基于光电器件的神经形态类脑芯片研究进展  

Progress in neuromorphic brain-like chips based on optoelectronic devices

作　　者：陈子龙 张欢 程传同 毛旭瑞 黄北举[2,3] CHEN Zilong;ZHANG Huan;CHENG Chuantong;MAO Xurui;HUANG Beiju(Syracuse University,New York 13244;Institute of brain-computer intelligence,Jiangsu industrial technology research institute,Suzhou,215133;Institute of Semiconductors,Chinese Academy of Sciences,Beijing,100083)

机构地区：[1]美国雪城大学,纽约13244 [2]江苏省产业技术研究院脑机智能研究所,苏州215133 [3]中国科学院半导体研究所,北京100083

出　　处：《微纳电子与智能制造》2021年第1期174-180,共7页Micro/nano Electronics and Intelligent Manufacturing

摘　　要：随着信息技术和大数据时代的迅猛发展,计算效率和速度逐渐受到冯·诺依曼架构的制约。神经形态类脑计算通过模拟大脑的运行机制实现存算一体化逐渐成为研究热点。光电器件和神经元遵从的动力学具有数学同构性,利用这种同构性可以实现基于光电器件的神经形态类脑芯片,实现类脑的学习、决策、认知等数据处理和分析功能。系统总结了光电类脑器件的工作原理,基于光电类脑器件构建的人工神经元,以及在此基础上构建的神经形态类脑芯片。随着基于光电器件的神经形态类脑芯片逐渐向着高集成度和低功耗方向发展,其必将会在神经形态类脑计算领域扮演越来越重要的角色。With the rapid development of information technology and big data era,the computational efficiency and speed are gradually restricted by the von Neumann architecture.The neuromorphic brain-like computing has become a hot topic by simulating the running mechanism of the brain.The kinetics of optoelectronic devices and neuron compliance have mathematical isomorphism.With this isomorphism,neuromorphic brain-like chips based on optoelectronic devices can be realized,and then data processing and analysis functions such as brain-like learning,decision-making,and cognition can be realized.This paper systematically summarizes the working principle of optoelectronic brain-like devices,artificial neurons based on optoelectronic brain-like devices,and neuromorphic brain-like chips built on this basis.As optoelectronic device-based neuromorphic brain chips are gradually moving toward high integration and low power consumption,they will play an increasingly important role in neuromorphic brain-like computing.

关 键 词：硅基光电子 人工神经元 类脑芯片 

分 类 号：TN29[电子电信—物理电子学]