Compare JDP results with LPP results
Closed this issue · 2 comments
In [1] the authors also partition hyperedges such that the number of replicas (i.e., vertices that are cut) is minimized. I implemented the algorithm today and at least my initial experiments on some of my hypergraphs indicate that the # of hyperedges incident to each of the blocks is reasonably balanced.
I've just pushed the LPP code to: https://github.com/SebastianSchlag/kahypar/tree/LPP
The binary can be built with: make HyperedgeLPP
You can call it like this:
./HyperedgeLPP -h /global_data/schlag/00publication/sea17/benchmark_set/ISPD98_ibm01.hgr --blocks=4 --epsilon=0.03 --mode=direct --objective=km1 --seed=42 -p ../../../config/km1_direct_kway_sea18.ini
Don't worry, the parameters epsilon, mode, objective, p won't actually be used.
The LPP partitioner then creates a file like /global_data/schlag/00publication/sea17/benchmark_set/ISPD98_ibm01.hgr.part4.epsilon0.03.seed42.LPP
that contains the assignment of each vertex.
Let me know if you have any questions.
I've just pushed the requested changes. Now you can use parameter --iterations=<int> to limit the number of LPP iterations.