Integrating surrogate evaluation model and asynchronous evolution in multi-objective evolutionary algorithm for expensive and different evaluation time
Misaki Kaidan, Tomohiro Harada, Ruck Thawonmas · Proceedings of the Genetic and Evolutionary Computation Conference Companion · 2017
This paper proposes Extreme Learning Surrogate assisted Asynchronous Multi-Objective Optimization Based on Decomposition (AELMOEA/D) that solves multi-objective optimization problems with expensive and different evaluation time by integrating a surrogate evaluation model and an asynchronous evolution method. Extreme Learning Surrogate assisted Multi-Objective Optimization Based on Decomposition (ELMOEA/D), which is a surrogate-assisted MOEA/D, was proposed to reduce the number of actual evaluations, while asynchronous evolution methods were proposed to reduce the waiting time for evaluation of solutions in a parallel evolutionary algorithm. This paper employs ELMOEA/D as a surrogate assisted EA and introduces an asynchronous manner into it. Our experiment proves that our proposed AELMOEA/D can obtain optimal solutions faster than ELMOEA/D without performance deterioration.