space-filling designs - gatech.edu
TRANSCRIPT
![Page 1: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/1.jpg)
Space-Filling Designs
C. F. Jeff Wu
Georgia Institute of Technology
Most commonly used designs for computer experiments
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 1 / 35
![Page 2: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/2.jpg)
Overview
1 Latin Hypercube DesignLatin Hypercube Design (LHD)Orthogonal Array-based Latin Hypercube Design (OA-LHD)
2 Designs Based on Measures of DistanceminiMax Distance Design (mM Design)Maximin Distance Design (Mm Design)
3 Maximin Latin Hypercube Design (MmLHD)
4 Maximum Projection Design (MaxPro)
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 2 / 35
![Page 3: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/3.jpg)
Latin Hypercube Design (LHD) - Introduction
Mckay, Beckman and Conover (1979).
An example: p = 2, n = 5:
••
••
D =
1 22 43 34 1
Each row and column has one and only one point.
Each factor has n levels.
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 3 / 35
![Page 4: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/4.jpg)
Factorial Design 22
On the other hand, a factorial 22 design is constructed as
• •
• •
D =
1 11 44 14 4
If x1 (or x2) is not significant, replication is wasted for computerexperiments.
“Effect sparsity” principle: only a few factors are expected to beimportant.
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 4 / 35
![Page 5: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/5.jpg)
Latin Hypercube Design (LHD) - Definition
Definition
A Latin hypercube design (LHD) with n runs and p inputs variables,denoted by LHD(n, p), is an n × p matrix, in which each column is arandom permutation of {1, 2, . . . , n}.
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 5 / 35
![Page 6: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/6.jpg)
Construct an LHD(n, p)
How to construct an LHD(n, p)?
Step1 Randomly permute {1, 2, . . . , n} for each x1, . . . , xp.
D =
x1 x2 x3
1 2 72 n 93 5 1...
......
n 3 2
Step2 D ′ ← D−0.5
n , where D ∈ {1, 2, . . . , n}p. Thus, D ′ ∈ [0, 1]p.
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 6 / 35
![Page 7: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/7.jpg)
Optimal LHD(n, p)
Thus, there are (n!)p−1 LHDs.
Not all of them are good. For example,
••
••
D =
1 12 23 34 4
This design is perfectly correlated and not space-filling.
Finding an “optimal” LHD is a challenge.
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 7 / 35
![Page 8: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/8.jpg)
Orthogonal Array-based LHD (OA-LHD) - Introduction
Owen (1992) and Tang (1993).
LHD: one-dimensional balancing property but two and higherdimensional projections can be very bad.
Orthogonal Array (OA) of strength 2 has two dimensional balancingproperty, so use it to generate an LHD.
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 8 / 35
![Page 9: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/9.jpg)
Orthogonal Array-based LHD (OA-LHD) - Example
For example, n = 9, p = 2,
OA
x1 x2
1 11 21 32 12 22 33 13 23 3
1→ {1, 2, 3} → {3, 2, 1}2→ {4, 5, 6} → {4, 5, 6}3→ {7, 8, 9} → {8, 7, 9}
OA-LHD
x1 x2
1 32 43 84 25 56 77 18 69 9
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 9 / 35
![Page 10: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/10.jpg)
Orthogonal Array-based LHD (OA-LHD) - Example
••
••
••
••
•
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 10 / 35
![Page 11: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/11.jpg)
Latin Hypercube Design (LHD) - Properties
McKay, Beckman, and Conover (1979) introduced Latin hypercubesampling and subsequently other authors (for example, Stein (1987)and Owen (1992)) have explored their properties. Roughly speaking,suppose we want to find the mean of some known function G (y(x))over X . Then the sample mean of G (y(x)) computed from a Latinhypercube design usually has smaller variance than the sample meancomputed from a simple random sample.
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 11 / 35
![Page 12: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/12.jpg)
Designs Based on Measures of Distance
Johnson, Moore, and Ylvisaker (1991).
Let ρ(·, ·) be a metric.
