节点文献

基于渗透原理迁移策略的并行遗传算法

Parallel Genetic Algorithms with Migration Scheme Based on Penetration Theory

  • 推荐 CAJ下载
  • PDF下载
  • 不支持迅雷等下载工具,请取消加速工具后下载。

【作者】 赖鑫生张明义

【Author】 LAI Xin-Sheng 1) ZHANG Ming-Yi 2) 1) (Department of Computer Science, Guizhou University, Guiyang 550025) 2 (Guizhou Academy of Sciences, Guiyang 550001)

【机构】 贵州大学计算机科学系贵州科学院 贵阳550025贵阳550001

【摘要】 通过分析影响并行遗传算法性能的诸多因素,以避免人为设置迁移代频、迁移率及迁移方向为问题的突破口,以减少通信量提高算法效率为主旨,提出一种基于渗透原理的迁移策略(MigrationSchemeBasedOnPenetration,PMS).PMS迁移策略源于渗透模型,引入渗透阈值控制相邻子群体的迁移,应用渗透原理自适应地确定迁移代频、迁移率及迁移方向,从而解决人为设置迁移代频、迁移率及迁移方向的关键问题,有效降低通信代价,进而提高算法效率.文中首先依据有限群体马尔可夫链模型对基于渗透原理的迁移策略算法的可行性进行了探讨,然后从理论角度给出了迁移代频期望、迁移率期望及通信代价,同时用实例验证了PMS在降低通信代价方面的巨大潜力.

【Abstract】 Analyzed several main factors that affect the performances of parallel genetic algorithms, and in order to solve the hard problem that migration interval and migration rate and migration direction must be set manually, and also in order to increase the performances of parallel genetic algorithms, authors propose a new migration scheme based on penetration theory. Introducing a threshold to control migration between any two sub-populations, authors use the penetration theory to set migration interval and migration rate and migration direction adaptively. The feasibility of this new migration scheme is discussed according to Markov Chain of finite population model. The expectation of migration interval and the expectation of migration rate are formalized then, the formalizing of communication costs is at the last, at the same time authors verify the PMS(Migration Scheme Based on Penetration) could greatly decrease the cost of communication by examples.

【基金】 国家自然科学基金(10161005)资助.
  • 【文献出处】 计算机学报 ,Chinese Journal of Computers , 编辑部邮箱 ,2005年07期
  • 【分类号】TP18
  • 【被引频次】21
  • 【下载频次】399
节点文献中: 

本文链接的文献网络图示:

本文的引文网络