← Papers

Solver-in-the-Loop

Learning corrections with a differentiable solver in the training loop

Paper title Solver-in-the-Loop: Learning from Differentiable Physics to Interact with Iterative PDE-Solvers

Um et al. (TUM) · Aero & CFD · General PDEs · Public: code

Paper arXiv 2007.00016PDF

A network corrects a coarse numerical solver step by step; during training the differentiable solver is unrolled for many steps, so the network sees the states its own corrections produce and is trained on the error after the whole rollout.

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. PhysicsNeMo-Mesh — GPU-Native Meshes and Simulation Cases for Physics AI
  2. Multi-fidelity scaling — Compute-optimal data scaling via multi-fidelity training
  3. CAE-UQ — Predictive Uncertainty for Neural CAE Surrogates