HASFL: Heterogeneity-aware Split Federated Learning over Edge Computing Systems

TL;DR

HASFL introduces a heterogeneity-aware split federated learning framework that adaptively manages batch sizes and model splitting to improve training efficiency and convergence on diverse edge devices.

Contribution

It provides a novel adaptive control mechanism for batch sizes and model splitting in SFL, backed by a convergence bound analysis for heterogeneous edge environments.

Findings

HASFL outperforms existing methods in training speed and accuracy.

Adaptive control of batch sizes and model splitting reduces straggler effects.

Experimental results validate the framework's effectiveness across datasets.

Abstract

Split federated learning (SFL) has emerged as a promising paradigm to democratize machine learning (ML) on edge devices by enabling layer-wise model partitioning. However, existing SFL approaches suffer significantly from the straggler effect due to the heterogeneous capabilities of edge devices. To address the fundamental challenge, we propose adaptively controlling batch sizes (BSs) and model splitting (MS) for edge devices to overcome resource heterogeneity. We first derive a tight convergence bound of SFL that quantifies the impact of varied BSs and MS on learning performance. Based on the convergence bound, we propose HASFL, a heterogeneity-aware SFL framework capable of adaptively controlling BS and MS to balance communication-computing latency and training convergence in heterogeneous edge networks. Extensive experiments with various datasets validate the effectiveness of HASFL and demonstrate its superiority over state-of-the-art benchmarks.