Federated Learning for Energy-balanced Client Selection in Mobile Edge Computing
Ref: CISTER-TR-210305 Publication Date: 28, Jun to 2, Jul, 2021
Federated Learning for Energy-balanced Client Selection in Mobile Edge Computing
Ref: CISTER-TR-210305 Publication Date: 28, Jun to 2, Jul, 2021Abstract:
Mobile edge computing (MEC) has been considered
as a promising technology to provide seamless integration of
multiple application services. Federated learning (FL) is carried
out at edge clients in MEC for privacy-preserving training of
data processing models. Despite that the edge clients with small
data payloads consume less energy on FL training, the small data
payload gives rise to a low learning accuracy due to insufficient
input to the FL training. Inadequate selection of the edge clients
can result in a large energy consumption at the edge clients,
or a low learning accuracy of the FL training. In this paper,
a new FL-based client selection optimization is proposed to
balance the trade-off between energy consumption of the edge
clients and the learning accuracy of FL. We first show that
this optimization problem is NP-complete. Next, we propose
a FL-based energy-accuracy balancing heuristic algorithm to
approximate the optimal client selection in polynomial time. The
numerical results show the advantage of our proposed algorithm.
Jingjing Zheng
Kai Li
Eduardo Tovar
Mohsen GuizaniEvents:
17th International Wireless Communications & Mobile Computing Conference (IWCMC 2021).
Harbin, China.
Notes: Jingjing Zheng, Kai Li, Eduardo Tovar, Mohsen Guizani
Record Date: 30, Mar, 2021