Package 'mlr3learners'

Title: Recommended Learners for 'mlr3'
Description: Recommended Learners for 'mlr3'. Extends 'mlr3' with interfaces to essential machine learning packages on CRAN. This includes, but is not limited to: (penalized) linear and logistic regression, linear and quadratic discriminant analysis, k-nearest neighbors, naive Bayes, support vector machines, and gradient boosting.
Authors: Michel Lang [aut] , Quay Au [aut] , Stefan Coors [aut] , Patrick Schratz [aut] , Marc Becker [cre, aut]
Maintainer: Marc Becker <[email protected]>
License: LGPL-3
Version: 0.9.0
Built: 2024-11-25 16:13:08 UTC
Source: CRAN

Help Index

mlr3learners: Recommended Learners for 'mlr3'

Description

More learners are implemented in the mlr3extralearners package. A guide on how to create custom learners is covered in the book: https://mlr3book.mlr-org.com. Feel invited to contribute a missing learner to the mlr3 ecosystem!

Author(s)

Maintainer: Marc Becker [email protected] (ORCID)

Authors:

See Also

Useful links:

GLM with Elastic Net Regularization Classification Learner

Description

Generalized linear models with elastic net regularization. Calls glmnet::cv.glmnet() from package glmnet.

The default for hyperparameter family is set to "binomial" or "multinomial", depending on the number of classes.

Dictionary

This mlr3::Learner can be instantiated via the dictionary mlr3::mlr_learners or with the associated sugar function mlr3::lrn(): 

mlr_learners$get("classif.cv_glmnet")
lrn("classif.cv_glmnet")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”

  • Required Packages: mlr3, mlr3learners, glmnet

Parameters

Id Type Default Levels Range
alignment character lambda lambda, fraction -
alpha numeric 1 [0,1][0, 1]
big numeric 9.9e+35 (−∞,∞)(-\infty, \infty)
devmax numeric 0.999 [0,1][0, 1]
dfmax integer - [0,∞)[0, \infty)
epsnr numeric 1e-08 [0,1][0, 1]
eps numeric 1e-06 [0,1][0, 1]
exclude integer - [1,∞)[1, \infty)
exmx numeric 250 (−∞,∞)(-\infty, \infty)
fdev numeric 1e-05 [0,1][0, 1]
foldid untyped NULL -
gamma untyped - -
grouped logical TRUE TRUE, FALSE -
intercept logical TRUE TRUE, FALSE -
keep logical FALSE TRUE, FALSE -
lambda.min.ratio numeric - [0,1][0, 1]
lambda untyped - -
lower.limits untyped - -
maxit integer 100000 [1,∞)[1, \infty)
mnlam integer 5 [1,∞)[1, \infty)
mxitnr integer 25 [1,∞)[1, \infty)
mxit integer 100 [1,∞)[1, \infty)
nfolds integer 10 [3,∞)[3, \infty)
nlambda integer 100 [1,∞)[1, \infty)
offset untyped NULL -
parallel logical FALSE TRUE, FALSE -
penalty.factor untyped - -
pmax integer - [0,∞)[0, \infty)
pmin numeric 1e-09 [0,1][0, 1]
prec numeric 1e-10 (−∞,∞)(-\infty, \infty)
predict.gamma numeric gamma.1se (−∞,∞)(-\infty, \infty)
relax logical FALSE TRUE, FALSE -
s numeric lambda.1se [0,∞)[0, \infty)
standardize logical TRUE TRUE, FALSE -
standardize.response logical FALSE TRUE, FALSE -
thresh numeric 1e-07 [0,∞)[0, \infty)
trace.it integer 0 [0,1][0, 1]
type.gaussian character - covariance, naive -
type.logistic character - Newton, modified.Newton -
type.measure character deviance deviance, class, auc, mse, mae -
type.multinomial character - ungrouped, grouped -
upper.limits untyped - -

Internal Encoding

Starting with mlr3 v0.5.0, the order of class labels is reversed prior to model fitting to comply to the stats::glm() convention that the negative class is provided as the first factor level.

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifCVGlmnet

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifCVGlmnet$new()

Method selected_features()

Returns the set of selected features as reported by glmnet::predict.glmnet() with type set to "nonzero".

Usage
LearnerClassifCVGlmnet$selected_features(lambda = NULL)
Arguments
lambda

(numeric(1))
Custom lambda, defaults to the active lambda depending on parameter set.

Returns

(character()) of feature names.

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifCVGlmnet$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

References

Friedman J, Hastie T, Tibshirani R (2010). “Regularization Paths for Generalized Linear Models via Coordinate Descent.” Journal of Statistical Software, 33(1), 1–22. doi:10.18637/jss.v033.i01.

See Also

Other Learner: mlr_learners_classif.glmnet, mlr_learners_classif.kknn, mlr_learners_classif.lda, mlr_learners_classif.log_reg, mlr_learners_classif.multinom, mlr_learners_classif.naive_bayes, mlr_learners_classif.nnet, mlr_learners_classif.qda, mlr_learners_classif.ranger, mlr_learners_classif.svm, mlr_learners_classif.xgboost, mlr_learners_regr.cv_glmnet, mlr_learners_regr.glmnet, mlr_learners_regr.kknn, mlr_learners_regr.km, mlr_learners_regr.lm, mlr_learners_regr.nnet, mlr_learners_regr.ranger, mlr_learners_regr.svm, mlr_learners_regr.xgboost

Examples

if (requireNamespace("glmnet", quietly = TRUE)) {
# Define the Learner and set parameter values
learner = lrn("classif.cv_glmnet")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

# importance method
if("importance" %in% learner$properties) print(learner$importance)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
}

GLM with Elastic Net Regularization Classification Learner

Description

Generalized linear models with elastic net regularization. Calls glmnet::glmnet() from package glmnet.

Details

Caution: This learner is different to learners calling glmnet::cv.glmnet() in that it does not use the internal optimization of parameter lambda. Instead, lambda needs to be tuned by the user (e.g., via mlr3tuning). When lambda is tuned, the glmnet will be trained for each tuning iteration. While fitting the whole path of lambdas would be more efficient, as is done by default in glmnet::glmnet(), tuning/selecting the parameter at prediction time (using parameter s) is currently not supported in mlr3 (at least not in efficient manner). Tuning the s parameter is, therefore, currently discouraged.

When the data are i.i.d. and efficiency is key, we recommend using the respective auto-tuning counterparts in mlr_learners_classif.cv_glmnet() or mlr_learners_regr.cv_glmnet(). However, in some situations this is not applicable, usually when data are imbalanced or not i.i.d. (longitudinal, time-series) and tuning requires custom resampling strategies (blocked design, stratification).

Dictionary

This mlr3::Learner can be instantiated via the dictionary mlr3::mlr_learners or with the associated sugar function mlr3::lrn(): 

mlr_learners$get("classif.glmnet")
lrn("classif.glmnet")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”

  • Required Packages: mlr3, mlr3learners, glmnet

Parameters

Id Type Default Levels Range
alpha numeric 1 [0,1][0, 1]
big numeric 9.9e+35 (−∞,∞)(-\infty, \infty)
devmax numeric 0.999 [0,1][0, 1]
dfmax integer - [0,∞)[0, \infty)
eps numeric 1e-06 [0,1][0, 1]
epsnr numeric 1e-08 [0,1][0, 1]
exact logical FALSE TRUE, FALSE -
exclude integer - [1,∞)[1, \infty)
exmx numeric 250 (−∞,∞)(-\infty, \infty)
fdev numeric 1e-05 [0,1][0, 1]
gamma numeric 1 (−∞,∞)(-\infty, \infty)
intercept logical TRUE TRUE, FALSE -
lambda untyped - -
lambda.min.ratio numeric - [0,1][0, 1]
lower.limits untyped - -
maxit integer 100000 [1,∞)[1, \infty)
mnlam integer 5 [1,∞)[1, \infty)
mxit integer 100 [1,∞)[1, \infty)
mxitnr integer 25 [1,∞)[1, \infty)
nlambda integer 100 [1,∞)[1, \infty)
newoffset untyped - -
offset untyped NULL -
penalty.factor untyped - -
pmax integer - [0,∞)[0, \infty)
pmin numeric 1e-09 [0,1][0, 1]
prec numeric 1e-10 (−∞,∞)(-\infty, \infty)
relax logical FALSE TRUE, FALSE -
s numeric 0.01 [0,∞)[0, \infty)
standardize logical TRUE TRUE, FALSE -
standardize.response logical FALSE TRUE, FALSE -
thresh numeric 1e-07 [0,∞)[0, \infty)
trace.it integer 0 [0,1][0, 1]
type.gaussian character - covariance, naive -
type.logistic character - Newton, modified.Newton -
type.multinomial character - ungrouped, grouped -
upper.limits untyped - -

Internal Encoding

Starting with mlr3 v0.5.0, the order of class labels is reversed prior to model fitting to comply to the stats::glm() convention that the negative class is provided as the first factor level.

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifGlmnet

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifGlmnet$new()

Method selected_features()

Returns the set of selected features as reported by glmnet::predict.glmnet() with type set to "nonzero".

Usage
LearnerClassifGlmnet$selected_features(lambda = NULL)
Arguments
lambda

(numeric(1))
Custom lambda, defaults to the active lambda depending on parameter set.

Returns

(character()) of feature names.

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifGlmnet$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

References

Friedman J, Hastie T, Tibshirani R (2010). “Regularization Paths for Generalized Linear Models via Coordinate Descent.” Journal of Statistical Software, 33(1), 1–22. doi:10.18637/jss.v033.i01.

