Toward the Analysis of Graph Neural Networks

TL;DR

This paper introduces a method to analyze Graph Neural Networks by converting them into Feed Forward Neural Networks, enabling the use of existing analysis techniques to better understand GNN behaviors.

Contribution

The paper proposes a novel approach to analyze GNNs through conversion to FFNNs, facilitating verification and property inference of GNNs.

Findings

Conversion method enables analysis of GNNs using FFNN tools

Scalability and accuracy of conversions are discussed and demonstrated

Case study on node classification validates the approach

Abstract

Graph Neural Networks (GNNs) have recently emerged as a robust framework for graph-structured data. They have been applied to many problems such as knowledge graph analysis, social networks recommendation, and even Covid19 detection and vaccine developments. However, unlike other deep neural networks such as Feed Forward Neural Networks (FFNNs), few analyses such as verification and property inferences exist, potentially due to dynamic behaviors of GNNs, which can take arbitrary graphs as input, whereas FFNNs which only take fixed size numerical vectors as inputs. This paper proposes an approach to analyze GNNs by converting them into FFNNs and reusing existing FFNNs analyses. We discuss various designs to ensure the scalability and accuracy of the conversions. We illustrate our method on a study case of node classification. We believe that our approach opens new research directions for understanding and analyzing GNNs.