Allow user to set seed for reproducibility about crossvalidation HOT 4 CLOSED

I think this is a good idea, but I'm not really sure how to go about it. The documentation for the generateCVRuns function doesn't list any seed arguments:

from crossvalidation.

Or does calling set.seed() at the beginning of the script set the seed for the duration of the script?

from crossvalidation.

Yes. You can simply run set.seed before the random portion of your code, which I believe is the call to generateCVRuns. We can choose a default seed so that the user can get run-over-run consistency. We can make the default random if required using Math.round(Math.random()*10e4) in javascript.

from crossvalidation.

As this example shows, set.seed() doesn't make generateCVRuns deterministic.

myLetters <- sample(c("A", "B", "C"), size = 100, replace = TRUE)
set.seed(1)
library(ROCR)
firstruns <- generateCVRuns(labels = myLetters, ntimes = 2, nfold = 10)
secondruns <- generateCVRuns(labels = myLetters, ntimes = 2, nfold = 10)

Can we do anything about this? If not, should I implement a "fold creating" function that can use the user-chosen seed myself?

from crossvalidation.