search

mlr-org / mlr3tuning

Hyperparameter optimization package of the mlr3 ecosystem
https://mlr3tuning.mlr-org.com/
GNU Lesser General Public License v3.0
53 stars 5 forks source link

Could we use a default measure for all tuning functions? #294

Closed jakob-r closed 2 years ago

jakob-r commented 3 years ago

Resampling in mlr3 does not need a specific measure because we can calculate all measures afterwards.

For tuning the measure can influence the location of the optimum but for most measures it will be the same (i.e. accuracy/mce). So it could make sense to have automatic defaults?

The function mlr3::as_measures(NULL, task_type) returns a default measure.

berndbischl commented 3 years ago

the optimum location is really no argument. it will certainly be different for different measures.

of course we can use the default measure of the task type, but i really don't see too much value in here? why not have it explicit?

berndbischl commented 2 years ago

we can use mlr3::default_measures(task_type)

mllg commented 2 years ago

as_measure(NULL, task_type) should also work