Large-Margin Softmax Loss for Convolutional Neural Networks

被引:0
|
作者
Liu, Weiyang [1 ]
Wen, Yandong [2 ]
Yu, Zhiding [3 ]
Yang, Meng [4 ]
机构
[1] Peking Univ, Sch ECE, Beijing, Peoples R China
[2] South China Univ Technol, Sch EIE, Guangzhou, Peoples R China
[3] Carnegie Mellon Univ, Dept ECE, Pittsburgh, PA 15213 USA
[4] Shenzhen Univ, Coll CS & SE, Shenzhen, Peoples R China
基金
美国国家科学基金会;
关键词
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Cross-entropy loss together with softmax is arguably one of the most common used supervision components in convolutional neural networks (CNNs). Despite its simplicity, popularity and excellent performance, the component does not explicitly encourage discriminative learning of features. In this paper, we propose a generalized large-margin softmax (L-Softmax) loss which explicitly encourages intra-class compactness and inter-class separability between learned features. Moreover, L-Softmax not only can adjust the desired margin but also can avoid overfitting. We also show that the L-Softmax loss can be optimized by typical stochastic gradient descent. Extensive experiments on four benchmark datasets demonstrate that the deeply-learned features with L-softmax loss become more discriminative, hence significantly boosting the performance on a variety of visual classification and verification tasks.
引用
收藏
页数:10
相关论文
共 50 条
  • [1] Convolutional Neural Networks with Large-Margin Softmax Loss Function for Cognitive Load Recognition
    Liu, Yuetian
    Liu, Qingshan
    [J]. PROCEEDINGS OF THE 36TH CHINESE CONTROL CONFERENCE (CCC 2017), 2017, : 4045 - 4049
  • [2] IMPROVED LARGE-MARGIN SOFTMAX LOSS FOR SPEAKER DIARISATION
    Fathullah, Y.
    Zhang, C.
    Woodland, P. C.
    [J]. 2020 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH, AND SIGNAL PROCESSING, 2020, : 7104 - 7108
  • [3] Investigation of Large-Margin Softmax in Neural Language Modeling
    Huo, Jingjing
    Gao, Yingbo
    Wang, Weiyue
    Schlueter, Ralf
    Ney, Hermann
    [J]. INTERSPEECH 2020, 2020, : 3645 - 3649
  • [4] Large-Margin Regularized Softmax Cross-Entropy Loss
    Li, Xiaoxu
    Chang, Dongliang
    Tian, Tao
    Cao, Jie
    [J]. IEEE ACCESS, 2019, 7 : 19572 - 19578
  • [5] Large-Margin Classification in Infinite Neural Networks
    Cho, Youngmin
    Saul, Lawrence K.
    [J]. NEURAL COMPUTATION, 2010, 22 (10) : 2678 - 2697
  • [6] Integrate Receptive Field Block into Large-margin Softmax Loss for Face Recognition
    Wei, Yi
    Pu, Haibo
    Zhu, Yu
    Li, XiaoFan
    [J]. 2019 3RD INTERNATIONAL CONFERENCE ON MACHINE VISION AND INFORMATION TECHNOLOGY (CMVIT 2019), 2019, 1229
  • [7] Advancing neural network calibration: The role of gradient decay in large-margin Softmax optimization
    Zhang, Siyuan
    Xie, Linbo
    [J]. NEURAL NETWORKS, 2024, 178
  • [8] Large Margin Softmax Loss for Speaker Verification
    Liu, Yi
    He, Liang
    Liu, Jia
    [J]. INTERSPEECH 2019, 2019, : 2873 - 2877
  • [9] Research on Additive Margin Softmax Speaker Recognition Based on Convolutional and Gated Recurrent Neural Networks
    Lan, Chaofeng
    Wang, Yuqiao
    Zhang, Lei
    Zhao, Hongyun
    [J]. AES: Journal of the Audio Engineering Society, 2022, 70 (7-8): : 611 - 620
  • [10] Research on Additive Margin Softmax Speaker Recognition Based on Convolutional and Gated Recurrent Neural Networks
    Lan, Chaofeng
    Wang, Yuqiao
    Zhang, Lei
    Zhao, Hongyun
    [J]. JOURNAL OF THE AUDIO ENGINEERING SOCIETY, 2022, 70 (7-8): : 611 - 620