Hyperparameter Tuning with Iterated Racing.

Subclass for iterated racing. Calls irace::irace() from package irace.

Source

Lopez-Ibanez M, Dubois-Lacoste J, Caceres LP, Birattari M, Stuetzle T (2016). “The irace package: Iterated racing for automatic algorithm configuration.” Operations Research Perspectives, 3, 43–58. doi:10.1016/j.orp.2016.09.002 .

Dictionary

This Tuner can be instantiated with the associated sugar function tnr():

tnr("irace")

Control Parameters

n_instances: integer(1)
Number of resampling instances.

For the meaning of all other parameters, see irace::defaultScenario(). Note that we have removed all control parameters which refer to the termination of the algorithm. Use bbotk::TerminatorEvals instead. Other terminators do not work with TunerIrace.

Result

The tuning result (instance$result) is the best-performing elite of the final race. The reported performance is the average performance estimated on all used instances.

Progress Bars

$optimize() supports progress bars via the package progressr combined with a bbotk::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::OptimizerBatchIrace 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.

Getting started with hyperparameter optimization.
An overview of all tuners can be found on our website.
Tune a support vector machine on the Sonar data set.
Learn about tuning spaces.
Estimate the model performance with nested resampling.
Learn about multi-objective optimization.
Simultaneously optimize hyperparameters and use early stopping with XGBoost.
Automate the tuning.

The gallery features a collection of case studies and demos about optimization.

Learn more advanced methods with the Practical Tuning Series.
Learn about hotstarting models.
Run the default hyperparameter configuration of learners as a baseline.
Use the Hyperband optimizer with different budget parameters.

The cheatsheet summarizes the most important functions of mlr3tuning.

Super classes

mlr3tuning::Tuner -> mlr3tuning::TunerBatch -> mlr3tuning::TunerBatchFromOptimizerBatch -> TunerBatchIrace

Methods

Inherited methods

Method `new()`

Creates a new instance of this R6 class.

Usage

TunerBatchIrace$new()

Method `optimize()`

Performs the tuning on a TuningInstanceBatchSingleCrit until termination. The single evaluations and the final results will be written into the ArchiveBatchTuning that resides in the TuningInstanceBatchSingleCrit. The final result is returned.

Usage

TunerBatchIrace$optimize(inst)

Arguments

inst: (TuningInstanceBatchSingleCrit).

Returns

data.table::data.table.

Method `clone()`

The objects of this class are cloneable with this method.

