遗传算法的群体分组并行研究  

Research on Population Grouping Parallelism of Genetic Algorithm

作　　者：张会福[1] 陈然[1] 

机构地区：[1]湖南科技大学计算机科学与工程学院,湖南湘潭411201

出　　处：《微电子学与计算机》2009年第8期140-143,共4页Microelectronics & Computer

基　　金：国家自然科学基金项目(50775070);湖南省重点实验室专项计划项目(2008TP4039-2)

摘　　要：遗传算法是解决最优化问题的一种有效方法,并行化处理是其重要发展方向.文中基于群体分组并行性提出一种引入精英保留机制和迁移操作的分布式并行遗传算法,弥补了基本遗传算法"过早收敛"和求解精度不高等不足.实验结果表明,该算法较之串行方式能获得更高的求解质量和更快的寻优速度.A genetic algorithm is an effective method to solve optimization problems. To make up for the lack of ＂premature convergence＂ and low precision, analyzed the four possible parallelisms of genetic algorithms, and based on the population grouping parallelism, we proposed a distributed parallel genetic algorithm （DPGA）. In this DPGA, we present an individual migration strategy, by which improve average fitness and rich individuals diversity of subpopulations; it can also enhance the evolutionary direction of genetic algorithm and improve the results accuracy. Experiences results show that the present DPGA has better search capability and faster convergence rate.

关 键 词：并行遗传算法 群体分组并行性 任务调度 迁移操作 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]