Optimizing High-Dimensional Physics Simulations via Composite Bayesian Optimization

TL;DR

This paper introduces a Bayesian optimization approach utilizing tensor-based Gaussian processes and trust region methods to efficiently optimize high-dimensional, image-based physical simulations in science and engineering.

Contribution

It presents a novel Bayesian optimization framework that effectively models tensor outputs and handles high-dimensional parameter spaces in physical simulations.

Findings

Successfully optimized radio-frequency tower configurations.

Enhanced optical design through the proposed method.

Demonstrated efficiency in high-dimensional, image-based simulation optimization.

Abstract

Physical simulation-based optimization is a common task in science and engineering. Many such simulations produce image- or tensor-based outputs where the desired objective is a function of those outputs, and optimization is performed over a high-dimensional parameter space. We develop a Bayesian optimization method leveraging tensor-based Gaussian process surrogates and trust region Bayesian optimization to effectively model the image outputs and to efficiently optimize these types of simulations, including a radio-frequency tower configuration problem and an optical design problem.