Thanks again! It does make sense but I think it gives more credit to the
disconnected nodes that it really should. If we comparing the closeness of
a connected node to another that's disconnected, it will not be a fair
comparison. Is it acceptable to multiply the returned closeness value by a
very small number as a workaround?

CONST = 9999999999.99
finalCloseness = igraphCloseness * (1/CONST)

Does this sound right?

-Ahmed




On Fri, Dec 6, 2013 at 5:24 PM, Tamás Nepusz <[email protected]> wrote:

> > I misspoke, I meant closeness
>
> Closeness is different; an isolated vertex will have a closeness
> centrality of 1/n where n is the number of vertices. This is due to how
> igraph defines closeness centrality for disconnected graphs: when two nodes
> are disconnected, igraph assumes that their distance is n because we cannot
> use infinity for obvious reasons, and n is always larger than the longest
> possible geodesic in a network so it sort of makes sense.
>
> T.
>
_______________________________________________
igraph-help mailing list
[email protected]
https://lists.nongnu.org/mailman/listinfo/igraph-help

Reply via email to