On Lipschitz Regularization of Convolutional Layers using Toeplitz Matrix Theory - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2021

On Lipschitz Regularization of Convolutional Layers using Toeplitz Matrix Theory

Résumé

This paper tackles the problem of Lipschitz regularization of Convolutional Neural Networks. Lipschitz regularity is now established as a key property of modern deep learning with implications in training stability, generalization, robustness against adversarial examples, etc. However, computing the exact value of the Lipschitz constant of a neural network is known to be NP-hard. Recent attempts from the literature introduce upper bounds to approximate this constant that are either efficient but loose or accurate but computationally expensive. In this work, by leveraging the theory of Toeplitz matrices, we introduce a new upper bound for convolutional layers that is both tight and easy to compute. Based on this result we devise an algorithm to train Lipschitz regularized Convolutional Neural Networks.

Dates et versions

hal-03107420 , version 1 (12-01-2021)

Identifiants

Citer

Alexandre Araujo, Benjamin Negrevergne, Yann Chevaleyre, Jamal Atif. On Lipschitz Regularization of Convolutional Layers using Toeplitz Matrix Theory. 35th AAAI Conference on Artificial Intelligence, Feb 2021, vancouver, Canada. ⟨hal-03107420⟩
85 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More