An ensemble of differential evolution algorithms with variable neighborhood search for constrained function optimization

TLDR: The proposed EDE-VNS algorithm employs multiple mutation operators and control parameters in its VNS loops to enhance the solution quality and utilizes opposition-based learning (OBL) to take advantages of opposite solutions to find a candidate solution which might be close to the global optimum.

Abstract: In this paper, an ensemble of differential evolution algorithms based on a variable neighborhood search algorithm (EDE-VNS) is proposed so as to solve the constrained real parameter-optimization problems. The performance of DE algorithms heavily depends on the mutation strategies, crossover operators and control parameters employed. The proposed EDE-VNS algorithm employs multiple mutation operators and control parameters in its VNS loops to enhance the solution quality. In addition, we utilize opposition-based learning (OBL) to take advantages of opposite solutions to find a candidate solution which might be close to the global optimum. In addition, we also present an idea of injecting some good dimensional values from promising areas in the population to the trial individual through the injection procedure. The computational results show that the EDE-VNS algorithm is very competitive to some of the best performing algorithms from the literature.