mb706
mb706
Similar to `mlr`'s `listLearners`, one should be able to use `listCPO` to query for all `CPO`s with certain properties.
Slice functions take the data as input and return the indices of affected columns as output. This enables user-defined slices and reduces cpo constructor parameter clutter. A simple constructor (`affect(type...
Tests are not covering some parts of CPO. TODO list: - [ ] dataformat.factor.with.ordered should always influence how strictly data form is checked in retrafo - [ ] make sure...
This would make things nicer, but also a bit more complicated. Should probably be solved only after #9 and #15 are done.
Imputation by Learner is a difficult beast, because it should be able to depend on more data than it imputes. E.g. it may use learners only fit to impute some...
Since CPORetrafo can also change target columns.
Since the prefix is also absent in CPOLearner inverter calls.
* number of rows must equal input rows * column names for non-response must be fitting the latter is especially important for "prob" inversion.