Research On Hybrid Multi-Objective Evolutionary Algorithm Based On Estimation Of Distribution

To solve all kinds of the optimization problem, it presents a variety of optimization algorithms. Evolutionary Algorithms have some features like good versatility, highly nonlinear, parallelism, etc., so that it can be obtained the problem of multi-objective Pareto optimal solution quickly and effectively. As a class of heuristic search algorithms, evolutionary algorithms have been successfully applied to multi-objective optimization field, and developed into a relatively hot research direction.In this paper, According to the features of MOEA and the advantages of EDA, I give a kind of research to MED A. Based on the advantages and disadvantages of EDA, now we combine EDA and DE, and then propose the research on adaptive hybrid Multi-Objective Evolutionary Algorithm Based on Estimation of Distribution. The basic idea is:In order to a strong global convergence, DE is introduced in MED A. When the change rate of function is larger, it generates new populations with MED A, and when the change rate of function is smaller, likely to when the algorithm falls into local convergence, it generates new populations with DE. Theoretical analysis and numerical experiments show that:The new improved algorithm not only has good global convergence, but also its distribution and uniformity of solutions have been improved to some extent than the hybrid multi-objective estimation of distribution algorithm without considering the change rate of function.So it has important theoretical significance and application value. The adaptive hybrid multi-objective evolutionary algorithm is applied to Shop scheduling optimization problem like Job-Shop problem, the superiority of the new algorithm has been verified.