Bregman learning for generative adversarial networks

Jian Gao, Hamidou Tembine

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

Abstract

In this paper we develop a game theoretic learning framework for deep generative models. Firstly, the problem of minimizing the dissimilarity between the generator distribution and real data is introduced based on f-divergence. Secondly, the optimization problem is transformed into a zero-sum game with two adversarial players, and the existence of Nash equilibrium is established in the quasi-concave-convex case under suitable conditions. Thirdly, a general Bregman-based learning algorithm is proposed to find the Nash equilibria. The algorithm is proved to have a doubly logarithmic convergence time with respect to the precision of the minimax value in potential convex games. Lastly, our methodology is implemented in three application scenarios and compared with several existing optimization algorithms. Both qualitative and quantitative evaluation show that the generative model trained by our algorithm has the state-of-art performance.

Original languageEnglish (US)
Title of host publicationProceedings of the 30th Chinese Control and Decision Conference, CCDC 2018
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages82-89
Number of pages8
ISBN (Electronic)9781538612439
DOIs
StatePublished - Jul 6 2018
Event30th Chinese Control and Decision Conference, CCDC 2018 - Shenyang, China
Duration: Jun 9 2018Jun 11 2018

Publication series

NameProceedings of the 30th Chinese Control and Decision Conference, CCDC 2018

Other

Other30th Chinese Control and Decision Conference, CCDC 2018
Country/TerritoryChina
CityShenyang
Period6/9/186/11/18

Keywords

  • Bregman learning
  • Convex Optimization
  • Deep Neural Network
  • GAN
  • Game Theory

ASJC Scopus subject areas

  • Artificial Intelligence
  • Computer Science Applications
  • Control and Optimization
  • Decision Sciences (miscellaneous)

Fingerprint

Dive into the research topics of 'Bregman learning for generative adversarial networks'. Together they form a unique fingerprint.

Cite this