Machine learning is a valuable tool in the study of materials and their properties. However, like other data-driven methods, its effectiveness is dependent on the quality and quantity of data that goes into a model. Materials data is diverse, siloed, and represented in many different formats so the tools that build on top of it must reconcile those differences if they are to be used most effectively. This talk covers work towards building a unified system for storing broad sets of materials and chemical data, generating machine learning models on top of that data, and the effectiveness of adding physical models to that predictive system.
Back to Long Programs