NeurIPS 2021poster14 citations
Interpolation can hurt robust generalization even when there is no noise
Konstantin Donhauser, Alexandru Tifrea, Michael Aerni, Reinhard Heckel, Fanny Yang
Abstract
Numerous recent works show that overparameterization implicitly reduces variance for min-norm interpolators and max-margin classifiers. These findings suggest that ridge regularization has vanishing benefits in high dimensions. We challenge this narrative by showing that, even in the absence of noise, avoiding interpolation through ridge regularization can significantly improve generalization. We prove this phenomenon for the robust risk of both linear regression and classification, and hence provide the first theoretical result on \emph{robust overfitting}.
regularizationhigh dimensional statisticslearning theoryrobustness
BibTeX
@inproceedings{
donhauser2021interpolation,
title={Interpolation can hurt robust generalization even when there is no noise},
author={Konstantin Donhauser and Alexandru Tifrea and Michael Aerni and Reinhard Heckel and Fanny Yang},
booktitle={Advances in Neural Information Processing Systems},
editor={A. Beygelzimer and Y. Dauphin and P. Liang and J. Wortman Vaughan},
year={2021},
url={https://openreview.net/forum?id=nNfj0pVn4Q}
}