partest laufen lassen

switch on version

#
##### mlr ######

- make test still produces output of learners even when configureMlr(show.learner.output=FALSE)

- b632+ test
 test b632 too

- document default perf measures

- checkTaskLEarner at beginning of resample, but then not each train call

- describe parallel in tutorial

- rd name of supervisedtask, report roxygen2 bug
- examples

- bbmisc function to print vector with names

- function to check if learner is applicable to task

- listMeasures

- checkFeature names for predict

- bug in describe /staticdocs

- allow task which only becomes true supervised task after preprocessing / feature
  extraction

- drop factor levels in makeTask
  schwierig. das ist beim subsetten des tasks?
  dann haben wir andere class levels? trat im allpairs exp auf.
auch doof:
task = makeClassifTask(data=iris, target="Species")
rin = makeFixedHoldoutInstance(train.inds=1:100, test.inds=101:150, size=150)
resample(makeLearner("classif.randomForest"), task, rin)


##### tune #####

- tune thresh in the end

- it is inconsistent to sometimes use controlobjects, sometimes ... in our control objects

- test mbo with trafo

- mlr tuneMBO minimize autom. setzen

- check that control and tune results are printed properly.

##### featsel #####

## FIXME: sbs and sfbs prefer higher number of features, when performance stays the same

add a function to get the selection path for features in forward selection

##### chains #####

- it's not clear how the changed descriptions of the files should be saved, e.g. when we change everything potentially during preproc
a) output of the model via print
b) when the task is forwarded via the training funs

check all print.funs

remove prints in mlrChains wrappers

add getLEarnerModel

add getters for all relevant stuff in wrappers

check that wrapper work with nfeatperc
