Fwd: Replying to Ahmed

[Roberto Zanoli]

---------- Forwarded message ----------
From: Vivi Nastase <vivi.a....@gmail.com>
Date: Fri, Jul 8, 2016 at 9:40 PM
Subject: Replying to Ahmed
To: Roberto Zanoli <zan...@fbk.eu>

_____________________________

Hi Ahmed

When doing N-fold crossvalidation (the -xval option for the Experimenter)
the system will first split the data into N folds, and then automatically
run the N experiments, where each of the combinations of N-1 folds are used
for training and the remaining one for testing. If you want to just split
the data into folds and not run, or just see if the split is OK you can use
the -fakeRun option. If your data is already split into folds, then you
don't need the Experimenter, just use the EOPRunner with the training and
testing that you want. You can write a wrapper around the EOPRunner to run
all the combination of the folds, similar to what the method
Experimenter.crossValidation() does.

best,
Vivi