Hello R masteRs,
I was wondering if somebody had already implemented a
parallel version of the function Nnet (with the SNOW
package for instance) and would be willing to share a
few pointers on how to achieve it. I have a training
set of dimensions 905,000 X 5. Should I just get more
RAM and run the nnet on one procesor? Or is there a
slick way to parallelize the computations?
I have tried to load the training set on each node and
have the nnet function run on the master, but it seems
that the master node will still put the whole training
set in memory....
Thank you in advance for your suggestions,
Ron Piccinini.
