Constrained Maximum Cross-Domain Likelihood for Domain Generalization

doi:10.1109/TNNLS.2023.3292242

CORC > 自动化研究所 > 中国科学院自动化研究所 > 多模态人工智能系统全国重点实验室

	Constrained Maximum Cross-Domain Likelihood for Domain Generalization
	Lin, Jianxin 1,2; Tang, Yongqiang 1,2; Wang, Junping 1,2; Zhang, Wensheng 1,2
刊名	IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
	2023-07-13
页码	15
关键词	Optimization Feature extraction Metalearning Entropy Training Hospitals Task analysis Distribution shift domain adaptation domain generalization domain-invariant representation joint distribution alignment
ISSN号	2162-237X
DOI	10.1109/TNNLS.2023.3292242
通讯作者	Tang, Yongqiang(yongqiang.tang@ia.ac.cn) ; Wang, Junping(junping.wang@ia.ac.cn)
英文摘要	As a recent noticeable topic, domain generalization aims to learn a generalizable model on multiple source domains, which is expected to perform well on unseen test domains. Great efforts have been made to learn domain-invariant features by aligning distributions across domains. However, existing works are often designed based on some relaxed conditions which are generally hard to satisfy and fail to realize the desired joint distribution alignment. In this article, we propose a novel domain generalization method, which originates from an intuitive idea that a domain-invariant classifier can be learned by minimizing the Kullback-Leibler (KL)-divergence between posterior distributions from different domains. To enhance the generalizability of the learned classifier, we formalize the optimization objective as an expectation computed on the ground-truth marginal distribution. Nevertheless, it also presents two obvious deficiencies, one of which is the side-effect of entropy increase in KL-divergence and the other is the unavailability of ground-truth marginal distributions. For the former, we introduce a term named maximum in-domain likelihood to maintain the discrimination of the learned domain-invariant representation space. For the latter, we approximate the ground-truth marginal distribution with source domains under a reasonable convex hull assumption. Finally, a constrained maximum cross-domain likelihood (CMCL) optimization problem is deduced, by solving which the joint distributions are naturally aligned. An alternating optimization strategy is carefully designed to approximately solve this optimization problem. Extensive experiments on four standard benchmark datasets, i.e., Digits-DG, PACS, Office-Home, and miniDomainNet, highlight the superior performance of our method.
资助项目	National Key Research and Development Program of China[2020AAA0109500] ; National Natural Science Foundation of China[62106266] ; National Natural Science Foundation of China[92167109] ; National Natural Science Foundation of China[U22B2048] ; National Natural Science Foundation of China[62173328]
WOS研究方向	Computer Science ; Engineering
语种	英语
出版者	IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
WOS记录号	WOS:001035824200001
资助机构	National Key Research and Development Program of China ; National Natural Science Foundation of China
内容类型	期刊论文
源URL	[http://ir.ia.ac.cn/handle/173211/53822]
专题	多模态人工智能系统全国重点实验室
通讯作者	Tang, Yongqiang; Wang, Junping
作者单位	1.Univ Chinese Acad Sci, Sch Artificial Intelligence, Beijing 100049, Peoples R China 2.Chinese Acad Sci, Inst Automat, State Key Lab Multimodal Artificial Intelligence S, Beijing 100190, Peoples R China
推荐引用方式 GB/T 7714	Lin, Jianxin,Tang, Yongqiang,Wang, Junping,et al. Constrained Maximum Cross-Domain Likelihood for Domain Generalization[J]. IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS,2023:15.
APA	Lin, Jianxin,Tang, Yongqiang,Wang, Junping,&Zhang, Wensheng.(2023).Constrained Maximum Cross-Domain Likelihood for Domain Generalization.IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS,15.
MLA	Lin, Jianxin,et al."Constrained Maximum Cross-Domain Likelihood for Domain Generalization".IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS (2023):15.