rival
rival copied to clipboard
per-user splitting
Implement a splitting strategy according to http://graphlab.com/files/lsrs2013/paper_12.pdf
Like in issue #26, I think this should be split in two:
- first, the splitting strategy that does the user filtering
- second, a splitting strategy that takes another splitting strategy and returns one split for each user (either a CV, random split, or a user-filtered one like in the paper)
Also, the cutoff N for the item selection should be introduced somewhere (probably in the first strategy)