Bayesian Optimization for Data Selection in Bragg Grating Design Space for Machine Learning Training

Data acquisition for training machine learning (ML) models in resource-intensive scientific fields is challenging [1]. Recognizing the high computational costs associated with generating extensive databases for training, we explore a Bayesian optimization (BO) approach to develop a minimal yet highly informative database. We aim to streamline data collection, reduce computational demands, and improve ML model performance in high-dimensional parameter spaces.