Coherent energy and force uncertainty in deep learning force fields
This work addresses uncertainty quantification for researchers in computational chemistry and materials science, offering a method to improve reliability in atomic system simulations, though it is incremental as it builds on existing deep learning and uncertainty modeling techniques.
The paper tackled the problem of quantifying aleatoric uncertainty in machine learning force fields, where energy uncertainty does not naturally translate to force uncertainty, by proposing a model that links energy and force uncertainty through a spatially correlated noise process, demonstrating it on equivariant neural networks with two molecular datasets.
In machine learning energy potentials for atomic systems, forces are commonly obtained as the negative derivative of the energy function with respect to atomic positions. To quantify aleatoric uncertainty in the predicted energies, a widely used modeling approach involves predicting both a mean and variance for each energy value. However, this model is not differentiable under the usual white noise assumption, so energy uncertainty does not naturally translate to force uncertainty. In this work we propose a machine learning potential energy model in which energy and force aleatoric uncertainty are linked through a spatially correlated noise process. We demonstrate our approach on an equivariant messages passing neural network potential trained on energies and forces on two out-of-equilibrium molecular datasets. Furthermore, we also show how to obtain epistemic uncertainties in this setting based on a Bayesian interpretation of deep ensemble models.