EvoMSN

This repo is the Pytorch implementation of our submitted paper to ICLR 2025: Evolving Multi-Scale Normalization for Time Series Forecasting Under Distribution Shifts

Usage

Environment and dataset setup

pip install -r requirements.txt
mkdir datasets

All the 9 datasets are available at the Google Driver provided by Autoformer. Many thanks to their efforts and devotion!

Running

We provide ready-to-use scripts for EvoMSN enhanced online forecasting and MSN enhanced offline forecasting.

sh run_<model_name>.sh 

Acknowledgement

This repo is built on the pioneer works. We appreciate the following GitHub repos a lot for their valuable code base or datasets:

SAN

FSNet