FAENet: Frame Averaging Equivariant GNN for Materials Modeling

Alexandre Agm Duval, Victor Schmidt, Alex Hernández-Garcı́a, Santiago Miret, Fragkiskos D. Malliaros, Yoshua Bengio, David Rolnick
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:9013-9033, 2023.

Abstract

Applications of machine learning techniques for materials modeling typically involve functions that are known to be equivariant or invariant to specific symmetries. While graph neural networks (GNNs) have proven successful in such applications, conventional GNN approaches that enforce symmetries via the model architecture often reduce expressivity, scalability or comprehensibility. In this paper, we introduce (1) a flexible, model-agnostic framework based on stochastic frame averaging that enforces E(3) equivariance or invariance, without any architectural constraints; (2) FAENet: a simple, fast and expressive GNN that leverages stochastic frame averaging to process geometric information without constraints. We prove the validity of our method theoretically and demonstrate its superior accuracy and computational scalability in materials modeling on the OC20 dataset (S2EF, IS2RE) as well as common molecular modeling tasks (QM9, QM7-X).

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-duval23a, title = {{FAEN}et: Frame Averaging Equivariant {GNN} for Materials Modeling}, author = {Duval, Alexandre Agm and Schmidt, Victor and Hern\'{a}ndez-Garc\'{\i}a, Alex and Miret, Santiago and Malliaros, Fragkiskos D. and Bengio, Yoshua and Rolnick, David}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {9013--9033}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/duval23a/duval23a.pdf}, url = {https://proceedings.mlr.press/v202/duval23a.html}, abstract = {Applications of machine learning techniques for materials modeling typically involve functions that are known to be equivariant or invariant to specific symmetries. While graph neural networks (GNNs) have proven successful in such applications, conventional GNN approaches that enforce symmetries via the model architecture often reduce expressivity, scalability or comprehensibility. In this paper, we introduce (1) a flexible, model-agnostic framework based on stochastic frame averaging that enforces E(3) equivariance or invariance, without any architectural constraints; (2) FAENet: a simple, fast and expressive GNN that leverages stochastic frame averaging to process geometric information without constraints. We prove the validity of our method theoretically and demonstrate its superior accuracy and computational scalability in materials modeling on the OC20 dataset (S2EF, IS2RE) as well as common molecular modeling tasks (QM9, QM7-X).} }
Endnote
%0 Conference Paper %T FAENet: Frame Averaging Equivariant GNN for Materials Modeling %A Alexandre Agm Duval %A Victor Schmidt %A Alex Hernández-Garcı́a %A Santiago Miret %A Fragkiskos D. Malliaros %A Yoshua Bengio %A David Rolnick %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-duval23a %I PMLR %P 9013--9033 %U https://proceedings.mlr.press/v202/duval23a.html %V 202 %X Applications of machine learning techniques for materials modeling typically involve functions that are known to be equivariant or invariant to specific symmetries. While graph neural networks (GNNs) have proven successful in such applications, conventional GNN approaches that enforce symmetries via the model architecture often reduce expressivity, scalability or comprehensibility. In this paper, we introduce (1) a flexible, model-agnostic framework based on stochastic frame averaging that enforces E(3) equivariance or invariance, without any architectural constraints; (2) FAENet: a simple, fast and expressive GNN that leverages stochastic frame averaging to process geometric information without constraints. We prove the validity of our method theoretically and demonstrate its superior accuracy and computational scalability in materials modeling on the OC20 dataset (S2EF, IS2RE) as well as common molecular modeling tasks (QM9, QM7-X).
APA
Duval, A.A., Schmidt, V., Hernández-Garcı́a, A., Miret, S., Malliaros, F.D., Bengio, Y. & Rolnick, D.. (2023). FAENet: Frame Averaging Equivariant GNN for Materials Modeling. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:9013-9033 Available from https://proceedings.mlr.press/v202/duval23a.html.

Related Material