Re: [igraph] Community similarity / dissimilarity

[Tamás Nepusz]

> vc = g.community_infomap()
You could try the following:

g2 = g.copy()
g2.contract_vertices(vc.membership)
g2.es["weight"] = 1
g2.simplify(combine_edges="sum")

This would give you a graph where the nodes represent the communities of the 
original graph, the edges are weighted, and the weight of an edge going between 
node A and B is the number of edges connecting community A and B. 
g2.get_adjacency(attribute="weight") would then give you the weighted adjacency 
matrix of g2, which is essentially a similarity matrix between the communities. 
Of course it is very well conceivable that other similarity or distance 
measures make more sense, but this is a good starting point.

> My ultimate goal is to use this similarity matrix and multidimensional
> scaling to make a nice 2d layout of the discovered communities
You can use any layout algorithm on g2 (including g2.layout_mds for what it's 
worth) to obtain a layout.

Cheers,
Tamas