Gephi forums

I found new layouts really powerful in a paper of Yifan Hu and Al untitled "A Maxent-Stress Model for Graph Layout" (2013).
The following algorithms experimented in this paper are named pivotMDS, PivotMDs(k), Maxent(k), sfdp, GRIP, and FSM (full stress model).
SFDP and GRIP are particularly impressive for visualization and factorization of large graphs, compared with standard spanning tree algorithms.
It would be a significant step beyond to find them implemented as plugins for gephi.

Hello cthovex.

Have you tried the Force Atlas 2 layout algorithm in Gephi? In my own experience it's been superior to other layout algorithms. It's fast and produces intuitive results. Especially when nodes have sizes and node overlapping is not desired.

Here are some examples I've done with Force Atlas 2:
http://85.25.226.110/mapper - 212600 nodes, 4045203 edges, read the description at the web page for more info
http://85.25.226.110/mapper/plrnet-1M-10M-color?noidle - 1 million nodes, 10 million edges
http://85.25.226.110/mapper/plrnet-10-1 ... g-b?noidle - example when node sizes are quite regular

Before finding FA2 I tried SFDP and OpenOrd and maybe few other layout algorithms. FA2 seems to perform the best in general both in layout and in speed. It's also great that Gephi provides easy interface for using the layout algorithm along with other network analyzing algorithms.


Is there some particular use case where you would want to use SFDP or GRIP? There are great examples of SFDP at http://www2.research.att.com/~yifanhu/G ... ndex1.html. Though my own opinion is that although the resulting images look great they might not provide much information about the underlying network.


Those who are interested in developing their own layout algorithms for Gephi can see the Gephi developer tutorials at https://gephi.org/developers/. It's possible to create your own plugins relavitely easily with Java.