Retrain or not retrain: conformal test martingales for change-point detection

Vladimir Vovk, Ivan Petej, Ilia Nouretdinov, Ernst Ahlberg, Lars Carlsson, Alex Gammerman
Proceedings of the Tenth Symposium on Conformal and Probabilistic Prediction and Applications, PMLR 152:191-210, 2021.

Abstract

We argue for supplementing the process of training a prediction algorithm by setting up a scheme for detecting the moment when the distribution of the data changes and the algorithm needs to be retrained. Our proposed schemes are based on exchangeability martingales, i.e., processes that are martingales under any exchangeable distribution for the data. Our method, based on conformal prediction, is general and can be applied on top of any modern prediction algorithm. Its validity is guaranteed, and in this paper we make first steps in exploring its efficiency.

Cite this Paper

BibTeX
@InProceedings{pmlr-v152-vovk21b, title = {Retrain or not retrain: conformal test martingales for change-point detection}, author = {Vovk, Vladimir and Petej, Ivan and Nouretdinov, Ilia and Ahlberg, Ernst and Carlsson, Lars and Gammerman, Alex}, booktitle = {Proceedings of the Tenth Symposium on Conformal and Probabilistic Prediction and Applications}, pages = {191--210}, year = {2021}, editor = {Carlsson, Lars and Luo, Zhiyuan and Cherubin, Giovanni and An Nguyen, Khuong}, volume = {152}, series = {Proceedings of Machine Learning Research}, month = {08--10 Sep}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v152/vovk21b/vovk21b.pdf}, url = {https://proceedings.mlr.press/v152/vovk21b.html}, abstract = {We argue for supplementing the process of training a prediction algorithm by setting up a scheme for detecting the moment when the distribution of the data changes and the algorithm needs to be retrained. Our proposed schemes are based on exchangeability martingales, i.e., processes that are martingales under any exchangeable distribution for the data. Our method, based on conformal prediction, is general and can be applied on top of any modern prediction algorithm. Its validity is guaranteed, and in this paper we make first steps in exploring its efficiency.} }
Endnote
%0 Conference Paper %T Retrain or not retrain: conformal test martingales for change-point detection %A Vladimir Vovk %A Ivan Petej %A Ilia Nouretdinov %A Ernst Ahlberg %A Lars Carlsson %A Alex Gammerman %B Proceedings of the Tenth Symposium on Conformal and Probabilistic Prediction and Applications %C Proceedings of Machine Learning Research %D 2021 %E Lars Carlsson %E Zhiyuan Luo %E Giovanni Cherubin %E Khuong An Nguyen %F pmlr-v152-vovk21b %I PMLR %P 191--210 %U https://proceedings.mlr.press/v152/vovk21b.html %V 152 %X We argue for supplementing the process of training a prediction algorithm by setting up a scheme for detecting the moment when the distribution of the data changes and the algorithm needs to be retrained. Our proposed schemes are based on exchangeability martingales, i.e., processes that are martingales under any exchangeable distribution for the data. Our method, based on conformal prediction, is general and can be applied on top of any modern prediction algorithm. Its validity is guaranteed, and in this paper we make first steps in exploring its efficiency.
APA
Vovk, V., Petej, I., Nouretdinov, I., Ahlberg, E., Carlsson, L. & Gammerman, A.. (2021). Retrain or not retrain: conformal test martingales for change-point detection. Proceedings of the Tenth Symposium on Conformal and Probabilistic Prediction and Applications, in Proceedings of Machine Learning Research 152:191-210 Available from https://proceedings.mlr.press/v152/vovk21b.html.

Related Material