← Papers
Latent models

CORAL

Operator learning with neural fields on general geometries

Paper title Operator Learning with Neural Fields: Tackling PDEs on General Geometries

Serrano et al. (Sorbonne Université) · Aero & CFD · Structures & crash · General PDEs · Public: code

Paper arXiv 2306.07266PDF

Each input and output field is compressed into a short latent code by a shared coordinate network (an implicit neural representation whose layers are shifted by the code), found with a few gradient steps. A small network then maps the input code to the output code, or a neural ODE steps the code forward in time, and the output network can be queried at any point of any mesh.

Published
NeurIPS

Share on LinkedIn, X or Bluesky, or or .

Know related work that belongs here?

Something wrong on this page?

More papers

All papers
  1. Latent modelsD-JEPA — One geometry latent, reused across flow conditions and physics
  2. Latent modelsCourant — A state-adaptive Perceiver surrogate with local support
  3. Latent modelsAeroJEPA — Semantic Latent Representations for 3D Aerodynamic Fields