Abstract:

	In this work, we present our efforts in developing techniques 
for detecting the onset and diagnosis of a fault. The Diagnosis domain 
is 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.