Re: [igraph] Generalization of Leading Eigenvector community detection

[Gábor Csárdi]

You want to do this from C? In C we usually don't use weighted graphs,
instead pass the weights as a separate argument. So if you want to
modify the igraph implementation of the leading eigenvector community
detection function, add an additional 'weights' argument. (I guess you
already did that.)

Then you can calculate the sum of the vector by calling
igraph_vector_sum, see
http://cneurocvs.rmki.kfki.hu/igraph/doc/html/ch06s02s10.html#igraph_vector_sum

Btw. the igraph implementation of the LEV has some serious problems,
because the version of the ARPACK library we used does not yield
consistent results in some cases. I am trying to correct this, but it
is not easy and I don't want to go too deep into ARPACK.

Gabor

On Tue, Feb 3, 2009 at 2:31 AM, Al <address@hidden> wrote:
> I would first like to thank the list and Gabor for some of my earlier
> posts to the list while I got my dev environment sorted out.
> My colleague and I are modifying the Newman Leading Eigenvector method
> to accept weighted graphs. We have had a dialogue with Dr Newman to
> understand what needs to be changed in the modularity algorithm to
> accept weighted graphs. One of the changes that needs to be made, is
> instead of using the number of edges, we need the sum of all of the
> edge weights. I was wondering the best way using the graph object's
> exposed interfaces to sum all of the edge weights.
>
> Thanks
> Alex
>
>
> _______________________________________________
> igraph-help mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/igraph-help
>



-- 
Gabor Csardi <address@hidden>     UNIL DGM