Graph Recurrent Neural Networks (GraphRNN)

Reproducing results from the paper "GraphRNN: Generating Realistic Graphs with Deep Auto-regressive Model".

Requirements

The following Python libraries are required:

• pytorch >= 1.10
• tensorboard >= 2.8
• networkx >= 2.6.3
• pyyaml >= 6.0
• pyemd >= 0.5.1

Extensions

• We added additional evaluation metrics:
  • Betweenness Centrality
  • Degree Centrality
  • Density
  • Triadic Closure
• We added support for generating directed graphs and a special mode for generating DAGs

Acknowledgement

The authors' original implementation can be found here.