Betweenness centrality with parallel edges

Hi,

I remember asking some time ago whether graph-tool computed BC with parallel
edges and Tiago pointing me to the BGL interface.

The answer is the BGL does take in account parallel edges in the computation
of the number of shortest paths. However, the normalization factor in
graph-tool becomes meaningless when networks have parallel edges as the star
is no longer the benchmark network.

Tiago, would it be possible to include a flag in graph-tool so we can choose
to compute normalized/non-normalized betweeness?

Many thanks,

Rui

attachment.html (2.42 KB)

Hello Rui,

Hi Tiago,

Thanks for considering it.

Until you implement it, my understanding is that to get absolute betweenness values we need to multiply node/edge betweenness by the following factors:
Node betweenness: (N-1)*(N-2)/2
Edge betweenness: N*(N-1)/2

Is this correct?

Cheers,
Rui

P.S.: good luck with the post-doc work :wink:

Hi Rui,

Until you implement it, my understanding is that to get absolute
betweenness values we need to multiply node/edge betweenness by the
following factors:
Node betweenness: (N-1)*(N-2)/2
Edge betweenness: N*(N-1)/2

Is this correct?

Yes, this is correct for undirected graphs. For directed graphs it
becomes (n-2)*(n-1) and n*(n-1), for vertex and edge betweenness
respectively.

P.S.: good luck with the post-doc work :wink:

Thanks. :wink:

Cheers,
Tiago