How to set the seed of cvpartition

I would like to partition a set of training data the same way (randomly) over different iterations of the code. How can I set the seed for the random index choice of cvpartition

Answer 1

[Akira Agata]

Please add rng function just before cvpartition to set seed of the random number generation. Here is an simple example:

n = 10000;
k = 5;
seed = 100;

rng(seed);
c1 = cvpartition(n,'KFold',k);

rng(seed);
c2 = cvpartition(n,'KFold',k);

Then, c1 and c2 becomes the same.

>> isequal(c1,c2)
ans =
logical
 1

But, to partition a set of training data the same way over different iterations, you may not need to set the seed for each iteration, like:

c0 = cvpartition(n,'KFold',k);

for Iteration = 1:M
  for kk = 1:5
    trainSet = Data(c0.training(kk),:);
    testSet  = Data(c0.test(kk),:);
    ... (Some process for each train/test dataset)
  end
  ... (Some process for each iteration)
end