Gephi forums

Hi all,

I've discovered Gephi recently and I've seen many video where the force atlas is turned on while the user is playing with the timeline to make new nodes appear or disappear. Overall, it is really wonderful.

What I would like to do is quite similar. I'm working in combustion and I have a network of chemical species with edges representing the link through chemical reactions. Throughout my simulations, new species and new connection appear, starting with a quite low number of nodes (around 10) and ending with around 250 nodes.

Here is my problem. When I specify the settings for the force atlas layout such that new nodes are added smoothly at the beginning (low number of nodes), it becomes really really slow when the number of nodes increases. Conversely, if everything is set to add new nodes at a decent speed for high number of nodes, when this number is reduced, they are added very quickly in a very stiff way.

I would really appreciate if you could help me on this.

Hi,

The problem with ForceAtlas is that its complexity is in n²: 2x more nodes makes a speed 4x slower. So switching from 10 nodes to 250 makes the speed 625x slower. This is a general problem, and the best we can do is to have a complexity of n.log(n), which means your speed would be "only" 65x slower. There will always be an impact of the number of nodes on the layout.

I just propose you to use the ForceAtlas 2 (work in progress) that is quite the same thing but with this optimization. Get it from this post:
http://forum.gephi.org/viewtopic.php?f= ... 3029#p3029

Thank you for the information. I will give it a try.

Another thing that I have noticed is the weird behavior of the algorithm once a node is settled.
For example, if I wait that my network has reached a steady state and I settle a node then the other nodes are much more attracted by this node than before. It is perhaps the right behavior but it looks strange to me.

What do you think?

Yes it's a wrong behavior. It has been fixed in ForceAtlas 2...