A Unified Perspective for Loss-Oriented Imbalanced Learning via Localization

doi:10.1109/TPAMI.2025.3609440

CSpace

	A Unified Perspective for Loss-Oriented Imbalanced Learning via Localization
	Wang, Zitai 1; Xu, Qianqian 1,2; Yang, Zhiyong 3; Xu, Zhikang 4,5; Zhang, Linchao 6; Cao, Xiaochun 7; Huang, Qingming 1,8,9
	2026
发表期刊	IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
ISSN	0162-8828
卷号	48 期号:1 页码:639-656
摘要	Due to the inherent imbalance in real-world datasets, na & iuml;ve Empirical Risk Minimization (ERM) tends to bias the learning process towards the majority classes, hindering generalization to minority classes. To rebalance the learning process, one straightforward yet effective approach is to modify the loss function via class-dependent terms, such as re-weighting and logit-adjustment. However, existing analysis of these loss-oriented methods remains coarse-grained and fragmented, failing to explain some empirical results. After reviewing prior work, we find that the properties used through their analysis are typically global, i.e., defined over the whole dataset. Hence, these properties fail to effectively capture how class-dependent terms influence the learning process. To bridge this gap, we turn to explore the localized versions of such properties i.e., defined within each class. Specifically, we employ localized calibration to provide consistency validation across a broader range of losses and localized Lipschitz continuity to provide a fine-grained generalization bound. In this way, we reach a unified perspective for improving and adjusting loss-oriented methods. Finally, a principled learning algorithm is developed based on these insights. Empirical results on both traditional ResNets and foundation models validate our theoretical analyses and demonstrate the effectiveness of the proposed method.
关键词	Calibration Training Safety Predictive models Data mining Computers Additives Smoothing methods Risk minimization Protocols Imbalanced learning re-weighting logit adjustment fisher consistency generalization analysis
DOI	10.1109/TPAMI.2025.3609440
收录类别	SCI
语种	英语
WOS研究方向	Computer Science ; Engineering
WOS类目	Computer Science, Artificial Intelligence ; Engineering, Electrical & Electronic
WOS记录号	WOS:001630351400004
出版者	IEEE COMPUTER SOC
引用统计
文献类型	期刊论文
条目标识符	http://119.78.100.204/handle/2XEOYT63/42810
专题	中国科学院计算技术研究所
通讯作者	Xu, Qianqian; Huang, Qingming
作者单位	1.Chinese Acad Sci, State Key Lab AI Safety, Inst Comp Technol, Beijing 100190, Peoples R China 2.Peng Cheng Lab, Shenzhen 518055, Peoples R China 3.Univ Chinese Acad Sci, Sch Comp Sci & Technol, Beijing 100049, Peoples R China 4.Chinese Acad Sci, Inst Informat Engn, Beijing 100093, Peoples R China 5.Univ Chinese Acad Sci, Sch Cyber Secur, Beijing 100049, Peoples R China 6.China Elect Technol Grp Corp, Artificial Intelligence Inst, Beijing 100041, Peoples R China 7.Sun Yat Sen Univ, Sch Cyber Sci & Technol, Shenzhen Campus, Shenzhen 518107, Peoples R China 8.Univ Chinese Acad Sci, Sch Comp Sci & Technol, Beijing 101408, Peoples R China 9.Univ Chinese Acad Sci, Key Lab Big Data Min & Knowledge Management BDKM, Beijing 101408, Peoples R China
推荐引用方式 GB/T 7714	Wang, Zitai,Xu, Qianqian,Yang, Zhiyong,et al. A Unified Perspective for Loss-Oriented Imbalanced Learning via Localization[J]. IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE,2026,48(1):639-656.
APA	Wang, Zitai.,Xu, Qianqian.,Yang, Zhiyong.,Xu, Zhikang.,Zhang, Linchao.,...&Huang, Qingming.(2026).A Unified Perspective for Loss-Oriented Imbalanced Learning via Localization.IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE,48(1),639-656.
MLA	Wang, Zitai,et al."A Unified Perspective for Loss-Oriented Imbalanced Learning via Localization".IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 48.1(2026):639-656.