Modeling graphs using dot product representations

Computational Statistics - Tập 25 - Trang 1-16 - 2009
Edward R. Scheinerman1, Kimberly Tucker2
1Department of Applied Mathematics and Statistics, Johns Hopkins University, Baltimore, USA
2Department of Mathematics, Harvey Mudd College, Claremont, USA

Tóm tắt

Given a simple (weighted) graph, or a collection of graphs on a common vertex set, we seek an assignment of vectors to the vertices such that the dot products of these vectors approximate the weight/frequency of the edges. By transforming vertices into (low dimensional) vectors, one can bring geometric methods to bear in the analysis of the graph(s). We illustrate our approach on the Mathematicians Collaboration Graph [Grossman (1996) The Erdős number project, http://www.oakland.edu/enp ] and the times series of Interstate Alliance Graphs (Gibler and Sarkees in J Peace Res 41(2):211–222, 2004).

Tài liệu tham khảo

Diehl P (2001) Correlates of war. http://www.correlatesofwar.org Gibler DM, Sarkees MR (2004) Measuring alliances: the correlates of war formal interstate alliance dataset, 1816-2000. J Peace Res 41(2): 211–222 Golub GH, van Loan CF (1996) Matrix computations. Johns Hopkins University Press, Baltimore Grossman J (1996) The Erdős number project. http://www.oakland.edu/enp Hoff PD, Raftery AE, Handcock MS (2002) Latent space approaches to social network analysis. J Am Stat Assoc 97(460): 1090–1098 Horn RA, Johnson CR (1985) Matrix analysis. Cambridge University Press, London Kraetzl M, Nickel C, Scheinerman ER (2009) Random dot product graphs: a model for social networks (submitted) Tucker K (2007) Exact and asymptotic dot product representations of graphs. Ph.D. Thesis, Johns Hopkins University