Analysis of the main components in a cluster via MPI

I am setting up a set of computers that run math programs on top of MPI. Do you know if there is any library that does PCA - core component analysis using MPI to use all the resources of networked computers? I'll take a look at Scalpack, but do you know any other libraries? My language is C ++ on linux, but if there is a good lib, also for Windows it is the same

thanks

+2


a source to share


1 answer


PCA is a fairly cheap operation, so your communication ratio (getting data to nodes) relative to computation (actual operation, here PCA) is likely to be relatively poor.

This means that clustering may not be a great solution for this particular problem.



Also, PCA is really a linear algebra operation, so you're better off looking at optimized BLASs like ATLAS, Goto, MKL, ... which (these days) can use multiple cores, giving you implicit parallelism that's easier to use. than explicit parallelism using MPI.

Don't get me wrong - I really like MPI (and I have some lessons here on using R ), but you need to keep in mind that not all tools are suitable for all problems.

+2


a source







All Articles