PageRank computation via a distributed randomized approach with lossy communication

For ranking web pages in search results, Google employs the so-called PageRank algorithm, which provides a measure of importance to each page based on the web structure. Recently, we have developed a distributed randomized approach for this algorithm, where pages compute their own PageRank by communicating over selected links. In this paper, the focus is on the effects of unreliability in communication channels. Specifically, we consider random data losses modeled as a Markov chain and introduce a generalized version of the distributed scheme. Its convergence properties and the error in the approximated PageRank are analyzed.