Boosting Graph Pooling with Persistent Homology

arXiv:2402.16346v1 Announce Type: new
Abstract: Recently, there has been an emerging trend to integrate persistent homology (PH) into graph neural networks (GNNs) to enrich expressive power. However, naively plugging PH features into GNN layers always results in marginal improvement with low interpretability. In this paper, we investigate a novel mechanism for injecting global topological invariance into pooling layers using PH, motivated by the observation that filtration operation in PH naturally aligns graph pooling in a cut-off manner. In this fashion, …

abstract arxiv boosting cs.lg features global gnn gnns graph graph neural networks improvement interpretability low math.at networks neural networks novel paper pooling power results trend type