more options for multilevel.community

[sds]

feature requst: make multilevel.community accept parameters which tune the algorithm,
e.g., specify the epsilon (the min modularity improvements to continue).
also, the returned community structure should be hierarchical.
thanks!

[Francis Markham (fmarkham)]

To be more specific, the Louvian algorithm can be extended to accept a "tunable" resolution parameter. By modifying the resolution parameter, a different number of communities will be detected. This parameter is always set to 1.0 in the original paper[1] and thus in its iGraph implementation.

A tunable resolution parameter is important because of the so-called "resolution limit" problem with modularity optimisation [2]. In short, the resolution limit means that optimising the modularity function of a large network may result in the detection of fewer communities than in the best partitioning. Other software packages such as Gephi expose this parameter in their implementation of the Louvian algorithm.

More details about the resolution parameter which would suggest an implementation are available in a second paper about the Louvain algorithm [3].

[1] Fast unfolding of communities in large networks, V.D. Blondel, J.-L. Guillaume, R. Lambiotte and E. Lefebvre, J. Stat. Mech., (2008) P10008
[2] Fortunato, S. & Barth´elemy,M. Resolution limit in community detection. Proc. Natl. Acad. Sci. USA 104, 36 (2007).
[3] Laplacian Dynamics and Multiscale Modular Structure in Networks, R. Lambiotte, J.-C. Delvenne and M. Barahona, arXiv:0812.1770

[Gábor Csárdi (gabor.csardi)]

The development of igraph has moved to github, so please do not comment on this bug here. You are of course welcome to comment on github, here:
https://github.com/igraph/igraph/issues/463