We give necessary and sufficient conditions for observability of a class of recurrent neural networks having a subsystem where the activation function is the identity. An algorithm for computing all pairs of indistinguishable states is also given.
Recurrent neural networks coupled with linear systems: Observability in continuous and discrete time
Francesca Albertini;Paolo Dai Pra
1996-01-01
Abstract
We give necessary and sufficient conditions for observability of a class of recurrent neural networks having a subsystem where the activation function is the identity. An algorithm for computing all pairs of indistinguishable states is also given.
File in questo prodotto:
| File | Dimensione | Formato | |
|---|---|---|---|
|
17.pdf
non disponibili
Dimensione
153.51 kB
Formato
Adobe PDF
|
153.51 kB | Adobe PDF | Visualizza/Apri Richiedi una copia |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