See Also

Other Learner: mlr_learners_classif.cv_glmnet, mlr_learners_classif.kknn, mlr_learners_classif.lda, mlr_learners_classif.log_reg, mlr_learners_classif.multinom, mlr_learners_classif.naive_bayes, mlr_learners_classif.nnet, mlr_learners_classif.qda, mlr_learners_classif.ranger, mlr_learners_classif.svm, mlr_learners_classif.xgboost, mlr_learners_regr.cv_glmnet, mlr_learners_regr.glmnet, mlr_learners_regr.kknn, mlr_learners_regr.km, mlr_learners_regr.lm, mlr_learners_regr.nnet, mlr_learners_regr.ranger, mlr_learners_regr.svm, mlr_learners_regr.xgboost

Examples

if (requireNamespace("glmnet", quietly = TRUE)) {
# Define the Learner and set parameter values
learner = lrn("classif.glmnet")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

# importance method
if("importance" %in% learner$properties) print(learner$importance)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
}

k-Nearest-Neighbor Classification Learner

Description

k-Nearest-Neighbor classification. Calls kknn::kknn() from package kknn.

Initial parameter values

  • store_model:

    • See note.

Dictionary

This mlr3::Learner can be instantiated via the dictionary mlr3::mlr_learners or with the associated sugar function mlr3::lrn(): 

mlr_learners$get("classif.kknn")
lrn("classif.kknn")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3learners, kknn

Parameters

Id Type Default Levels Range
k integer 7 [1,∞)[1, \infty)
distance numeric 2 [0,∞)[0, \infty)
kernel character optimal rectangular, triangular, epanechnikov, biweight, triweight, cos, inv, gaussian, rank, optimal -
scale logical TRUE TRUE, FALSE -
ykernel untyped NULL -
store_model logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifKKNN

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifKKNN$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifKKNN$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Note

There is no training step for k-NN models, just storing the training data to process it during the predict step. Therefore, ⁠$model⁠ returns a list with the following elements:

  • formula: Formula for calling kknn::kknn() during ⁠$predict()⁠.

  • data: Training data for calling kknn::kknn() during ⁠$predict()⁠.

  • pv: Training parameters for calling kknn::kknn() during ⁠$predict()⁠.

  • kknn: Model as returned by kknn::kknn(), only available after ⁠$predict()⁠ has been called. This is not stored by default, you must set hyperparameter store_model to TRUE.

References

Hechenbichler, Klaus, Schliep, Klaus (2004). “Weighted k-nearest-neighbor techniques and ordinal classification.” Technical Report Discussion Paper 399, SFB 386, Ludwig-Maximilians University Munich. doi:10.5282/ubm/epub.1769.

Samworth, J R (2012). “Optimal weighted nearest neighbour classifiers.” The Annals of Statistics, 40(5), 2733–2763. doi:10.1214/12-AOS1049.

Cover, Thomas, Hart, Peter (1967). “Nearest neighbor pattern classification.” IEEE transactions on information theory, 13(1), 21–27. doi:10.1109/TIT.1967.1053964.

See Also

Other Learner: mlr_learners_classif.cv_glmnet, mlr_learners_classif.glmnet, mlr_learners_classif.lda, mlr_learners_classif.log_reg, mlr_learners_classif.multinom, mlr_learners_classif.naive_bayes, mlr_learners_classif.nnet, mlr_learners_classif.qda, mlr_learners_classif.ranger, mlr_learners_classif.svm, mlr_learners_classif.xgboost, mlr_learners_regr.cv_glmnet, mlr_learners_regr.glmnet, mlr_learners_regr.kknn, mlr_learners_regr.km, mlr_learners_regr.lm, mlr_learners_regr.nnet, mlr_learners_regr.ranger, mlr_learners_regr.svm, mlr_learners_regr.xgboost

Examples

if (requireNamespace("kknn", quietly = TRUE)) {
# Define the Learner and set parameter values
learner = lrn("classif.kknn")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

# importance method
if("importance" %in% learner$properties) print(learner$importance)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
}

Linear Discriminant Analysis Classification Learner

Description

Linear discriminant analysis. Calls MASS::lda() from package MASS.

Details

Parameters method and prior exist for training and prediction but accept different values for each. Therefore, arguments for the predict stage have been renamed to predict.method and predict.prior, respectively.

Dictionary

This mlr3::Learner can be instantiated via the dictionary mlr3::mlr_learners or with the associated sugar function mlr3::lrn(): 

mlr_learners$get("classif.lda")
lrn("classif.lda")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3learners, MASS

Parameters

Id Type Default Levels Range
dimen untyped - -
method character moment moment, mle, mve, t -
nu integer - (−∞,∞)(-\infty, \infty)
predict.method character plug-in plug-in, predictive, debiased -
predict.prior untyped - -
prior untyped - -
tol numeric - (−∞,∞)(-\infty, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifLDA

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifLDA$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifLDA$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

References

Venables WN, Ripley BD (2002). Modern Applied Statistics with S, Fourth edition. Springer, New York. ISBN 0-387-95457-0, http://www.stats.ox.ac.uk/pub/MASS4/.

See Also

Other Learner: mlr_learners_classif.cv_glmnet, mlr_learners_classif.glmnet, mlr_learners_classif.kknn, mlr_learners_classif.log_reg, mlr_learners_classif.multinom, mlr_learners_classif.naive_bayes, mlr_learners_classif.nnet, mlr_learners_classif.qda, mlr_learners_classif.ranger, mlr_learners_classif.svm, mlr_learners_classif.xgboost, mlr_learners_regr.cv_glmnet, mlr_learners_regr.glmnet, mlr_learners_regr.kknn, mlr_learners_regr.km, mlr_learners_regr.lm, mlr_learners_regr.nnet, mlr_learners_regr.ranger, mlr_learners_regr.svm, mlr_learners_regr.xgboost

Examples

if (requireNamespace("MASS", quietly = TRUE)) {
# Define the Learner and set parameter values
learner = lrn("classif.lda")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

# importance method
if("importance" %in% learner$properties) print(learner$importance)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
}

Logistic Regression Classification Learner

Description

Classification via logistic regression. Calls stats::glm() with family set to "binomial".

Internal Encoding

Starting with mlr3 v0.5.0, the order of class labels is reversed prior to model fitting to comply to the stats::glm() convention that the negative class is provided as the first factor level.

Initial parameter values

  • model:

    • Actual default: TRUE.

    • Adjusted default: FALSE.

    • Reason for change: Save some memory.

Dictionary

This mlr3::Learner can be instantiated via the dictionary mlr3::mlr_learners or with the associated sugar function mlr3::lrn(): 

mlr_learners$get("classif.log_reg")
lrn("classif.log_reg")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “character”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3learners, 'stats'

Parameters

Id Type Default Levels Range
dispersion untyped NULL -
epsilon numeric 1e-08 (−∞,∞)(-\infty, \infty)
etastart untyped - -
maxit numeric 25 (−∞,∞)(-\infty, \infty)
model logical TRUE TRUE, FALSE -
mustart untyped - -
offset untyped - -
singular.ok logical TRUE TRUE, FALSE -
start untyped NULL -
trace logical FALSE TRUE, FALSE -
x logical FALSE TRUE, FALSE -
y logical TRUE TRUE, FALSE -

Contrasts

To ensure reproducibility, this learner always uses the default contrasts:

Setting the option "contrasts" does not have any effect. Instead, set the respective hyperparameter or use mlr3pipelines to create dummy features.

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifLogReg

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifLogReg$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifLogReg$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

See Also

Other Learner: mlr_learners_classif.cv_glmnet, mlr_learners_classif.glmnet, mlr_learners_classif.kknn, mlr_learners_classif.lda, mlr_learners_classif.multinom, mlr_learners_classif.naive_bayes, mlr_learners_classif.nnet, mlr_learners_classif.qda, mlr_learners_classif.ranger, mlr_learners_classif.svm, mlr_learners_classif.xgboost, mlr_learners_regr.cv_glmnet, mlr_learners_regr.glmnet, mlr_learners_regr.kknn, mlr_learners_regr.km, mlr_learners_regr.lm, mlr_learners_regr.nnet, mlr_learners_regr.ranger, mlr_learners_regr.svm, mlr_learners_regr.xgboost

Examples

if (requireNamespace("stats", quietly = TRUE)) {
# Define the Learner and set parameter values
learner = lrn("classif.log_reg")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

# importance method
if("importance" %in% learner$properties) print(learner$importance)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
}

Multinomial log-linear learner via neural networks

Description

Multinomial log-linear models via neural networks. Calls nnet::multinom() from package nnet.

