Argonne National Laboratory

Workshop on Machine Learning

February 6, 2018

Machine learning (ML) approaches are increasingly being used in science and engineering applications, with numerous successes reported by industry, academia, and research communities. Despite these successes, however, numerous challenges remain. Rigorous analytic techniques are needed, for example, for testing ML methods, understanding their approximation power and compute complexity, and improving their predictive capabilities.

To address these challenges, the DOE Office of Advanced Scientific Computing Research sponsored a Scientific Machine Learning Workshop in North Bethesda, Maryland, on January 30 - February 1, 2018. 

Stefan Wild, a computer scientist and deputy division director in Argonne’s Mathematics and Computer Science Division, was on the workshop organizing committee, which included scientists from seven other national laboratories and six universities. Testifying to the importance of the topic is the fact that the committee received 140 position papers.   

The three-day workshop featured plenary talks and multiple breakout sessions on topics including empirical modeling, probabilistic learning, numerical analysis, multifidelity and reduced-order models, optimization and complexity, and interpretability. The aim of the workshop is to produce a report on how applied mathematics can be used to increase the rigor, robustness, and reliability of machine learning for DOE science and engineering applications.

For further information see the website