Page 33 - 《软件学报》2020年第9期
P. 33
软件学报 ISSN 1000-9825, CODEN RUXUEW E-mail: jos@iscas.ac.cn
Journal of Software,2020,31(9):2654−2677 [doi: 10.13328/j.cnki.jos.005946] http://www.jos.org.cn
©中国科学院软件研究所版权所有. Tel: +86-10-62562563
∗
面向实时应用的深度学习研究综述
张政馗, 庞为光, 谢文静, 吕鸣松, 王 义
(东北大学 计算机科学与工程学院 智慧系统实验室,辽宁 沈阳 110819)
通讯作者: 张政馗, E-mail: zhangzhengkui@cse.neu.edu.cn
摘 要: 深度学习算法和 GPU 算力的不断进步,正促进着人工智能技术在包括计算机视觉、语音识别、自然语
言处理等领域得到广泛应用.与此同时,深度学习已经开始应用于以自动驾驶为代表的安全攸关领域.但是,近两年
接连发生了几起严重的交通事故表明,深度学习技术的成熟度还远未达到安全攸关应用的要求.因此,对可信人工智
能系统的研究已经成为了一个热点方向.对现有的面向实时应用的深度学习领域的研究工作进行了综述,首先介绍
了深度学习技术应用于实时嵌入式系统所面临的关键设计问题;然后,从深层神经网络的轻量化设计、GPU 时间分
析与任务调度、CPU+GPU SoC 异构平台的资源管理、深层神经网络与网络加速器的协同设计等多个方面对现有
的研究工作进行了分析和总结;最后展望了面向实时应用的深度学习领域进一步的研究方向.
关键词: 深度学习;深层神经网络;实时系统;时间分析;实时调度;共享资源冲突
中图法分类号: TP181
中文引用格式: 张政馗,庞为光,谢文静,吕鸣松,王义.面向实时应用的深度学习研究综述.软件学报,2020,31(9):2654−2677.
http://www.jos.org.cn/1000-9825/5946.htm
英文引用格式: Zhang ZK, Pang WG, Xie WJ, Lü MS, Wang Y. Deep learning for real-time applications: A survey. Ruan Jian
Xue Bao/Journal of Software, 2020,31(9):2654−2677 (in Chinese). http://www.jos.org.cn/1000-9825/5946.htm
Deep Learning for Real-time Applications: A Survey
ZHANG Zheng-Kui, PANG Wei-Guang, XIE Wen-Jing, LÜ Ming-Song, WANG Yi
(Smart System Laboratory, School of Computer Science and Engineering, Northeastern University, Shenyang 110819, China)
Abstract: The persistent advance of deep learning algorithms and GPU computing power have promoted artificial intelligence in
various fields including but not limited to compute vision, speech recognition, and natural language processing. Meanwhile, deep learning
already began exploiting its usage in safety-critical areas exemplified by self-driving vehicles. Unfortunately, the successive severe traffic
accidents in the past two years manifest that deep learning technology is still far from mature to fulfill safety-critical standards, and
consequently the trustworthy artificial intelligence starts to attract a lot of research interests worldwide. This article conveys a
state-of-the-art survey of the research on deep learning for real-time applications. It first introduces the main problems and challenges
when deploying deep learning on the real-time embedded systems. Then, a detailed review covering various topics is provided, such as
deep neural network lightweight design, GPU timing analysis and workload scheduling, shared resource management on the CPU+GPU
SoC platform, deep neural network and network accelerator co-design. Finally, open issues and research directions are identified to
conclude the survey.
Key words: deep learninig; deep neural network; real-time systems; timing analysis; real-time scheduling; shared-resource interference
∗ 基金项目: 国家自然科学基金(61532007, 61772123); 装备预研教育部联合基金青年人才基金(6141A020333)
Foundation item: National Natural Science Foundation of China (61532007, 61772123); Ministry of Education Joint Foundation for
Equipment Pre-Research (6141A020333)
本文由“智能嵌入式系统”专题特约编辑王泉教授、吴中海教授、陈仪香教授、苗启广教授推荐.
收稿时间: 2019-07-08; 修改时间: 2019-08-18; 采用时间: 2019-11-02;jos 在线出版时间: 2019-12-05
CNKI 网络优先出版: 2019-12-05 14:55:12, http://kns.cnki.net/kcms/detail/11.2560.TP.20191205.1454.006.html
