On Tue, Dec 22, 2009 at 5:59 AM, Tamas Nepusz
<address@hidden> wrote:
> After I reset the max iteration value, the program runs slowly....
It's no surprise, since the algorithm takes more steps in order to find an eigenvector.
> In order to find the best community division, should I use a loop to compute modularity for each step?
No, the algorithm does that automatically (i.e., it splits communities further and further as long as it can increase the modularity). Just set the steps value to the number of vertices.
By the way, I think that other community detection methods (i.e. igraph_community_fastgreedy) are faster and yield similar results, so if the leading eigenvector method is too slow for you, try another one.
--