Hi Mendhi, On 08/06/2011 04:17 AM, Mehdi Khoury wrote:
Hi Tiago,
thanks for the feedback.
I have followed your advice for the scale-free and er network and there are no pb for these. The watts-strogatz network version I have implemented works, but is very very slow. [...]
Yes, indeed. Doing it in C++ would be much faster. But there are ways to improve your code: - Do not call the range() function, since it is O(N). In loops, you should call xrange() instead. - To sample a random vertex, you can do something like: g.vertex(numpy.random.randint(g.num_vertices())) This is much faster than using choice().
I have noticed that the boost library has a small-world graph generator see: http://www.boost.org/doc/libs/1_40_0/libs/graph/doc/small_world_generator.ht... Would there be any way to call that inside graph-tool as a faster alternative?
Yes, I will implement this as soon as I have some time. Cheers, Tiago -- Tiago de Paula Peixoto <tiago@skewed.de>