v0.13.5

MLJBase v0.13.5

Diff since v0.13.4

• (bug fix) Fix issue with partition and stratified=... specifying a categorical vector (#274) (@ablaom)
• (code cleanup) (#279) (@aviatesk)
• Use cld instead of floor when spawning threads in resampling (#289) (@OkonSamuel)

v0.13.4

MLJBase v0.13.4

Diff since v0.13.3

#280, #281

Merged pull requests:

• implement schema interface (#280) (@aviatesk)
• no explicit MLJBase requirementfor creating @pipeline (#281) (@aviatesk)
• For a 0.13.4 release (#287) (@ablaom)
• For 0.13.4 release - Take 2 (#288) (@ablaom)

v0.13.3

MLJBase v0.13.3

Diff since v0.13.2

Remove use of channels for displaying progress meter in CPU1 and CPUProcesses versions of resampling (PR #283 )

Merged pull requests:

• Update train_test_pairs() for stratified CV (#276) (@CameronBieganek)
• Simplify stratified CV (#277) (@CameronBieganek)
• removed channels for CPU1 and CPUThreads (#283) (@OkonSamuel)
• For a 0.10.3 release (#284) (@ablaom)

v0.13.2

MLJBase v0.13.2

Diff since v0.13.1

Minor issues:

• #266 (PR #269)
• #220 (PR #271)

Closed issues:

• @mlj_model macro fails on negative default args [needs docs for way around] (#68)
• info/info_dict bottleneck (#76)
• possible test failure in upcoming Julia version 1.5 (#266)

Merged pull requests:

• Anticipate test break in julia 1.5 due to new rand! (#269) (@ablaom)
• For a 0.13.2 release (#270) (@ablaom)
• Better errors for operations with no data argument (#271) (@ablaom)

v0.13.1

MLJBase v0.13.1

Diff since v0.13.0

• Add machine support for supervised models that take X = nothing for input features, to fit a distribution or sampler to the target y.

• Bump compatibility requirement for MLJModelInterface to "^0.2.2" (to enable the above)

• Fix some transformer doc-strings

• Allow machines wrapping Static models to transform without first calling fit!

Merged pull requests:

• Remove error message for running fit! on a Static transformer (#263) (@ablaom)
• Enable fitting of supervised models which take X = nothing (#264) (@ablaom)
• For a 0.13.1 release (#265) (@ablaom)

v0.13.0

MLJBase v0.13.0

Diff since v0.12.9

• (mildly breaking enhancement) Spread remainder evenly among folds in train_test_pairs(::CV, ...) (#232 PR #243)

• (breaking) Restyle report and fitted_params for exported learning networks to include a dictionary of reports or fitted_params, keyed on the machines in the underlying learning network. Retain the machines key in both cases for user convenience (#254, PR #259)

• Update compatiblity for ScientificTypes to "0.1^,0.2"

Merged pull requests:

• Bump to 0.13.0 (#262) (@ablaom)

v0.12.8

MLJBase v0.12.8

Diff since v0.12.7

• Fix nrows issue (#246)

• Extend compatibility requirement for PrettyTables to "^0.8,^0.9"

• Fix evaluate hanging issue (#235)

Closed issues:

• suppress display of data in @load_iris, etc (#226)
• For a 0.12.6 release (#231)
• evaluate! hangs instead of throwing exceptions (flawed exception handling) (#235)
• nrows() of degenerate Matrix of Union{} is 0 (#239)
• nrows gives wrong answer for row tables (#246)

Merged pull requests:

• fixes issue 235 (#245) (@OkonSamuel)
• Fix nrows issue and extend compat for PrettyTables (#249) (@ablaom)
• For a 0.12.8 release (#250) (@ablaom)

v0.12.7

MLJBase v0.12.7

Diff since v0.12.6

• Update [compat] for LossFunctions.jl (PR #237)

Closed issues:

• join method for one-dimensional ranges (#211)

Merged pull requests:

• Bump LossFunctions v0.6 in Project.toml (#237) (@juliohm)
• Spread remainder evenly among folds in train_test_pairs(::CV, ...) (#240) (@CameronBieganek)
• Revert "Spread remainder evenly among folds in train_test_pairs(::CV, ...)" (#241) (@ablaom)
• For a 0.12.7 release (#244) (@ablaom)

v0.12.6

MLJBase v0.12.6

Diff since v0.12.5

• Add a new built-in dataset, loaded with X, y = @load_smarket or data = load_smarket()

Merged pull requests:

• Update [compat] JLSO = "^2.1,^2.2" (#230) (@ablaom)
• Add the "ISLR"/"Smarket" dataset and implement @load_smarket, load_smarket() (#233) (@ablaom)
• For a 0.12.6 release (#234) (@ablaom)

v0.12.5

MLJBase v0.12.5

Diff since v0.12.4

Update [compat] StatsBase = "^0.32,^0.33"

Merged pull requests:

• extend [compat] StatsBase = "^0.32,^0.33" (#229) (@ablaom)