Efficient spatial designs using Hausdorff distances and Bayesian optimization

An iterative Bayesian optimization technique is presented to find spatial designs of data that carry much information. We use the decision theoretic notion value information as design criterion. Gaussian process surrogate models enable fast calculations expected improvement for a large number designs, while full-scale evaluations are only done most promising designs. The Hausdorff distance used model similarity between in covariance representation, and this allows suggested algorithm learn across different study properties synthetic example real-world examples from forest conservation petroleum drilling operations. In we consider where exact solution available run under versions compare it with existing approaches such sequential selection an exchange algorithm.