v0.1.6

MathOptAI v0.1.6

Diff since v0.1.5

Merged pull requests:

• [docs] clarify instructions to load PythonCall (#163) (@odow)
• Prep for v0.1.6 (#164) (@odow)

Closed issues:

• Failed to reproduce the PytorchModel examples (#162)

v0.1.5

MathOptAI v0.1.5

Diff since v0.1.4

Merged pull requests:

• Add gray_box_device option for PytorchModel (#159) (@Robbybp)
• Prep for v0.1.5 (#160) (@odow)
• Fix flakey tests (#161) (@odow)

v0.1.4

MathOptAI v0.1.4

Diff since v0.1.3

Merged pull requests:

• Add bound constraints to the Formulation object (#155) (@odow)
• [docs] improvements to the docstrings (#156) (@odow)
• [docs] fix various typos (#157) (@odow)
• Prep for v0.1.4 (#158) (@odow)

v0.1.3

MathOptAI v0.1.3

Diff since v0.1.2

Merged pull requests:

• Delete .github/workflows/documentation-deploy.yml (#150) (@odow)
• [docs] improve manual for NN extensions (#152) (@odow)
• Fix documentation link in README.md (#153) (@odow)
• Prep for v0.1.3 (#154) (@odow)

Closed issues:

• Review feedback (#82)
• Tag did not trigger documentation build (#145)
• [docs] improve manual sections (#151)

v0.1.2

MathOptAI v0.1.2

Diff since v0.1.1

Merged pull requests:

• Add DOCUMENTER_KEY (#148) (@odow)
• Prep for v0.1.2 (#149) (@odow)

v0.1.1

MathOptAI v0.1.1

Diff since v0.1.0

Merged pull requests:

• Update installation instructions for registered package (#141) (@odow)
• Create documentation-deploy.yml (#146) (@odow)
• Prep for v0.1.1 (#147) (@odow)

v0.1.0

MathOptAI v0.1.0

Merged pull requests:

