Bilateral Transformation of Biased Pseudo-Labels under Distribution Inconsistency

doi:10.1007/s11263-025-02701-2

CSpace

	Bilateral Transformation of Biased Pseudo-Labels under Distribution Inconsistency
	Hou, Ruibing 1; Chang, Hong 1,2; Hu, Minyang 1,2; Ma, Bingpeng 1,2; Shan, Shiguang 1,2; Chen, Xilin 1,2
	2026-01-29
发表期刊	INTERNATIONAL JOURNAL OF COMPUTER VISION
ISSN	0920-5691
卷号	134 期号:3 页码:34
摘要	Typically self-training methods assign pseudo-labels by a source model and subsequently iteratively train the model with these pseudo-labeled samples. These approaches are widely employed to leverage vast reserves of unlabeled data, e.g., in semi-supervised learning (SSL) and unsupervised finetuning (UNF). However, our theoretical analysis reveals that the distribution inconsistency between source and unlabeled data could lead to a significant generation error bound for self-training methods. Motivated by this theoretical insight, we present a Bilateral Transformation Self-Training (BTST) learning approach to mitigate the distribution discrepancy and improve the generalization of self-training methods. Firstly, Representation Transformation Module (RTM) is designed to reduce representation distribution discrepancy by bidirectionally transforming high-level representations between source and unlabeled samples. Secondly, Logit Transformation Module (LTM) is designed to reduce class distribution discrepancy by aligning classifier's predictions with the unlabeled class distribution. Also, LTM incorporates a self-supervised regularization term to estimate unlabeled distribution, theoretically proven to effectively reduce estimation error bound. The two modules work complementary to reduce the generalization bound, ultimately achieving a more generalizable self-training model. Extensive experiments demonstrate that BTST can seamlessly integrate with self-training methods, improving their generalization across various SSL and UNF settings.
关键词	Self-training Semi-supervised learning Distribution inconsistency
DOI	10.1007/s11263-025-02701-2
收录类别	SCI
语种	英语
WOS研究方向	Computer Science
WOS类目	Computer Science, Artificial Intelligence
WOS记录号	WOS:001674370500005
出版者	SPRINGER
引用统计
文献类型	期刊论文
条目标识符	http://119.78.100.204/handle/2XEOYT63/42822
专题	中国科学院计算技术研究所
通讯作者	Hou, Ruibing
作者单位	1.Chinese Acad Sci, Inst Comp Technol, Key Lab Intelligent Informat Proc, Beijing 100190, Peoples R China 2.Univ Chinese Acad Sci, Beijing 100049, Peoples R China
推荐引用方式 GB/T 7714	Hou, Ruibing,Chang, Hong,Hu, Minyang,et al. Bilateral Transformation of Biased Pseudo-Labels under Distribution Inconsistency[J]. INTERNATIONAL JOURNAL OF COMPUTER VISION,2026,134(3):34.
APA	Hou, Ruibing,Chang, Hong,Hu, Minyang,Ma, Bingpeng,Shan, Shiguang,&Chen, Xilin.(2026).Bilateral Transformation of Biased Pseudo-Labels under Distribution Inconsistency.INTERNATIONAL JOURNAL OF COMPUTER VISION,134(3),34.
MLA	Hou, Ruibing,et al."Bilateral Transformation of Biased Pseudo-Labels under Distribution Inconsistency".INTERNATIONAL JOURNAL OF COMPUTER VISION 134.3(2026):34.