Re: [igraph] Community detection algorithm for (almost) full weighted graph

[Manuel J. Zetina-Rejón]

In this topic? What is best criterio to consider if a network is dense or sparse? I think is not easy to identify this frontier!

Manuel 

=================================
Dr. Manuel J.  Zetina-Rejón
Interdisciplinary Center of Marine Sciences National Polytechnic
Institute Marine Biology and Fisheries Department
Av. IPN S/N Col. Playa Palo de Sta. Rita La Paz, BCS, Mexico
AP 590
Tel. +52 (612) 1225344 Fax. +52(612) 1225322
http://www.cicimar.ipn.mx
=================================

El 24/10/2014, a las 10:57, Tamas Nepusz <address@hidden> escribió:

Then I want to apply a community detection algorithm which will evaluate
graph structure mainly on edges' weights.

All the community detection algorithms in igraph are designed for sparse
networks, so if you want to use any of them, you have to convert your matrix to
a sparse one first. This can be done in multiple ways; e.g., you could drop the
connections below a certain weight, or you could keep the best (i.e. highest
weight) connections for each of your vertices. Then you can use whichever
algorithm you prefer from igraph.

If you want to keep your dense graph, you could try MCL [1] or spectral
clustering [2], but these are not implemented in igraph.

[1]: http://micans.org/mcl
[2]: http://en.wikipedia.org/wiki/Spectral_clustering

T.

_______________________________________________
igraph-help mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/igraph-help