English

AutoBalance: An Automatic Balancing Framework for Training Physics-Informed Neural Networks

Machine Learning 2025-10-09 v1 Numerical Analysis Numerical Analysis Optimization and Control

Abstract

Physics-Informed Neural Networks (PINNs) provide a powerful and general framework for solving Partial Differential Equations (PDEs) by embedding physical laws into loss functions. However, training PINNs is notoriously difficult due to the need to balance multiple loss terms, such as PDE residuals and boundary conditions, which often have conflicting objectives and vastly different curvatures. Existing methods address this issue by manipulating gradients before optimization (a "pre-combine" strategy). We argue that this approach is fundamentally limited, as forcing a single optimizer to process gradients from spectrally heterogeneous loss landscapes disrupts its internal preconditioning. In this work, we introduce AutoBalance, a novel "post-combine" training paradigm. AutoBalance assigns an independent adaptive optimizer to each loss component and aggregates the resulting preconditioned updates afterwards. Extensive experiments on challenging PDE benchmarks show that AutoBalance consistently outperforms existing frameworks, achieving significant reductions in solution error, as measured by both the MSE and LL^{\infty} norms. Moreover, AutoBalance is orthogonal to and complementary with other popular PINN methodologies, amplifying their effectiveness on demanding benchmarks.

Keywords

Cite

@article{arxiv.2510.06684,
  title  = {AutoBalance: An Automatic Balancing Framework for Training Physics-Informed Neural Networks},
  author = {Kang An and Chenhao Si and Ming Yan and Shiqian Ma},
  journal= {arXiv preprint arXiv:2510.06684},
  year   = {2025}
}

Comments

23 pages

R2 v1 2026-07-01T06:23:08.922Z