Kriging Hyperparameter Tuning Strategies

Abstract

Response surfaces have been extensively used as a method of building effective surrogate models of high-fidelity<br/>computational simulations. Of the numerous types of response surface models, kriging is perhaps one of the most<br/>effective, due to its ability to model complicated responses through interpolation or regression of known data while<br/>providing an estimate of the error in its prediction. There is, however, little information indicating the extent to which<br/>the hyperparameters of a kriging model need to be tuned for the resulting surrogate model to be effective. The<br/>following paper addresses this issue by investigating how often and how well it is necessary to tune the<br/>hyperparameters of a kriging model as it is updated during an optimization process. To this end, an optimization<br/>benchmarking procedure is introduced and used to assess the performance of five different tuning strategies over a<br/>range of problem sizes. The results of this benchmark demonstrate the performance gains that can be associated with<br/>reducing the complexity of the hyperparameter tuning process for complicated design problems. The strategy of<br/>tuning hyperparameters only once after the initial design of experiments is shown to perform poorly.