Capsule Graph Neural Networks with EM Routing

TL;DR

This paper introduces Capsule Graph Neural Networks with EM routing, enhancing graph classification by capturing complex part-whole relationships, and demonstrates superior performance over existing models on real-world datasets.

Contribution

The paper proposes a novel Capsule GNN architecture using EM routing to improve graph embedding quality for classification tasks.

Findings

CapsGNNEM outperforms nine state-of-the-art models

Effective capture of part-whole relationships in graphs

Improved graph classification accuracy

Abstract

To effectively classify graph instances, graph neural networks need to have the capability to capture the part-whole relationship existing in a graph. A capsule is a group of neurons representing complicated properties of entities, which has shown its advantages in traditional convolutional neural networks. This paper proposed novel Capsule Graph Neural Networks that use the EM routing mechanism (CapsGNNEM) to generate high-quality graph embeddings. Experimental results on a number of real-world graph datasets demonstrate that the proposed CapsGNNEM outperforms nine state-of-the-art models in graph classification tasks.