Title: A hyperbolic tangent replacement by third order polynomial approximation
Authors: Dario Baptista , Morgado Dias
Abstract: Artificial Neural Networks have been widely used in different fields ranging from social sciences to engineering. Many of these applications have reached a hardware implementation phase and have been reported in scientific papers. Unfortunately most of the implementations suffer from a low precision of the hyperbolic tangent replacement which has been the most common problem and the most resource consuming block in terms of hardware. This paper proposes a high resolution hyperbolic tangent substitute which is far more modest in consumed resources than most of the solutions proposed in the literature, by using the simplest solution in order to obtain the lowest error proposed so far with a set of 25 polynomials of third order, obtained with Chebyshev approximations.
Publication date: 2012-07-16
Online entry date: 2013-05-08
Conference: CONTROLO’2012
Publication pdf: Click for full publication