VLSI Implementation of a Neural Associative Memory and its Application to Vector Quantization

Abstract

In previous papers we have proposed a new high capacity associative memory which we call the exponential correlation associative memory (ECAM). In this paper we describe the VLSI design of a programmable ECAM which has been implemented and tested in 3 micron CMOS. The prototype chip is capable of storing 32 memory vectors of 24 bits each. The high capacity of the ECAM is partly due to the use of special exponentiation neurons, which are implemented via sub-threshold MOS transistors in this design. The prototype chip is capable of performing associative recall in 3 μs,and we demonstrate its capabilities for real time processing using binary vector quantization as an example application.