Hi all! I received a question from a user, Peter Schofield (pcschof@gmail.com), that I’m unsure about. Do we handle stationarity in TimeGPT’s preprocessing pipeline? If not, do we have specific tools or methods for addressing non-stationary time series? Many thanks! Thanks for your help!

I’m curious if the proprietary preprocessing pipeline for TimeGPT includes steps to handle stationarity in the time series data. If so, could you share whatever you can about how that’s managed?