seqDesign.class: A Reference Class to generate a better Model2 or Model4...
In mathieucarmassi/CaliCo: Code Calibration in a Bayesian Framework

Description Usage Format Fields References

This class generates a new model.class for Model4 and Model2. Based on the previous estimation of the Gaussian process in the function model, the design of experiments previously used is improved according to [Damblin et al. 2018]. The aim is to reduce the error produced by the initial estimation of the Gaussian process by fortifying the initial DOE. The method consists in proposing new points based on the expectancy improvement criterion.

Fields should not be changed or manipulated by the user as they are updated internally during the estimation process.

1	seqDesign.class

An object of class R6ClassGenerator of length 24.

doe.init: the initial DOE used to fit the first Gaussian process
GP.init: the initial Gaussian process generated in model function
GP.new: the new Gaussian process fortified with the new design points
p: the number of parameter
md: the initial model
md.new: the new model
mdfit: the initial calibrated model
mdfit.new: the new calibrated model
X: the data set
m: minimum of the sum of squares used in the algorithm

DAMBLIN, Guillaume, BARBILLON, Pierre, KELLER, Merlin, et al. Adaptive numerical designs for the calibration of computer codes. SIAM/ASA Journal on Uncertainty Quantification, 2018, vol. 6, no 1, p. 151-179.

mathieucarmassi/CaliCo documentation built on Aug. 14, 2019, 11:32 a.m.