Weight-based Regularization for Improving Robustness in Image Classification

被引:0
|
作者
Yang, Hao [1 ]
Wang, Min [1 ]
Yu, Zhengfei [1 ]
Zhou, Yun [1 ]
机构
[1] Natl Univ Def Technol, Coll Syst Engn, Changsha, Peoples R China
来源
2023 IEEE INTERNATIONAL CONFERENCE ON MULTIMEDIA AND EXPO, ICME | 2023年
基金
中国国家自然科学基金;
关键词
stochastic neural network; adversarial robustness; adversarial examples;
D O I
10.1109/ICME55011.2023.00305
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Deep Neural Networks (DNNs) are known to be vulnerable to adversarial attacks. Recently, Stochastic Neural Networks (SNNs) have been proposed to enhance adversarial robustness by injecting uncertainty into the models. However, existing SNNs often inspired by intuition and rely on adversarial training, which is computationally costly. To address this issue, we propose a novel SNN called the Weight-based Stochastic Neural Network (WB-SNN), which is based on optimizing an error upper bound of adversarial robustness from the perspective of weight distribution. To the best of our knowledge, we are the first to propose a theoretically guaranteed weight-based stochastic neural network without relying on adversarial training. In comparison to normal adversarial training, our method saves about three times the computation cost. Extensive experiments on various datasets, networks, and adversarial attacks have demonstrated the effectiveness of the proposed method.
引用
收藏
页码:1775 / 1780
页数:6
相关论文
共 50 条
  • [21] Improving Adversarial Robustness of Detector via Objectness Regularization
    Bao, Jiayu
    Chen, Jiansheng
    Ma, Hongbing
    Ma, Huimin
    Yu, Cheng
    Huang, Yiqing
    PATTERN RECOGNITION AND COMPUTER VISION, PT IV, 2021, 13022 : 252 - 262
  • [22] Classification-based techniques for improving the robustness of CELP coders
    Jelinek, Milan
    Eksler, Vaclav
    Lemyre, Catherine
    Lefebvre, Roch
    CONFERENCE RECORD OF THE FORTY-FIRST ASILOMAR CONFERENCE ON SIGNALS, SYSTEMS & COMPUTERS, VOLS 1-5, 2007, : 1480 - 1484
  • [23] A Weight-Based Clustering Method
    Wang, Zhao-Yu
    Lee, Shie-Jue
    Pan, Shing-Tai
    4TH ANNUAL INTERNATIONAL CONFERENCE ON INFORMATION TECHNOLOGY AND APPLICATIONS (ITA 2017), 2017, 12
  • [24] Weight-based stigmatization in children
    Kraig, KA
    Keel, PK
    INTERNATIONAL JOURNAL OF OBESITY, 2001, 25 (11) : 1661 - 1666
  • [25] LANDFILL TAX TO BE WEIGHT-BASED
    不详
    CHEMICAL ENGINEER-LONDON, 1995, (594): : 9 - 9
  • [26] Weight-based quality specialization
    Lashkaripour, Ahmad
    JOURNAL OF INTERNATIONAL ECONOMICS, 2020, 127
  • [27] LANDFILL TAX TO BE WEIGHT-BASED
    不详
    PROCESS ENGINEERING, 1995, 76 (09) : 7 - 7
  • [28] Weight-based stigmatization in children
    KA Kraig
    PK Keel
    International Journal of Obesity, 2001, 25 : 1661 - 1666
  • [29] Weight-based low-molecular-weight heparin versus weight-based intravenous unfractionated heparin - Reponse
    Segal, Jodi B.
    Streiff, Michael B.
    Bass, Fric B.
    ANNALS OF INTERNAL MEDICINE, 2007, 147 (06) : 434 - 434
  • [30] A Universal Framework for Improving the Robustness of Coverless Image Steganography Based on Image Restoration
    Meng, Laijin
    Li, Fan
    Jiang, Xinghao
    Xu, Qiang
    IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY, 2025, 35 (01) : 922 - 937