v0.3.0

SossMLJ v0.3.0

Diff since v0.2.0

Closed issues:

• TagBot trigger issue (#126)

Merged pull requests:

• Drop support for Julia <1.6 (#138) (@DilumAluthge)

v0.2.0

SossMLJ v0.2.0

Diff since v0.1.0

Merged pull requests:

• Remove the cron jobs (#123) (@DilumAluthge)
• Remove some whitespace (#124) (@DilumAluthge)
• Bump version from "0.2.0-DEV" to "0.2.0" (#125) (@DilumAluthge)

v0.1.0

SossMLJ v0.1.0

Closed issues:

• add BayesianLogreg following Ridgereg example (#2)
• Particles for posterior? (#5)
• Similar projects (#6)
• Preprocessing transform (#15)
• Create DOCUMENTER_KEY (#18)
• TODO: Remove the dependency on MLJBase.jl (#30)
• Roadmap to enabling precompilation (#34)
• mainexample.jl: are the results correct? (#35)
• particles (#37)
• Names need qualification (#39)
• TODO: Turn mainexample.jl into actual documentation (#41)
• TODO: Add an way to specify which variable is observed, and which is predicted (#43)
• Deleted files (#47)
• TODO: add Bayesian binomial logistic regression example (#49)
• TODO: add Bayesian multinomial logistic regression example (#50)
• For classifiers, we need to figure out how to return a Vector{UnivariateFinite} (#51)
• TODO: Remove the dependency on Requires.jl (#58)
• Implement a simple loss function (#60)
• Weird behavior of particles w.r.t. mean vs median? (#61)
• Iris example for multinomial logistic regression (#63)
• "MultivariateFinite" distributions (#65)
• predict_joint method for Machines (#66)
• Extending For and UnivariateFinite (#68)
• TODO: compare accuracy of our Iris model to that of Turing's Iris model (#69)
• TODO: require the latest versions of MLJBase and MLJModelInterface (#70)
• TODO: remove the logpdf definition once it has been merged in MLJBase upstream (#71)
• TODO: add tests for the per-class accuracy of the Iris model (#72)
• TODO: increase code line coverage (#73)
• Better GitHub repository description (#79)
• TODO: use real data for the linear regression example (#81)
• @reexport using Soss? (#82)
• Bug: in multinomial logistic regression, MLJBase.predict throws an ArgumentError (#84)
• Branch protection settings for the master branch (#87)
• Tests are failing on Julia nightly (#112)

Merged pull requests:

• Speed up brr (test is now 2s on my machine) (#1) (@cscherrer)
• trasfer PR (#8) (@cscherrer)
• Fix a typo in a comment (#9) (@DilumAluthge)
• Set up CI, and add a brief description (#10) (@DilumAluthge)
• Fix CI (#11) (@DilumAluthge)
• CompatHelper: add new compat entry for "MLJModelInterface" at version "0.3" (#14) (@github-actions[bot])
• CompatHelper: add new compat entry for "Soss" at version "0.12" (#21) (@github-actions[bot])
• CompatHelper: add new compat entry for "Distributions" at version "0.23" (#22) (@github-actions[bot])
• Got it working! Sorta (#24) (@cscherrer)
• predict_joint (#25) (@cscherrer)
• A note to ourself to eventually use MMI.JointProbabilistic (once that type exists upstream) (#26) (@DilumAluthge)
• Temporarily disable CI on Julia nightly (#27) (@DilumAluthge)
• Re-enable CI on Julia nightly (#28) (@DilumAluthge)
• Add the helper code to make the predict_joint operation work on machines (#29) (@DilumAluthge)
• SossMLJPredictor{M} <: Distributions.Distribution (#36) (@cscherrer)
• Add (and export) the predict_particles function (#38) (@DilumAluthge)
• Debugging (#40) (@cscherrer)
• Convert mainexample.jl to documentation, add tests, and reorganize the source files (#46) (@DilumAluthge)
• Docs / examples: We don't really need to do const MMI = MLJModelInterface if we only use MLJModelInterface one time (#48) (@DilumAluthge)
• Linear regression example: add some using statements (#53) (@DilumAluthge)
• Examples: Improve some comments (#54) (@DilumAluthge)
• Implement the MLJModelInterface.predict_mean method (#55) (@DilumAluthge)
• A few small tweaks (#57) (@DilumAluthge)
• Use RMSE instead of L2 (#59) (@DilumAluthge)
• Multiple improvements: add RMSE loss, add response kwarg, add prior for σ, use the latest version of MLJBase.jl (#64) (@DilumAluthge)
• Some more updates (#67) (@cscherrer)
• CompatHelper: add new compat entry for "NamedTupleTools" at version "0.13" (#74) (@github-actions[bot])
• CompatHelper: add new compat entry for "CategoricalArrays" at version "0.8" (#75) (@github-actions[bot])
• Get code coverage to 100% (#76) (@DilumAluthge)
• Fix formatting in the docs (#77) (@DilumAluthge)
• Docs: Fix spelling (#78) (@DilumAluthge)
• Docs: Add link back to the GitHub repository (#80) (@DilumAluthge)
• Several improvements to the multinomial classification example (#83) (@DilumAluthge)
• Small tweaks to the CI configuration (#85) (@DilumAluthge)
• Require the latest versions of MLJBase and MLJModelInterface (#86) (@DilumAluthge)
• Docs: Fix a typo (#88) (@DilumAluthge)
• Make some text more concise (#90) (@DilumAluthge)
• A really hacky and incorrect implementation of predict for classifiers (#91) (@DilumAluthge)
• Docs: Fix a typo (#96) (@DilumAluthge)
• Fix a comment (#97) (@DilumAluthge)
• @reexport using Soss (#98) (@cscherrer)
• CompatHelper: add new compat entry for "Reexport" at version "0.2" (#99) (@github-actions[bot])
• Several small tweaks (#101) (@DilumAluthge)
• Fix a typo (#102) (@DilumAluthge)
• Linear regression example: Increase the number of rows in the synthetic data set (#103) (@DilumAluthge)
• Add a unit test (#104) (@DilumAluthge)
• Fix a typo (#105) (@DilumAluthge)
• Enable precompilation, and require Soss 0.14 (#106) (@DilumAluthge)
• Delete an empty page from the documentation (#107) (@DilumAluthge)
• Add some ".jl"s (#108) (@DilumAluthge)
• Set up VersionVigilante (#111) (@DilumAluthge)
• [docs] Remove italics (#113) (@DilumAluthge)
• CompatHelper: bump compat for "Soss" to "0.15" (#114) (@github-actions[bot])
• Don't re-export Soss (#115) (@DilumAluthge)
• Bump version from "0.1.0-DEV" to "0.1.0" (#116) (@DilumAluthge)