NEWS.md

In mlr3pipelines: Preprocessing Operators and Pipelines for 'mlr3'

mlr3pipelines 0.9.0

• Breaking change: Removed initialization of PipeOpImputeConstant's constant hyperparameter since it was incompatible with other defaults and would lead to not recommended usage (creating an empty level).
• Removed compatibility for old paradox versions pre-1.0.0.
• Added empty_level_control argument to PipeOpImpute allowing control over edge cases for factor/ordered columns.
• Set new construction argument empty_level_control to "param" for PipeOpImputeOOR and to "always" for PipeOpImputeConstant.
• Untrained PipeOps that take NULL as input during training now automatically perform training during prediction.
• PipeOpImputeConstant, PipeOpImputeMode, PipeOpImputeOOR, and PipeOpImputeLearner can now handle factor or ordered features with zero levels.
• PipeOpImputeConstant now gives a more informative error message if check_levels is TRUE and a new level would be created through imputation.
• Fix: PipeOpImputeOOR now imputes ".MISSING" for factor/ordered features with only NAs instead of sampling from the feature's levels.
• Fix: PipeOpImputeLearner no longer adds "factor" or "ordered" levels for these feature types arbitrarily and instead updates levels correctly in certain edge-cases.
• Fixed the error message for unexpected Multiplicities in the input and output type checking during PipeOps training and prediction.
• Fixed a grammatical error in PipeOp's error message wrapper: now correctly says "This happened in ...".

mlr3pipelines 0.8.0

• Added missing error for predicting with untrained PipeOps / Graphs.
• Fix: Corrected typo in the hyperparameter name use_parallel of PipeOpVtreat.
• Fix: Do not overwrite initial hyperparameter settings of bbotk::OptimizerBatchNLoptr in LearnerClassifAvg / LearnerRegrAvg's internal optimize_weights_learneravg function.
• Added new convenience function preproc() for easier training of or prediction with PipeOps or Graphs.
• Fix: PipeOpVtreat, PipeOpEncodeImpact, and PipeOpEncodeLmer now accept the more precise TaskSupervised instead of Task as input for training and prediction.
• Docs: Added missing documentation for the task_type of the input and output channels of PipeOps that inherit from PipeOpTaskPreproc and set a non-default task_type.
• Fix: PipeOpEncodeLmer, PipeOpADAS, PipeOpBLSmote, PipeOpSmote, and PipeOpSmoteNC no longer throw an error in case of empty target levels during training.
• Fix: PipeOpClassBalancing now handles unseen target levels by ignoring them during upsampling instead of producing NAs.

mlr3pipelines 0.7.2

• New parameter no_collapse_above_absolute for PipeOpCollapseFactors / po("collapse_factors").
• Fix: PipeOpCollapseFactors now correctly collapses levels of ordered factors.
• Fix: LearnerClassifAvg and LearnerRegrAvg hyperparameters get the "required" tag.
• New parameter use_groups (default TRUE) for PipeOpSubsampling to respect grouping (changed default behaviour for grouped data)
• New parameter new_role_direct for PipeOpColRoles / po("colroles") to change column roles by role instead of by column.
• Dictionary sugar functions po() / pos() / ppl() / ppls() now make suggestions for entries in both mlr_pipeops as well as mlr_graphs when an object by the given name could not be found in the respective dictionary.
• New PipeOp PipeOpDecode / po("decode") to reverse one-hot or treatment encoding.
• Fix: Columns that are feature and something else no longer lose the other column role during training or predicting of PipeOps inheriting from PipeOpTaskPreproc.
• Fix: Made tests for PipeOpBLSmote deterministic.
• Fix: Corrected hash calculation for PipeOpFilter.
• New PipeOps PipeOpEncodePLQuantiles and PipeOpEncodePLTree that implement piecewise linear encoding with two different binning methods.
• Compatibility with new R6 release.
• Docs: Performed cleanup and standardization.
• Docs: Performed cleanup of reference index page on website.
• Docs: Fixed parsing of examples on website for PipeOpNMF and PipeOpLearnerPICVPlus.
• Fix: PipeOpTargetMutate and PipeOpTargetTrafoScaleRange no longer drop unseen factor levels of features or targets during train and predict.
• Simplified parameter checks and added internal type checking for PipeOpTargetMutate.

