Meta Model Interface: DACE Kriging

Share:

Description

This Kriging meta model is based on DACE (Design and Analysis of Computer Experiments). It allows to choose different regression and correlation models. If multiple response variables are present, a DACE model for each will be created for the purpose of multi objective optimization.

Usage

1
spotPredictDace(rawB, mergedB, design, spotConfig, fit = NULL)

Arguments

rawB

unmerged data

mergedB

merged data (aggregation of repeated design points)

design

new design points which should be predicted

spotConfig

global list of all options, needed to provide data for calling functions. This also contains a list, with settings for Forrester:
spotConfig$seq.dace.budget Budget for MLE of parameters, default is 100. The value will be multiplied with the length of the model parameter vector to be optimized. (which means 100*i evaluations, where i depends on the problem dimension as well as the correlation function and whether a nugget value is to be determined)
spotConfig$seq.dace.tol Tolerance stopping criterion for MLE, default is 1e-6
spotConfig$seq.dace.regr Regression function to be used: regpoly0 (default), regpoly1, regpoly2
spotConfig$seq.dace.corr Correlation function to be used: corrnoisykriging (default), corrkriging, corrnoisygauss, corrgauss, correxp, correxpg, corrlin, corrcubic,corrspherical,corrspline
spotConfig$seq.dace.nugget Value for nugget. Default is -1, which means the nugget will be optimized during MLE. Else it can be fixed in a range between 0 and 1.

fit

if an existing model fit is supplied, the model will not be build based on data, but only evaluated with the model fit (on the design data). To build the model, this parameter has to be NULL. If it is not NULL the parameters mergedB and rawB will not be used at all in the function.

Value

returns the list spotConfig with two new entries:
spotConfig$seq.modelFit fit of the model used with dacePredictor
spotConfig$seq.largeDesignY the y values of the design, evaluated with the fit

Author(s)

The authors of the original DACE Matlab toolbox http://www2.imm.dtu.dk/projects/dace/ are Hans Bruun Nielsen hbn@imm.dtu.dk, Soren Nymand Lophaven and Jacob Sondergaard.
Extension of the Matlab code by Tobias Wagner wagner@isf.de.
Porting and adaptation to R and further extensions by Martin Zaefferer martin.zaefferer@fh-koeln.de.

References

S.~Lophaven, H.~Nielsen, and J.~Sondergaard. DACE—A Matlab Kriging Toolbox. Technical Report IMM-REP-2002-12, Informatics and Mathematical Modelling, Technical University of Denmark, Copenhagen, Denmark, 2002.

See Also

dacePredictor daceBuilder

Want to suggest features or report bugs for rdrr.io? Use the GitHub issue tracker.