find community classes/labels

Hi,

I am trying to use graph tool to extract communities, and I used the
minimize_blockmodel_dl function. However, after it has been executed
successfully, I wish to know each vertex and information of which community
it belongs to, and I can not figure out how.
For example,
state = graph_tool.community.minimize_blockmodel_dl(g, parallel=True)
##print state.b
print state.get_blocks().get_array()

it give me an array of all the vertex in zero. Could any one point me to the
right directions. Thank you very much.

Best Regards
Junjun

What you tried to do is correct. If the values of state.b are all zero,
it means that the best fit is a block model with only one group,
i.e. your graph looks like a random graph.

If you want to force a minimum number of groups (and risk overfitting),
you can specify the "min_B" option.

Best,
Tiago

Dear Tiago,

Many thanks for the information. Yes, you are right, after intentionally add
extra vertices, which should belong to other "communities", the function
works and provide difference community numbers in the output.

I have another question: Does graph tool provide support for community
detection on directed weighted graphs? I can see "eweight" can be provided
for minimize_blockmodel_dl, but it is only for "multigraph". Would you
please help confirm on this. Thanks again.

Best Regards
Junjun

Real-valued weights are not supported, but it is currently being
implemented and will be available soon.

Best,
Tiago

Hi

I'm looking for a way to find communities in directed and (real-valued)
weighted graphs and I just found this message in the mailing list. Is there
any update on this?

Regards,

I have another question: Does graph tool provide support for community
detection on directed weighted graphs? I can see "eweight" can be provided
for minimize_blockmodel_dl, but it is only for "multigraph". Would you
please help confirm on this. Thanks again.

Real-valued weights are not supported, but it is currently being
implemented and will be available soon.

Best,
Tiago

Support for real-valued weights is not yet done, but will be included
very soon.

Best,
Tiago