Improved Online Conformal Prediction via Strongly Adaptive Online Learning
This work addresses uncertainty quantification for online learning systems in non-stationary settings, representing an incremental improvement over prior online conformal prediction techniques.
The paper tackles the problem of uncertainty quantification in changing environments by developing online conformal prediction methods that minimize strongly adaptive regret, achieving near-optimal regret for all intervals and approximately valid coverage, with experiments showing better coverage and smaller prediction sets than existing methods on tasks like time series forecasting and image classification under distribution shift.
We study the problem of uncertainty quantification via prediction sets, in an online setting where the data distribution may vary arbitrarily over time. Recent work develops online conformal prediction techniques that leverage regret minimization algorithms from the online learning literature to learn prediction sets with approximately valid coverage and small regret. However, standard regret minimization could be insufficient for handling changing environments, where performance guarantees may be desired not only over the full time horizon but also in all (sub-)intervals of time. We develop new online conformal prediction methods that minimize the strongly adaptive regret, which measures the worst-case regret over all intervals of a fixed length. We prove that our methods achieve near-optimal strongly adaptive regret for all interval lengths simultaneously, and approximately valid coverage. Experiments show that our methods consistently obtain better coverage and smaller prediction sets than existing methods on real-world tasks, such as time series forecasting and image classification under distribution shift.