Circuit realization of a programmable neuron transfer function and its derivative

Author

Lu, Chun ; Shi, Bingxue

Author_Institution

Inst. of Microelectron., Tsinghua Univ., Beijing, China

Volume

4

fYear

2000

fDate

2000

Firstpage

47

Abstract

In on-chip back-propagation learning neural networks, both a sigmoidal transfer function and its derivative are required. A simple CMOS analog neuron circuit that can realizes both functions is proposed. The neuron is widely applicable because of its programmability. Based on this novel neuron, a two-layer feedforward artificial neural network (ANN) is designed. HSPICE simulation results has proved its ability to solve the XOR problem

Keywords

CMOS analogue integrated circuits; SPICE; backpropagation; feedforward neural nets; multilayer perceptrons; transfer functions; ANN; CMOS analog neuron circuit; HSPICE simulation; XOR problem; backpropagation; circuit realization; on-chip back-propagation learning neural networks; programmable neuron transfer function derivative; sigmoidal transfer function; two-layer feedforward artificial neural network; Artificial neural networks; CMOS analog integrated circuits; Circuit simulation; Inverters; Microelectronics; Network-on-a-chip; Neurons; Piecewise linear approximation; Transfer functions; Voltage;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks, 2000. IJCNN 2000, Proceedings of the IEEE-INNS-ENNS International Joint Conference on

Conference_Location

Como

ISSN

1098-7576

Print_ISBN

0-7695-0619-4

Type

conf

DOI

10.1109/IJCNN.2000.860747

Filename

860747