Fusion Graph Convolutional Networks

TL;DR

This paper analyzes the limitations of current graph convolutional networks in capturing multi-hop neighborhood information and proposes an improved model, F-GCN, which enhances representation capacity and outperforms existing methods on multiple datasets.

Contribution

The paper identifies the limited multi-hop representation capacity of existing models and introduces a simple, mathematically motivated extension to GCNs, improving their effectiveness.

Findings

F-GCN outperforms state-of-the-art models on six datasets.

F-GCN is highly competitive on two additional datasets.

Analysis reveals existing models have limited multi-hop information capture.

Abstract

Semi-supervised node classification in attributed graphs, i.e., graphs with node features, involves learning to classify unlabeled nodes given a partially labeled graph. Label predictions are made by jointly modeling the node and its' neighborhood features. State-of-the-art models for node classification on such attributed graphs use differentiable recursive functions that enable aggregation and filtering of neighborhood information from multiple hops. In this work, we analyze the representation capacity of these models to regulate information from multiple hops independently. From our analysis, we conclude that these models despite being powerful, have limited representation capacity to capture multi-hop neighborhood information effectively. Further, we also propose a mathematically motivated, yet simple extension to existing graph convolutional networks (GCNs) which has improved representation capacity. We extensively evaluate the proposed model, F-GCN on eight popular datasets from different domains. F-GCN outperforms the state-of-the-art models for semi-supervised learning on six datasets while being extremely competitive on the other two.