ρ(x1, x2) = ρ(x2, x1)
ρ(x1, x2) ≥ 0
ρ(x1, x2) = 0⇔ x1 = x2
ρ(x1, x2) ≤ ρ(x1, x3) + ρ(x3, x2)
For example,
ρ(u, v) = {p∑
j=1
|uj − vj |k}1/k
k = 1 : rectangular distance
k = 2 : Eulidean distance
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 12 / 35
![Page 13: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/13.jpg)
miniMax Distance Design (mM Design) - Introduction
Let ρ(x ,D) = minxi∈D ρ(x , xi ) be the minimum distance to thedesign.
Let χ = [0, 1]p andh = max
x∈χρ(x ,D)
be the maximum distance in χ.
h is called fill distance.
the largest gapthe radius of the largest ball that can be placed in χ which does notcontain any point in D.
Thus, find a D to minimize h. That is,
minD
maxx∈χ
ρ(x ,D)
→ miniMax distance design (mM).
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 13 / 35
![Page 14: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/14.jpg)
miniMax Distance Design (mM Design) - Examples
p = 1, n = 1:
0.0 0.2 0.4 0.6 0.8 1.0
0.6
0.8
1.0
1.2
1.4
c(0, 1)
c(1,
1)
0 11/2
p = 1, n = 2:
0.0 0.2 0.4 0.6 0.8 1.0
0.6
0.8
1.0
1.2
1.4
c(0, 1)
c(1,
1)
0 11/4 3/4
p = 2, n = 4:
1/4
3/4
1/4 3/4
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 14 / 35
![Page 15: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/15.jpg)
miniMax Distance Design (mM Design) - Properties
mM designs ensure that all points in χ are not too far from thedesign.
Consider the owner of a petroleum corporation who wants to opensome gas stations. mM design ensures that no customer is too farfrom one of the company’s gas stations.
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 15 / 35
![Page 16: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/16.jpg)
Maximin Distance Design (Mm Design) - Introduction
The minimum distance between any two points in D is
2q = minx1,x2∈D
ρ(x1, x2),
where q is the separation distance or packing radius - the radius ofthe largest ball that can be placed around every design point suchthat no two balls overlap.
A large q ensures numberical stability in Kriging.
A large q tends to decrease h.
Thus, find a D to maximize 2q. That is,
maxD
minx1,x2∈D
ρ(x1, x2)
→ Maximin distance design (Mm).
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 16 / 35
![Page 17: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/17.jpg)
Maximin Distance Design (Mm Design) - Example
p = 1, n = 2:
0.0 0.2 0.4 0.6 0.8 1.0
0.6
0.8
1.0
1.2
1.4
c(0, 1)
c(1,
1)
0 1
p = 2, n = 4:
⇒
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 17 / 35
![Page 18: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/18.jpg)
Maximin Distance Design (Mm Design) - Properties
Mm designs ensure that the points in D are as far apart from eachother as possible.
Gas station example: Mm design ensures that no two gas stations aretoo close to each other. It minimizes the competition from each otherby locating the stations as far apart as possible.
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 18 / 35
![Page 19: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/19.jpg)
Summary
In mathematics,
covering problems: cover Rp with spheres. → miniMaxpacking problems: pack spheres in Rp. → Maximin
Experimental design problem is different because we need to cover[0, 1]p or pack in [0, 1]p. → This introduces boundary effects.
1/4
3/4
1/4 3/4
(a) Minimum radius. (b) Maximum radius.
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 19 / 35
![Page 20: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/20.jpg)
Summary
Computationally,
miniMax:minD
maxx∈χ
minxi∈D
ρ(x , xi )→ very hard
Maximin:
maxD
minx1,x2∈D
ρ(x1, x2)→ hard, but easier than miniMax
Mak and Joseph (2017): a new hybrid algorithm combining particleswarm optimization and clustering for generating minimax designs onany convex and bounded design space.
R package: minimaxdesign
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 20 / 35
![Page 21: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/21.jpg)
Maximin Latin Hypercube Design (MmLHD) - Introduction
Morris and Mitchell (1995).
Advantage of mM/Mm designs: run size flexibility and “optimal”.
