Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2602.17918 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866912971838455808 |
|---|---|
| author | Yu, Jialin Blanchard, Moïse |
| author_facet | Yu, Jialin Blanchard, Moïse |
| contents | 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 $μ$ 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 $μ$ 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_17918 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Distribution-Free Sequential Prediction with Abstentions Yu, Jialin Blanchard, Moïse Machine Learning Data Structures and Algorithms 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 $μ$ 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 $μ$ 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. |
| title | Distribution-Free Sequential Prediction with Abstentions |
| topic | Machine Learning Data Structures and Algorithms |
| url | https://arxiv.org/abs/2602.17918 |