Stability of NNGD algorithm for nonlinear systemidentification
Danilo P. Mandic, Igor R. Krcmar · Electronics Letters · 2001
An analysis of stability and convergence for the recently proposed normalised nonlinear gradient descent algorithm is provided. It is shown that the value of the free parameter of the algorithm is a nonlinear function of the tap input power to the filter. Further, bounds on the value of the free parameter of the algorithm are derived. Simulation results show that such an algorithm outperforms the nonlinear gradient descent and normalised least mean square algorithm for nonlinear system identification.