Cluster-centroid-based mutation strategies for Differential Evolution

Mutation is the main operator of differential evolution (DE), as it responsible for combining information distinct solutions to generate a donor vector. Aiming at improving search effectivity DE, previous research incorporated calculation centroids into DE mutations. In some existing methods, are simply calculated center selected (or, entire population); in other cases, one-step clustering used perform local search, or themselves actual population. As opposed these this paper we extend traditional mutation strategies incorporate by deterministic hierarchical clustering. Experimental results on two sets well-known benchmark problems show that proposed cluster-centroid-based outperform, general, rand/1 strategy, well several metaheuristics from literature. Therefore, use an effective way improve performance could be exploited also population-based metaheuristics.