Optimizing High-Dimensional Physics Simulations via Composite Bayesian Optimization
This work addresses optimization challenges in science and engineering for simulations with high-dimensional parameter spaces and tensor outputs, representing an incremental improvement in Bayesian optimization methods.
The paper tackles the problem of optimizing high-dimensional physics simulations with image- or tensor-based outputs by developing a Bayesian optimization method using tensor-based Gaussian process surrogates and trust region Bayesian optimization, achieving effective modeling and efficient optimization as demonstrated in radio-frequency tower configuration and optical design problems.
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.