Closed dafnevk closed 5 years ago
This issue is specific to the mcfly-tutorial and not to the mcfly code itself.
The issue seem to have been addressed, because there now is a function split_data_random, which takes a random sub-sample as opposed to the user-specified sample as used before.
Right now, the preprocessing of Pamap2 for the tutuorial, separates some subjects for validation and test, but this gives not good performance. We want a random subset for validation, with a parameter for the validation set size.