Guardado en:
Detalles Bibliográficos
Autores principales: Koupaï, Armand Kassaï, Boudec, Lise Le, Serrano, Louis, Gallinari, Patrick
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2506.06158
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866908705173274624
author Koupaï, Armand Kassaï
Boudec, Lise Le
Serrano, Louis
Gallinari, Patrick
author_facet Koupaï, Armand Kassaï
Boudec, Lise Le
Serrano, Louis
Gallinari, Patrick
contents Solving time-dependent parametric partial differential equations (PDEs) remains a fundamental challenge for neural solvers, particularly when generalizing across a wide range of physical parameters and dynamics. When data is uncertain or incomplete-as is often the case-a natural approach is to turn to generative models. We introduce ENMA, a generative neural operator designed to model spatio-temporal dynamics arising from physical phenomena. ENMA predicts future dynamics in a compressed latent space using a generative masked autoregressive transformer trained with flow matching loss, enabling tokenwise generation. Irregularly sampled spatial observations are encoded into uniform latent representations via attention mechanisms and further compressed through a spatio-temporal convolutional encoder. This allows ENMA to perform in-context learning at inference time by conditioning on either past states of the target trajectory or auxiliary context trajectories with similar dynamics. The result is a robust and adaptable framework that generalizes to new PDE regimes and supports one-shot surrogate modeling of time-dependent parametric PDEs.
format Preprint
id arxiv_https___arxiv_org_abs_2506_06158
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle ENMA: Tokenwise Autoregression for Generative Neural PDE Operators
Koupaï, Armand Kassaï
Boudec, Lise Le
Serrano, Louis
Gallinari, Patrick
Machine Learning
Solving time-dependent parametric partial differential equations (PDEs) remains a fundamental challenge for neural solvers, particularly when generalizing across a wide range of physical parameters and dynamics. When data is uncertain or incomplete-as is often the case-a natural approach is to turn to generative models. We introduce ENMA, a generative neural operator designed to model spatio-temporal dynamics arising from physical phenomena. ENMA predicts future dynamics in a compressed latent space using a generative masked autoregressive transformer trained with flow matching loss, enabling tokenwise generation. Irregularly sampled spatial observations are encoded into uniform latent representations via attention mechanisms and further compressed through a spatio-temporal convolutional encoder. This allows ENMA to perform in-context learning at inference time by conditioning on either past states of the target trajectory or auxiliary context trajectories with similar dynamics. The result is a robust and adaptable framework that generalizes to new PDE regimes and supports one-shot surrogate modeling of time-dependent parametric PDEs.
title ENMA: Tokenwise Autoregression for Generative Neural PDE Operators
topic Machine Learning
url https://arxiv.org/abs/2506.06158