Machine learning for molecular simulation

Dynamical processes occurring in condensed matter systems can be modeled atomistically using first principles, but such calculations are very demanding, severely limiting accessible systems sizes and simulation times. Machine learning methods have recently emerged as a promising approach to address these challenges and to bridge the gap between ab initio calculations and large-scale molecular dynamics simulations of complex systems. We have recently developed a software package for the development and application of NNs, called n2p2, and have used it to create accurate ab initio quality potentials for water and other substances. Furthermore, we explore the application of machine learning methods for the analysis and rationalization of simulation results.

^{P. Geiger and C. Dellago. “Neural networks for local structure detection in polymorphic systems”, J. Chem. Phys. 139, 164105 (2013).
T. Morawietz, A. Singraber, C. Dellago, and J. Behler, „How Van der Waals interactions determine the unique properties of water“, Proc. Natl. Acad. Sci. USA 113, 8368-8373 (2016). 
A. Singraber, J. Behler, and C. Dellago, “Library-based LAMMPS implementation of high-dimensional neural network potentials”, J. Chem. Theo. Comp. 15, 1827 (2019).
A. Singraber, T. Morawietz, J. Behler, and C. Dellago, „Parallel multistream training of high-dimensional neural network potentials”, J. Chem. Theo. Comp, 15, 3075-3092 (2019).}

^{N2P2 on github gttps://github.com/CompPhysVienna/n2p2}