Efficient and decentralized PageRank approximation in a peer-to-peer web search network

• DOI: 10.5555/1182635.1164164
• Publication Date: 2006-09-01
• Authors: Josiane Xavier Pa...; Debora Donato; Sebastian Michel; Gerhard Weikum
• Keywords: PageRank; Link analysis

PageRank-style (PR) link analyses are a cornerstone of Web search engines and mining, but they computationally expensive. Recently, various techniques have been proposed for speeding up these by distributing the graph among multiple sites. However, none advanced methods is suitable fully decentralized PR computation in peer-to-peer (P2P) network with autonomous peers, where each peer can independently crawl fragments according to user's thematic interests. In such setting that different peers locally available or know about may arbitrarily overlap creating additional complexity computation.This paper presents JXP algorithm dynamically collaboratively computing scores pages distributed P2P network. The runs at every peer, it works combining computed random meetings It scalable as number on grows, experiments well theoretical arguments show converge true one would obtain centralized computation.