Distribution-Free Sequential Prediction with Abstentions

TL;DR

This paper introduces a distribution-free sequential prediction algorithm with abstentions that guarantees sublinear error rates against oblivious and adaptive adversaries, extending learning guarantees to more realistic, non-i.i.d. settings.

Contribution

It proposes AbstainBoost, a boosting-based algorithm for distribution-free abstention learning, achieving error bounds without prior knowledge of the data distribution.

Findings

Guarantees sublinear error for VC classes in distribution-free settings.

Works against both oblivious and adaptive adversaries.

Establishes lower bounds on error-abstention trade-offs.

Abstract

We study a sequential prediction problem in which an adversary is allowed to inject arbitrarily many adversarial instances in a stream of i.i.d. instances, but at each round, the learner may also abstain from making a prediction without incurring any penalty if the instance was indeed corrupted. This semi-adversarial setting naturally sits between the classical stochastic case with i.i.d. instances for which function classes with finite VC dimension are learnable; and the adversarial case with arbitrary instances, known to be significantly more restrictive. For this problem, Goel et al. (2023) showed that, if the learner knows the distribution $\mu$ of clean samples in advance, learning can be achieved for all VC classes without restrictions on adversary corruptions. This is, however, a strong assumption in both theory and practice: a natural question is whether similar learning guarantees can be achieved without prior distributional knowledge, as is standard in classical learning frameworks (e.g., PAC learning or asymptotic consistency) and other non-i.i.d. models (e.g., smoothed online learning). We therefore focus on the distribution-free setting where $\mu$ is unknown and propose an algorithm AbstainBoost based on a boosting procedure of weak learners, which guarantees sublinear error for general VC classes in distribution-free abstention learning for oblivious adversaries. These algorithms also enjoy similar guarantees for adaptive adversaries, for structured function classes including linear classifiers. These results are complemented with corresponding lower bounds, which reveal an interesting polynomial trade-off between misclassification error and number of erroneous abstentions.