GraphRNN Revisited

An Ablation Study and Extensions for Directed Acyclic Graphs

This project involved reproducing the GraphRNN model, achieving comparable performance in generating graphs on both qualitative and quantitative metrics. We also extended the paper by assessing model performance on additional graph similarity metrics, confirming the utility of the breadth-first search traversal with an ablation study, extending GraphRNN to directed graphs, and implementing a novel method to generate directed acyclic graphs.

This project was done for a course project for the Oxford MSc in Advanced Computer Science. It was accepted to the NeurIPS 2023 GLFrontiers workshop as well.

Paper link

Code link

References

2023

  1. GraphRNN Revisited: An Ablation Study and Extensions for Directed Acyclic Graphs
    M. Ravichandran, M. Koch, T. Das, and 1 more author
    In NeurIPS 2023: New Frontiers in Graph Learning Workshop, 2023