Help for package mlr3automl

Title:

Automated Machine Learning for 'mlr3'

Version:

0.1.0

Description:

Flexible automated machine learning (AutoML) system for the 'mlr3' ecosystem. Automatically selects a suitable machine learning algorithm and tunes its hyperparameters for a given task. Constructs preprocessing pipelines with multiple parallel branches using 'mlr3pipelines' and jointly optimizes them together with the learners using 'mlr3tuning'. The optimization is driven by asynchronous decentralized Bayesian optimization by Egele et al. (2023) <doi:10.1109/e-Science58273.2023.10254839>.

License:

LGPL-3

URL:

https://mlr3automl.mlr-org.com, https://github.com/mlr-org/mlr3automl

BugReports:

https://github.com/mlr-org/mlr3automl/issues

Depends:

mlr3 (≥ 1.6.0), mlr3tuning (≥ 1.6.0), R (≥ 3.3.0), rush (≥ 1.2.1)

Imports:

bbotk (≥ 1.7.1), checkmate, data.table, lhs, mlr3learners (≥ 0.14.0), mlr3mbo (≥ 1.2.0), mlr3misc (≥ 0.15.1), mlr3pipelines, paradox (≥ 1.0.1), R6, utils

Suggests:

callr, e1071, fastai, glmnet, kknn, lgr, lightgbm, MASS, mirai, mlr3extralearners, mlr3torch (≥ 0.3.3), mlr3viz, ranger, redux, reticulate, rpart, testthat (≥ 3.0.0), torch, xgboost (≥ 3.2.1.1)

Additional_repositories:

https://mlr-org.r-universe.dev

Config/roxygen2/version:

8.0.0.9000

Config/testthat/edition:

Config/testthat/parallel:

false

Encoding:

UTF-8

NeedsCompilation:

Collate:

'mlr_auto.R' 'Auto.R' 'AutoCatboost.R' 'AutoExtraTrees.R' 'AutoFTTransformer.R' 'AutoFastai.R' 'AutoGlmnet.R' 'AutoKKNN.R' 'AutoLda.R' 'AutoLightGBM.R' 'AutoMLP.R' 'AutoRanger.R' 'AutoResNet.R' 'AutoSVM.R' 'AutoTabPFN.R' 'AutoXgboost.R' 'aaa.R' 'LearnerAuto.R' 'LearnerClassifAuto.R' 'LearnerClassifAutoCatboost.R' 'LearnerClassifAutoFTTransformer.R' 'LearnerClassifAutoFastai.R' 'LearnerClassifAutoGlmnet.R' 'LearnerClassifAutoKKNN.R' 'LearnerClassifAutoLightGBM.R' 'LearnerClassifAutoMLP.R' 'LearnerClassifAutoRanger.R' 'LearnerClassifAutoResNet.R' 'LearnerClassifAutoSVM.R' 'LearnerClassifAutoTabPFN.R' 'LearnerClassifAutoXgboost.R' 'LearnerRegrAuto.R' 'LearnerRegrAutoCatboost.R' 'LearnerRegrAutoFTTransformer.R' 'LearnerRegrAutoGlmnet.R' 'LearnerRegrAutoKKNN.R' 'LearnerRegrAutoLightGBM.R' 'LearnerRegrAutoMLP.R' 'LearnerRegrAutoRanger.R' 'LearnerRegrAutoResNet.R' 'LearnerRegrAutoSVM.R' 'LearnerRegrAutoTabPFN.R' 'LearnerRegrAutoXgboost.R' 'helper.R' 'install_python_learners.R' 'isolated_model.R' 'mlr_callbacks.R' 'sugar.R' 'train_auto.R' 'zzz.R'

Packaged:

2026-08-20 09:05:18 UTC; marc

Author:

Marc Becker

[cre, aut, cph], Damir Pulatov [aut], Baisu Zhou [aut], Lona Koers [aut]

Maintainer:

Marc Becker <marcbecker@posteo.de>

Repository:

CRAN

Date/Publication:

2026-09-03 12:20:03 UTC

mlr3automl: Automated Machine Learning for 'mlr3'

Description

Author(s)

Maintainer: Marc Becker marcbecker@posteo.de (ORCID) [copyright holder]

Authors:

Marc Becker marcbecker@posteo.de (ORCID) [copyright holder]
Damir Pulatov damirpolat@protonmail.com
Baisu Zhou baisu.zhou@outlook.com
Lona Koers lona.koers@gmail.com

Auto Class

Description

This class is the base class for all autos.

Value

Object of class R6::R6Class and Auto.

Public fields

id: (character(1)).
properties: (character()).
task_types: (character()).
packages: (character()).
devices: (character()).

Methods

`Auto$new()`

Creates a new instance of this R6 class.

Usage

Auto$new(
  id,
  properties = character(0),
  task_types = character(0),
  packages = character(0),
  devices = character(0)
)

Arguments

id: (character(1)).
properties: (character()).
task_types: (character()).
packages: (character()).
devices: (character()).

`Auto$check()`

Check if the auto is compatible with the task.

Usage

Auto$check(task, memory_limit = Inf, large_data_set = FALSE, devices)

Arguments

task: (mlr3::Task).
memory_limit: (integer(1)).
large_data_set: (logical(1)).
devices: (character())
Devices to use. Allowed values are "cpu" and "cuda". Default is "cpu".

`Auto$graph()`

Create the graph for the auto.

Usage

Auto$graph(task, measure, n_threads, timeout, devices)

Arguments

task: (mlr3::Task).
measure: (mlr3::Measure).
n_threads: (integer(1)).
timeout: (integer(1)).
devices: (character())
Devices to use. Allowed values are "cpu" and "cuda". Default is "cpu".

`Auto$early_stopping_rounds()`

Estimate the number of early stopping rounds (the patience) for a learner. budget is the maximum number of training rounds (boosting iterations or epochs) the learner may use. The patience is capped well below the budget, otherwise early stopping and validation-based internal tuning can never trigger and the learner always trains for the full budget.

Usage

Auto$early_stopping_rounds(task, budget = Inf)

Arguments

task: (mlr3::Task).
budget: (integer(1))
Maximum number of training rounds (boosting iterations or epochs) the learner may use.

`Auto$estimate_memory()`

Estimate the memory for the auto.

Usage

Auto$estimate_memory(task)

Arguments

task: (mlr3::Task).

`Auto$finalize_model()`

Prepare the graph learner for the final model fit. Called after tuning to undo tuning-only setup (e.g., timeout callbacks).

Usage

Auto$finalize_model(graph_learner)

Arguments

graph_learner: (mlr3pipelines::GraphLearner).

`Auto$design_default()`

Default hyperparameters for the learner.

Usage

Auto$design_default(task)

Arguments

task: (mlr3::Task).

`Auto$design_set()`

Get the initial hyperparameter set for the learner.

Usage

Auto$design_set(task, measure, size)

Arguments

task: (mlr3::Task).
measure: (mlr3::Measure).
size: (integer(1)).

`Auto$search_space()`

Get the search space for the learner.

Usage

Auto$search_space(task)

Arguments

task: (mlr3::Task).

`Auto$clone()`

The objects of this class are cloneable with this method.

Usage

Auto$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Catboost Auto

Description

Catboost auto.

Value

Object of class R6::R6Class and AutoCatboost.

Super class

Auto -> AutoCatboost

Methods

Public methods

AutoCatboost$new()
AutoCatboost$graph()
AutoCatboost$estimate_memory()
AutoCatboost$internal_measure()
AutoCatboost$clone()

Inherited methods

`AutoCatboost$new()`

Creates a new instance of this R6 class.

Usage

AutoCatboost$new(id = "catboost")

Arguments

id: (character(1))
Identifier for the new instance.

`AutoCatboost$graph()`

Create the graph for the auto.

