Simple and Deep Graph Convolutional Networks

Ming Chen, Zhewei Wei, Zengfeng Huang, Bolin Ding, Yaliang Li
7/4/2020
cs.LGstat.ML

Official code: chennnM/GCNII (366 stars, last commit 2020-07-09).

Abstract

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 are shallow, due to the {\em over-smoothing} problem. In this paper, we study the problem of designing and analyzing deep graph convolutional networks. We propose the GCNII, an extension of the vanilla GCN model with two simple yet effective techniques: {\em Initial residual} and {\em Identity mapping}. We provide theoretical and empirical evidence that the two techniques effectively relieves the problem of over-smoothing. Our experiments show that the deep GCNII model outperforms the state-of-the-art methods on various semi- and full-supervised tasks. Code is available at https://github.com/chennnM/GCNII .

View on arXivView PDF

Code Implementations(1)

chennnM/GCNIIOfficial100%

PyTorch implementation of "Simple and Deep Graph Convolutional Networks"

36673Shell, PythonJun 10, 20206 years ago

Similar Papers

Transforming PageRank into an Infinite-Depth Graph Neural Network

86% match1 repo102022

Towards Deeper Graph Neural Networks

85% match1 repo682020

Graph Neural Networks Do Not Always Oversmooth

85% match1 repo2024

Towards a deeper GCN: Alleviate over-smoothing with iterative training and fine-tuning

84% match1 repo12025

Beyond Over-smoothing: Uncovering the Trainability Challenges in Deep Graph Neural Networks

84% match2024

DRGCN: Dynamic Evolving Initial Residual for Deep Graph Convolutional Networks

84% match1 repo142023

Cite this paper

@article{chen2020simple,
  title  = {Simple and Deep Graph Convolutional Networks},
  author = {Ming Chen and Zhewei Wei and Zengfeng Huang and Bolin Ding and Yaliang Li},
  year   = {2020},
  eprint = {2007.02133},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/2007.02133}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2007.02133.svg)](https://codeofpaper.com/paper/2007.02133)

Discussion