Home > Research > Publications & Outputs > Reversible Jump PDMP Samplers for Variable Sele...

Electronic data

  • RJ_PDMP_Samplers

    Accepted author manuscript, 7.55 MB, PDF document

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

  • 2010.11771v1

    Other version, 815 KB, PDF document

Links

Text available via DOI:

Keywords

View graph of relations

Reversible Jump PDMP Samplers for Variable Selection

Research output: Contribution to Journal/MagazineJournal articlepeer-review

Published

Standard

Reversible Jump PDMP Samplers for Variable Selection. / Chevallier, Augustin; Fearnhead, Paul; Sutton, Matthew.
In: Journal of the American Statistical Association, Vol. 118, No. 544, 31.07.2023, p. 2915-2927.

Research output: Contribution to Journal/MagazineJournal articlepeer-review

Harvard

Chevallier, A, Fearnhead, P & Sutton, M 2023, 'Reversible Jump PDMP Samplers for Variable Selection', Journal of the American Statistical Association, vol. 118, no. 544, pp. 2915-2927. https://doi.org/10.1080/01621459.2022.2099402

APA

Vancouver

Chevallier A, Fearnhead P, Sutton M. Reversible Jump PDMP Samplers for Variable Selection. Journal of the American Statistical Association. 2023 Jul 31;118(544):2915-2927. Epub 2022 Jul 11. doi: 10.1080/01621459.2022.2099402

Author

Chevallier, Augustin ; Fearnhead, Paul ; Sutton, Matthew. / Reversible Jump PDMP Samplers for Variable Selection. In: Journal of the American Statistical Association. 2023 ; Vol. 118, No. 544. pp. 2915-2927.

Bibtex

@article{f7d25767ddbc43ef8520df09a5be12c2,
title = "Reversible Jump PDMP Samplers for Variable Selection",
abstract = " A new class of Markov chain Monte Carlo (MCMC) algorithms, based on simulating piecewise deterministic Markov processes (PDMPs), have recently shown great promise: they are non-reversible, can mix better than standard MCMC algorithms, and can use subsampling ideas to speed up computation in big data scenarios. However, current PDMP samplers can only sample from posterior densities that are differentiable almost everywhere, which precludes their use for model choice. Motivated by variable selection problems, we show how to develop reversible jump PDMP samplers that can jointly explore the discrete space of models and the continuous space of parameters. Our framework is general: it takes any existing PDMP sampler, and adds two types of trans-dimensional moves that allow for the addition or removal of a variable from the model. We show how the rates of these trans-dimensional moves can be calculated so that the sampler has the correct invariant distribution. Simulations show that the new samplers can mix better than standard MCMC algorithms. Our empirical results show they are also more efficient than gradient-based samplers that avoid model choice through use of continuous spike-and-slab priors which replace a point mass at zero for each parameter with a density concentrated around zero. ",
keywords = "stat.CO, stat.ML",
author = "Augustin Chevallier and Paul Fearnhead and Matthew Sutton",
note = "Code available from https://github.com/matt-sutton/rjpdmp",
year = "2023",
month = jul,
day = "31",
doi = "10.1080/01621459.2022.2099402",
language = "English",
volume = "118",
pages = "2915--2927",
journal = "Journal of the American Statistical Association",
issn = "0162-1459",
publisher = "Taylor and Francis Ltd.",
number = "544",

}

RIS

TY - JOUR

T1 - Reversible Jump PDMP Samplers for Variable Selection

AU - Chevallier, Augustin

AU - Fearnhead, Paul

AU - Sutton, Matthew

N1 - Code available from https://github.com/matt-sutton/rjpdmp

PY - 2023/7/31

Y1 - 2023/7/31

N2 - A new class of Markov chain Monte Carlo (MCMC) algorithms, based on simulating piecewise deterministic Markov processes (PDMPs), have recently shown great promise: they are non-reversible, can mix better than standard MCMC algorithms, and can use subsampling ideas to speed up computation in big data scenarios. However, current PDMP samplers can only sample from posterior densities that are differentiable almost everywhere, which precludes their use for model choice. Motivated by variable selection problems, we show how to develop reversible jump PDMP samplers that can jointly explore the discrete space of models and the continuous space of parameters. Our framework is general: it takes any existing PDMP sampler, and adds two types of trans-dimensional moves that allow for the addition or removal of a variable from the model. We show how the rates of these trans-dimensional moves can be calculated so that the sampler has the correct invariant distribution. Simulations show that the new samplers can mix better than standard MCMC algorithms. Our empirical results show they are also more efficient than gradient-based samplers that avoid model choice through use of continuous spike-and-slab priors which replace a point mass at zero for each parameter with a density concentrated around zero.

AB - A new class of Markov chain Monte Carlo (MCMC) algorithms, based on simulating piecewise deterministic Markov processes (PDMPs), have recently shown great promise: they are non-reversible, can mix better than standard MCMC algorithms, and can use subsampling ideas to speed up computation in big data scenarios. However, current PDMP samplers can only sample from posterior densities that are differentiable almost everywhere, which precludes their use for model choice. Motivated by variable selection problems, we show how to develop reversible jump PDMP samplers that can jointly explore the discrete space of models and the continuous space of parameters. Our framework is general: it takes any existing PDMP sampler, and adds two types of trans-dimensional moves that allow for the addition or removal of a variable from the model. We show how the rates of these trans-dimensional moves can be calculated so that the sampler has the correct invariant distribution. Simulations show that the new samplers can mix better than standard MCMC algorithms. Our empirical results show they are also more efficient than gradient-based samplers that avoid model choice through use of continuous spike-and-slab priors which replace a point mass at zero for each parameter with a density concentrated around zero.

KW - stat.CO

KW - stat.ML

U2 - 10.1080/01621459.2022.2099402

DO - 10.1080/01621459.2022.2099402

M3 - Journal article

VL - 118

SP - 2915

EP - 2927

JO - Journal of the American Statistical Association

JF - Journal of the American Statistical Association

SN - 0162-1459

IS - 544

ER -