we CLEARLY need to document what "default" means in a learner param set

[berndbischl]

see the problems here #1479

[berndbischl]

what is the default in the ParamSet: it DOCUMENTS what is the default in the UNDERLYING function.
in this case, that is model = TRUE. it is not used in ANY way in code. so if you change it, it has no effect.
what is in there, does not refer to mlr in any way. just for glm.

if you want to change default behavior in a learner, you have to set par.vals.
(we should really not do this very often)

if you change par.vals, say in the note why.

i thought it would be written down here, but seems not to be the case
https://mlr-org.github.io/mlr-tutorial/devel/html/create_learner/index.html

[berndbischl]

we should also think about whether we want to place that info somewhere in the r docs

[berndbischl]

yes, this is kinda a tutorial issue, i created this here, so we solve this soon

[berndbischl]

actually we have to document what ALL things in the paramset mean.
constraints and requires settings.
and also how defaults should be specified when mlr introduces new parameters, like in regr.randomForest, the se.estimation stuff

[larskotthoff]

[schiffner]

I have written a lot of stuff, but not pushed yet. Give me a few more days!

[berndbischl]

@schiffner
cool. please push soon.

[schiffner]

Ok. Sorry. Was hit critically by SUDDEN RELEASE...

[berndbischl]

now, if you start complaining about TOO FREQUENT releases ..... :)

[larskotthoff]

I think the surprise of releases will decrease as they become more frequent.

[schiffner]

What Lars said. :)