Publication Type

Conference Proceeding Article

Version

publishedVersion

Publication Date

4-2026

Abstract

Deep neural networks often exhibit substantial disparities in class-wise accuracy, even when trained on class-balanced data—posing concerns for reliable deployment. While prior efforts have explored empirical remedies, a theoretical understanding of such performance disparities in classification remains limited. In this work, we present Margin Regularization for performance disparity Reduction (MR2 ), a theoretically principled regularization for classification by dynamically adjusting margins in both the logit and representation spaces. Our analysis establishes a margin-based, class-sensitive generalization bound that reveals how per-class feature variability contributes to error, motivating the use of larger margins for “hard” classes. Guided by this insight, MR2 optimizes per-class logit margins proportional to feature spread and penalizes excessive representation margins to enhance intra-class compactness. Experiments on seven datasets—including ImageNet—and diverse pre-trained backbones (MAE, MoCov2, CLIP) demonstrate that our MR2 not only improves overall accuracy but also significantly boosts “hard” class performance without trading off “easy” classes, thus reducing the performance disparity. Codes are available in https://github.com/BeierZhu/MR2.

Keywords

image classification, class-wise performance gap

Discipline

Artificial Intelligence and Robotics | Databases and Information Systems

Research Areas

Intelligent Systems and Optimization

Areas of Excellence

Digital transformation

Publication

Proceedings of the Fourteenth International Conference on Learning Representations (ICLR 2026), Rio de Janeiro, Brazil, April 23-27

First Page

1

Last Page

29

Publisher

ICLR

City or Country

Rio de Janeiro, Brazil

Additional URL

https://openreview.net/forum?id=KfjpyOcPQj

Download

Share

COinS