Timezone: »

FAENet: Frame Averaging Equivariant GNN for Materials Modeling
ALEXANDRE DUVAL · Victor Schmidt · Alex Hernandez-Garcia · Santiago Miret · Fragkiskos Malliaros · Yoshua Bengio · David Rolnick

Tue Jul 25 05:00 PM -- 06:30 PM (PDT) @ Exhibit Hall 1 #323

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).

Author Information


PhD Student working on Graph Machine Learning: GNN explicability, graph pooling and 3D molecular prediction for accelerated catalyst discovery. Currently at MILA in Montreal, under the supervision of David Rolnick & Yoshua Bengio.

Victor Schmidt (Mila / Université de Montréal)
Alex Hernandez-Garcia (Mila - Quebec AI Institute)
Santiago Miret (Intel Labs)
Fragkiskos Malliaros (CentraleSupelec, Paris-Saclay University)
Yoshua Bengio (Mila - Quebec AI Institute)
David Rolnick (McGill University, Mila)

More from the Same Authors