Building firmly nonexpansive convolutional neural networks

Matthieu Terris, Audrey Repetti, Jean-Christophe Pesquet, Yves Wiaux

Research output: Chapter in Book/Report/Conference proceedingConference contribution

33 Citations (Scopus)
587 Downloads (Pure)

Abstract

Building nonexpansive Convolutional Neural Networks (CNNs) is a challenging problem that has recently gained a lot of attention from the image processing community. In particular, it appears to be the key to obtain convergent Plugand- Play algorithms. This problem, which relies on an accurate control of the the Lipschitz constant of the convolutional layers, has also been investigated for Generative Adversarial Networks to improve robustness to adversarial perturbations. However, to the best of our knowledge, no efficient method has been developed yet to build nonexpansive CNNs. In this paper, we develop an optimization algorithm that can be incorporated in the training of a network to ensure the nonexpansiveness of its convolutional layers. This is shown to allow us to build firmly nonexpansive CNNs. We apply the proposed approach to train a CNN for an image denoising task and show its effectiveness through simulations.

Original languageEnglish
Title of host publicationICASSP 2020 - IEEE International Conference on Acoustics, Speech and Signal Processing
PublisherIEEE
Pages8658-8662
Number of pages5
ISBN (Electronic)9781509066315
DOIs
Publication statusPublished - 14 May 2020
Event45th IEEE International Conference on Acoustics, Speech and Signal Processing 2020 - Barcelona, Spain
Duration: 4 May 20208 May 2020
https://2020.ieeeicassp.org/

Publication series

NameIEEE International Conference on Acoustics, Speech and Signal Processing
ISSN (Electronic)2379-190X

Conference

Conference45th IEEE International Conference on Acoustics, Speech and Signal Processing 2020
Abbreviated titleICASSP 2020
Country/TerritorySpain
CityBarcelona
Period4/05/208/05/20
Internet address

Keywords

  • Image restoration
  • Monotone operators
  • Neural networks
  • Nonexpansive operator
  • Optimization

ASJC Scopus subject areas

  • Software
  • Signal Processing
  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'Building firmly nonexpansive convolutional neural networks'. Together they form a unique fingerprint.

Cite this