Force Atlas seems not to work properly

[seinecle]

Hi,

The Force Atlas seems to work awkwardly: very slow, the nodes do not spread: the isolated nodes remain at the centre when they should form a circle around the main component.

---- edit: this seems to depend on which data is used.


Best,

Clement

[mbastian]

Hi,

Is it with Undirected or Mixed Graphs only ?

[jacomyma]

I was the author of the first implementation but I didn't make this one. The algorithm was empirically designed for medium sized, small-world, graphs.
--edit : I previously told bullshit. Attraction works fine. Nevertheless attraction / repulsion are not as well balanced as previously (repulsion has to be high) but no big deal.

[mbastian]

The main difference I see is that now ForceAtlas use edge's weight when calculating forces. That could matter a lot, if you have edge weight different from 1.Probably needs more testing about this particular point and maybe specify it in the algorithm description snippet.