Expressive Losses for Verified Robustness via Convex Combinations - Inria - Institut national de recherche en sciences et technologies du numérique Accéder directement au contenu
Communication Dans Un Congrès Année : 2024

Expressive Losses for Verified Robustness via Convex Combinations

Résumé

In order to train networks for verified adversarial robustness, it is common to over-approximate the worst-case loss over perturbation regions, resulting in networks that attain verifiability at the expense of standard performance. As shown in recent work, better trade-offs between accuracy and robustness can be obtained by carefully coupling adversarial training with over-approximations. We hypothesize that the expressivity of a loss function, which we formalize as the ability to span a range of trade-offs between lower and upper bounds to the worst-case loss through a single parameter (the over-approximation coefficient), is key to attaining state-of-the-art performance. To support our hypothesis, we show that trivial expressive losses, obtained via convex combinations between adversarial attacks and IBP bounds, yield state-of-the-art results across a variety of settings in spite of their conceptual simplicity. We provide a detailed analysis of the relationship between the over-approximation coefficient and performance profiles across different expressive losses, showing that, while expressivity is essential, better approximations of the worst-case loss are not necessarily linked to superior robustness-accuracy trade-offs.
Fichier principal
Vignette du fichier
8030_expressive_losses_for_verified.pdf (525.99 Ko) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04511820 , version 1 (19-03-2024)

Licence

Paternité

Identifiants

  • HAL Id : hal-04511820 , version 1

Citer

Alessandro De Palma, Rudy Bunel, Krishnamurthy Dvijotham, M Pawan Kumar, Robert Stanforth, et al.. Expressive Losses for Verified Robustness via Convex Combinations. ICLR 2024 - International Conference on Learning Representations, May 2024, Vienna, Austria. ⟨hal-04511820⟩
0 Consultations
0 Téléchargements

Partager

Gmail Facebook X LinkedIn More