Study of the effects of nonidealities in multilayer analog neural networks with circuit level simulation

Author

Simsek, A. ; Civelek, M.L. ; Dündar, G.

Author_Institution

Dept. of Electr. & Electron. Eng., Bogazici Univ., Istanbul, Turkey

Volume

2

fYear

1996

fDate

13-16 May 1996

Firstpage

613

Abstract

In this study, an analog multilayer neural network was designed using 2 micron SCMOS technology and simulated via SPICE. The simulation results were compared with original training results which were obtained from a neural network simulator, and the discrepancies were investigated. Among the reasons for the discrepancies are limited precision in synapse weights, synapse function non-idealities, and neuron function non-idealities. Simulations were done to study the effects of various degrees of weight precision. Non-ideal operation of synapses and neurons were studied and this behavior was incorporated into the training software to reduce the error resulting from circuit non-idealities

Keywords

CMOS analogue integrated circuits; SPICE; multilayer perceptrons; neural chips; 2 micron SCMOS technology; SPICE; circuit level simulation; multilayer analog neural networks; neural network simulator; nonidealities; Circuit simulation; Intelligent networks; Multi-layer neural network; Neural networks; Neurons; Predictive models; Quantization; SPICE; Variable structure systems; Voltage;

fLanguage

English

Publisher

ieee

Conference_Titel

Electrotechnical Conference, 1996. MELECON '96., 8th Mediterranean

Conference_Location

Bari

Print_ISBN

0-7803-3109-5

Type

conf

DOI

10.1109/MELCON.1996.551295

Filename

551295