论文部分内容阅读
现代多目标进化算法在高维目标空间中遭遇性能危机,提出一种混合高维目标进化算法（Hybrid Many-Ob-jective Evolutionary Algorithm,HMOEA）以改善算法的解题性能.新的算法使用了新定义的w-支配关系替代Pa-reto支配关系;其次,为使算法在收敛性与多样性之间保持适当均衡,下一代种群个体的构成随当前进化世代动态调整;最后,算法使用了改进的拥挤距离赋值机制评估解个体密度以实施多样性保持操作.新算法在DTLZ2问题上进行测试,结果表明该算法可以获得很好的性能,而且新算法在收敛性和多样性之间也取得了较好的均衡.最后,从一般意义上分析了HMOEA算法的收敛性,分析结果表明HMOEA算法能够以概率1收敛.“,”There exsits a performance crisis for modern multi-objective evolutionary algorithms in high-dimensional objective space.In this paper,a hybird many-objective evolutionary algorithm（HMOEA） is proposed to improve MOEA＇s performance in solving many-objective optimization problems.In the HMOEA,w-dominance relation based on revaluation function is used to replace the Pareto-dominance relation,secondly,to balance the convergence and diversity in evolutionary population,the composition of the next population is varying with the current generation.Finally,an improved crowding distance assignment is used in HMOEA to evaluate individual＇s density to preserve diversity.HMOEA is examined on DTLZ2 to observe its performance in many-objective optimization problems.Experimental results illustrate that HMOEA exhibits a good performance in sense of convergence and diversity.At last,it is proven that the new algorithm can guarantee the convergence towards the global optimum under some conditions.