Decentralised Federated Learning over Temporal Networks: The Role of Heterogeneities

2026-07-03

AI Summary

This study investigates how the structural and temporal heterogeneity of communication networks affects model convergence in decentralized federated learning. By establishing a theoretical equivalence between local model averaging aggregation and the dynamics of lazy random walks on temporal networks, the work reveals, for the first time, the precise mapping mechanism underlying these two processes. Integrating temporal network modeling, diffusion theory, and empirical analysis of real-world time-varying communication graphs, the authors demonstrate that neglecting communication heterogeneity leads to a substantial overestimation of convergence rates. In practice, such heterogeneity in real networks typically slows convergence considerably, often producing misleadingly optimistic acceleration artifacts under standard experimental assumptions.

Abstract

Decentralised federated learning, based on peer-to-peer communication, is increasingly proposed for on-device training of machine learning models, promising a privacy-preserving, communication-efficient training process with no risk of single-point failure. However, the role of structural and temporal inhomogeneities in such fully decentralised settings remains poorly understood. Here, we investigate their effects when model parameters are locally averaged during aggregation. We show that the decentralised federated learning process is governed, both in the early phase and the late, stationary limit, by the same dynamics as a lazy random-walk diffusion process on temporal networks. Based on this mapping, we demonstrate that the typical experimental scenario used in decentralised federated learning leads to unrealistically rapid convergence because of ignoring the temporal and structural inhomogeneities inherent in the communication network. We analyse real-world temporal networks and find that inhomogeneities most often dramatically slow down diffusion, hence the convergence process.

Problem

• Decentralised Federated Learning
• Temporal Networks
• Heterogeneities
• Convergence
• Diffusion

Innovation

• Decentralised Federated Learning
• Temporal Networks
• Structural Heterogeneity
• Lazy Random Walk
• Convergence Dynamics