Local Learning Rules for Out-of-Equilibrium Physical Generative Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Bösch, Cyrill, Roeder, Geoffrey, Serra-Garcia, Marc, Adams, Ryan P.
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912556416761856
author Bösch, Cyrill
Roeder, Geoffrey
Serra-Garcia, Marc
Adams, Ryan P.
author_facet Bösch, Cyrill
Roeder, Geoffrey
Serra-Garcia, Marc
Adams, Ryan P.
contents We show that the out-of-equilibrium driving protocol of score-based generative models (SGMs) can be learned via local learning rules. The gradient with respect to the parameters of the driving protocol is computed directly from force measurements or from observed system dynamics. As a demonstration, we implement an SGM in a network of driven, nonlinear, overdamped oscillators coupled to a thermal bath. We first apply it to the problem of sampling from a mixture of two Gaussians in 2D. Finally, we train a 12x12 oscillator network on the MNIST dataset to generate images of handwritten digits 0 and 1.
format Preprint
id arxiv_https___arxiv_org_abs_2506_19136
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Local Learning Rules for Out-of-Equilibrium Physical Generative Models
Bösch, Cyrill
Roeder, Geoffrey
Serra-Garcia, Marc
Adams, Ryan P.
Machine Learning
Mesoscale and Nanoscale Physics
Emerging Technologies
Neural and Evolutionary Computing
We show that the out-of-equilibrium driving protocol of score-based generative models (SGMs) can be learned via local learning rules. The gradient with respect to the parameters of the driving protocol is computed directly from force measurements or from observed system dynamics. As a demonstration, we implement an SGM in a network of driven, nonlinear, overdamped oscillators coupled to a thermal bath. We first apply it to the problem of sampling from a mixture of two Gaussians in 2D. Finally, we train a 12x12 oscillator network on the MNIST dataset to generate images of handwritten digits 0 and 1.
title Local Learning Rules for Out-of-Equilibrium Physical Generative Models
topic Machine Learning
Mesoscale and Nanoscale Physics
Emerging Technologies
Neural and Evolutionary Computing
url https://arxiv.org/abs/2506.19136