CiteRank is an algorithm for ranking citation networks based purely on their topology. Similar to Google's well-known Pagerank, CiteRank works by simulating diffusion along the network. In CiteRank, a large number of random researchers are simulated. Each researcher selects a recent publication and follows citations to older publications until they have are satisfied. A general introduction to the model is given here, but for precise details, see our paper. For an analysis of the of the Physical Review dataset, see here.
Citations networks are information networks formed by citations between articles. Citation networks arise from scientific publications, patent filings, or other more exotic information systems.
Citation networks feature a distinct time-arrow, because researchers can only cite articles that have already been written. As a result, movement on a citation network is always backwards in time.
Citations are rarely altered after an article is published. Thus, unlike the www, links in citation networks are never updated. One known consequence of this is that citation networks are prone to serious ageing effects.