unit8co / darts

A python library for user-friendly forecasting and anomaly detection on time series.
https://unit8co.github.io/darts/
Apache License 2.0
7.56k stars 829 forks source link

window transforms as "encoders" #1515

Open hrzn opened 1 year ago

hrzn commented 1 year ago

Right now we have a transformer for window transforms. It'd be nice to also have an add_encoders-equivalent way to specify window transforms, in order to make them work out of the box with auto-regressive models. It can be re-using the add_encoders argument, or maybe introducing a new add_window one.

nagydavid commented 1 year ago

is there anyone working on this actively or is just in the pipeline. I am interested to add it, but I need definitely some initial guidance.

nagydavid commented 1 year ago

Hi All, is there any update on this task?

madtoinou commented 1 year ago

Hi @nagydavid,

This feature is still open to grab. Would you be interested in contributing?

nagydavid commented 1 year ago

@madtoinou I am still up for it, I am on holiday, when I get back I will look into it :)

wojnarabc commented 1 year ago

Would be great to have capability of do window calculations on the fly. Keeping fingers crossed for this one :)

GunZerker-data commented 9 months ago

Hi all, many thanks for your work, DarTS is very useful to me on a daily basis ! Can't wait to try out this new feature, which is very useful for purely autoregressive time series :)

Jonathan-87 commented 6 months ago

Hi @madtoinou do you think this feature will be available in a near future ?

madtoinou commented 6 months ago

It's currently being partially implemented in PR #2021 (for historical_forecasts() methods).

Once it's merged, we'll try to extend it to the models themselves but difficult to tell if it'll be part of the next release.