v0.21.6

MLJBase v0.21.6

Diff since v0.21.5

Closed issues:

• MethodError: fit(::StaticPipeline{NamedTuple... (#881)

Merged pull requests:

• Fix suspicious test that is failing on CI (#883) (@ablaom)

v0.21.5

MLJBase v0.21.5

Diff since v0.21.4

Closed issues:

• Typo in serializable docstring (#850)
• Improve error message for Stack (#861)
• Ensure private signatures of glb are marked private (#869)

Merged pull requests:

• Add check for base models in Stack (#877) (@ablaom)

v0.21.4

MLJBase v0.21.4

Diff since v0.21.3

Closed issues:

• Problems with display of PerformanceEvaluation object when using custom measures (functions) (#874)

Merged pull requests:

• fix bug in auc computation (#878) (@OkonSamuel)

v0.21.3

MLJBase v0.21.3

Diff since v0.21.2

Closed issues:

• What's wrong with the way we export learning networks as new model types (#831)
• serializing learning network machines (#844)
• export restore! (#851)
• Add transformer as alias for keyword target in TransformedTargetModel (#855)
• Confused about output of predict(MLJBase.Pipeline, fitresult, Xnew) (#870)
• evaluate errors displaying when using confusion_matrix as measure (#871)

Merged pull requests:

• Fixes to evaluate standard error calculation to fix display issue (#872) (@ablaom)
• Export restore! (#873) (@ablaom)
• For a 0.21.3 release (#875) (@ablaom)

v0.21.2

MLJBase v0.21.2

Diff since v0.21.1

Closed issues:

• Towards a 0.21 release (#852)
• predict on Stack returns a Tuple (#867)

Merged pull requests:

• For a 0.21.2 release (#868) (@ablaom)

v0.21.1

MLJBase v0.21.1

Diff since v0.21.0

• (bug fix) #865 @OkonSamuel

Closed issues:

• Latest release breaks MCMCDiagnosticTools (#863)
• Error when calling fallback operations for probabilistic networkcomposites (#864)

Merged pull requests:

• fix bug in fallback operations for probabilisticnetworkcomposites (#865) (@OkonSamuel)
• For a 0.21.1 release (#866) (@ablaom)

v0.21.0

MLJBase v0.21.0

Diff since v0.20.20

• (enhancement) Introduce new protocol for exporting learning networks via the new family of supertypes *NetworkComposite (#841, #831)
• (mildly breaking) Remove key-value pairs from reports and fitted parameters that have nothing or NamedTuple() as a value.
• (breaking) Re-implement Stack, TransformedTargetModel and pipelines (Pipeline) as NetworkComposite types. What is breaking is that some reports and fitted parameters will be missing some these keys: machines, fitted_params_given_machine, report_given_machine. However, the only information contained in those items that is not rendered redundant by the existing items are considered private.
• (breaking) Remove the formerly deprecated @pipeline macro, now rendered redundant by Pipeline constructor (or model1 |> model2 |> ... syntax) and TransformedTargetModel.

Merged pull requests:

• Add new option for exporting learning networks as stand-alone composite model types (#841) (@ablaom)
• For a 0.21 release (PR onto dev) (#853) (@ablaom)
• Re-implement Stack as a NetworkComposite model (#854) (@ablaom)
• Re-implement TransformedTargetModel as NetworkComposite model (#857) (@ablaom)
• Re-implement pipelines as NetworkComposite models (#858) (@ablaom)
• Remove previously deprecated @pipeline macro (#859) (@ablaom)
• Deprecate old ways of exporting learning networks (#860) (@ablaom)
• For a 0.21 release (#862) (@ablaom)

v0.20.20

MLJBase v0.20.20

Diff since v0.20.19

Closed issues:

• Issue in serializable (#843)
• allow train_test_pairs to work on generic vector y? (#847)

Merged pull requests:

• [Patch for older 0.19 series] Fix scitype checks for Annotator model types (such as detectors) (#842) (@ablaom)
• Relax type restriction on y in train_test_pairs(::StratifiedCV, rows, y) (#848) (@ablaom)
• For a 0.20.20 release (#849) (@ablaom)

v0.19.9

MLJBase v0.19.9 (added after 0.20.0 release)

Diff since v0.19.8

Closed issues:

• evaluate and evaluate! should return a struct instead of a NamedTuple (#408)
• Handling of NaN micro/macro aggregation (#630)
• Relax any checks that block transformers needing to see target in training. (#699)
• For a 0.20 release (#725)
• How to transform any deterministic regression model to a probabilistic binary classification model, and then evaluate it with k-fold cross validation (#760)
• Clean up: Remove data anonymization from from composite model apparatus (#764)
• unpack error (#777)
• Add Multiple targets to make_regression (#779)
• Add support in Stack for models that handle missing values (#781)
• typename method for AbstractDataFrame incorrect equality (#784)
• Typos in Stack constructor lead to uninformative error messages (#796)
• Incorrect display of one-dimensional range in case of scaling function (#797)
• Export serializable (#807)
• Improve warning of machine constructor (#817)

