Exploring Lottery Ticket Hypothesis in Neural Video Representations

Jiacong Chen, Qingyu Mao, Shuai Liu, Fanyang Meng, Shuangyan Yi, Yongsheng Liang

Abstract: Highlights•We present the first comprehensive exploration of applying the Lottery Ticket Hypothesis (LTH) within Neural Video Representations (NVR), aiming to discover a suitable pruning method.•This paper demonstrates the existence of winning tickets in NVR and explores an identify winning tickets method specifically for NVR.•We also explore the transferability properties of winning tickets and confirm their content-specific characteristics. Based on these characteristics and the video frames redundancy, a Progressive Cyclic Learning method is devised to expedite the process of finding winning tickets.