Certified Neural Approximations of Nonlinear Dynamics

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Mathiesen, Frederik Baymler, Vertovec, Nikolaus, Fabiano, Francesco, Laurenti, Luca, Abate, Alessandro
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908564564475904
author Mathiesen, Frederik Baymler
Vertovec, Nikolaus
Fabiano, Francesco
Laurenti, Luca
Abate, Alessandro
author_facet Mathiesen, Frederik Baymler
Vertovec, Nikolaus
Fabiano, Francesco
Laurenti, Luca
Abate, Alessandro
contents Neural networks hold great potential to act as approximate models of nonlinear dynamical systems, with the resulting neural approximations enabling verification and control of such systems. However, in safety-critical contexts, the use of neural approximations requires formal bounds on their closeness to the underlying system. To address this fundamental challenge, we propose a novel, adaptive, and parallelizable verification method based on certified first-order models. Our approach provides formal error bounds on the neural approximations of dynamical systems, allowing them to be safely employed as surrogates by interpreting the error bound as bounded disturbances acting on the approximated dynamics. We demonstrate the effectiveness and scalability of our method on a range of established benchmarks from the literature, showing that it significantly outperforms the state-of-the-art. Furthermore, we show that our framework can successfully address additional scenarios previously intractable for existing methods - neural network compression and an autoencoder-based deep learning architecture for learning Koopman operators for the purpose of trajectory prediction.
format Preprint
id arxiv_https___arxiv_org_abs_2505_15497
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Certified Neural Approximations of Nonlinear Dynamics
Mathiesen, Frederik Baymler
Vertovec, Nikolaus
Fabiano, Francesco
Laurenti, Luca
Abate, Alessandro
Machine Learning
Systems and Control
Neural networks hold great potential to act as approximate models of nonlinear dynamical systems, with the resulting neural approximations enabling verification and control of such systems. However, in safety-critical contexts, the use of neural approximations requires formal bounds on their closeness to the underlying system. To address this fundamental challenge, we propose a novel, adaptive, and parallelizable verification method based on certified first-order models. Our approach provides formal error bounds on the neural approximations of dynamical systems, allowing them to be safely employed as surrogates by interpreting the error bound as bounded disturbances acting on the approximated dynamics. We demonstrate the effectiveness and scalability of our method on a range of established benchmarks from the literature, showing that it significantly outperforms the state-of-the-art. Furthermore, we show that our framework can successfully address additional scenarios previously intractable for existing methods - neural network compression and an autoencoder-based deep learning architecture for learning Koopman operators for the purpose of trajectory prediction.
title Certified Neural Approximations of Nonlinear Dynamics
topic Machine Learning
Systems and Control
url https://arxiv.org/abs/2505.15497