Empirical comparison shows gradient-based explanations for GNN node similarities are actionable, consistent, and retain effects when sparsified, unlike mutual information explanations.
ter Hoeve, Gabriele Tolomei, Maarten de Rijke, and Fabrizio Silvestri
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Explaining Graph Neural Networks for Node Similarity on Graphs
Empirical comparison shows gradient-based explanations for GNN node similarities are actionable, consistent, and retain effects when sparsified, unlike mutual information explanations.