A question about the sub-community

[backbp]

I have a question about the sub-community. The statistics of the 10 sub-communities (table 10 in the paper) indicate the edges of each community are different, but I could not find any information about how to sample edges from the global graph.
Looking forward to your reply.

[BunsenFeng]

@whr000001 can you help with this question? I think its just: if A and B are both in this community, then an edge between node A and b counts as an edge of this community. But Herun can you help confirm?

[whr000001]

As @BunsenFeng said, if user A and user B are both in this community, then a 'following' or 'follower' edge between A and B is an edge of this community. Besides, other types of edges related to the community users are samples if the specific model needs them. For example, the tweet node and 'post' edge are sampled for BotRGCN.

[backbp]

I count the number of edges for each community as the above method, but the result has a big gap with the result in the paper (such as i get 375 edges in community 0, but it's 1,116,208 in paper). I'm not sure if there is any problem, and which will the result of Generalization Study in session 4.4.

[Skyorca]

Have you solved the problem? I'm also confused about how to build these 10 sub-communities

[lucas0716]

I count the number of edges for each community as the above method, but the result has a big gap with the result in the paper (such as i get 375 edges in community 0, but it's 1,116,208 in paper). I'm not sure if there is any problem, and which will the result of Generalization Study in session 4.4.

I meet the same problem. The number of edges among users in community0 is only 300+.