Transfer channel pruning for compressing deep domain adaptation models

doi:10.1007/s13042-019-01004-6

CORC > 计算技术研究所 > 中国科学院计算技术研究所 > 中国科学院计算技术研究所期刊论文 > 英文

	Transfer channel pruning for compressing deep domain adaptation models
	Yu, Chaohui 2,3; Wang, Jindong 1; Chen, Yiqiang 2,3; Qin, Xin 2,3
刊名	INTERNATIONAL JOURNAL OF MACHINE LEARNING AND CYBERNETICS
	2019-11-01
卷号	10 期号:11 页码:3129-3144
关键词	Unsupervised domain adaptation Transfer channel pruning Accelerating
ISSN号	1868-8071
DOI	10.1007/s13042-019-01004-6
英文摘要	Deep unsupervised domain adaptation has recently received increasing attention from researchers. However, existing methods are computationally intensive due to the computational cost of convolutional neural networks (CNN) adopted by most work. There is no effective network compression method for such problem. In this paper, we propose a unified transfer channel pruning (TCP) method for accelerating deep unsupervised domain adaptation (UDA) models. TCP method is capable of compressing the deep UDA model by pruning less important channels while simultaneously learning transferable features by reducing the cross-domain distribution divergence. Therefore, it reduces the impact of negative transfer and maintains competitive performance on the target task. To the best of our knowledge, TCP method is the first approach that aims at accelerating deep unsupervised domain adaptation models. TCP method is validated on two main kinds of UDA methods: the discrepancy-based methods and the adversarial-based methods. In addition, it is validated on two benchmark datasets: Office-31 and ImageCLEF-DA with two common backbone networks - VGG16 and ResNet50. Experimental results demonstrate that our TCP method achieves comparable or better classification accuracy than other comparison methods while significantly reducing the computational cost. To be more specific, in VGG16, we get even higher accuracy after pruning 26% floating point operations (FLOPs); in ResNet50, we also get higher accuracy on half of the tasks after pruning 12% FLOPs for both discrepancy-based methods and adversarial-based methods.
资助项目	National Key Research & Development Plan of China[2017YFB1002802] ; NSFC[61572471] ; Beijing Municipal Science & Technology Commission[Z171100000117017]
WOS研究方向	Computer Science
语种	英语
出版者	SPRINGER HEIDELBERG
WOS记录号	WOS:000494802500010
内容类型	期刊论文
源URL	[http://119.78.100.204/handle/2XEOYT63/14838]
专题	中国科学院计算技术研究所期刊论文_英文
通讯作者	Chen, Yiqiang
作者单位	1.Microsoft Res Asia, Beijing, Peoples R China 2.Chinese Acad Sci, Beijing Key Lab Mobile Comp & Pervas Device, Inst Comp Technol, Beijing, Peoples R China 3.Univ Chinese Acad Sci, Beijing, Peoples R China
推荐引用方式 GB/T 7714	Yu, Chaohui,Wang, Jindong,Chen, Yiqiang,et al. Transfer channel pruning for compressing deep domain adaptation models[J]. INTERNATIONAL JOURNAL OF MACHINE LEARNING AND CYBERNETICS,2019,10(11):3129-3144.
APA	Yu, Chaohui,Wang, Jindong,Chen, Yiqiang,&Qin, Xin.(2019).Transfer channel pruning for compressing deep domain adaptation models.INTERNATIONAL JOURNAL OF MACHINE LEARNING AND CYBERNETICS,10(11),3129-3144.
MLA	Yu, Chaohui,et al."Transfer channel pruning for compressing deep domain adaptation models".INTERNATIONAL JOURNAL OF MACHINE LEARNING AND CYBERNETICS 10.11(2019):3129-3144.