ts-kim / RevIN

RevIN: Reversible Instance Normalization For Accurate Time-series Forecasting Against Distribution Shift
MIT License
272 stars 23 forks source link

Is RevIN causal? #15

Open jaak-s opened 6 months ago

jaak-s commented 6 months ago

Thank you for the code.

I did a quick check with revin to see if it is causal along the sequence length axis, where by "causal" I mean that only the past and current tokens are used for transforming the token (or in other words that the future tokens cannot impact past). My results show that RevIN is not causal.

Can you confirm that this is correct? Thanks!