CF-GNNExplainer: Counterfactual Explanations for Graph Neural Networks

Ana Lucic, Maartje A. Ter Hoeve, Gabriele Tolomei, Maarten De Rijke, Fabrizio Silvestri
Proceedings of The 25th International Conference on Artificial Intelligence and Statistics, PMLR 151:4499-4511, 2022.

Abstract

Given the increasing promise of graph neural networks (GNNs) in real-world applications, several methods have been developed for explaining their predictions. Existing methods for interpreting predictions from GNNs have primarily focused on generating subgraphs that are especially relevant for a particular prediction. However, such methods are not counterfactual (CF) in nature: given a prediction, we want to understand how the prediction can be changed in order to achieve an alternative outcome. In this work, we propose a method for generating CF explanations for GNNs: the minimal perturbation to the input (graph) data such that the prediction changes. Using only edge deletions, we find that our method, CF-GNNExplainer, can generate CF explanations for the majority of instances across three widely used datasets for GNN explanations, while removing less than 3 edges on average, with at least $94%$ accuracy. This indicates that CF-GNNExplainer primarily removes edges that are crucial for the original predictions, resulting in minimal CF explanations.

Cite this Paper


BibTeX
@InProceedings{pmlr-v151-lucic22a, title = { CF-GNNExplainer: Counterfactual Explanations for Graph Neural Networks }, author = {Lucic, Ana and Ter Hoeve, Maartje A. and Tolomei, Gabriele and De Rijke, Maarten and Silvestri, Fabrizio}, booktitle = {Proceedings of The 25th International Conference on Artificial Intelligence and Statistics}, pages = {4499--4511}, year = {2022}, editor = {Camps-Valls, Gustau and Ruiz, Francisco J. R. and Valera, Isabel}, volume = {151}, series = {Proceedings of Machine Learning Research}, month = {28--30 Mar}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v151/lucic22a/lucic22a.pdf}, url = {https://proceedings.mlr.press/v151/lucic22a.html}, abstract = { Given the increasing promise of graph neural networks (GNNs) in real-world applications, several methods have been developed for explaining their predictions. Existing methods for interpreting predictions from GNNs have primarily focused on generating subgraphs that are especially relevant for a particular prediction. However, such methods are not counterfactual (CF) in nature: given a prediction, we want to understand how the prediction can be changed in order to achieve an alternative outcome. In this work, we propose a method for generating CF explanations for GNNs: the minimal perturbation to the input (graph) data such that the prediction changes. Using only edge deletions, we find that our method, CF-GNNExplainer, can generate CF explanations for the majority of instances across three widely used datasets for GNN explanations, while removing less than 3 edges on average, with at least $94%$ accuracy. This indicates that CF-GNNExplainer primarily removes edges that are crucial for the original predictions, resulting in minimal CF explanations. } }
Endnote
%0 Conference Paper %T CF-GNNExplainer: Counterfactual Explanations for Graph Neural Networks %A Ana Lucic %A Maartje A. Ter Hoeve %A Gabriele Tolomei %A Maarten De Rijke %A Fabrizio Silvestri %B Proceedings of The 25th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2022 %E Gustau Camps-Valls %E Francisco J. R. Ruiz %E Isabel Valera %F pmlr-v151-lucic22a %I PMLR %P 4499--4511 %U https://proceedings.mlr.press/v151/lucic22a.html %V 151 %X Given the increasing promise of graph neural networks (GNNs) in real-world applications, several methods have been developed for explaining their predictions. Existing methods for interpreting predictions from GNNs have primarily focused on generating subgraphs that are especially relevant for a particular prediction. However, such methods are not counterfactual (CF) in nature: given a prediction, we want to understand how the prediction can be changed in order to achieve an alternative outcome. In this work, we propose a method for generating CF explanations for GNNs: the minimal perturbation to the input (graph) data such that the prediction changes. Using only edge deletions, we find that our method, CF-GNNExplainer, can generate CF explanations for the majority of instances across three widely used datasets for GNN explanations, while removing less than 3 edges on average, with at least $94%$ accuracy. This indicates that CF-GNNExplainer primarily removes edges that are crucial for the original predictions, resulting in minimal CF explanations.
APA
Lucic, A., Ter Hoeve, M.A., Tolomei, G., De Rijke, M. & Silvestri, F.. (2022). CF-GNNExplainer: Counterfactual Explanations for Graph Neural Networks . Proceedings of The 25th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 151:4499-4511 Available from https://proceedings.mlr.press/v151/lucic22a.html.

Related Material