Timezone: »

 
Poster
Efficient Proximal Mapping of the 1-path-norm of Shallow Networks
Fabian Latorre · Paul Rolland · Shaul Nadav Hallak · Volkan Cevher

Thu Jul 16 01:00 PM -- 01:45 PM & Fri Jul 17 12:00 AM -- 12:45 AM (PDT) @
in Poster Session 46 »

We demonstrate two new important properties of the 1-path-norm of shallow neural networks. First, despite its non-smoothness and non-convexity it allows a closed form proximal operator which can be efficiently computed, allowing the use of stochastic proximal-gradient-type methods for regularized empirical risk minimization. Second, when the activation functions is differentiable, it provides an upper bound on the Lipschitz constant of the network. Such bound is tighter than the trivial layer-wise product of Lipschitz constants, motivating its use for training networks robust to adversarial perturbations. In practical experiments we illustrate the advantages of using the proximal mapping and we compare the robustness-accuracy trade-off induced by the 1-path-norm, L1-norm and layer-wise constraints on the Lipschitz constant (Parseval networks).

Keywords: Adversarial Examples · Algorithms · Non-convex Optimization · Optimization · Deep Learning - Algorithms

Author Information

Fabian Latorre (EPFL)
Paul Rolland (Ecole Polytechnique Fédérale de Lausanne)
Shaul Nadav Hallak (EPFL)
Volkan Cevher (EPFL)

More from the Same Authors