While instance-level explanation of GNN is a well-studied problem with plenty of approaches being developed, providing a global explanation for the behaviour of a GNN is much less explored, despite its potential in interpretability and debugging. Existing solutions either simply list local explanations for a given class, or generate a synthetic prototypical graph with maximal score for a given class, completely missing any combinatorial aspect that the GNN could have learned. In this work, we propose GLGExplainer (Global Logic-based GNN Explainer), the first Global Explainer capable of generating explanations as arbitrary Boolean combinations of learned graphical concepts. GLGExplainer is a fully differentiable architecture that takes local explanations as inputs and combines them into a logic formula over graphical concepts, represented as clusters of local explanations. Contrary to existing solutions, GLGExplainer provides accurate and human-interpretable global explanations that are perfectly aligned with ground-truth explanations (on synthetic data) or match existing domain knowledge (on real-world data). Extracted formulas are faithful to the model predictions, to the point of providing insights into some occasionally incorrect rules learned by the model, making GLGExplainer a promising diagnostic tool for learned GNNs.

Global Explainability of GNNs via Logic Combination of Learned Concepts / Azzolin, S.; Longa, A.; Barbiero, P.; Lio, P.; Passerini, A.. - (2023). (Intervento presentato al convegno 11th International Conference on Learning Representations, ICLR 2023 tenutosi a Kigali; rwa).

Global Explainability of GNNs via Logic Combination of Learned Concepts

Lio P.
;
2023

Abstract

While instance-level explanation of GNN is a well-studied problem with plenty of approaches being developed, providing a global explanation for the behaviour of a GNN is much less explored, despite its potential in interpretability and debugging. Existing solutions either simply list local explanations for a given class, or generate a synthetic prototypical graph with maximal score for a given class, completely missing any combinatorial aspect that the GNN could have learned. In this work, we propose GLGExplainer (Global Logic-based GNN Explainer), the first Global Explainer capable of generating explanations as arbitrary Boolean combinations of learned graphical concepts. GLGExplainer is a fully differentiable architecture that takes local explanations as inputs and combines them into a logic formula over graphical concepts, represented as clusters of local explanations. Contrary to existing solutions, GLGExplainer provides accurate and human-interpretable global explanations that are perfectly aligned with ground-truth explanations (on synthetic data) or match existing domain knowledge (on real-world data). Extracted formulas are faithful to the model predictions, to the point of providing insights into some occasionally incorrect rules learned by the model, making GLGExplainer a promising diagnostic tool for learned GNNs.
2023
11th International Conference on Learning Representations, ICLR 2023
Boolean combinations; Combinatorial aspect; Domain knowledge; Existing domains; Ground truth; Interpretability; Logic formulas; Model prediction; Real-world; Synthetic data
04 Pubblicazione in atti di convegno::04b Atto di convegno in volume
Global Explainability of GNNs via Logic Combination of Learned Concepts / Azzolin, S.; Longa, A.; Barbiero, P.; Lio, P.; Passerini, A.. - (2023). (Intervento presentato al convegno 11th International Conference on Learning Representations, ICLR 2023 tenutosi a Kigali; rwa).
File allegati a questo prodotto
File Dimensione Formato  
Azzolin_Global-explainability_2023.pdf

accesso aperto

Tipologia: Versione editoriale (versione pubblicata con il layout dell'editore)
Licenza: Creative commons
Dimensione 852.23 kB
Formato Adobe PDF
852.23 kB Adobe PDF

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11573/1726291
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 15
  • ???jsp.display-item.citation.isi??? ND
social impact