Certifiable Metric One Class Learning with Adversarially Trained Lipschitz Classifier

Abstract

We propose a new Novelty Detection and One Class classifier, based on the smoothness properties of orthogonal neural network, and on the properties of Hinge Kantorovich Rubinstein (HKR) function. The classifier benefits from robustness certificates against $l2$-attacks thanks to the Lipschitz constraint, whilst the HKR loss allows to provably approximate the signed distance function to the boundary of the distribution: the normality score induces by the classifier has a meaningful interpretation in term of distance to the support. Finally, gradient steps in the input space allows free generation of samples from the one class in a fashion that reminds GAN or VAE.

PDF NeurIPSW OpenReview Semantic Scholar

Cite

Text

Béthune and Serrurier. "Certifiable Metric One Class Learning with Adversarially Trained Lipschitz Classifier." NeurIPS 2022 Workshops: MLSW, 2022.

Markdown

[Béthune and Serrurier. "Certifiable Metric One Class Learning with Adversarially Trained Lipschitz Classifier." NeurIPS 2022 Workshops: MLSW, 2022.](https://mlanthology.org/neuripsw/2022/bethune2022neuripsw-certifiable/)

BibTeX

@inproceedings{bethune2022neuripsw-certifiable,
  title     = {{Certifiable Metric One Class Learning with Adversarially Trained Lipschitz Classifier}},
  author    = {Béthune, Louis and Serrurier, Mathieu},
  booktitle = {NeurIPS 2022 Workshops: MLSW},
  year      = {2022},
  url       = {https://mlanthology.org/neuripsw/2022/bethune2022neuripsw-certifiable/}
}