Graph Capsule Convolutional Neural Networks

TL;DR

This paper introduces Graph Capsule Networks (GCAPS-CNN), a novel deep learning model that enhances graph classification by addressing weaknesses in existing GCNNs, demonstrating superior performance on benchmark datasets.

Contribution

The paper proposes a new Graph Capsule Network model that improves graph classification accuracy and overcomes limitations of current GCNN architectures.

Findings

GCAPS-CNN significantly outperforms existing methods on benchmark datasets.

The model effectively addresses weaknesses in traditional GCNNs.

Extensive experiments validate the superiority of GCAPS-CNN.

Abstract

Graph Convolutional Neural Networks (GCNNs) are the most recent exciting advancement in deep learning field and their applications are quickly spreading in multi-cross-domains including bioinformatics, chemoinformatics, social networks, natural language processing and computer vision. In this paper, we expose and tackle some of the basic weaknesses of a GCNN model with a capsule idea presented in \cite{hinton2011transforming} and propose our Graph Capsule Network (GCAPS-CNN) model. In addition, we design our GCAPS-CNN model to solve especially graph classification problem which current GCNN models find challenging. Through extensive experiments, we show that our proposed Graph Capsule Network can significantly outperforms both the existing state-of-art deep learning methods and graph kernels on graph classification benchmark datasets.