I would like to ask the people in the NetworkX community what they
think about
incorporating a hierarchical graph datastructure into NetworkX.
Recently, there has been much research going on related to
clustering / modularity / community
in networks. E.g. the algorithm by Blondel et al. reconstructs a
hierarchical network representations [1]
I found GraphML [2] to have a very clear structure to represent such
graphs in XML.
What do you think of such an extension to NetworkX?
Regards,
Stephan
[1] http://sites.google.com/site/findcommunities/
[2] http://graphml.graphdrawing.org/
We would be glad to include community finding algorithms.
There are a few open tickets and some recent developments that you
might find interesting.
If you have a partition of the graph into communities (non-overlapping)
you can use the recently added blockmodel code to build a reduced graph.
See
http://networkx.lanl.gov/preview/reference/generated/networkx.blockmodel.html
http://networkx.lanl.gov/preview/examples/algorithms/blockmodel.html
Here are some other open and unfinished tickets related to community finding:
https://networkx.lanl.gov/trac/ticket/158
https://networkx.lanl.gov/trac/ticket/239
https://networkx.lanl.gov/trac/ticket/245
Aric
Stephan
On Mar 11, 10:21 pm, Aric Hagberg <ahagb...@gmail.com> wrote:
> On Thu, Mar 11, 2010 at 9:48 AM, Stephan Gerhard <enthou...@unidesign.ch> wrote:
> > Hi all,
>
> > I would like to ask the people in the NetworkX community what they
> > think about
> > incorporating a hierarchical graph datastructure into NetworkX.
>
> > Recently, there has been much research going on related to
> > clustering / modularity / community
> > in networks. E.g. the algorithm by Blondel et al. reconstructs a
> > hierarchical network representations [1]
>
> > I found GraphML [2] to have a very clear structure to represent such
> > graphs in XML.
>
> > What do you think of such an extension to NetworkX?
>
> We would be glad to include community finding algorithms.
>
> There are a few open tickets and some recent developments that you
> might find interesting.
>
> If you have a partition of the graph into communities (non-overlapping)
> you can use the recently added blockmodel code to build a reduced graph.