中国科学院机构知识库网格
Chinese Academy of Sciences Institutional Repositories Grid
首页 机构 成果 学者
  
  1. CAS IR Grid
  2. 数学与系统科学研究院
  3. 中国科学院数学与系统科学研究院
Training GANs with centripetal acceleration

文献类型：期刊论文

作者Peng, Wei2; Dai, Yu-Hong1,3; Zhang, Hui4; Cheng, Lizhi4
刊名OPTIMIZATION METHODS & SOFTWARE
出版日期2020-04-16
页码19
关键词Primary: 97N60 90C25 Secondary: 90C06 90C30
ISSN号1055-6788
DOI10.1080/10556788.2020.1754414
英文摘要Training generative adversarial networks (GANs) often suffers from cyclic behaviours of iterates. Based on a simple intuition that the direction of centripetal acceleration of an object moving in uniform circular motion is toward the centre of the circle, we present the Simultaneous Centripetal Acceleration (SCA) method and the Alternating Centripetal Acceleration (ACA) method to alleviate the cyclic behaviours. Under suitable conditions, gradient descent methods with either SCA or ACA are shown to be linearly convergent for bilinear games. Numerical experiments are conducted by applying ACA to existing gradient-based algorithms in a GAN setup scenario, which demonstrate the superiority of ACA.
资助项目Chinese NSF[11501569] ; Chinese NSF[61571008] ; Chinese NSF[11631013] ; Chinese NSF[11991021] ; Chinese NSF[11971372] ; Chinese NSF[11991020] ; Beijing Academy of Artificial Intelligence (BAAI)
WOS研究方向Computer Science ; Operations Research & Management Science ; Mathematics
语种英语
WOS记录号WOS:000527185500001
出版者TAYLOR & FRANCIS LTD
源URL[http://ir.amss.ac.cn/handle/2S8OKBNM/51187]  
专题中国科学院数学与系统科学研究院
通讯作者Dai, Yu-Hong
作者单位1.Univ Chinese Acad Sci, Sch Math Sci, Beijing 100049, Peoples R China
2.Chinese Acad Mil Sci, Natl Innovat Inst Def Technol, Beijing, Peoples R China
3.Chinese Acad Sci, LSEC, ICMSEC, Acad Math & Syst Sci, Beijing 100190, Peoples R China
4.Natl Univ Def Technol, Dept Math, Changsha, Hunan, Peoples R China
推荐引用方式
GB/T 7714		 Peng, Wei,Dai, Yu-Hong,Zhang, Hui,et al. Training GANs with centripetal acceleration[J]. OPTIMIZATION METHODS & SOFTWARE,2020:19.
APA Peng, Wei,Dai, Yu-Hong,Zhang, Hui,&Cheng, Lizhi.(2020).Training GANs with centripetal acceleration.OPTIMIZATION METHODS & SOFTWARE,19.
MLA Peng, Wei,et al."Training GANs with centripetal acceleration".OPTIMIZATION METHODS & SOFTWARE (2020):19.

入库方式： OAI收割

来源：数学与系统科学研究院

浏览0
下载0
收藏0
其他版本

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。