0.5 Fix the algorithm so lower level graphs can find a better layout as opposed to being slanted or having edges crossed.
1. My first target, the clumping implemented by Yifan Hu, only activated when a graph can no longer be coarsened efficiently. Using the different in vertices between each coarse level will trigger such a mechanism, which will then switch to a different matcher (or the same matcher with different parameters, dependant on how confident I feel).
2. Comparison of the theories of each algorithm against their describing papers, to ensure I have not missed anything and that they remain as close to the original implementation as possible.
3. Comparison of the implementations, for best and fairest results, I want the implementations of each to be as similar as possible (so using the same coarsening scheme, data structures, readers, drawers, even techniques where applicable).
4. Testing structure; as discussed in my previous meeting with supervisors, ideally I want a testing framework I can plug the algorithms into and expect results in a comparable and reliable form. For comparisons, I intend to follow suit and primarily compare run times, but also edge crossings (where possible and using a method to find the best viewing angle) and edge length ratio. Other smaller attributes may be monitored out of curiosity, such as the angle between edges, but this is heavy to compute and not necessarily a requirement.
5.Kill the Batman...
Though I know how my plans have panned out previously, it's hard to differentiate away from these but no doubt I will find a way. Assuming all goes well, these are my tasks, any others will be added to this list or will make a great reason to write another blog post. I will update with any news if I stumble across anything, for now, I hope to update later this week.
In the mean time, here is a image of some daunting finan512 output for you to enjoy:
No comments:
Post a Comment