Laplacian networks: bounding indicator function smoothness for neural networks robustness - Ecole Nationale d'Ingénieurs de Brest Accéder directement au contenu
Article Dans Une Revue APSIPA Transactions on Signal and Information Processing Année : 2021

Laplacian networks: bounding indicator function smoothness for neural networks robustness

Résumé

For the past few years, deep learning (DL) robustness (i.e. the ability to maintain the same decision when inputs are subject to perturbations) has become a question of paramount importance, in particular in settings where misclassification can have dramatic consequences. To address this question, authors have proposed different approaches, such as adding regularizers or training using noisy examples. In this paper we introduce a regularizer based on the Laplacian of similarity graphs obtained from the representation of training data at each layer of the DL architecture. This regularizer penalizes large changes (across consecutive layers in the architecture) in the distance between examples of different classes, and as such enforces smooth variations of the class boundaries. We provide theoretical justification for this regularizer and demonstrate its effectiveness to improve robustness on classical supervised learning vision datasets for various types of perturbations. We also show it can be combined with existing methods to increase overall robustness.
Fichier principal
Vignette du fichier
laplacian-networks-bounding-indicator-function-smoothness-for-neural-networks-robustness.pdf (733.81 Ko) Télécharger le fichier
Origine : Fichiers éditeurs autorisés sur une archive ouverte

Dates et versions

hal-03134179 , version 1 (28-07-2021)

Licence

Paternité

Identifiants

Citer

Carlos Lassance, Vincent Gripon, Antonio Ortega. Laplacian networks: bounding indicator function smoothness for neural networks robustness. APSIPA Transactions on Signal and Information Processing, 2021, 10, pp.e2. ⟨10.1017/ATSIP.2021.2⟩. ⟨hal-03134179⟩
53 Consultations
39 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More