【摘要】 为了提高大规模高维多目标优化算法的搜索效率，提出了两层优化策略实现最优解集的搜索。该方法中首先利用社会学习微粒群算法的种群多样性增加算法的探索能力以防止局部收敛，产生的新种群用于更新非支配解集增加非支配解集的多样性。然而，由于种群多样性的增加会降低算法的收敛速度，为此，在第二层优化操作中选择当前非支配解集的若干解进行遗传操作以增加算法的开采能力，产生的新种群进一步更新非支配解集，以期加快对大规模高维多目标问题最优解集的搜索。在500维MaF测试函数上进行了测试，目标空间维度分别为3,5,8,10,并且与近些年提出的相关算法对比。通过实验说明了两层优化策略在解决大规模高维多目标优化问题时是有效的。更多还原

【Abstract】 In order to improve the efficiency of large-scale many-objective optimization algorithm, a two-layer optimization strategy is proposed for the search on the optimal solutions of large-scale many-objective problems.In the method, the social learning particle swarm optimization is used to generate a new population, which is expected to improve the exploration capability of the algorithm and prevent the premature convergence.The new generated population will then be used to update the archive to improve the diversity of the non-dominated optimal solutions.However, the increase of the diversity may reduce the speed of the convergence.Therefore, in the second optimization strategy, genetic operations on a number of non-dominated solutions are used to improve the exploitation capability of the method.The new generated population will also be used to update the non-dominated optimal solution set so as to speed up finding the Pareto optimal solutions of the large-scale many-objective problems.A number of experiments on the 500-dimensional MaF benchmark problems with the number of objectives are 3,5,8 and 10 are conducted.The performance of the proposed method is compared to four algorithms proposed in recent years.Experimental results show that the proposed two-layer optimization strategy is effective for solving large-scale many-objective optimization problems.更多还原