A New Acquisition Function for Multi-objective Bayesian Optimization: Correlated Probability of Improvement

Publikation: Beitrag in Buch/Bericht/TagungsbandKonferenzbeitragBegutachtung

Abstract

Multi-objective Bayesian optimization is a sequential optimization strategy in which an optimizer searches for optimal solutions by maximizing an acquisition function. Most existing acquisition functions assume that objectives are independent, but none of them incorporates the correlations among objectives through an explicit formula for exact computation. This paper proposes a novel acquisition function, namely, correlated probability of improvement (cPoI), for bi-objective optimization problems. The cPoI method builds on the probability of improvement and addresses the correlations between objectives by utilizing 3 distinct approaches to compute the posterior covariance matrix from a multi-task Gaussian process. This paper presents both an explicit formula for exact computation of cPoI and a Monte Carlo method for approximating it. We evaluate the performance of the proposed cPoI against 4 state-of-the-art multi-objective optimization algorithms on 8 artificial benchmarks and 1 real-world problem. Our experimental results demonstrate the effectiveness of cPoI in achieving superior optimization performance.

OriginalspracheEnglisch
TitelGECCO 2023 Companion - Proceedings of the 2023 Genetic and Evolutionary Computation Conference Companion
Herausgeber (Verlag)Association for Computing Machinery, Inc
Seiten2308-2317
Seitenumfang10
ISBN (elektronisch)9798400701207
DOIs
PublikationsstatusVeröffentlicht - 15 Juli 2023
Veranstaltung2023 Genetic and Evolutionary Computation Conference Companion, GECCO 2023 Companion - Lisbon, Portugal
Dauer: 15 Juli 202319 Juli 2023

Publikationsreihe

NameGECCO 2023 Companion - Proceedings of the 2023 Genetic and Evolutionary Computation Conference Companion

Konferenz

Konferenz2023 Genetic and Evolutionary Computation Conference Companion, GECCO 2023 Companion
Land/GebietPortugal
OrtLisbon
Zeitraum15.07.202319.07.2023

Fingerprint

Untersuchen Sie die Forschungsthemen von „A New Acquisition Function for Multi-objective Bayesian Optimization: Correlated Probability of Improvement“. Zusammen bilden sie einen einzigartigen Fingerprint.

Zitieren