CNQ: Compressor-Based Non-uniform Quantization of Deep Neural NetworksInspec keywordsOther keywordsKey words

doi:10.1049/cje.2020.09.014

CORC > 自动化研究所 > 中国科学院自动化研究所 > 智能制造技术与系统研究中心 > 多维数据分析团队

	CNQ: Compressor-Based Non-uniform Quantization of Deep Neural NetworksInspec keywordsOther keywordsKey words
	Yuan, Yong 2,3; Chen, Chen 2,3; Hu, Xiyuan 4; Peng, Silong 1,2,3
刊名	CHINESE JOURNAL OF ELECTRONICS
	2020-11-01
卷号	29 期号:6 页码:1126-1133
关键词	entropy image classification learning (artificial intelligence) neural nets object detection optimisation quantisation (signal) network structure DNN low-bit quantization time-consuming training compressor-based fast nonuniform quantization method quantization model post-training quantization methods deep neural networks network quantization compressor-based nonuniform quantization CNQ Non-uniform quantization Knowledge distillation Unlabeled samples Network compression
ISSN号	1022-4653
DOI	10.1049/cje.2020.09.014
通讯作者	Chen, Chen(chen.chen@ia.ac.cn)
英文摘要	Deep neural networks (DNNs) have achieved state-of-the-art performance in a number of domains but suffer intensive complexity. Network quantization can effectively reduce computation and memory costs without changing network structure, facilitating the deployment of DNNs on mobile devices. While the existing methods can obtain good performance, low-bit quantization without time-consuming training or access to the full dataset is still a challenging problem. In this paper, we develop a novel method named Compressorbased non-uniform quantization (CNQ) method to achieve non-uniform quantization of DNNs with few unlabeled samples. Firstly, we present a compressor-based fast nonuniform quantization method, which can accomplish nonuniform quantization without iterations. Secondly, we propose to align the feature maps of the quantization model with the pre-trained model for accuracy recovery. Considering the property difference between different activation channels, we utilize the weighted-entropy perchannel to optimize the alignment loss. In the experiments, we evaluate the proposed method on image classification and object detection. Our results outperform the existing post-training quantization methods, which demonstrate the effectiveness of the proposed method.
资助项目	National Natural Science Foundation of China[61906194] ; National Natural Science Foundation of China[61571438]
WOS研究方向	Engineering
语种	英语
出版者	TECHNOLOGY EXCHANGE LIMITED HONG KONG
WOS记录号	WOS:000609935600016
资助机构	National Natural Science Foundation of China
内容类型	期刊论文
源URL	[http://ir.ia.ac.cn/handle/173211/42894]
专题	自动化研究所_智能制造技术与系统研究中心_多维数据分析团队
通讯作者	Chen, Chen
作者单位	1.Beijing Visyst Co Ltd, Beijing 100083, Peoples R China 2.Chinese Acad Sci, Inst Automat, Beijing 100190, Peoples R China 3.Univ Chinese Acad Sci, Beijing 100049, Peoples R China 4.Nanjing Univ Sci & Technol, Sch Comp Sci & Engn, Nanjing 210094, Peoples R China
推荐引用方式 GB/T 7714	Yuan, Yong,Chen, Chen,Hu, Xiyuan,et al. CNQ: Compressor-Based Non-uniform Quantization of Deep Neural NetworksInspec keywordsOther keywordsKey words[J]. CHINESE JOURNAL OF ELECTRONICS,2020,29(6):1126-1133.
APA	Yuan, Yong,Chen, Chen,Hu, Xiyuan,&Peng, Silong.(2020).CNQ: Compressor-Based Non-uniform Quantization of Deep Neural NetworksInspec keywordsOther keywordsKey words.CHINESE JOURNAL OF ELECTRONICS,29(6),1126-1133.
MLA	Yuan, Yong,et al."CNQ: Compressor-Based Non-uniform Quantization of Deep Neural NetworksInspec keywordsOther keywordsKey words".CHINESE JOURNAL OF ELECTRONICS 29.6(2020):1126-1133.