Improving the interpretability of GNN predictions through conformal-based graph sparsification

arXiv:2404.12356v1 Announce Type: cross
Abstract: Graph Neural Networks (GNNs) have achieved state-of-the-art performance in solving graph classification tasks. However, most GNN architectures aggregate information from all nodes and edges in a graph, regardless of their relevance to the task at hand, thus hindering the interpretability of their predictions. In contrast to prior work, in this paper we propose a GNN \emph{training} approach that jointly i) finds the most predictive subgraph by removing edges and/or nodes -- -\emph{without making assumptions about …

abstract architectures art arxiv classification contrast cs.lg cs.si gnn gnns graph graph neural networks however improving information interpretability networks neural networks nodes performance predictions state stat.ml tasks through type