Divide the train/val/test dataset, remove the redundant edges of each part of the dataset, but keep all the edges of the auxiliary view to satisfy the inductive setting;
Remove all edges of the main view in the val/test dataset to meet the "Strict Cold Start" scenario.
python preprocess.py --data_path data/
Experiments
About
CoMGL: Collaborative Multi-view Graph Learning for Strict Cold-start Tasks