Adversarial dynamical systems characterize when data-driven learning succeeds or fails
read the original abstract
Many systems resist analytical modeling, making data-driven inference of dynamics important. Yet data-driven methods can fail to converge or generalize, leaving open a central question: When can system behavior be learned reliably from data, and when is such learning impossible? We answer this question using adversarial dynamical systems to identify the boundary between accessible and inaccessible regimes. In Koopman operator learning, a leading framework for representing nonlinear dynamics through linear spectral objects, we design optimal data-driven spectral algorithms with convergence and certification guarantees under conditions arising broadly in physical systems. This yields a convergence theory for Koopman-operator approximations and resolves a longstanding open problem in Koopman spectral analysis. Conversely, by constructing adversarial systems, we prove matching impossibility results: without these conditions, no single-sequence limiting procedure can guarantee learning, regardless of data quality. These results sharply characterize when data-driven spectral learning can succeed and when it must fail. We validate the framework on oscillators, chaotic fluid flows and Arctic sea ice concentration forecasting. In the latter, we uncover hidden modes of Arctic sea ice decline, deliver long-range forecasts with geographic error bounds, and outperform state-of-the-art dynamical and deep learning models at substantially lower computational cost, enabling real-time deployment on standard CPUs.
This paper has not been read by Pith yet.
Forward citations
Cited by 5 Pith papers
-
Finite-Query Collapse and Modal Exact Bases in the SCI Hierarchy
Raw finite-query preorders collapse the CH23 spectral block to a single source while modal preorders under geometric admissibility yield exactly two minimal exact sources, reformulating the SCI exact-basis problem.
-
Sampling pseudospectrum for data-driven matrices
Introduces a sampling pseudospectrum P(λ) and estimator ˆP(λ) obtained by reprocessing finite data to statistically test the location of true eigenvalues versus sampling artifacts in data-driven matrices.
-
Nonparametric Sparse Online Learning of the Koopman Operator
Develops a nonparametric sparse online algorithm to learn the Koopman operator iteratively via stochastic approximation with explicit complexity control and convergence guarantees in misspecified RKHS settings via con...
-
Signature of glassy dynamics in dynamic modes decompositions
Dynamic mode decomposition identifies a vanishing gap between oscillatory and decaying Koopman modes as a signature of algebraic relaxation in glassy dynamics.
-
Residual SCI Upper Bounds And Lower Witnesses For Koopman Approximate Point Spectra In $L^p$ For $1<p<\infty$: Extended Version
Proves SCI upper bounds for regularized and closed approximate point ε-pseudospectra plus the approximate point spectrum of Koopman operators in L^p (1<p<∞) for four classes of maps via dictionaries and tagged quadrat...
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.