Graph Characteristics from the Ihara Zeta Function
This paper shows how to extract permutation invariant graph characteristics from the Ihara zeta function. In a previous paper, we have shown that the Ihara zeta function leads to a polynomial characterization of graph structure, and we have shown empirically that the coefficients of the polynomial can be used as to cluster graphs. The aim in this paper is to take this study further by showing how to select the most significant coefficients and how these can be used to gauge graph similarity. Experiments on real-world datasets reveal that the selected coefficients give results that are significantly better than those obtained with the Laplacian spectrum.
KeywordsZeta Function Cluster Performance Edit Distance Feature Distance Rand Index
- 4.Chung, F.K.: Spectral graph theory. American Mathematical Society (1997)Google Scholar
- 7.Ihara, Y.: Discrete subgroups of PL(2, k ϕ). In: Proceeding Symposium of Pure Mathematics, pp. 272–278 (1965)Google Scholar
- 11.Rosenberg, S.: The Laplacian on a Riemannian manifold. Cambridge University Press, Cambridge (2002)Google Scholar
- 12.Ren, P., Wilson, R.C., Hancock, E.R.: Pattern vectors from the Ihara zeta function. In: IEEE International Conference on Pattern Recognition (submitted, 2008)Google Scholar
- 15.Storm, C.K.: The zeta function of a hypergraph. Electronic Journal of Combinatorics 13(1) (2006)Google Scholar