Usage

TunerBatchIrace$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

# retrieve task
task = tsk("pima")

# load learner and set search space
learner = lrn("classif.rpart", cp = to_tune(1e-04, 1e-1, logscale = TRUE))

# runtime of the example is too long
# \donttest{
# hyperparameter tuning on the pima indians diabetes data set
instance = tune(
  tuner = tnr("irace"),
  task = task,
  learner = learner,
  resampling = rsmp("holdout"),
  measure = msr("classif.ce"),
  term_evals = 200
)
#> # 2024-12-18 10:09:57 UTC: Initialization
#> # Elitist race
#> # Elitist new instances: 1
#> # Elitist limit: 2
#> # nbIterations: 2
#> # minNbSurvival: 2
#> # nbParameters: 1
#> # seed: 1855097766
#> # confidence level: 0.95
#> # budget: 200
#> # mu: 5
#> # deterministic: FALSE
#> 
#> # 2024-12-18 10:09:57 UTC: Iteration 1 of 2
#> # experimentsUsedSoFar: 0
#> # remainingBudget: 200
#> # currentBudget: 100
#> # nbConfigurations: 16
#> # Markers:
#>      x No test is performed.
#>      c Configurations are discarded only due to capping.
#>      - The test is performed and some configurations are discarded.
#>      = The test is performed but no configuration is discarded.
#>      ! The test is performed and configurations could be discarded but elite configurations are preserved.
#>      . All alive configurations are elite and nothing is discarded.
#> 
#> +-+-----------+-----------+-----------+----------------+-----------+--------+-----+----+------+
#> | |   Instance|      Alive|       Best|       Mean best| Exp so far|  W time|  rho|KenW|  Qvar|
#> +-+-----------+-----------+-----------+----------------+-----------+--------+-----+----+------+
#> |x|          1|         16|          2|    0.2460937500|         16|00:00:00|   NA|  NA|    NA|
#> |x|          2|         16|          5|    0.2539062500|         32|00:00:00|+0.85|0.93|0.2432|
#> |x|          3|         16|          5|    0.2513020833|         48|00:00:00|+0.72|0.81|0.5409|
#> |x|          4|         16|          5|    0.2421875000|         64|00:00:00|+0.63|0.73|0.4926|
#> |-|          5|          6|          5|    0.2398437500|         80|00:00:00|+0.01|0.21|0.8365|
#> |=|          6|          6|          2|    0.2513020833|         86|00:00:00|-0.12|0.07|0.9438|
#> |=|          7|          6|          5|    0.2483258929|         92|00:00:00|-0.01|0.13|0.8680|
#> +-+-----------+-----------+-----------+----------------+-----------+--------+-----+----+------+
#> Best-so-far configuration:           5    mean value:     0.2483258929
#> Description of the best-so-far configuration:
#>   .ID.                cp .PARENT.
#> 5    5 -2.86582979189387       NA
#> 
#> # 2024-12-18 10:09:59 UTC: Elite configurations (first number is the configuration ID; listed from best to worst according to the sum of ranks):
#>                  cp
#> 5 -2.86582979189387
#> 2 -3.72929920176664
#> # 2024-12-18 10:09:59 UTC: Iteration 2 of 2
#> # experimentsUsedSoFar: 92
#> # remainingBudget: 108
#> # currentBudget: 108
#> # nbConfigurations: 15
#> # Markers:
#>      x No test is performed.
#>      c Configurations are discarded only due to capping.
#>      - The test is performed and some configurations are discarded.
#>      = The test is performed but no configuration is discarded.
#>      ! The test is performed and configurations could be discarded but elite configurations are preserved.
#>      . All alive configurations are elite and nothing is discarded.
#> 
#> +-+-----------+-----------+-----------+----------------+-----------+--------+-----+----+------+
#> | |   Instance|      Alive|       Best|       Mean best| Exp so far|  W time|  rho|KenW|  Qvar|
#> +-+-----------+-----------+-----------+----------------+-----------+--------+-----+----+------+
#> |x|          8|         15|         24|    0.2695312500|         15|00:00:00|   NA|  NA|    NA|
#> |x|          4|         15|         24|    0.2421875000|         28|00:00:00|+0.00|0.50|0.4667|
#> |x|          3|         15|         24|    0.2434895833|         41|00:00:00|+0.00|0.33|0.3111|
#> |x|          6|         15|         24|    0.2509765625|         54|00:00:00|-0.13|0.16|0.5342|
#> |=|          5|         15|         24|    0.2460937500|         67|00:00:00|-0.00|0.20|0.5544|
#> |=|          7|         15|         24|    0.2486979167|         80|00:00:00|-0.02|0.15|0.6385|
#> |=|          2|         15|         18|    0.2488839286|         93|00:00:00|+0.04|0.18|0.6580|
#> |=|          1|         15|         18|    0.2495117188|        106|00:00:00|-0.01|0.12|0.7102|
#> +-+-----------+-----------+-----------+----------------+-----------+--------+-----+----+------+
#> Best-so-far configuration:          18    mean value:     0.2495117188
#> Description of the best-so-far configuration:
#>    .ID.                cp .PARENT.
#> 18   18 -3.03505397360393        5
#> 
#> # 2024-12-18 10:10:02 UTC: Elite configurations (first number is the configuration ID; listed from best to worst according to the sum of ranks):
#>                   cp
#> 18 -3.03505397360393
#> 26 -3.07347033861101
#> # 2024-12-18 10:10:02 UTC: Stopped because there is not enough budget left to race more than the minimum (2).
#> # You may either increase the budget or set 'minNbSurvival' to a lower value.
#> # Iteration: 3
#> # nbIterations: 3
#> # experimentsUsedSoFar: 198
#> # timeUsed: 0
#> # remainingBudget: 2
#> # currentBudget: 2
#> # number of elites: 2
#> # nbConfigurations: 2
#> # Total CPU user time: 5.342, CPU sys time: 0.104, Wall-clock time: 5.457
#> # 2024-12-18 10:10:02 UTC: Starting post-selection:
#> # Configurations selected: 18, 26.
#> # Pending instances: 0, 0.
#> # Seed: 1855097766
#> # Configurations: 2
#> # Available experiments: 2
#> # minSurvival: 1
#> # Markers:
#>      x No test is performed.
#>      c Configurations are discarded only due to capping.
#>      - The test is performed and some configurations are discarded.
#>      = The test is performed but no configuration is discarded.
#>      ! The test is performed and configurations could be discarded but elite configurations are preserved.
#>      . All alive configurations are elite and nothing is discarded.
#> 
#> +-+-----------+-----------+-----------+----------------+-----------+--------+-----+----+------+
#> | |   Instance|      Alive|       Best|       Mean best| Exp so far|  W time|  rho|KenW|  Qvar|
#> +-+-----------+-----------+-----------+----------------+-----------+--------+-----+----+------+
#> |.|          1|          2|         18|    0.2539062500|          0|00:00:00|   NA|  NA|    NA|
#> |.|          8|          2|         18|    0.2636718750|          0|00:00:00|+1.00|1.00|0.0000|
#> |.|          4|          2|         18|    0.2473958333|          0|00:00:00|+1.00|1.00|0.0000|
#> |.|          3|          2|         18|    0.2470703125|          0|00:00:00|+1.00|1.00|0.0000|
#> |.|          7|          2|         18|    0.2476562500|          0|00:00:00|+1.00|1.00|0.0000|
#> |.|          2|          2|         18|    0.2486979167|          0|00:00:00|+1.00|1.00|0.0000|
#> |.|          5|          2|         18|    0.2460937500|          0|00:00:00|+1.00|1.00|0.0000|
#> |.|          6|          2|         18|    0.2495117188|          0|00:00:00|+1.00|1.00|0.0000|
#> |-|          9|          1|         18|    0.2513020833|          2|00:00:00|   NA|  NA|    NA|
#> +-+-----------+-----------+-----------+----------------+-----------+--------+-----+----+------+
#> Best-so-far configuration:          18    mean value:     0.2513020833
#> Description of the best-so-far configuration:
#>    .ID.                cp .PARENT.
#> 18   18 -3.03505397360393        5
#> 
#> # 2024-12-18 10:10:02 UTC: Elite configurations (first number is the configuration ID; listed from best to worst according to the sum of ranks):
#>                   cp
#> 18 -3.03505397360393
#> # Total CPU user time: 5.438, CPU sys time: 0.108, Wall-clock time: 5.556

