feature request - preprocessing power transform

[edenbaus]

Is it possible to include a similar preprocessing transformer to the scikit learn power transform for future versions of dask-ml?

In particular, I'm interested in an implementation of the Box Cox transformation for larger than memory datasets.

http://scikit-learn.org/dev/modules/generated/sklearn.preprocessing.power_transform.html

Thanks!

[TomAugspurger]

This block: https://github.com/scikit-learn/scikit-learn/blob/96a02f3/sklearn/preprocessing/data.py#L2721-L2724

seems to be parallizable by column (/row, depending on whether we're talking before or after transposing).

Do you know if there are blocked algorithms for power transformations? 30 seconds of googling didn't turn anything up :)