![[2602.15883] Distributed physics-informed neural networks via domain decomposition for fast flow reconstruction](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.15883] Distributed physics-informed neural networks via domain decomposition for fast flow reconstruction
Summary
This article presents a distributed framework for physics-informed neural networks (PINNs) aimed at efficient flow reconstruction, addressing computational challenges and pressure indeterminacy in large spatiotemporal domains.
Why It Matters
The research highlights advancements in using PINNs for flow reconstruction, which is crucial for applications in fluid dynamics. By overcoming computational bottlenecks and ensuring pressure consistency, this work paves the way for more accurate simulations in complex hydrodynamic scenarios, potentially impacting various engineering and scientific fields.
Key Takeaways
- Introduces a distributed PINNs framework for efficient flow reconstruction.
- Addresses pressure indeterminacy through innovative normalization strategies.
- Demonstrates near-linear strong scaling and high-fidelity reconstruction on complex benchmarks.
Computer Science > Machine Learning arXiv:2602.15883 (cs) [Submitted on 5 Feb 2026] Title:Distributed physics-informed neural networks via domain decomposition for fast flow reconstruction Authors:Yixiao Qian, Jiaxu Liu, Zewei Xia, Song Chen, Chao Xu, Shengze Cai View a PDF of the paper titled Distributed physics-informed neural networks via domain decomposition for fast flow reconstruction, by Yixiao Qian and 5 other authors View PDF Abstract:Physics-Informed Neural Networks (PINNs) offer a powerful paradigm for flow reconstruction, seamlessly integrating sparse velocity measurements with the governing Navier-Stokes equations to recover complete velocity and latent pressure fields. However, scaling such models to large spatiotemporal domains is hindered by computational bottlenecks and optimization instabilities. In this work, we propose a robust distributed PINNs framework designed for efficient flow reconstruction via spatiotemporal domain decomposition. A critical challenge in such distributed solvers is pressure indeterminacy, where independent sub-networks drift into inconsistent local pressure baselines. We address this issue through a reference anchor normalization strategy coupled with decoupled asymmetric weighting. By enforcing a unidirectional information flow from designated master ranks where the anchor point lies to neighboring ranks, our approach eliminates gauge freedom and guarantees global pressure uniqueness while preserving temporal continuity. Further...
