NeurIPS 2024poster2 citations

Provable Tempered Overfitting of Minimal Nets and Typical Nets

Itamar Harel, William M. Hoza, Gal Vardi, Itay Evron, Nathan Srebro, Daniel Soudry

Abstract

We study the overfitting behavior of fully connected deep Neural Networks (NNs) with binary weights fitted to perfectly classify a noisy training set. We consider interpolation using both the smallest NN (having the minimal number of weights) and a random interpolating NN. For both learning rules, we prove overfitting is tempered. Our analysis rests on a new bound on the size of a threshold circuit consistent with a partial function. To the best of our knowledge, ours are the first theoretical results on benign or tempered overfitting that: (1) apply to deep NNs, and (2) do not require a very high or very low input dimension.

Deep LearningTempered OverfittingGeneralization
BibTeX
@inproceedings{
harel2024provable,
title={Provable Tempered Overfitting of Minimal Nets and Typical Nets},
author={Itamar Harel and William M. Hoza and Gal Vardi and Itay Evron and Nathan Srebro and Daniel Soudry},
booktitle={The Thirty-eighth Annual Conference on Neural Information Processing Systems},
year={2024},
url={https://openreview.net/forum?id=QyR1dNDxRP}
}
Provable Tempered Overfitting of Minimal Nets and Typical Nets · NeurIPS 2024