A Fast Multi-Point Expected Improvement for Parallel Expensive Optimization

DOI number:10.1109/TEVC.2022.3168060

Affiliation of Author(s):School of Computing and Artificial Intelligence, Southwest Jiaotong University

Journal:IEEE Transactions on Evolutionary Computation

Key Words:Multi-point expected improvement,efficient global optimization,Kriging model,parallel computing,expensive optimization

Abstract:The multi-point expected improvement criterion is a well-defined parallel infill criterion for expensive optimization. However, the exact calculation of the classical multi-point expected improvement involves evaluating a significant amount of multivariate normal cumulative distribution functions, which makes the inner optimization of this infill criterion very time-consuming when the number of infill samples is large. To tackle this problem, we propose a novel fast multi-point expected improvement criterion in this work. The proposed infill criterion is calculated using only univariate normal cumulative distributions, thus is easier to implement and cheaper to compute than the classical multi-point expected improvement criterion. It is shown that the computational time of the proposed fast multi-point expected improvement is several orders lower than the classical multi-point expected improvement on the benchmark problems. In addition, we propose to use cooperative coevolutionary algorithms to solve the inner optimization problem of the proposed fast multi-point expected improvement by decomposing the optimization problem into multiple sub-problems with each sub-problem corresponding to one infill sample and solving these sub-problems cooperatively. Numerical experiments show that using cooperative coevolutionary algorithms can improve the performance of the proposed algorithm significantly compared with using standard evolutionary algorithms. This work provides a fast and efficient approach for parallel expensive optimization.

Co-author:Dawei Zhan,Yun Meng,Huanlai Xing

Document Code:10.1109/TEVC.2022.3168060

Page Number:1-1

ISSN No.:1089-778X

Translation or Not:no

Date of Publication:2022-04-18

Included Journals:SCI