Following the article Nodes groups visualization I was trying to implement this work in the new (very) experimental 0.7 version, just for fun. The purpose of this Potatoes module is to enable users to highlight group of nodes in a network.

When clustering (i.e. grouping) nodes, it is useful to leverage group perception on the network visualization. For instance with SNA (Social Network Analysis) it is a good way to display communities and better see interactions between them.

However it is a difficult problem, both for a software designer and a developer. There is no perfect way to visualize groups when the number of them increases and when they starts to hide and blend each other. This problem appears especially when using convex hulls solution. The presented solution belongs to implicit surfaces systems, also called Metaballs. I found the visual result very precise and beautiful, some qualify it organic. Unfortunately it costs much more resources than hulls to calculate these potatoes and may be applied only on small graphs.

Here you can see a very good example of this feature.

The following video presents it with movement. A small network is displayed with 3 different groups, one per color. A spatialization algorithm is running and moves nodes according to their links. Metaballs are refreshed continuously when moving.

Gephi – Potatoes work in progress from gephi on Vimeo.

Conclusions of this experimental implementation will guide us for official specification of further Gephi releases. We will implement this functionality for vectorial export but presumably not in the 3d engine.

Related Posts:

• No related posts.

Functionality ~

Trackback