Online Learning with Optimism and Delay
Flaspohler, Genevieve, Orabona, Francesco, Cohen, Judah, Mouatadid, Soukayna, Oprescu, Miruna, Orenstein, Paulo, Mackey, Lester
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.
Jun-16-2021
- Country:
- Europe > United Kingdom
- North America
- Canada > Ontario
- Toronto (0.14)
- United States > Massachusetts (0.14)
- Canada > Ontario
- Genre:
- Overview (0.67)
- Research Report (0.63)
- Industry:
- Education > Educational Setting > Online (1.00)