Multiple dynamic graph sequences used for the CrossLink paper to model graph evolution and link prediction. It provides the structural and temporal data necessary for conditioned link generation using transformer-decoder architectures.
Use Cases
- Train pattern-specific link prediction models using the graph evolution sequences
- Implement conditioned link generation by integrating structure and evolution modeling features
- Benchmark transformer-decoder performance on dynamic graph datasets
Strengths
- Raw dynamic graph data for temporal evolution modeling
- Specifically formatted for the CrossLink framework's conditioned link generation task
- Includes data suitable for transformer-decoder architectures in graph-based machine learning