Improved adaptive sampling method utilizing Gaussian process regression for prediction of spectral peak structures

Materials informatics exploiting machine learning techniques has the potential to afford high-throughput experiments. In the various spectroscopies widely used in the research fields of materials science and physics, detailed peak structures provide vitally important information. Here, we propose an improved adaptive sampling method for such spectroscopies, which utilizes Gaussian process regression with a new heuristic evaluation function. Application of this method to various spectroscopies reveals that it enables prediction of detailed peak structures using fewer sampling points while preserving the required accuracy. Our method reduces the measurement time and cost and thus increases the throughput of sample characterization.