Are High-Degree Representations Really Unnecessary in Equivariant Graph Neural Networks?

Open in new window