NEURAL NETWORKS FOR ARITHMETIC LOGIC FUNCTIONS

Abstract

The basic processing unit of a neural network is a linear threshold element. It has been known that neural networks can be much more powerful than traditional logic circuits in the computation of arithmetic logic functions. Whereas any logic circuit of polynomial size (in n) that computes the product of two n-bit numbers requires unbounded delay, such computations can be done in a neural network with "constant" delay. In this dissertation, application of neural networks in computation of arithmetic logic functions has been studied. Computer models of such neural networks have been developed which compute following common arithmetic logic functions in constant-depth. Functions considered are parity, binary count, multiplication, zero-mod-2c, divide by 2, comparision, majority function, maximum, sorting and merging.