04:00
2026-09-15
arxiv.org
machine-learning
A derivative-fidelity failure mode in physics-informed neural networks: strengthened benchmark evidence from function-value training
A paper posted to arXiv (arXiv:2609.13171v1) formulates derivative fidelity as a failure mode of physics-informed neural networks (PINNs), showing that multilayer perceptrons trained only on function …