[R] tune.svm (e1071) takes forever

[aramhovsepyan]

Hi,

I am in the process of migrating my machine learning scripts from bash +
libsvm to R and for the past weeks I have been battling a substantial
problem when using the tune.svm function from the e1071 package. The
tune.svm function that performs the grid-search for the best cost and gamma
parameters takes really long time compared to the grid search provided by
the libsvm library. The tune.svm takes 10 times longer than the grid search
from libsvm. For a very large data set this is too slow.
I would like to ask if there is any way to speed up the grid search. 
Here is the function call I use:

tune.obj <- tune.svm( formula, data=dataset, cost = 10^(0:2), gamma =
10^(-6:-4) )

Ideally I would like to search for cost 10^(0:3) and gamma = 10^(-6:-3)
however that takes even longer.
Thanks,
Aram.

--
View this message in context: http://r.789695.n4.nabble.com/tune-svm-e1071-takes-forever-tp4634415.html
Sent from the R help mailing list archive at Nabble.com.