Congratulations to undergraduate students Vidit Agrawal and Shixin Zhang for their contributions and co-authorship on their work on “Accelerating ensemble uncertainty estimates in supervised materials property regression models”! They worked closely with mentors Dr. Lane Schultz and Professor Dane Morgan throughout their work. Great Job!
Their work developed a faster way to estimate uncertainties in predicting material properties using machine learning. Normally, researchers use “ensembles,” which are groups of models working together, but these require much more computing power than a single model. The team created a method to train just one model that mimics the uncertainty estimates of an ensemble, achieving accurate uncertainty predictions with far less computational effort. The result is a more efficient and flexible tool for predicting material properties, making advanced techniques accessible for more researchers.
Check out the publication in Computational Materials Science!