Dictionary

This mlr3::Learner can be instantiated via the dictionary mlr3::mlr_learners or with the associated sugar function mlr3::lrn(): 

mlr_learners$get("classif.multinom")
lrn("classif.multinom")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3learners, nnet

Parameters

Id Type Default Levels Range
Hess logical FALSE TRUE, FALSE -
abstol numeric 1e-04 (−∞,∞)(-\infty, \infty)
censored logical FALSE TRUE, FALSE -
decay numeric 0 (−∞,∞)(-\infty, \infty)
entropy logical FALSE TRUE, FALSE -
mask untyped - -
maxit integer 100 [1,∞)[1, \infty)
MaxNWts integer 1000 [1,∞)[1, \infty)
model logical FALSE TRUE, FALSE -
linout logical FALSE TRUE, FALSE -
rang numeric 0.7 (−∞,∞)(-\infty, \infty)
reltol numeric 1e-08 (−∞,∞)(-\infty, \infty)
size integer - [1,∞)[1, \infty)
skip logical FALSE TRUE, FALSE -
softmax logical FALSE TRUE, FALSE -
summ character 0 0, 1, 2, 3 -
trace logical TRUE TRUE, FALSE -
Wts untyped - -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifMultinom

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifMultinom$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifMultinom$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

See Also

Other Learner: mlr_learners_classif.cv_glmnet, mlr_learners_classif.glmnet, mlr_learners_classif.kknn, mlr_learners_classif.lda, mlr_learners_classif.log_reg, mlr_learners_classif.naive_bayes, mlr_learners_classif.nnet, mlr_learners_classif.qda, mlr_learners_classif.ranger, mlr_learners_classif.svm, mlr_learners_classif.xgboost, mlr_learners_regr.cv_glmnet, mlr_learners_regr.glmnet, mlr_learners_regr.kknn, mlr_learners_regr.km, mlr_learners_regr.lm, mlr_learners_regr.nnet, mlr_learners_regr.ranger, mlr_learners_regr.svm, mlr_learners_regr.xgboost

Examples

if (requireNamespace("nnet", quietly = TRUE)) {
# Define the Learner and set parameter values
learner = lrn("classif.multinom")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

# importance method
if("importance" %in% learner$properties) print(learner$importance)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
}

Naive Bayes Classification Learner

Description

Naive Bayes classification. Calls e1071::naiveBayes() from package e1071.

Dictionary

This mlr3::Learner can be instantiated via the dictionary mlr3::mlr_learners or with the associated sugar function mlr3::lrn(): 

mlr_learners$get("classif.naive_bayes")
lrn("classif.naive_bayes")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3learners, e1071

Parameters

Id Type Default Range
eps numeric 0 (−∞,∞)(-\infty, \infty)
laplace numeric 0 [0,∞)[0, \infty)
threshold numeric 0.001 (−∞,∞)(-\infty, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifNaiveBayes

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifNaiveBayes$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifNaiveBayes$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

See Also

Other Learner: mlr_learners_classif.cv_glmnet, mlr_learners_classif.glmnet, mlr_learners_classif.kknn, mlr_learners_classif.lda, mlr_learners_classif.log_reg, mlr_learners_classif.multinom, mlr_learners_classif.nnet, mlr_learners_classif.qda, mlr_learners_classif.ranger, mlr_learners_classif.svm, mlr_learners_classif.xgboost, mlr_learners_regr.cv_glmnet, mlr_learners_regr.glmnet, mlr_learners_regr.kknn, mlr_learners_regr.km, mlr_learners_regr.lm, mlr_learners_regr.nnet, mlr_learners_regr.ranger, mlr_learners_regr.svm, mlr_learners_regr.xgboost

Examples

if (requireNamespace("e1071", quietly = TRUE)) {
# Define the Learner and set parameter values
learner = lrn("classif.naive_bayes")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

# importance method
if("importance" %in% learner$properties) print(learner$importance)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
}

Classification Neural Network Learner

Description

Single Layer Neural Network. Calls nnet::nnet.formula() from package nnet.

Note that modern neural networks with multiple layers are connected via package mlr3torch.

Dictionary

This mlr3::Learner can be instantiated via the dictionary mlr3::mlr_learners or with the associated sugar function mlr3::lrn(): 

mlr_learners$get("classif.nnet")
lrn("classif.nnet")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3learners, nnet

Parameters

Id Type Default Levels Range
Hess logical FALSE TRUE, FALSE -
MaxNWts integer 1000 [1,∞)[1, \infty)
Wts untyped - -
abstol numeric 1e-04 (−∞,∞)(-\infty, \infty)
censored logical FALSE TRUE, FALSE -
contrasts untyped NULL -
decay numeric 0 (−∞,∞)(-\infty, \infty)
mask untyped - -
maxit integer 100 [1,∞)[1, \infty)
na.action untyped - -
rang numeric 0.7 (−∞,∞)(-\infty, \infty)
reltol numeric 1e-08 (−∞,∞)(-\infty, \infty)
size integer 3 [0,∞)[0, \infty)
skip logical FALSE TRUE, FALSE -
subset untyped - -
trace logical TRUE TRUE, FALSE -
formula untyped - -

Initial parameter values

  • size:

    • Adjusted default: 3L.

    • Reason for change: no default in nnet().

Custom mlr3 parameters

  • formula: if not provided, the formula is set to task$formula().

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifNnet

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifNnet$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifNnet$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

References

Ripley BD (1996). Pattern Recognition and Neural Networks. Cambridge University Press. doi:10.1017/cbo9780511812651.

See Also

Other Learner: mlr_learners_classif.cv_glmnet, mlr_learners_classif.glmnet, mlr_learners_classif.kknn, mlr_learners_classif.lda, mlr_learners_classif.log_reg, mlr_learners_classif.multinom, mlr_learners_classif.naive_bayes, mlr_learners_classif.qda, mlr_learners_classif.ranger, mlr_learners_classif.svm, mlr_learners_classif.xgboost, mlr_learners_regr.cv_glmnet, mlr_learners_regr.glmnet, mlr_learners_regr.kknn, mlr_learners_regr.km, mlr_learners_regr.lm, mlr_learners_regr.nnet, mlr_learners_regr.ranger, mlr_learners_regr.svm, mlr_learners_regr.xgboost

Examples

if (requireNamespace("nnet", quietly = TRUE)) {
# Define the Learner and set parameter values
learner = lrn("classif.nnet")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

# importance method
if("importance" %in% learner$properties) print(learner$importance)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
}

Quadratic Discriminant Analysis Classification Learner

Description

Quadratic discriminant analysis. Calls MASS::qda() from package MASS.

Details

Parameters method and prior exist for training and prediction but accept different values for each. Therefore, arguments for the predict stage have been renamed to predict.method and predict.prior, respectively.

Dictionary

This mlr3::Learner can be instantiated via the dictionary mlr3::mlr_learners or with the associated sugar function mlr3::lrn(): 

mlr_learners$get("classif.qda")
lrn("classif.qda")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3learners, MASS

Parameters

Id Type Default Levels Range
method character moment moment, mle, mve, t -
nu integer - (−∞,∞)(-\infty, \infty)
predict.method character plug-in plug-in, predictive, debiased -
predict.prior untyped - -
prior untyped - -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifQDA

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifQDA$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifQDA$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

References

Venables WN, Ripley BD (2002). Modern Applied Statistics with S, Fourth edition. Springer, New York. ISBN 0-387-95457-0, http://www.stats.ox.ac.uk/pub/MASS4/.

See Also

Other Learner: mlr_learners_classif.cv_glmnet, mlr_learners_classif.glmnet, mlr_learners_classif.kknn, mlr_learners_classif.lda, mlr_learners_classif.log_reg, mlr_learners_classif.multinom, mlr_learners_classif.naive_bayes, mlr_learners_classif.nnet, mlr_learners_classif.ranger, mlr_learners_classif.svm, mlr_learners_classif.xgboost, mlr_learners_regr.cv_glmnet, mlr_learners_regr.glmnet, mlr_learners_regr.kknn, mlr_learners_regr.km, mlr_learners_regr.lm, mlr_learners_regr.nnet, mlr_learners_regr.ranger, mlr_learners_regr.svm, mlr_learners_regr.xgboost

Examples

if (requireNamespace("MASS", quietly = TRUE)) {
# Define the Learner and set parameter values
learner = lrn("classif.qda")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

# importance method
if("importance" %in% learner$properties) print(learner$importance)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
}

Ranger Classification Learner

Description

Random classification forest. Calls ranger::ranger() from package ranger.

Custom mlr3 parameters

  • mtry:

    • This hyperparameter can alternatively be set via our hyperparameter mtry.ratio as mtry = max(ceiling(mtry.ratio * n_features), 1). Note that mtry and mtry.ratio are mutually exclusive.

Initial parameter values

  • num.threads:

    • Actual default: 2, using two threads, while also respecting environment variable R_RANGER_NUM_THREADS, options(ranger.num.threads = N), or options(Ncpus = N), with precedence in that order.

    • Adjusted value: 1.

    • Reason for change: Conflicting with parallelization via future.

Dictionary

This mlr3::Learner can be instantiated via the dictionary mlr3::mlr_learners or with the associated sugar function mlr3::lrn(): 

mlr_learners$get("classif.ranger")
lrn("classif.ranger")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “character”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3learners, ranger

Parameters

Id Type Default Levels Range
always.split.variables untyped - -
class.weights untyped NULL -
holdout logical FALSE TRUE, FALSE -
importance character - none, impurity, impurity_corrected, permutation -
keep.inbag logical FALSE TRUE, FALSE -
max.depth integer NULL [1,∞)[1, \infty)
min.bucket untyped 1L -
min.node.size untyped NULL -
mtry integer - [1,∞)[1, \infty)
mtry.ratio numeric - [0,1][0, 1]
na.action character na.learn na.learn, na.omit, na.fail -
num.random.splits integer 1 [1,∞)[1, \infty)
node.stats logical FALSE TRUE, FALSE -
num.threads integer 1 [1,∞)[1, \infty)
num.trees integer 500 [1,∞)[1, \infty)
oob.error logical TRUE TRUE, FALSE -
regularization.factor untyped 1 -
regularization.usedepth logical FALSE TRUE, FALSE -
replace logical TRUE TRUE, FALSE -
respect.unordered.factors character - ignore, order, partition -
sample.fraction numeric - [0,1][0, 1]
save.memory logical FALSE TRUE, FALSE -
scale.permutation.importance logical FALSE TRUE, FALSE -
seed integer NULL (−∞,∞)(-\infty, \infty)
split.select.weights untyped NULL -
splitrule character gini gini, extratrees, hellinger -
verbose logical TRUE TRUE, FALSE -
write.forest logical TRUE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifRanger

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifRanger$new()

Method importance()

The importance scores are extracted from the model slot variable.importance. Parameter importance.mode must be set to "impurity", "impurity_corrected", or "permutation"

Usage
LearnerClassifRanger$importance()
Returns

Named numeric().

