GraphLIME: Local Interpretable Model Explanations for Graph Neural Networks

By Qiang Huang, Makoto Yamada, Yuan Tian, Dinesh Singh, Dawei Yin, Yi Chang. IEEE Transactions on Knowledge and Data Engineering, 2020

484 citations recorded by the source corpus.

Recently, graph neural networks (GNN) were shown to be successful in effectively representing graph structured data because of their good performance and generalization ability. However, explaining the effectiveness of GNN models is a challenging task because of the complex nonlinear transformations made over the iterations. In this paper, we propose GraphLIME, a local interpretable model explanation for graphs using the Hilbert-Schmidt Independence Criterion (HSIC) Lasso, which is a nonlinear feature selection method. GraphLIME is a generic GNN-model explanation framework that learns a nonlinear interpretable model locally in the subgraph of the node being explained. Through experiments on two real-world datasets, the explanations of GraphLIME are found to be of extraordinary degree and more descriptive in comparison to the existing explanation methods.

Subfields: Artificial Intelligence

View Paper: https://www.semanticscholar.org/paper/8819adc0f064d5293bd2b783a63e23f68109738a

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

On Explainability of Graph Neural Networks via Subgraph Explorations (2021, methodology): “In addition, existing methods (Huang et al., 2020; Vu & Thai, 2020) employ a simple and interpretable model as the surrogate method to capture local relationships of deep graph models around the input data.”
Interpretable and Generalizable Graph Learning via Stochastic Attention Mechanism (2022, methodology): “Other methods including PGM-Explainer (Vu & Thai, 2020) leveraging graphical models, Gem (Lin et al., 2021) checking Granger causality and Graphlime (Huang et al., 2020) using HSIC Lasso are only applied to node-level task interpretation.”
SurvNAM: The machine learning survival model explanation (2021, methodology): “Its success and simplicity motivated to develop several modifications, for example, ALIME [23], NormLIME [24], DLIME [25], Anchor LIME [26], LIME-SUP [27], LIME-Aleph [28], GraphLIME [29].”
A Perspective on Explainable Artificial Intelligence Methods: SHAP and LIME (2023, methodology): “Finally, GraphLIME [27] was proposed as an updated version of LIME to explain graph-based models where non-linear association is more appropriately considered.”
Graph Neural Networks in Network Neuroscience (2021, methodology): “GNNExplainer and GraphLIME [77], [78] aimed to explain a graph by extracting the crucial features in the graph that affect the prediction of the GNN.”
Explainability in Graph Neural Networks: A Taxonomic Survey (2020, methodology): “In addition, for methods like GNNExplainer [46] and GraphLime [61], the importance scores are continuous values, then the importance map mi can be obtained by normalization and thresholding or rank-”
Higher-Order Explanations of Graph Neural Networks via Relevant Walks (2020, methodology): “Other recently proposed methods thatmap theGNNprediction to graph substructures includeTrap2 [33] andGraphLime [34].”
Interpretable deep learning: interpretation, interpretability, trustworthiness, and beyond (2021, methodology): “Recently, more researches focus on the interpretations of GNN models, such as GraphLIME [76], CoGE [51], Counterfactual explanations on GNNs [18] and others [20, 111, 132].”
Survey of Explainable AI Techniques in Healthcare (2023, methodology): “It extends LIME to work in a non-linear environment by sampling N-hop network neighbors and using the Hilbert–Schmidt independence criterion Lasso (HSIC Lasso) as surrogate models [35].”