Optimizing High-Dimensional Physics Simulations via Composite Bayesian Optimization
arXiv:2111.14911
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.
arXiv commentFourth Workshop on Machine Learning and the Physical Sciences at NeurIPS 2021
References in corpus (8)
- Mixed Precision Training
- GPyTorch: Blackbox Matrix-Matrix Gaussian Process Inference with GPU Acceleration
- Bayesian optimization for materials design
- BoTorch: A Framework for Efficient Monte-Carlo Bayesian Optimization
- Bayesian optimization of a free-electron laser
- Scalable Global Optimization via Local Bayesian Optimization
- Multi-Objective Bayesian Optimization over High-Dimensional Search Spaces
- Bayesian Optimization with High-Dimensional Outputs