FedRolex: Model-Heterogeneous Federated Learning with Rolling Sub-Model Extraction

TL;DR

FedRolex introduces a novel model-heterogeneous federated learning approach that enables training larger global models than individual clients by employing a rolling sub-model extraction scheme, improving inclusiveness and performance.

Contribution

The paper proposes FedRolex, a partial training method with rolling sub-model extraction, allowing model-heterogeneous FL and training larger global models than client models.

Findings

FedRolex outperforms state-of-the-art PT-based methods like Federated Dropout.

It reduces the gap between model-heterogeneous and homogeneous FL.

FedRolex enhances inclusiveness and benefits low-end devices.

Abstract

Most cross-device federated learning (FL) studies focus on the model-homogeneous setting where the global server model and local client models are identical. However, such constraint not only excludes low-end clients who would otherwise make unique contributions to model training but also restrains clients from training large models due to on-device resource bottlenecks. In this work, we propose FedRolex, a partial training (PT)-based approach that enables model-heterogeneous FL and can train a global server model larger than the largest client model. At its core, FedRolex employs a rolling sub-model extraction scheme that allows different parts of the global server model to be evenly trained, which mitigates the client drift induced by the inconsistency between individual client models and server model architectures. We show that FedRolex outperforms state-of-the-art PT-based model-heterogeneous FL methods (e.g. Federated Dropout) and reduces the gap between model-heterogeneous and model-homogeneous FL, especially under the large-model large-dataset regime. In addition, we provide theoretical statistical analysis on its advantage over Federated Dropout and evaluate FedRolex on an emulated real-world device distribution to show that FedRolex can enhance the inclusiveness of FL and boost the performance of low-end devices that would otherwise not benefit from FL. Our code is available at: https://github.com/AIoT-MLSys-Lab/FedRolex