A0315
Title: Indicator-based Bayesian variable selection for Gaussian process models in computer experiments
Authors: Fan Zhang - Arizona State University (United States)
Ray-Bing Chen - National Cheng Kung University (Taiwan) [presenting]
Ying Hung - Rutgers University (United States)
Xinwei Deng - Virginia Tech (United States)
Abstract: Gaussian process (GP) models are commonly used to analyse computer experiments. Variable selection in GP models is of significant scientific interest, but existing solutions remain unsatisfactory. For each variable in a GP model, there are two potential effects with different implications: one is on the mean function, and the other is on the covariance function. However, most research on variable selection for GP models has focused only on one of the effects. To tackle this problem, an indicator-based Bayesian variable selection procedure is proposed to consider the effects of both the mean and covariance functions. A variable is defined as inactive if both effects are not significant, and an indicator is used to represent whether the variable is active or not. The proposed method adopts different prior assumptions for active variables to capture the two effects. Both simulations and real applications in computer experiments evaluate the performance of the proposed method.
