Supplementary Material for CrossGNN: Confronting Noisy Multivariate Time Series Via Cross Interaction Refinement Anonymous Author(s) Affiliation Address email

We conduct extensive experiments on 8 real-world datasets following [4]. Correlation mechanism to capture cross-time dependency for forecasting. Besides, the dimension of the channel is set to 16 based on efficiency considerations. The first row shows the performance when the prediction horizon is 96, while the second row shows the performance when the prediction horizon is 336. Figure 3: The MSE (left Y-axis) and MAE results (right Y-axis) of CrossGNN with different number of scales (X-axis) on ETTh2, ETTm2, Traffic, and Weather. Figure 4: The MSE (left Y-axis) and MAE results (right Y-axis) of CrossGNN with different K (X-axis) on ETTh2, ETTm2, Traffic, and Weather.