Data-driven evolutionary sampling optimization forexpensive problems

Surrogate models have shown to be effective in assisting evolutionary algorithms (EAs) for solving computationally expensive complex optimization problems. However, the effectiveness of existing surrogate-assisted still needs improved. A data-driven sampling (DESO) framework is proposed, where at each generation it randomly employs one two strategies, surrogate screening and local search based on historical data, effectively balance global search. In DESO, radial basis function (RBF) used as model strategy, different degrees process are sample candidate points. The sampled points by strategies evaluated, then added into database updating population next sampling. To get insight extensive experiments analysis DESO been performed. proposed algorithm presents superior computational efficiency robustness compared with five state-of-the-art benchmark problems from 20 200 dimensions. Besides, applied an airfoil design problem show its effectiveness.