• Add package skeleton and a basic implementation of LinearRegression (#2) (@odow)
• Add GLM package extension (#4) (@odow)
• Use Predictor instead of Model to distinguish from optimization model (#5) (@odow)
• Add support for LogisticRegression (#6) (@odow)
• Add docstrings (#7) (@odow)
• Add UnivariateNormalDistribution (#8) (@odow)
• Add preliminary NN support (#9) (@odow)
• Remove add_predictor! (#10) (@odow)
• Propagate variable bounds where possible (#11) (@odow)
• Pause actions until repo is open-sourced (#12) (@odow)
• Simplify to make extensions implement add_predictor (#13) (@odow)
• Add test for relu kwarg (#14) (@odow)
• Improve variable bound propagation in ReLU layers (#15) (@odow)
• Add Omelette.ReLU (#17) (@odow)
• Test and fix doctests (#18) (@odow)
• Add Sigmoid, SoftPlus, and Tanh layers (#19) (@odow)
• Fix test_GLM.jl (#20) (@odow)
• Support other activation functions in Lux extension (#21) (@odow)
• Tweak GLM logistic regression layer (#22) (@odow)
• Rename to MathOptAI.jl (#23) (@odow)
• Use predictor consistently instead of model (#24) (@odow)
• Add Flux.jl extension (#25) (@odow)
• Improve extension docstrings (#26) (@odow)
• Rename LinearRegression to Affine (#28) (@odow)
• Remove LogisticRegression predictor (#29) (@odow)
• Allow non-variable input vectors (#30) (@odow)
• Add matrix input/output method for add_predictor (#31) (@odow)
• Add StatsModels extension (#32) (@odow)
• Make DataFrames a weakdep (#33) (@odow)
• Fix code highlighting (#34) (@odow)
• Switch to LANL's BSD license (#36) (@odow)
• [docs] add MNIST tutorial (#37) (@odow)
• Add initial skeleton of the documentation (#38) (@odow)
• Rename added variables to moai_ (#39) (@odow)
• Add student ennrollment tutorial (#40) (@odow)
• Add SoftMax predictor (#41) (@odow)
• [docs] add manual section (#42) (@odow)
• Add kwarg option to extension methods (#46) (@odow)
• Fix Flux Float32 warning (#47) (@odow)
• [docs] add design principles (#48) (@odow)
• Remove usage of Lux.Experimental (#49) (@odow)
• Fix formatting (#50) (@odow)
• Fix various docstrings (#51) (@odow)
• Add logo (#52) (@odow)
• [docs] update mnist tutorial (#53) (@odow)
• Add BinaryDecisionTree (#54) (@odow)
• [docs] fix BinaryDecisionTree docstring (#55) (@odow)
• [docs] add docstrings for extensions (#56) (@odow)
• Standardize naming of test modules (#57) (@odow)
• Add tests for DecisionTreeExt (#58) (@odow)
• [docs] add dimension colum to predictor table (#59) (@odow)
• Add AbstractGPsExt (#60) (@odow)
• Fix test_DecisionTree for points near breakpoints (#61) (@odow)
• Remove UnivariateNormalDistribution in favor of Quantile (#62) (@odow)
• [docs] update docs with new extensions (#63) (@odow)
• Update README.md (#64) (@odow)
• Add PytorchModel and tutorial (#69) (@odow)
• Add ReducedSpace{<:AbstractPredictor} layer (#72) (@odow)
• Fix tests for update to MOI variable ordering (#73) (@odow)
• Add reduced_space option for PyTorchModel predictor (#76) (@Robbybp)
• Add methods for show(io::IO, ::AbstractPredictor) (#77) (@odow)
• Add build_predictor (#78) (@odow)
• Add CondaPkg to test/Project.toml (#79) (@odow)
• [breaking] return a formulation object (#80) (@odow)
• Various miscellaneous fixes (#81) (@odow)
• Relax type restrictions to support JuMP.AbstractModel (#83) (@odow)
• Allow nested ReducedSpace predictors (#84) (@odow)
• Various documentation improvements (#85) (@odow)
• [docs] add more design principles and add PyTorch manual page (#86) (@odow)
• Fix deprecation warning in tests (#88) (@odow)
• Add Scale predictor (#89) (@odow)
• [docs] remove Scaling section from design_principles.md (#91) (@odow)
• [docs] add inputs are Vectors to design principles (#93) (@odow)
• Fix typos in Scale doctest (#94) (@odow)
• Fix MethodError with unsupported layers in (F)lux (#95) (@odow)
• Add GrayBox predictor (#96) (@odow)
• Fix formatting in src/predictors/Scale.jl (#97) (@odow)
• Add test/test_PythonCall.jl (#98) (@odow)
• [docs] update sources of inspiration (#99) (@odow)
• Add Hessian support to GrayBox (#100) (@odow)
• [docs] fix doc build and avoid segfault via PythonCall (#103) (@odow)
• [docs] add DecisionTree tutorial, tweak other docs (#104) (@odow)
• Add a tutorial for Lux (#105) (@odow)
• [docs] minor tweaks to the documentation (#106) (@odow)
• s/Pytorch/PyTorch/ (#107) (@odow)
• Add tutorial for AbstractGPs (#108) (@odow)
• Add support for nn.Softplus layer in PyTorch (#111) (@odow)
• Add instructions for non-conda python environment (#113) (@Robbybp)
• Support beta parameter in SoftPlus (#114) (@Robbybp)
• Fix typo in SoftMax docs (#115) (@Robbybp)
• Tweak LICENSE to match wording from LANL (#117) (@odow)
• Add GitHub actions (#118) (@odow)
• Update README.md to mention project code (#119) (@odow)
• Clarify relationship to OMLT in README (#120) (@odow)
• Add documentation to CI (#121) (@odow)
• Run GitHub actions CI on pushes to main (#122) (@odow)
• Add badges to README (#123) (@odow)
• Add CODECOV_TOKEN to Github Actions (#124) (@odow)
• Build documentation on push to main (#125) (@odow)
• Add documentation link to the README (#126) (@odow)
• Make the tests less flakey (#128) (@odow)
• Improve code coverage (#129) (@odow)
• Improve code coverage (#130) (@odow)
• Add doc_cleanup to GitHub actions (#132) (@odow)
• Update tests to use Julia v1.11 (#133) (@odow)
• Update codecov badge in README (#134) (@odow)
• Remove explicit mention of Mo'ai (#135) (@odow)
• Install torch in GitHub actions (#136) (@odow)
• Add installation instructions to the README (#137) (@odow)
• Fix implied variable bounds in ReLU (#138) (@odow)

Closed issues:

• MVP planning (#1)
• Neural networks (#3)
• Logistic layer (#16)
• GLM.jl and DataFrames.jl (#27)
• Add a way to override layer choices (#35)
• Add GradientBoostedTrees (#43)
• Add GaussianProcess (#44)
• Tensorflow JSON (#65)
• [breaking] return config struct from each layer (#67)
• Add support for PyTorch (#68)
• Support reduced-space formulation (#70)
• Add a way to return predictor from extensions (#74)
• Doctests fail if CondaPkg is not installed (#75)
• Add OffsetScaling predictor (#87)
• [GrayBox] add Hessian support (#90)
• MOI Hessian evaluation is slow with reduced-space predictors (#102)
• Add soft plus layer for torch (#110)
• Threading and PyTorch (#116)
• Tests are flakey (#127)
• Add PyTorch to GitHub actions (#131)