This project provides research opportunities  to develop statistical methodology or computer software to address increasing needs at NIST on using machine learning to solve interesting engineering applications or physical measurement problems. Modern measuring devices often produce data in the form of spectra or 3d images (such as hyperspectral images and OCT),  the goal is to provide statistical methodology for measurements and uncertainty analysis based on such high throughput data. Functional data analysis in designed experiments settings is often encountered in standard developments. Both supervised learning and unsupervised learning including  singular value decomposition to extract spectral signatures are important  and are our current interests.  We're also interested in using nonaparmetric Bayes, including empirical Bayes methods for analyzing parallel and many data sets sampled from one or multiple populations.  

References: 

https://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.SP.260-228.pdf

https://doi.org/10.1557/s43578-021-00362-8

https://doi.org/10.1117/1.JBO.29.9.093503