Mechanism Design for Federated Learning with Non-Monotonic Network Effects

TL;DR

The paper tackles incentive design for federated learning (FL) under application-specific generalization requirements, revealing non-monotonic network effects in participant contributions. It introduces the Model Trading and Sharing (MoTS) framework, which allows participants to either join FL training or purchase trained models, and develops the SWAN mechanism to internalize network effects via pricing and rewards that depend on heterogeneity and error constraints. Theoretical results characterize when network effects are positive or negative and show how optimal social states can be achieved under MoTS, with budget-balanced operation when welfare is positive. Empirical evaluation on a hardware FL prototype demonstrates substantial gains over existing FL mechanisms, including up to 352.42% improvements in social welfare and up to 93.07% reductions in incentive costs across convex and non-convex, i.i.d. and non-i.i.d. data settings, validating the framework's practical impact.

Abstract

Mechanism design is pivotal to federated learning (FL) for maximizing social welfare by coordinating self-interested clients. Existing mechanisms, however, often overlook the network effects of client participation and the diverse model performance requirements (i.e., generalization error) across applications, leading to suboptimal incentives and social welfare, or even inapplicability in real deployments. To address this gap, we explore incentive mechanism design for FL with network effects and application-specific requirements of model performance. We develop a theoretical model to quantify the impact of network effects on heterogeneous client participation, revealing the non-monotonic nature of such effects. Based on these insights, we propose a Model Trading and Sharing (MoTS) framework, which enables clients to obtain FL models through either participation or purchase. To further address clients' strategic behaviors, we design a Social Welfare maximization with Application-aware and Network effects (SWAN) mechanism, exploiting model customer payments for incentivization. Experimental results on a hardware prototype demonstrate that our SWAN mechanism outperforms existing FL mechanisms, improving social welfare by up to $352.42\%$ and reducing extra incentive costs by $93.07\%$.

Figures (9)

• Figure 1: Network Effects of Client Participation in FL.
• Figure 2: The Model Trading and Sharing (MoTS) Framework.
• Figure 3: Network Effects of Type-$j$ Client Participation.
• Figure 4: Hardware Prototype of FL System with 20 Raspberry Pi Clients.
• Figure 5: Network Effects of Client Participation: Empirical Results (Blue) versus Theoretical Predictions (Red).