DNS: Data-driven Nonlinear Smoother for Complex Model-free Process
Fredrik Cumlin, Anubhab Ghosh, Saikat Chatterjee · 2026
Read DNS: Data-driven Nonlinear Smoother for Complex Model-free Process by Fredrik Cumlin, Anubhab Ghosh, Saikat Chatterjee (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-2602-08560.txt.
Description
Estimating a hidden state from noisy measurements is routine when the dynamics are known, but many real processes offer no model at all. Fredrik Cumlin and two co-authors propose DNS, a data-driven nonlinear smoother that works without any state-transition model. Their arXiv paper targets model-free dynamical processes observed through linear measurements.
A recurrent architecture yields a closed-form posterior for the hidden sequence given the measurements, and training needs only measurement data, never state data. Tests on several stochastic processes, including the Lorenz benchmark, show DNS beating a deep Kalman smoother and an iterative data-driven smoother by clear margins. Readers get an unsupervised smoother for processes too complex to model.
Catalog record
| CREATOR | Fredrik Cumlin, Anubhab Ghosh, Saikat Chatterjee |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 29 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-08560