Usage

AutoCatboost$graph(task, measure, n_threads, timeout, devices)

Arguments

task: (mlr3::Task).
measure: (mlr3::Measure).
n_threads: (integer(1)).
timeout: (integer(1)).
devices: (character())
Devices to use. Allowed values are "cpu" and "cuda". Default is "cpu".

`AutoCatboost$estimate_memory()`

Estimate the memory for the auto.

Usage

AutoCatboost$estimate_memory(task)

Arguments

task: (mlr3::Task).

`AutoCatboost$internal_measure()`

Get the internal measure for the auto.

Usage

AutoCatboost$internal_measure(measure, task)

Arguments

measure: (mlr3::Measure).
task: (mlr3::Task).

`AutoCatboost$clone()`

The objects of this class are cloneable with this method.

Usage

AutoCatboost$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

auto("catboost")

Extra Trees Auto

Description

Extra Trees auto.

Value

Object of class R6::R6Class and AutoExtraTrees.

Super class

Auto -> AutoExtraTrees

Methods

Public methods

AutoExtraTrees$new()
AutoExtraTrees$graph()
AutoExtraTrees$estimate_memory()
AutoExtraTrees$clone()

Inherited methods

`AutoExtraTrees$new()`

Creates a new instance of this R6 class.

Usage

AutoExtraTrees$new(id = "extra_trees")

Arguments

id: (character(1))
Identifier for the new instance.

`AutoExtraTrees$graph()`

Create the graph for the auto.

Usage

AutoExtraTrees$graph(task, measure, n_threads, timeout, devices)

Arguments

task: (mlr3::Task).
measure: (mlr3::Measure).
n_threads: (numeric(1)).
timeout: (numeric(1)).
devices: (character())
Devices to use. Allowed values are "cpu" and "cuda". Default is "cpu".

`AutoExtraTrees$estimate_memory()`

Estimate the memory for the auto.

Usage

AutoExtraTrees$estimate_memory(task)

Arguments

task: (mlr3::Task).

`AutoExtraTrees$clone()`

The objects of this class are cloneable with this method.

Usage

AutoExtraTrees$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

auto("extra_trees")

FTTransformer Auto

Description

FTTransformer auto.

Value

Object of class R6::R6Class and AutoFTTransformer.

Super class

Auto -> AutoFTTransformer

Methods

Public methods

AutoFTTransformer$new()
AutoFTTransformer$graph()
AutoFTTransformer$estimate_memory()
AutoFTTransformer$clone()

Inherited methods

`AutoFTTransformer$new()`

Creates a new instance of this R6 class.

Usage

AutoFTTransformer$new(id = "ft_transformer")

Arguments

id: (character(1))
Identifier for the new instance.

`AutoFTTransformer$graph()`

Create the graph for the auto.

Usage

AutoFTTransformer$graph(task, measure, n_threads, timeout, devices)

Arguments

task: (mlr3::Task).
measure: (mlr3::Measure).
n_threads: (integer(1)).
timeout: (integer(1)).
devices: (character())
Devices to use. Allowed values are "cpu" and "cuda". Default is "cpu".

`AutoFTTransformer$estimate_memory()`

Estimate the memory for the auto.

Usage

AutoFTTransformer$estimate_memory(task)

Arguments

task: (mlr3::Task).

`AutoFTTransformer$clone()`

The objects of this class are cloneable with this method.

Usage

AutoFTTransformer$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

auto("ft_transformer")

Fastai Auto

Description

Fastai auto.

Value

Object of class R6::R6Class and AutoFastai.

Python learners

Python learners like TabPFN and fastai run via reticulate and therefore need a Python installation with their required packages. There are two ways to provide it:

Do nothing and let reticulate::py_require() install the required packages into an ephemeral virtual environment automatically.
Point the RETICULATE_PYTHON environment variable to a Python installation that has the required packages installed.

We recommend option 2 when running on many workers, as it avoids the overhead of downloading and installing the packages on each worker. Use install_python_learners() to create a conda environment with the required packages and set RETICULATE_PYTHON to the returned Python binary.

The TabPFN learner additionally requires the TABPFN_TOKEN environment variable to download the model weights.

Super class

Auto -> AutoFastai

Methods

Public methods

AutoFastai$new()
AutoFastai$check()
AutoFastai$graph()
AutoFastai$estimate_memory()
AutoFastai$internal_measure()
AutoFastai$clone()

Inherited methods

`AutoFastai$new()`

Creates a new instance of this R6 class.

Usage

AutoFastai$new(id = "fastai")

Arguments

id: (character(1))
Identifier for the new instance.

`AutoFastai$check()`

Check if the auto is compatible with the task.

Usage

AutoFastai$check(
  task,
  memory_limit = Inf,
  large_data_set = FALSE,
  devices = "cpu"
)

Arguments

task: (mlr3::Task).
memory_limit: (integer(1)).
large_data_set: (logical(1)).
devices: (character())
Devices to use. Allowed values are "cpu" and "cuda". Default is "cpu".

`AutoFastai$graph()`

Create the graph for the auto.

Usage

AutoFastai$graph(task, measure, n_threads, timeout, devices)

Arguments

task: (mlr3::Task).
measure: (mlr3::Measure).
n_threads: (integer(1)).
timeout: (integer(1)).
devices: (character())
Devices to use. Allowed values are "cpu" and "cuda". Default is "cpu".

`AutoFastai$estimate_memory()`

Estimate the memory for the auto.

Usage

AutoFastai$estimate_memory(task)

Arguments

task: (mlr3::Task).

`AutoFastai$internal_measure()`

Get the internal measure for the auto.

Usage

AutoFastai$internal_measure(measure, task)

Arguments

measure: (mlr3::Measure).
task: (mlr3::Task).

`AutoFastai$clone()`

The objects of this class are cloneable with this method.

Usage

AutoFastai$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

auto("fastai")

Glmnet Auto

Description

Glmnet auto.

Value

Object of class R6::R6Class and AutoGlmnet.

Super class

Auto -> AutoGlmnet

Methods

Public methods

AutoGlmnet$new()
AutoGlmnet$graph()
AutoGlmnet$clone()

Inherited methods

`AutoGlmnet$new()`

Creates a new instance of this R6 class.

Usage

AutoGlmnet$new(id = "glmnet")

Arguments

id: (character(1))
Identifier for the new instance.

`AutoGlmnet$graph()`

Create the graph for the auto.

Usage

AutoGlmnet$graph(task, measure, n_threads, timeout, devices)

Arguments

task: (mlr3::Task).
measure: (mlr3::Measure).
n_threads: (integer(1)).
timeout: (integer(1)).
devices: (character())
Devices to use. Allowed values are "cpu" and "cuda". Default is "cpu".

`AutoGlmnet$clone()`

The objects of this class are cloneable with this method.

Usage

AutoGlmnet$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

auto("glmnet")

KKNN Auto

Description

Kknn auto.

Value

Object of class R6::R6Class and AutoKKNN.

Super class

Auto -> AutoKKNN

Methods

Public methods

AutoKKNN$new()
AutoKKNN$graph()
AutoKKNN$search_space()
AutoKKNN$clone()

Inherited methods

`AutoKKNN$new()`

Creates a new instance of this R6 class.

Usage

AutoKKNN$new(id = "kknn")

Arguments

id: (character(1))
Identifier for the new instance.

`AutoKKNN$graph()`

Create the graph for the auto.

Usage

AutoKKNN$graph(task, measure, n_threads, timeout, devices)

Arguments

task: (mlr3::Task).
measure: (mlr3::Measure).
n_threads: (integer(1)).
timeout: (integer(1)).
devices: (character())
Devices to use. Allowed values are "cpu" and "cuda". Default is "cpu".

`AutoKKNN$search_space()`

Get the search space for the auto.

Usage

AutoKKNN$search_space(task)

Arguments

task: (mlr3::Task).

