CMOS analog neural network systems based on oscillatory neurons

Abstract

This paper addresses the design of two neural network systems based on the use of pulsing neurons. Each neuron is built as a simple voltage controlled oscillator (VCO) whose control voltage makes the circuit to oscillate or not. The interconnecting synapses between neurons are made with programmable transconductance amplifiers. The weight of each synapse is represented by the transconductance gain of the amplifiers, and is externally adjustable for each synapse. Using this technique a 5-neurons Hopfield network and a 6- neurons BAM network were built and tested, using standard 2μm double-metal, double-polysilicon CMOS (MOSIS) chips.