Pull requests are welcome.

TODO

Catching up with Python

Write tests (translate the relevant ones from scikit-learn)
Support parallel execution. I'm not sure if njobs=2 is the right way to do this, but I don't have any experience with the Julia primitives
Review docstrings. They still contain Python code.
Support sparse matrices (grep for 'sparse' in /test). This might require work on PyCall.jl
Model serialization (possibly by implementing JLD serialization for PyObjects through pickling)

There are three ways to make a library compatible with ScikitLearnBase.jl

Make a PR to that library
Create a new library that wraps the type and implements the interface
Make a PR to ScikitLearn.jl, and use @require to avoid adding the library as a dependency

Option 1 is generally preferable where possible.

Mocha.jl (or MXNet, Merlin). It's a good candidate for being a separate library (SkMocha.jl?). sklearn-theano is a good starting point.

RegERM.jl (looks maintained + has an SVM implementation)
DiscriminantAnalysis.jl has an LDA and QDA implementation, and is currently maintained
SALSA: SVMs, maintained
Clustering.jl
GLM
MixedModels (also consider FixedEffectModels)
Mamba/Lora. Then we can have BayesianLinearRegression, BayesianHierarchicalModel, etc.
XGBoost.jl - actively maintained
BNP.jl is a recently-built package, pretty solid. Hard to tell if it will be maintained, but looks great. Not released though.
MultivariateStats.jl has some interesting algorithms. Eg. Lasso, ICA
Make some NLP utilities available? TextAnalysis
KernelDensity.jl
OnlineAI.jl
TopicModels has an LDA implementation. Not active, but still maintained
BayesianDataFusion is a solid package, documented and recently maintained.
DirichletProcessMixtures.jl would be cool, but not very active
BayesNets.jl is interesting and currently in development
Bootstrap.jl?? Is it possible to get bootstrap confidence intervals around arbitrary models?
Curvefit.jl?