TY - GEN
T1 - Enhanced Accuracy and Robustness via Multi-teacher Adversarial Distillation
AU - Zhao, Shiji
AU - Yu, Jie
AU - Sun, Zhenlong
AU - Zhang, Bo
AU - Wei, Xingxing
N1 - Publisher Copyright:
© 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.
PY - 2022
Y1 - 2022
N2 - Adversarial training is an effective approach for improving the robustness of deep neural networks against adversarial attacks. Although bringing reliable robustness, adversarial training (AT) will reduce the performance of identifying clean examples. Meanwhile, Adversarial training can bring more robustness for large models than small models. To improve the robust and clean accuracy of small models, we introduce the Multi-Teacher Adversarial Robustness Distillation (MTARD) to guide the adversarial training process of small models. Specifically, MTARD uses multiple large teacher models, including an adversarial teacher and a clean teacher to guide a small student model in the adversarial training by knowledge distillation. In addition, we design a dynamic training algorithm to balance the influence between the adversarial teacher and clean teacher models. A series of experiments demonstrate that our MTARD can outperform the state-of-the-art adversarial training and distillation methods against various adversarial attacks. Our code is available at https://github.com/zhaoshiji123/MTARD.
AB - Adversarial training is an effective approach for improving the robustness of deep neural networks against adversarial attacks. Although bringing reliable robustness, adversarial training (AT) will reduce the performance of identifying clean examples. Meanwhile, Adversarial training can bring more robustness for large models than small models. To improve the robust and clean accuracy of small models, we introduce the Multi-Teacher Adversarial Robustness Distillation (MTARD) to guide the adversarial training process of small models. Specifically, MTARD uses multiple large teacher models, including an adversarial teacher and a clean teacher to guide a small student model in the adversarial training by knowledge distillation. In addition, we design a dynamic training algorithm to balance the influence between the adversarial teacher and clean teacher models. A series of experiments demonstrate that our MTARD can outperform the state-of-the-art adversarial training and distillation methods against various adversarial attacks. Our code is available at https://github.com/zhaoshiji123/MTARD.
KW - Adversarial training
KW - DNNs
KW - Knowledge distillation
UR - https://www.scopus.com/pages/publications/85142686699
U2 - 10.1007/978-3-031-19772-7_34
DO - 10.1007/978-3-031-19772-7_34
M3 - 会议稿件
AN - SCOPUS:85142686699
SN - 9783031197710
T3 - Lecture Notes in Computer Science
SP - 585
EP - 602
BT - Computer Vision – ECCV 2022 - 17th European Conference, Proceedings
A2 - Avidan, Shai
A2 - Brostow, Gabriel
A2 - Cissé, Moustapha
A2 - Farinella, Giovanni Maria
A2 - Hassner, Tal
PB - Springer Science and Business Media Deutschland GmbH
T2 - 17th European Conference on Computer Vision, ECCV 2022
Y2 - 23 October 2022 through 27 October 2022
ER -