← Papers

Data-driven discretization

Learning data driven discretizations for partial differential equations

Paper title Learning data driven discretizations for partial differential equations

Bar-Sinai et al. (Harvard, Google Research) · General PDEs · Public: code

Paper arXiv 1808.04930PDF

A neural network reads the coarse solution around each cell boundary and outputs the finite-difference coefficients used to estimate the spatial derivatives there. The estimates feed a finite-volume flux, which keeps the scheme conservative, and training is end-to-end against coarse-grained high-resolution solutions of the known equation.

Published
PNAS

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