Diagnostic Certificates of Data Quality and Regression Identifiability for Koopman Identification
Yue Wu · 2026
Read Diagnostic Certificates of Data Quality and Regression Identifiability for Koopman Identification by Yue Wu (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-2605-09545.txt.
Description
Wu builds diagnostic certificates for data quality in Koopman identification, a method that lifts nonlinear dynamics into linear form for prediction and control. Classical excitation tests ask whether an input signal is rich enough. But in extended dynamic mode decomposition with control, data quality depends on lifted state features joined to control inputs. Rich inputs can still visit a narrow state region, and wide state samples can still yield degenerate lifted features. Either failure can wreck the final regression.
His certificates separate three layers: state spread and grouping, lifted feature soundness, and the regression spectrum. Only the last layer carries direct theory, controlling the smallest singular value with finite sample bounds. Examples and a Schur complement condition show the layers cannot stand in for each other. Tests on Duffing, Van der Pol, and Lorenz systems confirm the layers separate in practice, guiding data collection rather than promising universal optimality.
Catalog record
| CREATOR | Yue Wu |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 49 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.09545