[Boost-users] Implementing matrix_as_graph to convert a boost::numeric::ublas::matrix to a "graph"

Subject: [Boost-users] Implementing matrix_as_graph to convert a boost::numeric::ublas::matrix to a "graph"
From: John Burgoon (noogrub_at_[hidden])
Date: 2009-03-28 21:34:38

Hi all,

I wrote a small program in C++ using boost to read a neural network file
(directional graph of a feedforward multilayer perceptron) into
boost::numeric::ublas::matrix.  The typical file size creates matrices of
various sizes, anywhere from a 36x36 to a 130x130 matrix (of floats - edge
weights between 0 and 1 to four sig figs).  About half to two-thirds of the
values are 0 in a typical graph, so I think of them as sparse.  There are
thousands of such files, so I have been really thrilled with the speed of
boost.  (I was doing this work with Perl last year.  Ahem.)

Now I want to implement the boost matrix_as_graph library so that I can
convert my matrix into a graph and then implement various network metrics on
the graph (such as betweenness_centrality.hpp).  However, when I looked in
the docs at
http://www.boost.org/doc/libs/1_38_0/boost/graph/matrix_as_graph.hpp
I was dismayed to find only a code listing: no explanation, no examples.

I'm not the strongest C++ coder.  Can someone point me to an example or
tutorial of how to implement matrix_as_graph to turn a ublas::matrix into a
graph?  I think I could puzzle it out if I could find an example to emulate.

I'll repay the community with a simple blog tutorial of how I did it, if
someone can only help me get started.

Thanks in advance,
John
-- 
John Burgoon
Indiana University
Bloomington, IN, USA
-- 
>>Insight without implementation is worthless. Get to work.