On the Interpretability of Regularisation for Neural Networks Through Model Gradient Similarity

被引:0
|
作者
Szolnoky, Vincent [1 ]
Andersson, Viktor [2 ]
Kulcsar, Balazs [2 ]
Jornsten, Rebecka [1 ]
机构
[1] Chalmers Univ Technol, Dept Math Sci, Chalmers Tvargata 3, S-41296 Gothenburg, Sweden
[2] Chalmers Univ Technol, Dept Elect Engn, Chalmersplatsen 4, S-41296 Gothenburg, Sweden
基金
瑞典研究理事会;
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
Most complex machine learning and modelling techniques are prone to over-fitting and may subsequently generalise poorly to future data. Artificial neural networks are no different in this regard and, despite having a level of implicit regularisation when trained with gradient descent, often require the aid of explicit regularisers. We introduce a new framework, Model Gradient Similarity (MGS), that (1) serves as a metric of regularisation, which can be used to monitor neural network training, (2) adds insight into how explicit regularisers, while derived from widely different principles, operate via the same mechanism underneath by increasing MGS, and (3) provides the basis for a new regularisation scheme which exhibits excellent performance, especially in challenging settings such as high levels of label noise or limited sample sizes.
引用
收藏
页数:12
相关论文
共 50 条
  • [1] On the interpretability of quantum neural networks
    Pira, Lirande
    Ferrie, Chris
    QUANTUM MACHINE INTELLIGENCE, 2024, 6 (02)
  • [2] Radial Basis Function Networks for Convolutional Neural Networks to Learn Similarity Distance Metric and Improve Interpretability
    Amirian, Mohammadreza
    Schwenker, Friedhelm
    IEEE ACCESS, 2020, 8 : 123087 - 123097
  • [3] Regularisation of neural networks by enforcing Lipschitz continuity
    Henry Gouk
    Eibe Frank
    Bernhard Pfahringer
    Michael J. Cree
    Machine Learning, 2021, 110 : 393 - 416
  • [4] On Interpretability of Artificial Neural Networks: A Survey
    Fan, Feng-Lei
    Xiong, Jinjun
    Li, Mengzhou
    Wang, Ge
    IEEE TRANSACTIONS ON RADIATION AND PLASMA MEDICAL SCIENCES, 2021, 5 (06) : 741 - 760
  • [5] Regularisation of neural networks by enforcing Lipschitz continuity
    Gouk, Henry
    Frank, Eibe
    Pfahringer, Bernhard
    Cree, Michael J.
    MACHINE LEARNING, 2021, 110 (02) : 393 - 416
  • [6] Engravings, Secrets, and Interpretability of Neural Networks
    Hobbs, Nathaniel
    Papakonstantinou, Periklis A.
    Vaidya, Jaideep
    IEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTING, 2024, 12 (04) : 1093 - 1104
  • [7] Graph Convolutional Neural Networks Optimized by Momentum Cosine Similarity Gradient
    Yan, Jianhong
    Duan, Yunhui
    Computer Engineering and Applications, 2024, 60 (14) : 133 - 143
  • [8] Explainable Neural Networks: Achieving Interpretability in Neural Models
    Chakraborty, Manomita
    ARCHIVES OF COMPUTATIONAL METHODS IN ENGINEERING, 2024, 31 (06) : 3535 - 3550
  • [9] A canonical model for gradient frequency neural networks
    Large, Edward W.
    Almonte, Felix V.
    Velasco, Marc J.
    PHYSICA D-NONLINEAR PHENOMENA, 2010, 239 (12) : 905 - 911
  • [10] MaxGain: Regularisation of Neural Networks by Constraining Activation Magnitudes
    Gouk, Henry
    Pfahringer, Bernhard
    Frank, Eibe
    Cree, Michael J.
    MACHINE LEARNING AND KNOWLEDGE DISCOVERY IN DATABASES, ECML PKDD 2018, PT I, 2019, 11051 : 541 - 556