This message was deleted.

Hey. The only implemented target transforms are the differences and the local standard scaler and they're not documented. You can also define your own, you can find an example on how to do it on the end to end tutorial. I'll also add a guide soon with a class that will take an sklearn transformer to easily apply global transformations like log, boxcox, etc. For lag transforms all the functions in the window ops library work and you can also define your own, they can be any function that takes an array and returns an array of the same length which can be jit compiled with numba. I'll also add a guide with some examples soon. Sorry for the lack of documentation, we'll work on addressing this this month

Thank you. I think it's a great package, and some additional documentation and tutorials would go a long way.

Hi. We just added this guide on target transformations, hope you find it useful

The guide is terrific! Thanks for putting it together.

The window_ops are for lag_transforms and this guide focuses on target_transforms. They're a bit different because the lag_transforms use transformations that get updated with new data whilst the target_transforms use transformations where some parameters are learned to perform the transformation and then have to be inverted

Will you be working on a guide for lag transforms as well?

Sure. What do you think should be covered?

1. The different lag transforms available - rolling, expanding, ewm, etc. 2. How lag transforms are applied 3. Local vs. global transformations 4. How to build your own local and global transformations