10 Appendix

10.1 Integrated Learners

Learners are available from one of the following packages:

  • mlr3: debug learner and rpart learners.
  • mlr3learners: opinionated selection of some default learners.
  • mlr3proba: base learners for survival and probabilistic regression.
  • mlr3cluster: learners for unsupervised clustering.
  • mlr3extralearners: more experimental learners for regression, classification and survival.

Use the interactive search table to look through all our learners.

10.2 Integrated Performance Measures

Also see the overview on the website of mlr3measures.

Id Packages Task Type Predict Type
aic NA response
bic NA response
classif.acc mlr3measures classif response
classif.auc mlr3measures classif prob
classif.bacc mlr3measures classif response
classif.bbrier mlr3measures classif prob
classif.ce mlr3measures classif response
classif.costs classif response
classif.dor mlr3measures classif response
classif.fbeta mlr3measures classif response
classif.fdr mlr3measures classif response
classif.fn mlr3measures classif response
classif.fnr mlr3measures classif response
classif.fomr mlr3measures classif response
classif.fp mlr3measures classif response
classif.fpr mlr3measures classif response
classif.logloss mlr3measures classif prob
classif.mbrier mlr3measures classif prob
classif.mcc mlr3measures classif response
classif.npv mlr3measures classif response
classif.ppv mlr3measures classif response
classif.prauc mlr3measures classif prob
classif.precision mlr3measures classif response
classif.recall mlr3measures classif response
classif.sensitivity mlr3measures classif response
classif.specificity mlr3measures classif response
classif.tn mlr3measures classif response
classif.tnr mlr3measures classif response
classif.tp mlr3measures classif response
classif.tpr mlr3measures classif response
debug NA response
dens.logloss dens pdf
oob_error NA response
regr.bias mlr3measures regr response
regr.ktau mlr3measures regr response
regr.mae mlr3measures regr response
regr.mape mlr3measures regr response
regr.maxae mlr3measures regr response
regr.medae mlr3measures regr response
regr.medse mlr3measures regr response
regr.mse mlr3measures regr response
regr.msle mlr3measures regr response
regr.pbias mlr3measures regr response
regr.rae mlr3measures regr response
regr.rmse mlr3measures regr response
regr.rmsle mlr3measures regr response
regr.rrse mlr3measures regr response
regr.rse mlr3measures regr response
regr.rsq mlr3measures regr response
regr.sae mlr3measures regr response
regr.smape mlr3measures regr response
regr.srho mlr3measures regr response
regr.sse mlr3measures regr response
selected_features NA response
surv.brier surv distr
surv.calib_alpha surv distr
surv.calib_beta surv lp
surv.chambless_auc survAUC surv lp
surv.cindex surv crank
surv.dcalib surv distr
surv.graf surv distr
surv.hung_auc survAUC surv lp
surv.intlogloss distr6 surv distr
surv.logloss distr6 surv distr
surv.mae surv response
surv.mse surv response
surv.nagelk_r2 survAUC surv lp
surv.oquigley_r2 survAUC surv lp
surv.rmse surv response
surv.schmid distr6 surv distr
surv.song_auc survAUC surv lp
surv.song_tnr survAUC surv lp
surv.song_tpr survAUC surv lp
surv.uno_auc survAUC surv lp
surv.uno_tnr survAUC surv lp
surv.uno_tpr survAUC surv lp
surv.xu_r2 survAUC surv lp
time_both NA response
time_predict NA response
time_train NA response

10.3 Integrated Filter Methods

10.3.1 Standalone filter methods

Id Packages Task Types Feature Types
anova stats classif int, dbl
auc mlr3measures classif int, dbl
carscore care regr dbl
cmim praznik classif, regr int, dbl, fct, ord
correlation stats regr int, dbl
disr praznik classif, regr int, dbl, fct, ord
find_correlation stats classif, regr int, dbl
importance rpart classif lgl, int, dbl, fct, ord
information_gain FSelectorRcpp classif, regr int, dbl, fct, ord
jmi praznik classif, regr int, dbl, fct, ord
jmim praznik classif, regr int, dbl, fct, ord
kruskal_test stats classif int, dbl
mim praznik classif, regr int, dbl, fct, ord
mrmr praznik classif, regr int, dbl, fct, ord
njmim praznik classif, regr int, dbl, fct, ord
performance mlr3measures, rpart classif lgl, int, dbl, fct, ord
permutation mlr3measures, rpart classif lgl, int, dbl, fct, ord
relief FSelectorRcpp classif, regr int, dbl, fct, ord
variance stats classif, regr int, dbl

