Relational Pooling for Graph Representations

TL;DR

Relational Pooling (RP) is a novel framework that enhances graph neural networks' representation power, surpassing traditional methods like WL tests, and enables the use of diverse neural architectures for graph classification.

Contribution

This paper introduces Relational Pooling, a new framework that generalizes GNNs beyond existing spectral and combinatorial methods, increasing their expressive capacity.

Findings

RP outperforms state-of-the-art methods on several graph tasks.

RP can incorporate architectures like RNNs and CNNs for graph classification.

RP provides a theoretically sound foundation for more powerful graph representations.

Abstract

This work generalizes graph neural networks (GNNs) beyond those based on the Weisfeiler-Lehman (WL) algorithm, graph Laplacians, and diffusions. Our approach, denoted Relational Pooling (RP), draws from the theory of finite partial exchangeability to provide a framework with maximal representation power for graphs. RP can work with existing graph representation models and, somewhat counterintuitively, can make them even more powerful than the original WL isomorphism test. Additionally, RP allows architectures like Recurrent Neural Networks and Convolutional Neural Networks to be used in a theoretically sound approach for graph classification. We demonstrate improved performance of RP-based graph representations over state-of-the-art methods on a number of tasks.