Robust Neural Networks Inspired by Strong Stability Preserving Runge-Kutta Methods

Byungjoo Kim, Bryce Chudomelka, Jinyoung Park, Jaewoo Kang, Youngjoon Hong, Hyunwoo J. Kim

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

2 Citations (Scopus)

Abstract

Deep neural networks have achieved state-of-the-art performance in a variety of fields. Recent works observe that a class of widely used neural networks can be viewed as the Euler method of numerical discretization. From the numerical discretization perspective, Strong Stability Preserving (SSP) methods are more advanced techniques than the explicit Euler method that produce both accurate and stable solutions. Motivated by the SSP property and a generalized Runge-Kutta method, we proposed Strong Stability Preserving networks (SSP networks) which improve robustness against adversarial attacks. We empirically demonstrate that the proposed networks improve the robustness against adversarial examples without any defensive methods. Further, the SSP networks are complementary with a state-of-the-art adversarial training scheme. Lastly, our experiments show that SSP networks suppress the blow-up of adversarial perturbations. Our results open up a way to study robust architectures of neural networks leveraging rich knowledge from numerical discretization literature.

Original languageEnglish
Title of host publicationComputer Vision – ECCV 2020 - 16th European Conference, 2020, Proceedings
EditorsAndrea Vedaldi, Horst Bischof, Thomas Brox, Jan-Michael Frahm
PublisherSpringer Science and Business Media Deutschland GmbH
Pages416-432
Number of pages17
ISBN (Print)9783030585440
DOIs
Publication statusPublished - 2020
Event16th European Conference on Computer Vision, ECCV 2020 - Glasgow, United Kingdom
Duration: 2020 Aug 232020 Aug 28

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume12354 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference16th European Conference on Computer Vision, ECCV 2020
Country/TerritoryUnited Kingdom
CityGlasgow
Period20/8/2320/8/28

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint

Dive into the research topics of 'Robust Neural Networks Inspired by Strong Stability Preserving Runge-Kutta Methods'. Together they form a unique fingerprint.

Cite this