Exploring uncertainty in pseudo-label guided unsupervised domain adaptation

doi:10.1016/j.patcog.2019.106996

CASIA OpenIR > 智能感知与计算研究中心

	Exploring uncertainty in pseudo-label guided unsupervised domain adaptation
	Liang, Jian1,2,3 ; He, Ran1,3,4 ; Sun, Zhenan1,3,4 ; Tan, Tieniu1,3,4
发表期刊	PATTERN RECOGNITION
ISSN	0031-3203
	2019-12-01
卷号	96 页码:11
通讯作者	Liang, Jian(liangjian92@gmail.com)
摘要	Due to the unavailability of labeled target data, most existing unsupervised domain adaptation (UDA) methods alternately classify the unlabeled target samples and discover a low-dimensional subspace by mitigating the cross-domain distribution discrepancy. During the pseudo-label guided subspace discovery step, however, the posterior probabilities (uncertainties) from the previous target label estimation step are totally ignored, which may promote the error accumulation and degrade the adaptation performance. To address this issue, we propose to progressively increase the number of target training samples and incorporate the uncertainties to accurately characterize both cross-domain distribution discrepancy and other infra-domain relations. Specifically, we exploit maximum mean discrepancy (MMD) and within class variance minimization for these relations, yet, these terms merely focus on the global class structure while ignoring the local structure. Then, a triplet-wise instance-to-center margin is further maximized to push apart target instances and source class centers of different classes and bring closer them of the same class. Generally, an EM-style algorithm is developed by alternating between inferring uncertainties, progressively selecting certain training target samples, and seeking the optimal feature transformation to bridge two domains. Extensive experiments on three popular visual domain adaptation datasets demonstrate that our method significantly outperforms recent state-of-the-art approaches. (C) 2019 Elsevier Ltd. All rights reserved.
关键词	Unsupervised domain adaptation Pseudo labeling Feature transformation Progressive learning Transfer learning
DOI	10.1016/j.patcog.2019.106996
关键词[WOS]	KERNEL
收录类别	SCI
语种	英语
资助项目	State Key Development Program[2016YFB1001001] ; National Natural Science Foundation of China[61622310] ; National Natural Science Foundation of China[61473289] ; State Key Development Program[2016YFB1001001] ; National Natural Science Foundation of China[61622310] ; National Natural Science Foundation of China[61473289]
项目资助者	State Key Development Program ; National Natural Science Foundation of China
WOS研究方向	Computer Science ; Engineering
WOS类目	Computer Science, Artificial Intelligence ; Engineering, Electrical & Electronic
WOS记录号	WOS:000487569700042
出版者	ELSEVIER SCI LTD
七大方向——子方向分类	模式识别基础
引用统计	被引频次：82[WOS] [WOS记录] [WOS相关记录]
文献类型	期刊论文
条目标识符	http://ir.ia.ac.cn/handle/173211/26442
专题	智能感知与计算研究中心
通讯作者	Liang, Jian
作者单位	1.Chinese Acad Sci CASIA, Inst Automat, Natl Lab Pattern Recognit, Ctr Res Intelligent Percept & Comp, Beijing, Peoples R China 2.NUS, Dept ECE, Singapore, Singapore 3.UCAS, Beijing, Peoples R China 4.Chinese Acad Sci, Ctr Excellence Brain Sci & Intelligence Technol, Beijing, Peoples R China
第一作者单位	模式识别国家重点实验室
通讯作者单位	模式识别国家重点实验室
推荐引用方式 GB/T 7714	Liang, Jian,He, Ran,Sun, Zhenan,et al. Exploring uncertainty in pseudo-label guided unsupervised domain adaptation[J]. PATTERN RECOGNITION,2019,96:11.
APA	Liang, Jian,He, Ran,Sun, Zhenan,&Tan, Tieniu.(2019).Exploring uncertainty in pseudo-label guided unsupervised domain adaptation.PATTERN RECOGNITION,96,11.
MLA	Liang, Jian,et al."Exploring uncertainty in pseudo-label guided unsupervised domain adaptation".PATTERN RECOGNITION 96(2019):11.