MotifNet: a motif-based Graph Convolutional Network for directed graphs

TL;DR

MotifNet introduces a novel graph convolutional neural network that effectively handles directed graphs by leveraging local graph motifs, overcoming limitations of traditional spectral methods that assume undirected graphs.

Contribution

This work presents MotifNet, a new graph CNN architecture that incorporates local motifs to process directed graphs, expanding the applicability of graph neural networks.

Findings

MotifNet outperforms existing methods on real directed graph data.

The approach effectively captures directionality in graph structures.

Experimental results demonstrate improved accuracy over undirected graph models.

Abstract

Deep learning on graphs and in particular, graph convolutional neural networks, have recently attracted significant attention in the machine learning community. Many of such techniques explore the analogy between the graph Laplacian eigenvectors and the classical Fourier basis, allowing to formulate the convolution as a multiplication in the spectral domain. One of the key drawback of spectral CNNs is their explicit assumption of an undirected graph, leading to a symmetric Laplacian matrix with orthogonal eigendecomposition. In this work we propose MotifNet, a graph CNN capable of dealing with directed graphs by exploiting local graph motifs. We present experimental evidence showing the advantage of our approach on real data.