Broken / unused chebai features #160
Description
When making changes to chebai, I did not update all features accordingly, meaning some of them are probably broken at this point. This includes
- cross-validation (see Crossvalidation needs to be generalised #11 ) - I implemented this but I don't think it was ever functional / used
- ChEBIOverXPartial - this requires selecting a subset of both labels and samples, should be easy to get running again if needed (prior experiments suggest that this is not a good idea)
- ChEBI with different train / test versions - I implemented that for the Chebifier paper. Apart from that I don't see a reoccuring use case
I would suggest not to fix these features (or maybe even remove them entirely to avoid confusion) and only get them running again if necessary.