Merged pull requests:

• Serialization (#733) (@olivierlabayle)
• Add suggestion to err_incompatible_prediction_types message (#748) (@ablaom)
• Add a brief reminder that the fields of the PerformanceEvaluation struct are part of the public API (#749) (@DilumAluthge)
• For a 0.20 release (#751) (@ablaom)
• Fix a problem with confmat and CategoricalValue eltype (#752) (@ablaom)
• Allow user to control the level of scitype checks in machine constructor (#753) (@ablaom)
• For a 0.20 release (#754) (@ablaom)
• Bump version. (#755) (@ablaom)
• Improve short version of show for measures (#757) (@ablaom)
• Fix wrong field name in a show method (#758) (@rikhuijzer)
• For a 0.20.1 release (#761) (@ablaom)
• For a 0.20.2 release (#763) (@ablaom)
• Add std to show for PerformanceEvaluation (#766) (@rikhuijzer)
• Make running tests via TestEnv easier (#769) (@rikhuijzer)
• For a 0.20.3 release (#770) (@ablaom)
• Fix typo (#771) (@KronosTheLate)
• Bump LossFunctions.jl (#773) (@juliohm)
• bump 0.20.4 (#774) (@OkonSamuel)
• For 0.20.4 release (#776) (@OkonSamuel)
• Stack cache and acceleration (rebased) (#785) (@ablaom)
• closes #784 by fixing typename + tests (#786) (@tlienart)
• For a 0.20.5 release (#787) (@ablaom)
• For a 0.20.6 release (#788) (@ablaom)
• Skip NaN in averaging in MulticlassFScore (#792) (@ablaom)
• Fix bugs and inconsistencies in documentation on evaluate!s resampling keyword (#793) (@wolthom)
• For a 0.20.7 release (#794) (@ablaom)
• Remove apparatus to anonymize composite model fitresults (#795) (@ablaom)
• Add feature_importances machine inspection method. (#798) (@OkonSamuel)
• For a 0.20.8 release (#799) (@ablaom)
• Bump compat MLJModelInterface = "1.5" (#800) (@ablaom)
• For a 0.20.9 release (#801) (@ablaom)
• Fix inverse_transform bug for pipelines with Static components (#802) (@ablaom)
• Fix show for one-dimensional ranges (#803) (@ablaom)
• For a 0.20.10 release (#804) (@ablaom)
• For a 0.20.11 release (#805) (@ablaom)
• Proposal to add support to allow non-generalizing models to contribute to a machine's report (#806) (@ablaom)
• For a 0.20.12 release (#808) (@ablaom)
• Fix bug around training losses for pipelines (#809) (@ablaom)
• For a 0.20.13 release (#810) (@ablaom)
• Add multi-target feature to make_regression (#811) (@ablaom)
• For a 0.20.14 release (#812) (@ablaom)
• Overload all operations (predict, predict_mode, etc) for use on Static machines (#813) (@ablaom)
• For a 0.20.15 release (#814) (@ablaom)
• Fix typo probab i listic (#815) (@svilupp)
• Get rid of long lines in stacking.jl (#816) (@ablaom)
• Adapt stack to allow missings in a basemodel target_scitype (#818) (@ablaom)
• Add check_ismodel method and use it in Stack to improve error messages (#819) (@ablaom)
• Improve warning of machine constructor (#820) (@ablaom)
• For a 0.20.16 release (#821) (@ablaom)
• Test suite: modify some uses of deepcopy (#823) (@DilumAluthge)
• Adapt some tests for Julia nightly. (#824) (@ablaom)
• Fix a bug with feature_importances(::Machine) (#825) (@ablaom)
• For a 0.20.17 release (#826) (@ablaom)
• Tweak display of machines (#828) (@ablaom)
• Remove redundant code (#830) (@ablaom)
• remove unbound type parameters (#836) (@nsajko)
• For 0.20.18 release (#838) (@ablaom)
• Fix corner case bugs for operations on machines (#839) (@ablaom)
• For a 0.20.19 release (#840) (@ablaom)
• [Patch for older 0.19 series] Fix scitype checks for Annotator model types (such as detectors) (#842) (@ablaom)

v0.20.19

MLJBase v0.20.19

Diff since v0.20.18

Merged pull requests:

• Fix corner case bugs for operations on machines (#839) (@ablaom)
• For a 0.20.19 release (#840) (@ablaom)