多目标进化算法中收敛性能的实验研究-计算机应用技术专业论文.docxVIP

Abstract Multi-objective optimization problem is one of the main areas of research in optimization, most of the optimization problem have multi-objective characteristics in reality, and these goals often conflict with each other. Unlike single-objective optimization problems have a unique optimal solution, multi-objective optimization problem of the optimal solution has a compromise solution set, that is, Pareto optimal solution set. Evolutionary algorithm is a kind of simulation of biological natural selection and the natural evolution of the random search algorithm, its good at solving highly complex and nonlinear problems has been widely used, and can be run as a result of a group of non-dominated solution set, thus very suitable for solving multi-objective optimization problem. In recent years, researchers have applied for different problems,put forward its own multi-objective evolutionary algorithm,more representative algorithm: NSGA-II, SPEA-II, PESA-II and so on. However, because these algorithms use the Classic method, which determines to the complexity of some large-scale multi-extremal problem, even if it is based on the evolution of global search algorithms are often limited to local optimal. In addition, the traditional multi-objective evolutionary algorithm vulnerable to random disturbances. In this paper, in view of the current multi-objective evolutionary algorithm easily susceptible to a local convergence and random interference with the work of the shortcomings ,the study mainly included the following two aspects: First, Statement an estimate based on the local convergence of the multi-objective evolutionary algorithm.In the run of the algorithm , determining whether it has been in a local convergence is extremely difficult.So a method of estimating local convergence is proposed. when reached a relatively weak conditions, we believe that there is a certain probability that the algorithm has already fallen into a local convergence. We can under cert