A Differential Evolution Algorithm to Semivectorial Bilevel Problems
A Differential Evolution Algorithm to Semivectorial Bilevel Problems
复制标题
DOI：
10.1007/978-3-319-72926-8_15
复制
发表时间：
2017-09
期刊：
影响因子：
--
通讯作者：
M. J. Alves;C. H. Antunes
中科院分区：
文献类型：
--
作者：
M. J. Alves;C. H. Antunes
半矢量型双层问题(SVBLP)是在递阶决策的上层处理单个函数的优化问题，在下层处理多个目标函数的优化问题。因此，下层决策者(跟随者)存在一组非支配解，并且应该为上层决策者(领导者)控制的每一组决策变量利用这些非支配解。本文提出了一种新的基于差分进化的算法来计算SVBLP的四个最优解。这些解决方案不仅捕捉到乐观与悲观领导者的态度，还捕捉到了较低级别的非支配解集合中可能的追随者或多或少对领导者有利的反应。将差分进化算法与粒子群算法进行了比较。在这个实验比较中，我们提请注意与SVBLP中的结果解释和算法性能评估相关的陷阱。
Semivectorial bilevel problems (SVBLP) deal with the optimization of a single function at the upper level and multiple objective functions at the lower level of hierarchical decisions. Therefore, a set of nondominated solutions to the lower level decision maker (the follower) exists and should be exploited for each setting of decision variables controlled by the upper level decision maker (the leader). This paper presents a new algorithmic approach based on differential evolution to compute a set of fourextremesolutions to the SVBLP. These solutions capture not just the optimistic vs. pessimistic leader’s attitude but also possible follower’s reactions more or less favorable to the leader within the lower level nondominated solution set. The differential evolution approach is compared with a particle swarm optimization algorithm. In this experimental comparison we draw attention to pitfalls associated with the interpretation of results and assessment of the performance of algorithms in SVBLP.