Using opposition-based learning to enhance differential evolution: A comparative study
Wenjun Wang, Hui Wang, Hui Sun, Shahryar Rahnamayan · 2016
Opposition-based learning (OBL) is a recently proposed method, which is successfully used to accelerate the search process of some well-known techniques in soft computing, such as swarm and evolutionary algorithms, artificial neural networks, reinforcement learning, and fuzzy logic systems. Among these opposition-based algorithms, opposition-based differential evolution (ODE) is one of the most popular algorithm. In the past several years, several variants of OBL scheme have been proposed. This paper presents a comparative study conducted on various OBL schemes, all utilized in differential evolution (DE) in order to enhance its accuracy or convergence rate. In the experiments, eight different OBL versions, namely the original OBL, quasi opposition, quasi reflection opposition, current optimum opposition, generalized opposition, centroid opposition, extended opposition, and reflected extended opposition, are embedded in DE algorithm and studied. Results on the CEC-2014 benchmark set for dimensions 10, 30, and 50 are reported.