Koopman Autoencoders with Continuous-Time Latent Dynamics for Fluid Dynamics Forecasting
Rares Grozavescu, Pengyu Zhang, Mark Girolami, Etienne Meunier · 2026
Read Koopman Autoencoders with Continuous-Time Latent Dynamics for Fluid Dynamics Forecasting by Rares Grozavescu, Pengyu Zhang, Mark Girolami, Etienne Meunier (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-2602-02832.txt.
Description
Forecasting a physical system and recovering its hidden state from sparse data are paired in a 2026 paper by Rares Grozavescu and 3 co-authors. Learned surrogate models usually excel at forecasting only. The authors test whether one model can do both jobs well.
Their continuous-time Koopman autoencoder compresses the state and moves it with a simple linear rule. That rule gives forecasts at any time horizon in one step. Experiments use a chaotic equation and a transient flow. The model ran much faster than strong diffusion baselines. It also recovered starting states more accurately than comparable forecasters.
Catalog record
| CREATOR | Rares Grozavescu, Pengyu Zhang, Mark Girolami, Etienne Meunier |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 50 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-02832