[Networkit] NetworKit

Lars Gottesbüren lars.gottesbueren at gmail.com
Mon Feb 9 23:02:42 CET 2015


Dear NetworKit developers,

I am currently working on a student project about matching-based multilevel
graph partitioning for Professor Meyerhenke's optimization problems class
and have a few questions and suggestions.

We use the MatchingContractor (subfolder coarsening), which when assigning
weights to the coarsened graph counts each original edge twice (thereby
doubling the total edge weight of the original graph). Now I am wondering
if that is unwanted behaviour or a feature. And most importantly if we're
allowed to modify it.

Secondly, we use the path-growing matching algorithm. Our graphs are
supposed to have selfloops but these are not supposed to be eligible for a
matching. As far as I can see, the path growing algorithm allows for that
to happen.
In this case, we would only like to add another function (so as not to mess
up other people's code), which essentially performs the same algorithm but
instead forbids selfloops as matching edges. In addition we would like the
new function to accept an EdgeScoring instance in addition to the graph,
because we want to easily deploy different edge ratings without having to
copy the graphs.

So, what's your opinion on this?

Regards,
Lars Gottesbüren
-------------- n�chster Teil --------------
Ein Dateianhang mit HTML-Daten wurde abgetrennt...
URL: <https://lists.ira.uni-karlsruhe.de/mailman/private/networkit/attachments/20150209/c7400dc5/attachment.html>


More information about the NetworKit mailing list