Symposium MT02-Closing the Loop—Using Machine Learning in High-Throughput Discovery of New Materials

Automation of high-throughput experiments and computations has resulted in the creation of materials data at a scale hitherto unknown. Meanwhile, 7 years of efforts within the Materials Genome Initiative have provided tremendous advances in materials data science, empowering researchers to analyze and extract knowledge from data at unprecedented rates. This has created a field at the boundary of materials science and computer science in which artificial intelligence (A.I.) is being used to drive materials research. In fact, with the prevalence of open-source machine learning platforms, a materials researcher can quickly and effectively apply A.I. to their research. Initial efforts have already led to reports of ten to thousand-fold acceleration in the rate of materials discovery, especially when these approaches are augmented with high-throughput approaches. However, most of the general materials community lacks formal computer science training. This leads to concerns about the uniformed application of A.I. and hinders community acceptance of these powerful tools potentially resulting in a slower discovery of important materials. This symposium will provide a platform for machine learning materials science early adopters, materials data platform creators, and cutting-edge computer scientists to present their recent work exploring the interface between computer science and materials science.

Topics will include:

Define and highlight the current state of the art in artificial intelligence assisted materials science.
Enable a discussion of how physical laws can be used as constraints that allow “Big Data” algorithms to provide meaningful insights from small “deep” materials data.
Build mutually beneficial interactions with computer scientists
Highlight the emergence of autonomous “closed-loop” systems which merge automated experimental and theoretical studies with artificial intelligence
Materials knowledge representation, ontologies, and artificial intelligence
Uncertainty quantification, prior formation, and Bayesian methods in materials research
High-throughput experimental and computational approaches to novel materials discovery
Materials by Design

Invited Speakers:

John Gregoire (California Institute of Technology, USA)
Kristin Persson (Lawrence Berkeley National Laboratory, USA)
Carla Gomes (Cornell University, USA)
Vladan Stevanovic (Colorado School of Mines, USA)
Benji Maruyama (Air Force Research Laboratory, USA)
Anatole von Lilienfeld (University of Basel, Switzerland)
Olga Wodo (University at Buffalo, USA)
Rampi Ramprasad (Georgia Institute of Technology, USA)
Gabor Csanyi (University of Cambridge, United Kingdom)
Elsa Olivetti (Massachusetts Institute of Technology, USA)
Alan Aspuru-Guzik (University of Toronto, Canada)
Lee Cronin (University of Glasgow, United Kingdom)
Brian DeCost (National Institute of Standards and Technology, USA)
Roman Garnett (Washington University in St. Loius, USA)
Olexandr Isayev (University of North Carolina, USA)
Heather Kulik (Massachusetts Institute of Technology, USA)
Jatin Kumar (Nanyang Technological University, Singapore)
Julia Ling (Citrine, USA)
Klaus-Robert Muller (Technische Universität Berlin, Germany)
Matthias Scheffler (Fritz-Haber-Institute of the Max-Planck-Society, Germany)
Yibin Xu (National Institute for Materials Science, Japan)
Lusann Yang (Google, USA)