Model management is a critical component in data-driven surrogate-assisted evolutionaryoptimization. It is mainly used for selecting candidate samples to be evaluated using exact functions to balance the exploration ...
详细信息
ISBN:
(纸本)9781728190488
Model management is a critical component in data-driven surrogate-assisted evolutionaryoptimization. It is mainly used for selecting candidate samples to be evaluated using exact functions to balance the exploration and exploitation of algorithms. However, there are some potential limitations for typical infill sampling criteria. An adaptive model management strategy is proposed to overcome the limitations. It adaptively adjusts the selection probability of candidate samples, and avoids potentially unreliability and the burden of selecting parameters of typical infill sampling criteria. The comparison results on benchmark problems demonstrate the competitiveness of the strategy in solving data-drivenoptimization problems.
暂无评论