02/02/2021

Amata: An Annealing Mechanism for Adversarial Training Acceleration

Nanyang Ye, Qianxiao Li, Xiao-Yun Zhou, Zhanxing Zhu

Keywords:

Abstract Paper Similar Papers

Abstract: Despite the empirical success in various domains, it has been revealed that deep neural networks are vulnerable to maliciously perturbed input data that much degrade their performance. This is known as adversarial attacks. To counter adversarial attacks, adversarial training formulated as a form of robust optimization has been demonstrated to be effective. However, conducting adversarial training brings much computational overhead compared with standard training. In order to reduce the computational cost, we propose an annealing mechanism, Amata, to reduce the overhead associated with adversarial training. The proposed Amata is provably convergent, well-motivated from the lens of optimal control theory and can be combined with existing acceleration methods to further enhance performance. It is demonstrated that on standard datasets, Amata can achieve similar or better robustness with around 1/3 to 1/2 the computational time compared with traditional methods. In addition, Amata can be incorporated into other adversarial training acceleration algorithms (e.g. YOPO, Free, Fast, and ATTA), which leads to further reduction in computational time on large-scale problems.

The video of this talk cannot be embedded. You can watch it here:
https://slideslive.com/38948352
(Link will open in new window)
 0
 0
 0
 0
  Share    
This is an embedded video. Talk and the respective paper are published at AAAI 2021 virtual conference. If you are one of the authors of the paper and want to manage your upload, see the question "My papertalk has been externally embedded..." in the FAQ section.

Comments

Post Comment
no comments yet
code of conduct: tbd Characters remaining: 140

Similar Papers

 0
 0
 0
 0
 16:42 
 1
 0
 0
 0
 4:45 
 0
 0
 0
 0
 15:25 
 0
 0
 0
 0
 18:26 
 0
 0
 0
 0
 13:21 
 0
 0
 0
 0
 5:44 
 0
 0
 0
 0
 5:21 
 0
 0
 0
 0
 1:01 
 0
 0
 0
 0
 5:22 
 0
 0
 0
 0
 1:01 
 0
 0
 0
 0
 1:01 
 0
 0
 0
 0
 5:12 
 0
 0
 0
 0
 4:30 
 0
 0
 0
 0
 3:29 
 0
 0
 0
 0
 10:55 
 0
 0
 0
 0
 5:20 
 0
 0
 0
 0
 14:56 
 0
 0
 0
 0
 3:19 
 0
 0
 0
 0
 14:48 
 0
 0
 0
 0
 14:39 
 0
 0
 0
 0
 5:02 
 0
 0
 0
 0
 3:24 
 0
 0
 0
 0
 3:06 
 0
 0
 0
 0
 4:56 
 0
 0
 0
 0
 11:05 
 1
 0
 0
 0
 1:01 
 0
 0
 0
 0
 14:07 
 0
 1
 1
 0
 4:24 
 0
 0
 0
 0
 3:16 
 0
 0
 0
 0
 15:19 
 0
 0
 0
 0
 14:04 
 0
 0
 0
 0
 1:01 
 0
 0
 0
 0
 1:00 
 0
 0
 0
 0
 9:12 
 0
 0
 0
 0
 11:54 
 0
 0
 0
 0
 3:22 
 0
 0
 0
 0
 3:20 
 0
 0
 0
 0
 5:13 
 0
 0
 0
 0
 3:01 
 0
 0
 0
 0
 3:21