Poster
in
Workshop: 2nd Annual Workshop on Topology, Algebra, and Geometry in Machine Learning (TAG-ML)
Learning to Explain Hypergraph Neural Networks
Sepideh Maleki · Ehsan Hajiramezanali · Gabriele Scalia · Tommaso Biancalani · Kangway Chuang
Hypergraphs are expressive structures for describing higher-order relationships among entities, with widespread applications across biology and drug discovery. Hypergraph neuralnetworks (HGNNs) have recently emerged as apromising representation learning approach onthese structures for clustering, classification, andmore. However, despite their promising performance, HGNNs remain a black box, and explaining how they make predictions remains an open challenge. To address this problem, we proposeHyperEX, a post-hoc explainability frameworkfor hypergraphs that can be applied to any trainedHGNN. HyperEX computes node-hyperedge pairimportance to identify sub-hypergraphs as explanations. Our experiments demonstrate how HyperEX learns important sub-hypergraphs responsible for driving node classification to give usefulinsight into HGNNs.