[Gephi-plugins] clustering plugins and nodes with multiple parents

Mathieu Bastian mathieu.bastian at gmail.com
Thu Jul 7 14:10:52 CEST 2011


On Wed, Jul 6, 2011 at 2:49 PM, Chris Esposito <chris.esposito at gmail.com>wrote:

> Hello All,
>
>
> I've recently finished the development of 2 clustering plugins, so I'll be
> submitting those shortly. An issue that can arise, both during the execution
> of a single algorithm and even more commonly when different algorithms are
> applied by the user in sequence, is that clusters may overlap. While this
> may not happen in a single run if the clustering algorithm in use during
> that run does a partial or complete partitioning of the graph, a different
> algorithm that partitions in a different way or isn't partiton-based will
> wind up producing groups that overlap others, which means that a node would
> be a member of multiple clusters, something that that graph API in Gephi
> prohibits. For large networks and large numbers of clusters, even asking the
> user to keep track of what cluster a node is or should be in doesn't work
> very well, I think.
>
> I've built this multiple-parent capability into other graph systems that
> I've developed over the years, but I haven't yet had the time to look into
> what's involved in modifying gephi's underlying data structures to support a
> node having multiple parents.
>
I think this can be done using list attributes. Though really visible yet on
the UI side, you can define list attributes values [1]. For instance,
instead of a single INTEGER for the group identifier, one could have a list
of ints. So it's totally doable to a valid data structure to host the result
of your algorithm in a good way. However, the UI and in particular the
Partition module is not using these lists yet so it your attribute
wont'appear there.

Do you have suggestions on how to adapt Gephi GUI to support this, as the
purpose of partition is to output a single color per node.

[1]
http://gephi.org/docs/api/org/gephi/data/attributes/api/AttributeType.html

Mathieu

>  Does anybody have a rough order-of-magnitude estimate of how hard this
> would be in Gephi?
>
> Thanks,
>
> Chris
>
> _______________________________________________
> gephi-plugins mailing list
> gephi-plugins at lists.gephi.org
> http://gephi.org/mailman/listinfo/gephi-plugins
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://gephi.org/pipermail/gephi-plugins/attachments/20110707/4e742714/attachment.html>


More information about the gephi-plugins mailing list