Support for multipartite/multimode networks

[jaroslav.kuchar]

Dear Gephi community,

I would like to propose an idea for GSoC project. There is no support for networks with more than one type of nodes in Gephi: bipartite/multipartite/multimode networks.
This extension will be useful for many researches of affiliation networks, recommendation systems, networks of collaborations, gene networks, peer to peer networks and many others.

Here is the initial list of missing features:
1) Statistics: degree, density, distances, redundancy, paths, centralities, clustering coefficient, …
2) Clustering: modularity, detecting communities,
3) Layouts: layered layout, radial, cyclic, Sugiyama method, 2D, 3D, force-directed, …
4) Transformations/Projections: general projections using adjacency matrices, weighted, overlap weighted, …

I think, there are more possible features.

Thank you for your ideas, comments and help with definition of possible objectives…

JK

[mbastian]

Hi Jaroslav,

Thanks for your suggestion. Multipartite networks are something very important indeed, and an area we haven't really explored yet. My knowledge is limited about them and would like to learn more in order to think about the necessary tooling. First of all, is the study of multipartite networks discussed in the literature and what are the few papers to read to get a better idea? I've seem many papers where multipartite networks are used but not generic enough usually.

In terms of features you list important points. Do you links to papers or code implementing those for multipartite networks?

In terms of implementation it's not something that should be too difficult to do. One can add attributs to nodes and edges easily in Gephi.

[jaroslav.kuchar]

Hi Mathieu,

Thanks for your suggestion. Multipartite networks are something very important indeed, and an area we haven't really explored yet. My knowledge is limited about them and would like to learn more in order to think about the necessary tooling. First of all, is the study of multipartite networks discu

I am not "expert" in multipartite networks, but I am a bit interested in this type of networks. That's true, there is a problem with papers/literature about multipartite networks in general. From my point of view, the good starting point are bipartite/affiliation networks.
http://arxiv.org/pdf/cond-mat/0611631v1.pdf
http://www.steveborgatti.com/papers/bhaffiliations.pdf
http://toreopsahl.com/
http://people.maths.ox.ac.uk/porterm/re ... _final.pdf
http://arxiv.org/pdf/1104.3184v2.pdf

In terms of features you list important points. Do you links to papers or code implementing those for multipartite networks?

I don't know about implementations of all of these features. Many tools focus only on bipartite networks http://networkx.lanl.gov/reference/algo ... rtite.html.

I already started implementation of plug-in which enables general projection(transformation) of multipartite graphs. It of course covers projection of bipartite networks to monopartite networks. It was implemented to support our lectures about SNA and it is not yet in public plugin repository.

In terms of implementation it's not something that should be too difficult to do. One can add attributs to nodes and edges easily in Gephi.

I know, but I am not sure, if it is good idea to have for each node(edge) type new attribute for each future...

JK