Data-driven Kriging models based on FANOVA-decomposition
- 371 Downloads
Kriging models have been widely used in computer experiments for the analysis of time-consuming computer codes. Based on kernels, they are flexible and can be tuned to many situations. In this paper, we construct kernels that reproduce the computer code complexity by mimicking its interaction structure. While the standard tensor-product kernel implicitly assumes that all interactions are active, the new kernels are suited for a general interaction structure, and will take advantage of the absence of interaction between some inputs. The methodology is twofold. First, the interaction structure is estimated from the data, using a first initial standard Kriging model, and represented by a so-called FANOVA graph. New FANOVA-based sensitivity indices are introduced to detect active interactions. Then this graph is used to derive the form of the kernel, and the corresponding Kriging model is estimated by maximum likelihood. The performance of the overall procedure is illustrated by several 3-dimensional and 6-dimensional simulated and real examples. A substantial improvement is observed when the computer code has a relatively high level of complexity.
KeywordsSensitivity analysis Computer experiments Functional ANOVA decomposition Graph Kriging
Unable to display preview. Download preview PDF.
- Bach, F.: High-dimensional non-linear variable selection through hierarchical kernel learning. Technical report (2009). http://arxiv.org/abs/0909.0844
- Cressie, N.: Statistics for Spatial Data. Wiley Series in Probability and Mathematical Statistics. Wiley, New York (1993) Google Scholar
- Diestel, R.: Graph Theory. Springen, New York (2000) Google Scholar
- Durrande, N., Ginsbourger, D., Roustant, O.: Additive kernels for high-dimensional Gaussian process modeling. Technical report (2010). http://hal.archives-ouvertes.fr/hal-00446520/en/
- Goesling, M., Kracker, H., Brosius, A., Kuhnt, S., Tekkaya, A.: Simulation und kompensation rueckfederungsbedingter formabweichungen. In: Tillmann, W. (ed.) SFB 708–3 oeffentliches Kolloquium, pp. 155–170. Praxiswissen, Dortmund (2009) Google Scholar
- Krige, D.: A statistical approach to some basic mine valuation problems on the witwatersrand. J. Chem. Metall. Min. Soc. S. Afr. 52(6), 119–139 (1951) Google Scholar
- Matheron, G.: Les cahiers du centre de morphologie mathématique de fontainebleau. Tech. Rep. 1, Ecole Nationale Supérieure des Mines de Paris (1969) Google Scholar
- Rasmussen, C., Williams, C.: Gaussian processes for machine learning. Adaptive Computation and Machine Learning. MIT, Cambridge (2006) Google Scholar
- Roustant, O., Ginsbourger, D., Deville, Y.: DiceKriging, DiceOptim: Two R packages for the analysis of computer experiments by kriging-based metamodeling and optimization. Technical report (2010). http://hal.archives-ouvertes.fr/hal-00495766/fr/