# best performing hyperparameter configuration
instance$result
#>           cp configuration learner_param_vals  x_domain classif.ce
#>        <num>         <int>             <list>    <list>      <num>
#> 1: -3.035054            18          <list[2]> <list[1]>  0.2513021

# all evaluated hyperparameter configuration
as.data.table(instance$archive)
#>             cp classif.ce runtime_learners           timestamp  race  step
#>          <num>      <num>            <num>              <POSc> <int> <int>
#>   1: -7.183177  0.3085938            0.007 2024-12-18 10:09:57     1     1
#>   2: -3.729299  0.2460938            0.006 2024-12-18 10:09:57     1     1
#>   3: -8.910116  0.3085938            0.007 2024-12-18 10:09:57     1     1
#>   4: -5.456238  0.3085938            0.007 2024-12-18 10:09:57     1     1
#>   5: -2.865830  0.2539062            0.007 2024-12-18 10:09:57     1     1
#>  ---                                                                      
#> 196: -3.782776  0.2578125            0.006 2024-12-18 10:10:02     2     1
#> 197: -2.949281  0.2539062            0.006 2024-12-18 10:10:02     2     1
#> 198: -2.692814  0.2812500            0.006 2024-12-18 10:10:02     2     1
#> 199: -3.035054  0.2656250            0.007 2024-12-18 10:10:02     3     1
#> 200: -3.073470  0.2656250            0.006 2024-12-18 10:10:02     3     1
#>      instance configuration warnings errors  x_domain batch_nr  resample_result
#>         <int>         <int>    <int>  <int>    <list>    <int>           <list>
#>   1:       10             1        0      0 <list[1]>        1 <ResampleResult>
#>   2:       10             2        0      0 <list[1]>        1 <ResampleResult>
#>   3:       10             3        0      0 <list[1]>        1 <ResampleResult>
#>   4:       10             4        0      0 <list[1]>        1 <ResampleResult>
#>   5:       10             5        0      0 <list[1]>        1 <ResampleResult>
#>  ---                                                                           
#> 196:       10            27        0      0 <list[1]>       15 <ResampleResult>
#> 197:       10            28        0      0 <list[1]>       15 <ResampleResult>
#> 198:       10            29        0      0 <list[1]>       15 <ResampleResult>
#> 199:        2            18        0      0 <list[1]>       16 <ResampleResult>
#> 200:        2            26        0      0 <list[1]>       16 <ResampleResult>

# fit final model on complete data set
learner$param_set$values = instance$result_learner_param_vals
learner$train(task)
# }

Source

Dictionary

Control Parameters

Archive

Result

Progress Bars

Logging

Optimizer

Resources

See also

Super classes

Methods

Public methods

Method new()

Usage

Method optimize()

Usage

Arguments

Returns

Method clone()

Usage

Arguments

Examples

Method `new()`

Method `optimize()`

Method `clone()`