Generates a hyperparameter tuning control object

Tries to automatically create a suitable hyperparameter tuning control.

generateHyperControl(
  task,
  par.config = NULL,
  learner = NULL,
  budget.evals = 250
)

Arguments

task	[Task] The Task
par.config	[ParConfig] The Parameter Configuration
learner	[Learner] Learner
budget.evals	[integer] How many train-test iterations do you want to allow?

Value

[HyperControl]

Examples

par.config = getDefaultParConfig("regr.randomForest")
hyper.control = generateHyperControl(bh.task, par.config)
# get what is inside
getHyperControlMeasures(hyper.control)
#> [[1]]
#> Name: Mean of squared errors
#> Performance measure: mse
#> Properties: regr,req.pred,req.truth
#> Minimize: TRUE
#> Best: 0; Worst: Inf
#> Aggregated by: test.mean
#> Arguments: 
#> Note: Defined as: mean((response - truth)^2)
#> 
getHyperControlMlrControl(hyper.control)
#> Tune control: TuneControlMBO
#> Same resampling instance: TRUE
#> Imputation value: <worst>
#> Start: <NULL>
#> 
#> Tune threshold: FALSE
#> Further arguments: 
getHyperControlResampling(hyper.control)
#> Resample description: cross-validation with 10 iterations.
#> Predict: test
#> Stratification: FALSE
# change what is inside
hyper.control = setHyperControlMeasures(hyper.control, measures = medse)
hyper.control = setHyperControlMlrControl(
  hyper.control,
  mlr.control = makeTuneControlRandom(maxit = 10))
hyper.control = setHyperControlResampling(hyper.control, resampling = cv3)
hyperopt(task = bh.task, par.config = par.config, hyper.control = hyper.control)
#> [Tune] Started tuning learner regr.randomForest for parameter set:
#>             Type len Def  Constr Req Tunable Trafo
#> nodesize integer   -   1 1 to 10   -    TRUE     -
#> mtry     integer   -   4 1 to 13   -    TRUE     -
#> With control class: TuneControlRandom
#> Imputation value: Inf
#> [Tune-x] 1: nodesize=5; mtry=10
#> [Tune-y] 1: medse.test.mean=2.6536466; time: 0.0 min
#> [Tune-x] 2: nodesize=5; mtry=11
#> [Tune-y] 2: medse.test.mean=2.5917132; time: 0.0 min
#> [Tune-x] 3: nodesize=1; mtry=8
#> [Tune-y] 3: medse.test.mean=2.5762609; time: 0.0 min
#> [Tune-x] 4: nodesize=10; mtry=2
#> [Tune-y] 4: medse.test.mean=3.0611292; time: 0.0 min
#> [Tune-x] 5: nodesize=8; mtry=2
#> [Tune-y] 5: medse.test.mean=3.0070662; time: 0.0 min
#> [Tune-x] 6: nodesize=9; mtry=4
#> [Tune-y] 6: medse.test.mean=2.6248005; time: 0.0 min
#> [Tune-x] 7: nodesize=4; mtry=4
#> [Tune-y] 7: medse.test.mean=2.5734068; time: 0.0 min
#> [Tune-x] 8: nodesize=4; mtry=6
#> [Tune-y] 8: medse.test.mean=2.5121071; time: 0.0 min
#> [Tune-x] 9: nodesize=7; mtry=13
#> [Tune-y] 9: medse.test.mean=2.7329064; time: 0.0 min
#> [Tune-x] 10: nodesize=5; mtry=2
#> [Tune-y] 10: medse.test.mean=2.7246530; time: 0.0 min
#> [Tune] Result: nodesize=4; mtry=6 : medse.test.mean=2.5121071
#> Tune result:
#> Op. pars: nodesize=4; mtry=6
#> medse.test.mean=2.5121071