Efficient time series adaptive representation learning via Dynamic Routing Sparse Attention

Wenyan Wang, Enguang Zuo, Chen Chen, Cheng Chen, Jie Zhong, Ziwei Yan, Xiaoyi Lv

Abstract: Highlights•Developed a DRSA model for time series analysis greatly improves time efficiency.•Introduced a block-based approach that shifts the focus to local views.•Dynamic Route Filter achieves dynamic self-adaptation of locally aware contexts.•DRSA achieves SOTA on long series prediction tasks on six real-world public datasets.