Method oob_error()

The out-of-bag error, extracted from model slot prediction.error.

Usage
LearnerClassifRanger$oob_error()
Returns

numeric(1).

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifRanger$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

References

Wright, N. M, Ziegler, Andreas (2017). “ranger: A Fast Implementation of Random Forests for High Dimensional Data in C++ and R.” Journal of Statistical Software, 77(1), 1–17. doi:10.18637/jss.v077.i01.

Breiman, Leo (2001). “Random Forests.” Machine Learning, 45(1), 5–32. ISSN 1573-0565, doi:10.1023/A:1010933404324.

See Also

Other Learner: mlr_learners_classif.cv_glmnet, mlr_learners_classif.glmnet, mlr_learners_classif.kknn, mlr_learners_classif.lda, mlr_learners_classif.log_reg, mlr_learners_classif.multinom, mlr_learners_classif.naive_bayes, mlr_learners_classif.nnet, mlr_learners_classif.qda, mlr_learners_classif.svm, mlr_learners_classif.xgboost, mlr_learners_regr.cv_glmnet, mlr_learners_regr.glmnet, mlr_learners_regr.kknn, mlr_learners_regr.km, mlr_learners_regr.lm, mlr_learners_regr.nnet, mlr_learners_regr.ranger, mlr_learners_regr.svm, mlr_learners_regr.xgboost

Examples

if (requireNamespace("ranger", quietly = TRUE)) {
# Define the Learner and set parameter values
learner = lrn("classif.ranger")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

# importance method
if("importance" %in% learner$properties) print(learner$importance)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
}

Support Vector Machine

Description

Support vector machine for classification. Calls e1071::svm() from package e1071.

Dictionary

This mlr3::Learner can be instantiated via the dictionary mlr3::mlr_learners or with the associated sugar function mlr3::lrn(): 

mlr_learners$get("classif.svm")
lrn("classif.svm")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”

  • Required Packages: mlr3, mlr3learners, e1071

Parameters

Id Type Default Levels Range
cachesize numeric 40 (−∞,∞)(-\infty, \infty)
class.weights untyped NULL -
coef0 numeric 0 (−∞,∞)(-\infty, \infty)
cost numeric 1 [0,∞)[0, \infty)
cross integer 0 [0,∞)[0, \infty)
decision.values logical FALSE TRUE, FALSE -
degree integer 3 [1,∞)[1, \infty)
epsilon numeric 0.1 [0,∞)[0, \infty)
fitted logical TRUE TRUE, FALSE -
gamma numeric - [0,∞)[0, \infty)
kernel character radial linear, polynomial, radial, sigmoid -
nu numeric 0.5 (−∞,∞)(-\infty, \infty)
scale untyped TRUE -
shrinking logical TRUE TRUE, FALSE -
tolerance numeric 0.001 [0,∞)[0, \infty)
type character C-classification C-classification, nu-classification -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifSVM

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifSVM$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifSVM$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

References

Cortes, Corinna, Vapnik, Vladimir (1995). “Support-vector networks.” Machine Learning, 20(3), 273–297. doi:10.1007/BF00994018.

See Also

Other Learner: mlr_learners_classif.cv_glmnet, mlr_learners_classif.glmnet, mlr_learners_classif.kknn, mlr_learners_classif.lda, mlr_learners_classif.log_reg, mlr_learners_classif.multinom, mlr_learners_classif.naive_bayes, mlr_learners_classif.nnet, mlr_learners_classif.qda, mlr_learners_classif.ranger, mlr_learners_classif.xgboost, mlr_learners_regr.cv_glmnet, mlr_learners_regr.glmnet, mlr_learners_regr.kknn, mlr_learners_regr.km, mlr_learners_regr.lm, mlr_learners_regr.nnet, mlr_learners_regr.ranger, mlr_learners_regr.svm, mlr_learners_regr.xgboost

Examples

if (requireNamespace("e1071", quietly = TRUE)) {
# Define the Learner and set parameter values
learner = lrn("classif.svm")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

# importance method
if("importance" %in% learner$properties) print(learner$importance)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
}

Extreme Gradient Boosting Classification Learner

Description

eXtreme Gradient Boosting classification. Calls xgboost::xgb.train() from package xgboost.

If not specified otherwise, the evaluation metric is set to the default "logloss" for binary classification problems and set to "mlogloss" for multiclass problems. This was necessary to silence a deprecation warning.

Note that using the watchlist parameter directly will lead to problems when wrapping this mlr3::Learner in a mlr3pipelines GraphLearner as the preprocessing steps will not be applied to the data in the watchlist. See the section Early Stopping and Validation on how to do this.

Initial parameter values

  • nrounds:

    • Actual default: no default.

    • Adjusted default: 1000.

    • Reason for change: Without a default construction of the learner would error. The lightgbm learner has a default of 1000, so we use the same here.

  • nthread:

    • Actual value: Undefined, triggering auto-detection of the number of CPUs.

    • Adjusted value: 1.

    • Reason for change: Conflicting with parallelization via future.

  • verbose:

    • Actual default: 1.

    • Adjusted default: 0.

    • Reason for change: Reduce verbosity.

Early Stopping and Validation

In order to monitor the validation performance during the training, you can set the ⁠$validate⁠ field of the Learner. For information on how to configure the valdiation set, see the Validation section of mlr3::Learner. This validation data can also be used for early stopping, which can be enabled by setting the early_stopping_rounds parameter. The final (or in the case of early stopping best) validation scores can be accessed via ⁠$internal_valid_scores⁠, and the optimal nrounds via ⁠$internal_tuned_values⁠. The internal validation measure can be set via the eval_metric parameter that can be a mlr3::Measure, a function, or a character string for the internal xgboost measures. Using an mlr3::Measure is slower than the internal xgboost measures, but allows to use the same measure for tuning and validation.

Dictionary

This mlr3::Learner can be instantiated via the dictionary mlr3::mlr_learners or with the associated sugar function mlr3::lrn(): 

mlr_learners$get("classif.xgboost")
lrn("classif.xgboost")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”

  • Required Packages: mlr3, mlr3learners, xgboost

Parameters

Id Type Default Levels Range
alpha numeric 0 [0,∞)[0, \infty)
approxcontrib logical FALSE TRUE, FALSE -
base_score numeric 0.5 (−∞,∞)(-\infty, \infty)
base_margin untyped NULL -
booster character gbtree gbtree, gblinear, dart -
callbacks untyped list() -
colsample_bylevel numeric 1 [0,1][0, 1]
colsample_bynode numeric 1 [0,1][0, 1]
colsample_bytree numeric 1 [0,1][0, 1]
device untyped "cpu" -
disable_default_eval_metric logical FALSE TRUE, FALSE -
early_stopping_rounds integer NULL [1,∞)[1, \infty)
eta numeric 0.3 [0,1][0, 1]
eval_metric untyped - -
feature_selector character cyclic cyclic, shuffle, random, greedy, thrifty -
gamma numeric 0 [0,∞)[0, \infty)
grow_policy character depthwise depthwise, lossguide -
interaction_constraints untyped - -
iterationrange untyped - -
lambda numeric 1 [0,∞)[0, \infty)
lambda_bias numeric 0 [0,∞)[0, \infty)
max_bin integer 256 [2,∞)[2, \infty)
max_delta_step numeric 0 [0,∞)[0, \infty)
max_depth integer 6 [0,∞)[0, \infty)
max_leaves integer 0 [0,∞)[0, \infty)
maximize logical NULL TRUE, FALSE -
min_child_weight numeric 1 [0,∞)[0, \infty)
missing numeric NA (−∞,∞)(-\infty, \infty)
monotone_constraints untyped 0 -
nrounds integer - [1,∞)[1, \infty)
normalize_type character tree tree, forest -
nthread integer 1 [1,∞)[1, \infty)
ntreelimit integer NULL [1,∞)[1, \infty)
num_parallel_tree integer 1 [1,∞)[1, \infty)
objective untyped "binary:logistic" -
one_drop logical FALSE TRUE, FALSE -
outputmargin logical FALSE TRUE, FALSE -
predcontrib logical FALSE TRUE, FALSE -
predinteraction logical FALSE TRUE, FALSE -
predleaf logical FALSE TRUE, FALSE -
print_every_n integer 1 [1,∞)[1, \infty)
process_type character default default, update -
rate_drop numeric 0 [0,1][0, 1]
refresh_leaf logical TRUE TRUE, FALSE -
reshape logical FALSE TRUE, FALSE -
seed_per_iteration logical FALSE TRUE, FALSE -
sampling_method character uniform uniform, gradient_based -
sample_type character uniform uniform, weighted -
save_name untyped NULL -
save_period integer NULL [0,∞)[0, \infty)
scale_pos_weight numeric 1 (−∞,∞)(-\infty, \infty)
skip_drop numeric 0 [0,1][0, 1]
strict_shape logical FALSE TRUE, FALSE -
subsample numeric 1 [0,1][0, 1]
top_k integer 0 [0,∞)[0, \infty)
training logical FALSE TRUE, FALSE -
tree_method character auto auto, exact, approx, hist, gpu_hist -
tweedie_variance_power numeric 1.5 [1,2][1, 2]
updater untyped - -
verbose integer 1 [0,2][0, 2]
watchlist untyped NULL -
xgb_model untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifXgboost

Active bindings

internal_valid_scores

(named list() or NULL) The validation scores extracted from model$evaluation_log. If early stopping is activated, this contains the validation scores of the model for the optimal nrounds, otherwise the nrounds for the final model.

internal_tuned_values

(named list() or NULL) If early stopping is activated, this returns a list with nrounds, which is extracted from ⁠$best_iteration⁠ of the model and otherwise NULL.

validate

(numeric(1) or character(1) or NULL) How to construct the internal validation data. This parameter can be either NULL, a ratio, "test", or "predefined".

