In this paper, we propose a dynamic sparse network (DSN) with sparse connections for TSC, which can learn to cover various RF without cumbersome hyper-parameters tuning.

Thus, the robustness of 3D deep learning models against adversarial attacks becomes a major consideration.