Saved in:
Bibliographic Details
Main Authors: Semler, Phillip, Weiser, Martin
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2404.01864
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910396088057856
author Semler, Phillip
Weiser, Martin
author_facet Semler, Phillip
Weiser, Martin
contents Generating simulated training data needed for constructing sufficiently accurate surrogate models to be used for efficient optimization or parameter identification can incur a huge computational effort in the offline phase. We consider a fully adaptive greedy approach to the computational design of experiments problem using gradient-enhanced Gaussian process regression as surrogates. Designs are incrementally defined by solving an optimization problem for accuracy given a certain computational budget. We address not only the choice of evaluation points but also of required simulation accuracy, both of values and gradients of the forward model. Numerical results show a significant reduction of the computational effort compared to just position-adaptive and static designs as well as a clear benefit of including gradient information into the surrogate training.
format Preprint
id arxiv_https___arxiv_org_abs_2404_01864
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Adaptive Gradient Enhanced Gaussian Process Surrogates for Inverse Problems
Semler, Phillip
Weiser, Martin
Numerical Analysis
65N21, 65K10, 65N30, 90C31
Generating simulated training data needed for constructing sufficiently accurate surrogate models to be used for efficient optimization or parameter identification can incur a huge computational effort in the offline phase. We consider a fully adaptive greedy approach to the computational design of experiments problem using gradient-enhanced Gaussian process regression as surrogates. Designs are incrementally defined by solving an optimization problem for accuracy given a certain computational budget. We address not only the choice of evaluation points but also of required simulation accuracy, both of values and gradients of the forward model. Numerical results show a significant reduction of the computational effort compared to just position-adaptive and static designs as well as a clear benefit of including gradient information into the surrogate training.
title Adaptive Gradient Enhanced Gaussian Process Surrogates for Inverse Problems
topic Numerical Analysis
65N21, 65K10, 65N30, 90C31
url https://arxiv.org/abs/2404.01864