A1347
Title: Online monitoring for distributional changes with energy distances
Authors: B Cooper Boniece - Drexel University (United States) [presenting]
Lorenzo Trapani - University of Leicester (United Kingdom)
Lajos Horvath - University of Utah (USA)
Abstract: The aim is to propose a nonparametric sequential monitoring procedure for detecting distributional breaks in otherwise stationary time series. The method is based on a class of degenerate U-statistics that includes energy distances and kernel-based maximum mean discrepancy (MMD). The framework accommodates weak dependence and general data types, including multivariate and functional observations. Asymptotic behavior is established under the null, and the distribution of detection delays under alternatives is characterized. Simulations and real-data examples demonstrate strong performance in identifying subtle and mixed-type changes.
