您的位置:山东大学 -> 科技期刊社 -> 《山东大学学报(工学版)》

山东大学学报(工学版) ›› 2011, Vol. 41 ›› Issue (3): 1-6.

• 机器学习与数据挖掘 •    下一篇

求解公式关键文字集的信息传播算法

王晓峰,许道云,秦永彬   

  1. 贵州大学计算机科学系, 贵州 贵阳 550025
  • 收稿日期:2011-03-14 出版日期:2011-06-16 发布日期:2011-03-14
  • 作者简介:王晓峰(1980- ),男,甘肃会宁人,博士研究生,主要研究方向为算法分析与设计.Email:wxf-gzu@163.com
  • 基金资助:

    国家自然科学基金资助项目(60863005, 61011130038);贵州大学自然科学青年科研基金资助项目((2009)021);贵州大学研究生创新基金资助项目

A message propagation algorithm computing set of key literals of a formula

WANG Xiao-feng, XU Dao-yun, QIN Yong-bin   

  1. Department of Computer Science, Guizhou University, Guiyang 550025, China
  • Received:2011-03-14 Online:2011-06-16 Published:2011-03-14

摘要:

关键文字集影响判定布尔公式可满足性的判定难度。如果能找到公式的关键文字集或关键文字集的子集,就可使公式的可满足性判定变得容易。通过对警示传播算法的原理分析,发现高概率决定的部分变元对公式的求解难度有一定的影响。当某个子句与变元之间的警示信息具有一致性时,表明该子句的可满足性完全由该变元的取值决定,其它变元的取值都不能使得该子句可满足。进一步研究发现,当该算法收敛时,具有一致性的警示信息对应的变元集合是关键文字集的子集,这就佐证了警示传播算法可用于求解公式的关键文字集。基于该算法的特征,给出了一个寻找公式关键文字集的信息传播算法。

关键词: 信息传递, 警示传播算法, 原理分析, 关键文字

Abstract:

The difficulty of solving Boolean formulae is mainly decided by the key literals. If the key literals or subset of formulas could be found, they  would be tractable to solve the formula. The principle of the warning propagation was studied, and it was found that the decision part variables of high probability were important for solving the Boolean formulae. It was also found that if the warning message had consistency between the clause and variable, the variable could decide the clause satisfiability, while other variables did not contribute to satisfiability of the clause. Furthermore, it was proved that the part variables were the subset of the key literals when the warning propagation was convergent, which showed that the warning propagation could be used to find the key literals for the formulae. Finally, the massage passing algorithm to find the key literals was given by the warning propagation characterization.
 

Key words: message passing, warning propagation algorithm, analysis of principle, key literal

No related articles found!
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
No Suggested Reading articles found!