人工智能芯片技术体系研究综述
|
摘要
人工智能技术是当前各国关注的新焦点。人工智能技术的发展对计算芯片提出了新的需求,深度学习算法需要海量数据的训练,而传统计算架构无法支撑深度学习算法的大规模计算需求,因此新架构的人工智能芯片层出不穷。分析了人工智能芯片不同的技术路线,比较了不同路线的特点,研究了人工智能芯片产业全球及我国的发展态势,分析了我国人工智能芯片发展面临的机遇与挑战,并对未来人工智能芯片技术发展趋势进行了展望。
Artificial intelligence technology is the new focus of current countries. The development of artificial intelligence technology has put forward new requirements for computing chips. Deep learning algorithms require the training of massive data, while traditional computing architectures can't support the large-scale computing requirements of deep learning algorithms. Therefore, artificial intelligence chips of new architectures emerge one after another. The different technical routes of artificial intelligence chips were analyzed, the characteristics of different routes were compared, the development trend of artificial intelligence chip industry and studied, the opportunities and challenges of artificial intelligence chip development in China were analyzed, and the future development of artificial intelligence chip technology was forecasted.
Artificial intelligence technology is the new focus of current countries. The development of artificial intelligence technology has put forward new requirements for computing chips. Deep learning algorithms require the training of massive data, while traditional computing architectures can't support the large-scale computing requirements of deep learning algorithms. Therefore, artificial intelligence chips of new architectures emerge one after another. The different technical routes of artificial intelligence chips were analyzed, the characteristics of different routes were compared, the development trend of artificial intelligence chip industry and studied, the opportunities and challenges of artificial intelligence chip development in China were analyzed, and the future development of artificial intelligence chip technology was forecasted.
引文
[1]JOUPPI N P.In-datacenter performance analysis of a tensor processing unit[Z].2017.
[2]CHEN Y H,KRISHNA T,EMER J S,et al.Eyeriss:an energy-efficient reconfigurable accelerator for deep convolutional neural networks[J].IEEE Journal of Solid-State Circuits,2017,52(1):127-138.
[3]LECUN Y,BENGIO Y,HINTON G.Deep learning[J].Nature,2015(5):436-444.
[4]ABADI M,AGARWAL A,BARHAM P,et al.TensorFlow:large-scale machine learning on heterogeneous distributed systems[J].arXiv:1603.04467.
[5]SUNDARARAJAN N,SARATCHANDRAN P.Parallel architectures for artificial networks:paradigms and implementations[M].New York:ACM Press,1998:23-27.
[6]NVIDIA DGX-1.Deep learning[Z].2018.
[7]BENGIO Y,COURVILLE A,VINCENT P.Representation learning:a review and new perspectives[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2013,35(8):1798-1828.
[8]BENJAMIN B V,GAO P,MCQUINN E,et al.Neurogrid:a mixed-analog-digital multichip system for large-scale neural simulations[J].Proceedings of the IEEE,2014,102(5):699-716.
[9]DIKOV G,FIROUZI M,ROHRBEIN F,et al.Spiking cooperative stereo-matching at 2 ms latency with neuromorphic hardware[M]//Biomimetic and Biohybrid Systems.Berlin:Springer,2017:119-137.
[10]施羽暇.人工智能芯片技术研究[J].电信网技术,2016(12):11-13.SHI Y X.Research on artificial intelligence process chip technology[J].Network Technology,2016(12):11-13.
[11]施羽暇.促进人工智能芯片发展,助力产业基础实力提升[J].信息通信技术与政策,2018(11):76-79.SHI Y X.Promoting the development of artificial intelligence chips,enhancing the industrial foundation[J].Information and Communications Technology and Policy,2018(11):76-79.
[2]CHEN Y H,KRISHNA T,EMER J S,et al.Eyeriss:an energy-efficient reconfigurable accelerator for deep convolutional neural networks[J].IEEE Journal of Solid-State Circuits,2017,52(1):127-138.
[3]LECUN Y,BENGIO Y,HINTON G.Deep learning[J].Nature,2015(5):436-444.
[4]ABADI M,AGARWAL A,BARHAM P,et al.TensorFlow:large-scale machine learning on heterogeneous distributed systems[J].arXiv:1603.04467.
[5]SUNDARARAJAN N,SARATCHANDRAN P.Parallel architectures for artificial networks:paradigms and implementations[M].New York:ACM Press,1998:23-27.
[6]NVIDIA DGX-1.Deep learning[Z].2018.
[7]BENGIO Y,COURVILLE A,VINCENT P.Representation learning:a review and new perspectives[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2013,35(8):1798-1828.
[8]BENJAMIN B V,GAO P,MCQUINN E,et al.Neurogrid:a mixed-analog-digital multichip system for large-scale neural simulations[J].Proceedings of the IEEE,2014,102(5):699-716.
[9]DIKOV G,FIROUZI M,ROHRBEIN F,et al.Spiking cooperative stereo-matching at 2 ms latency with neuromorphic hardware[M]//Biomimetic and Biohybrid Systems.Berlin:Springer,2017:119-137.
[10]施羽暇.人工智能芯片技术研究[J].电信网技术,2016(12):11-13.SHI Y X.Research on artificial intelligence process chip technology[J].Network Technology,2016(12):11-13.
[11]施羽暇.促进人工智能芯片发展,助力产业基础实力提升[J].信息通信技术与政策,2018(11):76-79.SHI Y X.Promoting the development of artificial intelligence chips,enhancing the industrial foundation[J].Information and Communications Technology and Policy,2018(11):76-79.