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifXgboost$new()

Method importance()

The importance scores are calculated with xgboost::xgb.importance().

Usage
LearnerClassifXgboost$importance()
Returns

Named numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifXgboost$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Note

To compute on GPUs, you first need to compile xgboost yourself and link against CUDA. See https://xgboost.readthedocs.io/en/stable/build.html#building-with-gpu-support.

References

Chen, Tianqi, Guestrin, Carlos (2016). “Xgboost: A scalable tree boosting system.” In Proceedings of the 22nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 785–794. ACM. doi:10.1145/2939672.2939785.

See Also

Other Learner: mlr_learners_classif.cv_glmnet, mlr_learners_classif.glmnet, mlr_learners_classif.kknn, mlr_learners_classif.lda, mlr_learners_classif.log_reg, mlr_learners_classif.multinom, mlr_learners_classif.naive_bayes, mlr_learners_classif.nnet, mlr_learners_classif.qda, mlr_learners_classif.ranger, mlr_learners_classif.svm, mlr_learners_regr.cv_glmnet, mlr_learners_regr.glmnet, mlr_learners_regr.kknn, mlr_learners_regr.km, mlr_learners_regr.lm, mlr_learners_regr.nnet, mlr_learners_regr.ranger, mlr_learners_regr.svm, mlr_learners_regr.xgboost

Examples

## Not run: 
if (requireNamespace("xgboost", quietly = TRUE)) {
# Define the Learner and set parameter values
learner = lrn("classif.xgboost")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

# importance method
if("importance" %in% learner$properties) print(learner$importance)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
}

## End(Not run)

## Not run: 
# Train learner with early stopping on spam data set
task = tsk("spam")

# use 30 percent for validation
# Set early stopping parameter
learner = lrn("classif.xgboost",
  nrounds = 100,
  early_stopping_rounds = 10,
  validate = 0.3
)

# Train learner with early stopping
learner$train(task)

# Inspect optimal nrounds and validation performance
learner$internal_tuned_values
learner$internal_valid_scores

## End(Not run)

GLM with Elastic Net Regularization Regression Learner

Description

Generalized linear models with elastic net regularization. Calls glmnet::cv.glmnet() from package glmnet.

The default for hyperparameter family is set to "gaussian".

Dictionary

This mlr3::Learner can be instantiated via the dictionary mlr3::mlr_learners or with the associated sugar function mlr3::lrn(): 

mlr_learners$get("regr.cv_glmnet")
lrn("regr.cv_glmnet")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”

  • Required Packages: mlr3, mlr3learners, glmnet

Parameters

Id Type Default Levels Range
alignment character lambda lambda, fraction -
alpha numeric 1 [0,1][0, 1]
big numeric 9.9e+35 (−∞,∞)(-\infty, \infty)
devmax numeric 0.999 [0,1][0, 1]
dfmax integer - [0,∞)[0, \infty)
eps numeric 1e-06 [0,1][0, 1]
epsnr numeric 1e-08 [0,1][0, 1]
exclude integer - [1,∞)[1, \infty)
exmx numeric 250 (−∞,∞)(-\infty, \infty)
family character gaussian gaussian, poisson -
fdev numeric 1e-05 [0,1][0, 1]
foldid untyped NULL -
gamma untyped - -
grouped logical TRUE TRUE, FALSE -
intercept logical TRUE TRUE, FALSE -
keep logical FALSE TRUE, FALSE -
lambda untyped - -
lambda.min.ratio numeric - [0,1][0, 1]
lower.limits untyped - -
maxit integer 100000 [1,∞)[1, \infty)
mnlam integer 5 [1,∞)[1, \infty)
mxit integer 100 [1,∞)[1, \infty)
mxitnr integer 25 [1,∞)[1, \infty)
nfolds integer 10 [3,∞)[3, \infty)
nlambda integer 100 [1,∞)[1, \infty)
offset untyped NULL -
parallel logical FALSE TRUE, FALSE -
penalty.factor untyped - -
pmax integer - [0,∞)[0, \infty)
pmin numeric 1e-09 [0,1][0, 1]
prec numeric 1e-10 (−∞,∞)(-\infty, \infty)
predict.gamma numeric gamma.1se (−∞,∞)(-\infty, \infty)
relax logical FALSE TRUE, FALSE -
s numeric lambda.1se [0,∞)[0, \infty)
standardize logical TRUE TRUE, FALSE -
standardize.response logical FALSE TRUE, FALSE -
thresh numeric 1e-07 [0,∞)[0, \infty)
trace.it integer 0 [0,1][0, 1]
type.gaussian character - covariance, naive -
type.logistic character - Newton, modified.Newton -
type.measure character deviance deviance, class, auc, mse, mae -
type.multinomial character - ungrouped, grouped -
upper.limits untyped - -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrCVGlmnet

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrCVGlmnet$new()

Method selected_features()

Returns the set of selected features as reported by glmnet::predict.glmnet() with type set to "nonzero".

Usage
LearnerRegrCVGlmnet$selected_features(lambda = NULL)
Arguments
lambda

(numeric(1))
Custom lambda, defaults to the active lambda depending on parameter set.

Returns

(character()) of feature names.

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrCVGlmnet$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

References

Friedman J, Hastie T, Tibshirani R (2010). “Regularization Paths for Generalized Linear Models via Coordinate Descent.” Journal of Statistical Software, 33(1), 1–22. doi:10.18637/jss.v033.i01.

See Also

Other Learner: mlr_learners_classif.cv_glmnet, mlr_learners_classif.glmnet, mlr_learners_classif.kknn, mlr_learners_classif.lda, mlr_learners_classif.log_reg, mlr_learners_classif.multinom, mlr_learners_classif.naive_bayes, mlr_learners_classif.nnet, mlr_learners_classif.qda, mlr_learners_classif.ranger, mlr_learners_classif.svm, mlr_learners_classif.xgboost, mlr_learners_regr.glmnet, mlr_learners_regr.kknn, mlr_learners_regr.km, mlr_learners_regr.lm, mlr_learners_regr.nnet, mlr_learners_regr.ranger, mlr_learners_regr.svm, mlr_learners_regr.xgboost

Examples

if (requireNamespace("glmnet", quietly = TRUE)) {
# Define the Learner and set parameter values
learner = lrn("regr.cv_glmnet")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

# importance method
if("importance" %in% learner$properties) print(learner$importance)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
}

GLM with Elastic Net Regularization Regression Learner

Description

Generalized linear models with elastic net regularization. Calls glmnet::glmnet() from package glmnet.

The default for hyperparameter family is set to "gaussian".

Details

Caution: This learner is different to learners calling glmnet::cv.glmnet() in that it does not use the internal optimization of parameter lambda. Instead, lambda needs to be tuned by the user (e.g., via mlr3tuning). When lambda is tuned, the glmnet will be trained for each tuning iteration. While fitting the whole path of lambdas would be more efficient, as is done by default in glmnet::glmnet(), tuning/selecting the parameter at prediction time (using parameter s) is currently not supported in mlr3 (at least not in efficient manner). Tuning the s parameter is, therefore, currently discouraged.

When the data are i.i.d. and efficiency is key, we recommend using the respective auto-tuning counterparts in mlr_learners_classif.cv_glmnet() or mlr_learners_regr.cv_glmnet(). However, in some situations this is not applicable, usually when data are imbalanced or not i.i.d. (longitudinal, time-series) and tuning requires custom resampling strategies (blocked design, stratification).

Dictionary

This mlr3::Learner can be instantiated via the dictionary mlr3::mlr_learners or with the associated sugar function mlr3::lrn(): 

mlr_learners$get("regr.glmnet")
lrn("regr.glmnet")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”

  • Required Packages: mlr3, mlr3learners, glmnet

Parameters

Id Type Default Levels Range
alignment character lambda lambda, fraction -
alpha numeric 1 [0,1][0, 1]
big numeric 9.9e+35 (−∞,∞)(-\infty, \infty)
devmax numeric 0.999 [0,1][0, 1]
dfmax integer - [0,∞)[0, \infty)
eps numeric 1e-06 [0,1][0, 1]
epsnr numeric 1e-08 [0,1][0, 1]
exact logical FALSE TRUE, FALSE -
exclude integer - [1,∞)[1, \infty)
exmx numeric 250 (−∞,∞)(-\infty, \infty)
family character gaussian gaussian, poisson -
fdev numeric 1e-05 [0,1][0, 1]
gamma numeric 1 (−∞,∞)(-\infty, \infty)
grouped logical TRUE TRUE, FALSE -
intercept logical TRUE TRUE, FALSE -
keep logical FALSE TRUE, FALSE -
lambda untyped - -
lambda.min.ratio numeric - [0,1][0, 1]
lower.limits untyped - -
maxit integer 100000 [1,∞)[1, \infty)
mnlam integer 5 [1,∞)[1, \infty)
mxit integer 100 [1,∞)[1, \infty)
mxitnr integer 25 [1,∞)[1, \infty)
newoffset untyped - -
nlambda integer 100 [1,∞)[1, \infty)
offset untyped NULL -
parallel logical FALSE TRUE, FALSE -
penalty.factor untyped - -
pmax integer - [0,∞)[0, \infty)
pmin numeric 1e-09 [0,1][0, 1]
prec numeric 1e-10 (−∞,∞)(-\infty, \infty)
relax logical FALSE TRUE, FALSE -
s numeric 0.01 [0,∞)[0, \infty)
standardize logical TRUE TRUE, FALSE -
standardize.response logical FALSE TRUE, FALSE -
thresh numeric 1e-07 [0,∞)[0, \infty)
trace.it integer 0 [0,1][0, 1]
type.gaussian character - covariance, naive -
type.logistic character - Newton, modified.Newton -
type.multinomial character - ungrouped, grouped -
upper.limits untyped - -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrGlmnet

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrGlmnet$new()

Method selected_features()

Returns the set of selected features as reported by glmnet::predict.glmnet() with type set to "nonzero".

