Initiale Lösung für KaHyPar

[poettig]

It might be that Sebastian's partitioner will require a good initial solution to give good results: think about ways to get a good initial solution and ideally have some randomized component in there to generate alternative good initial solutions such that the partitioner can start optimizing from different good starting points to maybe better escape local optima.

[adrianzap]

Just use PaToH? :-)

[lukashuebner]

Wer mag das testen?

[adrianzap]

Klassische Graph-Balancierung (Knotengewichte = Grad; oder Kantengewichte = sum(1/|v|) für v in e) mit KaHyPar berechnen und als initiale Lösung verwenden?