Aggregating dynamic graph sequences representing temporal evolution patterns for the CrossLink link prediction framework. It provides structured data for training transformer-decoder models to perform conditioned link generation based on integrated evolution and structural modeling.
Use Cases
- Train a transformer-decoder for link prediction using the evolution and structure modeling components
- Implement conditioned link generation to predict future edges based on specific graph evolution patterns
- Benchmark dynamic graph algorithms on their ability to capture pattern-specific evolution
Strengths
- Features dynamic graph sequences for modeling temporal evolution patterns
- Optimized for conditioned link generation using transformer-decoder architectures
- Combines structural and evolutionary data points for link prediction tasks