Concurrent implementation techniques using differential evolution for multi-core CPUs: A comparative study using statistical tests

Abstract

In order to utilize multi-core CPUs effectively, a concurrent version of a recently developed evolutionary algorithm, i.e., Differential Evolution (DE), is described. The concurrent version of DE is called Concurrent DE (CDE). CDE is designed based on a programming model known as “MapReduce” and implemented in Java. Two implementations of CDE, namely CDE/D and CDE/S, are proposed and compared from the viewpoint of both quality of solutions and execution time. Through the numerical experiments and the statistical tests conducted on two kinds of popular multi-core CPUs, it is shown that CDE/S uses multi-core CPUs more effectively than CDE/D. However, the quality of solutions obtained by CDE/S tends to fluctuate with the number of threads and the kind of benchmark problems.