Cross-FCL:Cross-edge Federated Continual Learning
Cross-FCL explores the enhancement of continuous learning capabilities for cross-edge devices through the method of “parameter separation + cross-edge knowledge fusion” to balance memory and adaptability.
Model Architecture Diagram：
In ubiquitous environments, there are often mobile devices spanning multiple independent edge federated learning systems, learning a series of tasks. Due to the differences in scenarios and tasks between different FL systems, cross-edge devices will forget past tasks after learning new ones, which is unacceptable for devices that have incurred system costs to participate in FL. To address this issue, Cross-FCL, a cross-edge federated continual learning framework, is proposed. By using a parameter-separation-based federated continual learning model, devices can retain past learned knowledge when participating in new task training. Multiple cross-edge strategies are introduced, including biased global aggregation and local optimization, to balance memory and adaptability.