Principal Component Analysis

[manustone]

Hi All
I am newbie to ScaLAPACK.
I need to perform a big PCA ( Principal Compoent Analysis ) ove a huge amount of data.
I would like to distribute it over my cluster of computer using MPI.
Do you know if ScaLAPACK is able to do that?
Thanks mnstn

[Julien Langou]

pretty much certain, what the dimension of your matrix?
If it's tall and skinny, there might be better algorithm around,
let us know.
--j

[manustone]

Hi Julien
Many Thanks for your reply.

My matrix is done by 65000( might go up to 95000 ) rows,
but I have 3 scenario for the columns and I need to cover all of the 3

a: 44 columns
b: 660 columns
c: 8030 columns

Is this matrix a good candidate for Scalapack PCA?
If not what could be a good size where Scalapack is more valuable?
Mnstn