Nested maximin Latin hypercube designs

Research output: Contribution to journalArticleScientificpeer-review

268 Downloads (Pure)

Abstract

Abstract In the field of design of computer experiments (DoCE), Latin hypercube designs are frequently used for the approximation and optimization of blackboxes. In certain situations, we need a special type of designs consisting of two separate designs, one being a subset of the other. These nested designs can be used to deal with training and test sets, models with different levels of accuracy, linking parameters, and sequential evaluations. In this paper, we construct nested maximin Latin hypercube designs for up to ten dimensions. We show that different types of grids should be considered when constructing nested designs and discuss how to determine which grid to use for a specific application. To determine nested maximin designs for dimensions higher than two, four variants of the ESE algorithm of Jin et al. (J Stat Plan Inference 134(1):268–287, 2005) are introduced and compared. Our main focus is on GROUPRAND, the most successful of these four variants. In the numerical comparison, we consider the calculation times, space-fillingness of the obtained designs and the performance of different grids. Maximin distances for different numbers of points are provided; the corresponding nested maximin designs can be found on the website http://www.spacefillingdesigns.nl.
Original languageEnglish
Pages (from-to)371-395
JournalStructural and Multidisciplinary Optimization
Volume41
Publication statusPublished - 2010

Fingerprint

Latin Hypercube Design
Maximin
Nested Design
Grid
Computer Experiments
Numerical Comparisons
Test Set
Black Box
Linking
Higher Dimensions
Design
Subset
Optimization
Evaluation
Approximation
Websites

Cite this

@article{7f0703e806bc45b4886c359aa115101c,
title = "Nested maximin Latin hypercube designs",
abstract = "Abstract In the field of design of computer experiments (DoCE), Latin hypercube designs are frequently used for the approximation and optimization of blackboxes. In certain situations, we need a special type of designs consisting of two separate designs, one being a subset of the other. These nested designs can be used to deal with training and test sets, models with different levels of accuracy, linking parameters, and sequential evaluations. In this paper, we construct nested maximin Latin hypercube designs for up to ten dimensions. We show that different types of grids should be considered when constructing nested designs and discuss how to determine which grid to use for a specific application. To determine nested maximin designs for dimensions higher than two, four variants of the ESE algorithm of Jin et al. (J Stat Plan Inference 134(1):268–287, 2005) are introduced and compared. Our main focus is on GROUPRAND, the most successful of these four variants. In the numerical comparison, we consider the calculation times, space-fillingness of the obtained designs and the performance of different grids. Maximin distances for different numbers of points are provided; the corresponding nested maximin designs can be found on the website http://www.spacefillingdesigns.nl.",
author = "G. Rennen and B.G.M. Husslage and {van Dam}, E.R. and {den Hertog}, D.",
note = "Appeared earlier as CentER DP 2009-06",
year = "2010",
language = "English",
volume = "41",
pages = "371--395",
journal = "Structural and Multidisciplinary Optimization",
issn = "1615-147X",
publisher = "Springer Verlag",

}

Nested maximin Latin hypercube designs. / Rennen, G.; Husslage, B.G.M.; van Dam, E.R.; den Hertog, D.

In: Structural and Multidisciplinary Optimization, Vol. 41, 2010, p. 371-395.

Research output: Contribution to journalArticleScientificpeer-review

TY - JOUR

T1 - Nested maximin Latin hypercube designs

AU - Rennen, G.

AU - Husslage, B.G.M.

AU - van Dam, E.R.

AU - den Hertog, D.

N1 - Appeared earlier as CentER DP 2009-06

PY - 2010

Y1 - 2010

N2 - Abstract In the field of design of computer experiments (DoCE), Latin hypercube designs are frequently used for the approximation and optimization of blackboxes. In certain situations, we need a special type of designs consisting of two separate designs, one being a subset of the other. These nested designs can be used to deal with training and test sets, models with different levels of accuracy, linking parameters, and sequential evaluations. In this paper, we construct nested maximin Latin hypercube designs for up to ten dimensions. We show that different types of grids should be considered when constructing nested designs and discuss how to determine which grid to use for a specific application. To determine nested maximin designs for dimensions higher than two, four variants of the ESE algorithm of Jin et al. (J Stat Plan Inference 134(1):268–287, 2005) are introduced and compared. Our main focus is on GROUPRAND, the most successful of these four variants. In the numerical comparison, we consider the calculation times, space-fillingness of the obtained designs and the performance of different grids. Maximin distances for different numbers of points are provided; the corresponding nested maximin designs can be found on the website http://www.spacefillingdesigns.nl.

AB - Abstract In the field of design of computer experiments (DoCE), Latin hypercube designs are frequently used for the approximation and optimization of blackboxes. In certain situations, we need a special type of designs consisting of two separate designs, one being a subset of the other. These nested designs can be used to deal with training and test sets, models with different levels of accuracy, linking parameters, and sequential evaluations. In this paper, we construct nested maximin Latin hypercube designs for up to ten dimensions. We show that different types of grids should be considered when constructing nested designs and discuss how to determine which grid to use for a specific application. To determine nested maximin designs for dimensions higher than two, four variants of the ESE algorithm of Jin et al. (J Stat Plan Inference 134(1):268–287, 2005) are introduced and compared. Our main focus is on GROUPRAND, the most successful of these four variants. In the numerical comparison, we consider the calculation times, space-fillingness of the obtained designs and the performance of different grids. Maximin distances for different numbers of points are provided; the corresponding nested maximin designs can be found on the website http://www.spacefillingdesigns.nl.

M3 - Article

VL - 41

SP - 371

EP - 395

JO - Structural and Multidisciplinary Optimization

JF - Structural and Multidisciplinary Optimization

SN - 1615-147X

ER -