Universal approximation property of neural stochastic differential equations

We identify various classes of neural networks that are able to approximate continuous functions locally uniformly subject to fixed global linear growth constraints. For such neural networks the associated neural stochastic differential equations can approximate general stochastic differential equat...

Full description

Saved in:
Bibliographic Details
Main Authors Kwossek, Anna P, Prömel, David J, Teichmann, Josef
Format Journal Article
LanguageEnglish
Published 20.03.2025
Subjects
Computer Science - Learning
Mathematics - Functional Analysis
Mathematics - Probability
Quantitative Finance - Mathematical Finance
Statistics - Machine Learning
Online AccessGet full text
DOI10.48550/arxiv.2503.16696

Cover

More Information
Summary:We identify various classes of neural networks that are able to approximate continuous functions locally uniformly subject to fixed global linear growth constraints. For such neural networks the associated neural stochastic differential equations can approximate general stochastic differential equations, both of Itô diffusion type, arbitrarily well. Moreover, quantitative error estimates are derived for stochastic differential equations with sufficiently regular coefficients.
DOI:10.48550/arxiv.2503.16696