Analysis of G-type model exploited for online ZLE solving

In this paper, the performance analysis of the model of gradient neural network (or termed G-type model), which was designed originally for solving constant linear equation, is investigated, analyzed and simulated for online solution of Zhang linear equation (ZLE or termed time-varying linear equation). Compared with the constant case, G-type model for online ZLE solving can only approximately approach its time-varying theoretical solution, instead of converging to it exactly. That is, the steady-state error between the solution of G-type model and the theoretical solution cannot vanish to zero. In order to understand this situation better, the upper bound of such an error is estimated firstly, and then the global exponential convergence rate is investigated for such a G-type model when approaching the error bound. Computer simulations substantiate the performance analysis of the G-type model exploited for online ZLE solving.