gephi.team team mailing list archive
-
gephi.team team
-
Mailing list archive
-
Message #01509
Re: [Question #147488]: modularity optimisation method?
Question #147488 on Gephi changed:
https://answers.launchpad.net/gephi/+question/147488
Gautier Krings gave more information on the question:
Sure,
I have an undirected graph composed of a clique of 16 nodes on one hand, and 4 cliques of 4 nodes on the other hand.
Every pair of cliques of 4 nodes are linked together by 2 random edges.
Finally, there is one edge from one random node of the large clique to one random node taken in the small cliques.
This is a case where the resolution limit of modularity occurs; the
"intuitive" partition should give 5 communities: the large clique and
the 4 small cliques. However, the modularity of this partition is
0.3146, and there exists a better partition in two communities: the
large clique on one hand, and the 4 small cliques grouped together on
the other hand. This partition has a modularity of 0.3505.
The code provided online (matlab or python) finds the optimal partition
(2 communities) while gephi returns either a partition in 4 or 5
communities.
Here's the dataset:
1 clique of 16 nodes (IDs: 0 to 15)
1 clique of 4 nodes (IDs: 16 to 19)
1 clique of 4 nodes (IDs: 20 to 23)
1 clique of 4 nodes (IDs: 24 to 27)
1 clique of 4 nodes (IDs: 28 to 31)
inter-clique edges:
16 22
17 20
17 28
18 26
18 28
19 24
21 25
21 30
22 27
23 30
24 29
25 29
4 22
I can share a gephi file if needed, if you tell me how to share a file.
--
You received this question notification because you are a member of
Gephi Team, which is an answer contact for Gephi.