A counter-propagation neural network for function approximation

Abstract

A counterpropagation network architecture for continuous function approximation is introduced. The paradigm consists of a splitting Kohonen layer architecture, functional-link network, continuous activation functions, and a modified training procedure. The network mapping capabilities are analyzed. To demonstrate the applicability of the network, simulation results for the robot inverse kinematic problem are provided. They show an improved function approximation accuracy compared to standard counterpropagation networks.< <ETX xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">&gt;</ETX>