摘要
为了改善多目标粒子群优化算法生成的最终Pareto前端的多样性和收敛性,提出了一种针对多目标粒子群算法进化状态的检测机制.通过对外部Pareto解集的更新情况进行检测,进而评估算法的进化状态,获取反馈信息来动态调整进化策略,使得算法在进化过程中兼顾近似Pareto前端的多样性和收敛性.最后,在ZDT系列测试函数中,将本文算法与其他4种对等算法比较,证明了本文算法生成的最终Pareto前端在多样性和收敛性上均有显著的优势.
To improve the diversity and convergence of Pareto front generated by multi objective particle swarm optimization, a detection mechanism for evolutionary state of multi objective particle swarm optimization is presented in this paper. The evolutionary state of the algorithm is assumed by detecting the updating situation of the external Pareto set to get the feedback information to adjust the evolutionary strategy of the algorithm dynamically. It enables the algorithm to take the diversity and convergence of the approximate Pareto front into account in the process of the evolution. Finally,the proposed algorithm shows a good performance compared with other four kinds of equivalence algorithms in the ZDT series test function.
出处
《计算机系统应用》
2017年第7期161-166,共6页
Computer Systems & Applications
基金
山东省自然科学基金(ZR2013FL034)
关键词
多目标优化
粒子群算法
反馈信息
进化状态
multi objective optimization
particle swarm optimization
feedback information
evolutionary state