Usage
LearnerRegrGlmnet$selected_features(lambda = NULL)
Arguments
lambda

(numeric(1))
Custom lambda, defaults to the active lambda depending on parameter set.

Returns

(character()) of feature names.

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrGlmnet$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

References

Friedman J, Hastie T, Tibshirani R (2010). “Regularization Paths for Generalized Linear Models via Coordinate Descent.” Journal of Statistical Software, 33(1), 1–22. doi:10.18637/jss.v033.i01.

See Also

Other Learner: mlr_learners_classif.cv_glmnet, mlr_learners_classif.glmnet, mlr_learners_classif.kknn, mlr_learners_classif.lda, mlr_learners_classif.log_reg, mlr_learners_classif.multinom, mlr_learners_classif.naive_bayes, mlr_learners_classif.nnet, mlr_learners_classif.qda, mlr_learners_classif.ranger, mlr_learners_classif.svm, mlr_learners_classif.xgboost, mlr_learners_regr.cv_glmnet, mlr_learners_regr.kknn, mlr_learners_regr.km, mlr_learners_regr.lm, mlr_learners_regr.nnet, mlr_learners_regr.ranger, mlr_learners_regr.svm, mlr_learners_regr.xgboost

Examples

if (requireNamespace("glmnet", quietly = TRUE)) {
# Define the Learner and set parameter values
learner = lrn("regr.glmnet")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

# importance method
if("importance" %in% learner$properties) print(learner$importance)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
}

k-Nearest-Neighbor Regression Learner

Description

k-Nearest-Neighbor regression. Calls kknn::kknn() from package kknn.

Initial parameter values

  • store_model:

    • See note.

Dictionary

This mlr3::Learner can be instantiated via the dictionary mlr3::mlr_learners or with the associated sugar function mlr3::lrn(): 

mlr_learners$get("regr.kknn")
lrn("regr.kknn")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3learners, kknn

Parameters

Id Type Default Levels Range
k integer 7 [1,∞)[1, \infty)
distance numeric 2 [0,∞)[0, \infty)
kernel character optimal rectangular, triangular, epanechnikov, biweight, triweight, cos, inv, gaussian, rank, optimal -
scale logical TRUE TRUE, FALSE -
ykernel untyped NULL -
store_model logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrKKNN

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrKKNN$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrKKNN$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Note

There is no training step for k-NN models, just storing the training data to process it during the predict step. Therefore, ⁠$model⁠ returns a list with the following elements:

  • formula: Formula for calling kknn::kknn() during ⁠$predict()⁠.

  • data: Training data for calling kknn::kknn() during ⁠$predict()⁠.

  • pv: Training parameters for calling kknn::kknn() during ⁠$predict()⁠.

  • kknn: Model as returned by kknn::kknn(), only available after ⁠$predict()⁠ has been called. This is not stored by default, you must set hyperparameter store_model to TRUE.

References

Hechenbichler, Klaus, Schliep, Klaus (2004). “Weighted k-nearest-neighbor techniques and ordinal classification.” Technical Report Discussion Paper 399, SFB 386, Ludwig-Maximilians University Munich. doi:10.5282/ubm/epub.1769.

Samworth, J R (2012). “Optimal weighted nearest neighbour classifiers.” The Annals of Statistics, 40(5), 2733–2763. doi:10.1214/12-AOS1049.

Cover, Thomas, Hart, Peter (1967). “Nearest neighbor pattern classification.” IEEE transactions on information theory, 13(1), 21–27. doi:10.1109/TIT.1967.1053964.

See Also

Other Learner: mlr_learners_classif.cv_glmnet, mlr_learners_classif.glmnet, mlr_learners_classif.kknn, mlr_learners_classif.lda, mlr_learners_classif.log_reg, mlr_learners_classif.multinom, mlr_learners_classif.naive_bayes, mlr_learners_classif.nnet, mlr_learners_classif.qda, mlr_learners_classif.ranger, mlr_learners_classif.svm, mlr_learners_classif.xgboost, mlr_learners_regr.cv_glmnet, mlr_learners_regr.glmnet, mlr_learners_regr.km, mlr_learners_regr.lm, mlr_learners_regr.nnet, mlr_learners_regr.ranger, mlr_learners_regr.svm, mlr_learners_regr.xgboost

Examples

if (requireNamespace("kknn", quietly = TRUE)) {
# Define the Learner and set parameter values
learner = lrn("regr.kknn")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

# importance method
if("importance" %in% learner$properties) print(learner$importance)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
}

Kriging Regression Learner

Description

Kriging regression. Calls DiceKriging::km() from package DiceKriging.

  • The predict type hyperparameter "type" defaults to "sk" (simple kriging).

  • The additional hyperparameter nugget.stability is used to overwrite the hyperparameter nugget with nugget.stability * var(y) before training to improve the numerical stability. We recommend a value of 1e-8.

  • The additional hyperparameter jitter can be set to add ⁠N(0, [jitter])⁠-distributed noise to the data before prediction to avoid perfect interpolation. We recommend a value of 1e-12.

Dictionary

This mlr3::Learner can be instantiated via the dictionary mlr3::mlr_learners or with the associated sugar function mlr3::lrn(): 

mlr_learners$get("regr.km")
lrn("regr.km")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”

  • Feature Types: “logical”, “integer”, “numeric”

  • Required Packages: mlr3, mlr3learners, DiceKriging

Parameters

Id Type Default Levels Range
bias.correct logical FALSE TRUE, FALSE -
checkNames logical TRUE TRUE, FALSE -
coef.cov untyped NULL -
coef.trend untyped NULL -
coef.var untyped NULL -
control untyped NULL -
cov.compute logical TRUE TRUE, FALSE -
covtype character matern5_2 gauss, matern5_2, matern3_2, exp, powexp -
estim.method character MLE MLE, LOO -
gr logical TRUE TRUE, FALSE -
iso logical FALSE TRUE, FALSE -
jitter numeric 0 [0,∞)[0, \infty)
kernel untyped NULL -
knots untyped NULL -
light.return logical FALSE TRUE, FALSE -
lower untyped NULL -
multistart integer 1 (−∞,∞)(-\infty, \infty)
noise.var untyped NULL -
nugget numeric - (−∞,∞)(-\infty, \infty)
nugget.estim logical FALSE TRUE, FALSE -
nugget.stability numeric 0 [0,∞)[0, \infty)
optim.method character BFGS BFGS, gen -
parinit untyped NULL -
penalty untyped NULL -
scaling logical FALSE TRUE, FALSE -
se.compute logical TRUE TRUE, FALSE -
type character SK SK, UK -
upper untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrKM

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrKM$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrKM$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

References

Roustant O, Ginsbourger D, Deville Y (2012). “DiceKriging, DiceOptim: Two R Packages for the Analysis of Computer Experiments by Kriging-Based Metamodeling and Optimization.” Journal of Statistical Software, 51(1), 1–55. doi:10.18637/jss.v051.i01.

See Also

Other Learner: mlr_learners_classif.cv_glmnet, mlr_learners_classif.glmnet, mlr_learners_classif.kknn, mlr_learners_classif.lda, mlr_learners_classif.log_reg, mlr_learners_classif.multinom, mlr_learners_classif.naive_bayes, mlr_learners_classif.nnet, mlr_learners_classif.qda, mlr_learners_classif.ranger, mlr_learners_classif.svm, mlr_learners_classif.xgboost, mlr_learners_regr.cv_glmnet, mlr_learners_regr.glmnet, mlr_learners_regr.kknn, mlr_learners_regr.lm, mlr_learners_regr.nnet, mlr_learners_regr.ranger, mlr_learners_regr.svm, mlr_learners_regr.xgboost

Examples

if (requireNamespace("DiceKriging", quietly = TRUE)) {
# Define the Learner and set parameter values
learner = lrn("regr.km")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

# importance method
if("importance" %in% learner$properties) print(learner$importance)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
}

Linear Model Regression Learner

Description

Ordinary linear regression. Calls stats::lm().

