FedViT: Federated continual learning of vision transformer at edge

Xiaojiang Zuo, Yaxin Luopan, Rui Han, Qinglong Zhang, Chi Harold Liu, Guoyin Wang, Lydia Y. Chen

Abstract: Highlights•Propose a lightweight and client-side federated continual learning method.•Prevent catastrophic forgetting by integrating signature tasks’ knowledge on client.•Mitigate negative knowledge transfer without frequent communication to server.•Theoretically guarantees faster training convergency of our method.•Applicable to the scenarios of multiple tasks, clients, and various ViTs.