`AutoKKNN$clone()`

The objects of this class are cloneable with this method.

Usage

AutoKKNN$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

auto("kknn")

Lda Auto

Description

Lda auto.

Value

Object of class R6::R6Class and AutoLda.

Super class

Auto -> AutoLda

Methods

Public methods

AutoLda$new()
AutoLda$graph()
AutoLda$clone()

Inherited methods

`AutoLda$new()`

Creates a new instance of this R6 class.

Usage

AutoLda$new(id = "lda")

Arguments

id: (character(1))
Identifier for the new instance.

`AutoLda$graph()`

Create the graph for the auto.

Usage

AutoLda$graph(task, measure, n_threads, timeout, devices)

Arguments

task: (mlr3::Task).
measure: (mlr3::Measure).
n_threads: (integer(1)).
timeout: (integer(1)).
devices: (character())
Devices to use. Allowed values are "cpu" and "cuda". Default is "cpu".

`AutoLda$clone()`

The objects of this class are cloneable with this method.

Usage

AutoLda$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

auto("lda")

LightGBM Auto

Description

Lightgbm auto.

Value

Object of class R6::R6Class and AutoLightGBM.

Super class

Auto -> AutoLightGBM

Methods

Public methods

AutoLightGBM$new()
AutoLightGBM$graph()
AutoLightGBM$finalize_model()
AutoLightGBM$estimate_memory()
AutoLightGBM$internal_measure()
AutoLightGBM$clone()

Inherited methods

`AutoLightGBM$new()`

Creates a new instance of this R6 class.

Usage

AutoLightGBM$new(id = "lightgbm")

Arguments

id: (character(1))
Identifier for the new instance.

`AutoLightGBM$graph()`

Create the graph for the auto.

Usage

AutoLightGBM$graph(task, measure, n_threads, timeout, devices)

Arguments

task: (mlr3::Task).
measure: (mlr3::Measure).
n_threads: (integer(1)).
timeout: (integer(1)).
devices: (character())
Devices to use. Allowed values are "cpu" and "cuda". Default is "cpu".

`AutoLightGBM$finalize_model()`

Prepare the graph learner for the final model fit.

Usage

AutoLightGBM$finalize_model(graph_learner)

Arguments

graph_learner: (mlr3pipelines::GraphLearner).

`AutoLightGBM$estimate_memory()`

Estimate the memory for the auto.

Usage

AutoLightGBM$estimate_memory(task)

Arguments

task: (mlr3::Task).

`AutoLightGBM$internal_measure()`

Get the internal measure for the auto.

Usage

AutoLightGBM$internal_measure(measure, task)

Arguments

measure: (mlr3::Measure).
task: (mlr3::Task).

`AutoLightGBM$clone()`

The objects of this class are cloneable with this method.

Usage

AutoLightGBM$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

auto("lightgbm")

MLP Auto

Description

Mlp auto.

Value

Object of class R6::R6Class and AutoMLP.

Super class

Auto -> AutoMLP

Methods

Public methods

AutoMLP$new()
AutoMLP$graph()
AutoMLP$estimate_memory()
AutoMLP$clone()

Inherited methods

`AutoMLP$new()`

Creates a new instance of this R6 class.

Usage

AutoMLP$new(id = "mlp")

Arguments

id: (character(1))
Identifier for the new instance.

`AutoMLP$graph()`

Create the graph for the auto.

Usage

AutoMLP$graph(task, measure, n_threads, timeout, devices)

Arguments

task: (mlr3::Task).
measure: (mlr3::Measure).
n_threads: (integer(1)).
timeout: (integer(1)).
devices: (character())
Devices to use. Allowed values are "cpu" and "cuda". Default is "cpu".

`AutoMLP$estimate_memory()`

Estimate the memory for the auto.

Usage

AutoMLP$estimate_memory(task)

Arguments

task: (mlr3::Task).

`AutoMLP$clone()`

The objects of this class are cloneable with this method.

Usage

AutoMLP$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

auto("mlp")

Ranger Auto

Description

Ranger auto.

Value

Object of class R6::R6Class and AutoRanger.

Super class

Auto -> AutoRanger

Methods

Public methods

AutoRanger$new()
AutoRanger$graph()
AutoRanger$estimate_memory()
AutoRanger$clone()

Inherited methods

`AutoRanger$new()`

Creates a new instance of this R6 class.

Usage

AutoRanger$new(id = "ranger")

Arguments

id: (character(1))
Identifier for the new instance.

`AutoRanger$graph()`

Create the graph for the auto.

Usage

AutoRanger$graph(task, measure, n_threads, timeout, devices)

Arguments

task: (mlr3::Task).
measure: (mlr3::Measure).
n_threads: (integer(1)).
timeout: (integer(1)).
devices: (character())
Devices to use. Allowed values are "cpu" and "cuda". Default is "cpu".

`AutoRanger$estimate_memory()`

Estimate the memory for the auto.

Usage

AutoRanger$estimate_memory(task)

Arguments

task: (mlr3::Task).

`AutoRanger$clone()`

The objects of this class are cloneable with this method.

Usage

AutoRanger$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

auto("ranger")

ResNet Auto

Description

ResNet auto.

Value

Object of class R6::R6Class and AutoResNet.

Super class

Auto -> AutoResNet

Methods

Public methods

AutoResNet$new()
AutoResNet$graph()
AutoResNet$estimate_memory()
AutoResNet$clone()

Inherited methods

`AutoResNet$new()`

Creates a new instance of this R6 class.

Usage

AutoResNet$new(id = "resnet")

Arguments

id: (character(1))
Identifier for the new instance.

`AutoResNet$graph()`

Create the graph for the auto.

Usage

AutoResNet$graph(task, measure, n_threads, timeout, devices)

Arguments

task: (mlr3::Task).
measure: (mlr3::Measure).
n_threads: (integer(1)).
timeout: (integer(1)).
devices: (character())
Devices to use. Allowed values are "cpu" and "cuda". Default is "cpu".

`AutoResNet$estimate_memory()`

Estimate the memory for the auto.

Usage

AutoResNet$estimate_memory(task)

Arguments

task: (mlr3::Task).

`AutoResNet$clone()`

The objects of this class are cloneable with this method.

Usage

AutoResNet$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

auto("resnet")

SVM Auto

Description

Svm auto.

Value

Object of class R6::R6Class and AutoSVM.

Super class

Auto -> AutoSVM

Methods

Public methods

AutoSVM$new()
AutoSVM$graph()
AutoSVM$design_default()
AutoSVM$clone()

Inherited methods

`AutoSVM$new()`

Creates a new instance of this R6 class.

Usage

AutoSVM$new(id = "svm")

Arguments

id: (character(1))
Identifier for the new instance.

`AutoSVM$graph()`

Create the graph for the auto.

Usage

AutoSVM$graph(task, measure, n_threads, timeout, devices)

Arguments

task: (mlr3::Task).
measure: (mlr3::Measure).
n_threads: (integer(1)).
timeout: (integer(1)).
devices: (character())
Devices to use. Allowed values are "cpu" and "cuda". Default is "cpu".

`AutoSVM$design_default()`

Default hyperparameters for the learner.

Usage

AutoSVM$design_default(task)

Arguments

task: (mlr3::Task).

`AutoSVM$clone()`

The objects of this class are cloneable with this method.

Usage

AutoSVM$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

auto("svm")

TabPFN Auto

Description

Tabpfn auto.

Value

Object of class R6::R6Class and AutoTabPFN.

Python learners

Python learners like TabPFN and fastai run via reticulate and therefore need a Python installation with their required packages. There are two ways to provide it:

Do nothing and let reticulate::py_require() install the required packages into an ephemeral virtual environment automatically.
Point the RETICULATE_PYTHON environment variable to a Python installation that has the required packages installed.

The TabPFN learner additionally requires the TABPFN_TOKEN environment variable to download the model weights.

