Online Classification with Predictions
Vinod Raman, Ambuj Tewari
TL;DR
This work addresses online classification when the learner has access to predictions of future inputs via a consistent, lazy Predictor. It develops adaptive online learners that interpolate between offline (transductive) learnability and worst-case online performance, with mistake/regret bounds that degrade gracefully as predictor quality, measured by $M_{\\mathcal{P}}(x_{1:T})$, improves. The key contributions include a concrete upper-bound framework that combines Littlestone dimension, predictor- and offline-learner error terms, and a refined bound using a concave offline-regret term, plus lower bounds showing the bounds are tight up to logarithmic factors. The results imply that offline learnable classes become online learnable under predictable streams and connect online learning with predictions to transductive online learning, extending the AwP and smoothed online classification literature. Overall, the paper demonstrates that incorporating machine-learned predictions can dramatically ease online classification tasks when data streams exhibit predictability, while preserving worst-case guarantees when predictions are poor.
Abstract
We study online classification when the learner has access to predictions about future examples. We design an online learner whose expected regret is never worse than the worst-case regret, gracefully improves with the quality of the predictions, and can be significantly better than the worst-case regret when the predictions of future examples are accurate. As a corollary, we show that if the learner is always guaranteed to observe data where future examples are easily predictable, then online learning can be as easy as transductive online learning. Our results complement recent work in online algorithms with predictions and smoothed online classification, which go beyond a worse-case analysis by using machine-learned predictions and distributional assumptions respectively.