TITLE:   Systems Identification using Reccurent 
	 Asymptotically Stable Neural Networks
	 
AUTHORS: Chris M. Jubien, Nikitas J. Dimopoulos 
	 
IN:      Proceedings of the 1993 IEEE Pacific Rim 
	 Conference on Communications, Computers and 
	 Signal Processing, Victoria, Canada, 
	 May 19-21, 1993, pp. 610-613. 

	 

ABSTRACT 

A training procedure for a class of neural networks 
that are asymptotically stable is presented. The 
training procedure is gradient method which adapts 
the interconnection weights as well as the relaxation 
constants and the slopes of the activation function 
used so as the error between the expected and obtained 
responses is minimized. A method for assuring that 
stability is maintained throughout the training procedure
is also given. Such a network was used to identify the 
dynamic behavior of several nonlinear dynamical systems
which included a PUMA-560 robot and boat based on 
collected rudder/heading data.