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 <[email protected]> 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 > [email protected] > https://lists.nongnu.org/mailman/listinfo/igraph-help > > -- Gabor Csardi <[email protected]> MTA KFKI RMKI
_______________________________________________ igraph-help mailing list [email protected] https://lists.nongnu.org/mailman/listinfo/igraph-help
