Wireless Decentralized Federated Learning via Device Clustering and Inter-Cluster Link Enhancement

2026-07-08

AI Summary

This work addresses the challenges of slow model aggregation and poor convergence in wireless decentralized federated learning (DFL), which stem from uncoordinated devices, high communication latency, and straggler nodes. To overcome these limitations, the authors propose a budget-aware two-tier aggregation framework: first, devices are clustered into densely connected groups where cluster heads perform fast intra-cluster parallel aggregation; second, a small number of reliable backhaul links are strategically deployed among cluster heads to enable infrequent global synchronization. By jointly optimizing network clustering, cluster head selection, and constrained backhaul allocation, the proposed method significantly accelerates model consensus in large-scale device-to-device (D2D) networks under limited backhaul budgets, achieving an O(1/t) convergence rate. Experiments on image classification tasks demonstrate that with only a few carefully placed backhaul links, the approach substantially outperforms existing DFL methods in convergence efficiency.

Abstract

Decentralized federated learning (DFL) dispenses with the central server of classical FL by utilizing peer-to-peer model exchanges among edge devices. This server-free architecture enables ad-hoc, flexible distributed learning in large device-to-device (D2D) networks. However, wireless DFL converges slowly because peer-to-peer model aggregation incurs high delays and errors. Each DFL training round involves many-to-many gradient sharing over wireless channels, resulting in uncoordinated channel access, large communication errors from stragglers, and slow model consensus, especially in large-scale D2D networks with pronounced clustering structures. We address these aggregation bottlenecks by provisioning a few reliable backhaul links at straggling nodes to enhance network connectivity. Building on this idea, our budget-aware, cluster-centric DFL framework first partitions the network into densely connected clusters, and then allocates the limited backhaul budget to selected cluster heads. The resulting two-tier protocol executes fast, parallel model aggregation within clusters and infrequent inter-cluster exchanges among the heads, yielding an O(1/t) convergence rate in t iterations. Numerical experiments on image-classification tasks confirm that our approach accelerates convergence compared to state-of-the-art DFL baselines with only a few strategically placed backhaul links.

Problem

• Decentralized Federated Learning
• Wireless Communication
• Convergence Rate
• Device Clustering
• Straggler Nodes

Innovation

• Decentralized Federated Learning
• Device Clustering
• Backhaul Enhancement
• Wireless D2D Networks
• Convergence Acceleration