v0.20.18

MLJBase v0.20.18

Diff since v0.20.17

Merged pull requests:

• Tweak display of machines (#828) (@ablaom)
• Remove redundant code (#830) (@ablaom)
• remove unbound type parameters (#836) (@nsajko)
• For 0.20.18 release (#838) (@ablaom)

v0.20.17

MLJBase v0.20.17

Diff since v0.20.16

Merged pull requests:

• 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)

v0.20.16

MLJBase v0.20.16

Diff since v0.20.15

Closed issues:

• evaluate and evaluate! should return a struct instead of a NamedTuple (#408)
• Clean up: Remove data anonymization from from composite model apparatus (#764)
• Add Multiple targets to make_regression (#779)
• Add support in Stack for models that handle missing values (#781)
• Typos in Stack constructor lead to uninformative error messages (#796)
• Improve warning of machine constructor (#817)

Merged pull requests:

• 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)

v0.20.15

MLJBase v0.20.15

Diff since v0.20.14

• Support Static models that have a predict or other operation implemented (in addition to transform already supported) #813

Merged pull requests:

• Overload all operations (predict, predict_mode, etc) for use on Static machines (#813) (@ablaom)
• For a 0.20.15 release (#814) (@ablaom)

v0.20.14

MLJBase v0.20.14

Diff since v0.20.13

• (enhancement) Add option in make_regression to generate multi-target data, as in make_regression(100, 3, n_targets=2) (#811) @josephsdavid

Merged pull requests:

• Add multi-target feature to make_regression (#811) (@ablaom)
• For a 0.20.14 release (#812) (@ablaom)

v0.20.13

MLJBase v0.20.13

Diff since v0.20.12

Closed issues:

• Export serializable (#807)

Merged pull requests:

• Fix bug around training losses for pipelines (#809) (@ablaom)
• For a 0.20.13 release (#810) (@ablaom)

v0.20.12

MLJBase v0.20.12

Diff since v0.20.11

• (enhancement) Add support for model operations (such as transform) that return extra information for merging into machine reports (#806)

Merged pull requests:

• 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)

v0.20.11

MLJBase v0.20.11

Diff since v0.20.10

Merged pull requests:

• For a 0.20.11 release (#805) (@ablaom)

v0.20.10

MLJBase v0.20.10

Diff since v0.20.9

Closed issues:

• Incorrect display of one-dimensional range in case of scaling function (#797)

Merged pull requests:

• 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)

v0.20.9

MLJBase v0.20.9

Diff since v0.20.8

Merged pull requests:

• Bump compat MLJModelInterface = "1.5" (#800) (@ablaom)
• For a 0.20.9 release (#801) (@ablaom)