Inspired by the Lottery Ticket Hypothesis (LTH), which highlights the existence of efficient subnetworks within larger, dense networks, a high-performing Winning Subnetwork (WSN) in terms of task performance under appropriate sparsity conditions is considered for various continual learning tasks. It leverages pre-existing weights from dense networks to achieve efficient learning in Task Incremental Learning (TIL) scenarios. In Few-Shot Class Incremental Learning (FSCIL), a variation of WSN referred to as the Soft subnetwork (SoftNet) is designed to prevent overfitting when the data samples are scarce. Furthermore, the sparse reuse of WSN weights is considered for Video Incremental Learning (VIL). The use of Fourier Subneural Operator (FSO) within WSN is considered. It enables compact encoding of videos and identifies reusable subnetworks across varying bandwidths. We have integrated FSO into different architectural frameworks for continual learning, including VIL, TIL, and FSCIL. Our comprehensive experiments demonstrate FSO's effectiveness, significantly improving task performance at various convolutional representational levels. Specifically, FSO enhances higher-layer performance in TIL and FSCIL and lower-layer performance in VIL

翻译：受彩票假说（Lottery Ticket Hypothesis，LTH）启发，该假说揭示了大型稠密网络中存在高效的子网络，本文在适当稀疏条件下，针对各类持续学习任务，考虑了一种在任务性能方面表现优异的优胜子网络（Winning Subnetwork，WSN）。该子网络利用稠密网络的预训练权重，在任务增量学习（Task Incremental Learning，TIL）场景中实现高效学习。在少样本类增量学习（Few-Shot Class Incremental Learning，FSCIL）中，设计了一种名为软子网络（Soft Subnetwork，SoftNet）的WSN变体，以防止数据样本稀缺时的过拟合问题。此外，本文还考虑了视频增量学习（Video Incremental Learning，VIL）中WSN权重的稀疏复用。我们进一步探讨了在WSN中引入傅里叶子神经算子（Fourier Subneural Operator，FSO），该算子能够实现视频的紧凑编码，并识别不同带宽下可复用的子网络。我们将FSO集成到包括VIL、TIL和FSCIL在内的不同持续学习架构框架中。综合实验表明，FSO在多种卷积表征层级上显著提升了任务性能，具体而言，FSO在TIL和FSCIL中增强了高层性能，在VIL中则增强了低层性能。