Online Learning with Optimism and Delay

Genevieve E Flaspohler, Francesco Orabona, Judah Cohen, Soukayna Mouatadid, Miruna Oprescu, Paulo Orenstein, Lester Mackey
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:3363-3373, 2021.

Abstract

Inspired by the demands of real-time climate and weather forecasting, we develop optimistic online learning algorithms that require no parameter tuning and have optimal regret guarantees under delayed feedback. Our algorithms—DORM, DORM+, and AdaHedgeD—arise from a novel reduction of delayed online learning to optimistic online learning that reveals how optimistic hints can mitigate the regret penalty caused by delay. We pair this delay-as-optimism perspective with a new analysis of optimistic learning that exposes its robustness to hinting errors and a new meta-algorithm for learning effective hinting strategies in the presence of delay. We conclude by benchmarking our algorithms on four subseasonal climate forecasting tasks, demonstrating low regret relative to state-of-the-art forecasting models.

Cite this Paper

BibTeX
@InProceedings{pmlr-v139-flaspohler21a, title = {Online Learning with Optimism and Delay}, author = {Flaspohler, Genevieve E and Orabona, Francesco and Cohen, Judah and Mouatadid, Soukayna and Oprescu, Miruna and Orenstein, Paulo and Mackey, Lester}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {3363--3373}, year = {2021}, editor = {Meila, Marina and Zhang, Tong}, volume = {139}, series = {Proceedings of Machine Learning Research}, month = {18--24 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v139/flaspohler21a/flaspohler21a.pdf}, url = {https://proceedings.mlr.press/v139/flaspohler21a.html}, abstract = {Inspired by the demands of real-time climate and weather forecasting, we develop optimistic online learning algorithms that require no parameter tuning and have optimal regret guarantees under delayed feedback. Our algorithms—DORM, DORM+, and AdaHedgeD—arise from a novel reduction of delayed online learning to optimistic online learning that reveals how optimistic hints can mitigate the regret penalty caused by delay. We pair this delay-as-optimism perspective with a new analysis of optimistic learning that exposes its robustness to hinting errors and a new meta-algorithm for learning effective hinting strategies in the presence of delay. We conclude by benchmarking our algorithms on four subseasonal climate forecasting tasks, demonstrating low regret relative to state-of-the-art forecasting models.} }
Endnote
%0 Conference Paper %T Online Learning with Optimism and Delay %A Genevieve E Flaspohler %A Francesco Orabona %A Judah Cohen %A Soukayna Mouatadid %A Miruna Oprescu %A Paulo Orenstein %A Lester Mackey %B Proceedings of the 38th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2021 %E Marina Meila %E Tong Zhang %F pmlr-v139-flaspohler21a %I PMLR %P 3363--3373 %U https://proceedings.mlr.press/v139/flaspohler21a.html %V 139 %X Inspired by the demands of real-time climate and weather forecasting, we develop optimistic online learning algorithms that require no parameter tuning and have optimal regret guarantees under delayed feedback. Our algorithms—DORM, DORM+, and AdaHedgeD—arise from a novel reduction of delayed online learning to optimistic online learning that reveals how optimistic hints can mitigate the regret penalty caused by delay. We pair this delay-as-optimism perspective with a new analysis of optimistic learning that exposes its robustness to hinting errors and a new meta-algorithm for learning effective hinting strategies in the presence of delay. We conclude by benchmarking our algorithms on four subseasonal climate forecasting tasks, demonstrating low regret relative to state-of-the-art forecasting models.
APA
Flaspohler, G.E., Orabona, F., Cohen, J., Mouatadid, S., Oprescu, M., Orenstein, P. & Mackey, L.. (2021). Online Learning with Optimism and Delay. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:3363-3373 Available from https://proceedings.mlr.press/v139/flaspohler21a.html.

Related Material