BitStream: An efficient framework for inference of binary neural networks on CPUs

doi:10.1016/j.patrec.2019.04.016

	BitStream: An efficient framework for inference of binary neural networks on CPUs
	Jiang, Yanshu 1; Zhao, Tianli 1,2; He, Xiangyu 2; Leng, Cong 2; Cheng, Jian 2
刊名	PATTERN RECOGNITION LETTERS
	2019-07-01
卷号	125 页码:303-309
关键词	Convolutional neural networks Binary neural networks Image classification
ISSN号	0167-8655
DOI	10.1016/j.patrec.2019.04.016
通讯作者	Zhao, Tianli(tianli.zhao@nlpr.ia.ac.cn)
英文摘要	Convolutional Neural Networks (CNN) has been well-studied and widely used in the field of pattern recognition. Many pattern recognition algorithms need features extracted from CNN models to adapt to complex tasks, such as image classification, object detection, natural language processing and so on. However, to deal with more and more complex tasks, modern CNN models are becoming larger and larger, contain large number of parameters and computation, leading to high consumption of memory, computational and power resources during inference. This makes it difficult to run CNN based applications in real time on mobile devices, where memory, computational and power resources are limited. Binarization of neural networks is proposed to reduce memory and computational complexity of CNN. However, traditional implementations of Binary Neural Networks (BNN) follow the conventional im2col-based convolution computation flow, which is widely used in floating-point networks but not friendly enough to cache when it comes to binarized neural networks. In this paper, we propose BitStream, a general architecture for efficient inference of BNN on CPUs. In BitStream, we propose a simple but novel computation flow for BNN. Unlike existing implementations of BNN, in BitStream, all the layers, including convolutional layers, binarization layers and pooling layers are all calculated in binary precision. Comprehensive analyses demonstrate that our proposed computation flow consumes less memory during inference of BNN, and it's friendly to cache because of its continuous memory access. (C) 2019 Published by Elsevier B.V.
WOS研究方向	Computer Science
语种	英语
出版者	ELSEVIER
WOS记录号	WOS:000482374500042
内容类型	期刊论文
源URL	[http://ir.ia.ac.cn/handle/173211/27269]
专题	中国科学院自动化研究所
通讯作者	Zhao, Tianli
作者单位	1.Harbin Univ Sci & Technol, Dept Automat, Harbin, Heilongjiang, Peoples R China 2.Chinese Acad Sci, Natl Lab Pattern Recoginit, Inst Automat, Beijing, Peoples R China
推荐引用方式 GB/T 7714	Jiang, Yanshu,Zhao, Tianli,He, Xiangyu,et al. BitStream: An efficient framework for inference of binary neural networks on CPUs[J]. PATTERN RECOGNITION LETTERS,2019,125:303-309.
APA	Jiang, Yanshu,Zhao, Tianli,He, Xiangyu,Leng, Cong,&Cheng, Jian.(2019).BitStream: An efficient framework for inference of binary neural networks on CPUs.PATTERN RECOGNITION LETTERS,125,303-309.
MLA	Jiang, Yanshu,et al."BitStream: An efficient framework for inference of binary neural networks on CPUs".PATTERN RECOGNITION LETTERS 125(2019):303-309.