Parseval Networks: Improving Robustness to Adversarial Examples

Moustapha Cisse, Piotr Bojanowski, Edouard Grave, Yann Dauphin, Nicolas Usunier
Proceedings of the 34th International Conference on Machine Learning, PMLR 70:854-863, 2017.

Abstract

We introduce Parseval networks, a form of deep neural networks in which the Lipschitz constant of linear, convolutional and aggregation layers is constrained to be smaller than $1$. Parseval networks are empirically and theoretically motivated by an analysis of the robustness of the predictions made by deep neural networks when their input is subject to an adversarial perturbation. The most important feature of Parseval networks is to maintain weight matrices of linear and convolutional layers to be (approximately) Parseval tight frames, which are extensions of orthogonal matrices to non-square matrices. We describe how these constraints can be maintained efficiently during SGD. We show that Parseval networks match the state-of-the-art regarding accuracy on CIFAR-10/100 and Street View House Numbers (SVHN), while being more robust than their vanilla counterpart against adversarial examples. Incidentally, Parseval networks also tend to train faster and make a better usage of the full capacity of the networks.

Cite this Paper


BibTeX
@InProceedings{pmlr-v70-cisse17a, title = {Parseval Networks: Improving Robustness to Adversarial Examples}, author = {Moustapha Cisse and Piotr Bojanowski and Edouard Grave and Yann Dauphin and Nicolas Usunier}, booktitle = {Proceedings of the 34th International Conference on Machine Learning}, pages = {854--863}, year = {2017}, editor = {Precup, Doina and Teh, Yee Whye}, volume = {70}, series = {Proceedings of Machine Learning Research}, month = {06--11 Aug}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v70/cisse17a/cisse17a.pdf}, url = { http://proceedings.mlr.press/v70/cisse17a.html }, abstract = {We introduce Parseval networks, a form of deep neural networks in which the Lipschitz constant of linear, convolutional and aggregation layers is constrained to be smaller than $1$. Parseval networks are empirically and theoretically motivated by an analysis of the robustness of the predictions made by deep neural networks when their input is subject to an adversarial perturbation. The most important feature of Parseval networks is to maintain weight matrices of linear and convolutional layers to be (approximately) Parseval tight frames, which are extensions of orthogonal matrices to non-square matrices. We describe how these constraints can be maintained efficiently during SGD. We show that Parseval networks match the state-of-the-art regarding accuracy on CIFAR-10/100 and Street View House Numbers (SVHN), while being more robust than their vanilla counterpart against adversarial examples. Incidentally, Parseval networks also tend to train faster and make a better usage of the full capacity of the networks.} }
Endnote
%0 Conference Paper %T Parseval Networks: Improving Robustness to Adversarial Examples %A Moustapha Cisse %A Piotr Bojanowski %A Edouard Grave %A Yann Dauphin %A Nicolas Usunier %B Proceedings of the 34th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2017 %E Doina Precup %E Yee Whye Teh %F pmlr-v70-cisse17a %I PMLR %P 854--863 %U http://proceedings.mlr.press/v70/cisse17a.html %V 70 %X We introduce Parseval networks, a form of deep neural networks in which the Lipschitz constant of linear, convolutional and aggregation layers is constrained to be smaller than $1$. Parseval networks are empirically and theoretically motivated by an analysis of the robustness of the predictions made by deep neural networks when their input is subject to an adversarial perturbation. The most important feature of Parseval networks is to maintain weight matrices of linear and convolutional layers to be (approximately) Parseval tight frames, which are extensions of orthogonal matrices to non-square matrices. We describe how these constraints can be maintained efficiently during SGD. We show that Parseval networks match the state-of-the-art regarding accuracy on CIFAR-10/100 and Street View House Numbers (SVHN), while being more robust than their vanilla counterpart against adversarial examples. Incidentally, Parseval networks also tend to train faster and make a better usage of the full capacity of the networks.
APA
Cisse, M., Bojanowski, P., Grave, E., Dauphin, Y. & Usunier, N.. (2017). Parseval Networks: Improving Robustness to Adversarial Examples. Proceedings of the 34th International Conference on Machine Learning, in Proceedings of Machine Learning Research 70:854-863 Available from http://proceedings.mlr.press/v70/cisse17a.html .

Related Material