A Surrogate-based Optimization Algorithm with Local Search

In 5th generation (5G) network, the optimized design of high-performance antenna has been an important and complicated issue. In this paper, to further improve the performance of surrogate-based global optimization algorithms (EGO) for 'black-book' problem, a neighborhood field search strategy is incorporated into the original EGO algorithm to make up for the insufficient local search. The resulted mimetic algorithm is called NFSEGO. In the evolution process, a valid local search is performed near certain promising candidate solutions. The new solution obtained by the local search will replace the current better candidate solution. To validate the proposed algorithm, five well-known benchmark functions, and one antenna optimization design engineering problem are studied. The presented results show that NFSEGO is able to excavate more excellent solution than original algorithm in terms of accuracy.