Event info

Date: Wednesday, 8 July 2026 16:45 - 17:45

Place: IMI Auditorium(W1-D-413) and Live streaming with Zoom

Speaker: Andre Fujita (Division of Network AI Statistics - Medical Institute of Bioregulation - Kyushu University)

Title: Nonparametric inference for network generative mechanisms via graph spectra

Abstract:
Complex systems are composed of many interacting components whose collective behavior often cannot be inferred from the properties of individual parts. Such systems arise across biology, sociology, economics, and technology and are naturally represented as networks that encode interactions among their components. Traditional approaches model these networks as random graphs and compare them using graph-level features such as edge counts, motifs, and centrality measures. However, these features are often incidental and do not directly capture the underlying generative mechanisms that define the systems of interest. I will present a nonparametric statistical framework for inference on unobserved network generation mechanisms using graph spectra, the eigenvalues of graph adjacency matrices. Because graph spectra are closely linked to both network structure and generative processes, they provide a principled basis for parameter estimation, model selection, graph comparison, correlation analysis, Granger causality for time-series graphs, clustering, and extensions to hypergraphs. The proposed methods are illustrated through applications.

IMI Colloquium Organizers
IKEMATSU, Yasuhiko
KURATA, Sumito
TAGAMI, Daisuke