Simple and Deep Graph Convolutional Networks

• DOI: 10.48550/arxiv.2007.02133
• Publication Date: 2020-01-01
• Authors: Ming Chen; Zhewei Wei; Zengfeng Huang; Bolin Ding; Yaliang Li
• Keywords: Smoothing; Code (set theory)

Graph convolutional networks (GCNs) are a powerful deep learning approach for graph-structured data. Recently, GCNs and subsequent variants have shown superior performance in various application areas on real-world datasets. Despite their success, most of the current GCN models shallow, due to {\em over-smoothing} problem. In this paper, we study problem designing analyzing graph networks. We propose GCNII, an extension vanilla model with two simple yet effective techniques: Initial residual} Identity mapping}. provide theoretical empirical evidence that techniques effectively relieves over-smoothing. Our experiments show GCNII outperforms state-of-the-art methods semi- full-supervised tasks. Code is available at https://github.com/chennnM/GCNII .