About this Abstract |
| Meeting |
MS&T25: Materials Science & Technology
|
| Symposium
|
Enhancing the Accessibility of Machine Learning-Enabled Experiments
|
| Presentation Title |
From automated to autonomous – creating a general active learning service for self-driving laboratories |
| Author(s) |
Stephen J. DeWitt, Ankit Shrivastava, Marshall McDonnell, Singanallur Venkatakrishnan, Chris Fancher, Jorge Ramirez Osorio, Paul Laiu, Lance Drane, Ayana Ghosh, David Joy |
| On-Site Speaker (Planned) |
Stephen J. DeWitt |
| Abstract Scope |
A key step in most scientific workflows is answering the question “Given the data I already have, what should I measure next?”. The field of sequential experimental design is targeted toward answering this question and is critical to convert automated experiments into autonomous (self-driving) experiments. In this presentation, we discuss the creation of Dial, a new open-source tool that applies active learning methods for sequential experimental design. Dial operates as a service in the broader INTERSECT ecosystem for autonomous experiments. After presenting the design and structure of Dial, we show examples where Dial was used to drive experiments, including neutron diffraction experiments to map residual strain in 3D-printed metal parts and to identify the temperature of the “spin flop” transition in hematite nanoparticles.
This abstract has been authored by UT-Battelle, LLC under Contract No. DE-AC05-00OR22725 with the U.S. Department of Energy. |