Need to rename HalLearners assembly and nuget

[eerhardt]

The name HalLearners isn't a great public name. We should come up with a better name.

Current issues:

1. What does Hal mean?
2. We are using the term "trainers" instead of "learners" everywhere else.
3. This assembly contains more than just learners/trainers. It also has vector whitening, which isn't a trainer.

Possible suggestions to get the naming started:

1. Microsoft.ML.HardwareAccelerated
2. Microsoft.ML.HardwareAccelerated.Components
3. Microsoft.ML.Mkl
   • (we already have Microsoft.ML.Mkl.Redist. this would be an analog to ML.TensorFlow and ML.TensorFlow.Redist)
4. Microsoft.ML.Mkl.Components

@TomFinley @sfilipi @Ivanidzo4ka @shauheen @yaeldekel

[TomFinley]

I think it was a joke to refer to those learners that had this relatively heavyweight MKL library as a dependency for one reason or another.

I like Microsoft.ML.MklTrainers or Microsoft.ML.MklComponents or Microsoft.ML.Mkl or something like this. (If possible Trainers would be good, but there's more in there that depends on MKL than merely trainer estimators -- so maybe the more general name would be good).

"Hardware accelerated" is far too prone to misunderstanding in my opinion. Without knowing what I know, I think if I were to look at this library, and I saw, "hardware accelerated," my natural thought would be "OK, so, they are saying the rest the stuff isn't hardware accelerated." And, that would be incorrect while also making us look kind of silly.

[codemzs]

Even Microsoft.ML.MKLTrainers is a bit controversial because TimeSeries for example uses MKL but is not in this nuget.

[TomFinley]

I see. Well, I'm not sure that's a blocker. We have after all the "transforms" assembly without the suggestion that it refers to all transforms. (Indeed it does not.) How about Microsoft.ML.Mkl.Components then, anyway?

[codemzs]

Microsoft.ML.Mkl.Components it is! :)