mlr3pipelines 0.7.1

• Compatibility fix for upcoming mlr3
• New down-sampling PipeOps for inbalanced data: PipeOpTomek / po("tomek") and PipeOpNearmiss / po("nearmiss")
• New PipeOp PipeOpLearnerPICVPlus / po("learner_pi_cvplus")
• New PipeOp for Quantile Regression PipeOpLearnerQuantiles / po(learner_quantiles)
• GraphLearner has new active bindings/methods as shortcuts for active bindings/methods of the underlying Graph: $pipeops, $edges, $pipeops_param_set, and $pipeops_param_set_values as well as $ids() and $plot().

mlr3pipelines 0.7.0

• New PipeOp PipeOpRowApply / po("rowapply")
• Empty PipeOp IDs now explicitly forbidden.
• Bugfix: Graph$tran() / Graph$predict() with single_input = FALSE now correctly handles PipeOps with multiple inputs.
• GraphLearner$base_learner() now works with PipeOpBranch, and is generally more robust.
• GraphLearner now supports $importance, $selected_features(), $oob_error(), and $loglik(). These are computed from the underlying Learner.
• GraphLearner$impute_selected_features option added: $selected_features() is reported even if the underlying base learner does not report it; in this case, the full feature set as seen by that learner is returned.
• GraphLearner$predict_type handling more robust now.
• PipeOpThreshold and PipeOpTuneThreshold now have the $predict_type "prob". They can be set to "response", in which case the probability predictions are discarded, potentially saving memory.
• Bugfix for handling multiplicities in PipeOps with vararg channels.
• Bugfix: PipeOpImputeOOR now retains the .MISSING level in factors during prediction that were imputed during training, but had no missing values during prediction.
• as_data_table(po()) now works even when some PipeOps can not be constructed. For these PipeOps, NA is reported in most columns.
• Compatibility with upcoming mlr3 release.
• New PipeOps for handling inbalanced data: PipeOpADAS / po("adas"), PipeOpBLSmote / po("blsmote") and PipeOpSmoteNC / po("smotenc")

mlr3pipelines 0.6.0

• Compatibility with new bbotk release.
• Added marshaling support to GraphLearner
• Support internal tuning and validation

mlr3pipelines 0.5.2

• Added new ppl("convert_types").
• Minor documentation fixes.
• Test helpers are now available in inst/. These are considered experimental and unstable.

mlr3pipelines 0.5.1

• Changed the ID of PipeOpFeatureUnion used in ppl("robustify") and ppl("stacking").
• pipeline_bagging() gets the replace argument (old behaviour FALSE by default).
• Feature: The $add_pipeop() method got an argument clone (old behaviour TRUE by default).
• Bugfix: PipeOpFeatureUnion in some rare cases dropped variables called "x".
• Compatibility with upcoming paradox release.

mlr3pipelines 0.5.0-2

• Avoid unnecessarily large serializations of ppl("robustify") pipelines.
• Made tests and examples compatible with mlr3 update.

mlr3pipelines 0.5.0-1

• Bugfix: PipeOpTuneThreshold was not overloading the correct .train and .predict functions.

mlr3pipelines 0.5.0

• New way of computing $hash and $phash for GraphLearner and all PipeOps. This could break users that inherit from PipeOp and make use of $hash in the future (but is ultimately in their interest!).
• Neater plots.
• Bugfix: phash of GraphLearner now considers content of Graph, not only IDs.
• One vignette removed for version 0.1.3 added back here. Welcome home!
• Bugfix: Make Graph work that have PipeOps with more than one output, where one output was linked to multiple inputs.

