2009
DOI: 10.1002/qre.1044
|View full text |Cite
|
Sign up to set email alerts
|

Design and analysis for the Gaussian process model

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

0
46
0

Year Published

2009
2009
2016
2016

Publication Types

Select...
7
1

Relationship

1
7

Authors

Journals

citations
Cited by 61 publications
(46 citation statements)
references
References 34 publications
0
46
0
Order By: Relevance
“…Here we use a maximin Latin hypercube (McKay et al, 1979) to fill the uncertainty space of the parameters, which has been shown to be the best design for Gaussian process emulation (Jones and Johnson, 2009). The maximin Latin hypercube is based on the Latin square, a common example of which is the sudoku puzzle, which consists of nine overlying Latin squares; that is there is precisely one of each number in each row and column.…”
Section: Designing To Inform the Emulatormentioning
confidence: 99%
“…Here we use a maximin Latin hypercube (McKay et al, 1979) to fill the uncertainty space of the parameters, which has been shown to be the best design for Gaussian process emulation (Jones and Johnson, 2009). The maximin Latin hypercube is based on the Latin square, a common example of which is the sudoku puzzle, which consists of nine overlying Latin squares; that is there is precisely one of each number in each row and column.…”
Section: Designing To Inform the Emulatormentioning
confidence: 99%
“…The number of runs was selected based on the rule of thumb to use 10 times the number of input parameters. Since the results of computer experiments are deterministic, no replications were included in the design [17]. …”
Section: Space Filling Designmentioning
confidence: 99%
“…A maximum likelihood estimation of , and is performed and predicted values for the responses ( ) are calculated within the design space [17]. A more detailed description of the GP interpolation model can be found in [17].…”
Section: Gaussian Process Modeling and Validationmentioning
confidence: 99%
See 1 more Smart Citation
“…The first consideration is that the response, based on previous information, is expected to be highly nonlinear and may require the use of a nonlinear polynomial model or a special type of spatial correlation model, such as the kriging model, which is a special form of the Gaussian process model (see Jones and Johnson 2009;Santner et al 2003). The use of these more complicated empirical models potentially warrants the use of an experimental design that has more levels than the factorial or CCDs.…”
Section: Advanced Doementioning
confidence: 99%