A feed-forward neural circuit for ranking N numbers using O (N) neurons

Abstract

Ranking or ordering of a set of numbers based on their relative magnitudes is a fundamental operation in computing. In this paper, we propose a neural circuit for ranking a given set of numbers. The proposed network does not require any feedback connection, requires fewer neurons, and fewer interconnections between neurons as compared to existing schemes. Moreover, the numbers to be sorted are applied as inputs to the circuit unlike some existing schemes which impress the numbers to be sorted as initial conditions on the network. Results of PSPICE simulation confirm the theory proposed.