mlr3pipelines 0.4.3

• po(), pos() can now construct PipeOps with ID postfix _<number> to avoid ID clashes.
• GraphLearner now has method $base_learner() that returns the underlying Learner, if it can be found by a simple heuristic.
• Fix S3 function signatures

mlr3pipelines 0.4.2

• Documentation: Clarified PipeOpHistBin operation.
• Documentation: Fixed PipeOpPCA documentation of center default.
• Added $label active binding, setting it to the help()-page title by default.
• Made tests compatible with upcoming mlr3misc update.

mlr3pipelines 0.4.1

• $help() function for all PipeOps as well as Graph, GraphLearner and all Learners.
• GraphLearner can be created without cloning Graph (for internal use).
• predict.Graph throws helpful error when it cannot create a fitting Task.
• PipeOpLearner packages slot is set to the Learner's packages.
• Bugfix: PipeOp train() and predict() report correct channel name when output has wrong type.
• Bugfix: More accurate type inference when constructing Graphs.
• Stability fix for interaction with packages such as mlr3spatiotempcv that extend existing Task types.

mlr3pipelines 0.4.0

• New operator %>>!% that modifies Graphs in-place.
• New methods chain_graphs(), concat_graphs(), Graph$chain() as alternatives for %>>% and %>>!%.
• New methods pos() and ppls() which create lists of PipeOps/Graphs and can be seen as "plural" forms of po() and ppl().
• po() S3-method for PipeOp class that clones a PipeOp object and optionally modifies its attributes.
• Graph$add_pipeop() now clones the PipeOp being added.
• Documentation: Clarified documentation about cloning of input arguments in several places.
• Performance enhancements for Graph concatenation.
• More informative error outputs.
• New attribute graph_model in GraphLearner class, which gets the trained Graph.
• as_learner() S3-method for PipeOp class that wraps a PipeOp in a Graph and turns that into a Learner.
• Changed PipeOps:
• PipeOpHistBin: renamed bins Param to breaks
• PipeOpImputeHist: fix handling of integer features spanning the entire represented integer range
• PipeOpImputeOOR: fix handling of integer features spanning the entire represented integer range
• PipeOpProxy: Avoid unnecessary clone
• PipeOpScale: Performance improvement

mlr3pipelines 0.3.6-1

• Fix numerics problem in tests

mlr3pipelines 0.3.6

• Bugfix: Make empty Multiplicities work (unless they are nested)
• Fixed: Compatibility with upcoming bbotk version.
• New mlr_graphs: pipeline_stacking
• Added JMLR-Citation

mlr3pipelines 0.3.5-1

• Fixed: Compatibility with upcoming mlr3 version.

mlr3pipelines 0.3.5

• Changed PipeOp: PipeOpFilter gets additional filter.permuted hyperparameter.
• Bugfix: Make add_edge of Graphs work with Multiplicities.
• Bugfix: Make GraphLearner hash depend on id.
• Documentation: Clarify documentation of LearnerAvg.
• Internals: Using more idiomatic internal helper functions.
• Compatibility with upcoming mlr3 version.

mlr3pipelines 0.3.4

• Stability: PipeOps don't crash when they have python/reticulate hyperparameter values.
• Documentation: Titles of PipeOp documentation articles reworked.

mlr3pipelines 0.3.3

• Bugfix: fix rare issue in randomized test
• Compatibility with bbotk 0.3.0

mlr3pipelines 0.3.2

• Bugfix: Make as.data.table(mlr_pipeops) work with paradox 0.6
• Changed PipeOps:
• PipeOpColApply: now allows for an applicator function with multiple columns as a return value; also inherits from PipeOpTaskPreprocSimple now

mlr3pipelines 0.3.1

• Changed PipeOps:
• PipeOpMissInd now also allows for setting type = integer
• PipeOpNMF: now exposes all parameters previously in .options
• Changed mlr_graphs:
• pipeline_bagging now uses multiplicities internally
• fix how pipeline_robustify determines the type of newly created columns when using PipeOpMissInd
• PipeOpFeatureUnion: Fixed a minor bug when checking for duplicates
• added an autotest for ParamSets of PipeOps: expect_valid_pipeop_param_set
• More informative error message when PipeOp input value has wrong type
• Fix automatic detection of R6 type hierarchy
• Performance improvements for GraphLearner
• GraphLearner allows custom id
• Use parallel tests
• Removed bibtex dependency

