Design of a 4-valued digital multiplier using an artificial heterogeneous two-layered neural network

Abstract

A two-layered parallel cascaded neural network using heterogeneous binary and ternary artificial neurons is used to implement the four-valued digital multiplication process. The hardware design is derived from a general M-valued perceptron mapping theory. It requires only three binary neurons and one ternary neuron. The use of heterogeneous perceptron mapping techniques allows the design to be reasonably simple and accurate.<>