Scalable mean rate signal encoding analog neural network

Author

Berg, Yngvar ; Ruth, Jon-Erik ; Lande, Tor Sverre

Author_Institution

Dept. of Inf., Oslo Univ., Norway

Volume

3

fYear

1995

fDate

30 Apr-3 May 1995

Firstpage

1668

Abstract

A micropower feedforward neural network implementation in standard CMOS using pulse frequency modulated signal representation is presented in this paper. The implementation includes a modified backpropagation algorithm using analog UV-programmable floating gate memories for weight storage. Measurements of a 3-6-2 layered feedforward neural network implementation solving the 3-parity problem are shown

Keywords

CMOS analogue integrated circuits; analogue processing circuits; analogue storage; backpropagation; encoding; feedforward neural nets; neural chips; pulse frequency modulation; signal representation; 3-parity problem solution; PFM signal representation; analog UV-programmable floating gate memories; analog neural network; micropower feedforward neural network; modified backpropagation algorithm; pulse frequency modulation; scalable mean rate signal encoding; standard CMOS; weight storage; Biological neural networks; Circuits; Encoding; Feedforward neural networks; Feedforward systems; Neural networks; Neurons; Pulse measurements; Stochastic processes; Stochastic resonance;

fLanguage

English

Publisher

ieee

Conference_Titel

Circuits and Systems, 1995. ISCAS '95., 1995 IEEE International Symposium on

Conference_Location

Seattle, WA

Print_ISBN

0-7803-2570-2

Type

conf

DOI

10.1109/ISCAS.1995.523731

Filename

523731