pyg-team/pytorch_geometric

Node2Vec implementation is degree agnostic of the starting vertex

Open

#5,600 opened on 2022幎10月3日

GitHub で芋る
 (4 comments) (0 reactions) (0 assignees)Python (3,514 forks)batch import
1 - Priority P1featurehelp wantednn

Repository metrics

Stars
 (19,985 stars)
PR merge metrics
 (平均マヌゞ 35d 1h) (30d で 14 merged PRs)

説明

🚀 The feature, motivation and pitch

In the original Node2vec paper probability of picking a start, a node is proportional to its out degree. While looking at existing code in this project to me it appears that it is not in this case and the probability of picking a start node is uniformly distributed across all nodes.

Attaching ss from paper. If you want I can create a quick PR to fix this. It should be easy since we already take edge_index from the user. Not sure if this is a feature request or a bug because the original implementation of authors also considered weighted networks but I don't think that is the case here. So not sure if this was also left out on purpose. Please feel free to point out any gaps in my understanding of this code base or the paper itself.

Screenshot from 2022-10-03 12-40-11

Alternatives

No response

Additional context

No response

コントリビュヌタヌガむド