Theory and Algorithm for Batch Distribution Drift Problems

Pranjal Awasthi, Corinna Cortes, Christopher Mohri
Proceedings of The 26th International Conference on Artificial Intelligence and Statistics, PMLR 206:9826-9851, 2023.

Abstract

We study a problem of batch distribution drift motivated by several applications, which consists of determining an accurate predictor for a target time segment, for which a moderate amount of labeled samples are at one’s disposal, while leveraging past segments for which substantially more labeled samples are available. We give new algorithms for this problem guided by a new theoretical analysis and generalization bounds derived for this scenario. We further extend our results to the case where few or no labeled data is available for the period of interest. Finally, we report the results of extensive experiments demonstrating the benefits of our drifting algorithm, including comparisons with natural baselines. A by-product of our study is a principled solution to the problem of multiple-source adaptation with labeled source data and a moderate amount of target labeled data, which we briefly discuss and compare with.

Cite this Paper


BibTeX
@InProceedings{pmlr-v206-awasthi23b, title = {Theory and Algorithm for Batch Distribution Drift Problems}, author = {Awasthi, Pranjal and Cortes, Corinna and Mohri, Christopher}, booktitle = {Proceedings of The 26th International Conference on Artificial Intelligence and Statistics}, pages = {9826--9851}, year = {2023}, editor = {Ruiz, Francisco and Dy, Jennifer and van de Meent, Jan-Willem}, volume = {206}, series = {Proceedings of Machine Learning Research}, month = {25--27 Apr}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v206/awasthi23b/awasthi23b.pdf}, url = {https://proceedings.mlr.press/v206/awasthi23b.html}, abstract = {We study a problem of batch distribution drift motivated by several applications, which consists of determining an accurate predictor for a target time segment, for which a moderate amount of labeled samples are at one’s disposal, while leveraging past segments for which substantially more labeled samples are available. We give new algorithms for this problem guided by a new theoretical analysis and generalization bounds derived for this scenario. We further extend our results to the case where few or no labeled data is available for the period of interest. Finally, we report the results of extensive experiments demonstrating the benefits of our drifting algorithm, including comparisons with natural baselines. A by-product of our study is a principled solution to the problem of multiple-source adaptation with labeled source data and a moderate amount of target labeled data, which we briefly discuss and compare with.} }
Endnote
%0 Conference Paper %T Theory and Algorithm for Batch Distribution Drift Problems %A Pranjal Awasthi %A Corinna Cortes %A Christopher Mohri %B Proceedings of The 26th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2023 %E Francisco Ruiz %E Jennifer Dy %E Jan-Willem van de Meent %F pmlr-v206-awasthi23b %I PMLR %P 9826--9851 %U https://proceedings.mlr.press/v206/awasthi23b.html %V 206 %X We study a problem of batch distribution drift motivated by several applications, which consists of determining an accurate predictor for a target time segment, for which a moderate amount of labeled samples are at one’s disposal, while leveraging past segments for which substantially more labeled samples are available. We give new algorithms for this problem guided by a new theoretical analysis and generalization bounds derived for this scenario. We further extend our results to the case where few or no labeled data is available for the period of interest. Finally, we report the results of extensive experiments demonstrating the benefits of our drifting algorithm, including comparisons with natural baselines. A by-product of our study is a principled solution to the problem of multiple-source adaptation with labeled source data and a moderate amount of target labeled data, which we briefly discuss and compare with.
APA
Awasthi, P., Cortes, C. & Mohri, C.. (2023). Theory and Algorithm for Batch Distribution Drift Problems. Proceedings of The 26th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 206:9826-9851 Available from https://proceedings.mlr.press/v206/awasthi23b.html.

Related Material