Parametric definition of the influence of a paper in a citation network using communicability functions

• Published in: Journal of complex networks 2019-08, Vol.7 (4), p.623-640
• Main Authors: Pichardo-Corpus, Juan A, Contreras, J Guillermo, de la Peña, José A
• Format: Article
• Language: English
• ISSN: 2051-1329; 2051-1329
• DOI: 10.1093/comnet/cny037

Communicability functions quantify the flow of information between two nodes of a network. In this work, we use them to explore the concept of the influence of a paper in a citation network. These functions depend on a parameter. By varying the parameter in a continuous way we explore different definitions of influence. We study six citation networks, three from physics and three from computer science. As a benchmark, we compare our results against two frequently used measures: the number of citations of a paper and the PageRank algorithm. We show that the ranking of the articles in a network can be varied from being equivalent to the ranking obtained from the number of citations to a behaviour tending to the eigenvector centrality, these limits correspond to small and large values of the communicability-function parameter, respectively. At an intermediate value of the parameter a PageRank-like behaviour is recovered. As a test case, we apply communicability functions to two sets of articles, where at least one author of each paper was awarded a Nobel Prize for the research presented in the corresponding article.