L. Dumortier

Managing Expectations and Imbalanced Training Data in Reactive Force Field Development: An Application to Water Adsorption on Alumina

L. Dumortier, C. Chizallet, B. Creton, T. De Bruin, T. Verstraelen
Journal of Chemical Theory and Computation (JCTC)


ReaxFF is a computationally efficient model for reactive molecular dynamics simulations that has been applied to a wide variety of chemical systems. When ReaxFF parameters are not yet available for a chemistry of interest, they must be (re)optimized, for which one defines a set of training data that the new ReaxFF parameters should reproduce. ReaxFF training sets typically contain diverse properties with different units, some of which are more abundant (by orders of magnitude) than others. To find the best parameters, one conventionally minimizes a weighted sum of squared errors over all of the data in the training set. One of the challenges in such numerical optimizations is to assign weights so that the optimized parameters represent a good compromise among all the requirements defined in the training set. This work introduces a new loss function, called Balanced Loss, and a workflow that replaces weight assignment with a more manageable procedure. The training data are divided into categories with corresponding “tolerances”, i.e., acceptable root-mean-square errors for the categories, which define the expectations for the optimized ReaxFF parameters. Through the Log-Sum-Exp form of Balanced Loss, the parameter optimization is also a validation of one’s expectations, providing meaningful feedback that can be used to reconfigure the tolerances if needed. The new methodology is demonstrated with a nontrivial parametrization of ReaxFF for water adsorption on alumina. This results in a new force field that reproduces both the rare and frequent properties of a validation set not used for training. We also demonstrate the robustness of the new force field with a molecular dynamics simulation of water desorption from a γ-Al2O3 slab model.

This publication is licensed under the terms of your institutional subscription. Request reuse permissions.

From Ionic Surfactants to Nafion through Convolutional Neural Networks

L. Dumortier, S. Mossa
Journal of Physical Chemistry B
Volume 124 Issue 40 Pages 8918-8927
Published while none of the authors were employed at the CMM


We have applied recent machine learning advances-deep convolutional neural networks-to three-dimensional (voxels) soft matter data, generated by molecular dynamics computer simulation. We have focused on the structural and phase properties of a coarse-grained model of hydrated ionic surfactants. We have trained a classifier able to automatically detect the water quantity absorbed in the system, therefore associating to each hydration level the corresponding most representative nanostructure. On the basis of the notion of transfer learning, we have next applied the same network to the related polymeric ionomer Nafion and have extracted a measure of the similarity of these configurations with those above. We demonstrate that on this basis it is possible to express the static structure factor of the polymer at fixed hydration level as a superposition of those of the surfactants at multiple water contents. We suggest that such a procedure can provide a useful, agnostic, data-driven, and precise description of the multiscale structure of disordered materials, without resorting to any a priori model picture.

Subscribe to RSS - L. Dumortier