High-dimensional real-parameter optimization using Self-Adaptive Differential Evolution algorithm with population size reduction

In this paper we investigate a Self-Adaptive Differential Evolution algorithm (jDEdynNP-F) where F and CR control parameters are self-adapted and a population size reduction method is used. Additionally the proposed jDEdynNPF algorithm uses a mechanism for sign changing of F control parameter with some probability based on the fitness values of randomly chosen vectors, which are multiplied by the F control parameter (scaling factor) in the mutation operation of DE algorithm. The performance of the jDEdynNP-F algorithm is evaluated on the set of 7 benchmark functions provided for the CEC’2008 special session on high-dimensional real-parameter optimization.