Support matching: A novel regularization to escape from mode collapse in GANs

Yinghua YAO, Yuangang PAN, Ivor W. TSANG, Xin YAO*

*Corresponding author for this work

Research output: Book Chapters | Papers in Conference ProceedingsConference paper (refereed)Researchpeer-review

2 Citations (Scopus)

Abstract

Generative adversarial network (GAN) is an implicit generative model known for its ability to generate sharp images. However, it is poor at generating diverse data, which refers to the mode collapse problem. It turns out that GAN is prone to emphasizing the quality of samples but ignoring their diversity. When mode collapse happens, the support of the generated data distribution is not aligned with that of the real data distribution. We thus propose Support Regularized-GAN (SR-GAN) to address such a mode collapse issue by matching their support. Our experiments on synthetic and real-world datasets show that our regularization can mitigate the mode collapse and also improve the data quality. © Springer Nature Switzerland AG 2019.
Original languageEnglish
Title of host publicationNeural Information Processing : 26th International Conference, ICONIP 2019, Sydney, NSW, Australia, December 12–15, 2019, Proceedings, Part IV
EditorsTom GEDEON, Kok Wai WONG, Minho LEE
PublisherSpringer
Pages40-48
Number of pages9
ISBN (Electronic)9783030368081
ISBN (Print)9783030368074
DOIs
Publication statusPublished - 2019
Externally publishedYes
Event26th International Conference on Neural Information Processing - Sydney, Australia
Duration: 12 Dec 201915 Dec 2019

Publication series

NameCommunications in Computer and Information Science
PublisherSpringer
Volume1142
ISSN (Print)1865-0929
ISSN (Electronic)1865-0937

Conference

Conference26th International Conference on Neural Information Processing
Abbreviated titleICONIP 2019
Country/TerritoryAustralia
CitySydney
Period12/12/1915/12/19

Keywords

  • GANs
  • Mode collapse
  • Support matching

Fingerprint

Dive into the research topics of 'Support matching: A novel regularization to escape from mode collapse in GANs'. Together they form a unique fingerprint.

Cite this