PIVONet: A Physically-Informed Variational Neuro ODE Model for Efficient Advection-Diffusion Fluid Simulation
Hei Shing Cheung, Qicheng Long, Zhiyue Lin · 2026
Read PIVONet: A Physically-Informed Variational Neuro ODE Model for Efficient Advection-Diffusion Fluid Simulation by Hei Shing Cheung, Qicheng Long, Zhiyue Lin (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-2601-03397.txt.
Description
Hei Shing Cheung, Qicheng Long, and Zhiyue Lin present PIVONet, a learned simulator for fluids governed by advection and diffusion. The model keeps the physics inside it. It unifies neural ordinary differential equations with continuous normalizing flows. A physics-based model, parameterized through the flow, is trained offline. It becomes a fast stand-in for one specific fluid system. Later runs no longer need to integrate the full dynamics step by step.
A second part of the model captures the hidden randomness in observed fluid paths. The network output is treated as a variational distribution, and a pathwise evidence lower bound is optimized. That allows stochastic integration of the equations. The randomness lets the stand-in reproduce turbulence and random fluctuations, not only the average flow. It stays cheap enough for visualization and repeated runs.
Catalog record
| CREATOR | Hei Shing Cheung, Qicheng Long, Zhiyue Lin |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-03397