Trustworthy AI in numerics: On verification algorithms for neural network-based PDE solvers

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Haugen, Emil, Stepanenko, Alexei, Hansen, Anders C.
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866908568770314240
author Haugen, Emil
Stepanenko, Alexei
Hansen, Anders C.
author_facet Haugen, Emil
Stepanenko, Alexei
Hansen, Anders C.
contents We present new algorithms for a posteriori verification of neural networks (NNs) approximating solutions to PDEs. These verification algorithms compute accurate estimates of $L^p$ norms of NNs and their derivatives. When combined with residual bounds for specific PDEs, the algorithms provide guarantees of $\eps$-accuracy (in a suitable norm) with respect to the true, but unknown, solution of the PDE -- for arbitrary $\eps >0$. In particular, if the NN fails to meet the desired accuracy, our algorithms will detect that and reject it, whereas any NN that passes the verification algorithms is certified to be $\eps$-accurate. This framework enables trustworthy algorithms for NN-based PDE solvers, regardless of how the NN is initially computed. Such a posteriori verification is essential, since a priori error bounds in general cannot guarantee the accuracy of computed solutions, due to algorithmic undecidability of the optimization problems used to train NNs.
format Preprint
id arxiv_https___arxiv_org_abs_2509_26122
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Trustworthy AI in numerics: On verification algorithms for neural network-based PDE solvers
Haugen, Emil
Stepanenko, Alexei
Hansen, Anders C.
Numerical Analysis
65M15, 68T07 (Primary), 26D10, 41A55, 65D32 (Secondary)
We present new algorithms for a posteriori verification of neural networks (NNs) approximating solutions to PDEs. These verification algorithms compute accurate estimates of $L^p$ norms of NNs and their derivatives. When combined with residual bounds for specific PDEs, the algorithms provide guarantees of $\eps$-accuracy (in a suitable norm) with respect to the true, but unknown, solution of the PDE -- for arbitrary $\eps >0$. In particular, if the NN fails to meet the desired accuracy, our algorithms will detect that and reject it, whereas any NN that passes the verification algorithms is certified to be $\eps$-accurate. This framework enables trustworthy algorithms for NN-based PDE solvers, regardless of how the NN is initially computed. Such a posteriori verification is essential, since a priori error bounds in general cannot guarantee the accuracy of computed solutions, due to algorithmic undecidability of the optimization problems used to train NNs.
title Trustworthy AI in numerics: On verification algorithms for neural network-based PDE solvers
topic Numerical Analysis
65M15, 68T07 (Primary), 26D10, 41A55, 65D32 (Secondary)
url https://arxiv.org/abs/2509.26122