Explain what you would like to see improved and how.
When using the RDataLoader for splitting a dataset into train/test subsets, the first fraction of every cluster is always assigned to training and the last one is always assigned to testing, this does not pose a problem within one complete training loop over N epochs, but this does in fact mean that across different runs, the resulting train/test datasets will always be the same.
ROOT version
v6.40-rc
Installation method
all
Operating system
all
Additional context
No response
Explain what you would like to see improved and how.
When using the
RDataLoaderfor splitting a dataset into train/test subsets, the first fraction of every cluster is always assigned to training and the last one is always assigned to testing, this does not pose a problem within one complete training loop over N epochs, but this does in fact mean that across different runs, the resulting train/test datasets will always be the same.ROOT version
v6.40-rc
Installation method
all
Operating system
all
Additional context
No response