Re: [igraph] betweenness centrality -- negative and floating values?

[Tamas Nepusz]

Hi Yong,

I tested your graph and it looks like the issue is platform-dependent;  
e.g., it doesn't happen in openSuSE 11.1 (gcc 4.3.2, igraph 0.5.2),  
but I can reproduce it with OS X (gcc 4.0.1, igraph 0.5.2 and igraph  
0.6). I suspect some kind of an internal overflow problem, but it  
takes more time to track down the bug than I expected. I filed a bug  
report on Launchpad; if you want to get notified when there's some  
progress, subscribe to the bug report here:

https://bugs.launchpad.net/igraph/+bug/385747

--
Tamas

On 2009.06.10., at 11:43, Yong Zou wrote:

> Hi Tamas,
>
> Could you please help me to understand those extremely big values for
> the betweenness centrality?
> 58743768046350440071168.000000 for a network of 10000 nodes.
>
> Thank you very much.
> Yong
>
> On Thu, May 28, 2009 at 1:38 PM, Yong Zou<address@hidden>  
> wrote:
> > Thanks Tamas,
> >
> > Now it is clear for those floating values. But what about the very  
> > big
> > or even negative values, then?
> >
> > Best,
> > Yong
> >
> > On Thu, May 28, 2009 at 1:33 PM, Tamas Nepusz <address@hidden>  
> > wrote:
> > > > Ok, I can understand the floating values for the BC then because of
> > > > this weight of 1/k. It also means that the more shortest paths  
> > > > between
> > > > A and B, the less weight will be counted.
> > > >
> > > > However from the original definition for BC,
> > > > http://en.wikipedia.org/wiki/Centrality, this sounds to be  
> > > > different
> > > > for me.
> > >
> > > No, it isn't. The formula on that Wikipedia page has the number of  
> > > shortest
> > > paths between s and t in the denumerator, so if there is more than  
> > > a single
> > > shortest path between s and t, the vertices participating in those  
> > > paths
> > > will receive a contribution of 1/k due to this path, where k is  
> > > \sigma_{s,t}
> > > - the number of shortest paths between s and t.
> > >
> > > > I'm quite confused by this now. The values should be
> > > > normalized to the interval [0 1]?
> > >
> > > It can be normalized if you wish, but igraph doesn't do that by  
> > > default. As
> > > it says on the Wikipedia page you mentioned, "This may be  
> > > normalised by
> > > dividing through the number of pairs of vertices not including v,  
> > > which is
> > > (n - 1)(n - 2) / 2.". So if you want to get values between 0 and  
> > > 1, divide
> > > the betweenness centrality values by (n-1)(n-2)/2.
> > >
> > > --
> > > Tamas
> > >
> > > _______________________________________________
> > > igraph-help mailing list
> > > address@hidden
> > > http://lists.nongnu.org/mailman/listinfo/igraph-help
>
>
> _______________________________________________
> igraph-help mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/igraph-help