Cross-Validation Based Single Response Adaptive Design of Experiments for Deterministic Computer Simulations
Vikrant Aute, Omar Abd Elaziz, Shapour Azarm, Reinhard Radermacher · 12th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference · 2008
A new approach for single response adaptive design of deterministic computer experiments is presented. The approach is called SFCVT, for Space-Filling Cross-Validation Tradeoff. SFCVT uses metamodeling to obtain an estimate of cross-validation errors, which is then maximized, combined with a space filling criterion (treated as a constraint) to judiciously allocate sample points in the design space. The proposed method is compared, using a set of twenty four numerical examples, with two DOE methods from the literature and then applied to an air-to-refrigerant heat exchanger problem. Based on the comparison results for the numerical examples, it is shown that SFCVT performs better than an existing adaptive and a non-adaptive DOE method for the majority of the examples.