Re: [igraph] Re: Suggestion for a new community detection function

[Tamas Nepusz]

> first step. As far as I have understood, in a interactive way, both
> algorithms join communities together choosing among those with higher
> modularity value.
This is a little bit more complicated in the algorithm of  
Lancichinetti et al. They calculate a so-called "natural community"  
for every vertex (or to randomly selected vertices). The natural  
community of a vertex is grown from a community consisting of only the  
single vertex being considered. New vertices are added one by one to  
the community based on their fitness with respect to the current state  
of the community as it grows. (Sometimes they also delete vertices  
from the community if their fitness becomes negative, see the paper  
for details). After the natural community of a given vertex has been  
determined, they re-start the process from scratch with another vertex  
- so what they obtain is not a partition of the network into disjoint  
communities, but a union of natural communities that can overlap.

--
Tamas