摘要
针对传统PageRank算法平均分配PageRank值给每个超链接网页这一缺陷,提出了改进的PageRank算法,并证明如果Web网的邻接矩阵P包含至少2个不可约闭子集,则非周期不可约矩阵的次特征值为d且至少2重.为了降低解PageRank近似解的误差和提高幂法的收敛速度,用lingo算得d取0.71,且知若采用改进的PageRank算法用小于0.85的d值可以达到传统Pag-eRank算法的计算结果.
Based on the average distribution of the traditional PageRank algorithm Pag- eRank value to each Web page hyperlink, this paper presents an improved PageRank al- gorithm, and proves that if the Web hyperlink matrix P used by Google for computing PageRank contains at least two irreducible closed subsets, the second eigenvalue for ma- trix is d, and the multiplicity of the eigenvalue d is 2. In order to reduce the error of the approximate PageRank solutions and improve the convergence speed, d with the lingo calculated is to take 0.71. And if using the improved PageRank algorithm, the results of traditional PageRank algorithm can be achieved with the value of d less than 0.85.
出处
《华中师范大学学报（自然科学版）》
CAS
CSCD
北大核心
2011年第4期534-537,共4页
Journal of Central China Normal University：Natural Sciences
基金
云南省教育厅科学研究基金项目(09y0423)
