Improving the Quality of Rule-Based GNN Explanations

From LRDE

Abstract

Recent works have proposed to explain GNNs using activation rules. Activation rules allow to capture specific configurations in the embedding space of a given layer that is discriminant for the GNN decision. These rules also catch hidden features of input graphs. This requires to associate these rules to representative graphs. In this paper, we propose on the one hand an analysis of heuristic-based algorithms to extract the activation rulesand on the other hand the use of transport-based optimal graph distances to associate each rule with the most specific graph that triggers them.


Bibtex (lrde.bib)

@InProceedings{	  kamal.22.xkdd,
  author	= {Ataollah Kamal and Elouan Vincent and Marc Plantevit and
		  C\'{e}line Robardet},
  booktitle	= {ECML PKDD International Workshop on eXplainable Knowledge
		  Discovery in Data Mining},
  title		= {Improving the Quality of Rule-Based {GNN} Explanations},
  year		= {2022},
  address	= {Grenoble, France},
  month		= sep,
  abstract	= {Recent works have proposed to explain GNNs using
		  activation rules. Activation rules allow to capture
		  specific configurations in the embedding space of a given
		  layer that is discriminant for the GNN decision. These
		  rules also catch hidden features of input graphs. This
		  requires to associate these rules to representative graphs.
		  In this paper, we propose on the one hand an analysis of
		  heuristic-based algorithms to extract the activation rules,
		  and on the other hand the use of transport-based optimal
		  graph distances to associate each rule with the most
		  specific graph that triggers them.},
  pages		= {1--16},
  note		= {accepted}
}