Multilevel Gaussian Processes with Applications to Inverse Problems

DOI10.48550/arXiv.2610.23202MaRDI QIDQ7395308arXivFDO

Authors Sebastian Krumscheid, Stjepan Salatovic, Aretha L. Teckentrup

Publication date 16 October 2026

Published in arXiv

Tldr We develop a multilevel Gaussian process regression framework that leverages hierarchies of approximations with increasing accuracy and cost to efficiently model functions from expensive simulations. Novel is the derivation of convergence rates for posterior mean and variance, which allows for maximum likelihood estimation and kernel misspecification under adaptive sequential-design active learning criteria.

Full work available at URL https://doi.org/10.48550/arXiv.2610.23202

arXiv classification

math.NA

This page was built for publication: Multilevel Gaussian Processes with Applications to Inverse Problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7395308)