Thanks you for your answer. I know that without the graph is impossible to
know what happened, but with knowing that it can happen even in a rare a
remote case, it's enough. Thank you Tamas.
El 2/6/2016 21:21, "Tamas Nepusz" <[email protected]> escribió:

> Hi,
>
> In the absence of an example that reproduces the problem, I can only
> say that _in theory_, it is possible that some rare cases result in
> communities being split when using the multilevel algorithm, although
> it is definitely not common (you are the first one to report it).
>
> For instance, consider a subgraph consisting of 3 vertices (a, b and
> c) such that a is connected to b, b is connected to c, but a is not
> connected to c and there are no other short paths leading from a to c.
> In that case, if b belongs to a community X at a given iteration of
> the algorithm, it _might_ happen (in rare cases) that both a and c
> decide to move themselves to community X while, at the same time, b
> decides to leave community X for another one. This would result in
> community X being split into two disconnected components. Of course
> this is not enough because further iterations of the algorithm might
> annihilate one of the components (or both).
>
> T.
>
>
> On Thu, Jun 2, 2016 at 1:53 PM, Kuu <[email protected]> wrote:
> > Unfortunately I work with sensible data and I don't have the option to
> > extract information from my PC (I don't even have internet connection,
> this
> > is sent from my smart phone), so I can not send it to you.
> >
> > El 2/6/2016 13:42, "Tamas Nepusz" <[email protected]> escribió:
> >
> > Hello,
> >
> > Can you send us a graph that reproduces this behaviour?
> > T.
> >
> >
> > On Thu, Jun 2, 2016 at 1:31 PM, Kuu <[email protected]> wrote:
> >> Hello,
> >>
> >> I have a graph that has disconnected components and I want to find
> >> communities over it. I'm applying directly the community multilevel
> >> algorithm, but I'm obtaining communities that aren't connected, which
> from
> >> my point of view, those shouldn't be considered as a valid community.
> >>
> >> Should I run the algorithm only over connected components?? By reading
> the
> >> original paper I understand that only two communities are merged if
> >> they're
> >> connected, but maybe I'm not understanding it correctly.
> >>
> >> Thanks for your help.
> >>
> >> Regards,
> >> Javier
> >>
> >>
> >> _______________________________________________
> >> igraph-help mailing list
> >> [email protected]
> >> https://lists.nongnu.org/mailman/listinfo/igraph-help
> >>
> >
> > _______________________________________________
> > igraph-help mailing list
> > [email protected]
> > https://lists.nongnu.org/mailman/listinfo/igraph-help
> >
> >
> > _______________________________________________
> > igraph-help mailing list
> > [email protected]
> > https://lists.nongnu.org/mailman/listinfo/igraph-help
> >
>
> _______________________________________________
> igraph-help mailing list
> [email protected]
> https://lists.nongnu.org/mailman/listinfo/igraph-help
>
_______________________________________________
igraph-help mailing list
[email protected]
https://lists.nongnu.org/mailman/listinfo/igraph-help

Reply via email to