基于双向维度抽取的协同进化存档算法

A Coevolution Archive Based on Bidirectional Dimension Extraction

为改进协同进化算法的可靠性和有效性,本文结合问题域内固有的维度结构特性,提出了一个基于双向维度抽取的存档算法.该算法简化了DECA中的维度抽取方法,并提出从测试个体和候选个体两端分别实施维度抽取(仅选取每个维度上代表当前进展的测试个体和带有维度信息特征的高性能候选个体保留存档),用于维持进化在各维度上的全局进展.实验表明,与同类算法相比,本算法使用的两个档案在进化中均保持了较小的存档量,性能高于其他同类算法.

To improve the reliability and efficiency of coevolutionary algorithm, an archive-based coevolutionary algorithm is proposed which is based on the dimension structures implicit in problem domain, and a method of bidirectional dimension extraction is incorporated. The proposed algorithm simplifies the dimension extraction method in DECA algorithm. It extracts dimensions from both sides of candidate solutions and tests, respectively. Then, for the purpose of maintaining progress in various dimensions, it only selects tests and high-performance candidate solutions that possess dimension representative features. The experimental results show that the proposed algorithm maintains relatively small archives, and outperforms three other existing archive-based algorithms.