From: Luc H. <luc...@in...> - 2013-02-21 09:47:51
|
Hi, Because of this problem I developped a fast implementation of page rank in the Grph library. Feel free to have a try with it. Don't hesiste to contact me if needed. I'll be happy to help you. http://www-sop.inria.fr/members/Luc.Hogie/grph/ Best regards, Luc. On 2/21/13 10:44 AM, ga...@gm... wrote: > Hello all, > > I was trying out JUNG2 for calculating Pagerank for larger datasets, exactly for dbpedia-datasets. > > Well using Strings with larger datasets is of course not very smart, so a Integer representation is necessary. > > I use a DirectedSparseGraph<Integer, Integer> and the following amount of (almost Integer) data: > preferenceVector as HashMap<Integer, Double> containing 18*10^6 entries > edges: 130*10^6 > vertices: 18*10^6 > > Trying out calculation of pagerank with DirectedSparseGraph<Integer, Integer> leads to huge RAM consumption (graph.addEdge(Integer, Integer, Integer, EdgeType.DIRECTED)). > > Why is that this way? Is there an more efficient way? Any important archetectural background? I allocated 30GB RAM and its not possible to calculate Pagerank with the above amount of edges and vertices with simple Integer values. > > best regards > ------------------------------------------------------------------------------ > Everyone hates slow websites. So do we. > Make your web apps faster with AppDynamics > Download AppDynamics Lite for free today: > http://p.sf.net/sfu/appdyn_d2d_feb > _______________________________________________ > Jung-support mailing list > Jun...@li... > https://lists.sourceforge.net/lists/listinfo/jung-support -- Luc Hogie - CNRS Research Engineer COMRED Research Unit (I3S(CNRS-UNS) INRIA) http://www-sop.inria.fr/members/Luc.Hogie/ luc...@i3... +33 4 89 73 24 25 (office) +33 6 80 91 40 71 (mobile) Skype ID: luchogie |