Discounted Adaptive Online Learning: Towards Better Regularization

Zhiyu Zhang, David Bombara, Heng Yang
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:58631-58661, 2024.

Abstract

We study online learning in adversarial nonstationary environments. Since the future can be very different from the past, a critical challenge is to gracefully forget the history while new data comes in. To formalize this intuition, we revisit the discounted regret in online convex optimization, and propose an adaptive (i.e., instance optimal), FTRL-based algorithm that improves the widespread non-adaptive baseline – gradient descent with a constant learning rate. From a practical perspective, this refines the classical idea of regularization in lifelong learning: we show that designing better regularizers can be guided by the principled theory of adaptive online optimization. Complementing this result, we also consider the (Gibbs & Candes, 2021)-style online conformal prediction problem, where the goal is to sequentially predict the uncertainty sets of a black-box machine learning model. We show that the FTRL nature of our algorithm can simplify the conventional gradient-descent-based analysis, leading to instance-dependent performance guarantees.

Cite this Paper

BibTeX
@InProceedings{pmlr-v235-zhang24e, title = {Discounted Adaptive Online Learning: Towards Better Regularization}, author = {Zhang, Zhiyu and Bombara, David and Yang, Heng}, booktitle = {Proceedings of the 41st International Conference on Machine Learning}, pages = {58631--58661}, year = {2024}, editor = {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix}, volume = {235}, series = {Proceedings of Machine Learning Research}, month = {21--27 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v235/main/assets/zhang24e/zhang24e.pdf}, url = {https://proceedings.mlr.press/v235/zhang24e.html}, abstract = {We study online learning in adversarial nonstationary environments. Since the future can be very different from the past, a critical challenge is to gracefully forget the history while new data comes in. To formalize this intuition, we revisit the discounted regret in online convex optimization, and propose an adaptive (i.e., instance optimal), FTRL-based algorithm that improves the widespread non-adaptive baseline – gradient descent with a constant learning rate. From a practical perspective, this refines the classical idea of regularization in lifelong learning: we show that designing better regularizers can be guided by the principled theory of adaptive online optimization. Complementing this result, we also consider the (Gibbs & Candes, 2021)-style online conformal prediction problem, where the goal is to sequentially predict the uncertainty sets of a black-box machine learning model. We show that the FTRL nature of our algorithm can simplify the conventional gradient-descent-based analysis, leading to instance-dependent performance guarantees.} }
Endnote
%0 Conference Paper %T Discounted Adaptive Online Learning: Towards Better Regularization %A Zhiyu Zhang %A David Bombara %A Heng Yang %B Proceedings of the 41st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2024 %E Ruslan Salakhutdinov %E Zico Kolter %E Katherine Heller %E Adrian Weller %E Nuria Oliver %E Jonathan Scarlett %E Felix Berkenkamp %F pmlr-v235-zhang24e %I PMLR %P 58631--58661 %U https://proceedings.mlr.press/v235/zhang24e.html %V 235 %X We study online learning in adversarial nonstationary environments. Since the future can be very different from the past, a critical challenge is to gracefully forget the history while new data comes in. To formalize this intuition, we revisit the discounted regret in online convex optimization, and propose an adaptive (i.e., instance optimal), FTRL-based algorithm that improves the widespread non-adaptive baseline – gradient descent with a constant learning rate. From a practical perspective, this refines the classical idea of regularization in lifelong learning: we show that designing better regularizers can be guided by the principled theory of adaptive online optimization. Complementing this result, we also consider the (Gibbs & Candes, 2021)-style online conformal prediction problem, where the goal is to sequentially predict the uncertainty sets of a black-box machine learning model. We show that the FTRL nature of our algorithm can simplify the conventional gradient-descent-based analysis, leading to instance-dependent performance guarantees.
APA
Zhang, Z., Bombara, D. & Yang, H.. (2024). Discounted Adaptive Online Learning: Towards Better Regularization. Proceedings of the 41st International Conference on Machine Learning, in Proceedings of Machine Learning Research 235:58631-58661 Available from https://proceedings.mlr.press/v235/zhang24e.html.

Related Material