计算机应用 ›› 2010, Vol. 30 ›› Issue (8): 2073-2076.

• 先进计算 • 上一篇    下一篇

改进的最大和分散式协调算法

耿蕾蕾1,蔚承建2   

  1. 1. 南京工业大学信息工程与科学学院
    2.
  • 收稿日期:2010-01-29 修回日期:2010-03-16 发布日期:2010-07-30 出版日期:2010-08-01
  • 通讯作者: 耿蕾蕾

Improved max-sum decentralized coordination algorithm

  • Received:2010-01-29 Revised:2010-03-16 Online:2010-07-30 Published:2010-08-01

摘要: 大规模信息系统的协调问题正成为新兴的研究热点,最大和协调算法显示了在该领域的应用前景,然而其收敛速度和鲁棒性有待进一步提高。针对以上问题,提出一种可用于合作系统中的基于混沌的分散式信息传递算法,并通过图形着色问题验证了该算法的有效性。在该算法中,首先利用混沌序列丰富的时空动态性,产生最大和协调算法的节点信息;然后依据优化后的效用值表和更新规则对节点信息进行交互,完成算法的迭代。通过与传统最大和协调算法的对比实验结果分析,改进的算法不仅解决了传统最大和协调算法的计算瓶颈,而且在收敛速度及鲁棒性方面有着更好的表现。

关键词: 分散式协调, 和积算法, 因子图, 混沌序列, 效用值表

Abstract: Coordination problems of large scale information system gradually become a hot spot, and the maxsum product algorithm shows a bright application future but with improvement of the convergence speed and the robustness. In this paper a decentralized message passing algorithm based on chaotic sequence was presented which can be used in cooperative systems and was operated in a canonical coordination problem (graph coloring ) performing decentralized coordination. The spacetime dynamic messages of factor node were generated using chaotic sequence and updated following optimized utilitymap and simple rules. Compared to maxsum product algorithm, the solution resolves computational bottleneck and shows a better performance in both convergence speed and robustness.

Key words: decentralized coordination, sum-product algorithm, factor graph, chaotic sequence, utility-map