Models to wrap

Relatively easy wraps are the SckitLearn.jl models, presently being wrapped one-at-a-time. Checkout regressor metadata for scikit-learn learners that are regressors (needs checking), which are the easiest ones to wrap. Can use ElasticNetCV (already done) in MLJModels/src/ScikitLearn.jl as a template, which is where new wraps will also live. Check the models do not already have implementations by entering models() in MLJ.

A template for deterministic classifiers is SVMClassifer in MLJModels/src/ScikitLearn.jl.

ManifoldLearning.jl #161 contains transformers, Isomap, Diffusion maps, Laplacian Eigenmaps, Locally Linear Embedding, Hessian Eigenmaps, Local Tangent Space Alignment.

LowRankModels.jl #44 contains transformers PCA (iterative version) QPCA, Means, robust PCA, among others.

MultivariateStats.jl #167 contains transformers kernalPCA, ICA, among others

ElasticNet.jl is a pure Julia elastic net implementation (familiarity with the method will be needed)

Stheno.jl #57 for Guassian processes (regression, as a start; classification is trickier) - more challenging wrap

Turing.jl #157 for probabilistic programming, subject of special session Wednesday

Flux.jl work-in-progress Ayush/Anthony, subject of special session Thursday

LightGBM.jl - packages needs TLC (binaries not automatically built for this wrap of C)

XGBoost.jl - has existing MLJ wraps but package needs TLC (binaries not building reliably)

Also suggestions for models in other packages eagerly sought.