GraphRNN Revisited: An Ablation Study and Extensions for Directed Acyclic Graphs
This work provides incremental insights into graph generation methods, particularly for directed acyclic graphs, benefiting researchers in graph machine learning.
The paper replicated GraphRNN's results and conducted an ablation study, finding that BFS traversal is crucial for performance, and extended it to generate directed acyclic graphs using topological sort, showing significant improvement over a baseline on a real-world dataset.
GraphRNN is a deep learning-based architecture proposed by You et al. for learning generative models for graphs. We replicate the results of You et al. using a reproduced implementation of the GraphRNN architecture and evaluate this against baseline models using new metrics. Through an ablation study, we find that the BFS traversal suggested by You et al. to collapse representations of isomorphic graphs contributes significantly to model performance. Additionally, we extend GraphRNN to generate directed acyclic graphs by replacing the BFS traversal with a topological sort. We demonstrate that this method improves significantly over a directed-multiclass variant of GraphRNN on a real-world dataset.