Learning Hamiltonian Dynamics with Bayesian Data Assimilation
Taehyeun Kim, Tae-Geun Kim, Anouck Girard, Ilya Kolmanovsky · 2025
Read Learning Hamiltonian Dynamics with Bayesian Data Assimilation by Taehyeun Kim, Tae-Geun Kim, Anouck Girard, Ilya Kolmanovsky (2025) 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-2501-18808.txt.
Description
Predicting how a physical system moves far into the future is hard when its rules are only partly known. Taehyeun Kim and co-authors build a learning method for Hamiltonian systems, the class that covers many mechanical and orbital problems. Their network learns from positions and momenta while keeping the system's total energy constant by design.
A plain version of such a network drifts over long horizons. The authors add an autoregressive twist: past prediction errors feed back into training, which steadies long runs. They also blend in live measurements using a Bayesian updating method, so the forecast corrects itself as new data arrive. Tests on a spring-mass system and on stretched orbits under gravity pulls show accurate, robust long-term predictions. The method suits settings where sensors report while the system runs.
Catalog record
| CREATOR | Taehyeun Kim, Tae-Geun Kim, Anouck Girard, Ilya Kolmanovsky |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 39 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-18808