cispa_all_3595.pdf (956.48 kB)

Relative Flatness and Generalization

Download (956.48 kB)
conference contribution
posted on 2023-11-29, 18:18 authored by Henning Petzka, Michael Kamp, Linara Adilova, Cristian Sminchisescu, Mario Boley
Flatness of the loss curve is conjectured to be connected to the generalization ability of machine learning models, in particular neural networks. While it has been empirically observed that flatness measures consistently correlate strongly with generalization, it is still an open theoretical problem why and under which circumstances flatness is connected to generalization, in particular in light of reparameterizations that change certain flatness measures but leave generalization unchanged. We investigate the connection between flatness and generalization by relating it to the interpolation from representative data, deriving notions of representativeness, and feature robustness. The notions allow us to rigorously connect flatness and generalization and to identify conditions under which the connection holds. Moreover, they give rise to a novel, but natural relative flatness measure that correlates strongly with generalization, simplifies to ridge regression for ordinary least squares, and solves the reparameterization issue.


Preferred Citation

Henning Petzka, Michael Kamp, Linara Adilova, Cristian Sminchisescu and Mario Boley. Relative Flatness and Generalization. In: Conference on Neural Information Processing Systems (NeurIPS). 2021.

Primary Research Area

  • Empirical and Behavioral Security

Name of Conference

Conference on Neural Information Processing Systems (NeurIPS)

Legacy Posted Date


Open Access Type

  • Green


@inproceedings{cispa_all_3595, title = "Relative Flatness and Generalization", author = "Petzka, Henning and Kamp, Michael and Adilova, Linara and Sminchisescu, Cristian and Boley, Mario", booktitle="{Conference on Neural Information Processing Systems (NeurIPS)}", year="2021", }

Usage metrics


    No categories selected


    Ref. manager