Home > Research > Publications & Outputs > Hug and Hop

Electronic data

  • HugHop

    Rights statement: This is a pre-copy-editing, author-produced PDF of an article accepted for publication in British Journal for the Philosophy of Science following peer review. The definitive publisher-authenticated version M Ludkin, C Sherlock, Hug and Hop: a discrete-time, nonreversible Markov chain Monte Carlo algorithm, Biometrika, 2022;, asac039 is available online at: https://academic.oup.com/biomet/advance-article/doi/10.1093/biomet/asac039/6633932

    Accepted author manuscript, 2.22 MB, PDF document

    Embargo ends: 8/07/23

    Available under license: CC BY-NC: Creative Commons Attribution-NonCommercial 4.0 International License

Text available via DOI:

View graph of relations

Hug and Hop: a discrete-time, nonreversible Markov chain Monte Carlo algorithm

Research output: Contribution to Journal/MagazineJournal articlepeer-review

E-pub ahead of print

Standard

Hug and Hop : a discrete-time, nonreversible Markov chain Monte Carlo algorithm. / Ludkin, Matthew; Sherlock, Chris.

In: Biometrika, 08.07.2022.

Research output: Contribution to Journal/MagazineJournal articlepeer-review

Harvard

APA

Vancouver

Ludkin M, Sherlock C. Hug and Hop: a discrete-time, nonreversible Markov chain Monte Carlo algorithm. Biometrika. 2022 Jul 8. Epub 2022 Jul 8. doi: 10.1093/biomet/asac039

Author

Bibtex

@article{b42b132beda046e2925b7ecf0592ba53,
title = "Hug and Hop: a discrete-time, nonreversible Markov chain Monte Carlo algorithm",
abstract = "We introduced the Hug and Hop Markov chain Monte Carlo algorithm for estimating expectations with respect to an intractable distribution. The algorithm alternates between two kernels: Hug and Hop. Hug is a non-reversible kernel that repeatedly applies the bounce mechanism from the recently proposed Bouncy Particle Sampler to produce a proposal point far from the current position, yet on almost the same contour of the target density, leading to a high acceptance probability. Hug is complemented by Hop, which deliberately proposes jumps between contours and has an efficiency that degrades very slowly with increasing dimension. There are many parallels between Hug and Hamiltonian Monte Carlo using a leapfrog integrator, including the order of the integration scheme, however Hug is also able to make use of local Hessian information without requiring implicit numerical integration steps, and its performance is not terminally affected by unbounded gradients of the log-posterior. We test Hug and Hop empirically on a variety of toy targets and real statistical models and find that it can, and often does, outperform Hamiltonian Monte Carlo.",
author = "Matthew Ludkin and Chris Sherlock",
note = "This is a pre-copy-editing, author-produced PDF of an article accepted for publication in British Journal for the Philosophy of Science following peer review. The definitive publisher-authenticated version M Ludkin, C Sherlock, Hug and Hop: a discrete-time, nonreversible Markov chain Monte Carlo algorithm, Biometrika, 2022;, asac039 is available online at: https://academic.oup.com/biomet/advance-article/doi/10.1093/biomet/asac039/6633932",
year = "2022",
month = jul,
day = "8",
doi = "10.1093/biomet/asac039",
language = "English",
journal = "Biometrika",
issn = "0006-3444",
publisher = "Oxford University Press",

}

RIS

TY - JOUR

T1 - Hug and Hop

T2 - a discrete-time, nonreversible Markov chain Monte Carlo algorithm

AU - Ludkin, Matthew

AU - Sherlock, Chris

N1 - This is a pre-copy-editing, author-produced PDF of an article accepted for publication in British Journal for the Philosophy of Science following peer review. The definitive publisher-authenticated version M Ludkin, C Sherlock, Hug and Hop: a discrete-time, nonreversible Markov chain Monte Carlo algorithm, Biometrika, 2022;, asac039 is available online at: https://academic.oup.com/biomet/advance-article/doi/10.1093/biomet/asac039/6633932

PY - 2022/7/8

Y1 - 2022/7/8

N2 - We introduced the Hug and Hop Markov chain Monte Carlo algorithm for estimating expectations with respect to an intractable distribution. The algorithm alternates between two kernels: Hug and Hop. Hug is a non-reversible kernel that repeatedly applies the bounce mechanism from the recently proposed Bouncy Particle Sampler to produce a proposal point far from the current position, yet on almost the same contour of the target density, leading to a high acceptance probability. Hug is complemented by Hop, which deliberately proposes jumps between contours and has an efficiency that degrades very slowly with increasing dimension. There are many parallels between Hug and Hamiltonian Monte Carlo using a leapfrog integrator, including the order of the integration scheme, however Hug is also able to make use of local Hessian information without requiring implicit numerical integration steps, and its performance is not terminally affected by unbounded gradients of the log-posterior. We test Hug and Hop empirically on a variety of toy targets and real statistical models and find that it can, and often does, outperform Hamiltonian Monte Carlo.

AB - We introduced the Hug and Hop Markov chain Monte Carlo algorithm for estimating expectations with respect to an intractable distribution. The algorithm alternates between two kernels: Hug and Hop. Hug is a non-reversible kernel that repeatedly applies the bounce mechanism from the recently proposed Bouncy Particle Sampler to produce a proposal point far from the current position, yet on almost the same contour of the target density, leading to a high acceptance probability. Hug is complemented by Hop, which deliberately proposes jumps between contours and has an efficiency that degrades very slowly with increasing dimension. There are many parallels between Hug and Hamiltonian Monte Carlo using a leapfrog integrator, including the order of the integration scheme, however Hug is also able to make use of local Hessian information without requiring implicit numerical integration steps, and its performance is not terminally affected by unbounded gradients of the log-posterior. We test Hug and Hop empirically on a variety of toy targets and real statistical models and find that it can, and often does, outperform Hamiltonian Monte Carlo.

U2 - 10.1093/biomet/asac039

DO - 10.1093/biomet/asac039

M3 - Journal article

JO - Biometrika

JF - Biometrika

SN - 0006-3444

ER -