hi Grant, The original version of psvm is build on top of google's map/reduce implementation. But the author move it to mpi. So I think the best way is to contact the author to see whether he can share some code or move the implementation to hadoop or give some clue on it. Here is the author Kaihua Zhu's email: [EMAIL PROTECTED]
On 5/6/08, Grant Ingersoll <[EMAIL PROTECTED]> wrote: > > Cool! Are you suggesting you would like to donate it? I see it uses MPI > (boy, that takes me back...) This is probably something we will be using > eventually for things that don't run on Hadoop, but I understand that SVM > can run on Map Reduce. Do you have a M/R version? Have you compared the > two versions? > > -Grant > > On May 6, 2008, at 12:06 AM, phonechen wrote: > > ** > > *Parallel svm | a parallel version of support vector > > machine<http://code.google.com/p/psvm/> > > * > > here is the paper: > > > > http://psvm.googlegroups.com/web/NIPS2007-435.pdf?gda=B3ZXxkEAAAC0-d9XK49vdMfIT-gxvP9xH1vxRghNOvOxp_YnhJOW52G1qiJ7UbTIup-M2XPURDTxHA03ey30XW-34-8ERa4l5koy8jtfjFBmuoKXCx0UNA > > and here is the mpi version source code: > > http://code.google.com/p/psvm/ > > > > > > > > -- > > --~--~---------~--~----~------------~-------~-- > > > > Best Regards, > > > > Yours > > Phonechen > > > > -~----------~----~----~----~------~----~------ > > > > > -- --~--~---------~--~----~------------~-------~-- Best Regards, Yours Phonechen -~----------~----~----~----~------~----~------
