
Hyperparameter Tuning with Covariance Matrix Adaptation Evolution Strategy
Source:R/TunerCmaes.R
mlr_tuners_cmaes.Rd
Subclass for Covariance Matrix Adaptation Evolution Strategy (CMA-ES).
Calls adagio::pureCMAES()
from package adagio.
Control Parameters
start_values
character(1)
Createrandom
start values or based oncenter
of search space? In the latter case, it is the center of the parameters before a trafo is applied.
For the meaning of the control parameters, see adagio::pureCMAES()
.
Note that we have removed all control parameters which refer to the termination of the algorithm and where our terminators allow to obtain the same behavior.
Progress Bars
$optimize()
supports progress bars via the package progressr
combined with a Terminator. Simply wrap the function in
progressr::with_progress()
to enable them. We recommend to use package
progress as backend; enable with progressr::handlers("progress")
.
Logging
All Tuners use a logger (as implemented in lgr) from package
bbotk.
Use lgr::get_logger("bbotk")
to access and control the logger.
Optimizer
This Tuner is based on bbotk::OptimizerCmaes which can be applied on any black box optimization problem. See also the documentation of bbotk.
Resources
There are several sections about hyperparameter optimization in the mlr3book.
Learn more about tuners.
The gallery features a collection of case studies and demos about optimization.
Use the Hyperband optimizer with different budget parameters.
Super classes
mlr3tuning::Tuner
-> mlr3tuning::TunerFromOptimizer
-> TunerCmaes
Examples
# Hyperparameter Optimization
# load learner and set search space
learner = lrn("classif.rpart",
cp = to_tune(1e-04, 1e-1, logscale = TRUE),
minsplit = to_tune(p_dbl(2, 128, trafo = as.integer)),
minbucket = to_tune(p_dbl(1, 64, trafo = as.integer))
)
# run hyperparameter tuning on the Palmer Penguins data set
instance = tune(
tuner = tnr("cmaes"),
task = tsk("penguins"),
learner = learner,
resampling = rsmp("holdout"),
measure = msr("classif.ce"),
term_evals = 10)
# best performing hyperparameter configuration
instance$result
#> cp minsplit minbucket learner_param_vals x_domain classif.ce
#> 1: -4.318533 96.99061 46.16052 <list[4]> <list[3]> 0.06086957
# all evaluated hyperparameter configuration
as.data.table(instance$archive)
#> cp minsplit minbucket classif.ce x_domain_cp x_domain_minsplit
#> 1: -5.099205 107.83171 1.000000 0.08695652 0.0061015937 107
#> 2: -3.137953 2.00000 33.319814 0.08695652 0.0433714695 2
#> 3: -9.210340 55.41243 11.843087 0.08695652 0.0001000000 55
#> 4: -7.691137 96.07633 54.174181 0.07826087 0.0004568585 96
#> 5: -2.302585 128.00000 38.764034 0.08695652 0.1000000000 128
#> 6: -4.318533 96.99061 46.160516 0.06086957 0.0133194135 96
#> 7: -2.302585 49.46354 61.350979 0.18260870 0.1000000000 49
#> 8: -7.638317 86.16346 49.889635 0.06956522 0.0004816384 86
#> 9: -2.302585 128.00000 64.000000 0.24347826 0.1000000000 128
#> 10: -7.238470 113.92537 9.580762 0.08695652 0.0007184104 113
#> x_domain_minbucket runtime_learners timestamp batch_nr warnings
#> 1: 1 0.008 2023-03-09 07:21:39 1 0
#> 2: 33 0.008 2023-03-09 07:21:39 2 0
#> 3: 11 0.008 2023-03-09 07:21:39 3 0
#> 4: 54 0.007 2023-03-09 07:21:39 4 0
#> 5: 38 0.008 2023-03-09 07:21:39 5 0
#> 6: 46 0.007 2023-03-09 07:21:39 6 0
#> 7: 61 0.007 2023-03-09 07:21:39 7 0
#> 8: 49 0.008 2023-03-09 07:21:39 8 0
#> 9: 64 0.006 2023-03-09 07:21:39 9 0
#> 10: 9 0.006 2023-03-09 07:21:39 10 0
#> errors resample_result
#> 1: 0 <ResampleResult[21]>
#> 2: 0 <ResampleResult[21]>
#> 3: 0 <ResampleResult[21]>
#> 4: 0 <ResampleResult[21]>
#> 5: 0 <ResampleResult[21]>
#> 6: 0 <ResampleResult[21]>
#> 7: 0 <ResampleResult[21]>
#> 8: 0 <ResampleResult[21]>
#> 9: 0 <ResampleResult[21]>
#> 10: 0 <ResampleResult[21]>
# fit final model on complete data set
learner$param_set$values = instance$result_learner_param_vals
learner$train(tsk("penguins"))