Generalized integrated Brownian fields for simulation metamodeling

Peter Salemi, Jeremy Staum, Barry L Nelson

Research output: Contribution to journalArticle

Abstract

We introduce a novel class of Gaussian random fields (GRFs), called generalized integrated Brownian fields (GIBFs), focusing on the use of GIBFs for Gaussian process regression in deterministic and stochastic simulation metamodeling. We build GIBFs from the well-known Brownian motion and discuss several of their properties, including differentiability that can differ in each coordinate, no mean reversion, and the Markov property. We explain why we desire to use GRFs with these properties and provide formal definitions of mean reversion and the Markov property for real-valued, differentiable random fields. We show how to use GIBFs with stochastic kriging, covering trend modeling and parameter fitting, discuss their approximation capability, and show that the resulting metamodel also has differentiability that can differ in each coordinate. Last, we use several examples to demonstrate superior prediction capability as compared with the GRFs corresponding to the Gaussian and Matérn covariance functions.

Original languageEnglish (US)
Pages (from-to)874-891
Number of pages18
JournalOperations Research
Volume67
Issue number3
DOIs
StatePublished - Jan 1 2019

Fingerprint

Brownian movement
Random field
Integrated
Simulation
Metamodeling
Differentiability
Mean reversion

Keywords

  • Gaussian process regression
  • Gaussian random fields
  • Kriging
  • Markov property
  • Mean reversion
  • Simulation metamodeling
  • Stochastic kriging

ASJC Scopus subject areas

  • Computer Science Applications
  • Management Science and Operations Research

Cite this

Salemi, Peter ; Staum, Jeremy ; Nelson, Barry L. / Generalized integrated Brownian fields for simulation metamodeling. In: Operations Research. 2019 ; Vol. 67, No. 3. pp. 874-891.
@article{d6ee0c096ed644eebc61fee93bb7a0d1,
title = "Generalized integrated Brownian fields for simulation metamodeling",
abstract = "We introduce a novel class of Gaussian random fields (GRFs), called generalized integrated Brownian fields (GIBFs), focusing on the use of GIBFs for Gaussian process regression in deterministic and stochastic simulation metamodeling. We build GIBFs from the well-known Brownian motion and discuss several of their properties, including differentiability that can differ in each coordinate, no mean reversion, and the Markov property. We explain why we desire to use GRFs with these properties and provide formal definitions of mean reversion and the Markov property for real-valued, differentiable random fields. We show how to use GIBFs with stochastic kriging, covering trend modeling and parameter fitting, discuss their approximation capability, and show that the resulting metamodel also has differentiability that can differ in each coordinate. Last, we use several examples to demonstrate superior prediction capability as compared with the GRFs corresponding to the Gaussian and Mat{\'e}rn covariance functions.",
keywords = "Gaussian process regression, Gaussian random fields, Kriging, Markov property, Mean reversion, Simulation metamodeling, Stochastic kriging",
author = "Peter Salemi and Jeremy Staum and Nelson, {Barry L}",
year = "2019",
month = "1",
day = "1",
doi = "10.1287/opre.2018.1804",
language = "English (US)",
volume = "67",
pages = "874--891",
journal = "Operations Research",
issn = "0030-364X",
publisher = "INFORMS Inst.for Operations Res.and the Management Sciences",
number = "3",

}

Generalized integrated Brownian fields for simulation metamodeling. / Salemi, Peter; Staum, Jeremy; Nelson, Barry L.

In: Operations Research, Vol. 67, No. 3, 01.01.2019, p. 874-891.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Generalized integrated Brownian fields for simulation metamodeling

AU - Salemi, Peter

AU - Staum, Jeremy

AU - Nelson, Barry L

PY - 2019/1/1

Y1 - 2019/1/1

N2 - We introduce a novel class of Gaussian random fields (GRFs), called generalized integrated Brownian fields (GIBFs), focusing on the use of GIBFs for Gaussian process regression in deterministic and stochastic simulation metamodeling. We build GIBFs from the well-known Brownian motion and discuss several of their properties, including differentiability that can differ in each coordinate, no mean reversion, and the Markov property. We explain why we desire to use GRFs with these properties and provide formal definitions of mean reversion and the Markov property for real-valued, differentiable random fields. We show how to use GIBFs with stochastic kriging, covering trend modeling and parameter fitting, discuss their approximation capability, and show that the resulting metamodel also has differentiability that can differ in each coordinate. Last, we use several examples to demonstrate superior prediction capability as compared with the GRFs corresponding to the Gaussian and Matérn covariance functions.

AB - We introduce a novel class of Gaussian random fields (GRFs), called generalized integrated Brownian fields (GIBFs), focusing on the use of GIBFs for Gaussian process regression in deterministic and stochastic simulation metamodeling. We build GIBFs from the well-known Brownian motion and discuss several of their properties, including differentiability that can differ in each coordinate, no mean reversion, and the Markov property. We explain why we desire to use GRFs with these properties and provide formal definitions of mean reversion and the Markov property for real-valued, differentiable random fields. We show how to use GIBFs with stochastic kriging, covering trend modeling and parameter fitting, discuss their approximation capability, and show that the resulting metamodel also has differentiability that can differ in each coordinate. Last, we use several examples to demonstrate superior prediction capability as compared with the GRFs corresponding to the Gaussian and Matérn covariance functions.

KW - Gaussian process regression

KW - Gaussian random fields

KW - Kriging

KW - Markov property

KW - Mean reversion

KW - Simulation metamodeling

KW - Stochastic kriging

UR - http://www.scopus.com/inward/record.url?scp=85068444128&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85068444128&partnerID=8YFLogxK

U2 - 10.1287/opre.2018.1804

DO - 10.1287/opre.2018.1804

M3 - Article

AN - SCOPUS:85068444128

VL - 67

SP - 874

EP - 891

JO - Operations Research

JF - Operations Research

SN - 0030-364X

IS - 3

ER -