混合SETMOS神经元分段线性输出函数的实现  

Implementation of Piecewise Linear Output Function of a Neuron by Hybrid SETMOS

作　　者：冯朝文[1] 蔡理[1] 吴刚[1] 

机构地区：[1]空军工程大学理学院,西安710051

出　　处：《微纳电子技术》2008年第5期255-259,共5页Micronanoelectronic Technology

基　　金：陕西省自然科学基础研究计划项目(2005F20);空军工程大学理学院科研资助项目(2005ZK19)

摘　　要：提出了一种结构简单的神经元分段线性输出函数SETMOS实现方式。理论分析了恒流源偏置下单电子晶体管(SET)器件的特性和神经元分段线性输出函数。利用SET和金属氧化物场效应晶体管(MOSFET)混合结构实现了分段线性输出函数电路,并通过仿真分析得到了分段线性特性,提出了具体相应的调节方法,验证了该混合结构功能的正确性。结果表明,该混合电路具有结构简单,nm级特征尺寸,分段线性度好,静态功耗极低,约200nW,驱动负载工作能力强,输出电压可达几百毫伏,易于大规模神经网络电路的实现及应用和集成度的进一步提高。A simple implementation method of the piecewise linear output function of a neuron by SETMOS structure was proposed. The characteristic of a single electron transistor （SET） device biased with a constant current source and the piecewise linear output function of a neuron were analyzed theoretically. Then the circuit of the piecewise linear output function was realized by the hybrid structure of SETs and metal oxide semiconductor field effect transistors （MOSFETs）. The piecewise linear characteristic was obtained by the simulation and the concrete method of adjustment was proposed correspondingly. The function of the hybrid structure was verified correctly. The result shows that the hybrid circuit has the following characteristics： simple structure, nanolevel feature size, fine piecewise linearity, ultra-low static power dissipation of 200 nW or so, and strength capability of driving loads with several hundred millivolts of output. So the large-scale neural networks can be realized and implemented easily. The integrated density of the circuit is further improved.

关 键 词：单电子晶体管 库仑阻塞振荡 双栅极 神经元 分段线性输出函数 

分 类 号：TN402[电子电信—微电子学与固体电子学] TN389