling badly scaled distributions. Computational tests show that the affine invariant methods can be significantly faster than standard MCMC methods on highly skewed distributions.

AB - We propose a family of Markov chain Monte Carlo methods whose performance is unaffected by affine tranformations of space. These algorithms are easy to construct and require little or no additional computational overhead. They should be particularly useful for sampling badly scaled distributions. Computational tests show that the affine invariant methods can be significantly faster than standard MCMC methods on highly skewed distributions.

KW - Affine invariance

KW - Ensemble samplers

KW - Markov chain Monte Carlo

UR - http://www.scopus.com/inward/record.url?scp=85017381725&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85017381725&partnerID=8YFLogxK

U2 - 10.2140/camcos.2010.5.65

DO - 10.2140/camcos.2010.5.65

M3 - Article

AN - SCOPUS:85017381725

VL - 5

SP - 65

EP - 80

JO - Communications in Applied Mathematics and Computational Science

JF - Communications in Applied Mathematics and Computational Science

SN - 1559-3940

IS - 1

ER -