Counterfactual Explanations for Equivariant GNNs
CE-EGNN for molecular property regression.
CE-EGNN studies counterfactual explanations for equivariant graph neural networks used in molecular property regression. The work connects geometric deep learning with explainable AI for molecular modeling.
This project was published with Fatia Lekbour and Guillaume Renton at the EDBT/ICDT 2026 Workshops.