Alleviating limit cycling in training GANs with an optimization technique

扫码查看

原文链接

万方数据
维普

外文摘要：In this paper,we undertake further investigation to alleviate the issue of limit cycling behavior in training generative adversarial networks(GANs)through the proposed predictive centripetal acceleration algorithm(PCAA).Specifically,we first derive the upper and lower complexity bounds of PCAA for a general bilinear game,with the last-iterate convergence rate notably improving upon previous results.Then,we combine PCAA with the adaptive moment estimation algorithm(Adam)to propose PCAA-Adam,for practical training of GANs to enhance their generalization capability.Finally,we validate the effectiveness of the proposed algorithm through experiments conducted on bilinear games,multivariate Gaussian distributions,and the CelebA dataset,respectively.

外文关键词：

GANsgeneral bilinear gamepredictive centripetal acceleration algorithmlower and upper complexity boundsPCAA-Adam

作者：

Keke Li、Liping Tang、Xinmin Yang

展开 >

作者单位：

National Center for Applied Mathematics in Chongqing,Chongqing Normal University,Chongqing 401331,China

School of Mathematical Sciences,University of Electronic Science and Technology of China,Chengdu 611731,China

基金：

Major Program of National Natural Science Foundation of ChinaMajor Program of National Natural Science Foundation of ChinaTeam Project of Innovation Leading Talent in ChongqingNSFC-RGC(Hong Kong)Joint Research ProgramScientific and Technological Research Program of Chongqing Municipal Education Commission

项目编号：

1199102011991024CQYC2021030953612261160365KJQN202300528

出版年：

2024

DOI：

10.1007/s11425-023-2296-5

中国科学:数学(英文版)

中国科学院

中国科学:数学(英文版)

CSTPCD

影响因子：0.36

ISSN：1674-7283

年,卷(期)：2024.67(6)