luoleicn / psvm Goto Github PK

Dear all!

I'm trying to implement your algorithm and I find something interesting. 
During the IPM, you set the vector of x to be 0 as initial value. Will the 
result be improved a little if you set some initial value like c/2?

Thanks
yours,
Zeyuan Zhu

I spawn the code on windows platform and using MPICH2 to develop.
My hardware platform is on two workstation with multicore(intel Xeon 2CPU 
with E5435).So I have 16 cores to use.The communication between two 
workstaion is using high speed lan. But I got low speedup on it. Could you 
tell me the reason? Thanks.
The speedup figures can be found in following figs. The communication time 
sometimes are longer than the computation time.

Hello,
   I am noticing that the speedup is very bad as the number of processors
increases beyond 4. It is very easy to see what the problem is. The serial
Cholesky Factorization (Not the PICF) used in the IPM will always take the
same amount of time for a fixed problem size independent of the number of
processors. Thus  If this were done in parallel, the speedup might be improved.
Here are the results I got for a cluster of 8-core Xeon Nodes using an
infiniband network.
 CPUS   TIME    TRAIN   IO  ICF IPM B EST.  OUT EFF   SpdUp
2   16.507  13.684  1.547   3.529   3.344   6.810   1.277   1.000 2.000
4   13.955  12.398  0.755   3.197   2.749   6.451   0.803   0.591 2.366
8   12.015  10.929  0.515   2.466   2.490   5.974   0.570   0.343 2.748
16  6.425   5.437   0.529   1.137   2.356   1.944   0.459   0.321 5.138
24  5.082   4.181   0.163   0.512   2.396   1.272   0.739   0.271 6.496
32  5.684   4.269   1.050   0.968   2.307   0.994   0.365   0.182 5.808

Also , there seems to be some issues if the input file get too large..
The program never get through the PICF.

Thanks, 
  Patrick Nichols

Hello,

  We are testing PSVM using PBS and MOAB schedulers. In the code, you guys
are using exit(1) when an error occurs. This causes that particular job to 
quit, however all the other processes keep going. The job then goes zombie
and has to be killed manually ( we usually do not get root access so it's a
touchy subject to ask our sys administrator to do this).
  Would it be possible to add a function for this use, something like:

void paralllel_interface:exit()
{
  MPI_Abort(MPI_COMM_WORLD,EXIT_FAILURE);
}
which would get around this problem?

Thanks,
Patrick Nichols

Hi, do we have the cross-validation function in psvm, like the -v option in
libsvm? Or, how can I wrap the cross-validation from psvm source code?

thanks