PIRSA:25120038

Video URL

https://pirsa.org/25120038

Machine Learning through Equilibrium Propagation: Extending the framework to quantum, thermal, and time dependent systems. - Quantum Foundations Seminar

Massar, S. (2025). Machine Learning through Equilibrium Propagation: Extending the framework to quantum, thermal, and time dependent systems. - Quantum Foundations Seminar. Perimeter Institute for Theoretical Physics. https://pirsa.org/25120038

Massar, Serge. Machine Learning through Equilibrium Propagation: Extending the framework to quantum, thermal, and time dependent systems. - Quantum Foundations Seminar. Perimeter Institute for Theoretical Physics, Dec. 01, 2025, https://pirsa.org/25120038 

          @misc{ scivideos_PIRSA:25120038,
            doi = {10.48660/25120038},
            url = {https://pirsa.org/25120038},
            author = {Massar, Serge},
            keywords = {Quantum Foundations},
            language = {en},
            title = {Machine Learning through Equilibrium Propagation: Extending the framework to quantum, thermal, and time dependent systems. - Quantum Foundations Seminar},
            publisher = {Perimeter Institute for Theoretical Physics},
            year = {2025},
            month = {dec},
            note = {PIRSA:25120038 see, \url{https://scivideos.org/pirsa/25120038}}
          }

Serge Massar Université Libre de Bruxelles - Physique théorique

Talk numberPIRSA:25120038
DOI10.48660/25120038
Source RepositoryPIRSA
Collection
Talk Type Scientific Series
Subject

Abstract

In 2017, Scellier and Bengio introduced the machine learning algorithm known as Equilibrium Propagation, in which the stationary state of a dynamical system is exploited for computation. A canonical example involves a network of nonlinear coupled elements -such as springs- described by a Hamiltonian H, where learning is carried out by tuning the parameters of H (for instance, the spring constants) so that its ground state implements the desired computation.
 
In this seminar, I will present the basic principles of the Equilibrium Propagation algorithm and discuss several recent extensions. These include the case of systems at finite temperature and of quantum systems, the extension to time-dependent systems governed by Lagrangian dynamics, and the case of nonreciprocal dynamical systems. These developments broaden the applicability of Equilibrium Propagation, making some new connections between machine learning, quantum mechanics, statistical physics, and dynamical systems theory.