Abstract
Molecular dynamics simulations are a cornerstone in science, allowing toinvestigate from the system's thermodynamics to analyse intricate molecularinteractions. In general, to create extended molecular trajectories can be acomputationally expensive process, for example, when running $ab-initio$simulations. Hence, repeating such calculations to either obtain more accuratethermodynamics or to get a higher resolution in the dynamics generated by afine-grained quantum interaction can be time- and computationally-consuming. Inthis work, we explore different machine learning (ML) methodologies to increasethe resolution of molecular dynamics trajectories on-demand within apost-processing step. As a proof of concept, we analyse the performance ofbi-directional neural networks such as neural ODEs, Hamiltonian networks,recurrent neural networks and LSTMs, as well as the uni-directional variants asa reference, for molecular dynamics simulations (here: the MD17 dataset). Wehave found that Bi-LSTMs are the best performing models; by utilizing the localtime-symmetry of thermostated trajectories they can even learn long-rangecorrelations and display high robustness to noisy dynamics across molecularcomplexity. Our models can reach accuracies of up to 10$^{-4}$ angstroms intrajectory interpolation, while faithfully reconstructing several full cyclesof unseen intricate high-frequency molecular vibrations, rendering thecomparison between the learned and reference trajectories indistinguishable.The results reported in this work can serve (1) as a baseline for largersystems, as well as (2) for the construction of better MD integrators.