← Papers

PINN failure modes

Characterizing possible failure modes in physics-informed neural networks

Paper title Characterizing possible failure modes in physics-informed neural networks

Krishnapriyan et al. · General PDEs · Public: code

Paper arXiv 2109.01050PDF

Trains PINNs on simple convection, reaction and reaction–diffusion equations as the coefficients grow, and studies the loss landscape. Proposes two fixes: a curriculum that starts from an easy coefficient and raises it, and sequence-to-sequence learning that solves one time window at a time.

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