Disadvantage of mM/Mm designs: projections are poor.
Optimal Design Criterion for Computer Experiments
6
Maximin Distance Design Minimax Distance Design
Poor Projections
Use Latin Hypercube Designs (McKay, Conover, Beckman 1979)
(a) Maximin design
Optimal Design Criterion for Computer Experiments
6
Maximin Distance Design Minimax Distance Design
Poor Projections
Use Latin Hypercube Designs (McKay, Conover, Beckman 1979)
(b) Minimax design.
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 21 / 35
![Page 22: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/22.jpg)
Maximin Latin Hypercube Design (MmLHD) - Introduction
LHD: good 1-dimensional projections, but can be poor in terms ofspace-filling in higher dimensions.
Optimal Design Criterion for Computer Experiments
6
Maximin Distance Design Minimax Distance Design
Poor Projections
Use Latin Hypercube Designs (McKay, Conover, Beckman 1979)
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 22 / 35
![Page 23: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/23.jpg)
Maximin Latin Hypercube Design (MmLHD) - Introduction
Combine these two ideas → Maximin Latin hypercube designs(MmLHD)
maxD
minx1,x2∈D
ρ(x1, x2),
where D is an LHD.
It is the same as
minD
n−1∑i=1
n∑j=i+1
1
ρk(xi , xj)
1/k
and k →∞, where D is an LHD.
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 23 / 35
![Page 24: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/24.jpg)
Maximin Latin Hypercube Design (MmLHD) - Examples
Optimal Design Criterion for Computer Experiments
8
• Maximin Latin hypercube design (MmLHD) (Morris and Mitchell 1995)
where is an LHD.
MmLHD(20,2) MmLHD(20,10)
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 24 / 35
![Page 25: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/25.jpg)
Maximum Projection Design (MaxPro) - Introduction
Joseph, Gul, and Ba (2015).
MmLHDs only ensure good space-fillingness in p dimensions anduniform projections in a single dimension, but projection properties in2, 3, . . . , p − 1 dimensions may not be good.
In practice, we often have 1<the number of important factors< p.
Optimal Design Criterion for Computer Experiments
8
• Maximin Latin hypercube design (MmLHD) (Morris and Mitchell 1995)
where is an LHD.
MmLHD(20,2) MmLHD(20,10)
Figure : Projection in two dimensions for a 10 dimensional MmLHD.
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 25 / 35
![Page 26: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/26.jpg)
Maximum Projection Design (MaxPro)
How to calculate distances in a projected subspace? Answer: Putweights of 1 on the defining factors and 0 on the remaining factors.
Weighted Euclidean Distance:
d(xi , xj ;θ) = {p∑
l=1
θl(xil − xjl)2}1/2.
Let 0 ≤ θl ≤ 1 be the weight assigned to the factor l and let∑pl=1 θl = 1.
Then the MmLHD criterion can be modified to
minDφk(D;θ) =
n−1∑i=1
n∑j=i+1
1
dk(xi , xj ;θ),
where θ = (θ1, . . . , θp−1)′ and θp = 1−∑p−1
l=1 θl .
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 26 / 35
![Page 27: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/27.jpg)
Maximum Projection Design (MaxPro)
In practice, we often have no idea about the importance of the factorsbefore the experiment.
Assigning non-informative prior:
p(θ) =1
(p − 1)!,θ ∈ Sp−1,
where Sp−1 = {θ : θ1, . . . , θp−1 ≥ 0,∑p−1
i=1 θi ≤ 1}.The design criterion becomes
minD
E{φk(D;θ)} =
∫Sp−1
n−1∑i=1
n∑j=i+1
1
dk(xi , xj ;θ)p(θ)dθ.
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 27 / 35
![Page 28: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/28.jpg)
Maximum Projection Design (MaxPro) - New Criterion
Theorem
If k = 2p, then under the given prior
E{φk(D;θ)} =1
{(p − 1)!}2
n−1∑i=1
n∑j=i+1
1∏pl=1(xil − xjl)2
.
