Differential evolution with a relational neighbourhood-based strategy for numerical optimization

Abstract

Differential Evolution is a competitive optimizer, with a simplified framework, for numerical optimization problems. Many research works have been done to enhance the performance of Differential Evolution by developing the evolutionary operators. One of the major challenges in DE is performing intelligent search based on population topology. To maintain the diversity in the population as well as to improve the convergence rate, we have introduced a mutation strategy based on relative mapping of the members in population topology. Also Gamma and Cauchy distribution have been adapted in the control parameter framework to include randomness and thorough search. The proposed DE framework is referred to as the Relational Neighbourhood Differential Evolution (ReNbd-DE) and its performance is reported on the set of CEC2005 benchmark functions. © 2012 Springer-Verlag.