Super class

Auto -> AutoTabPFN

Methods

Public methods

AutoTabPFN$new()
AutoTabPFN$check()
AutoTabPFN$graph()
AutoTabPFN$estimate_memory()
AutoTabPFN$design_default()
AutoTabPFN$search_space()
AutoTabPFN$clone()

Inherited methods

`AutoTabPFN$new()`

Creates a new instance of this R6 class.

Usage

AutoTabPFN$new(id = "tabpfn")

Arguments

id: (character(1))
Identifier for the new instance.

`AutoTabPFN$check()`

Check if the auto is compatible with the task.

Usage

AutoTabPFN$check(
  task,
  memory_limit = Inf,
  large_data_set = FALSE,
  devices = "cpu"
)

Arguments

task: (mlr3::Task).
memory_limit: (integer(1)).
large_data_set: (logical(1)).
devices: (character())
Devices to use. Allowed values are "cpu" and "cuda". Default is "cpu".

`AutoTabPFN$graph()`

Create the graph for the auto.

Usage

AutoTabPFN$graph(task, measure, n_threads, timeout, devices)

Arguments

task: (mlr3::Task).
measure: (mlr3::Measure).
n_threads: (integer(1)).
timeout: (integer(1)).
devices: (character())
Devices to use. Allowed values are "cpu" and "cuda". Default is "cpu".

`AutoTabPFN$estimate_memory()`

Estimate the memory for the auto.

Usage

AutoTabPFN$estimate_memory(task)

Arguments

task: (mlr3::Task).

`AutoTabPFN$design_default()`

Default hyperparameters for the learner.

Usage

AutoTabPFN$design_default(task)

Arguments

task: (mlr3::Task).

`AutoTabPFN$search_space()`

Get the search space for the auto.

Usage

AutoTabPFN$search_space(task)

Arguments

task: (mlr3::Task).

`AutoTabPFN$clone()`

The objects of this class are cloneable with this method.

Usage

AutoTabPFN$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

auto("tabpfn")

Xgboost Auto

Description

Xgboost auto.

Value

Object of class R6::R6Class and AutoXgboost.

Super class

Auto -> AutoXgboost

Methods

Public methods

AutoXgboost$new()
AutoXgboost$graph()
AutoXgboost$finalize_model()
AutoXgboost$estimate_memory()
AutoXgboost$internal_measure()
AutoXgboost$clone()

Inherited methods

`AutoXgboost$new()`

Creates a new instance of this R6 class.

Usage

AutoXgboost$new(id = "xgboost")

Arguments

id: (character(1))
Identifier for the new instance.

`AutoXgboost$graph()`

Create the graph for the auto.

Usage

AutoXgboost$graph(task, measure, n_threads, timeout, devices)

Arguments

task: (mlr3::Task).
measure: (mlr3::Measure).
n_threads: (integer(1)).
timeout: (integer(1)).
devices: (character())
Devices to use. Allowed values are "cpu" and "cuda". Default is "cpu".

`AutoXgboost$finalize_model()`

Prepare the graph learner for the final model fit.

Usage

AutoXgboost$finalize_model(graph_learner)

Arguments

graph_learner: (mlr3pipelines::GraphLearner).

`AutoXgboost$estimate_memory()`

Estimate the memory for the auto.

Usage

AutoXgboost$estimate_memory(task)

Arguments

task: (mlr3::Task).

`AutoXgboost$internal_measure()`

Get the internal measure for the auto.

Usage

AutoXgboost$internal_measure(measure, task)

Arguments

measure: (mlr3::Measure).
task: (mlr3::Task).

`AutoXgboost$clone()`

The objects of this class are cloneable with this method.

Usage

AutoXgboost$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

auto("xgboost")

Base AutoML Learner

Description

Abstract base class for AutoML learners. Contains the shared logic for LearnerClassifAuto and LearnerRegrAuto.

The following learners are supported:

Algorithm	Package
catboost	mlr3extralearners
extra_trees	mlr3learners
fastai	mlr3extralearners
ft_transformer	mlr3torch
glmnet	mlr3learners
kknn	mlr3learners
lda	mlr3learners
lightgbm	mlr3extralearners
mlp	mlr3torch
ranger	mlr3learners
resnet	mlr3torch
svm	mlr3learners
tabpfn	mlr3extralearners
xgboost	mlr3learners

Debugging

Set options(bbotk.debug = TRUE) to run the tuning in the main session. Set encapsulate_learner = FALSE to remove encapsulation of the learner. Set encapsulate_mbo = FALSE to catch no errors in mbo.

Parameters

learner_timeout: (integer(1))
Timeout for training and predicting with a single learner.
n_threads: (integer(1))
Number of threads used for training a single learner.
memory_limit: (integer(1))
Memory limit for training a single learner in MB. The limit is shared across the parallel workers, i.e. divided by the number of workers.
devices: (character())
Devices to use for model training. Possible values are "cpu" and "cuda". If "cuda", the learner will be trained on a GPU.
large_data_size: (integer(1))
Threshold for the data set size (number of rows times number of columns) above which large-data rules apply. Beyond this threshold the number of parallel workers is reduced and each remaining worker is given proportionally more threads and memory.
small_data_size: (integer(1))
Threshold value for the data set size from which special rules apply.
small_data_resampling: (mlr3::Resampling)
Resampling strategy to use for model training on small data sets.
initial_design_default: (logical(1))
Whether to use the default design of the learner.
initial_design_set: (integer(1))
Number of points to use for the initial design set.
initial_design_size: (integer(1))
Size of the random, sobol or lhs initial design.
initial_design_type: (character(1))
Type of the initial design used for mbo. Possible values are "lhs", "sobol", "random". "lhs" uses a Latin Hypercube Sampling design. "sobol" uses a Sobol sequence design. "random" uses a random design.
initial_design_fraction: (numeric(1))
Fraction of the budget to use for the initial design.
resampling: (mlr3::Resampling)
Resampling strategy used for tuning.
terminator: (bbotk::Terminator)
Terminator criterion for tuning.
measure: (mlr3::Measure)
Measure used for tuning.
callbacks: (mlr3tuning::CallbackAsyncTuning)
Callbacks used for tuning.
store_benchmark_result: (logical(1))
Whether to store the benchmark result.
store_models: (logical(1))
Whether to store the models.
encapsulate_learner: (logical(1))
Whether to encapsulate the learner. Change to FALSE to debug.
encapsulate_mbo: (logical(1))
Whether to encapsulate the tuning. Change to FALSE to debug.
check_learners: (logical(1))
Whether to check if the learners are compatible with the task. Change to FALSE to debug.

Super class

mlr3::Learner -> LearnerAuto

Public fields

instance: (mlr3tuning::TuningInstanceAsyncSingleCrit).
rush: (rush::Rush)
Rush instance for parallel tuning.

Methods

Public methods

LearnerAuto$new()
LearnerAuto$encapsulate()
LearnerAuto$clone()

Inherited methods

`LearnerAuto$new()`

Creates a new instance of this R6 class.

Usage

LearnerAuto$new(
  id,
  learner_ids,
  task_type,
  predict_types,
  properties,
  rush = NULL
)

Arguments

id: (character(1))
Identifier for the new instance.
learner_ids: (character())
Learner that should be used.
task_type: (character(1))
The task type, either "classif" or "regr".
predict_types: (character())
Supported predict types.
properties: (character())
Learner properties.
rush: rush::Rush
Rush instance.

`LearnerAuto$encapsulate()`

Redirects encapsulation to the final model fit.

The AutoML learner itself always trains in the main session, because the rush-based parallel tuning cannot run inside an encapsulated session. The encapsulation method and fallback learner are instead applied to the final model fit that follows the tuning phase. If the final model fit fails, the fallback learner is trained instead. Without encapsulation, a failed final model fit raises an error. The tuning phase is guarded by the encapsulate_learner and encapsulate_mbo parameters instead.

