类脑计算新发展——“TrueNorth”神经元芯片  

New Development of Artificial Cognitive Computation:TrueNorth Neuron Chip

作　　者：王雨辰[1] 胡华[1] 

机构地区：[1]清华大学精密仪器系,北京100084

出　　处：《计算机科学》2016年第S1期17-20 24,共5页Computer Science

摘　　要：类脑计算是一种基于神经网络的全新数据存储和计算技术,通过模拟大脑的工作机理,可以突破传统计算机处理大型问题时遇到的冯·诺依曼瓶颈,在显著提高信息处理速度的同时大幅降低功耗,并且具有自我学习和自适应能力。介绍了IBM最新研究的TrueNorth神经元芯片技术,包括其基本架构、工作原理、芯片性能、应用成果等,并展望了类脑计算技术的未来发展前景。Inspired by the brain's operating mechanism,artificial cognitive memory is a novel method based on neural networks and is integrated with self-learning and self-adaptive ability.In principle,the artificial cognitive memory can break the Von Neumann bottlenecks,and then significantly accelerate the information processing speed and reduce the power consumption.In this paper,the TrueNorth neuron chip developed by IBM Research was introduced in detail,including the fundamental architecture,the computing principle,the chip characteristics,and application results.In the end,the future development prospects of artificial cognitive memory were discussed.

关 键 词：类脑计算 神经突触 TrueNorth芯片 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP333[自动化与计算机技术—控制科学与工程]