Rank-based self-training for graph convolutional networks

作者：

Highlights：

• A novel Rank-Based Self-Training approach for Graph Convolutional Networks.

• A margin-based score proposed to select predictions as pseudo-labeled data.

• The approach is not restricted to a specific GCN model and can used to combine models.

• Evaluation on semi-supervised classification considering four GCN models.

• Results on representative benchmarks and comparison with state-of-the-art methods.

摘要

•A novel Rank-Based Self-Training approach for Graph Convolutional Networks.•A margin-based score proposed to select predictions as pseudo-labeled data.•The approach is not restricted to a specific GCN model and can used to combine models.•Evaluation on semi-supervised classification considering four GCN models.•Results on representative benchmarks and comparison with state-of-the-art methods.

论文关键词：Graph convolutional networks,Self-training,Rank model,Semi-supervised learning

论文评审过程：Received 31 July 2020, Revised 30 October 2020, Accepted 22 November 2020, Available online 1 December 2020, Version of Record 1 December 2020.

论文官网地址：https://doi.org/10.1016/j.ipm.2020.102443