(Not true CV of course as there's no training involved; just tested the method separately on each test fold).
Based on @bramiozo's initial implementation (this is a copy of bram_branch, which should now be safe to delete). Had to be changed because his implementation used the older version of the dataset splitter. It became a bit of a hodgepodge as I tried to keep most of his code in tact, but not all, so feel free to change stuff (back). But as a first shot I think this ought to do. Main thing I left out is the KDE / ridgeline plots, as we now only have 10 splits and those didn't look right any more.
I also reverted a bunch of files that were changed in his initial commit but that seemed accidental and/or not directly related to the purpose of this branch/PR.
(Not true CV of course as there's no training involved; just tested the method separately on each test fold).
Based on @bramiozo's initial implementation (this is a copy of
bram_branch
, which should now be safe to delete). Had to be changed because his implementation used the older version of the dataset splitter. It became a bit of a hodgepodge as I tried to keep most of his code in tact, but not all, so feel free to change stuff (back). But as a first shot I think this ought to do. Main thing I left out is the KDE / ridgeline plots, as we now only have 10 splits and those didn't look right any more.I also reverted a bunch of files that were changed in his initial commit but that seemed accidental and/or not directly related to the purpose of this branch/PR.