Summary
2011 International Symposium on Nonlinear Theory and Its Applications
2011
Session Number:A3L-C
Session:
Number:A3L-C1
Graph zeta function and loopy belief propagation
Yusuke Watanabe, Kenji Fukumizu,
pp.192-195
Publication Date:2011/9/4
Online ISSN:2188-5079
Summary:
This paper discusses a link between the loopy belief propagation (LBP) algorithm and the Graph zeta function. The LBP algorithm is a nonlinear iteration to approximate the marginal or posterior probabilities required for various statistical inference, using the graph structure to define the joint probability. The theoretical properties of the LBP algorithm are not easy to analyze because of the complex nonlinearity and the graph structure. The derived connection with Graph zeta function involves the mathematical relation with the properties of the graph, leading various theoretical analysis of the LBP algorithm.