Therefore, we propose a new criterion
minDψ(D) =
1(n2
) n−1∑i=1
n∑j=i+1
1∏pl=1(xil − xjl)2
1/p
.
It is called MaxPro Criterion.
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 28 / 35
![Page 29: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/29.jpg)
Maximum Projection Design (MaxPro)
MaxPro Criterion:
minDψ(D) =
1(n2
) n−1∑i=1
n∑j=i+1
1∏pl=1(xil − xjl)2
1/p
.
Maximum projection (MaxPro) design: maximizes space-fillingproperties on projections to all possible subsets of factors.
MaxPro criterion can be computed at a cost no more than a designcriterion that ignores projection properties.
For any l , if xil = xjl , then ψ(D) =∞;→ n distinct levels for eachfactor → MaxPro designs automatically have the LHD property!
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 29 / 35
![Page 30: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/30.jpg)
Maximum Projection Design (MaxPro) - Algorithm
Design construction algorithm:
Step1 Find an optimal MaxProLHD with simulated annealing (Morris andMitchell 1995).
Step2 Apply a fast derivative-based continuous optimization algorithm to findlocally optimal MaxPro design in the neighborhood of the optimalMaxProLHD:
∂ψp(D)
∂xrs=
2(n2
) ∑i 6=r
1∏pl=1(xil − xjl)2
1
(xis − xjs).
R package: MaxPro.
Available in JMP 12.
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 30 / 35
![Page 31: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/31.jpg)
Maximum Projection Design (MaxPro) - Examples
MaxProLHD(20,2) MaxPro(20,2)
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
Factor 1
Fact
or 2
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 31 / 35
![Page 32: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/32.jpg)
Maximum Projection Design (MaxPro) - Examples
MaxProLHD(20,2) MaxPro(20,2)
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
Factor 1
Fact
or 2
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
Factor 1
Fact
or 2
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 32 / 35
![Page 33: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/33.jpg)
Maximum Projection Design (MaxPro) - Examples
MaxProLHD(20,2) MaxPro(20,2)
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
Factor 1
Fact
or 2
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
Factor 1
Fact
or 2
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
Factor 1
Fact
or 2
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 33 / 35
![Page 34: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/34.jpg)
Maximum Projection Design (MaxPro) - Numerical Results
Maximin Criterion:
Mmq = minr=1,...,(p
q)
1(n2
) n−1∑i=1
n∑j=i+1
1
d2qqr (xi , xj)
−1/(2q)
Numerical Results – Maximin Criterion
21
• Incorporate the index of the design:
𝑴𝑴𝑴𝑴𝒒𝒒 (larger-the-better)
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 34 / 35
![Page 35: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/35.jpg)
Maximum Projection Design (MaxPro) - Numerical Results
miniMax Criterion:
mMq = maxr=1,...,(p
q)maxu∈χq
(1
n
n∑i=1
1
d2qqr (u, xi )
)−1/(2q)
,
where χq is the set of sample points.
Numerical Results – miniMax Criterion
22
where is the set of sample points.
𝑴𝑴𝑴𝑴𝒒𝒒 (smaller-the-better)
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 35 / 35
![Page 36: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/36.jpg)
Maximum Projection Design (MaxPro) - Properties
Easy to compute MaxPro criterion.
Good space-filling properties in projections to all subsets of factors.
Available in R Package MaxPro and JMP 12.
Advantageous in Gaussian process modeling.
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 36 / 35
![Page 37: Space-Filling Designs - gatech.edu](https://reader031.vdocument.in/reader031/viewer/2022012518/619223c3fd7f253a2e74dbb6/html5/thumbnails/37.jpg)
Conclusion
Latin hypercube design:
Good 1-dimensional projections, but might not be space-filling in pdimensions.
miniMax/Maximin distance design:
Run size flexibility and good space-fillingness in p dimensions, but badprojections.
Maximin Latin Hypercube design:
Good space-fillingness in p dimensions and uniform projections in asingle dimension.
Maximum projection design:
Good space-filling properties in projections to all subsets of factors.
C. F. Jeff Wu (Georgia Tech) Space-Filling Designs 37 / 35