第八章 并行演化研究进展.pdf

并行演化算法研究进展 许有准1 曾文华2 （厦门大学计算机科学系 厦门 361005) （厦门大学软件学院 厦门 361005) 摘 要 并行计算与演化计算同为复杂问题寻优的现代技术.两者的结合不仅减少运行时间、提高解的质 量，还增加计算能力，即增加应用领域.并行演化算法除了，主从式、粗粒度、细粒度和分层四类并行化 模型外，还有改进的并行演化算法.本文分析介绍加速比定义分类、迁移策略与并行演化算法的应用情 况，总结前人对并行演化算法开展的理论分析与实践应用研究进展，并就该领域的进一步研究方向提出了 一些看法，指出细胞状模型和并行遗传程序设计将成为两大研究热点. 关键词 演化计算 并行计算 加速比 串行成分 效力 中图法分类号 THE DEVELOPMENT OF PARALLEL EVOLUTIONARY ALGORITHMS Xu Youzhun, Zeng Wenhua (Department of Computer Science, Xiamen University, Xiamen 361005) Abstract Parallel Computation and Evolutionary Computation both have attracted much attention recently as the modern techniques used for solving problems of high complexity. Combing the two, Parallel Evolutionary Computation improves greatly the efficiency, accuracy and computing ability of Evolutionary Computation. Parallel evolutionary algorithms (PEAs) are including not only master-slave, coarse grained, fine-grained and hierarchical parallel avenues, but also improved parallel evolutionary algorithms. In this paper, taxonomy of speedup definition, the migration policy and successful applications in PEAs are analyzed and generalized; a structured and extensive overview on theoretical analysis and implementation in PEAs are discussed. In addition, future research trends in this growing area are discussed. It is shown that researches of PEAs will focus on cellular model and parallel genetic programming. Key words Evolutionary Computation, Parallel Computation, Speedup, Serial Fraction, Efficacy 1 引 言 [1] 演化计算是基于生物进化思想而发展起来的一种通用问