create_autotuner: Create an AutoTuner with a single line of code

In a-hanf/mlr3automl: AutoML in mlr3

Description

Small utility function, which creates an AutoTuner for a given learner.
Uses the same interface as the AutoTuner, but provides defaults for all arguments.
Parameter spaces are identical to the ones used in mlr3automl.

Usage

create_autotuner(
  learner = lrn("classif.xgboost"),
  resampling = rsmp("cv", folds = 10),
  measure = NULL,
  terminator = trm("run_time", secs = 60),
  tuner = tnr("random_search"),
  num_effective_vars = NULL
)

Arguments

learner	(character(1) | Learner) Learner inside the AutoTuner. Parameter sets are predefined for ranger, xgboost, liblinear, svm and cv_glmnet learners for both prediction and regression. Other learners will obtain empty parameter sets.
resampling	(Resampling)
measure	(Measure)
terminator	(Terminator)
tuner	(Tuner | TunerHyperband) Tuner. Hyperband is supported by creating a GraphLearner with PipeOpSubsample.
num_effective_vars	Integer giving the number of features in the dataset. Only required for parameter transformation of mtry in Random Forest (we are tuning over num_effective_vars^0.1 to num_effective_vars^0.9)

Value

AutoTuner

Examples

library(mlr3automl)
my_autotuner = create_autotuner(lrn("classif.svm"))

a-hanf/mlr3automl documentation built on Feb. 21, 2022, 1:06 a.m.