PIECEWISE-LINEAR APPROXIMATION APPLIED TO NONLINEAR FUNCTION OF A NEURAL-NETWORK

An efficient piecewise linear approximation of a nonlinear function (P LAN) is proposed. This uses a simple digital gate design to perform a direct transformation from X to Y, where X is the input and Y is the a pproximated sigmoidal output. This PLAN is then used within the output s of an artificial neural network to perform the nonlinear approximati on. :The comparison of this technique with two other sigmoidal approxi mation techniques for digital circuits is presented and the results sh ow that the fast and compact digital circuit proposed produces the clo sest approximation to the sigmoid function. The hardware implementatio n of PLAN has been verified by a VHDL simulation with Mentor Graphics running under the UNIX operating system.