A novel surrogate-assisted evolutionary algorithm with an uncertainty grouping based infill criterion

For tackling expensive optimization problems (EOPs), surrogate-assisted evolutionary algorithms (SAEAs) will run the search and then select some promising solutions to be evaluated as predicted by surrogate models. Different model management criteria for models, such improvement probability, expected improvement, lower confidence bound, have shown their effectiveness when solving EOPs. In this paper, a novel SAEA with an uncertainty grouping based infill criterion, called SAEA-UGC, is proposed, in which treated indicator training The selected are adopted train ensemble radial basis function model, global local respectively. After obtaining value all search, they evenly grouped according best solution from each group form new population. cooperative find optimal target EOP, i.e., or continually if improved EOP can found iteration; otherwise, switch between happen. performance of SAEA-UGC validated tacking 20 widely used test various properties. experimental results confirm superiority over four representative SAEAs majority