Generalization Comparison of Deep Neural Networks via Output Sensitivity

被引:7
|
作者
Forouzesh, Mahsa [1 ]
Salehi, Farnood [2 ]
Thiran, Patrick [1 ]
机构
[1] Ecole Polytech Fed Lausanne, Lausanne, Switzerland
[2] DisneyRes Studios, Zurich, Switzerland
关键词
deep neural networks; generalization; sensitivity; bias-variance decomposition;
D O I
10.1109/ICPR48806.2021.9412496
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Although recent works have brought some insights into the performance improvement of techniques used in state-of-the-art deep-learning models, more work is needed to understand their generalization properties. We shed light on this matter by linking the loss function to the output's sensitivity to its input. We find a rather strong empirical relation between the output sensitivity and the variance in the bias-variance decomposition of the loss function, which hints on using sensitivity as a metric for comparing the generalization performance of networks, without requiring labeled data. We find that sensitivity is decreased by applying popular methods which improve the generalization performance of the model, such as (1) using a deep network rather than a wide one, (2) adding convolutional layers to baseline classifiers instead of adding fully-connected layers, (3) using batch normalization, dropout and max-pooling, and (4) applying parameter initialization techniques.
引用
收藏
页码:7411 / 7418
页数:8
相关论文
共 50 条
  • [31] Hybrid pooling for enhancement of generalization ability in deep convolutional neural networks
    Tong, Zhiqiang
    Tanaka, Gouhei
    NEUROCOMPUTING, 2019, 333 : 76 - 85
  • [32] Theoretical Investigation of Generalization Bounds for Adversarial Learning of Deep Neural Networks
    Qingyi Gao
    Xiao Wang
    Journal of Statistical Theory and Practice, 2021, 15
  • [33] On Filter Generalization for Music Bandwidth Extension Using Deep Neural Networks
    Sulun, Serkan
    Davies, Matthew E. P.
    IEEE JOURNAL OF SELECTED TOPICS IN SIGNAL PROCESSING, 2021, 15 (01) : 132 - 142
  • [34] Generalization bounds for neural ordinary differential equations and deep residual networks
    Marion, Pierre
    ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 36 (NEURIPS 2023), 2023,
  • [35] Improving Generalization Ability of Deep Neural Networks for Visual Recognition Tasks
    Okatani, Takayuki
    Liu, Xing
    Suganuma, Masanori
    COMPUTATIONAL COLOR IMAGING, CCIW 2019, 2019, 11418 : 3 - 13
  • [36] Generalization Bounds of Stochastic Gradient Descent for Wide and Deep Neural Networks
    Cao, Yuan
    Gu, Quanquan
    ADVANCES IN NEURAL INFORMATION PROCESSING SYSTEMS 32 (NIPS 2019), 2019, 32
  • [37] Investigating latent representations and generalization in deep neural networks for tabular data
    Couplet, Edouard
    Lambert, Pierre
    Verleysen, Michel
    Lee, John A.
    de Bodt, Cyril
    NEUROCOMPUTING, 2024, 597
  • [38] Theoretical Investigation of Generalization Bounds for Adversarial Learning of Deep Neural Networks
    Gao, Qingyi
    Wang, Xiao
    JOURNAL OF STATISTICAL THEORY AND PRACTICE, 2021, 15 (02)
  • [39] Efficient architecture for deep neural networks with heterogeneous sensitivity
    Cho, Hyunjoong
    Jang, Jinhyeok
    Lee, Chanhyeok
    Yang, Seungjoon
    NEURAL NETWORKS, 2021, 134 : 95 - 106
  • [40] Load Forecasting via Deep Neural Networks
    He, Wan
    5TH INTERNATIONAL CONFERENCE ON INFORMATION TECHNOLOGY AND QUANTITATIVE MANAGEMENT, ITQM 2017, 2017, 122 : 308 - 314