Titre :

Realization Theory Of Recurrent Neural ODEs Using Polynomial System Embeddings

Auteur(s) :

Gonzalez, Martin [Auteur]
IRT SystemX
Defourneau, Thibault [Auteur]
Hajri, Hatem [Auteur]
IRT SystemX
Petreczky, Mihaly [Auteur]
Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189 [CRIStAL]

Titre de la revue :

Systems & Control Letters

Pagination :

105468

Date de publication :

2023-03-01

Mot(s)-clé(s) en anglais :

Realization theory
Neural ODEs
Recurrent Neural Networks
Long Short-Term Memory
System Identification

Discipline(s) HAL :

Informatique [cs]/Intelligence artificielle [cs.AI]
Informatique [cs]/Automatique

Résumé en anglais : [en]

In this paper we show that neural ODE analogs of recurrent (ODE-RNN) and Long Short-Term Memory (ODE-LSTM) networks can be algorithmically embedded into the class of polynomial systems. This embedding preserves input-output ...
Lire la suite >In this paper we show that neural ODE analogs of recurrent (ODE-RNN) and Long Short-Term Memory (ODE-LSTM) networks can be algorithmically embedded into the class of polynomial systems. This embedding preserves input-output behavior and can suitably be extended to other neural DE architectures. We then use realization theory of polynomial systems to provide necessary conditions for an input-output map to be realizable by an ODE-LSTM and sufficient conditions for minimality of such systems. These results represent the first steps towards realization theory of recurrent neural ODE architectures, which is is expected be useful for model reduction and learning algorithm analysis of recurrent neural ODEs.Lire moins >

Langue :

Anglais

Vulgarisation :

Non

Commentaire :

10 pages. Corrected typos and added references

Collections :

• Centre de Recherche en Informatique, Signal et Automatique de Lille (CRIStAL) - UMR 9189

Source :

Harvested from HAL