TY - JOUR
T1 - Distributed Adversarial Training to Robustify Deep Neural Networks at Scale
AU - Zhang, Gaoyuan
AU - Lu, Songtao
AU - Zhang, Yihua
AU - Chen, Xiangyi
AU - Chen, Pin Yu
AU - Fan, Quanfu
AU - Martie, Lee
AU - Horesh, Lior
AU - Hong, Mingyi
AU - Liu, Sijia
N1 - Publisher Copyright:
© 2022 UAI. All Rights Reserved.
PY - 2022
Y1 - 2022
N2 - Current deep neural networks (DNNs) are vulnerable to adversarial attacks, where adversarial perturbations to the inputs can change or manipulate classification. To defend against such attacks, an effective and popular approach, known as adversarial training (AT), has been shown to mitigate the negative impact of adversarial attacks by virtue of a min-max robust training method. While effective, it remains unclear whether it can successfully be adapted to the distributed learning context. The power of distributed optimization over multiple machines enables us to scale up robust training over large models and datasets. Spurred by that, we propose distributed adversarial training (DAT), a large-batch adversarial training framework implemented over multiple machines. We show that DAT is general, which supports training over labeled and unlabeled data, multiple types of attack generation methods, and gradient compression operations favored for distributed optimization. Theoretically, we provide, under standard conditions in the optimization theory, the convergence rate of DAT to the first-order stationary points in general non-convex settings. Empirically, we demonstrate that DAT either matches or outperforms state-of-the-art robust accuracies and achieves a graceful training speedup (e.g., on ResNet-50 under ImageNet). Codes are available at https://github.com/dat-2022/dat.
AB - Current deep neural networks (DNNs) are vulnerable to adversarial attacks, where adversarial perturbations to the inputs can change or manipulate classification. To defend against such attacks, an effective and popular approach, known as adversarial training (AT), has been shown to mitigate the negative impact of adversarial attacks by virtue of a min-max robust training method. While effective, it remains unclear whether it can successfully be adapted to the distributed learning context. The power of distributed optimization over multiple machines enables us to scale up robust training over large models and datasets. Spurred by that, we propose distributed adversarial training (DAT), a large-batch adversarial training framework implemented over multiple machines. We show that DAT is general, which supports training over labeled and unlabeled data, multiple types of attack generation methods, and gradient compression operations favored for distributed optimization. Theoretically, we provide, under standard conditions in the optimization theory, the convergence rate of DAT to the first-order stationary points in general non-convex settings. Empirically, we demonstrate that DAT either matches or outperforms state-of-the-art robust accuracies and achieves a graceful training speedup (e.g., on ResNet-50 under ImageNet). Codes are available at https://github.com/dat-2022/dat.
UR - http://www.scopus.com/inward/record.url?scp=85163356102&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85163356102&partnerID=8YFLogxK
M3 - Conference article
AN - SCOPUS:85163356102
SN - 2640-3498
VL - 180
SP - 2353
EP - 2363
JO - Proceedings of Machine Learning Research
JF - Proceedings of Machine Learning Research
T2 - 38th Conference on Uncertainty in Artificial Intelligence, UAI 2022
Y2 - 1 August 2022 through 5 August 2022
ER -