Prof. Billinge has more than 25 years of experience developing and applying techniques to study local structure in materials using X-ray, neutron and electron diffraction including the development of novel data analysis methods including graph theoretic, Artificial Intelligence and Machine Learning approaches. He earned his Ph.D in Materials Science and Engineering from University of Pennsylvania in 1992. After 13 years as a faculty member at Michigan State University, in 2008 he took up his current position as Professor of Materials Science and Applied Physics and Applied Mathematics at Columbia University and Physicist at Brookhaven National Laboratory.
Prof. Billinge has published more than 300 papers in scholarly journals. He is a fellow of the American Physical Society and the Neutron Scattering Society of America, a former Fulbright and Sloan fellow and has earned a number of awards including the 2018 Warren Award of the American Crystallographic Association and being honored in 2011 for contributions to the nation as an immigrant by the Carnegie Corporation of New York, the 2010 J. D. Hanawalt Award of the International Center for Diffraction Data, University Distinguished Faculty award at Michigan State, the Thomas H. Osgood Undergraduate Teaching Award. He is Section Editor of Acta Crystallographica Section A: Advances and Foundations. He regularly chairs and participates in reviews of major facilities and federally funded programs.
Lecture 13: Simon Billinge
Do materials have a genome, and if they do, what can we do with it?
Simon Billinge
Professor of Materials Science and Applied Physics and Applied Mathematics,
School of Engineering and Applied Science, Columbia University, USA
E-mail: sb2896@columbia.edu>
Abstract:
The materials genome initiative (MGI) is a US government initiative from 2011 to speed up, and lower the cost of, materials discovery, development and deployment in technology. The idea is to explore whether we can learn from lessons of the human genome project where the application of data mining and more broadly data analytics is used to speed up drug discovery. The approach has been viewed as being successful and was recently renewed. It is clear that data mining and data analytics can help in this process, even if materials don't actually have a genome. But don't they? We have been exploring whether the atomic pair distribution function can act as a defacto materials genome, and whether this insight allows novel mathematical and computational methods to help in the goals of the MGI in general. In this talk Iwill describe recent efforts to apply machine learning to the analysis of total scattering and PDF data and modeling. This includes the use of deep neural nets for finding the unknown space-group of a measured PDF. I will also present preliminary results of pplying variational conditional autoencoders to do structure solution from PDF data, at least in a very limited way. Some of these tools are available as cloud-hosted web services at PDFitc.org and are free to use by the community.