Nonlinear system identification using discrete-time neural networks with stable learning algorithm

This paper presents a stable neural sytem identification for nonlinear systems. An input output discrete time representation is considered. No a priori knowledge about the nonlinearities of the system is assumed. The proposed learning rule is the backpropagation algorithm under the condition that the learning rate belongs to a specified range defining the stability domain. Satisfying such condition, unstable phenomenon during the learning process is a-voided. A Lyapunov analysis is made in order to extract the new updating formulation which contain a set of inequality constraints. In the constrained learning rate algorithm, the learning rate is updated at each iteration by an equation derived using the stability conditions. As a case study, identification of two discrete time systems is considered to demonstrate the effectiveness of the proposed algorithm. Simulation results concerning the considered systems are presented.