Machine Learning Hits Molecular Simulations
Machine learning (ML) is changing the way we approach molecular simulations. By translating ab-initio calculations into neural networks, it is possible to reach sizes and time scales not possible before and to analyze in detail the performance of the ab-initio approaches. ML can also lead to new force fields specially targeted to reproduce certain properties or to develop coarse grained models, opening new frontiers in numerical studies of collective molecular behavior. In addition, ML can be useful in analyzing trajectories and searching for efficient order parameters and reaction coordinates, in achieving enhanced sampling of rare events, and in discovering unexpected connections between structural and dynamic properties, including the evaluation of correlation and memory functions. This special issue will focus on applications of ML in molecular simulation with a particular focus on biomolecular and materials systems.
Guest Editors: Laura Filion, Frank Noe, Christine Peter, Pratyush Tiwary, Christoph Dellago, with JCP Editors Carlos Vega, Michele Ceriotti, Francesco Sciortino, and John Straub.