Okon Samuel

Sorry wanted to find out the status of this issue

I guess it's still distinct as we have `TableTraits.jl` and `Tables.jl`. There has been no unification as to the best of my knowledge but `Tables.jl` no longer has a dependency...

> What is the current state of this ?? I need feature importance support ! Am still working on this. Will be done soon.

Strange, I don't experience the same error for `MultinomialClassifier` on my PC when running julia with 4 threads. ```julia julia> using MLJ julia> function _make_binary() data = MLJ.load_crabs() y_, X...

Allowing input to models to be an instance `AbstractMatrix` is a good solution to the sparsity issue.(just let the model implementers worry about it). My only concern is the inconsistency...

yes `MLJModels` implements LIBSVM models as `Deterministic` models. hence `AUC` measure doesn't apply. Although am not sure what the `probability` keyword in `SVC` does

@Leonardbcm . You have pretty much figured this out. I just have a the following addition to make. In MLJ, Composite models have a fallback `update` method for user defined...

> However, in your example, it seems that the Sub-Models are not retrained : new machines are created and they are fitted from scratch (In your second call to `fit!(m)`...

> So the only option is to re-implement an update method? Yes

> However, this also causes the reconstruction of a new learning network. This is a consequence of the `update` fall back method particularly this line ` if !issubset(submodel_ids, network_model_ids)`