On vector field reconstruction from noisy ODE in high ambient dimension
Electronic Journal of Statistics · 01.01.2026 · 20 (2) · 5470-5500
DOI: 10.1214/26-ejs2613 ↗Kurz gefasst
Fachartikel in der Zeitschrift Electronic Journal of Statistics, erschienen am 01.01.2026. Verfasst von Hugo Henneuse. Die Arbeit nennt 36 Literaturangaben.
Abstract (Originalsprache)
This work addresses the nonparametric estimation of the vector field of a noisy autonomous Ordinary Differential Equation (ODE) in a high-dimensional ambient space, under the assumption that initial conditions lie on a lower-dimensional structure. We observe trajectories of the system at discrete times, up to additive noise, each starting from a randomly sampled initial condition. From a minimax perspective, we study the reconstruction of the vector field. We propose an estimator that combines flow reconstruction with nonparametric derivative estimation. Under mild regularity assumptions, we establish convergence rates that depend only on the temporal resolution, the number of initial conditions, and the intrinsic dimension of the space on which the initial conditions are sampled. These rates are shown to be minimax optimal up to logarithmic factors. Finally, we demonstrate the estimator’s statistical and computational performance through numerical experiments.
Zitieren
APA
Henneuse, H. (2026). On vector field reconstruction from noisy ODE in high ambient dimension. Electronic Journal of Statistics, 20(2), 5470-5500. https://doi.org/10.1214/26-ejs2613
Vancouver
Henneuse H. On vector field reconstruction from noisy ODE in high ambient dimension. Electronic Journal of Statistics. 2026;20(2):5470-5500. doi:10.1214/26-ejs2613
BibTeX
@article{Henneuse2026,
author = {Henneuse, Hugo},
title = {On vector field reconstruction from noisy ODE in high ambient dimension},
journal = {Electronic Journal of Statistics},
year = {2026},
volume = {20},
number = {2},
pages = {5470--5500},
doi = {10.1214/26-ejs2613}
}
EUVIMED