Computer Science · Computational Theory and Mathematics
Comparison of parallel infill sampling criteria for optimization problems
Tangtao Feng, Cong Chen, Ping Song, Yongquan Li, Liao Yi, Cong Wang
See on the mapAbstract
The efficient global optimization (EGO) algorithm based on the Kriging surrogate model is popular. One of the main problems is to determine the infill sampling criteria. An adaptive distance function is proposed and applied to the EI, PI, and MP criteria. The criteria base on a fix distance is also investigated. Seven test problems were used to evaluate these criteria. The results show that the MMP criterion has poor global search ability. PEI, KB and DMP criteria perform better and have better convergence speed and stability.