Release v0.15.0 · JuliaAI/MLJBase.jl

MLJBase v0.15.0

(enhancement) Add fitted_params_per_fold and report_per_fold properties to the object returned by evaluate/evaluate! to give user access to the outcomes of training for each train/test pair in resampling (#400, JuliaAI/MLJ.jl#616)
(enhancement) Implement logpdf for UnivariateFinite distributions (#411)
(code organization) Remove ScientificTypes as explicit dependency (#403)
(bug fix) Fix bug related to creating new composite models by hand in special case of non-model hyper-parameters (not an issue with @pipeline or @from_network models). Introduce new return! syntax for doing this and deprecate calling of learning network machines ( #390, #391, #377)
(breaking) Change the behavior of evaluate/evaluate! so that weights are only passed to measures if explicitly passed using the key-word argument weights=... (#405)

Closed issues:

possible test failure (again!) in upcoming Julia version 1.5 (#286)
Prohibit distinct fields in composite models pointing to two models that are === (#377)
Update logic flawed in case of composite models with non-model fields (#390)
Towards a 0.15 release (#404)
Decouple interface points for training weights and weights used in measures. (#405)
Release that doesn't have ScientificTypes as a hard dep? (#412)

Merged pull requests:

Fix update logic for Composite models (#391) (@ablaom)
Add report_per_fold and machine_per_fold to results of evaluate!() (#400) (@ablaom)
Remove ScientificTypes as explicit dependency and as a test dependency (#403) (@ablaom)
CompatHelper: open PRs against dev (#407) (@DilumAluthge)
Add logpdf method for UnivariateFinite (#411) (@cscherrer)
Decouple training weights from evaluation weights (#413) (@ablaom)
For a 0.15 release (#417) (@ablaom)