ftheberge
commented
4 years ago -
you are quite right that inverting those would correspond to the paper, thanks, I will make the change! In practice, when looking for communities, I would always take the 2-core anyways, and Louvain will tend to cluster tendrils (1-core trees) with its root anyways, so this is more of a 'marker'.
-
you can certainly experiment with other algorithms for the last step. I did for example try InfoMap as the last step.
best, F. Theberge
Le ven. 1 nov. 2019, à 11 h 11, Karlis Kanders notifications@github.com a écrit :
Thank you very much for publishing your code!
1.
I wanted to ask if there is any significance to the fact that in the community_ecg() method, the forcing of the weights that are outside the 2-core, to the min_weight happens after the final clustering step that yields the consensus partition? From reading your paper, it seems that the forcing should come before the final clustering step... 2.
With regard to the final clustering step that yields the consensus partition, would you perhaps advise to perform not just one multilevel Louvain clustering but do in fact another consensus clustering run with the newly obtained weights? This might help to further reduce the ambiguity regarding the final partition.
Thank you!
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/ftheberge/Ensemble-Clustering-for-Graphs/issues/1?email_source=notifications&email_token=AKFEKTABJHKFS4LGHOGZDFLQRRBLRA5CNFSM4JH4CHSKYY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4HWEFTPA, or unsubscribe https://github.com/notifications/unsubscribe-auth/AKFEKTDTD5DPO2EPD2TWRULQRRBLRANCNFSM4JH4CHSA .
beingkk
Thank you very much for publishing your code!
1) I wanted to ask if there is any significance to the fact that in the community_ecg() method, the forcing of the weights that are outside the 2-core, to the min_weight happens after the final clustering step that yields the consensus partition? From reading your paper, it seems that the forcing should come before the final clustering step...
2) With regard to the final clustering step that yields the consensus partition, would you perhaps advise to perform not just one multilevel Louvain clustering but do in fact another consensus clustering run with the newly obtained weights? This might help to further reduce the ambiguity regarding the final partition.
Thank you!