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
  • [1] Robust Neural Networks Learning via a Minimization of Stochastic Output Sensitivity
    Li, Jincheng
    Ng, Wing W. Y.
    IEEE ACCESS, 2020, 8 (08) : 205455 - 205466
  • [2] Sensitivity Analysis of Deep Neural Networks
    Shu, Hai
    Zhu, Hongtu
    THIRTY-THIRD AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE / THIRTY-FIRST INNOVATIVE APPLICATIONS OF ARTIFICIAL INTELLIGENCE CONFERENCE / NINTH AAAI SYMPOSIUM ON EDUCATIONAL ADVANCES IN ARTIFICIAL INTELLIGENCE, 2019, : 4943 - 4950
  • [3] Generalization Bounds of Deep Neural Networks With τ -Mixing Samples
    Liu, Liyuan
    Chen, Yaohui
    Li, Weifu
    Wang, Yingjie
    Gu, Bin
    Zheng, Feng
    Chen, Hong
    IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, 2025,
  • [4] VOVU: A Method for Predicting Generalization in Deep Neural Networks
    Wang, Juan
    Ge, Liangzhu
    Liu, Guorui
    Li, Guoyan
    MATHEMATICAL PROBLEMS IN ENGINEERING, 2021, 2021
  • [5] Abstraction Mechanisms Predict Generalization in Deep Neural Networks
    Gain, Alex
    Siegelmann, Hava
    INTERNATIONAL CONFERENCE ON MACHINE LEARNING, VOL 119, 2020, 119
  • [6] A generalization bound of deep neural networks for dependent data
    Do, Quan Huu
    Nguyen, Binh T.
    Ho, Lam Si Tung
    STATISTICS & PROBABILITY LETTERS, 2024, 208
  • [7] Abstraction Mechanisms Predict Generalization in Deep Neural Networks
    Gain, Alex
    Siegelmann, Hava
    25TH AMERICAS CONFERENCE ON INFORMATION SYSTEMS (AMCIS 2019), 2019,
  • [8] Deep Neural Networks with Cascaded Output Layers
    Cui H.
    Bai J.
    Bi X.
    Huang L.
    Bai, Jie (baijie@tongji.edu.cn), 1600, Science Press (45): : 19 - 23
  • [9] Towards Better Generalization of Deep Neural Networks via Non-Typicality Sampling Scheme
    Peng, Xinyu
    Wang, Fei-Yue
    Li, Li
    IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, 2023, 34 (10) : 7910 - 7920
  • [10] Spectral Pruning: Compressing Deep Neural Networks via Spectral Analysis and its Generalization Error
    Suzuki, Taiji
    Abe, Hiroshi
    Murata, Tomoya
    Horiuchi, Shingo
    Ito, Kotaro
    Wachi, Tokuma
    Hirai, So
    Yukishima, Masatoshi
    Nishimura, Tomoaki
    PROCEEDINGS OF THE TWENTY-NINTH INTERNATIONAL JOINT CONFERENCE ON ARTIFICIAL INTELLIGENCE, 2020, : 2839 - 2846