Surrogate models for mixed discrete-continuous variables

Laura P. Swiler, Patricia D. Hough, Peter Qian, Xu Xu, Curtis Storlie, Herbert Lee

Research output: Chapter in Book/Report/Conference proceedingConference contribution

13 Citations (Scopus)

Abstract

Large-scale computational models have become common tools for analyzing complex man-made systems. However, when coupled with optimization or uncertainty quantification methods in order to conduct extensive model exploration and analysis, the computational expense quickly becomes intractable. Furthermore, these models may have both continuous and discrete parameters. One common approach to mitigating the computational expense is the use of response surface approximations. While well developed for models with continuous parameters, they are still new and largely untested for models with both continuous and discrete parameters. In this work, we describe and investigate the performance of three types of response surfaces developed for mixed-variable models: Adaptive Component Selection and Shrinkage Operator, Treed Gaussian Process, and Gaussian Process with Special Correlation Functions. We focus our efforts on test problems with a small number of parameters of interest, a characteristic of many physics-based engineering models. We present the results of our studies and offer some insights regarding the performance of each response surface approximation method.

Original languageEnglish (US)
Title of host publicationConstraint Programming and Decision Making
PublisherSpringer Verlag
Pages181-202
Number of pages22
Volume539
ISBN (Print)9783319042794
DOIs
StatePublished - 2014
Externally publishedYes

Publication series

NameStudies in Computational Intelligence
Volume539
ISSN (Print)1860949X

Fingerprint

Physics
Uncertainty

ASJC Scopus subject areas

  • Artificial Intelligence

Cite this

Swiler, L. P., Hough, P. D., Qian, P., Xu, X., Storlie, C., & Lee, H. (2014). Surrogate models for mixed discrete-continuous variables. In Constraint Programming and Decision Making (Vol. 539, pp. 181-202). (Studies in Computational Intelligence; Vol. 539). Springer Verlag. https://doi.org/10.1007/978-3-319-04280-0_21

Surrogate models for mixed discrete-continuous variables. / Swiler, Laura P.; Hough, Patricia D.; Qian, Peter; Xu, Xu; Storlie, Curtis; Lee, Herbert.

Constraint Programming and Decision Making. Vol. 539 Springer Verlag, 2014. p. 181-202 (Studies in Computational Intelligence; Vol. 539).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Swiler, LP, Hough, PD, Qian, P, Xu, X, Storlie, C & Lee, H 2014, Surrogate models for mixed discrete-continuous variables. in Constraint Programming and Decision Making. vol. 539, Studies in Computational Intelligence, vol. 539, Springer Verlag, pp. 181-202. https://doi.org/10.1007/978-3-319-04280-0_21
Swiler LP, Hough PD, Qian P, Xu X, Storlie C, Lee H. Surrogate models for mixed discrete-continuous variables. In Constraint Programming and Decision Making. Vol. 539. Springer Verlag. 2014. p. 181-202. (Studies in Computational Intelligence). https://doi.org/10.1007/978-3-319-04280-0_21
Swiler, Laura P. ; Hough, Patricia D. ; Qian, Peter ; Xu, Xu ; Storlie, Curtis ; Lee, Herbert. / Surrogate models for mixed discrete-continuous variables. Constraint Programming and Decision Making. Vol. 539 Springer Verlag, 2014. pp. 181-202 (Studies in Computational Intelligence).
@inproceedings{3eabebf7ee6e4ad7a7d4c8bdd4102ae4,
title = "Surrogate models for mixed discrete-continuous variables",
abstract = "Large-scale computational models have become common tools for analyzing complex man-made systems. However, when coupled with optimization or uncertainty quantification methods in order to conduct extensive model exploration and analysis, the computational expense quickly becomes intractable. Furthermore, these models may have both continuous and discrete parameters. One common approach to mitigating the computational expense is the use of response surface approximations. While well developed for models with continuous parameters, they are still new and largely untested for models with both continuous and discrete parameters. In this work, we describe and investigate the performance of three types of response surfaces developed for mixed-variable models: Adaptive Component Selection and Shrinkage Operator, Treed Gaussian Process, and Gaussian Process with Special Correlation Functions. We focus our efforts on test problems with a small number of parameters of interest, a characteristic of many physics-based engineering models. We present the results of our studies and offer some insights regarding the performance of each response surface approximation method.",
author = "Swiler, {Laura P.} and Hough, {Patricia D.} and Peter Qian and Xu Xu and Curtis Storlie and Herbert Lee",
year = "2014",
doi = "10.1007/978-3-319-04280-0_21",
language = "English (US)",
isbn = "9783319042794",
volume = "539",
series = "Studies in Computational Intelligence",
publisher = "Springer Verlag",
pages = "181--202",
booktitle = "Constraint Programming and Decision Making",

}

TY - GEN

T1 - Surrogate models for mixed discrete-continuous variables

AU - Swiler, Laura P.

AU - Hough, Patricia D.

AU - Qian, Peter

AU - Xu, Xu

AU - Storlie, Curtis

AU - Lee, Herbert

PY - 2014

Y1 - 2014

N2 - Large-scale computational models have become common tools for analyzing complex man-made systems. However, when coupled with optimization or uncertainty quantification methods in order to conduct extensive model exploration and analysis, the computational expense quickly becomes intractable. Furthermore, these models may have both continuous and discrete parameters. One common approach to mitigating the computational expense is the use of response surface approximations. While well developed for models with continuous parameters, they are still new and largely untested for models with both continuous and discrete parameters. In this work, we describe and investigate the performance of three types of response surfaces developed for mixed-variable models: Adaptive Component Selection and Shrinkage Operator, Treed Gaussian Process, and Gaussian Process with Special Correlation Functions. We focus our efforts on test problems with a small number of parameters of interest, a characteristic of many physics-based engineering models. We present the results of our studies and offer some insights regarding the performance of each response surface approximation method.

AB - Large-scale computational models have become common tools for analyzing complex man-made systems. However, when coupled with optimization or uncertainty quantification methods in order to conduct extensive model exploration and analysis, the computational expense quickly becomes intractable. Furthermore, these models may have both continuous and discrete parameters. One common approach to mitigating the computational expense is the use of response surface approximations. While well developed for models with continuous parameters, they are still new and largely untested for models with both continuous and discrete parameters. In this work, we describe and investigate the performance of three types of response surfaces developed for mixed-variable models: Adaptive Component Selection and Shrinkage Operator, Treed Gaussian Process, and Gaussian Process with Special Correlation Functions. We focus our efforts on test problems with a small number of parameters of interest, a characteristic of many physics-based engineering models. We present the results of our studies and offer some insights regarding the performance of each response surface approximation method.

UR - http://www.scopus.com/inward/record.url?scp=84958532603&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84958532603&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-04280-0_21

DO - 10.1007/978-3-319-04280-0_21

M3 - Conference contribution

AN - SCOPUS:84958532603

SN - 9783319042794

VL - 539

T3 - Studies in Computational Intelligence

SP - 181

EP - 202

BT - Constraint Programming and Decision Making

PB - Springer Verlag

ER -