Dictionary

This mlr3::Learner can be instantiated via the dictionary mlr3::mlr_learners or with the associated sugar function mlr3::lrn(): 

mlr_learners$get("regr.lm")
lrn("regr.lm")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”

  • Feature Types: “logical”, “integer”, “numeric”, “character”, “factor”

  • Required Packages: mlr3, mlr3learners, 'stats'

Parameters

Id Type Default Levels Range
df numeric Inf (−∞,∞)(-\infty, \infty)
interval character - none, confidence, prediction -
level numeric 0.95 (−∞,∞)(-\infty, \infty)
model logical TRUE TRUE, FALSE -
offset logical - TRUE, FALSE -
pred.var untyped - -
qr logical TRUE TRUE, FALSE -
scale numeric NULL (−∞,∞)(-\infty, \infty)
singular.ok logical TRUE TRUE, FALSE -
x logical FALSE TRUE, FALSE -
y logical FALSE TRUE, FALSE -
rankdeficient character - warnif, simple, non-estim, NA, NAwarn -
tol numeric 1e-07 (−∞,∞)(-\infty, \infty)
verbose logical FALSE TRUE, FALSE -

Contrasts

To ensure reproducibility, this learner always uses the default contrasts:

Setting the option "contrasts" does not have any effect. Instead, set the respective hyperparameter or use mlr3pipelines to create dummy features.

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrLM

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrLM$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrLM$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

See Also

Other Learner: mlr_learners_classif.cv_glmnet, mlr_learners_classif.glmnet, mlr_learners_classif.kknn, mlr_learners_classif.lda, mlr_learners_classif.log_reg, mlr_learners_classif.multinom, mlr_learners_classif.naive_bayes, mlr_learners_classif.nnet, mlr_learners_classif.qda, mlr_learners_classif.ranger, mlr_learners_classif.svm, mlr_learners_classif.xgboost, mlr_learners_regr.cv_glmnet, mlr_learners_regr.glmnet, mlr_learners_regr.kknn, mlr_learners_regr.km, mlr_learners_regr.nnet, mlr_learners_regr.ranger, mlr_learners_regr.svm, mlr_learners_regr.xgboost

Examples

if (requireNamespace("stats", quietly = TRUE)) {
# Define the Learner and set parameter values
learner = lrn("regr.lm")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

# importance method
if("importance" %in% learner$properties) print(learner$importance)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
}

Neural Network Regression Learner

Description

Single Layer Neural Network. Calls nnet::nnet.formula() from package nnet.

Note that modern neural networks with multiple layers are connected via package mlr3torch.

Dictionary

This mlr3::Learner can be instantiated via the dictionary mlr3::mlr_learners or with the associated sugar function mlr3::lrn(): 

mlr_learners$get("regr.nnet")
lrn("regr.nnet")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3learners, nnet

Parameters

Id Type Default Levels Range
Hess logical FALSE TRUE, FALSE -
MaxNWts integer 1000 [1,∞)[1, \infty)
Wts untyped - -
abstol numeric 1e-04 (−∞,∞)(-\infty, \infty)
censored logical FALSE TRUE, FALSE -
contrasts untyped NULL -
decay numeric 0 (−∞,∞)(-\infty, \infty)
mask untyped - -
maxit integer 100 [1,∞)[1, \infty)
na.action untyped - -
rang numeric 0.7 (−∞,∞)(-\infty, \infty)
reltol numeric 1e-08 (−∞,∞)(-\infty, \infty)
size integer 3 [0,∞)[0, \infty)
skip logical FALSE TRUE, FALSE -
subset untyped - -
trace logical TRUE TRUE, FALSE -
formula untyped - -

Initial parameter values

  • size:

    • Adjusted default: 3L.

    • Reason for change: no default in nnet().

Custom mlr3 parameters

  • formula: if not provided, the formula is set to task$formula().

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrNnet

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrNnet$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrNnet$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

References

Ripley BD (1996). Pattern Recognition and Neural Networks. Cambridge University Press. doi:10.1017/cbo9780511812651.

See Also

Other Learner: mlr_learners_classif.cv_glmnet, mlr_learners_classif.glmnet, mlr_learners_classif.kknn, mlr_learners_classif.lda, mlr_learners_classif.log_reg, mlr_learners_classif.multinom, mlr_learners_classif.naive_bayes, mlr_learners_classif.nnet, mlr_learners_classif.qda, mlr_learners_classif.ranger, mlr_learners_classif.svm, mlr_learners_classif.xgboost, mlr_learners_regr.cv_glmnet, mlr_learners_regr.glmnet, mlr_learners_regr.kknn, mlr_learners_regr.km, mlr_learners_regr.lm, mlr_learners_regr.ranger, mlr_learners_regr.svm, mlr_learners_regr.xgboost

Examples

if (requireNamespace("nnet", quietly = TRUE)) {
# Define the Learner and set parameter values
learner = lrn("regr.nnet")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

# importance method
if("importance" %in% learner$properties) print(learner$importance)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
}

Ranger Regression Learner

Description

Random regression forest. Calls ranger::ranger() from package ranger.

Dictionary

This mlr3::Learner can be instantiated via the dictionary mlr3::mlr_learners or with the associated sugar function mlr3::lrn(): 

mlr_learners$get("regr.ranger")
lrn("regr.ranger")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”, “quantiles”

  • Feature Types: “logical”, “integer”, “numeric”, “character”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3learners, ranger

Parameters

Id Type Default Levels Range
always.split.variables untyped - -
holdout logical FALSE TRUE, FALSE -
importance character - none, impurity, impurity_corrected, permutation -
keep.inbag logical FALSE TRUE, FALSE -
max.depth integer NULL [1,∞)[1, \infty)
min.bucket integer 1 [1,∞)[1, \infty)
min.node.size integer 5 [1,∞)[1, \infty)
mtry integer - [1,∞)[1, \infty)
mtry.ratio numeric - [0,1][0, 1]
na.action character na.learn na.learn, na.omit, na.fail -
node.stats logical FALSE TRUE, FALSE -
num.random.splits integer 1 [1,∞)[1, \infty)
num.threads integer 1 [1,∞)[1, \infty)
num.trees integer 500 [1,∞)[1, \infty)
oob.error logical TRUE TRUE, FALSE -
poisson.tau numeric 1 (−∞,∞)(-\infty, \infty)
regularization.factor untyped 1 -
regularization.usedepth logical FALSE TRUE, FALSE -
replace logical TRUE TRUE, FALSE -
respect.unordered.factors character - ignore, order, partition -
sample.fraction numeric - [0,1][0, 1]
save.memory logical FALSE TRUE, FALSE -
scale.permutation.importance logical FALSE TRUE, FALSE -
se.method character infjack jack, infjack -
seed integer NULL (−∞,∞)(-\infty, \infty)
split.select.weights untyped NULL -
splitrule character variance variance, extratrees, maxstat, beta, poisson -
verbose logical TRUE TRUE, FALSE -
write.forest logical TRUE TRUE, FALSE -

Custom mlr3 parameters

  • mtry:

    • This hyperparameter can alternatively be set via our hyperparameter mtry.ratio as mtry = max(ceiling(mtry.ratio * n_features), 1). Note that mtry and mtry.ratio are mutually exclusive.

Initial parameter values

  • num.threads:

    • Actual default: 2, using two threads, while also respecting environment variable R_RANGER_NUM_THREADS, options(ranger.num.threads = N), or options(Ncpus = N), with precedence in that order.

    • Adjusted value: 1.

    • Reason for change: Conflicting with parallelization via future.

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrRanger

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrRanger$new()

Method importance()

The importance scores are extracted from the model slot variable.importance. Parameter importance.mode must be set to "impurity", "impurity_corrected", or "permutation"

Usage
LearnerRegrRanger$importance()
Returns

Named numeric().

Method oob_error()

The out-of-bag error, extracted from model slot prediction.error.

Usage
LearnerRegrRanger$oob_error()
Returns

numeric(1).

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrRanger$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

References

Wright, N. M, Ziegler, Andreas (2017). “ranger: A Fast Implementation of Random Forests for High Dimensional Data in C++ and R.” Journal of Statistical Software, 77(1), 1–17. doi:10.18637/jss.v077.i01.

Breiman, Leo (2001). “Random Forests.” Machine Learning, 45(1), 5–32. ISSN 1573-0565, doi:10.1023/A:1010933404324.

See Also

Other Learner: mlr_learners_classif.cv_glmnet, mlr_learners_classif.glmnet, mlr_learners_classif.kknn, mlr_learners_classif.lda, mlr_learners_classif.log_reg, mlr_learners_classif.multinom, mlr_learners_classif.naive_bayes, mlr_learners_classif.nnet, mlr_learners_classif.qda, mlr_learners_classif.ranger, mlr_learners_classif.svm, mlr_learners_classif.xgboost, mlr_learners_regr.cv_glmnet, mlr_learners_regr.glmnet, mlr_learners_regr.kknn, mlr_learners_regr.km, mlr_learners_regr.lm, mlr_learners_regr.nnet, mlr_learners_regr.svm, mlr_learners_regr.xgboost

Examples

if (requireNamespace("ranger", quietly = TRUE)) {
# Define the Learner and set parameter values
learner = lrn("regr.ranger")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

# importance method
if("importance" %in% learner$properties) print(learner$importance)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
}

Support Vector Machine

Description

Support vector machine for regression. Calls e1071::svm() from package e1071.

