Gaussian class-conditional simplex loss for accurate, adversarially robust deep classifier training

被引:0
|
作者
Ali, Arslan [1 ]
Migliorati, Andrea [1 ]
Bianchi, Tiziano [1 ]
Magli, Enrico [1 ]
机构
[1] Politecn Torino, DET Dept Elect & Telecommun, Turin, Italy
关键词
Deep neural networks; Classification; Adversarial robustness; Adversarial training; Adversarial defense; NETWORKS;
D O I
10.1186/s13635-023-00137-0
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
In this work, we present the Gaussian Class-Conditional Simplex (GCCS) loss: a novel approach for training deep robust multiclass classifiers that improves over the state-of-the-art in terms of classification accuracy and adversarial robustness, with little extra cost for network training. The proposed method learns a mapping of the input classes onto Gaussian target distributions in a latent space such that a hyperplane can be used as the optimal decision surface. Instead of maximizing the likelihood of target labels for individual samples, our loss function pushes the network to produce feature distributions yielding high inter-class separation and low intra-class separation. The mean values of the learned distributions are centered on the vertices of a simplex such that each class is at the same distance from every other class. We show that the regularization of the latent space based on our approach yields excellent classification accuracy. Moreover, GCCS provides improved robustness against adversarial perturbations, outperforming models trained with conventional adversarial training (AT). In particular, our model learns a decision space that minimizes the presence of short paths toward neighboring decision regions. We provide a comprehensive empirical evaluation that shows how GCCS outperforms state-of-the-art approaches over challenging datasets for targeted and untargeted gradient-based, as well as gradient-free adversarial attacks, both in terms of classification accuracy and adversarial robustness.
引用
收藏
页数:17
相关论文
共 18 条
  • [1] Gaussian class-conditional simplex loss for accurate, adversarially robust deep classifier training
    Arslan Ali
    Andrea Migliorati
    Tiziano Bianchi
    Enrico Magli
    EURASIP Journal on Information Security, 2023
  • [2] Naive and Robust: Class-Conditional Independence in Human Classification Learning
    Jarecki, Jana B.
    Meder, Bjoern
    Nelson, Jonathan D.
    COGNITIVE SCIENCE, 2018, 42 (01) : 4 - 42
  • [3] A comparative study of PCA, ICA and class-conditional ICA for Naive Bayes classifier
    Fan, Liwei
    Poh, Kim Leng
    COMPUTATIONAL AND AMBIENT INTELLIGENCE, 2007, 4507 : 16 - +
  • [4] Robust Class-Conditional Distribution Alignment for Partial Domain Adaptation
    Choudhuri, Sandipan
    Sen, Arunabha
    FIFTY-SEVENTH ASILOMAR CONFERENCE ON SIGNALS, SYSTEMS & COMPUTERS, IEEECONF, 2023, : 530 - 536
  • [5] Class-conditional Importance Weighting for Deep Learning with Noisy Labels
    Nagarajan, Bhalaji
    Marques, Ricardo
    Mejia, Marcos
    Radeva, Petia
    PROCEEDINGS OF THE 17TH INTERNATIONAL JOINT CONFERENCE ON COMPUTER VISION, IMAGING AND COMPUTER GRAPHICS THEORY AND APPLICATIONS (VISAPP), VOL 5, 2022, : 679 - 686
  • [6] CCGG: A Deep Autoregressive Model for Class-Conditional Graph Generation
    Ommi, Yassaman
    Yousefabadi, Matin
    Faez, Faezeh
    Sabour, Amirmojtaba
    Baghshah, Mahdieh Soleymani
    Rabiee, Hamid R.
    COMPANION PROCEEDINGS OF THE WEB CONFERENCE 2022, WWW 2022 COMPANION, 2022, : 1092 - 1098
  • [7] Weighted class-conditional Bayesian network classifier parameter learning of chaos quantum particle swarm
    Liu J.
    Ding X.
    Zheng R.
    Wang B.
    Liu H.
    Wang Z.
    Xi Tong Gong Cheng Yu Dian Zi Ji Shu/Systems Engineering and Electronics, 2019, 41 (10): : 2304 - 2309
  • [8] Neural network based class-conditional probability density function using kernel trick for supervised classifier
    Gopi, E. S.
    Palanisamy, P.
    NEUROCOMPUTING, 2015, 154 : 225 - 229
  • [9] Class-Conditional Sharpness-Aware Minimization for Deep Long-Tailed Recognition
    Zhou, Zhipeng
    Li, Lanqing
    Zhao, Peilin
    Heng, Pheng-Ann
    Gong, Wei
    2023 IEEE/CVF CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION, CVPR, 2023, : 3499 - 3509
  • [10] A linear classifier for Gaussian class conditional distributions with unequal covariance matrices
    Vaswani, N
    16TH INTERNATIONAL CONFERENCE ON PATTERN RECOGNITION, VOL II, PROCEEDINGS, 2002, : 60 - 63