Good news, I'm in a happy mood. Also, I have managed to get the "smooth" Fuchterman and Reingold spring embedder to work with the multilevel scheme (mentioned previously and used to visualise the placement of vertices in a smooth fashion similar to Eades). Even luckier, I have managed to get it working with the MGF approximation - so much so that we can semi-visualise the layout being given to graphs as big as 4elt.
Now the bad news, there is only a 5% chance of it working due to some unknown issues generating positions which are NaN. Previously a problem with the distance between them being zero but it doesn't appear to be the cause here, so that's for me to investigate further. As always, there's some bug in there that is waiting for my attention.
Now the tricky part, the visualisations observed this far show the generation of good local layout but lacking in the global layout side of things (see the image below - when visualised this was moving but had slowed so much that it was unlikely to find a good global layout). The reason? the cooling schedule is great for smooth visualisation but makes generation of layout much longer as we wait for it to overcome all the local minima. This just gave me an idea to try with the original Frucherman and Reingold algorithm and see how that works out.
So the aim now? fix this damn NaN problem. After? play around with the visualisation, see if it can be improved and then get on with testing the dynamic matching stuff. For now, that's all I want to say until it works appropriately, so I'll keep you up to date, enjoy your Wednesday!
No comments:
Post a Comment