Opposition-based adaptive differential evolution

Differential evolution (DE) is a simple and efficient evolutionary algorithm. It contains three parameters which need to be predefined by users. These parameters are sensitive to specific problems and difficult to set. Opposition-based computing (OBC) is a new scheme for computational intelligence. OBC is helpful to existing techniques by making better decisions through simultaneous consideration of entities and opposite entities. The opposition phenomenon exists in the literature concerning parameter control of DE. In this paper, OBC is employed to assist with the solving of parameter control problem in DE. Employing OBC to parameter control problem in DE has not been reported previously to our knowledge. The proposed approach is called opposition-based adaptive DE (OADE). It uses two pools to respectively store parameters and opposite parameters. The parameters and their opposites are used at the same time to generate trial vectors in DE. During the evolutionary process, fitness improvement at a generation serves as a filter to detect proper parameters for optimization problems. The detected proper parameters and their opposites are stored in pools, whereas the improper parameters and their opposites are replaced by new randomly generated ones. The utilization of parameters and their opposites can balance the exploration and exploitation behavior of DE in one generation. The performance of OADE is compared with three other DE algorithms. The experimental results show that OADE significantly outperforms the benchmark algorithms. Moreover, OADE is not sensitive to the pool size.