The authors introduce a new method called Clifford Group Equivariant Simplicial Message Passing Networks. This method combines Clifford group-equivariant layers with simplicial message passing to express geometric features. The method outperforms both equivariant and simplicial graph neural networks on a variety of geometric tasks. The authors share parameters of the message network across different dimensions for efficient simplicial message passing. The final message is restricted to an aggregation of incoming messages from different dimensions, leading to shared simplicial message passing. The implementation is available on GitHub.
Publication date: 16 Feb 2024
Project Page: https://arxiv.org/abs/2402.1001
Paper: https://arxiv.org/pdf/2402.10011
