PAINT: Parallel-in-time Neural Twins for Dynamical System Reconstruction
Andreas Radler, Vincent Seyfried, Stefan Pirker, Johannes Brandstetter, Thomas Lichtenegger · 2025
Read PAINT: Parallel-in-time Neural Twins for Dynamical System Reconstruction by Andreas Radler, Vincent Seyfried, Stefan Pirker, Johannes Brandstetter, Thomas Lichtenegger (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-2510-16004.txt.
Description
Andreas Radler, Vincent Seyfried, Johannes Brandstetter, and Thomas Lichtenegger argue that a neural twin must stay on the true system's path, not merely start well. Neural surrogates simulate dynamical systems fast, and a twin goes further by eating live measurements at test time to update its state for context-specific decisions. Drifting off trajectory ruins that promise over time.
Their family of methods, PAINT, trains a generative network to model the distribution of states in parallel across time, rather than stepping forward one state at a time. At test time, states are predicted from measurements inside a sliding window. Theory shows PAINT stays on-trajectory while autoregressive models generally do not. On a hard two-dimensional turbulent fluid problem, PAINT tracks the true state from sparse measurements with high fidelity, the four authors report.
Catalog record
| CREATOR | Andreas Radler, Vincent Seyfried, Stefan Pirker, Johannes Brandstetter, Thomas Lichtenegger |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 43 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2510-16004