A Datasets Following [5, 7], we run experiments on five popular datasets for time series forecasting, including, T

Dimension, domain, frequency, total training timesteps and prediction length properties of the training datasets used in the experiments. For human motion prediction, we run experiments on two datasets, namely, Human3.6M[2] and HumanEva-I [8], following [10]. As described in Section 5, Human3.6 is a large-scale dataset with 11 subjects performing 15 actions, totaling 3.6 million video frames recorded at 50Hz. To be consistent with previous work [4, 10], we adopt a 17-joint skeleton and train on 5 subjects (S1, S5, S6, S7, S8) and test on two subjects (S9, S11). For HumanEva-I, we adopt a 15-joint skeleton and use the same training and test split provided in the dataset.