AUTHORS: Nikitas J. Dimopoulos, Stephen W. Neville, Andrew Watkins, 
	 Kin F. Li, Eric G. Manning
	 
TITLE:   Neural Networks in Fault Identification in Large 
	 Communication Networks
	 
IN: 	 2nd IEEE Mediterranean Symposium on New Directions 
	 in Control and Automation, Maleme-Chania, Crete, 
	 Greece, pp. 25 - 31, June 19-22, 1994. 

ABSTRACT

In this work, we present our efforts in developing techniques for 
detecting the onset and diagnosis of a fault. The diagnosis domain 
in that of large Cable Television Networks. We are using stable 
recurrent neural networks to model the dynamic behavior of some of 
the measured parameters both for normal operation and during a fault.
Deviations from the model indicate the onset of a fault, while the 
properties of the behavior during a fault are indicative of the fault 
modality.