An Inverse Distance Weighting and Radial BasisFunction Based Particle Swarm Optimizer forHigh-dimensional Expensive Optimization

Surrogate-assisted evolutionary algorithms is one of the effective solutions tohigh-dimensional expensive optimization problems. It builds a surrogate modelto reduce the computational cost of expensive problems. However, the quality ofmodel management will directly affect the optimization performance of the algorithm. Inverse distance weighting and radial basis function (IDW-RBF) modelis constructed with low time cost and provides relatively estimated fitness valueand estimated uncertainty respectively. In order to reduce the calculation cost,this paper adopts a social learning particle swarm optimization algorithm namedIR-SLPSO which based on IDW-RBF model. In addition, IR-SLPSO proposes asimplified multiobjective infill criterion (SMIC), selects the last layer and the firstlayer of two prominent particles into the expensive real function operation in theprocess of nondominated sorting. We simulate IR-SLPSO with the existing fourSAEAs on the six commonly used benchmark functions from 50 to 100 dimension.Experimental results show that IR-SLPSO is competitive on high-dimensionalexpensive optimization problems.