Adaptive Gaussian Process Change Point Detection

Edoardo Caldarelli, Philippe Wenk, Stefan Bauer, Andreas Krause
Proceedings of the 39th International Conference on Machine Learning, PMLR 162:2542-2571, 2022.

Abstract

Detecting change points in time series, i.e., points in time at which some observed process suddenly changes, is a fundamental task that arises in many real-world applications, with consequences for safety and reliability. In this work, we propose ADAGA, a novel Gaussian process-based solution to this problem, that leverages a powerful heuristics we developed based on statistical hypothesis testing. In contrast to prior approaches, ADAGA adapts to changes both in mean and covariance structure of the temporal process. In extensive experiments, we show its versatility and applicability to different classes of change points, demonstrating that it is significantly more accurate than current state-of-the-art alternatives.

Cite this Paper


BibTeX
@InProceedings{pmlr-v162-caldarelli22a, title = {Adaptive {G}aussian Process Change Point Detection}, author = {Caldarelli, Edoardo and Wenk, Philippe and Bauer, Stefan and Krause, Andreas}, booktitle = {Proceedings of the 39th International Conference on Machine Learning}, pages = {2542--2571}, year = {2022}, editor = {Chaudhuri, Kamalika and Jegelka, Stefanie and Song, Le and Szepesvari, Csaba and Niu, Gang and Sabato, Sivan}, volume = {162}, series = {Proceedings of Machine Learning Research}, month = {17--23 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v162/caldarelli22a/caldarelli22a.pdf}, url = {https://proceedings.mlr.press/v162/caldarelli22a.html}, abstract = {Detecting change points in time series, i.e., points in time at which some observed process suddenly changes, is a fundamental task that arises in many real-world applications, with consequences for safety and reliability. In this work, we propose ADAGA, a novel Gaussian process-based solution to this problem, that leverages a powerful heuristics we developed based on statistical hypothesis testing. In contrast to prior approaches, ADAGA adapts to changes both in mean and covariance structure of the temporal process. In extensive experiments, we show its versatility and applicability to different classes of change points, demonstrating that it is significantly more accurate than current state-of-the-art alternatives.} }
Endnote
%0 Conference Paper %T Adaptive Gaussian Process Change Point Detection %A Edoardo Caldarelli %A Philippe Wenk %A Stefan Bauer %A Andreas Krause %B Proceedings of the 39th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2022 %E Kamalika Chaudhuri %E Stefanie Jegelka %E Le Song %E Csaba Szepesvari %E Gang Niu %E Sivan Sabato %F pmlr-v162-caldarelli22a %I PMLR %P 2542--2571 %U https://proceedings.mlr.press/v162/caldarelli22a.html %V 162 %X Detecting change points in time series, i.e., points in time at which some observed process suddenly changes, is a fundamental task that arises in many real-world applications, with consequences for safety and reliability. In this work, we propose ADAGA, a novel Gaussian process-based solution to this problem, that leverages a powerful heuristics we developed based on statistical hypothesis testing. In contrast to prior approaches, ADAGA adapts to changes both in mean and covariance structure of the temporal process. In extensive experiments, we show its versatility and applicability to different classes of change points, demonstrating that it is significantly more accurate than current state-of-the-art alternatives.
APA
Caldarelli, E., Wenk, P., Bauer, S. & Krause, A.. (2022). Adaptive Gaussian Process Change Point Detection. Proceedings of the 39th International Conference on Machine Learning, in Proceedings of Machine Learning Research 162:2542-2571 Available from https://proceedings.mlr.press/v162/caldarelli22a.html.

Related Material