SupplementaryMaterials: Autoformer: DecompositionTransformerswithAuto-Correlation forLong-termSeriesForecasting

Autoformer achieves sharp improvement over the state-of-the-art on various forecasting horizons. This design is to provide recent past information to the decoder. Our model gives the best performance among different models. Compared toInformer [14], Autoformer can precisely capture the periods of the future horizon. We also apply our model to the COVID-19 real-world data [4].