On Fri, Apr 25, 2008 at 2:24 PM, Oliver Kullmann
<O.Kullmann at swansea.ac.uk> wrote:
> Further to the bug:
> It's worse than I first thought:
> I had the impression that only the
> extreme case of one vertex caused problems, but
> actually max_clique became completely defunct:
>
> max_clique(complete_graph(4));
>
> max_degree: no max degree in an empty graph.
This has been fixed in cvs. Thanks for reporting it.
> When cleaning this up, perhaps one could also finally
> actually specify what this function is supposed to
> return: If just some maximal clique (w.r.t. subset-inclusion),
> then which heuristics is used, or otherwise, if actually
> a clique of maximum size is returned then this should be
> stated.
I think you already asked this and I already answered it. It returns a
clique of maximum size and the documentation in 5.15 states this. The
algorithm is branch and bound and the bound is based on greedy
coloring.
--
Andrej