On a new class of neural adaptive FIR filters
Mauro Forti, A. Liberatore, S. Manetti · 1991
Basic aspects of a recently introduced class of neural networks for adaptive FIR filtering are analyzed. The main quantities characterizing the transient motion, such as setting time and errors of computation, are analytically evaluated. In addition, an interpretation is given of the adaptive neural filter computation in terms of both correlation functions and a continuous-time gradient-search technique applied on the least squares error surface.>