10.3.2 Learners With Embedded Filter Methods

##  [1] "classif.featureless" "classif.ranger"      "classif.rpart"      
##  [4] "classif.xgboost"     "regr.featureless"    "regr.ranger"        
##  [7] "regr.rpart"          "regr.xgboost"        "surv.ranger"        
## [10] "surv.rpart"          "surv.xgboost"

10.4 Integrated Pipe Operators

Id Packages Tags Train Predict
boxcox bestNormalize data transform Task → Task Task→Task
colapply data transform Task → Task Task→Task
collapsefactors data transform Task → Task Task→Task
colroles data transform Task → Task Task→Task
datefeatures data transform Task → Task Task→Task
histbin graphics data transform Task → Task Task→Task
ica fastICA data transform Task → Task Task→Task
kernelpca kernlab data transform Task → Task Task→Task
modelmatrix stats data transform Task → Task Task→Task
mutate data transform Task → Task Task→Task
nmf MASS, NMF data transform Task → Task Task→Task
pca data transform Task → Task Task→Task
quantilebin stats data transform Task → Task Task→Task
randomprojection data transform Task → Task Task→Task
renamecolumns data transform Task → Task Task→Task
scale data transform Task → Task Task→Task
scalemaxabs data transform Task → Task Task→Task
scalerange data transform Task → Task Task→Task
spatialsign data transform Task → Task Task→Task
subsample data transform Task → Task Task→Task
textvectorizer quanteda, stopwords data transform Task → Task Task→Task
yeojohnson bestNormalize data transform Task → Task Task→Task
encode stats encode , data transform Task → Task Task→Task
encodeimpact encode , data transform Task → Task Task→Task
encodelmer lme4, nloptr encode , data transform Task → Task Task→Task
vtreat vtreat encode , missings , data transform Task → Task Task→Task
classifavg stats ensemble NULL → NULL PredictionClassif→PredictionClassif
featureunion ensemble Task → Task Task→Task
regravg ensemble NULL → NULL PredictionRegr→PredictionRegr
survavg ensemble NULL → NULL PredictionSurv→PredictionSurv
filter feature selection, data transform Task → Task Task→Task
select feature selection, data transform Task → Task Task→Task
classbalancing imbalanced data, data transform TaskClassif → TaskClassif TaskClassif→TaskClassif
classweights imbalanced data, data transform TaskClassif → TaskClassif TaskClassif→TaskClassif
smote smotefamily imbalanced data, data transform Task → Task Task→Task
learner learner TaskClassif → NULL TaskClassif→PredictionClassif
learner_cv learner , ensemble , data transform TaskClassif → TaskClassif TaskClassif→TaskClassif
imputeconstant missings Task → Task Task→Task
imputehist graphics missings Task → Task Task→Task
imputelearner missings Task → Task Task→Task
imputemean missings Task → Task Task→Task
imputemedian stats missings Task → Task Task→Task
imputemode missings Task → Task Task→Task
imputeoor missings Task → Task Task→Task
imputesample missings Task → Task Task→Task
missind missings , data transform Task → Task Task→Task
multiplicityexply multiplicity [*] → * [*]→*
multiplicityimply multiplicity * → [*] *→[*]
ovrunite multiplicity, ensemble [NULL] → NULL [PredictionClassif]→PredictionClassif
replicate multiplicity * → [*] *→[*]
fixfactors robustify , data transform Task → Task Task→Task
removeconstants robustify , data transform Task → Task Task→Task
ovrsplit target transform, multiplicity TaskClassif → [TaskClassif] TaskClassif→[TaskClassif]
targetmutate target transform Task → NULL, Task Task→function, Task
targettrafoscalerange target transform TaskRegr → NULL, TaskRegr TaskRegr→function, TaskRegr
threshold target transform NULL → NULL PredictionClassif→PredictionClassif
tunethreshold bbotk target transform Task → NULL Task→Prediction
branch meta * → *
chunk meta Task → Task Task→Task
copy meta * → *
nop meta * → *
proxy meta * → *
unbranch meta * → *
compose_crank distr6 abstract NULL → NULL PredictionSurv→PredictionSurv
compose_distr distr6 abstract NULL, NULL → NULL PredictionSurv, PredictionSurv→PredictionSurv
compose_probregr distr6 abstract NULL, NULL → NULL PredictionRegr, PredictionRegr→PredictionRegr
crankcompose distr6 abstract NULL → NULL PredictionSurv→PredictionSurv
distrcompose distr6 abstract NULL, NULL → NULL PredictionSurv, PredictionSurv→PredictionSurv
randomresponse abstract NULL → NULL Prediction→Prediction
targetinvert abstract NULL, NULL → NULL function, Prediction→Prediction
trafopred_regrsurv abstract NULL, NULL → NULL PredictionRegr, *→PredictionSurv
trafopred_survregr abstract NULL → NULL PredictionSurv→PredictionRegr
trafotask_regrsurv abstract TaskRegr, * → TaskSurv TaskRegr, *→TaskSurv
trafotask_survregr abstract TaskSurv, * → TaskRegr TaskSurv, *→TaskRegr

