Molecular relaxation by reverse diffusion with time step prediction

arXiv:2404.10935v1 Announce Type: cross
Abstract: Molecular relaxation, finding the equilibrium state of a non-equilibrium structure, is an essential component of computational chemistry to understand reactivity. Classical force field methods often rely on insufficient local energy minimization, while neural network force field models require large labeled datasets encompassing both equilibrium and non-equilibrium structures. As a remedy, we propose MoreRed, molecular relaxation by reverse diffusion, a conceptually novel and purely statistical approach where non-equilibrium structures are treated as noisy instances of their …

abstract arxiv chemistry computational computational chemistry cs.lg datasets diffusion energy equilibrium network neural network physics.chem-ph physics.comp-ph prediction state stat.ml type