Minor MLJ docstring fixes

[ablaom]

I've left the descriptions for input X unchanged, despite the incongruence with input_scitype, as noted here.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 97.00%. Comparing base (d3f0ed5) to head (c53c2f3).
Report is 2 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #133   +/-   ##
=======================================
  Coverage   97.00%   97.00%           
=======================================
  Files          22       22           
  Lines         734      734           
=======================================
  Hits          712      712           
  Misses         22       22           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[pasq-cat]

@ablam you are right with the scytipe error. i guess i must change

    input_scitype=Union{
        AbstractMatrix{<:Union{MLJBase.Finite,MLJBase.Continuous}}, # matrix with mixed types
        MLJBase.Table(MLJBase.Finite, MLJBase.Continuous), # table with mixed types

in

    input_scitype=Union{
        AbstractMatrix{<:Union{MLJBase.Finite,MLJBase.Infinite}}, # matrix with mixed types
        MLJBase.Table(MLJBase.Finite, MLJBase.Infinite), # table with mixed types

right? afterall i don't see why it shouldn't work with Count types

[pasq-cat]

@ablaom regarding μ₀ and P₀ , i can change the structs but i will have to keep them in the Laplacereduc.laplace objects like this one for regression but also for classification

    la = LaplaceRedux.Laplace(
        copied_model;
        likelihood=:regression,
        subset_of_weights=m.subset_of_weights,
        subnetwork_indices=m.subnetwork_indices,
        hessian_structure=m.hessian_structure,
        backend=m.backend,
        σ=m.σ,
        μ₀=m.mu0,
        P₀=m.p0
    )

is this enough?

[ablaom]

your input_scitype correction looks consistent with the doc-string now. as to whether Count will actually work, I expect so, but that is your responsibility to check 😉

re the notation change: my suggestion is to modify only the keyword constructor to look something like LaplaceRegressor(; m0=1.0, μ₀=m0, ....) = LaplaceRegressor(..., μ₀, ....) but only do this if you are also prepared, in documentation, to always write m0 instead of μ₀ because otherwise this option is not discoverable. The alternative is to completely rename, which would simplify the code but also break it.

i am only making a suggestion, and I expect @pat-alt will want to chime in here before you proceed with a change. you may want to kick the can and raise in a new issue tagged "next breaking release" or something .

[pasq-cat]

i am only making a suggestion, and I expect @pat-alt will want to chime in here before you proceed with a change. you may want to kick the can and raise in a new issue tagged "next breaking release" or something.

ok the scitype isn't a problem, it's just another type of feature. I ill fix the input_scitype. Regarding the other problem,
i asked him via ms teams, i guess he will look as soon as he can.

[pat-alt]

Thanks @ablaom and sorry for my late response, been preoccupied with PhD research these past few weeks.

I'm in favor of removing all unicode characters in a breaking release (#134).