Re: [igraph] two questions from a new member

[Gábor Csárdi]

On Wed, Dec 10, 2008 at 8:45 AM, Zhang Meng <address@hidden> wrote:
> Hi all,
>
> This is a new member of igraph solving the calculation of average shortest
> path length, diameter, cluster coefficient, k-cliques etc with
> size=100000(nodes) and edge higher 5000000(that is 100000*50), recommended
> by CFinder team for high efficient program.
>
> Here I downloaded igraph and found it is mainly for Unix environment. For
> me the network model is created using my program in matlab, and want to call
> igraph's functions in matlab under windows environment to calculate its
> features such as average shortest path length, k-Cliques..., so can you tell
> me how to compile the igraph using MinGW or other compilers? (As I know,
> there are so many compilers in MinGW and I don't know which one to use.)

Hi,

the easiest would be to install GNU R (www.r-project.org) and then the
igraph R package. There is some Matlab package (google for it) to call
R directly from Matlab, personally I have never tried it. It is surely
possible to call R as an external program from Matlab and R can read
and write Matlab files with the R.matlab package.

Note that igraph is a library, so you cannot just call it like you
would call a program from Matlab. To use igraph in Matlab you would
need to write a glue layer that translates Matlab functions into
igraph function calls. Currently we have such glue layers to  R and
Python only.

> Another question is, how long will it take to calculate the average shortest
> path length and k-cliques(where I want to know the number of 3-clique,
> 4-clique, 5-clique,...) using functions in igraph?

I doubt that you can calculate the average shortest path on such a big
graph, it is an O(n^2) algorithm. Similarly, K-cliques might not be
feasible in general, probably it is fine for k=3,4,5.

I would suggest to install R, and the igraph R package, generate some
example graphs with the same density or degree distribution as yours,
but in different sizes, and then see how the calculations you want to
do scale with the graph size and whether it is feasible to do them.

Gabor

> thanks
>
> --
> BRs
>
> ZHANG MENG
>
> _______________________________________________
> igraph-help mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/igraph-help
>
>



-- 
Gabor Csardi <address@hidden>     UNIL DGM