Neural semi-Lagrangian method for high-dimensional advection-diffusion problems

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Franck, Emmanuel, Michel-Dansac, Victor, Navoret, Laurent, Vigon, Vincent
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917101672857600
author Franck, Emmanuel
Michel-Dansac, Victor
Navoret, Laurent
Vigon, Vincent
author_facet Franck, Emmanuel
Michel-Dansac, Victor
Navoret, Laurent
Vigon, Vincent
contents This work is devoted to the numerical approximation of high-dimensional advection-diffusion equations. It is well-known that classical methods, such as the finite volume method, suffer from the curse of dimensionality, and that their time step is constrained by a stability condition. The semi-Lagrangian method is known to overcome the stability issue, while recent time-discrete neural network-based approaches overcome the curse of dimensionality. In this work, we propose a novel neural semi-Lagrangian method that combines these last two approaches. It relies on projecting the initial condition onto a finite-dimensional neural space, and then solving an optimization problem, involving the backwards characteristic equation, at each time step. It is particularly well-suited for implementation on GPUs, as it is fully parallelizable and does not require a mesh. We provide rough error estimates present several high-dimensional numerical experiments to assess the performance of our approach, and compare it to other neural methods.
format Preprint
id arxiv_https___arxiv_org_abs_2504_20715
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Neural semi-Lagrangian method for high-dimensional advection-diffusion problems
Franck, Emmanuel
Michel-Dansac, Victor
Navoret, Laurent
Vigon, Vincent
Numerical Analysis
2020: 65M25, 76R05, 65M15, 68T07
This work is devoted to the numerical approximation of high-dimensional advection-diffusion equations. It is well-known that classical methods, such as the finite volume method, suffer from the curse of dimensionality, and that their time step is constrained by a stability condition. The semi-Lagrangian method is known to overcome the stability issue, while recent time-discrete neural network-based approaches overcome the curse of dimensionality. In this work, we propose a novel neural semi-Lagrangian method that combines these last two approaches. It relies on projecting the initial condition onto a finite-dimensional neural space, and then solving an optimization problem, involving the backwards characteristic equation, at each time step. It is particularly well-suited for implementation on GPUs, as it is fully parallelizable and does not require a mesh. We provide rough error estimates present several high-dimensional numerical experiments to assess the performance of our approach, and compare it to other neural methods.
title Neural semi-Lagrangian method for high-dimensional advection-diffusion problems
topic Numerical Analysis
2020: 65M25, 76R05, 65M15, 68T07
url https://arxiv.org/abs/2504.20715