An adaptive analog synapse circuit that implements the least-mean-square learning rule

In this paper a compact adaptive analog synapse circuit that implements the least-mean-square (LMS) learning rule is described. Basic simulation results demonstrate the LMS learning rule in the proposed circuit. An adaptive linear combiner that uses the proposed synapse is shown to learn a square wave that matches closely with the desired target. Issues of weight decay and its implications to the design of the synapse circuit are presented as well. The synapse is designed in a 0.5 /spl mu/m CMOS technology.

I. Introduction

An adaptive system is one whose structure is adjustable such that its performance improves through its contact with the environment. These systems are inherently non-linear and time varying in nature. The key advantages to such systems are that these are self-optimizing and can be trained to perform specific filtering or decision-making tasks. Adaptive filters find use in a variety of applications such as adaptive equalization, prediction, system identification and interference cancellation.