Publication
AAAI 2021
Conference paper

Fast Training of Provably Robust Neural Networks by SingleProp

AAAI 2021
Download paper

Abstract

Recent works have developed several methods of defending neural networks against adversarial attacks with certified guarantees. However, these techniques can be computationally costly due to the use of certification during training. We develop a new regularizer that is both more efficient than existing certified defenses, requiring only one additional forward propagation through a network, and can be used to train networks with similar certified accuracy. Through experiments on MNIST and CIFAR-10 we demonstrate improvements in training speed and comparable certified accuracy compared to state-of-the-art certified defenses.

Date

02 Feb 2021

Publication

AAAI 2021

Authors

Tags

Resources

Share