Parameter Space Exploration with Gaussian Process Trees

Authors: Robert B. Gramacy, Herbert K. H. Lee, and William G. Macready - UCSC and NASA

Complete Citation

• Robert B. Gramacy, Herbert K. H. Lee, and William G. Macready. Parameter Space Exploration with Gaussian Process Trees. Proc. International Conference on Machine Learning, 2004.

Abstract

Computer experiments often require dense sweeps over input parameters to obtain a qualitative understanding of their response. Such sweeps can be prohibitively expensive, and are unnecessary in regions where the response is easy predicted; well-chosen designs could allow a mapping of the response with far fewer simulation runs. Thus, there is need for computationally inexpensive surrogate models and an accompanying method for selecting small designs. We explore a general methodology for addressing this need that uses non-stationary Gaussian processes. Binary trees partition the input space to facilitate non-stationarity and a Bayesian interpretation provides an explicit measure of predictive uncertainty that can be used to guide sampling. Our methods are illustrated on several examples, including a motivating example involving computational fluid dynamics simulation of a NASA reentry vehicle.

Annotations

• The authors apply a machine learning strategy to improve the efficiency of a parameter sweep.
• By applying an Gaussian Process Tree model, they are able to decrease error in the resulting model with reduced function evaluations.
• The authors apply the method to a CFD example.

-- JustinWozniak - 12 Dec 2007