Memetic differential evolutions using adaptive golden section search and their concurrent implementation techniques

This paper proposes a Golden Section Search (GSS) based adaptive Local Search (LS) for enhancing the performance of two kinds of Differential Evolutions (DEs), namely Synchronous DE and Asynchronous DE. GSS is used to find the minimum between two existing solutions. Therefore, GSS-based LS can be regarded as a Crossover-based LS (XLS). The aim of GSS is not only to improve solutions obtained by DE but also break the stagnation of search. In order to balance between DE and GSS, the frequency and the intensity of GSS are adaptively controlled. Performance comparison between GSS-based LS and an existing XLS is also presented. Furthermore, in order to make the best use of multi-core CPUs, which have been widely used even in personal computers, concurrent implementation techniques of the two DEs coupled with GSS-based LS are proposed.