加速PageRank计算的方法研究

摘要

网络矩阵的规模以及稀疏性导致了对求解方法的限制，并使得幂法占据了主导地位。但是幂法的收敛速度是缓慢的，尤其在网络规模的矩阵上运行的每次幂法迭代的时间和成本是高昂的。因此，其他加速PageRank计算的方法逐渐得到研究者的重视。文中首先对布尔搜索引擎、向量空间模型引擎、概率模型搜索引擎、元搜索引擎等基本搜索引擎模型进行综述，总结各基本搜索引擎模型的特征和优缺点。文中立足于加速PageRank计算的方法研究，并总结出自适应幂法、外插方法、BlockRank聚合方法的特征和优缺点。%Scale and sparse network matrix led to restrictions on solving method, and makes the power method to occupy the dominant position. But convergence speed is slow, especially in the network every time power method to run on the size of the matrix iterative time and cost is high. Therefore, other accelerate PageRank calculation method gradually got the attention of the researchers. This paper summarizes the basic characteristics and advantages and disadvantages of search engine model, including the Boolean search engine, vector space model engines, search engines, metasearch engines. Finally, this paper analyzes the accelerate the PageRank calculation method of study, Summed up the characteristics and advantages and disadvantages including adaptive power method, extrapolation method, polymerization method.