Usage

LearnerAuto$encapsulate(method, fallback = NULL, when = NULL)

Arguments

method: (character(1))
One of "none", "try", "evaluate", "callr", or "mirai".
fallback: (mlr3::Learner)
Learner to train when the final model fit fails.
when: (⁠function()⁠)
Optional condition handler passed to the ⁠$encapsulate()⁠ method of the final model.

Returns

self (invisibly).

`LearnerAuto$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerAuto$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Classification AutoML Learner

Description

The LearnerClassifAuto is an automated machine learning (AutoML) system for classification tasks. It combines preprocessing, a switch between multiple learners, and hyperparameter tuning to find the best model for the given task.

Value

Object of class R6::R6Class and LearnerClassifAuto.

Debugging

Parameters

learner_timeout: (integer(1))
Timeout for training and predicting with a single learner.
n_threads: (integer(1))
Number of threads used for training a single learner.
memory_limit: (integer(1))
Memory limit for training a single learner in MB. The limit is shared across the parallel workers, i.e. divided by the number of workers.
devices: (character())
Devices to use for model training. Possible values are "cpu" and "cuda". If "cuda", the learner will be trained on a GPU.
large_data_size: (integer(1))
Threshold for the data set size (number of rows times number of columns) above which large-data rules apply. Beyond this threshold the number of parallel workers is reduced and each remaining worker is given proportionally more threads and memory.
small_data_size: (integer(1))
Threshold value for the data set size from which special rules apply.
small_data_resampling: (mlr3::Resampling)
Resampling strategy to use for model training on small data sets.
initial_design_default: (logical(1))
Whether to use the default design of the learner.
initial_design_set: (integer(1))
Number of points to use for the initial design set.
initial_design_size: (integer(1))
Size of the random, sobol or lhs initial design.
initial_design_type: (character(1))
Type of the initial design used for mbo. Possible values are "lhs", "sobol", "random". "lhs" uses a Latin Hypercube Sampling design. "sobol" uses a Sobol sequence design. "random" uses a random design.
initial_design_fraction: (numeric(1))
Fraction of the budget to use for the initial design.
resampling: (mlr3::Resampling)
Resampling strategy used for tuning.
terminator: (bbotk::Terminator)
Terminator criterion for tuning.
measure: (mlr3::Measure)
Measure used for tuning.
callbacks: (mlr3tuning::CallbackAsyncTuning)
Callbacks used for tuning.
store_benchmark_result: (logical(1))
Whether to store the benchmark result.
store_models: (logical(1))
Whether to store the models.
encapsulate_learner: (logical(1))
Whether to encapsulate the learner. Change to FALSE to debug.
encapsulate_mbo: (logical(1))
Whether to encapsulate the tuning. Change to FALSE to debug.
check_learners: (logical(1))
Whether to check if the learners are compatible with the task. Change to FALSE to debug.

Python learners

Python learners like TabPFN and fastai run via reticulate and therefore need a Python installation with their required packages. There are two ways to provide it:

Do nothing and let reticulate::py_require() install the required packages into an ephemeral virtual environment automatically.
Point the RETICULATE_PYTHON environment variable to a Python installation that has the required packages installed.

The TabPFN learner additionally requires the TABPFN_TOKEN environment variable to download the model weights.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerClassifAuto

Methods

Public methods

LearnerClassifAuto$new()
LearnerClassifAuto$clone()

Inherited methods

`LearnerClassifAuto$new()`

Creates a new instance of this R6 class.

Usage

LearnerClassifAuto$new(id = "classif.auto", learner_ids, rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
learner_ids: (character())
Learner that should be used.
rush: rush::Rush
Rush instance.

`LearnerClassifAuto$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerClassifAuto$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

packages = c(
  "mlr3extralearners", "catboost", "ranger", "callr", "mlr3torch",
  "glmnet", "kknn", "MASS", "lightgbm", "e1071", "xgboost"
)
if (mlr3misc::require_namespaces(packages, quietly = TRUE)) {
  learner = lrn("classif.auto")
  learner
}

Classification Gradient Boosted Decision Trees Auto Learner

Description

Classification auto learner.

Value

Object of class R6::R6Class and LearnerClassifAutoCatboost.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerClassifAuto -> LearnerClassifAutoCatboost

Methods

Public methods

LearnerClassifAutoCatboost$new()
LearnerClassifAutoCatboost$clone()

Inherited methods

`LearnerClassifAutoCatboost$new()`

Creates a new instance of this R6 class.

Usage

LearnerClassifAutoCatboost$new(id = "classif.auto_catboost", rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerClassifAutoCatboost$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerClassifAutoCatboost$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces(c("mlr3extralearners", "catboost"), quietly = TRUE)) {
  learner = lrn("classif.auto_catboost")
  learner
}

Classification FT-Transformer Auto Learner

Description

Classification auto learner.

Value

Object of class R6::R6Class and LearnerClassifAutoFTTransformer.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerClassifAuto -> LearnerClassifAutoFTTransformer

Methods

Public methods

LearnerClassifAutoFTTransformer$new()
LearnerClassifAutoFTTransformer$clone()

Inherited methods

`LearnerClassifAutoFTTransformer$new()`

Creates a new instance of this R6 class.

Usage

LearnerClassifAutoFTTransformer$new(
  id = "classif.auto_ft_transformer",
  rush = NULL
)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerClassifAutoFTTransformer$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerClassifAutoFTTransformer$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces("mlr3torch", quietly = TRUE)) {
  learner = lrn("classif.auto_ft_transformer")
  learner
}

Classification Fastai Auto Learner

Description

Classification auto learner.

Value

Object of class R6::R6Class and LearnerClassifAutoFastai.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerClassifAuto -> LearnerClassifAutoFastai

Methods

Public methods

LearnerClassifAutoFastai$new()
LearnerClassifAutoFastai$clone()

Inherited methods

`LearnerClassifAutoFastai$new()`

Creates a new instance of this R6 class.

Usage

LearnerClassifAutoFastai$new(id = "classif.auto_fastai", rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerClassifAutoFastai$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerClassifAutoFastai$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces(c("mlr3extralearners", "callr"), quietly = TRUE)) {
  learner = lrn("classif.auto_fastai")
  learner
}

Classification GLM with Elastic Net Regularization Auto Learner

Description

Classification auto learner.

Value

Object of class R6::R6Class and LearnerClassifAutoGlmnet.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerClassifAuto -> LearnerClassifAutoGlmnet

Methods

Public methods

LearnerClassifAutoGlmnet$new()
LearnerClassifAutoGlmnet$clone()

Inherited methods

`LearnerClassifAutoGlmnet$new()`

Creates a new instance of this R6 class.

Usage

LearnerClassifAutoGlmnet$new(id = "classif.auto_glmnet", rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerClassifAutoGlmnet$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerClassifAutoGlmnet$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces("glmnet", quietly = TRUE)) {
  learner = lrn("classif.auto_glmnet")
  learner
}

Classification k-Nearest-Neighbor Auto Learner

Description

Classification auto learner.

Value

Object of class R6::R6Class and LearnerClassifAutoKKNN.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerClassifAuto -> LearnerClassifAutoKKNN

Methods

Public methods

LearnerClassifAutoKKNN$new()
LearnerClassifAutoKKNN$clone()

Inherited methods

`LearnerClassifAutoKKNN$new()`

Creates a new instance of this R6 class.

Usage

LearnerClassifAutoKKNN$new(id = "classif.auto_kknn", rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerClassifAutoKKNN$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerClassifAutoKKNN$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces("kknn", quietly = TRUE)) {
  learner = lrn("classif.auto_kknn")
  learner
}

Classification LightGBM Auto Learner

Description

Classification auto learner.

