摘要
基于忆阻突触器件的硬件神经网络是神经形态计算的重要发展方向,是后摩尔时代突破传统冯·诺依曼计算架构的有力技术候选。综述了国内外忆阻硬件神经网络的近期发展现状,从器件发展和神经网络两个方面,详细阐述了忆阻器这一新兴信息器件在神经形态计算中所发挥的角色作用,讨论了依然存在的关键问题和技术挑战。忆阻器为实现存算一体化架构和超越摩尔定律提供了技术障碍突破的可行方案。
Hardware neural networks based on memristive synaptic devices,as a key development paradigm for neuromorphic computing,is a strong technology candidate to break the Von Neumann bottleneck in traditional computing architecture.In this paper,we present a survey of recent works in developing memristive hardware neural networks or neuromorphic systems.In particular,the role and prospect of memristive synaptic devices in neuromorphic computing is illustrated from two aspects,in terms of device and neural network.The challenges and outlook are also discussed.This emerging basic information device provides possible solutions to realize the computing-in-memory architecture beyond the Moore’s Law.
作者
陈佳
潘文谦
秦一凡
王峰
李灏阳
李祎
缪向水
CHEN Jia;PANWenqian;QIN Yifan;WANG Feng;LI Haoyang;LI Yi;MIAO Xiangshui(School of Optical and Electronic Information,and Wuhan National Laboratory for Optoelectronics,Huazhong University of Science and Technology,Wuhan 430074,China)
出处
《微纳电子与智能制造》
2019年第4期24-38,共15页
Micro/nano Electronics and Intelligent Manufacturing
基金
国家自然科学基金(61874164,61841404,51732003,61674061)项目资助.
关键词
神经形态计算
忆阻器
突触器件
硬件神经网络
neuromorphic computing
memristor
synaptic device
hardware neural network
作者简介
陈佳,博士,主要研究方向为忆阻突触器件及其神经网络应用。E-mail:chenjia_0816@163.com;通信作者:李祎,副教授,主要研究方向为高性能忆阻器及存算一体化技术。E-mail:liyi@hust.edu.cn