Dictionary

This mlr3::Learner can be instantiated via the dictionary mlr3::mlr_learners or with the associated sugar function mlr3::lrn(): 

mlr_learners$get("regr.svm")
lrn("regr.svm")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”

  • Required Packages: mlr3, mlr3learners, e1071

Parameters

Id Type Default Levels Range
cachesize numeric 40 (−∞,∞)(-\infty, \infty)
coef0 numeric 0 (−∞,∞)(-\infty, \infty)
cost numeric 1 [0,∞)[0, \infty)
cross integer 0 [0,∞)[0, \infty)
degree integer 3 [1,∞)[1, \infty)
epsilon numeric 0.1 [0,∞)[0, \infty)
fitted logical TRUE TRUE, FALSE -
gamma numeric - [0,∞)[0, \infty)
kernel character radial linear, polynomial, radial, sigmoid -
nu numeric 0.5 (−∞,∞)(-\infty, \infty)
scale untyped TRUE -
shrinking logical TRUE TRUE, FALSE -
tolerance numeric 0.001 [0,∞)[0, \infty)
type character eps-regression eps-regression, nu-regression -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrSVM

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrSVM$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrSVM$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

References

Cortes, Corinna, Vapnik, Vladimir (1995). “Support-vector networks.” Machine Learning, 20(3), 273–297. doi:10.1007/BF00994018.

See Also

Other Learner: mlr_learners_classif.cv_glmnet, mlr_learners_classif.glmnet, mlr_learners_classif.kknn, mlr_learners_classif.lda, mlr_learners_classif.log_reg, mlr_learners_classif.multinom, mlr_learners_classif.naive_bayes, mlr_learners_classif.nnet, mlr_learners_classif.qda, mlr_learners_classif.ranger, mlr_learners_classif.svm, mlr_learners_classif.xgboost, mlr_learners_regr.cv_glmnet, mlr_learners_regr.glmnet, mlr_learners_regr.kknn, mlr_learners_regr.km, mlr_learners_regr.lm, mlr_learners_regr.nnet, mlr_learners_regr.ranger, mlr_learners_regr.xgboost

Examples

if (requireNamespace("e1071", quietly = TRUE)) {
# Define the Learner and set parameter values
learner = lrn("regr.svm")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

# importance method
if("importance" %in% learner$properties) print(learner$importance)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
}

Extreme Gradient Boosting Regression Learner

Description

eXtreme Gradient Boosting regression. Calls xgboost::xgb.train() from package xgboost.

To compute on GPUs, you first need to compile xgboost yourself and link against CUDA. See https://xgboost.readthedocs.io/en/stable/build.html#building-with-gpu-support.

Note that using the watchlist parameter directly will lead to problems when wrapping this mlr3::Learner in a mlr3pipelines GraphLearner as the preprocessing steps will not be applied to the data in the watchlist. See the section Early Stopping and Validation on how to do this.

Dictionary

This mlr3::Learner can be instantiated via the dictionary mlr3::mlr_learners or with the associated sugar function mlr3::lrn(): 

mlr_learners$get("regr.xgboost")
lrn("regr.xgboost")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”

  • Required Packages: mlr3, mlr3learners, xgboost

Parameters

Id Type Default Levels Range
alpha numeric 0 [0,∞)[0, \infty)
approxcontrib logical FALSE TRUE, FALSE -
base_score numeric 0.5 (−∞,∞)(-\infty, \infty)
base_margin untyped NULL -
booster character gbtree gbtree, gblinear, dart -
callbacks untyped list() -
colsample_bylevel numeric 1 [0,1][0, 1]
colsample_bynode numeric 1 [0,1][0, 1]
colsample_bytree numeric 1 [0,1][0, 1]
device untyped "cpu" -
disable_default_eval_metric logical FALSE TRUE, FALSE -
early_stopping_rounds integer NULL [1,∞)[1, \infty)
eta numeric 0.3 [0,1][0, 1]
eval_metric untyped "rmse" -
feature_selector character cyclic cyclic, shuffle, random, greedy, thrifty -
gamma numeric 0 [0,∞)[0, \infty)
grow_policy character depthwise depthwise, lossguide -
interaction_constraints untyped - -
iterationrange untyped - -
lambda numeric 1 [0,∞)[0, \infty)
lambda_bias numeric 0 [0,∞)[0, \infty)
max_bin integer 256 [2,∞)[2, \infty)
max_delta_step numeric 0 [0,∞)[0, \infty)
max_depth integer 6 [0,∞)[0, \infty)
max_leaves integer 0 [0,∞)[0, \infty)
maximize logical NULL TRUE, FALSE -
min_child_weight numeric 1 [0,∞)[0, \infty)
missing numeric NA (−∞,∞)(-\infty, \infty)
monotone_constraints untyped 0 -
normalize_type character tree tree, forest -
nrounds integer - [1,∞)[1, \infty)
nthread integer 1 [1,∞)[1, \infty)
ntreelimit integer NULL [1,∞)[1, \infty)
num_parallel_tree integer 1 [1,∞)[1, \infty)
objective untyped "reg:squarederror" -
one_drop logical FALSE TRUE, FALSE -
outputmargin logical FALSE TRUE, FALSE -
predcontrib logical FALSE TRUE, FALSE -
predinteraction logical FALSE TRUE, FALSE -
predleaf logical FALSE TRUE, FALSE -
print_every_n integer 1 [1,∞)[1, \infty)
process_type character default default, update -
rate_drop numeric 0 [0,1][0, 1]
refresh_leaf logical TRUE TRUE, FALSE -
reshape logical FALSE TRUE, FALSE -
sampling_method character uniform uniform, gradient_based -
sample_type character uniform uniform, weighted -
save_name untyped NULL -
save_period integer NULL [0,∞)[0, \infty)
scale_pos_weight numeric 1 (−∞,∞)(-\infty, \infty)
seed_per_iteration logical FALSE TRUE, FALSE -
skip_drop numeric 0 [0,1][0, 1]
strict_shape logical FALSE TRUE, FALSE -
subsample numeric 1 [0,1][0, 1]
top_k integer 0 [0,∞)[0, \infty)
training logical FALSE TRUE, FALSE -
tree_method character auto auto, exact, approx, hist, gpu_hist -
tweedie_variance_power numeric 1.5 [1,2][1, 2]
updater untyped - -
verbose integer 1 [0,2][0, 2]
watchlist untyped NULL -
xgb_model untyped NULL -

Early Stopping and Validation

In order to monitor the validation performance during the training, you can set the ⁠$validate⁠ field of the Learner. For information on how to configure the valdiation set, see the Validation section of mlr3::Learner. This validation data can also be used for early stopping, which can be enabled by setting the early_stopping_rounds parameter. The final (or in the case of early stopping best) validation scores can be accessed via ⁠$internal_valid_scores⁠, and the optimal nrounds via ⁠$internal_tuned_values⁠. The internal validation measure can be set via the eval_metric parameter that can be a mlr3::Measure, a function, or a character string for the internal xgboost measures. Using an mlr3::Measure is slower than the internal xgboost measures, but allows to use the same measure for tuning and validation.

Initial parameter values

  • nrounds:

    • Actual default: no default.

    • Adjusted default: 1000.

    • Reason for change: Without a default construction of the learner would error. The lightgbm learner has a default of 1000, so we use the same here.

  • nthread:

    • Actual value: Undefined, triggering auto-detection of the number of CPUs.

    • Adjusted value: 1.

    • Reason for change: Conflicting with parallelization via future.

  • verbose:

    • Actual default: 1.

    • Adjusted default: 0.

    • Reason for change: Reduce verbosity.

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrXgboost

Active bindings

internal_valid_scores

(named list() or NULL) The validation scores extracted from model$evaluation_log. If early stopping is activated, this contains the validation scores of the model for the optimal nrounds, otherwise the nrounds for the final model.

internal_tuned_values

(named list() or NULL) If early stopping is activated, this returns a list with nrounds, which is extracted from ⁠$best_iteration⁠ of the model and otherwise NULL.

validate

(numeric(1) or character(1) or NULL) How to construct the internal validation data. This parameter can be either NULL, a ratio, "test", or "predefined". Returns the ⁠$best_iteration⁠ when early stopping is activated.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrXgboost$new()

Method importance()

The importance scores are calculated with xgboost::xgb.importance().

Usage
LearnerRegrXgboost$importance()
Returns

Named numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrXgboost$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Note

To compute on GPUs, you first need to compile xgboost yourself and link against CUDA. See https://xgboost.readthedocs.io/en/stable/build.html#building-with-gpu-support.

References

Chen, Tianqi, Guestrin, Carlos (2016). “Xgboost: A scalable tree boosting system.” In Proceedings of the 22nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 785–794. ACM. doi:10.1145/2939672.2939785.

See Also

Other Learner: mlr_learners_classif.cv_glmnet, mlr_learners_classif.glmnet, mlr_learners_classif.kknn, mlr_learners_classif.lda, mlr_learners_classif.log_reg, mlr_learners_classif.multinom, mlr_learners_classif.naive_bayes, mlr_learners_classif.nnet, mlr_learners_classif.qda, mlr_learners_classif.ranger, mlr_learners_classif.svm, mlr_learners_classif.xgboost, mlr_learners_regr.cv_glmnet, mlr_learners_regr.glmnet, mlr_learners_regr.kknn, mlr_learners_regr.km, mlr_learners_regr.lm, mlr_learners_regr.nnet, mlr_learners_regr.ranger, mlr_learners_regr.svm

Examples

## Not run: 
if (requireNamespace("xgboost", quietly = TRUE)) {
# Define the Learner and set parameter values
learner = lrn("regr.xgboost")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

# importance method
if("importance" %in% learner$properties) print(learner$importance)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()
}

## End(Not run)

## Not run: 
# Train learner with early stopping on spam data set
task = tsk("mtcars")

# use 30 percent for validation
# Set early stopping parameter
learner = lrn("regr.xgboost",
  nrounds = 100,
  early_stopping_rounds = 10,
  validate = 0.3
)

# Train learner with early stopping
learner$train(task)

# Inspect optimal nrounds and validation performance
learner$internal_tuned_values
learner$internal_valid_scores

## End(Not run)