v1.12.1

MLJBase v1.12.1

Diff since v1.12.0

Merged pull requests:

• Re-organize code: Break up resampling.jl (#1039) (@ablaom)
• Remove a name conflict for int and decoder (#1040) (@ablaom)
• Stratify shuffle fix (#1041) (@tlienart)
• For 1.12.1 release (#1042) (@ablaom)

v1.12.0

MLJBase v1.12.0

Diff since v1.11.0

• (enhancement) Arrange that the objects returned by evaluate/evaluate! store the currently displayed "1.96SE" estimates of the radius of the confidence interval for each performance estimate, to be accessible via a new property uncertainty_radius_95 (#1034)
• (enhancement) Allow users to specify a tag for a model or machine they are evaluating, as in evaluate("my_model"=>model, ...) in place of evaluate(model, ...), to be accessible in the returned object via a new property, tag. Include the tag in the display of the returned object and vectors of the same (#1034)
• (enhancement) Allow users to pass a vector of (possibly tagged) models or machines in evaluate/evaluate! calls, as in evaluate([model1, model2], ...) or evaluate(["knn with K=5"=>model1, "random forest"=>model2], ...) (#1034).

Merged pull requests:

• Enhance performance evaluations (#1034) (@ablaom)
• Fix show for CompactPerformanceEvaluation (#1036) (@ablaom)
• Disambiguate internal conflict: MMI.classes and CategoricalDistributions.classes. (#1037) (@ablaom)
• For a 1.12 release (#1038) (@ablaom)

Closed issues:

• Typo in error messages (#1025)
• Allow evaluate to have multiple models as argument (#1031)

v1.11.0

MLJBase v1.11.0

Diff since v1.10.1

• Add evaluate/evaluate! support for density estimators (#1026)

Merged pull requests:

• Add evaluate support for density estimators (#1026) (@ablaom)
• For a 1.10 release (#1035) (@ablaom)

v1.10.1

MLJBase v1.10.1

Diff since v1.10.0

Merged pull requests:

• Bump compat, PrettyTables="3" (#1032) (@ablaom)
• For a 1.10.1 release (#1033) (@ablaom)

v1.10.0

MLJBase v1.10.0

Diff since v1.9.2

• Add support for CategoricalArrays 1.0 and CategoricalDistributions 0.2. This changes the behaviour of the re-exported methods levels and unique, as applied to CategoricalArray and UnivariateFinite. Whereas previously these returned a vector of "raw" values, they now return a CategoricalVector. See these CategoricalArrays.jl release notes.

Merged pull requests:

• Bump the compat requirements for CategoricalArrays and CategoricalDistributions (#1028) (@ablaom)
• For a 1.10 release (#1030) (@ablaom)

v1.9.2

MLJBase v1.9.2

Diff since v1.9.1

Merged pull requests:

• Fix show for the Node type (#1023) (@ablaom)
• For a 1.9.2 release (#1027) (@ablaom)

Closed issues:

• show for nodes broken in Julia 1.12 (#1022)

v1.9.1

MLJBase v1.9.1

Diff since v1.9.0

Merged pull requests:

• Remove type piracy for show(::NamedTuple) (#1019) (@ablaom)
• For a 1.9.1 release (#1020) (@ablaom)

Closed issues:

• Stop pirating show for NamedTuples (#1018)

v1.9.0

MLJBase v1.9.0

Diff since v1.8.2

For the release notes:

• Allow tuples, in addition to vectors, when specifying resampling=... in evaluate/evaluate!, and fix a mistake in the error message for invalid resampling values (#1010)

Merged pull requests:

• Allow tuples in the resampling option for evaluate (#1010) (@ablaom)
• For a 1.9 release (#1011) (@ablaom)

v1.8.2

MLJBase v1.8.2

Diff since v1.8.1

Merged pull requests:

• Ensure source() != source() and clean up some tests (#1003) (@ablaom)
• For a 1.8.2 release (#1004) (@ablaom)

v1.8.1

MLJBase v1.8.1

Diff since v1.8.0

Merged pull requests:

• Fix deprecated Vararg expression (#1000) (@devmotion)
• For a 1.8.1 release (#1001) (@ablaom)