Asyn-SimRank：一种可异步执行的大规模 SimRank 算法.pdf

DOI ：１０ ．７５４４/issn１０００‐１２３９ ．２０１５ Journal of Computer Research and Development ５２（７）：１５６７ １５７９ ，２０１５ Asyn‐SimRank ：SimRank １ ２ １ ２ １ ３ 　 　 　 　 　 　 １ （　 　 １１０８１９） ２ （　 　 １１０８１９） ３ （　 　 ０１００３） （zhangyf＠ cc ．neu ．edu ．cn ） Asyn‐SimRank ：An Asynchronous Large‐Scale SimRank Algorithm １ ２ １ ２ １ ３ Wang Chunlei ，Zhang Yanf eng ，Bao Y ubin ，Zhao Chang kuan ，Y u Ge ，and Gao Lixin １ （I ns titute of Comp ute r Sof tw are ， College of I nf ormation Sc ience and Eng inee r ing ， N o rtheas te rn Univ e rs ity ， Sheny ang １１０８１９） ２ （Comp ute r Cente r ， N ortheas te rn Univ e rs ity ， Sheny ang １１０８１９） ３ （Dep artment of E lectr ical and Comp ute r Eng inee r ing ，Univ e rs ity of M ass achusetts A mhe rs t ，A mhe rs t ，U ．S ．０１００３） Abstract 　 T he Sim Rank algorithm ， w hich ex ploit s netw ork structure to measure the similarity betw een node pairs ， has been w idely used in many areas ， such as online social netw orks and link prediction ．In recent y ears ， w ith the development of big data ， the input data set of the Sim Rank algorithm is constantly increasing ． People are utilizing distributed computing models ， such as M apReduce ， to desig n large‐scale Sim Rank algorithm f or solving the big data problems ． How ever ， since Sim Rank algorithm contains a high ‐co st iterative process w ith sy nchronization barriers betw een iterations and the computational complex ity is high in each iteration ， the large‐scale Sim Rank computation does not result in the satisf actory perf ormance ． In this paper ， １ ）We propo