中文版 | English
题名

Value Function Based Difference-of-Convex Algorithm for Bilevel Hyperparameter Selection Problems

作者
通讯作者Zhang, Jin
发表日期
2022
会议名称
38th International Conference on Machine Learning (ICML)
ISSN
2640-3498
会议录名称
会议日期
JUL 17-23, 2022
会议地点
null,Baltimore,MD
出版地
1269 LAW ST, SAN DIEGO, CA, UNITED STATES
出版者
摘要
Gradient-based optimization methods for hyper-parameter tuning guarantee theoretical convergence to stationary solutions when for fixed upperlevel variable values, the lower level of the bilevel program is strongly convex (LLSC) and smooth (LLS). This condition is not satisfied for bilevel programs arising from tuning hyperparameters in many machine learning algorithms. In this work, we develop a sequentially convergent Value Function based Difference-of-Convex Algorithm with inexactness (VF-iDCA). We show that this algorithm achieves stationary solutions without LLSC and LLS assumptions for bilevel programs from a broad class of hyperparameter tuning applications. Our extensive experiments confirm our theoretical findings and show that the proposed VF-iDCA yields superior performance when applied to tune hyperparameters.
学校署名
通讯
语种
英语
相关链接[来源记录]
收录类别
资助项目
National Natural Science Foundation of China[11971220] ; Shenzhen Science and Technology Program[RCYX20200714114700072] ; Guangdong Basic and Applied Basic Research Foundation[2022B1515020082]
WOS研究方向
Computer Science
WOS类目
Computer Science, Artificial Intelligence
WOS记录号
WOS:000922378802009
来源库
Web of Science
引用统计
被引频次[WOS]:0
成果类型会议论文
条目标识符http://sustech.caswiz.com/handle/2SGJ60CL/502107
专题理学院_数学系
作者单位
1.Univ Waterloo, Dept Stat & Actuarial Sci, Waterloo, ON, Canada
2.Univ Victoria, Dept Math & Stat, Victoria, BC, Canada
3.Southern Univ Sci & Technol, Dept Math, SUSTech Int Natl Ctr Math, Shenzhen, Guangdong, Peoples R China
4.Natl Ctr Appl Math Shenzhen, Shenzhen, Guangdong, Peoples R China
通讯作者单位数学系
推荐引用方式
GB/T 7714
Gao, Lucy,Ye, Jane J.,Yin, Haian,et al. Value Function Based Difference-of-Convex Algorithm for Bilevel Hyperparameter Selection Problems[C]. 1269 LAW ST, SAN DIEGO, CA, UNITED STATES:JMLR-JOURNAL MACHINE LEARNING RESEARCH,2022.
条目包含的文件
条目无相关文件。
个性服务
原文链接
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
导出为Excel格式
导出为Csv格式
Altmetrics Score
谷歌学术
谷歌学术中相似的文章
[Gao, Lucy]的文章
[Ye, Jane J.]的文章
[Yin, Haian]的文章
百度学术
百度学术中相似的文章
[Gao, Lucy]的文章
[Ye, Jane J.]的文章
[Yin, Haian]的文章
必应学术
必应学术中相似的文章
[Gao, Lucy]的文章
[Ye, Jane J.]的文章
[Yin, Haian]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
[发表评论/异议/意见]
暂无评论

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。