Answers are grounded in this paper and cite their sources.
Hierarchical federated learning (HFL) incorporates edge aggregation into federated learning (FL), thereby reducing communication overhead and improving scalability in wireless edge networks. Incentive design in HFL must account for how payments across the cloud–edge–device hierarchy affect the data contributions of participants with private costs. Moreover, participation in HFL consumes local device resources, making effective incentives essential for motivating self-interested devices to participate. To address these issues, we propose a two-tier incentive mechanism for HFL, namely TIM-HFL, in which the lower layer addresses information asymmetry between access points (APs) and devices through a contract-theoretic mechanism and derives closed-form expressions for the data contributions optimized under the proposed reward rule and the corresponding contract rewards, with the resulting contract satisfying individual rationality and incentive compatibility constraints. At the upper layer, we formulate a Stackelberg game between the server and APs and derive a closed-form expression for the optimal reward coefficient, thereby enabling coordinated cross-layer optimization without requiring iterative alternating optimization between the two incentive layers within each global training round. Simulation results demonstrate that TIM-HFL improves model performance and system utility under both independent and identically distributed (IID) and non-IID settings, while showing consistency between the analytical solutions and numerical optimization results.
No takes yet. Share an insight, caveat, or question.
Xu et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: