Re: [igraph] edge betweenness after scaling edge weights

[Gábor Csárdi]

My guess is that this a floating point rounding error bug in igraph. You cannot represent 0.1 exactly on a computer (well, on most computers), and this causes some problems. You also need multiple shortest paths between some pairs for this to be a problem, which you do have I guess.

I am not sure how to fix this, though. I could probably fix this single bug, but there are several similar bugs in igraph, I am sure. 

G.

On Tue, May 28, 2013 at 6:24 PM, Guido Scarabelli <address@hidden> wrote:

Hi everyone!

I am using igraph version 0.6.5-2 with R (version 2.15.0) on a Linux machine.

My question is: is it possible that the betweenness of the edges change if the weights of the network are scaled all by the same factor?

In my case I started with a file containing the following adjacency matrix:

0  1  0  0  0
1  0  1  1  0
0  1  0  0  1
0  1  0  0  1
0  0  1  1  0

and in the R session I used these commands to build the network (it has 5 nodes and 5 edges):

library(igraph)
test.network <- data.matrix(read.table("matrix.dat"))
test.graphic <- graph.adjacency(test.network,mode="undirected",weighted=TRUE,diag=FALSE)

The command "edge.betweenness(test.graphic)" gives this output:
4.0 3.5 3.5 2.5 2.5

Now, if I use the following matrix (I divided the weights all by 10):

0.0 0.1 0.0 0.0 0.0
0.1 0.0 0.1 0.1 0.0
0.0 0.1 0.0 0.0 0.1
0.0 0.1 0.0 0.0 0.1
0.0 0.0 0.1 0.1 0.0

the command "edge.betweenness" gives this output:

4.00 2.25 3.75 2.25 3.75

Also, if I divide by 100 the original matrix I get other edge betweenness values:
4 5 3 3 1

I do not know what I am doing wrong, as I suppose the edge betweenness should not change if I scale all the weights of the edges by the same value...
Does anyone have an idea?

Thanks so much in advance for the help!

Guido

P.S. Counting by hand the betweenness values, the ones of the first matrix should be the correct ones.

_______________________________________________
igraph-help mailing list
address@hidden
https://lists.nongnu.org/mailman/listinfo/igraph-help

--
Gabor Csardi <address@hidden>     MTA KFKI RMKI