E(n) Equivariant Graph Neural Networks
This provides a more efficient and scalable approach for tasks like dynamical systems modelling and molecular property prediction, with incremental improvements over existing equivariant methods.
The paper tackles the problem of learning graph neural networks equivariant to rotations, translations, reflections, and permutations by introducing E(n)-Equivariant Graph Neural Networks (EGNNs), achieving competitive or better performance without computationally expensive higher-order representations and scaling to higher-dimensional spaces.
This paper introduces a new model to learn graph neural networks equivariant to rotations, translations, reflections and permutations called E(n)-Equivariant Graph Neural Networks (EGNNs). In contrast with existing methods, our work does not require computationally expensive higher-order representations in intermediate layers while it still achieves competitive or better performance. In addition, whereas existing methods are limited to equivariance on 3 dimensional spaces, our model is easily scaled to higher-dimensional spaces. We demonstrate the effectiveness of our method on dynamical systems modelling, representation learning in graph autoencoders and predicting molecular properties.