mlr3pipelines 0.3.0

• compatibility with mlr3 0.6
• NULL input channels accept any kind of input
• print() method of Graphs now also allows for printing a DOT representation on the console
• state of PipeOps is now reset to NULL when training fails
• implemented as_learner.PipeOp
• LearnerClassifAvg, LearnerRegrAvg use bbotk now
• Changed PPLs:
• fix how ppl_robustify detects whether a learner can handle factors
• Changed PipeOps:
• PipeOpTextVectorizer can now return an "integer sequence representation".
• New PipeOps:
• PipeOpNMF
• PipeOpColRoles
• PipeOpVtreat
• various bugfixes

mlr3pipelines 0.2.1

• New feature: Multiplicities: implicit repetition of operations
• new mlr_graphs:
• pipeline_bagging
• pipeline_branch
• pipeline_greplicate
• pipeline_robustify
• pipeline_targettrafo
• pipeline_ovr
• New PipeOps:
• PipeOpOVRSplit, PipeOpOVRUnite
• PipeOpReplicate
• PipeOpMultiplicityExply, PipeOpMultiplicityImply
• PipeOpTargetTrafo, PipeOpTargetInvert
• PipeOpTargetMutate
• PipeOpTargetTrafoScaleRange
• PipeOpProxy
• PipeOpDateFeatures
• PipeOpImputeConstant
• PipeOpImputeLearner
• PipeOpMode
• PipeOpRandomResponse
• PipeOpRenameColumns
• PipeOpTextVectorizer
• PipeOpThreshold
• Renamed PipeOps:
• PipeOpImputeNewlvl --> PipeOpImputeOOR (with additional functionality for continuous values)
• Changed PipeOps:
• PipeOpFeatureUnion: Bugfix: avoid silently overwriting features when names clash
• PipeOpHistBin: Bugfix: handle test set data out of training set range
• PipeOpLearnerCV: Allow returning trainingset prediction during train()
• PipeOpMutate: Allow referencing newly created columns
• PipeOpScale: Allow robust scaling
• PipeOpLearner, PipeOpLearnerCV: learner_models for access to learner with model slot
• New Selectors:
• selector_missing
• selector_cardinality_greater_than
• NULL is neutral element of %>>%
• PipeOpTaskPreproc now has feature_types slot
• PipeOpTaskPreproc(Simple) internal API changed: use .train_task(), .predict_task(), .train_dt(), .predict_dt(), .select_cols(), .get_state(), .transform(), .get_state_dt(), .transform_dt() instead of the old methods without dot prefix
• PipeOp now has tags slot
• PipeOp internal API changed: use .train(), .predict() instead of train_internal(), predict_internal()
• Graph new method update_ids()
• Graph methods train(single_input = FALSE) and predict(single_input = FALSE) now handle vararg channels correctly.
• Obsoleted greplicate(); use pipeline_greplicate / ppl("greplicate") instead.
• po() now automatically converts Selector to PipeOpSelect
• po() prints available mlr_pipeops dictionary content
• mlr_graphs dictionary of useful Graphs, with short form accessor ppl()
• Work with new mlr3 version 0.4.0

mlr3pipelines 0.1.3

• small test fix for R 4.0 (necessary for stringsAsFactors option default change in 3.6 -> 4.0)
• predict() generic for Graph
• Migrated last vignette to "mlr3 Book"
• Compact in-memory representation of R6 objects to save space when saving objects via saveRDS(), serialize() etc.

mlr3pipelines 0.1.2

• Work with new mlr3 version 0.1.5 (handling of character columns changed)

mlr3pipelines 0.1.1

• Better html graphics for linear Graphs
• New PipeOps:
• PipeOpEncodeImpact
• Changed PipeOp Behaviour:
• PipeOpEncode: handle NAs

mlr3pipelines 0.1.0

• Initial upload to CRAN.

mlr3pipelines documentation built on Aug. 8, 2025, 7:19 p.m.