[igraph] calculating centrality separately for each connected component

[Michael Bishop]

Even when I use sparse matrices, I'm having difficulty calculating measures of centrality on a large graph (less than 100,000 vertices and 300,000 edges)

I thought of a workaround though which would be helpful to me in its own right.  Due to the way the data was connected, there are no large connected components.  Therefore, if I can run functions like alpha.centrality() and bonpow() separately on each connected component my system should be able to handle it without crashing.  Even if this results in a slightly different measure, I don't think the difference is substantively important.

  

Given that I'm starting with a giant igraph object, how should I proceed?

For others with similar interests, I include Gabor's previous wisdom and links on how to implement these functions with sparse matrices:

 Hmmmm, 'alpha.centrality' is not very good, it uses the adjacency

 matrix, it is thus inappropriate for large graphs. Here is what

 you can do using the 'Matrix' package:

 http://lists.gnu.org/archive/html/igraph-help/2008-04/msg00049.html

and it is possible to use sparse matrices for bonpow as well, see e.g.

http://igraph.wikidot.com/r-recipes#toc6