Efficient acquisition rules for model-based approximate Bayesian computation

Research output: Contribution to journalArticle

Standard

Efficient acquisition rules for model-based approximate Bayesian computation. / Järvenpää, Marko; Gutmann, Michael U.; Pleska, Arijus; Vehtari, Aki; Marttinen, Pekka.

In: Bayesian Analysis, Vol. 14, No. 2, 06.2019, p. 595-622.

Research output: Contribution to journalArticle

Harvard

APA

Vancouver

Author

Bibtex - Download

@article{2030ee3c7c7c4b6f9352d2cf06c695e7,
title = "Efficient acquisition rules for model-based approximate Bayesian computation",
abstract = "Approximate Bayesian computation (ABC) is a method for Bayesian inference when the likelihood is unavailable but simulating from the model is possible. However, many ABC algorithms require a large number of simulations, which can be costly. To reduce the computational cost, Bayesian optimisation (BO) and surrogate models such as Gaussian processes have been proposed. Bayesian optimisation enables one to intelligently decide where to evaluate the model next but common BO strategies are not designed for the goal of estimating the posterior distribution. Our paper addresses this gap in the literature. We propose to compute the uncertainty in the ABC posterior density, which is due to a lack of simulations to estimate this quantity accurately, and define a loss function that measures this uncertainty. We then propose to select the next evaluation location to minimise the expected loss. Experiments show that the proposed method often produces the most accurate approximations as compared to common BO strategies.",
keywords = "Approximate Bayesian computation, Bayesian optimisation, Gaussian processes, Intractable likelihood, Sequential experiment design, MONTE-CARLO, STATISTICAL-INFERENCE, REDUCTION, sequential experiment design, approximate Bayesian computation, intractable likelihood",
author = "Marko J{\"a}rvenp{\"a}{\"a} and Gutmann, {Michael U.} and Arijus Pleska and Aki Vehtari and Pekka Marttinen",
year = "2019",
month = "6",
doi = "10.1214/18-BA1121",
language = "English",
volume = "14",
pages = "595--622",
journal = "Bayesian Analysis",
issn = "1936-0975",
number = "2",

}

RIS - Download

TY - JOUR

T1 - Efficient acquisition rules for model-based approximate Bayesian computation

AU - Järvenpää, Marko

AU - Gutmann, Michael U.

AU - Pleska, Arijus

AU - Vehtari, Aki

AU - Marttinen, Pekka

PY - 2019/6

Y1 - 2019/6

N2 - Approximate Bayesian computation (ABC) is a method for Bayesian inference when the likelihood is unavailable but simulating from the model is possible. However, many ABC algorithms require a large number of simulations, which can be costly. To reduce the computational cost, Bayesian optimisation (BO) and surrogate models such as Gaussian processes have been proposed. Bayesian optimisation enables one to intelligently decide where to evaluate the model next but common BO strategies are not designed for the goal of estimating the posterior distribution. Our paper addresses this gap in the literature. We propose to compute the uncertainty in the ABC posterior density, which is due to a lack of simulations to estimate this quantity accurately, and define a loss function that measures this uncertainty. We then propose to select the next evaluation location to minimise the expected loss. Experiments show that the proposed method often produces the most accurate approximations as compared to common BO strategies.

AB - Approximate Bayesian computation (ABC) is a method for Bayesian inference when the likelihood is unavailable but simulating from the model is possible. However, many ABC algorithms require a large number of simulations, which can be costly. To reduce the computational cost, Bayesian optimisation (BO) and surrogate models such as Gaussian processes have been proposed. Bayesian optimisation enables one to intelligently decide where to evaluate the model next but common BO strategies are not designed for the goal of estimating the posterior distribution. Our paper addresses this gap in the literature. We propose to compute the uncertainty in the ABC posterior density, which is due to a lack of simulations to estimate this quantity accurately, and define a loss function that measures this uncertainty. We then propose to select the next evaluation location to minimise the expected loss. Experiments show that the proposed method often produces the most accurate approximations as compared to common BO strategies.

KW - Approximate Bayesian computation

KW - Bayesian optimisation

KW - Gaussian processes

KW - Intractable likelihood

KW - Sequential experiment design

KW - MONTE-CARLO

KW - STATISTICAL-INFERENCE

KW - REDUCTION

KW - sequential experiment design

KW - approximate Bayesian computation

KW - intractable likelihood

UR - http://www.scopus.com/inward/record.url?scp=85064907418&partnerID=8YFLogxK

U2 - 10.1214/18-BA1121

DO - 10.1214/18-BA1121

M3 - Article

VL - 14

SP - 595

EP - 622

JO - Bayesian Analysis

JF - Bayesian Analysis

SN - 1936-0975

IS - 2

ER -

ID: 33938263