Re: [igraph] community detection

[Sam Steingold]

> * Gábor Csárdi <address@hidden> [2012-08-15 15:23:08 -0400]:
>
> Maybe you don't want community detection at all, but just some kind of
> clustering with a given number of clusters. If you can define some
> meaningful distance measure for the vertices,

the obvious distance is the resistance one.

> then you can run k-means clustering.

cool! is there a function for that?
(kmeans requires a numeric dataset, i.e., it works only in R^n).

> Another option is to run the fast-greedy algorithm in igraph and cut
> the merge dendrogram at the desired number of clusters. This does not
> ensure equal cluster sizes, though.

will try.
thanks.
(btw, why does multilevel.community return a non-hierarchical community
structure?)


> Btw. if you don't care much about the graph structure in your division
> (= you don't care about modularity), then why don't you just create
> subgraphs of equal sizes?

well, I _do_ need the subgraph elements to be "similar" to each other.

-- 
Sam Steingold (http://sds.podval.org/) on Ubuntu 12.04 (precise) X 11.0.11103000
http://www.childpsy.net/ http://camera.org http://dhimmi.com http://memri.org
http://thereligionofpeace.com http://mideasttruth.com
main(a){a="main(a){a=%c%s%c;printf(a,34,a,34);}";printf(a,34,a,34);}