Anthony Blaom, PhD

I agree it would be neat but I'm not sure it makes sense at the present time. Here are few reasons: - I presume StatsAPI `predict` has a different contract...

Oh, and reversing our decision about ownership is definitely breaking.

I was not aware of `Base.nameof`. It's doc-string suggests a different purpose. `human_name` is basically for documentation, some of which is generated automatically. So, for example, ```julia using MLJTSVDInterface julia>...

See the `fit_only!` docstring for the detailed logic on whether `fit` or`update` is dispatched. The user typically interacts via `fit!` only (which calls `fit_only!` on the machine, and in the...

This is tricky. I think I need a pretty good argument for dodging two core Julia libraries. REPL was added because `Base.Docs.doc` stopped working after a recent Julia release, which...

That's possible. I'd be happy to support such an effort. I think it will be a bit of work. At this point a lot of MLJ predictors do not have...

@CameronBieganek Thanks for pitching in with this idea. Here are a few drawbacks that come to mind. I wonder what you think about them: - Essentially you are replacing an...

This sounds very reasonable. For the ContinuousEncoder, perhaps we can just call `Float32.(...)` after the coercion. Be great to have a PR, thanks! Would it be enough to have a...

After thinking about this more, my vote would be to add this to MLJFlux, as this is, for now, the only package fussy about float types. In the rare we...

> In MLJFlux, I wouldn't implement the conversion by adding a layer to the Flux model. Rather do it in the data pre-processing step, so that it only needs to...