Time-varying moore-penrose inverse solving shows different zhang functions leading to different ZNN models

Abstract

A novel class of recurrent neural network (RNN), termed Zhang neural network (ZNN), has been proposed for solving online time-varying problems by Zhang et al since 2001. In this paper, by defining different Zhang functions (ZFs), we construct different ZNN models correspondingly solving for time-varying Moore-Penrose inverse (MPI). As an error-monitoring function, ZF is the basis of the ZNN design method and can be positive, zero, negative, bounded or even unbounded (including lower-unbounded). Computer simulation results further illustrate the excellent convergence performance of the proposed ZNN models for online time-varying MPI solving. © 2012 Springer-Verlag.