Symposium GI01-Machine Learning and Data-Driven Materials Development and Design

Materials are an important contributor to technological progress, and yet the process of materials discovery and development has historically been inefficient. In general, the current innovation workflow is human-centered, where researchers design, conduct, analyze and interpret results obtained through experiments, simulations or literature review. Such results are often high-dimensional, large in number and heterogeneous in nature, which hinders a researcher’s ability to draw insight from this data manually. Efforts such as the Materials Genome Initiative address this issue through data-driven techniques, increasingly utilizing methods from statistics, machine learning and artificial intelligence to interpret and model materials data. Such techniques have allowed researchers to make predictions of materials structures and properties, perform autonomous research with robot scientists, and speed up computational work. The result is an acceleration of materials discovery, and a strengthened ability to draw scientific insights from complex data. This symposium will explore the synthesis of machine learning with materials research, highlighting a broad spectrum of topics in which machine learning, artificial intelligence, or statistics play a significant role in addressing problems in experimental and theoretical materials science. It will also spur discussion on the fundamental connection between machine learning and material science, and its future application and impact. The combination of machine learning with materials science has broad implications for the future of research processes, the roles of humans and intelligent systems in a coordinated human-machine partnership, and ultimately the rate at which materials research and scientific understanding progresses.

Topics will include:

Closed-loop, fully autonomous experimental planning and research
Computational materials augmented with machine learning
Uncertainty quantification, prior formation and Bayesian methods in materials research
Synthesis of physics-oriented models, knowledge, and constraints with statistical and machine learning techniques
Automated or computer-assisted learning of physical models and basic scientific discovery
Optimal experimental design under uncertainty
High performance computing and big data approaches to machine learning-driven materials development
Materials knowledge representation, ontologies and artificial intelligence
Human-machine interfaces and interactions in materials research

Invited Speakers:

Kristin Persson (Lawrence Berkeley National Laboratory, USA)
Patrick Riley (Google Accelerated Science, USA)
Juan de Pablo (University of Chicago, USA)
Leonardo Ajdelsztajn (General Electric, USA)
Placidus Amama (Kansas State University, USA)
Kieron Burke (University of California, Irvine, USA)
Lucy Colwell (University of Cambridge, United Kingdom)
Yu Ding (Texas A&M University, USA)
A. John Hart (Massachusetts Institute of Technology, USA)
Ross King (University of Manchester, United Kingdom)
Michael Krein (Lockheed Martin, USA)
A. Gilad Kusne (National Institute of Standards and Technology, USA)
Turab Lookman (Los Alamos National Laboratory, USA)
Noa Marom (Carnegie Mellon University, USA)
Benji Maruyama (Air Force Research Laboratory, USA)
Marco Buongiorno Nardelli (University of North Texas, USA)
Ruth Pachter (Air Force Research Laboratory, USA)
Krishna Rajan (University at Buffalo, USA)
Mark Rummeli (Leibniz Institute for Solid State and Materials Research Desden, Germany)
Larisa Soldatova (Goldsmiths University of London, United Kingdom)
Brian Storey (Toyota Research Institute, USA)
R. Bruce von Dover (Cornell University, USA)
Anatole von Lilienfeld (University of Basel, Switzerland)
Olga Wodo (University at Buffalo, USA)