mlr3tuningspaces

Package website: release

Collection of search spaces for hyperparameter tuning. Includes various search spaces that can be directly applied to an mlr3 learner. Additionally, meta information about the search space can be queried.

Installation

Install the development version from GitHub:

remotes::install_github("mlr-org/mlr3tuningspaces")

Example

Quick tuning

library(mlr3tuningspaces)

# tune learner with default search space
instance = tune(
  method = "random_search",
  task = tsk("pima"),
  learner = lts(lrn("classif.rpart")),
  resampling = rsmp ("holdout"),
  measure = msr("classif.ce"),
  term_evals = 10
)

# best performing hyperparameter configuration
instance$result

##    minsplit minbucket        cp learner_param_vals  x_domain classif.ce
## 1: 4.174471 0.5070691 -4.542023          <list[4]> <list[3]>  0.1953125

Tuning search spaces

library("data.table")

# print keys and learners
as.data.table(mlr_tuning_spaces)

##                         key         learner n_values
##  1:     classif.glmnet.rbv2  classif.glmnet        2
##  2:       classif.kknn.rbv2    classif.kknn        1
##  3:  classif.ranger.default  classif.ranger        3
##  4:     classif.ranger.rbv2  classif.ranger        7
##  5:   classif.rpart.default   classif.rpart        3
##  6:      classif.rpart.rbv2   classif.rpart        4
##  7:     classif.svm.default     classif.svm        4
##  8:        classif.svm.rbv2     classif.svm        5
##  9: classif.xgboost.default classif.xgboost        9
## 10:    classif.xgboost.rbv2 classif.xgboost       13
## 11:        regr.glmnet.rbv2     regr.glmnet        2
## 12:          regr.kknn.rbv2       regr.kknn        1
## 13:     regr.ranger.default     regr.ranger        3
## 14:        regr.ranger.rbv2     regr.ranger        6
## 15:      regr.rpart.default      regr.rpart        3
## 16:         regr.rpart.rbv2      regr.rpart        4
## 17:        regr.svm.default        regr.svm        4
## 18:           regr.svm.rbv2        regr.svm        5
## 19:    regr.xgboost.default    regr.xgboost        9
## 20:       regr.xgboost.rbv2    regr.xgboost       13

# get tuning space and view tune token
tuning_space = lts("classif.rpart.default")
tuning_space$values

## $minsplit
## Tuning over:
## range [2, 128] (log scale)
## 
## 
## $minbucket
## Tuning over:
## range [1, 64] (log scale)
## 
## 
## $cp
## Tuning over:
## range [1e-04, 0.1] (log scale)

# get learner with tuning space
learner = tuning_space$get_learner()

# tune learner
instance = tune(
  method = "random_search",
  task = tsk("pima"),
  learner = learner,
  resampling = rsmp ("holdout"),
  measure = msr("classif.ce"),
  term_evals = 10)

instance$result

##    minsplit minbucket        cp learner_param_vals  x_domain classif.ce
## 1: 3.009338  2.506336 -8.291878          <list[4]> <list[3]>  0.2421875