Granularity-based surrogate-assisted particle swarm optimization for high-dimensional expensive optimization

Surrogate-assisted meta-heuristic algorithms have won more and more attention for solving computationally expensive problems over past decades. However, most existing surrogate-assisted meta-heuristic algorithms either require thousands of expensive exact function evaluations to obtain acceptable solutions, or focus on solving only low-dimensional expensive optimization problems. In this paper, we attempt to propose a new method to solve high-dimensional expensive optimization problems, in which the population will firstly be granulated into two subsets, i.e., coarse-grained individuals and fine-grained ones, then different approximation methods are proposed for each category, and finally a new infill criteria is adopted to select solutions that have maximum uncertainty among all coarse-grained individuals and that among all fine-grained individuals, and the solution that has minimal approximated fitness value, to be re-evaluated using the exact objective function. Experimental results comparing the proposed algorithm with a few state-of-the-art surrogate-assisted evolutionary algorithms on benchmark problems with 50 and 100 dimensions show that the proposed algorithm is able to achieve better results when solving high-dimensional multi-modal expensive problems with a limited budget on exact fitness evaluations. (C) 2019 Elsevier B.V. All rights reserved.