一种新的进化多目标优化算法中的文档策略

在采用带精英保留策略的进化算法求解多目标优化问题时,我们需要一个文档来保存在进化过程中产生的不被占优的解。当互不占优的解的数量超过文档的大小时,如何在互不占优的解之间取舍将变得非常重要。针对传统的文档更新策略只考虑解的分布性的情况,提出了一种能够同时保证分布性和收敛性的文档更新策略。该文档更新策略以超体积为基础,与目前经典算法NSGA-Ⅱ进行比较,结果表明新算法拥有良好的分布性,同时也较好地保证了收敛性。

An Archiving Strategy in Evolutionary Multi-Objective Optionization Algorithm

When evolutionary algorithm with elitist strategy is used to solve multi-objective optimization problems,we need an archive to maintain nondominated solutions generated in the evolutionary process.When the number of mutually nondominated solutions is larger than archive＇s capacity,how to decide which solutions should be maintained and which discarded will become very important.For the traditional methods only concentrate on the distribution of solutions,an archiving updating strategy with both diversity and convergence con-sidered is proposed in this paper.This archiving updating strategy is based on hypervolume.Compared with NSGA-Ⅱ,the experimental results demonstrate that the new algorithm can make the solutions as spread as possible in the Pareto front and ensure convergence.