Every Node Counts: Self-Ensembling Graph Convolutional Networks for Semi-Supervised Learning

TL;DR

This paper introduces Self-Ensembling GCN (SEGCN), a semi-supervised learning framework that enhances graph convolutional networks by leveraging a teacher-student model to utilize unlabeled data more effectively, improving classification accuracy.

Contribution

The paper proposes SEGCN, combining GCN with Mean Teacher to better exploit unlabeled data in semi-supervised learning for graph tasks, which was not addressed by prior GCN models.

Findings

SEGCN achieves state-of-the-art accuracy on Citeseer, Cora, and Pubmed datasets.

The method effectively utilizes unlabeled data through mutual teacher-student training.

SEGCN improves robustness under high dropout and graph collapse scenarios.

Abstract

Graph convolutional network (GCN) provides a powerful means for graph-based semi-supervised tasks. However, as a localized first-order approximation of spectral graph convolution, the classic GCN can not take full advantage of unlabeled data, especially when the unlabeled node is far from labeled ones. To capitalize on the information from unlabeled nodes to boost the training for GCN, we propose a novel framework named Self-Ensembling GCN (SEGCN), which marries GCN with Mean Teacher - another powerful model in semi-supervised learning. SEGCN contains a student model and a teacher model. As a student, it not only learns to correctly classify the labeled nodes, but also tries to be consistent with the teacher on unlabeled nodes in more challenging situations, such as a high dropout rate and graph collapse. As a teacher, it averages the student model weights and generates more accurate predictions to lead the student. In such a mutual-promoting process, both labeled and unlabeled samples can be fully utilized for backpropagating effective gradients to train GCN. In three article classification tasks, i.e. Citeseer, Cora and Pubmed, we validate that the proposed method matches the state of the arts in the classification accuracy.