Value

Object of class R6::R6Class and LearnerClassifAutoLightGBM.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerClassifAuto -> LearnerClassifAutoLightGBM

Methods

Public methods

LearnerClassifAutoLightGBM$new()
LearnerClassifAutoLightGBM$clone()

Inherited methods

`LearnerClassifAutoLightGBM$new()`

Creates a new instance of this R6 class.

Usage

LearnerClassifAutoLightGBM$new(id = "classif.auto_lightgbm", rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerClassifAutoLightGBM$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerClassifAutoLightGBM$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces(c("mlr3extralearners", "lightgbm"), quietly = TRUE)) {
  learner = lrn("classif.auto_lightgbm")
  learner
}

Classification MLP Auto Learner

Description

Classification auto learner.

Value

Object of class R6::R6Class and LearnerClassifAutoMLP.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerClassifAuto -> LearnerClassifAutoMLP

Methods

Public methods

LearnerClassifAutoMLP$new()
LearnerClassifAutoMLP$clone()

Inherited methods

`LearnerClassifAutoMLP$new()`

Creates a new instance of this R6 class.

Usage

LearnerClassifAutoMLP$new(id = "classif.auto_mlp", rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerClassifAutoMLP$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerClassifAutoMLP$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces("mlr3torch", quietly = TRUE)) {
  learner = lrn("classif.auto_mlp")
  learner
}

Classification Ranger Auto Learner

Description

Classification auto learner.

Value

Object of class R6::R6Class and LearnerClassifAutoRanger.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerClassifAuto -> LearnerClassifAutoRanger

Methods

Public methods

LearnerClassifAutoRanger$new()
LearnerClassifAutoRanger$clone()

Inherited methods

`LearnerClassifAutoRanger$new()`

Creates a new instance of this R6 class.

Usage

LearnerClassifAutoRanger$new(id = "classif.auto_ranger", rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerClassifAutoRanger$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerClassifAutoRanger$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces("ranger", quietly = TRUE)) {
  learner = lrn("classif.auto_ranger")
  learner
}

Classification ResNet Auto Learner

Description

Classification auto learner.

Value

Object of class R6::R6Class and LearnerClassifAutoResNet.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerClassifAuto -> LearnerClassifAutoResNet

Methods

Public methods

LearnerClassifAutoResNet$new()
LearnerClassifAutoResNet$clone()

Inherited methods

`LearnerClassifAutoResNet$new()`

Creates a new instance of this R6 class.

Usage

LearnerClassifAutoResNet$new(id = "classif.auto_resnet", rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerClassifAutoResNet$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerClassifAutoResNet$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces("mlr3torch", quietly = TRUE)) {
  learner = lrn("classif.auto_resnet")
  learner
}

Classification SVM Auto Learner

Description

Classification auto learner.

Value

Object of class R6::R6Class and LearnerClassifAutoSVM.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerClassifAuto -> LearnerClassifAutoSVM

Methods

Public methods

LearnerClassifAutoSVM$new()
LearnerClassifAutoSVM$clone()

Inherited methods

`LearnerClassifAutoSVM$new()`

Creates a new instance of this R6 class.

Usage

LearnerClassifAutoSVM$new(id = "classif.auto_svm", rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerClassifAutoSVM$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerClassifAutoSVM$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces("e1071", quietly = TRUE)) {
  learner = lrn("classif.auto_svm")
  learner
}

Classification TabPFN Auto Learner

Description

Classification auto learner.

Value

Object of class R6::R6Class and LearnerClassifAutoTabPFN.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerClassifAuto -> LearnerClassifAutoTabPFN

Methods

Public methods

LearnerClassifAutoTabPFN$new()
LearnerClassifAutoTabPFN$clone()

Inherited methods

`LearnerClassifAutoTabPFN$new()`

Creates a new instance of this R6 class.

Usage

LearnerClassifAutoTabPFN$new(id = "classif.auto_tabpfn", rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerClassifAutoTabPFN$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerClassifAutoTabPFN$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces(c("mlr3extralearners", "callr"), quietly = TRUE)) {
  learner = lrn("classif.auto_tabpfn")
  learner
}

Classification XGBoost Auto Learner

Description

Classification auto learner.

Value

Object of class R6::R6Class and LearnerClassifAutoXgboost.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerClassifAuto -> LearnerClassifAutoXgboost

Methods

Public methods

LearnerClassifAutoXgboost$new()
LearnerClassifAutoXgboost$clone()

Inherited methods

`LearnerClassifAutoXgboost$new()`

Creates a new instance of this R6 class.

Usage

LearnerClassifAutoXgboost$new(id = "classif.auto_xgboost", rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerClassifAutoXgboost$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerClassifAutoXgboost$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces("xgboost", quietly = TRUE)) {
  learner = lrn("classif.auto_xgboost")
  learner
}

Fastai Learner Isolated

Description

A subclass of mlr3extralearners::LearnerClassifFastai that isolates the Python environment in a callr session.

Value

Object of class R6::R6Class and LearnerClassifFastaiIsolated.

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> mlr3extralearners::LearnerClassifFastai -> LearnerClassifFastaiIsolated

Methods

Public methods

LearnerClassifFastaiIsolated$new()
LearnerClassifFastaiIsolated$clone()

Inherited methods

`LearnerClassifFastaiIsolated$new()`

Creates a new instance of this R6 class.

Usage

LearnerClassifFastaiIsolated$new()

`LearnerClassifFastaiIsolated$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerClassifFastaiIsolated$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

TabPFN Learner Isolated

Description

A subclass of mlr3extralearners::LearnerClassifTabPFN that isolates the Python environment in a callr session.

Value

Object of class R6::R6Class and LearnerClassifTabPFNIsolated.

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> mlr3extralearners::LearnerClassifTabPFN -> LearnerClassifTabPFNIsolated

Methods

Public methods

LearnerClassifTabPFNIsolated$new()
LearnerClassifTabPFNIsolated$clone()

Inherited methods

`LearnerClassifTabPFNIsolated$new()`

Creates a new instance of this R6 class.

Usage

LearnerClassifTabPFNIsolated$new()

`LearnerClassifTabPFNIsolated$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerClassifTabPFNIsolated$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Regression AutoML Learner

Description

The LearnerRegrAuto is an automated machine learning (AutoML) system for regression tasks. It combines preprocessing, a switch between multiple learners, and hyperparameter tuning to find the best model for the given task.

Value

Object of class R6::R6Class and LearnerRegrAuto.

Debugging

Parameters

learner_timeout: (integer(1))
Timeout for training and predicting with a single learner.
n_threads: (integer(1))
Number of threads used for training a single learner.
memory_limit: (integer(1))
Memory limit for training a single learner in MB. The limit is shared across the parallel workers, i.e. divided by the number of workers.
devices: (character())
Devices to use for model training. Possible values are "cpu" and "cuda". If "cuda", the learner will be trained on a GPU.
large_data_size: (integer(1))
Threshold for the data set size (number of rows times number of columns) above which large-data rules apply. Beyond this threshold the number of parallel workers is reduced and each remaining worker is given proportionally more threads and memory.
small_data_size: (integer(1))
Threshold value for the data set size from which special rules apply.
small_data_resampling: (mlr3::Resampling)
Resampling strategy to use for model training on small data sets.
initial_design_default: (logical(1))
Whether to use the default design of the learner.
initial_design_set: (integer(1))
Number of points to use for the initial design set.
initial_design_size: (integer(1))
Size of the random, sobol or lhs initial design.
initial_design_type: (character(1))
Type of the initial design used for mbo. Possible values are "lhs", "sobol", "random". "lhs" uses a Latin Hypercube Sampling design. "sobol" uses a Sobol sequence design. "random" uses a random design.
initial_design_fraction: (numeric(1))
Fraction of the budget to use for the initial design.
resampling: (mlr3::Resampling)
Resampling strategy used for tuning.
terminator: (bbotk::Terminator)
Terminator criterion for tuning.
measure: (mlr3::Measure)
Measure used for tuning.
callbacks: (mlr3tuning::CallbackAsyncTuning)
Callbacks used for tuning.
store_benchmark_result: (logical(1))
Whether to store the benchmark result.
store_models: (logical(1))
Whether to store the models.
encapsulate_learner: (logical(1))
Whether to encapsulate the learner. Change to FALSE to debug.
encapsulate_mbo: (logical(1))
Whether to encapsulate the tuning. Change to FALSE to debug.
check_learners: (logical(1))
Whether to check if the learners are compatible with the task. Change to FALSE to debug.