10.5 Framework Comparison

Below, we collected some examples, where mlr3pipelines is compared to different other software packages, such as mlr, recipes and sklearn.

Before diving deeper, we give a short introduction to PipeOps.

10.5.1 An introduction to “PipeOp”s

In this example, we create a linear Pipeline. After scaling all input features, we rotate our data using principal component analysis. After this transformation, we use a simple Decision Tree learner for classification.

As exemplary data, we will use the “iris” classification task. This object contains the famous iris dataset and some meta-information, such as the target variable.

task = tsk("iris")

We quickly split our data into a train and a test set:

test.idx = sample(seq_len(task$nrow), 30)
train.idx = setdiff(seq_len(task$nrow), test.idx)
# Set task to only use train indexes
task$row_roles$use = train.idx

A Pipeline (or Graph) contains multiple pipeline operators (“PipeOp”s), where each PipeOp transforms the data when it flows through it. For this use case, we require 3 transformations:

  • A PipeOp that scales the data
  • A PipeOp that performs PCA
  • A PipeOp that contains the Decision Tree learner

A list of available PipeOps can be obtained from

## <DictionaryPipeOp> with 64 stored values
## Keys: boxcox, branch, chunk, classbalancing, classifavg, classweights,
##   colapply, collapsefactors, colroles, copy, datefeatures, encode,
##   encodeimpact, encodelmer, featureunion, filter, fixfactors, histbin,
##   ica, imputeconstant, imputehist, imputelearner, imputemean,
##   imputemedian, imputemode, imputeoor, imputesample, kernelpca,
##   learner, learner_cv, missind, modelmatrix, multiplicityexply,
##   multiplicityimply, mutate, nmf, nop, ovrsplit, ovrunite, pca, proxy,
##   quantilebin, randomprojection, randomresponse, regravg,
##   removeconstants, renamecolumns, replicate, scale, scalemaxabs,
##   scalerange, select, smote, spatialsign, subsample, targetinvert,
##   targetmutate, targettrafoscalerange, textvectorizer, threshold,
##   tunethreshold, unbranch, vtreat, yeojohnson

First we define the required PipeOps:

op1 = po("scale")
op2 = po("pca")
op3 = po("learner", learner = lrn("classif.rpart")) A quick glance into a PipeOp

In order to get a better understanding of what the respective PipeOps do, we quickly look at one of them in detail:

The most important slots in a PipeOp are:

  • $train(): A function used to train the PipeOp.
  • $predict(): A function used to predict with the PipeOp.

The $train() and $predict() functions define the core functionality of our PipeOp. In many cases, in order to not leak information from the training set into the test set it is imperative to treat train and test data separately. For this we require a $train() function that learns the appropriate transformations from the training set and a $predict() function that applies the transformation on future data.

In the case of PipeOpPCA this means the following:

  • $train() learns a rotation matrix from its input and saves this matrix to an additional slot, $state. It returns the rotated input data stored in a new Task.
  • $predict() uses the rotation matrix stored in $state in order to rotate future, unseen data. It returns those in a new Task. Constructing the Pipeline

We can now connect the PipeOps constructed earlier to a Pipeline. We can do this using the %>>% operator.

linear_pipeline = op1 %>>% op2 %>>% op3

The result of this operation is a “Graph”. A Graph connects the input and output of each PipeOp to the following PipeOp. This allows us to specify linear processing pipelines. In this case, we connect the output of the scaling PipeOp to the input of the PCA PipeOp and the output of the PCA PipeOp to the input of PipeOpLearner.

We can now train the Graph using the iris Task.

