Robust designs for Gaussian process emulation of computer experiments

TL;DR

This paper introduces support points and projected support points as robust, efficient experimental designs for Gaussian process emulation of computer experiments, effective across various response surface complexities.

Contribution

It proposes new design methods that are robust and scalable for Gaussian process emulation, with a theoretical framework and practical algorithms for high-dimensional problems.

Findings

Designs perform well across diverse response surfaces

Efficient generation for large, high-dimensional designs

Validated through numerical experiments

Abstract

We study in this paper two classes of experimental designs, support points and projected support points, which can provide robust and effective emulation of computer experiments with Gaussian processes. These designs have two important properties that are appealing for surrogate modeling of computer experiments. First, the proposed designs are robust: they enjoy good emulation performance over a wide class of smooth and rugged response surfaces. Second, they can be efficiently generated for large designs in high dimensions using difference-of-convex programming. In this work, we present a theoretical framework that investigates the above properties, then demonstrate their effectiveness for Gaussian process emulation in a suite of numerical experiments.