MAR-FL: A Communication Efficient Peer-to-Peer Federated Learning System

2025-12-04

AI Summary

To address the high communication overhead and poor robustness of federated learning (FL) in wireless environments with dynamic topologies and large-scale node populations, this paper proposes a decentralized peer-to-peer (P2P) FL framework. The core method introduces a hierarchical group-based iterative aggregation mechanism: nodes are organized into multi-level groups, enabling local intra-group aggregation followed by progressive inter-group synchronization—reducing communication complexity from conventional O(N²) to O(N log N). Additionally, the framework incorporates a lightweight communication protocol, fault-tolerant model update mechanisms, and privacy-preserving computation interfaces. Experimental results demonstrate that the proposed framework significantly reduces communication load in highly dynamic networks, improves training efficiency and system scalability, and maintains model convergence and accuracy stability even under frequent node join/leave events.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Multiagent Systems: Agent Communication
• Search and Optimization: Distributed Search

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Federated Web and WoT systems, including distributed, federated and edge-based data processing
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

The convergence of next-generation wireless systems and distributed Machine Learning (ML) demands Federated Learning (FL) methods that remain efficient and robust with wireless connected peers and under network churn. Peer-to-peer (P2P) FL removes the bottleneck of a central coordinator, but existing approaches suffer from excessive communication complexity, limiting their scalability in practice. We introduce MAR-FL, a novel P2P FL system that leverages iterative group-based aggregation to substantially reduce communication overhead while retaining resilience to churn. MAR-FL achieves communication costs that scale as O(N log N), contrasting with the O(N^2) complexity of previously existing baselines, and thereby maintains effectiveness especially as the number of peers in an aggregation round grows. The system is robust towards unreliable FL clients and can integrate private computing.

Problem

• Reduces communication overhead in peer-to-peer federated learning
• Enhances scalability by lowering complexity from O(N^2) to O(N log N)
• Maintains resilience to network churn and unreliable clients

Innovation

• Peer-to-peer federated learning system
• Iterative group-based aggregation reduces overhead
• Communication scales O(N log N) not O(N^2)