## $classif.rpart.output

When we now train the graph, the data flows through the graph as follows:

  • The Task flows into the PipeOpScale. The PipeOp scales each column in the data contained in the Task and returns a new Task that contains the scaled data to its output.
  • The scaled Task flows into the PipeOpPCA. PCA transforms the data and returns a (possibly smaller) Task, that contains the transformed data.
  • This transformed data then flows into the learner, in our case classif.rpart. It is then used to train the learner, and as a result saves a model that can be used to predict new data.

In order to predict on new data, we need to save the relevant transformations our data went through while training. As a result, each PipeOp saves a state, where information required to appropriately transform future data is stored. In our case, this is mean and standard deviation of each column for PipeOpScale, the PCA rotation matrix for PipeOpPCA and the learned model for PipeOpLearner.

# predict on test.idx
task$row_roles$use = test.idx
## $classif.rpart.output
## <PredictionClassif> for 30 observations:
##     row_ids      truth   response
##          96 versicolor versicolor
##          66 versicolor versicolor
##         125  virginica  virginica
## ---                              
##         119  virginica  virginica
##         102  virginica  virginica
##           2     setosa     setosa

10.5.2 mlr3pipelines vs. mlr

In order to showcase the benefits of mlr3pipelines over mlr’s Wrapper mechanism, we compare the case of imputing missing values before filtering the top 2 features and then applying a learner.

While mlr wrappers are generally less verbose and require a little less code, this heavily inhibits flexibility. As an example, wrappers can generally not process data in parallel. mlr

# We first create a learner
lrn = makeLearner("classif.rpart")
# Wrap this learner in a FilterWrapper
lrn.wrp = makeFilterWrapper(lrn, fw.abs = 2L)
# And wrap the resulting wrapped learner into an ImputeWrapper.
lrn.wrp = makeImputeWrapper(lrn.wrp, classes = list(factor = imputeConstant("missing")))

# Afterwards, we can train the resulting learner on a task
train(lrn, iris.task) mlr3pipelines


impute = po("imputeoor")
filter = po("filter", filter = flt("variance"), filter.nfeat = 2L)
rpart = po("learner", lrn("classif.rpart"))

# Assemble the Pipeline
pipeline = impute %>>% filter %>>% rpart
# And convert to a 'GraphLearner'
learner = as_learner(pipeline)

The fact that mlr’s wrappers have to be applied inside-out, i.e. in the reverse order is often confusing. This is way more straight-forward in mlr3pipelines, where we simply chain the different methods using %>>%. Additionally, mlr3pipelines offers way greater possibilities with respect to the kinds of Pipelines that can be constructed. In mlr3pipelines, we allow for the construction of parallel and conditional pipelines. This was previously not possible.

10.5.3 mlr3pipelines vs. sklearn.pipeline.Pipeline

In order to broaden the horizon, we compare to Python sklearn’s Pipeline methods. sklearn.pipeline.Pipeline sequentially applies a list of transforms before fitting a final estimator. Intermediate steps of the pipeline are transforms, i.e. steps that can learn from the data, but also transform the data while it flows through it. The purpose of the pipeline is to assemble several steps that can be cross-validated together while setting different parameters. For this, it enables setting parameters of the various steps.

It is thus conceptually very similar to mlr3pipelines. Similarly to mlr3pipelines, we can tune over a full Pipeline using various tuning methods. Pipeline mainly supports linear pipelines. This means, that it can execute parallel steps, such as for example Bagging, but it does not support conditional execution, i.e. PipeOpBranch. At the same time, the different transforms in the pipeline can be cached, which makes tuning over the configuration space of a Pipeline more efficient, as executing some steps multiple times can be avoided.

We compare functionality available in both mlr3pipelines and sklearn.pipeline.Pipeline to give a comparison.

