A Flat Minima Perspective on Understanding Augmentations and Model Robustness
Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (33) · 27879-27889
DOI: 10.1609/aaai.v40i33.40011 ↗Kurz gefasst
Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Weebum Yoo und 1 weiteren Autor:innen.
Abstract (Originalsprache)
Model robustness indicates a model’s capability to generalize well on unforeseen distributional shifts, including data corruptions and adversarial attacks. Data augmentation is one of the most prevalent and effective ways to enhance robustness. Despite the great success of the diverse augmentations in different fields, a unified theoretical understanding of their efficacy in improving model robustness is lacking. We theoretically reveal a general condition for label-preserving augmentations to bring robustness to diverse distribution shifts through the lens of flat minima and generalization bound, which de facto turns out to be strongly correlated with robustness against different distribution shifts in practice. Unlike most earlier works, our theoretical framework accommodates all the label-preserving augmentations and is not limited to particular distribution shifts. We substantiate our theories through different simulations on the existing common corruption and adversarial robustness benchmarks based on the CIFAR and ImageNet datasets.
Zitieren
APA
Yoo, W., & Yoon, S. W. (2026). A Flat Minima Perspective on Understanding Augmentations and Model Robustness. Proceedings of the AAAI Conference on Artificial Intelligence, 40(33), 27879-27889. https://doi.org/10.1609/aaai.v40i33.40011
Vancouver
Yoo W, Yoon SW. A Flat Minima Perspective on Understanding Augmentations and Model Robustness. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(33):27879-27889. doi:10.1609/aaai.v40i33.40011
BibTeX
@article{Yoo2026,
author = {Yoo, Weebum and Yoon, Sung Whan},
title = {A Flat Minima Perspective on Understanding Augmentations and Model Robustness},
journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
year = {2026},
volume = {40},
number = {33},
pages = {27879--27889},
doi = {10.1609/aaai.v40i33.40011}
}
EUVIMED