PEST: Physics-Enhanced Swin Transformer for 3D Turbulence Simulation DOI: XXXXXXX.XXXXXXX Conference: Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining; August 3–7, 2025; Toronto, Canada ISBN: 978-1-4503-XXXX-X/2025/08
Yilong Dai, Shengyu Chen, Xiaowei Jia, Peyman Givi, Runlong Yu · 2026
Read PEST: Physics-Enhanced Swin Transformer for 3D Turbulence Simulation DOI: XXXXXXX.XXXXXXX Conference: Proceedings of the 31st ACM SIGKDD Conference on Knowledge Discovery and Data Mining; August 3–7, 2025; Toronto, Canada ISBN: 978-1-4503-XXXX-X/2025/08 by Yilong Dai, Shengyu Chen, Xiaowei Jia, Peyman Givi, Runlong Yu (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-10150.txt.
Description
Turbulent flows strain both direct number methods, which cost far too much, and data-driven methods, which wobble on long rollouts. A 2026 paper led by Yilong Dai, with four co-authors, offers PEST, a physics-boosted Swin transformer for 3D turbulence. Window-based self-attention models local equation links while keeping compute cheap.
A scale-tuned adaptive loss stresses small-scale structures for faithful high-end dynamics. Navier-Stokes residual constraints and divergence-free smoothing enter the learning goal itself, keeping physics consistent. Tests on two typical turbulent setups show accurate, physically steady, stable long-term simulations that beat current data-driven baselines.
Catalog record
| CREATOR | Yilong Dai, Shengyu Chen, Xiaowei Jia, Peyman Givi, Runlong Yu |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 73 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-10150