• 中国精品科技期刊
  • CCF推荐A类中文期刊
  • 计算领域高质量科技期刊T1类
Advanced Search
Zhang Qiang, Yang Jibin, Zhang Xiongwei, Cao Tieyong, Zheng Changyan. CS-Softmax: A Cosine Similarity-Based Softmax Loss Function[J]. Journal of Computer Research and Development, 2022, 59(4): 936-949. DOI: 10.7544/issn1000-1239.20200879
Citation: Zhang Qiang, Yang Jibin, Zhang Xiongwei, Cao Tieyong, Zheng Changyan. CS-Softmax: A Cosine Similarity-Based Softmax Loss Function[J]. Journal of Computer Research and Development, 2022, 59(4): 936-949. DOI: 10.7544/issn1000-1239.20200879

CS-Softmax: A Cosine Similarity-Based Softmax Loss Function

Funds: This work was supported by the National Natural Science Foundation of China (61602031), the Fundamental Research Funds for the Central Universities (FRF-BD-19-012A, FRF-IDRY-19-023), and the National Key 
More Information
  • Published Date: March 31, 2022
  • Convolutional neural networks (CNNs)-based classification framework has achieved significant effects in pattern classification tasks, where the Softmax function with the cross-entropy loss (Softmax loss) can make CNNs learn separable embeddings. However, for some multi-classification problems, training with Softmax loss does not encourage increasing intra-class compactness and inter-class separability, which means it hardly generates the embedding with strong discriminability, making it hard to improve the performance further. In order to enhance the discriminability of learned embeddings, a cosine similarity-based Softmax (CS-Softmax) loss function is proposed. Without changing the network structure, the CS-Softmax loss introduces some parameters such as margin factor, scale factor and weight update factor to calculate the positive similarity and negative similarity between embeddings and different class weights based on the Softmax loss, so as to achieve the objectives of enhancing intra-class compactness and inter-class separability. Furthermore, the size of classification decision margin can be modified flexibly. These characteristics further enhance the discriminability of learned embeddings in CNNs. Classification experimental results on typical audio and image datasets show that the CS-Softmax loss can effectively improve the classification performance without increasing the computational complexity. The classification accuracies of the proposed loss are 99.81%, 95.46%, and 76.46% on the MNIST, CIFAR10, and CIFAR100 classification tasks, respectively.
  • Related Articles

    [1]Chen Kejia, Lu Hao, Zhang Jiajun. Conditional Variational Time-Series Graph Auto-Encoder[J]. Journal of Computer Research and Development, 2020, 57(8): 1663-1673. DOI: 10.7544/issn1000-1239.2020.20200202
    [2]Geng Pu, Zhu Yuefei. A Code Encrypt Technique Based on Branch Condition Obfuscation[J]. Journal of Computer Research and Development, 2019, 56(10): 2183-2192. DOI: 10.7544/issn1000-1239.2019.20190368
    [3]Shi Haihe, Zhou Weixing. Design and Implementation of Pairwise Sequence Alignment Algorithm Components Based on Dynamic Programming[J]. Journal of Computer Research and Development, 2019, 56(9): 1907-1917. DOI: 10.7544/issn1000-1239.2019.20180835
    [4]Zhou Yanhong, Zhang Xianyong, Mo Zhiwen. Conditional Neighborhood Entropy with Granulation Monotonicity and Its Relevant Attribute Reduction[J]. Journal of Computer Research and Development, 2018, 55(11): 2395-2405. DOI: 10.7544/issn1000-1239.2018.20170607
    [5]Guo Xi, Wang Pan. Variable Dependent Relation Analysis in Program State Condition Merging[J]. Journal of Computer Research and Development, 2018, 55(10): 2331-2342. DOI: 10.7544/issn1000-1239.2018.20170545
    [6]Zhao Liang, Wang Yongli, Du Zhongshu, Chen Guangsheng. HL-DAQ: A Dynamic Adaptive Quantization Coding for Hash Learning[J]. Journal of Computer Research and Development, 2018, 55(6): 1294-1307. DOI: 10.7544/issn1000-1239.2018.20170238
    [7]Liu Xianmin, Li Jianzhong. Discovering Extended Conditional Functional Dependencies[J]. Journal of Computer Research and Development, 2015, 52(1): 130-140. DOI: 10.7544/issn1000-1239.2015.20130691
    [8]Yang Wenguo, Guo Tiande, Zhao Tong. Routing Algorithms of the Wireless Sensor Network Based on Dynamic Programming[J]. Journal of Computer Research and Development, 2007, 44(5): 890-897.
    [9]Xia Yimin, Luo Jun, and Zhang Minxuan. Detecting Out-of-Bounds Accesses with Conditional Range Constraint[J]. Journal of Computer Research and Development, 2006, 43(10): 1760-1766.
    [10]Jiang Yuncheng, Shi Zhongzhi, Tang Yong, Wang Ju. A Distributed Dynamic Description Logic[J]. Journal of Computer Research and Development, 2006, 43(9): 1603-1608.
  • Cited by

    Periodical cited type(5)

    1. 蒋志荣,李亚男. 跨层安全访问多模态异构网络数据的数学建模. 计算机仿真. 2024(04): 475-479 .
    2. 褚治广,李俊燕,陈昊,张兴. 基于分布式多关联属性的高维数据差分隐私保护方法. 计算机工程与设计. 2024(04): 967-973 .
    3. 卢晓天,朴春慧,杨兴雨,白英杰. 基于贝叶斯网络的差分隐私高维数据发布技术研究. 计算机工程. 2024(05): 167-181 .
    4. 梁丹凝,梁坚. 基于Relief算法的不平衡数据分类分级算法仿真. 计算机仿真. 2024(06): 477-480+497 .
    5. 张小玉,沈国华,杨阳. 基于属性分割的差分隐私异构多属性数据发布. 计算机系统应用. 2022(10): 225-235 .

    Other cited types(12)

Catalog

    Article views (965) PDF downloads (383) Cited by(17)

    /

    DownLoad:  Full-Size Img  PowerPoint
    Return
    Return