Loading [MathJax]/extensions/MathZoom.js
Reinforced Causal Explainer for Graph Neural Networks | IEEE Journals & Magazine | IEEE Xplore

Reinforced Causal Explainer for Graph Neural Networks


Abstract:

Explainability is crucial for probing graph neural networks (GNNs), answering questions like “Why the GNN model makes a certain prediction?”. Feature attribution is a pre...Show More

Abstract:

Explainability is crucial for probing graph neural networks (GNNs), answering questions like “Why the GNN model makes a certain prediction?”. Feature attribution is a prevalent technique of highlighting the explanatory subgraph in the input graph, which plausibly leads the GNN model to make its prediction. Various attribution methods have been proposed to exploit gradient-like or attention scores as the attributions of edges, then select the salient edges with top attribution scores as the explanation. However, most of these works make an untenable assumption — the selected edges are linearly independent — thus leaving the dependencies among edges largely unexplored, especially their coalition effect. We demonstrate unambiguous drawbacks of this assumption — making the explanatory subgraph unfaithful and verbose. To address this challenge, we propose a reinforcement learning agent, Reinforced Causal Explainer (RC-Explainer). It frames the explanation task as a sequential decision process — an explanatory subgraph is successively constructed by adding a salient edge to connect the previously selected subgraph. Technically, its policy network predicts the action of edge addition, and gets a reward that quantifies the action’s causal effect on the prediction. Such reward accounts for the dependency of the newly-added edge and the previously-added edges, thus reflecting whether they collaborate together and form a coalition to pursue better explanations. It is trained via policy gradient to optimize the reward stream of edge sequences. As such, RC-Explainer is able to generate faithful and concise explanations, and has a better generalization power to unseen graphs. When explaining different GNNs on three graph classification datasets, RC-Explainer achieves better or comparable performance to state-of-the-art approaches w.r.t. two quantitative metrics: predictive accuracy, contrastivity, and safely passes sanity checks and visual inspections. Codes and datasets are avai...
Published in: IEEE Transactions on Pattern Analysis and Machine Intelligence ( Volume: 45, Issue: 2, 01 February 2023)
Page(s): 2297 - 2309
Date of Publication: 26 April 2022

ISSN Information:

PubMed ID: 35471869

Funding Agency:


1 Introduction

Graph neural networks (GNNs) [1], [2] have exhibited impressive performance in a variety of tasks, where the data are graph-structured. Their success comes mainly from the powerful representation learning, which incorporates graph structure in an end-to-end fashion. Alongside performance, explainability becomes central to the practical impact of GNNs, especially in real-world applications on fairness, security, and robustness [3], [4], [5]. Aiming to answer questions like “Why the GNN model made a certain prediction?,” we focus on post-hoc [6], local [7], model-agnostic [8] explainability — that is, considering the target GNN model as a black-box (i.e., post-hoc), an explainer interprets its predictions of individual instances (i.e., local), which is applicable to any GNNs (i.e., model-agnostic). Towards this end, a prevalent paradigm is feature attribution and selection [9], [10], [11], [12]. Typically, given an input graph, it distributes the model’s outcome prediction to the input features (i.e., edges), and then selects the salient substructure (i.e., the subset of edges) as an explanatory subgraph. Such an explanatory subgraph is expected to provide insight into the model workings.

References

References is not available for this document.