v0.6.0

MLJTuning v0.6.0

Diff since v0.5.5

• (enhancement) Add LatinHypercube tuning (#78, #96 @ludoro)
• (API change) Make a tuned model's number of iterations, n, available to setup method (#106). This change is not breaking for users of MLJTuning but will break third party implementations of the the tuning strategy API (TreeParzen.jl being the only case-in-point of which @blaom is aware)

Closed issues:

• Add LatinHypercube tuning (#78)
• Make a tuned model's number of iterations, n, available to setup method (#106)

Merged pull requests:

• Latin cube 2 (#96) (@ablaom)
• Move from Travis CI to GitHub Actions CI (#102) (@DilumAluthge)
• switch ci to gh action (#103) (@ablaom)
• Set NUM_THREADS=2 in ci.yml (#104) (@ablaom)
• In CI set NUM_THREADS=2 (#105) (@ablaom)
• API change: Make n from a TunedModel instance available to setup (#109) (@ablaom)
• For a 0.6.0 release (#110) (@ablaom)

v0.5.5

MLJTuning v0.5.5

Diff since v0.5.4

Merged pull requests:

• extend [compat] MLJBase = "^0.15,^0.16" (#100) (@ablaom)
• For a 0.5.5 release (#101) (@ablaom)

v0.5.4

MLJTuning v0.5.4

Diff since v0.5.3

Closed issues:

• fit! on TunedModel instances fails with verbosity > 1 (#97)

Merged pull requests:

• Reolve issue 97 (#98) (@ablaom)
• For a 0.5.4 release (#99) (@ablaom)

v0.5.3

MLJTuning v0.5.3

Diff since v0.5.2

Merged pull requests:

• Latin hypercube (#83) (@ludoro) WIP
• For a 0.5.3 release (#93) (@ablaom)
• Add forgotten compat entry for LatinHypercubeSampling.jl (#94) (@ablaom)

v0.5.2

MLJTuning v0.5.2

Diff since v0.5.1

Closed issues:

• Score measure maximization double signal flip (#87)

Merged pull requests:

• Extend [compat] Distributions = "^0.22,^0.23,^0.24" (#84) (@ablaom)
• For a 0.5.2 release (#85) (@ablaom)
• Rm stray text from readme. No new release. (#86) (@ablaom)
• Fix orientation bug (#88) (@ablaom)
• For a 0.5.3 release (#89) (@ablaom)
• fix version (#90) (@ablaom)
• fix compat entry (#91) (@ablaom)
• fix compat (#92) (@ablaom)

v0.5.1

MLJTuning v0.5.1

Diff since v0.5.0

Merged pull requests:

• delete multiple inclusion of same file (#79) (@OkonSamuel)
• For a 0.5.1 release (#80) (@ablaom)
• "Bump version to 0.5.1" (#81) (@ablaom)

v0.5.0

MLJTuning v0.5.0

Diff since v0.4.3

Closed issues:

• For a 0.5 release (#76)

Merged pull requests:

• readme fixes. no new release (#73) (@ablaom)
• Allow user to specify a selection heuristic (#75) (@ablaom)
• For a 0.5 release (#77) (@ablaom)

v0.4.3

MLJTuning v0.4.3

Diff since v0.4.2

Closed issues:

• Plots warning for scale when visualising tuning results. (#65)
• Decouple training weights from weights for measures, in TunedModel (#70)

Merged pull requests:

• ScientificTypes -> MLJScientificTypes in testing (#69) (@ablaom)
• Okonsfixes (#71) (@OkonSamuel)
• For a 0.4.3 release (#72) (@ablaom)

v0.4.2

MLJTuning v0.4.2

Diff since v0.4.1

Closed issues:

• Intermittent failure of CI for Julia 1.3 (#51)
• Allow rows=... specification in learning_curve (#61)

Merged pull requests:

• Readme update (add tree parzen). Not for new release. (#64) (@ablaom)
• fix ambiguous scope warning (#66) (@OkonSamuel)
• Fix plot warning for custom scale (#67) (@ablaom)
• For a 0.4.2 release (#68) (@ablaom)

v0.4.1

MLJTuning v0.4.1

Diff since v0.4.0

Closed issues:

• Issue for triggering new releases (#53)

Merged pull requests:

• Allow specification rows=... in learning curve() (#62) (@ablaom)
• For a 0.4.1 release (#63) (@ablaom)