Skip to content

Latest commit

 

History

History
11 lines (6 loc) · 399 Bytes

README.md

File metadata and controls

11 lines (6 loc) · 399 Bytes

EFGNN

For a quick demo run best_model_for('Squirrel') or best_model_for('Chameleon') from best_model.py

The model's code is in EFGNN.py

MultiHopTransform.py just augments the data with some precomputed stuff, like CSR adjacencies, etc.

train.py contains a train function

AttLayer adds a learnable attention-based filter, but it doesn't seem to bring any performance benefits at this point.