[Networkit] NetworKit Sprint
matteo at cs.brown.edu
Fri Dec 2 17:08:45 CET 2016
> On Dec 2, 2016, at 9:34 AM, Elisabetta Bergamini <elisabetta.bergamini at kit.edu> wrote:
> Hi everyone,
> Matteo, thank you for your e-mail and for bringing this up.
> In my opinion, the “pseudo-optimization” is not necessary: it’s not really faster and when using it we lose RK's theoretical guarantee. If one doesn’t need a guarantee, then it would probably make more sense to use ApproxBetweenness2 (algorithm by Geisberger et al.), which doesn’t have any theoretical guarantee but seems to give a good estimation even with a few samples.
This makes sense to me.
> By the way, the names ApproxBetweenness and ApproxBetweenness2 are not very meaningful. How about we call them BetweennessApproximation and BetweennessEstimation, respectively?
I would be careful with renaming, as you may break applications that use those classes/algos.
You may want to rename thm but create compatibility aliases and print a warning that using the old names may be deprecated soon.
I’m not going to do that myself =).
> If nobody has anything against that, I would suggest to just remove the pseudo-optimization and use the “pedantic” computation all the time.
I’ll see what I can do about it in the next few days.
More information about the NetworKit