The following example obtained from the sklearn documentation showcases a Pipeline that first Selects a feature and performs PCA on the original data, concatenates the resulting datasets and applies a Support Vector Machine. sklearn

from sklearn.pipeline import Pipeline, FeatureUnion
from sklearn.model_selection import GridSearchCV
from sklearn.svm import SVC
from sklearn.datasets import load_iris
from sklearn.decomposition import PCA
from sklearn.feature_selection import SelectKBest

iris = load_iris()

X, y = iris.data, iris.target

# This dataset is way too high-dimensional. Better do PCA:
pca = PCA(n_components=2)

# Maybe some original features where good, too?
selection = SelectKBest(k=1)

# Build estimator from PCA and Univariate selection:
combined_features = FeatureUnion([("pca", pca), ("univ_select", selection)])

# Use combined features to transform dataset:
X_features = combined_features.fit(X, y).transform(X)

svm = SVC(kernel="linear")

# Do grid search over k, n_components and C:
pipeline = Pipeline([("features", combined_features), ("svm", svm)])

param_grid = dict(features__pca__n_components=[1, 2, 3],
                  features__univ_select__k=[1, 2],
                  svm__C=[0.1, 1, 10])

grid_search = GridSearchCV(pipeline, param_grid=param_grid, cv=5, verbose=10)
grid_search.fit(X, y) mlr3pipelines

iris = tsk("iris")

# Build the steps
copy = po("copy", 2)
pca = po("pca")
selection = po("filter", filter = flt("variance"))
union = po("featureunion", 2)
svm = po("learner", lrn("classif.svm", kernel = "linear", type = "C-classification"))

# Assemble the Pipeline
pipeline = copy %>>% gunion(list(pca, selection)) %>>% union %>>% svm
learner = as_learner(pipeline)

# For tuning, we define the resampling and the Parameter Space
resampling = rsmp("cv", folds = 5L)

search_space = ps(
  classif.svm.cost = p_dbl(lower = 0.1, upper = 1),
  pca.rank. = p_int(lower = 1, upper = 3),
  variance.filter.nfeat = p_int(lower = 1, upper = 2)

instance = TuningInstanceSingleCrit$new(
  task = iris,
  learner = learner,
  resampling = resampling,
  measure = msr("classif.ce"),
  terminator = trm("none"),
  search_space = search_space

tuner = tnr("grid_search", resolution = 10)

Set the learner to the optimal values and train
learner$param_set$values = instance$result_learner_param_vals

In summary, we can achieve similar results with a comparable number of lines, while at the same time offering greater flexibility with respect to which kinds of pipelines we want to optimize over. At the same time, experiments using mlr3 can now be arbitrarily parallelized using futures.

10.5.4 mlr3pipelines vs recipes

recipes is a new package, that covers some of the same applications steps as mlr3pipelines. Both packages feature the possibility to connect different pre- and post-processing methods using a pipe-operator. As the recipes package tightly integrates with the tidymodels ecosystem, much of the functionality integrated there can be used in recipes. We compare recipes to mlr3pipelines using an example from the recipes vignette.

The aim of the analysis is to predict whether customers pay back their loans given some information on the customers. In order to do this, we build a model that does the following:

  1. It first imputes missing values using k-nearest neighbors
  2. All factor variables are converted to numerics using dummy encoding
  3. The data is first centered then scaled.

In order to validate the algorithm, data is first split into a train and test set using initial_split, training, testing. The recipe trained on the train data (see steps above) is then applied to the test data. recipes

data("credit_data", package = "modeldata")

train_test_split = initial_split(credit_data)
credit_train = training(train_test_split)
credit_test = testing(train_test_split)

rec = recipe(Status ~ ., data = credit_train) %>%
  step_knnimpute(all_predictors()) %>%
  step_dummy(all_predictors(), -all_numeric()) %>%
  step_center(all_numeric()) %>%

trained_rec = prep(rec, training = credit_train)

# Apply to train and test set
train_data <- bake(trained_rec, new_data = credit_train)
test_data  <- bake(trained_rec, new_data = credit_test)

Afterwards, the transformed data can be used during train and predict:

# Train
rf = rand_forest(mtry = 12, trees = 200, mode = "classification") %>%
  set_engine("ranger", importance = 'impurity') %>%
  fit(Status ~ ., data = train_data)

# Predict
prds = predict(rf, test_data) mlr3pipelines

The same analysis can be performed in mlr3pipelines. Note, that for now we do not impute via knn but instead via sampling.

data("credit_data", package = "modeldata")

# Create the task
task = as_task_classif(credit_data, target = "Status")

# Build up the Pipeline:
g = po("imputesample", id = "impute") %>>%
  po("encode", method = "one-hot") %>>%
  po("scale") %>>%
  po("learner", lrn("classif.ranger", num.trees = 200, mtry = 12))

# We can visualize what happens to the data using the `plot` function:

# And we can use `mlr3's` full functionality be wrapping the Graph into a GraphLearner.
glrn = as_learner(g)
resample(task, glrn, rsmp("holdout"))