Realtime Personalized PageRank Query for Social Network Search
As an improvement of classical PageRank,the personalized PageRank soon became one of the most major ranking algorithm in graph computation.However,it suffers from a severe efficiency issue and there are many studies focus on enhancing its precision and lowering down its complexity,among which the Mento Carlo random approximation estimation performs well in time and the power iteration method handles with accuracy better.In this paper,several optimization methods are combined to make it practical for personalized social network searching.It leverages the localization property of social network with local searching and quick dot product computing for the sparse matrix.It also optimizes the searching performance by pre-computing the intermediate data and parallelizing for the real-time computation.The experimental result shows that our improved personalized PageRank works well in the individual online search of the social network which can meet the online search time requirement with a practical storage usage.
personalized pagerank social network localization mento carlo
Jiang Wu
Shanghai Jiao Tong University Shanghai,China
国际会议
重庆
英文
647-651
2017-03-25（万方平台首次上网日期，不代表论文的发表时间）