Online Non-parametric Change-point Detection via Graph-based Likelihood-Ratio Estimation

Download PDF
Open Webpage

Alejandro de la Concha, Nicolas Vayatis, Argyris Kalogeratos

Published: 2025, Last Modified: 28 Jan 2026EUSIPCO 2025EveryoneRevisionsBibTeXCC BY-SA 4.0
Abstract: Consider each node of a graph to be generating a data stream that is synchronized and observed at near real-time. At a change-point <tex>$\tau$</tex>, a change occurs for a subset of nodes <tex>$C$</tex>, which affects the probability distribution of their associated node streams. In this paper, we propose the Online Centralized Kernel-and Graph-based (OCKG) detection method to both detect <tex>$\tau$</tex> and localize <tex>$C$</tex>, based on the direct estimation of the likelihood-ratio between the post-change and the pre-change distributions of the node streams. Our main working hypothesis is the smoothness of the likelihood-ratio estimates over the graph, i.e. connected nodes are expected to have similar likelihood-ratios. The proposed method is evaluated in synthetic experiments.