Adversarial Robustness via Label-Smoothing - Archive ouverte HAL Accéder directement au contenu
Pré-Publication, Document De Travail Année : 2020

Adversarial Robustness via Label-Smoothing

Robustesse Adversaire via Label-Smoothing

Morgane Goibert
Elvis Dohmatob
  • Fonction : Auteur
  • PersonId : 956319

Résumé

We study Label-Smoothing as a means for improving adversarial robustness of supervised deep-learning models. After establishing a thorough and unified framework, we propose several variations to this general method: adversarial, Boltzmann and second-best Label-Smoothing methods, and we explain how to construct your own one. On various datasets (MNIST, CIFAR10, SVHN) and models (linear models, MLPs, LeNet, ResNet), we show that Label-Smoothing in general improves adversarial robustness against a variety of attacks (FGSM, BIM, DeepFool, Carlini-Wagner) by better taking account of the dataset geometry. The proposed Label-Smoothing methods have two main advantages: they can be implemented as a modified cross-entropy loss, thus do not require any modifications of the network architecture nor do they lead to increased training times, and they improve both standard and adversarial accuracy.
Fichier principal
Vignette du fichier
Article_adversarial_label_smoothing.pdf (3.59 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

hal-02437752 , version 1 (13-01-2020)

Identifiants

  • HAL Id : hal-02437752 , version 1

Citer

Morgane Goibert, Elvis Dohmatob. Adversarial Robustness via Label-Smoothing. 2020. ⟨hal-02437752⟩
179 Consultations
222 Téléchargements

Partager

Gmail Facebook X LinkedIn More