Python learners

Python learners like TabPFN and fastai run via reticulate and therefore need a Python installation with their required packages. There are two ways to provide it:

Do nothing and let reticulate::py_require() install the required packages into an ephemeral virtual environment automatically.
Point the RETICULATE_PYTHON environment variable to a Python installation that has the required packages installed.

The TabPFN learner additionally requires the TABPFN_TOKEN environment variable to download the model weights.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerRegrAuto

Methods

Public methods

LearnerRegrAuto$new()
LearnerRegrAuto$clone()

Inherited methods

`LearnerRegrAuto$new()`

Creates a new instance of this R6 class.

Usage

LearnerRegrAuto$new(id = "regr.auto", learner_ids, rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
learner_ids: (character())
Learner that should be used.
rush: rush::Rush
Rush instance.

`LearnerRegrAuto$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerRegrAuto$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

packages = c(
  "mlr3extralearners", "catboost", "ranger", "callr", "mlr3torch",
  "glmnet", "kknn", "MASS", "lightgbm", "e1071", "xgboost"
)
if (mlr3misc::require_namespaces(packages, quietly = TRUE)) {
  learner = lrn("regr.auto")
  learner
}

Regression Gradient Boosted Decision Trees Auto Learner

Description

Regression auto learner.

Value

Object of class R6::R6Class and LearnerRegrAutoCatboost.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerRegrAuto -> LearnerRegrAutoCatboost

Methods

Public methods

LearnerRegrAutoCatboost$new()
LearnerRegrAutoCatboost$clone()

Inherited methods

`LearnerRegrAutoCatboost$new()`

Creates a new instance of this R6 class.

Usage

LearnerRegrAutoCatboost$new(id = "regr.auto_catboost", rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerRegrAutoCatboost$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerRegrAutoCatboost$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces(c("mlr3extralearners", "catboost"), quietly = TRUE)) {
  learner = lrn("regr.auto_catboost")
  learner
}

Regression FT-Transformer Auto Learner

Description

Regression auto learner.

Value

Object of class R6::R6Class and LearnerRegrAutoFTTransformer.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerRegrAuto -> LearnerRegrAutoFTTransformer

Methods

Public methods

LearnerRegrAutoFTTransformer$new()
LearnerRegrAutoFTTransformer$clone()

Inherited methods

`LearnerRegrAutoFTTransformer$new()`

Creates a new instance of this R6 class.

Usage

LearnerRegrAutoFTTransformer$new(id = "regr.auto_ft_transformer", rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerRegrAutoFTTransformer$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerRegrAutoFTTransformer$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces("mlr3torch", quietly = TRUE)) {
  learner = lrn("regr.auto_ft_transformer")
  learner
}

Regression GLM with Elastic Net Regularization Auto Learner

Description

Regression auto learner.

Value

Object of class R6::R6Class and LearnerRegrAutoGlmnet.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerRegrAuto -> LearnerRegrAutoGlmnet

Methods

Public methods

LearnerRegrAutoGlmnet$new()
LearnerRegrAutoGlmnet$clone()

Inherited methods

`LearnerRegrAutoGlmnet$new()`

Creates a new instance of this R6 class.

Usage

LearnerRegrAutoGlmnet$new(id = "regr.auto_glmnet", rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerRegrAutoGlmnet$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerRegrAutoGlmnet$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces("glmnet", quietly = TRUE)) {
  learner = lrn("regr.auto_glmnet")
  learner
}

Regression k-Nearest-Neighbor Auto Learner

Description

Regression auto learner.

Value

Object of class R6::R6Class and LearnerRegrAutoKKNN.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerRegrAuto -> LearnerRegrAutoKKNN

Methods

Public methods

LearnerRegrAutoKKNN$new()
LearnerRegrAutoKKNN$clone()

Inherited methods

`LearnerRegrAutoKKNN$new()`

Creates a new instance of this R6 class.

Usage

LearnerRegrAutoKKNN$new(id = "regr.auto_kknn", rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerRegrAutoKKNN$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerRegrAutoKKNN$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces("kknn", quietly = TRUE)) {
  learner = lrn("regr.auto_kknn")
  learner
}

Regression LightGBM Auto Learner

Description

Regression auto learner.

Value

Object of class R6::R6Class and LearnerRegrAutoLightGBM.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerRegrAuto -> LearnerRegrAutoLightGBM

Methods

Public methods

LearnerRegrAutoLightGBM$new()
LearnerRegrAutoLightGBM$clone()

Inherited methods

`LearnerRegrAutoLightGBM$new()`

Creates a new instance of this R6 class.

Usage

LearnerRegrAutoLightGBM$new(id = "regr.auto_lightgbm", rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerRegrAutoLightGBM$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerRegrAutoLightGBM$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces(c("mlr3extralearners", "lightgbm"), quietly = TRUE)) {
  learner = lrn("regr.auto_lightgbm")
  learner
}

Regression MLP Auto Learner

Description

Regression auto learner.

Value

Object of class R6::R6Class and LearnerRegrAutoMLP.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerRegrAuto -> LearnerRegrAutoMLP

Methods

Public methods

LearnerRegrAutoMLP$new()
LearnerRegrAutoMLP$clone()

Inherited methods

`LearnerRegrAutoMLP$new()`

Creates a new instance of this R6 class.

Usage

LearnerRegrAutoMLP$new(id = "regr.auto_mlp", rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerRegrAutoMLP$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerRegrAutoMLP$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces("mlr3torch", quietly = TRUE)) {
  learner = lrn("regr.auto_mlp")
  learner
}

Regression Ranger Auto Learner

Description

Regression auto learner.

Value

Object of class R6::R6Class and LearnerRegrAutoRanger.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerRegrAuto -> LearnerRegrAutoRanger

Methods

Public methods

LearnerRegrAutoRanger$new()
LearnerRegrAutoRanger$clone()

Inherited methods

`LearnerRegrAutoRanger$new()`

Creates a new instance of this R6 class.

Usage

LearnerRegrAutoRanger$new(id = "regr.auto_ranger", rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerRegrAutoRanger$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerRegrAutoRanger$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces("ranger", quietly = TRUE)) {
  learner = lrn("regr.auto_ranger")
  learner
}

Regression ResNet Auto Learner

Description

Regression auto learner.

Value

Object of class R6::R6Class and LearnerRegrAutoResNet.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerRegrAuto -> LearnerRegrAutoResNet

Methods

Public methods

LearnerRegrAutoResNet$new()
LearnerRegrAutoResNet$clone()

Inherited methods

`LearnerRegrAutoResNet$new()`

Creates a new instance of this R6 class.

Usage

LearnerRegrAutoResNet$new(id = "regr.auto_resnet", rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerRegrAutoResNet$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerRegrAutoResNet$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces("mlr3torch", quietly = TRUE)) {
  learner = lrn("regr.auto_resnet")
  learner
}

Regression SVM Auto Learner

Description

Regression auto learner.

Value

Object of class R6::R6Class and LearnerRegrAutoSVM.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerRegrAuto -> LearnerRegrAutoSVM

Methods

Public methods

LearnerRegrAutoSVM$new()
LearnerRegrAutoSVM$clone()

Inherited methods

`LearnerRegrAutoSVM$new()`

Creates a new instance of this R6 class.

