Comparison of High-Dimensional Bayesian Optimization Algorithms on BBOB
ACM Transactions on Evolutionary Learning and Optimization · 30.09.2024 · 4 (3) · 1-33
DOI: 10.1145/3670683 ↗Kurz gefasst
Fachartikel in der Zeitschrift ACM Transactions on Evolutionary Learning and Optimization, online erschienen am 29.07.2024. Verfasst von Maria Laura Santoni und 3 weiteren Autor:innen. Die Arbeit wurde bisher 35-mal zitiert (Crossref) und nennt 64 Literaturangaben. Gefördert von: Sorbonne Center for Artificial Intelligence (SCAI) of Sorbonne University, ANR T-ERC project VARIATION, PRIME programme of the German Academic Exchange Service (DAAD) with funds from the German Federal Ministry of Education and Research.
Abstract (Originalsprache)
Bayesian Optimization (BO) is a class of surrogate-based black-box optimization heuristics designed to efficiently locate high-quality solutions for problems that are expensive to evaluate, and therefore allow only small evaluation budgets. BO is particularly popular for solving numerical optimization problems in industry, where the evaluation of objective functions often relies on time-consuming simulations or physical experiments. However, many industrial problems depend on a large number of parameters. This poses a challenge for BO algorithms, whose performance is often reported to suffer when the dimension grows beyond 15 decision variables. Although many new algorithms have been proposed to address this, it remains unclear which one is best suited for a specific optimization problem. In this work, we compare five state-of-the-art high-dimensional BO algorithms with vanilla BO, CMA-ES, and random search on the 24 BBOB functions of the COCO environment at increasing dimensionality, ranging from 10 to 60 variables. Our results confirm the superiority of BO over CMA-ES for limited evaluation budgets and suggest that the most promising approach to improve BO is the use of trust regions. However, we also observe significant performance differences for different function landscapes and budget exploitation phases, indicating improvement potential, e.g., through hybridization of algorithmic components.
Förderung
- Sorbonne Center for Artificial Intelligence (SCAI) of Sorbonne University
- ANR T-ERC project VARIATION
- PRIME programme of the German Academic Exchange Service (DAAD) with funds from the German Federal Ministry of Education and Research
- European Union–NextGenerationEU under the Italian Ministry of University and Research (MUR) National Innovation Ecosystem
- INdAM–GNCS Project 2023
Zitieren
APA
Santoni, M. L., Raponi, E., Leone, R. D., & Doerr, C. (2024). Comparison of High-Dimensional Bayesian Optimization Algorithms on BBOB. ACM Transactions on Evolutionary Learning and Optimization, 4(3), 1-33. https://doi.org/10.1145/3670683
Vancouver
Santoni ML, Raponi E, Leone RD, Doerr C. Comparison of High-Dimensional Bayesian Optimization Algorithms on BBOB. ACM Transactions on Evolutionary Learning and Optimization. 2024;4(3):1-33. doi:10.1145/3670683
BibTeX
@article{Santoni2024,
author = {Santoni, Maria Laura and Raponi, Elena and Leone, Renato De and Doerr, Carola},
title = {Comparison of High-Dimensional Bayesian Optimization Algorithms on BBOB},
journal = {ACM Transactions on Evolutionary Learning and Optimization},
year = {2024},
volume = {4},
number = {3},
pages = {1--33},
doi = {10.1145/3670683}
}
EUVIMED