Nixtla's Open Source Time Series Ecosystem.

Nixtla Community

In issue <https://github.com/Nixtla/neuralforecast/issues/217|#217> `f_cols` is said to be used to identify future available exogenous variables. But I can't find anywhere in the whole code that `f_cols` and `f_idxs` are actually used. How are exogenous variables that are available in the future and the ones that are not, treated differently?

Hi <@U03D9LFSDEK>, thanks for reaching out. We are currently working on that feature. The `f_cols`is sort of a placeholder right now.

<@U038ZT9UDGA>, anything you might want to add?

Hi <@U03D9LFSDEK>,

For the moment the method operates with exogenous variables available at the time of the prediction (future exogenous).
Checkout the <https://arxiv.org/abs/2104.05522|N-BEATSx paper> for a more precise explanation.

Examples of them: Calendar features, controlled prices and promotions, knowledge of future events.
We intend to have next versions of N-BEATSx, and neuralforecast models to operate with past temporal variables and future temporal variables separatedly.

read the N-BEATSx paper, I was misusing N-HiTS. thanks. looking forward to the next version of N-BEATSx.