Usage

LearnerRegrAutoSVM$new(id = "regr.auto_svm", rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerRegrAutoSVM$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerRegrAutoSVM$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces("e1071", quietly = TRUE)) {
  learner = lrn("regr.auto_svm")
  learner
}

Regression TabPFN Auto Learner

Description

Regression auto learner.

Value

Object of class R6::R6Class and LearnerRegrAutoTabPFN.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerRegrAuto -> LearnerRegrAutoTabPFN

Methods

Public methods

LearnerRegrAutoTabPFN$new()
LearnerRegrAutoTabPFN$clone()

Inherited methods

`LearnerRegrAutoTabPFN$new()`

Creates a new instance of this R6 class.

Usage

LearnerRegrAutoTabPFN$new(id = "regr.auto_tabpfn", rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerRegrAutoTabPFN$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerRegrAutoTabPFN$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces(c("mlr3extralearners", "callr"), quietly = TRUE)) {
  learner = lrn("regr.auto_tabpfn")
  learner
}

Regression XGBoost Auto Learner

Description

Regression auto learner.

Value

Object of class R6::R6Class and LearnerRegrAutoXgboost.

Super classes

mlr3::Learner -> LearnerAuto -> LearnerRegrAuto -> LearnerRegrAutoXgboost

Methods

Public methods

LearnerRegrAutoXgboost$new()
LearnerRegrAutoXgboost$clone()

Inherited methods

`LearnerRegrAutoXgboost$new()`

Creates a new instance of this R6 class.

Usage

LearnerRegrAutoXgboost$new(id = "regr.auto_xgboost", rush = NULL)

Arguments

id: (character(1))
Identifier for the new instance.
rush: rush::Rush
Rush instance.

`LearnerRegrAutoXgboost$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerRegrAutoXgboost$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Examples

if (mlr3misc::require_namespaces("xgboost", quietly = TRUE)) {
  learner = lrn("regr.auto_xgboost")
  learner
}

TabPFN Regressor Learner Isolated

Description

A subclass of mlr3extralearners::LearnerRegrTabPFN that isolates the Python environment in a callr session.

Value

Object of class R6::R6Class and LearnerRegrTabPFNIsolated.

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> mlr3extralearners::LearnerRegrTabPFN -> LearnerRegrTabPFNIsolated

Methods

Public methods

LearnerRegrTabPFNIsolated$new()
LearnerRegrTabPFNIsolated$clone()

Inherited methods

`LearnerRegrTabPFNIsolated$new()`

Creates a new instance of this R6 class.

Usage

LearnerRegrTabPFNIsolated$new()

`LearnerRegrTabPFNIsolated$clone()`

The objects of this class are cloneable with this method.

Usage

LearnerRegrTabPFNIsolated$clone(deep = FALSE)

Arguments

deep: Whether to make a deep clone.

Install Python Learners

Description

Creates a conda environment and installs the packages required by the Python learners. Installs the dependencies of the fastai auto learner and the tabpfn auto learner. The learners share a single environment, so one RETICULATE_PYTHON covers both.

Usage

install_python_learners(
  learners = c("fastai", "tabpfn"),
  envname = file.path(getwd(), ".conda", "mlr3automl-python"),
  python_version = "3.12"
)

Arguments

learners

(character())
Learners to install the Python dependencies for. One or more of "fastai" and "tabpfn".

envname

(character(1))
Path to the conda environment directory. Defaults to .conda/mlr3automl-python in the current working directory.

python_version

(character(1))
Python version to use. Pinned to "3.12" by default so the environment is reproducible across machines.

Details

The environment is created in the project directory by default. The function downloads and installs Python packages and must therefore be called explicitly by the user. It is never called by any other function of the package.

Value

Invisibly returns the path to the Python binary in the environment.

Examples

# The example is wrapped in \dontrun{} because it cannot be executed during checks.
# It downloads and installs conda, a Python interpreter, and several Python packages,
# which requires network access, several GB of disk space, and a few minutes of runtime.
## Not run: 
python = install_python_learners()
Sys.setenv(RETICULATE_PYTHON = python)

## End(Not run)

Encapsulation Daemon Callback

Description

This mlr3misc::Callback starts a persistent mirai daemon on each rush worker that is reused for the "mirai" encapsulation of the tuned learners. Reusing a single daemon avoids the overhead of starting a new daemon for every learner evaluation. The daemon is started when the worker begins and its liveness is checked before every evaluation. If the daemon has died, it is restarted.

Parameters

n_daemons :: integer(1)
Number of daemons to start on each worker. Defaults to 1.

Examples

clbk("mlr3automl.encapsulation_daemon", n_daemons = 1)

Initial Design Runtime Limit Callback

Description

This mlr3misc::Callback drops the remaining tasks of the initial design from the queue when a configurable fraction (initial_design_fraction) of the runtime limit is reached. The initial_design_fraction is set to 0.25 by default.

Examples

clbk("mlr3automl.initial_design_runtime", initial_design_fraction = 0.5)

Dictionary of Auto Objects

Description

A dictionary of Auto objects.

Sugar function to retrieve Auto objects from mlr_auto.

Usage

mlr_auto

auto(.key, ...)

Arguments

.key

(character(1))
Key of the object to retrieve. If missing, the dictionary itself is returned.

...

(named list())
Additional arguments passed to the constructor.

Value

Auto

Examples

auto("catboost")

Package {mlr3automl}

mlr3automl: Automated Machine Learning for 'mlr3'

Description

Author(s)

See Also

Auto Class

Description

Value

Public fields

Methods

Public methods

Auto$new()

Usage

Arguments

Auto$check()

Usage

Arguments

Auto$graph()

Usage

Arguments

Auto$early_stopping_rounds()

Usage

Arguments

Auto$estimate_memory()

Usage

Arguments

Auto$finalize_model()

Usage

Arguments

Auto$design_default()

Usage

Arguments

Auto$design_set()

Usage

Arguments

Auto$search_space()

Usage

Arguments

Auto$clone()

Usage

Arguments

Catboost Auto

Description

Value

Super class

Methods

Public methods

AutoCatboost$new()

Usage

Arguments

AutoCatboost$graph()

Usage

Arguments

AutoCatboost$estimate_memory()

Usage

Arguments

AutoCatboost$internal_measure()

Usage

Arguments

AutoCatboost$clone()

Usage

Arguments

Examples

Extra Trees Auto

Description

Value

Super class

Methods

Public methods

AutoExtraTrees$new()

Usage

Arguments

AutoExtraTrees$graph()

Usage

Arguments

AutoExtraTrees$estimate_memory()

Usage

Arguments

AutoExtraTrees$clone()

Usage

`Auto$new()`

`Auto$check()`

`Auto$graph()`

`Auto$early_stopping_rounds()`

`Auto$estimate_memory()`

`Auto$finalize_model()`

`Auto$design_default()`

`Auto$design_set()`

`Auto$search_space()`

`Auto$clone()`

`AutoCatboost$new()`

`AutoCatboost$graph()`

`AutoCatboost$estimate_memory()`

`AutoCatboost$internal_measure()`

`AutoCatboost$clone()`

`AutoExtraTrees$new()`

`AutoExtraTrees$graph()`

`AutoExtraTrees$estimate_memory()`

`AutoExtraTrees$clone()`

`AutoFTTransformer$new()`

`AutoFTTransformer$graph()`

`AutoFTTransformer$estimate_memory()`

`AutoFTTransformer$clone()`

`AutoFastai$new()`

`AutoFastai$check()`

`AutoFastai$graph()`

`AutoFastai$estimate_memory()`

`AutoFastai$internal_measure()`

`AutoFastai$clone()`

`AutoGlmnet$new()`

`AutoGlmnet$graph()`

`AutoGlmnet$clone()`