Cooker: Self-supervised Adaptive Aggregator Learning on Graph

Paper

Self-supervised Adaptive Aggregator Learning on Graph, Pacific-Asia Conference on Knowledge Discovery and Data Mining, 2021

Please cite this paper.

@inproceedings{lin2021self,
 title={Self-supervised Adaptive Aggregator Learning on Graph},
 author={Lin, Bei and Luo, Binli and He, Jiaojiao and Gui, Ning},
 booktitle={Pacific-Asia Conference on Knowledge Discovery and Data Mining},
 pages={29--41},
 year={2021},
 organization={Springer}
}

Usage

Here we provide an implementation of Cooker in Python, along with a minimal execution example (on the Cora dataset). The repository is organised as follows:

• input/ contains the necessary dataset files for Cora;
• model.py contains the implementation of the Cooker pipeline and a random walk layer;
• utils.py contains the necessary processing subroutines.
• main.py puts all of the above together and may be used to execute a full training run on Cora.

Installation

• Requirement
  • Python==3.6
  • numpy==1.18.5
  • pandas==1.0.4
  • scikit-learn==0.23.1
  • tensorflow==2.2.0
  • networkx==2.4