A multi-model assisted differential evolution algorithm for computationally expensive optimization problems

Abstract Surrogate models are commonly used to reduce the number of required expensive fitness evaluations in optimizing computationally problems. Although many competitive surrogate-assisted evolutionary algorithms have been proposed, it remains a challenging issue develop an effective model management strategy address problems with different landscape features under limited computational budget. This paper adopts coarse-to-fine evaluation scheme basing on two surrogate models, i.e., coarse Gaussian process and fine radial basis function, for assisting differential evolution algorithm solve optimization The is meant capture general contour estimate its degree uncertainty. A environmental selection then developed according non-dominance relationship between approximated estimated Meanwhile, function aims learn details local refine approximation quality new parent population find optima real-evaluations. performance scalability proposed method extensively evaluated sets widely benchmark Experimental results show that can outperform several state-of-the-art within