Efficiency Comparison on SBM: Benchmark Results for Dynamic Graph Neural Networks

Evolving Graph Attention Networks (EGAT) are proposed for learning from dynamic graphs where nodes and edges change over time. The 5.5 KB XLS file, authored by Yucai Jiang and last updated in June 2026, likely contains performance metrics from experiments comparing EGAT to other models. The results demonstrate that the proposed model outperforms state-of-the-art baselines on benchmark datasets.

Use Cases

• Benchmarking new dynamic graph neural network architectures based on the reported efficiency comparison.
• Analyzing the performance of attention mechanisms over time based on the described weight-evolving paradigm.
• Studying model adaptation to topological changes based on the described temporal relational dynamics.

Strengths

• File size is precisely 5.5 KB, indicating a compact and focused results dataset.
• Dataset has a clear, specific purpose related to benchmarking a novel GNN framework.
• License is explicitly CC-BY-4.0, permitting broad reuse with attribution.

Limitations

• Row count and specific column definitions are unknown, limiting assessment of data scope and structure.
• Description metadata focuses on the model methodology rather than detailing the specific benchmark data content.
• The dataset's small size suggests it contains summary results, not the raw training data or graphs used in experiments.

Provenance

Source

Yucai Jiang via figshare

Collection Method

Likely generated from experiments comparing the proposed EGAT model to other baselines.

Freshness

Last updated 2026-06-01 17:40:57