[Networkit] community EPP

Christian Staudt christian.staudt at kit.edu
Fri Mar 20 11:24:13 CET 2015


Hi Damiano,
two remarks:

- On the Python level, an instance of EPP has to be created with the EPPFactory helper class. This is a workaround due to difficulties of mapping certain OOP constructs from C++ to Python.
- Our experiments have shown that EPP has no practical advantage over PLM. EPP is therefore of academic interest only, and I don’t recommend using it for actual community detection tasks.

Best,
Chris

On 18 Mar 2015, at 14:20, Damiano Rossato <damiano_rossato at alice.it> wrote:

> Hi Max,
> 
> I show you the incriminated snippet of code:
> 
> communitiesuEPP = community.detectCommunities(Tuk, algo=community.EPP(Tuk))
> 
> instead everything works fine when I use this call:
> 
> communitiesuPLP = comsmunity.detectCommunities(Tuk,algo=community.PLP(Tuk))
> 
> Tuk is an undirected graph. I don't use the inizialization with EPPFactory, because I don't know how to use it. Can you show me how?
> 
> Damiano
> 
> ----Messaggio originale----
> Da: maximilian.vogel at student.kit.edu
> Data: 18-mar-2015 12.47
> A: "Damiano Rossato"<damiano_rossato at alice.it>, <NetworKit:a toolkit for high-performance network analysis <networkit at ira.uni-karlsruhe.de>;>
> Ogg: Re: [Networkit] community EPP
> 
> Hi Damiano,
> 
> Does the crash really happen everytime? For me at least the first call
> to run() after initiliazing the ensemble works just fine.
> Also, how do you initialize the ensemble? Do you use the EPP factory?
> The default base clusterer PLP can only be run once after
> initializiation, which is the reason why it crashes the kernel on my
> system when I try a second call to epp.run(). This crash should
> definitely be avoided. I'll look into this.
> 
> Max
> 
> On 18.03.2015 12:22, Damiano Rossato wrote:
> > Hi,
> > I'm trying to use the community.EPP algorithm on iPython3 but it
> > crashs the kernel everytime. In file EPP.cpp in community folder I
> > found this:
> >
> > // run base clusterers in parallel
> >     #pragma omp parallel for
> >     for (index b = 0; b < baseClusterers.size(); b += 1) {
> >         // FIXME: initialization of base clusterer?
> >         baseClusterers.at(b)->run();
> >         baseClusterings.at(b) = baseClusterers.at(b)->getPartition();
> >     }
> >
> > so I suppose that the issue I already know to Network's community.
> > Someone knows the solution?
> >
> > Thank you to everybody and have a nice day
> > Damiano Rossato
> 
> 
> 
> <ATT00001.c>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ira.uni-karlsruhe.de/mailman/private/networkit/attachments/20150320/f33cee95/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 495 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <https://lists.ira.uni-karlsruhe.de/mailman/private/networkit/attachments/20150320/f33cee95/attachment.sig>


More information about the NetworKit mailing list