Pré-Publication, Document De Travail Année : 2024

Training of Physical Neural Networks

Ali Momeni
  • Fonction : Auteur
Babak Rahmani
  • Fonction : Auteur
Benjamin Scellier
  • Fonction : Auteur
Logan G. Wright
  • Fonction : Auteur
Peter L. Mcmahon
  • Fonction : Auteur
Clara C. Wanjura
  • Fonction : Auteur
Yuhang Li
  • Fonction : Auteur
Natalia G. Berloff
  • Fonction : Auteur
Tatsuhiro Onodera
  • Fonction : Auteur
Ilker Oguz
  • Fonction : Auteur
Francesco Morichetti
  • Fonction : Auteur
Manuel Le Gallo
  • Fonction : Auteur
Abu Sebastian
  • Fonction : Auteur
Azalia Mirhoseini
  • Fonction : Auteur
Cheng Zhang
  • Fonction : Auteur
Christophe Moser
  • Fonction : Auteur
Florian Marquardt
  • Fonction : Auteur
Aydogan Ozcan
  • Fonction : Auteur
Andrea J. Liu
  • Fonction : Auteur
Demetri Psaltis
  • Fonction : Auteur
Andrea Alù
  • Fonction : Auteur
Romain Fleury
  • Fonction : Auteur

Résumé

Physical neural networks (PNNs) are a class of neural-like networks that leverage the properties of physical systems to perform computation. While PNNs are so far a niche research area with small-scale laboratory demonstrations, they are arguably one of the most underappreciated important opportunities in modern AI. Could we train AI models 1000x larger than current ones? Could we do this and also have them perform inference locally and privately on edge devices, such as smartphones or sensors? Research over the past few years has shown that the answer to all these questions is likely "yes, with enough research": PNNs could one day radically change what is possible and practical for AI systems. To do this will however require rethinking both how AI models work, and how they are trained - primarily by considering the problems through the constraints of the underlying hardware physics. To train PNNs at large scale, many methods including backpropagation-based and backpropagation-free approaches are now being explored. These methods have various trade-offs, and so far no method has been shown to scale to the same scale and performance as the backpropagation algorithm widely used in deep learning today. However, this is rapidly changing, and a diverse ecosystem of training techniques provides clues for how PNNs may one day be utilized to create both more efficient realizations of current-scale AI models, and to enable unprecedented-scale models.
Fichier principal
Vignette du fichier
2024 - Training of Physical Neural Networks.pdf (2) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04802998 , version 1 (17-02-2025)

Identifiants

Citer

Ali Momeni, Babak Rahmani, Benjamin Scellier, Logan G. Wright, Peter L. Mcmahon, et al.. Training of Physical Neural Networks. 2025. ⟨hal-04802998⟩
0 Consultations
0 Téléchargements

Altmetric

Partager

More