Static features in paper not supported in repo or lack of clarity?

[Thomas2419]

There is no documentation in the paper, or in the repo on the expected usage of static features, while it is mentioned in paper the # of static features per dataset. Is the expected usage just to broadcast a static feature across every single timestep related to said static feature?

[GGA-PERSO]

I am also interested with this question @abdulfatir. Thank you.

[Thomas2419]

I've done some digging, in fact to go even further on this point, assuming I've read this correctly, static features you just expand currently are entirely uninformative and useless. Categoricals get turned into numerics, if said value is the same across the entire series, instance norm takes the value and normalizes it to be completely uninformative noise. Not sure why I thought that in light of instance norm that this would function at all, but it in fact doesn't seem to currently support static features, or even categorical if a specified time slide doesn't include other options.

Failure case example: If I have 48 hours of 10 minute intervals of sales, and the random time slicing excludes the previous day, and we have a categorical feature for day of the week, said feature will get normalized to a single uninformative value.

I will now raise this to a bug as I'm fairly certain this is not the intended setup. Or the paper has some completely random BS in it that in fact the model isn't designed to support.

[abdulfatir]

@Thomas2419 @GGA-PERSO I responded to this in #352. Please let me know if you have further questions.