Molecular Simulations Meets Machine Learning and Artificial Intelligence

Description and aim

Rational design of advanced materials, such as smart energy materials, cognitive materials, bio-inspired materials, etc., requires the use of dedicated multiscale simulation methods. The hierarchy of spatio-temporal scales giving rise to material properties necessitates the use of multi-scale simulation techniques, from ab initio quantum mechanics (QM) via all-atom and coarse-grain molecular dynamics (MD) up to the continuum level. Currently, a limiting factor is handling of the big data that either goes into such models or results from the predictions. Here the use of machine learning techniques (ML) (including kernel methods and deep neural networks) and methods borrowed from the realm of artificial intelligence (AI) (neuromorphic computing, data analytics, and robotic technologies) are needed to bring the field of computational material design ahead.

The goal of this workshop is to connect several disciplinary lines and to bring together top researchers in the fields of QM, MD, ML, and AI to develop unified methodology, covering the whole simulation pipeline, for the rational design of advanced materials. In this workshop (i) practical connections between molecular simulations (QM, MD) on the one hand and ML/AI on the other hand will be identified, (ii) the major challenges and existing bottlenecks as well as ideas on how to improve these connections will be discussed, from different disciplinary aspects, and (iii) new collaborative networks will be delineated. Establishing such an interactive scientific network will serve as a solid foundation toward the ultimate goal of developing novel materials needed for efficient neuromorphic computers, improved photovoltaics, or biomedical applications.

CECAM-Lorentz call
This workshop is winner of the CECAM-Lorentz call 2017. For more information see our CECAM-Lorentz program.