Kernel-based machine learning for efficient simulations of molecular liquids.

Current machine learning (ML) models aimed at learning force fields are plagued by their high computational cost at every integration time step. We describe a number of practical and computationall...