Discussion: file storage

[dsuplica]

Currently, files are stored as a derivative bids-ish dataset. They are stored in with the same root as the preprocessed data.

I thought this was advantageous because it allows us to keep all files in the same place, and allows for easy loading/saving of only one part of the pipeline (say, you just want accuracy). It also allows us to select data from subjects more easily.

Open to suggestions on if we should be keeping this, or other ways to store the output data

[henrymj]

I'm in favor of storing them as derivatives as well. I think they meet the requirements of being a derivative, and are easy to find there.