FLrce: Resource-Efficient Federated Learning with Early-Stopping Strategy

TL;DR

This paper tackles resource scarcity and client heterogeneity in IoT federated learning by introducing FLrce, a framework that combines relationship-based client selection with an early-stopping mechanism. The method estimates client importance through a relationship map, using synchronous cosine similarity and asynchronous orthogonal-distance modeling to drive selective participation, while stopping training when conflicts among updates threaten efficiency and convergence. Empirical results on EMNIST, Google Speech, CIFAR-10, and CIFAR-100 across 100 simulated clients show FLrce achieves higher final accuracy with significantly lower computation and communication costs (average gains of at least 30% and 43%, respectively) compared with baselines. The work demonstrates practical gains for resource-constrained IoT deployments and outlines avenues for extending FLrce with model pruning and theoretical analyses.

Abstract

Federated Learning (FL) achieves great popularity in the Internet of Things (IoT) as a powerful interface to offer intelligent services to customers while maintaining data privacy. Under the orchestration of a server, edge devices (also called clients in FL) collaboratively train a global deep-learning model without sharing any local data. Nevertheless, the unequal training contributions among clients have made FL vulnerable, as clients with heavily biased datasets can easily compromise FL by sending malicious or heavily biased parameter updates. Furthermore, the resource shortage issue of the network also becomes a bottleneck. Due to overwhelming computation overheads generated by training deep-learning models on edge devices, and significant communication overheads for transmitting deep-learning models across the network, enormous amounts of resources are consumed in the FL process. This encompasses computation resources like energy and communication resources like bandwidth. To comprehensively address these challenges, in this paper, we present FLrce, an efficient FL framework with a relationship-based client selection and early-stopping strategy. FLrce accelerates the FL process by selecting clients with more significant effects, enabling the global model to converge to a high accuracy in fewer rounds. FLrce also leverages an early stopping mechanism that terminates FL in advance to save communication and computation resources. Experiment results show that, compared with existing efficient FL frameworks, FLrce improves the computation and communication efficiency by at least 30% and 43% respectively.

Figures (18)

• Figure 1: FL faces the dilemma of expensive computation and communication costs in IoT networks.
• Figure 2: A simple 2-dimensional demonstration of unequal training contributions among FL clients.
• Figure 3: Existing efficient FL methods easily become vulnerable to non-iid client data.
• Figure 4: Due to the vulnerability to heterogeneous client contributions, existing efficient FL methods consume more communication and computation resources to reach the same target accuracy in the non-iid case.
• Figure 5: Overview of the FLrce framework.