A0905
Title: Discovering underlying dynamics in time series of networks
Authors: Avanti Athreya - Johns Hopkins University (United States) [presenting]
Zachary Lubberts - University of Virginia (United States)
Carey Priebe - Johns Hopkins University (United States)
Youngser Park - Johns Hopkins University (United States)
Abstract: Understanding dramatic changes in the evolution of networks is central to statistical network inference. A joint network model has been considered in which each node has an associated time-varying low-dimensional latent vector of feature data, and connection probabilities are functions of these vectors. Under mild assumptions, the time-varying evolution of the constellation of latent vectors exhibits a low-dimensional manifold structure under a suitable notion of distance. This distance can be approximated by a measure of separation between the observed networks. Euclidean representations exist for the underlying network structure, characterized by this distance, at any given time. These Euclidean representations and their data-driven estimates permit the visualization of network evolution and transform network inference questions such as change-point and anomaly detection into a classical setting. The methodology is illustrated with real and synthetic data, and change points are identified corresponding to shifts in pandemic policies in a communication network of a large organization.
