Conditional Score Learning for Quickest Change Detection in Markov Transition Kernels
IEEE Transactions on Signal Processing · 2026 · 74 · 3271-3282
DOI: 10.1109/tsp.2026.3718274 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Signal Processing, 2026. Verfasst von Wuxia Chen und 2 weiteren Autor:innen. Die Arbeit nennt 29 Literaturangaben. Gefördert von: U.S. National Science Foundation.
Abstract (Originalsprache)
We address the problem of quickest change detection in Markov processes with unknown transition kernels. The key idea is to learn the conditional score $\mathbf{\nabla}_{\mathbf{y}}\mathbf{log} \boldsymbol{p}\mathbf{(}\mathbf{y}\mathbf{|}\mathbf{x}\mathbf{)}$ directly from sample pairs $\mathbf{(}\mathbf{x},\mathbf{y}\mathbf{)}$ , where both $\mathbf{x}$ and $\mathbf{y}$ are high-dimensional data generated by the same transition kernel. In this way, we avoid explicit likelihood evaluation and provide a practical way to learn the transition dynamics. Based on this estimation, we develop a score-based CUSUM procedure that uses conditional Hyvärinen score differences to detect changes in the kernel. To ensure bounded increments, we propose a truncated version of the statistic. With Hoeffding’s inequality for uniformly ergodic Markov processes, we prove exponential lower bounds on the mean time to false alarm. We also prove asymptotic upper bounds on detection delay. These results give both theoretical guarantees and practical feasibility for score-based detection in high-dimensional Markov models.
Förderung
- U.S. National Science Foundation
Zitieren
APA
Chen, W., Banerjee, T., & Tarokh, V. (2026). Conditional Score Learning for Quickest Change Detection in Markov Transition Kernels. IEEE Transactions on Signal Processing, 74, 3271-3282. https://doi.org/10.1109/tsp.2026.3718274
Vancouver
Chen W, Banerjee T, Tarokh V. Conditional Score Learning for Quickest Change Detection in Markov Transition Kernels. IEEE Transactions on Signal Processing. 2026;74:3271-3282. doi:10.1109/tsp.2026.3718274
BibTeX
@article{Chen2026,
author = {Chen, Wuxia and Banerjee, Taposh and Tarokh, Vahid},
title = {Conditional Score Learning for Quickest Change Detection in Markov Transition Kernels},
journal = {IEEE Transactions on Signal Processing},
year = {2026},
volume = {74},
pages = {3271--3282},
doi = {10.1109/tsp.2026.3718274}
}
EUVIMED