Cuml: [FEA] Implement SKL preprocessing.normalize

[miguelusque]

Is your feature request related to a problem? Please describe.
I am porting existing code from SKL to cuML

A clear and concise description of what the problem is. Ex. I wish I could use cuML to do [...]
It would be handy to have implemented preprocessing.normalize method from SKL.
https://scikit-learn.org/stable/modules/preprocessing.html#normalization

Describe the solution you'd like
Implement preprocessing.normalize (https://scikit-learn.org/stable/modules/preprocessing.html#normalization)

Describe alternatives you've considered
In the meanwhile, I will normalize by looping each column, subtracting the mean and dividing by the standard deviation.

[oyilmaz-nvidia]

We have this algorithm in the C++ and CUDA level and use it for GLM algorithms. For instance, if you set normalize=True in OLS, the data will be automatically normalized.

If you want to normalize the data in the python level, cuDF might help you actually.

[miguelusque]

Thank you Onur,

If it is in the C++ layer, I assume it would be easy to expose it in the python layer.

Would you consider adding it in future releases?

It would be great to be able to normalize (l1 and l2) a dataframe the same way I do now with SKL.

Regards,
Miguel

[oyilmaz-nvidia]

There is actually a plan to expose these kind of primitives in the python layer in the future releases. Don't know when though.

[miguelusque]

This is great news. Thank you!

[miguelusque]

Hi!

I was wondering if there is any ETA for this.

Thanks!
Miguel

[JohnZed]

Timing is TBD but hopefully by 0.13

[ChristofHenkel]

bump

[ChristofHenkel]

working on a PR