This is the final project of MATH8013.
- How does #conv-layers affect performance?
- How dows Batch Normalization affect performance?
- How does the aggregation method affect performance?
- How does the normalized A affect performance?
- How does the order of Linear and A affect performance?
- pytorch 1.9.0
- pyg 2.0.3 (pytorch geometric)
- torchmetrics 0.7.3
- seaborn 0.11.2
- matplotlib 3.5.1
We focus on three semi-supervised node classification tasks.
- CiteSeer
- Cora
- PubMed
Simply use bash run.sh
to produce 54 experiments.
Check core/visual.ipynb