GraphSVX: Shapley Value Explanations for Graph Neural Networks - Archive ouverte HAL Access content directly
Conference Papers Year : 2021

GraphSVX: Shapley Value Explanations for Graph Neural Networks

Abstract

Graph Neural Networks (GNNs) achieve significant performance for various learning tasks on geometric data due to the incorporation of graph structure into the learning of node representations, which renders their comprehension challenging. In this paper, we first propose a unified framework satisfied by most existing GNN explainers. Then, we introduce GraphSVX, a post hoc local model-agnostic explanation method specifically designed for GNNs. GraphSVX is a decomposition technique that captures the "fair" contribution of each feature and node towards the explained prediction by constructing a surrogate model on a perturbed dataset. It extends to graphs and ultimately provides as explanation the Shapley Values from game theory. Experiments on real-world and synthetic datasets demonstrate that GraphSVX achieves state-ofthe-art performance compared to baseline models while presenting core theoretical and human-centric properties.
Fichier principal
Vignette du fichier
GraphSVX_ECML_PKDD_2021.pdf (2.7 Mo) Télécharger le fichier
Origin Files produced by the author(s)

Dates and versions

hal-03539183 , version 1 (21-01-2022)

Identifiers

  • HAL Id : hal-03539183 , version 1

Cite

Alexandre Duval, Fragkiskos D. Malliaros. GraphSVX: Shapley Value Explanations for Graph Neural Networks. European Conference on Machine Learning and Knowledge Discovery in Databases (ECML PKDD), Sep 2021, Bilbao, Spain. ⟨hal-03539183⟩
142 View
190 Download

Share

Gmail Mastodon Facebook X LinkedIn More