Convergence Acceleration in Wireless Federated Learning: A Stackelberg Game Approach
Kaidi Wang, Yi Ma, Mahdi Boloursaz Mashhadi, Chuan Heng Foh, Rahim Tafazolli, Zhi Ding
TL;DR
The paper tackles convergence-time issues in wireless federated learning by casting joint global loss and latency minimization as a Stackelberg game between a leader (global loss) and a follower (latency). It introduces AoU-based device selection to prioritize informative updates and decouples the follower into resource/power optimization via monotonic optimization and a matching-based sub-channel assignment, with a derived convergence-rate upper bound showing the benefit of selecting devices with large data sizes. The proposed MO-RA and M-SA algorithms demonstrate improved convergence and efficient sub-channel usage on MNIST, CIFAR-10, and SST-2 datasets, under energy constraints. Overall, the framework provides a practical, energetically aware method to balance convergence speed and latency in FLOWN, adaptable to channel variations and hardware budgets.
Abstract
This paper studies issues that arise with respect to the joint optimization for convergence time in federated learning over wireless networks (FLOWN). We consider the criterion and protocol for selection of participating devices in FLOWN under the energy constraint and derive its impact on device selection. In order to improve the training efficiency, age-of-information (AoI) enables FLOWN to assess the freshness of gradient updates among participants. Aiming to speed up convergence, we jointly investigate global loss minimization and latency minimization in a Stackelberg game based framework. Specifically, we formulate global loss minimization as a leader-level problem for reducing the number of required rounds, and latency minimization as a follower-level problem to reduce time consumption of each round. By decoupling the follower-level problem into two sub-problems, including resource allocation and sub-channel assignment, we achieve an optimal strategy of the follower through monotonic optimization and matching theory. At the leader-level, we derive an upper bound of convergence rate and subsequently reformulate the global loss minimization problem and propose a new age-of-update (AoU) based device selection algorithm. Simulation results indicate the superior performance of the proposed AoU based device selection scheme in terms of the convergence rate, as well as efficient utilization of available sub-channels.