Repository logo

Infoscience

  • English
  • French
Log In
Logo EPFL, École polytechnique fédérale de Lausanne

Infoscience

  • English
  • French
Log In
  1. Home
  2. Academic and Research Output
  3. Conferences, Workshops, Symposiums, and Seminars
  4. Backpropagation-Friendly Eigendecomposition
 
conference paper

Backpropagation-Friendly Eigendecomposition

Wang, Wei  
•
Dang, Zheng
•
Hu, Yinlin  
Show more
January 1, 2019
Advances In Neural Information Processing Systems 32 (Nips 2019)
Conference on Neural Information Processing Systems (NeurIPS)

Eigendecomposition (ED) is widely used in deep networks. However, the backpropagation of its results tends to be numerically unstable, whether using ED directly or approximating it with the Power Iteration method, particularly when dealing with large matrices. While this can be mitigated by partitioning the data in small and arbitrary groups, doing so has no theoretical basis and makes its impossible to exploit the power of ED to the full. In this paper, we introduce a numerically stable and differentiable approach to leveraging eigenvectors in deep networks. It can handle large matrices without requiring to split them. We demonstrate the better robustness of our approach over standard ED and PI for ZCA whitening, an alternative to batch normalization, and for PCA denoising, which we introduce as a new normalization strategy for deep networks, aiming to further denoise the network's features.

  • Files
  • Details
  • Metrics
Type
conference paper
Web of Science ID

WOS:000534424303018

Author(s)
Wang, Wei  
Dang, Zheng
Hu, Yinlin  
Fua, Pascal  
Salzmann, Mathieu  
Date Issued

2019-01-01

Publisher

NEURAL INFORMATION PROCESSING SYSTEMS (NIPS)

Publisher place

La Jolla

Published in
Advances In Neural Information Processing Systems 32 (Nips 2019)
Series title/Series vol.

Advances in Neural Information Processing Systems

Volume

32

Subjects

Computer Science, Artificial Intelligence

•

Computer Science

•

decomposition

Editorial or Peer reviewed

REVIEWED

Written at

EPFL

EPFL units
CVLAB  
Event nameEvent placeEvent date
Conference on Neural Information Processing Systems (NeurIPS)

Vancouver, CANADA

Dec 08-14, 2019

Available on Infoscience
June 14, 2020
Use this identifier to reference this record
https://infoscience.epfl.ch/handle/20.500.14299/169262
Logo EPFL, École polytechnique fédérale de Lausanne
  • Contact
  • infoscience@epfl.ch

  • Follow us on Facebook
  • Follow us on Instagram
  • Follow us on LinkedIn
  • Follow us on X
  • Follow us on Youtube
AccessibilityLegal noticePrivacy policyCookie settingsEnd User AgreementGet helpFeedback

Infoscience is a service managed and provided by the Library and IT Services of EPFL. © EPFL, tous droits réservés