Edge Contraction Pooling for Graph Neural Networks

TL;DR

This paper introduces EdgePool, a novel graph pooling layer based on edge contraction, which enhances GNN performance by enabling effective abstraction over node groups, outperforming existing methods.

Contribution

The paper presents a new edge contraction-based pooling layer for GNNs that is easy to integrate and improves classification accuracy.

Findings

EdgePool outperforms alternative pooling methods.

EdgePool improves performance on node and graph classification.

EdgePool is easily integrated into existing GNN models.

Abstract

Graph Neural Network (GNN) research has concentrated on improving convolutional layers, with little attention paid to developing graph pooling layers. Yet pooling layers can enable GNNs to reason over abstracted groups of nodes instead of single nodes. To close this gap, we propose a graph pooling layer relying on the notion of edge contraction: EdgePool learns a localized and sparse hard pooling transform. We show that EdgePool outperforms alternative pooling methods, can be easily integrated into most GNN models, and improves performance on both node and graph classification.