GPU-Accelerated Infill Criterion for Multi-Objective Efficient Global Optimization Algorithm and Its Applications

In this work, a novel multi-objective efficient global optimization (EGO) algorithm, namely GMOEGO, is presented by proposing an approach of available threads’ infill criterion. The work applies the outstanding hypervolume-based expected improvement criterion to enhance Pareto solutions in view accuracy and their distribution on front, values sophisticated hypervolume (HVI) are technically approximated counting Monte Carlo sampling points under modern GPU (graphics processing unit) architecture. As compared with traditional methods, such as slice-based integration, programing complexity present greatly reduced due counting-like simple operations. That is, calculation HVI, which has proven be most time-consuming part many objectives, can light programed implementation. Meanwhile, time consumption massive computing associated Carlo-based HVI approximation (MCHVI) alleviated parallelizing GPU. A set mathematical function cases real engineering airfoil shape problem that appeared literature taken validate proposed approach. All results show that, less time-consuming, up around 13.734 times speedup achieved when appropriate captured.