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
  • [41] Optimizing for Interpretability in Deep Neural Networks with Tree Regularization
    Wu, Mike
    Parbhoo, Sonali
    Hughes, Michael C.
    Roth, Volker
    Doshi-Velez, Finale
    JOURNAL OF ARTIFICIAL INTELLIGENCE RESEARCH, 2021, 72 : 1 - 37
  • [42] Improving the Interpretability of Deep Neural Networks with Knowledge Distillation
    Liu, Xuan
    Wang, Xiaoguang
    Matwin, Stan
    2018 18TH IEEE INTERNATIONAL CONFERENCE ON DATA MINING WORKSHOPS (ICDMW), 2018, : 905 - 912
  • [43] Similarity learning with neural networks
    Sanfins, G.
    Ramos, F.
    Naiff, D.
    PHYSICAL REVIEW E, 2025, 111 (02)
  • [44] Optimizing Deep Neural Networks Through Neuroevolution With Stochastic Gradient Descent
    Zhang, Haichao
    Hao, Kuangrong
    Gao, Lei
    Wei, Bing
    Tang, Xuesong
    IEEE TRANSACTIONS ON COGNITIVE AND DEVELOPMENTAL SYSTEMS, 2023, 15 (01) : 111 - 121
  • [45] Training of photonic neural networks through in situ backpropagation and gradient measurement
    Hughes, Tyler W.
    Minkov, Momchil
    Shi, Yu
    Fan, Shanhui
    OPTICA, 2018, 5 (07): : 864 - 871
  • [46] Enhancing Brain Tumor Detection Through Custom Convolutional Neural Networks and Interpretability-Driven Analysis
    Dewage, Kavinda Ashan Kulasinghe Wasalamuni
    Hasan, Raza
    Rehman, Bacha
    Mahmood, Salman
    INFORMATION, 2024, 15 (10)
  • [47] MonoNet: enhancing interpretability in neural networks via monotonic features
    Nguyen, An-Phi
    Moreno, Dana Lea
    Le-Bel, Nicolas
    Martinez, Maria Rodriguez
    BIOINFORMATICS ADVANCES, 2023, 3 (01):
  • [48] UNCERTAINTY MODELING AND INTERPRETABILITY IN CONVOLUTIONAL NEURAL NETWORKS FOR POLYP SEGMENTATION
    Wickstrom, Kristoffer
    Kampffmeyer, Michael
    Jenssen, Robert
    2018 IEEE 28TH INTERNATIONAL WORKSHOP ON MACHINE LEARNING FOR SIGNAL PROCESSING (MLSP), 2018,
  • [49] Interpretability of Neural Networks Based on Game-theoretic Interactions
    Zhou, Huilin
    Ren, Jie
    Deng, Huiqi
    Cheng, Xu
    Zhang, Jinpeng
    Zhang, Quanshi
    MACHINE INTELLIGENCE RESEARCH, 2024, 21 (04) : 718 - 739
  • [50] Functional network: A novel framework for interpretability of deep neural networks
    Zhang, Ben
    Dong, Zhetong
    Zhang, Junsong
    Lin, Hongwei
    NEUROCOMPUTING, 2023, 519 : 94 - 103