On the Two Sides of Redundancy in Graph Neural Networks

arXiv:2310.04190v2 Announce Type: replace
Abstract: Message passing neural networks iteratively generate node embeddings by aggregating information from neighboring nodes. With increasing depth, information from more distant nodes is included. However, node embeddings may be unable to represent the growing node neighborhoods accurately and the influence of distant nodes may vanish, a problem referred to as oversquashing. Information redundancy in message passing, i.e., the repetitive exchange and encoding of identical information amplifies oversquashing. We develop a novel aggregation scheme based on …

abstract arxiv cs.lg embeddings generate graph graph neural networks however influence information networks neural networks node nodes redundancy type