Specifies a general single-criteria tuning scenario, including objective function and archive for Tuners to act upon. This class stores an ObjectiveTuning object that encodes the black box objective function which a Tuner has to optimize. It allows the basic operations of querying the objective at design points ($eval_batch()), storing the evaluations in the internal Archive and accessing the final result ($result).

Evaluations of hyperparameter configurations are performed in batches by calling mlr3::benchmark() internally. Before a batch is evaluated, the bbotk::Terminator is queried for the remaining budget. If the available budget is exhausted, an exception is raised, and no further evaluations can be performed from this point on.

The tuner is also supposed to store its final result, consisting of a selected hyperparameter configuration and associated estimated performance values, by calling the method instance$assign_result.

Super classes

bbotk::OptimInstance -> bbotk::OptimInstanceSingleCrit -> TuningInstanceSingleCrit

Active bindings

result_learner_param_vals

(list())
Param values for the optimal learner call.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

This defines the resampled performance of a learner on a task, a feasibility region for the parameters the tuner is supposed to optimize, and a termination criterion.

Usage

TuningInstanceSingleCrit$new(
  task,
  learner,
  resampling,
  measure,
  terminator,
  search_space = NULL,
  store_benchmark_result = TRUE,
  store_models = FALSE,
  check_values = FALSE
)

Arguments

task

(mlr3::Task)
Task to operate on.

learner

(mlr3::Learner).

resampling

(mlr3::Resampling)
Uninstantiated resamplings are instantiated during construction so that all configurations are evaluated on the same data splits. If a new resampling is passed, it is instantiated with new data splits. Already instantiated resamplings are kept unchanged.

measure

(mlr3::Measure)
Measure to optimize.

terminator

(Terminator).

search_space

(paradox::ParamSet).

store_benchmark_result

(logical(1))
If TRUE (default), stores the mlr3::BenchmarkResult in archive.

store_models

(logical(1))
If FALSE (default), the fitted models are not stored in the mlr3::BenchmarkResult. If store_benchmark_result = FALSE, the models are only stored temporarily and not accessible after the tuning. This combination might be useful for measures that require a model.

check_values

(logical(1))
Should parameters before the evaluation and the results be checked for validity?


Method assign_result()

The Tuner object writes the best found point and estimated performance value here. For internal use.

Usage

TuningInstanceSingleCrit$assign_result(xdt, y, learner_param_vals = NULL)

Arguments

xdt

(data.table::data.table())
x values as data.table. Each row is one point. Contains the value in the search space of the TuningInstanceMultiCrit object. Can contain additional columns for extra information.

y

(numeric(1))
Optimal outcome.

learner_param_vals

(list())
Fixed parameter values of the learner that are neither part of the


Method clone()

The objects of this class are cloneable with this method.

Usage

TuningInstanceSingleCrit$clone(deep = FALSE)

Arguments

deep

Whether to make a deep clone.

Examples

library(data.table) library(paradox) library(mlr3) # Objects required to define the performance evaluator: task = tsk("iris") learner = lrn("classif.rpart") resampling = rsmp("holdout") measure = msr("classif.ce") param_set = ParamSet$new(list( ParamDbl$new("cp", lower = 0.001, upper = 0.1), ParamInt$new("minsplit", lower = 1, upper = 10)) ) terminator = trm("evals", n_evals = 5) inst = TuningInstanceSingleCrit$new( task = task, learner = learner, resampling = resampling, measure = measure, search_space = param_set, terminator = terminator ) # first 4 points as cross product design = CJ(cp = c(0.05, 0.01), minsplit = c(5, 3)) inst$eval_batch(design) inst$archive
#> <ArchiveTuning> #> cp minsplit classif.ce timestamp batch_nr #> 1: 0.01 3 0.02 2021-04-11 04:26:29 1 #> 2: 0.01 5 0.02 2021-04-11 04:26:29 1 #> 3: 0.05 3 0.06 2021-04-11 04:26:29 1 #> 4: 0.05 5 0.06 2021-04-11 04:26:29 1
# try more points, catch the raised terminated message tryCatch( inst$eval_batch(data.table(cp = 0.01, minsplit = 7)), terminated_error = function(e) message(as.character(e)) ) # try another point although the budget is now exhausted # -> no extra evaluations tryCatch( inst$eval_batch(data.table(cp = 0.01, minsplit = 9)), terminated_error = function(e) message(as.character(e)) )
#> Error: Objective (obj:classif.rpart_on_iris, term:<TerminatorEvals>) terminated
inst$archive
#> <ArchiveTuning> #> cp minsplit classif.ce timestamp batch_nr #> 1: 0.01 3 0.02 2021-04-11 04:26:29 1 #> 2: 0.01 5 0.02 2021-04-11 04:26:29 1 #> 3: 0.05 3 0.06 2021-04-11 04:26:29 1 #> 4: 0.05 5 0.06 2021-04-11 04:26:29 1 #> 5: 0.01 7 0.02 2021-04-11 04:26:29 2
### Error handling # get a learner which breaks with 50% probability # set encapsulation + fallback learner = lrn("classif.debug", error_train = 0.5) learner$encapsulate = c(train = "evaluate", predict = "evaluate") learner$fallback = lrn("classif.featureless") param_set = ParamSet$new(list( ParamDbl$new("x", lower = 0, upper = 1) )) inst = TuningInstanceSingleCrit$new( task = tsk("wine"), learner = learner, resampling = rsmp("cv", folds = 3), measure = msr("classif.ce"), search_space = param_set, terminator = trm("evals", n_evals = 5) ) tryCatch( inst$eval_batch(data.table(x = 1:5 / 5)), terminated_error = function(e) message(as.character(e)) ) archive = as.data.table(inst$archive) # column errors: multiple errors recorded print(archive)
#> x classif.ce uhash timestamp #> 1: 0.2 0.6908663 86a32799-f292-404a-a240-189992530293 2021-04-11 04:26:30 #> 2: 0.4 0.6011299 473bc8e5-179e-4d5e-8a6a-a638a6bbf9f6 2021-04-11 04:26:30 #> 3: 0.6 0.6011299 26387fd0-91c7-4f22-9821-af5a962dc367 2021-04-11 04:26:30 #> 4: 0.8 0.6011299 5788231a-5415-4ce9-ae39-6d601e61afc5 2021-04-11 04:26:30 #> 5: 1.0 0.6739171 1438a109-e3f7-4445-a45e-794563dd64ea 2021-04-11 04:26:30 #> batch_nr x_domain_x #> 1: 1 0.2 #> 2: 1 0.4 #> 3: 1 0.6 #> 4: 1 0.8 #> 5: 1 1.0