Thread: Towards fine-grained precision reconfiguration in variable-precision neural network accelerator

doi:10.1587/elex.16.20190145

	Thread: Towards fine-grained precision reconfiguration in variable-precision neural network accelerator
	Zhang, Shichang 1,2; Wang, Ying 1; Chen, Xiaoming 1; Han, Yinhe 1,2; Wang, Yujie 1; Li, Xiaowei 1,2
	2019-07-25
发表期刊	IEICE ELECTRONICS EXPRESS
ISSN	1349-2543
卷号	16 期号:14 页码:6
摘要	In this work, we propose a neural network accelerator that supports finer-grained precision tuning for both activations and weights. To the best of our knowledge, this is the first neural network accelerator supporting arbitrary bit widths within 8 bits for both neural weights and activations. Our accelerator combines the features of bit-parallel and bit-serial design methods so that we can accomplish high arithmetic-unit utilization and precision tuning flexibility simultaneously. According to the cycle accurate simulation and the synthesized result, the proposed accelerator achieves up to 1.77x energy-efficiency improvement over the hit-serial design of Stripes for the evaluated workloads.
关键词	DNN accelerator variable bit-precision bit-serial bit-parallel fine-grained precision
DOI	10.1587/elex.16.20190145
收录类别	SCI
语种	英语
资助项目	National Natural Science Foundation of China (NSFC)[61834006] ; National Natural Science Foundation of China (NSFC)[61804155] ; National Natural Science Foundation of China (NSFC)[61874124] ; Beijing Municipal Science & Technology Commission[Z171100000117019] ; Beijing Municipal Science & Technology Commission[Z181100008918006] ; Strategic Priority Research Program of the Chinese Academy of Sciences[XDPB12] ; Innovative Project of ICT, CAS[5120186140] ; Youth Innovation Promotion Association CAS ; Young Elite Scientists Sponsorship Program by CAST[2018QNRC001] ; Young Elite Scientists Sponsorship Program by CAST[2016QNRC001]
WOS研究方向	Engineering
WOS类目	Engineering, Electrical & Electronic
WOS记录号	WOS:000480417500001
出版者	IEICE-INST ELECTRONICS INFORMATION COMMUNICATIONS ENG
引用统计	被引频次：9[WOS] [WOS记录] [WOS相关记录]
文献类型	期刊论文
条目标识符	http://119.78.100.204/handle/2XEOYT63/4452
专题	中国科学院计算技术研究所期刊论文_英文
通讯作者	Han, Yinhe
作者单位	1.Chinese Acad Sci, Inst Comp Technol, State Key Lab Comp Architecture, Beijing 100190, Peoples R China 2.Univ Chinese Acad Sci, Beijing 100190, Peoples R China
推荐引用方式 GB/T 7714	Zhang, Shichang,Wang, Ying,Chen, Xiaoming,et al. Thread: Towards fine-grained precision reconfiguration in variable-precision neural network accelerator[J]. IEICE ELECTRONICS EXPRESS,2019,16(14):6.
APA	Zhang, Shichang,Wang, Ying,Chen, Xiaoming,Han, Yinhe,Wang, Yujie,&Li, Xiaowei.(2019).Thread: Towards fine-grained precision reconfiguration in variable-precision neural network accelerator.IEICE ELECTRONICS EXPRESS,16(14),6.
MLA	Zhang, Shichang,et al."Thread: Towards fine-grained precision reconfiguration in variable-precision neural network accelerator".IEICE ELECTRONICS EXPRESS 16.14(2019):6.