Nonstationary Gaussian Process Surrogates

TL;DR

This paper surveys nonstationary Gaussian process surrogate models, including various adaptations and implementations, and demonstrates their application through an 8-dimensional satellite drag experiment with publicly available code.

Contribution

It provides a comprehensive overview of nonstationary GP surrogate models, their software implementations, and a practical benchmark example.

Findings

Effective nonstationary GP models for high-dimensional problems

Comparison of different nonstationary kernel methods

Open-source code for satellite drag simulation

Abstract

We provide a survey of nonstationary surrogate models which utilize Gaussian processes (GPs) or variations thereof, including nonstationary kernel adaptations, partition and local GPs, and spatial warpings through deep Gaussian processes. We also overview publicly available software implementations and conclude with a bake-off involving an 8-dimensional satellite drag computer experiment. Code for this example is provided in a public git repository.