Synaptic learning in VLSI-based artificial nerve cells

Author

Laffely, Andrew J. ; Wolpert, Seth

Author_Institution

Dept. of Electr. Eng., Maine Univ., Orono, ME, USA

fYear

1993

fDate

18-19 Mar 1993

Firstpage

103

Lastpage

105

Abstract

A VLSI method for analog synaptic learning in an electronic neuronal model is presented. This method reduces the size and complexity involved in implementing adaptive neuronally based controllers for robotic motion. It also provides for a continuous range of synaptic weights at both excitatory and inhibitory inputs while anticipating the need to interface to a pulse-driven system. The system is described, and test results indicate that it is able to alter the synaptic coupling on an inhibitory or an excitory input over a wide range

Keywords

CMOS analogue integrated circuits; VLSI; buffer circuits; learning (artificial intelligence); neural chips; neurocontrollers; robot vision; sample and hold circuits; CMOS; Neuromime; VLSI-based artificial nerve cells; Widlar current source; adaptive neuronally based controllers; analog synaptic learning; buffer circuit; complexity; continuous range; electronic neuronal model; excitory input; inhibitory input; interface; one-chip neuronal controller; pulse-driven system; robotic motion; sample and hold circuit; synaptic coupling; synaptic weights; Adaptive control; Biological control systems; Cells (biology); Control systems; Motion control; Neurofeedback; Neurons; Programmable control; Pulse circuits; Robot control;

fLanguage

English

Publisher

ieee

Conference_Titel

Bioengineering Conference, 1993., Proceedings of the 1993 IEEE Nineteenth Annual Northeast

Conference_Location

Newark, NJ

Print_ISBN

0-7803-0925-1

Type

conf

DOI

10.1109/NEBC.1993.404399

Filename

404399