Importance- and Channel-Aware Scheduling in Cellular Federated Edge Learning
Jinke Ren, Yinghui He, Dingzhu Wen, Guanding Yu, Kaibin Huang, Dongning Guo · 2020
This paper proposes a novel scheduling policy for federated edge learning, which exploits both diversity in multiuser channels and diversity in the "importance" of the edge devices’ learning updates. A probabilistic scheduling framework is first developed to yield unbiased update aggregation in federated edge learning. The importance of a local learning update is measured by its gradient divergence. Considering the tradeoff between channel quality and update importance, the optimal scheduling policy is developed in closed form. The convergence analysis is also provided. Numerical results demonstrate the effectiveness of the proposed scheduling policy as compared with some benchmark policies.