Re: [igraph] semi-supervized mincut

[Gábor Csárdi]

Hi Renato,

this was added to the development tree already, s-t minimum cuts
return the cut itself now, and the partitions as well. You can get the
development tree from launchpad, tell us if you need assistance with
that.

Best Regards,
Gabor

On Tue, Jul 6, 2010 at 1:28 AM, Renato Fabbri <address@hidden> wrote:
> hi list,
>
> if
> g=igraph.Graph()
>
> we have
> g.mincut(capacity=None) and g.mincut_value(source=-1, target=-1, 
> capacity=None)
> methods.
>
> the thing is that mincut() returns the partitions, but you cannot
> specify the nodes one wants in each partition.
>
> On the other hand, mincut_value() returns only the value, not the partitions.
>
> this makes it very odd (afaik impossible) to implement the mincut
> semi-supervised learning method such as described here:
> http://www.google.com/url?sa=t&source=web&cd=1&ved=0CBIQFjAA&url=http%3A%2F%2Fwww.cs.cmu.edu%2F~avrim%2FPapers%2Fmincut.ps&ei=OmkyTOd4wZyWB46z3MAL&usg=AFQjCNHWepNiPTGMk2FdUM-X5VVEtKk2Hg
>
> For a two class problem, one needs to specify at least that one node
> of each class should be in each partition, and have the partitions,
> not the value of the cut itself.
>
> It seems to me that the g.mincut_value() finds the partitions but
> discards it instead of returning the nodes related to each partition.
>
> any idea?
>
> cheers,
> rf
>
> --
> GNU/Linux User #479299
> skype: fabbri.renato
>
> _______________________________________________
> igraph-help mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/igraph-help
>



-- 
Gabor Csardi <address@hidden>     UNIL DGM