A problem of estimating the intrinsic graph structure from observed data is considered. The observed data in this study is a matrix with elements representing dependency between nodes in the graph. Each element of the observed matrix represents, for example, co-occurrence of events at two nodes, or correlation of variables corresponding to two nodes. The dependency does not represent direct connections and includes influences of various paths, and spurious correlations make the estimation of direct connection difficult. To alleviate this difficulty, digraph Laplacian is used for characterizing a graph. A generative model of an observed matrix is proposed, and a parameter estimation algorithm for the model is also proposed. The proposed method is capable of dealing with directed graphs, while conventional graph structure estimation methods from an observed matrix are only applicable to undirected graphs. Experimental result shows that the proposed algorithm is able to identify the intrinsic graph structure. © 2014 Springer International Publishing Switzerland.
CITATION STYLE
Hino, H., Noda, A., Tatsuno, M., Akaho, S., & Murata, N. (2014). An algorithm for directed graph estimation. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 8681 LNCS, pp. 145–152). Springer Verlag. https://doi.org/10.1007/978-3-319-11179-7_19
Mendeley helps you to discover research relevant for your work.