Client Selection for Federated Learning with Heterogeneous Resources in Mobile Edge

TLDR

Mobile edge computing enables privacy‑preserving machine learning by training models on distributed client data and computation resources, but heterogeneous client capabilities and poor wireless links can make federated learning inefficient. This work introduces FedCS, a federated learning protocol designed to efficiently train models in practical cellular networks by adapting to heterogeneous client resources. FedCS formulates client selection as a resource‑constrained optimization that chooses the maximum number of clients per round based on their computation and upload capacities, and is evaluated on large‑scale image datasets in MEC simulations. Experiments show that FedCS reduces training time substantially compared to the standard federated learning protocol.

Abstract

We envision a mobile edge computing (MEC) framework for machine learning (ML) technologies, which leverages distributed client data and computation resources for training high-performance ML models while preserving client privacy. Toward this future goal, this work aims to extend Federated Learning (FL), a decentralized learning framework that enables privacy-preserving training of models, to work with heterogeneous clients in a practical cellular network. The FL protocol iteratively asks random clients to download a trainable model from a server, update it with own data, and upload the updated model to the server, while asking the server to aggregate multiple client updates to further improve the model. While clients in this protocol are free from disclosing own private data, the overall training process can become inefficient when some clients are with limited computational resources (i.e., requiring longer update time) or under poor wireless channel conditions (longer upload time). Our new FL protocol, which we refer to as FedCS, mitigates this problem and performs FL efficiently while actively managing clients based on their resource conditions. Specifically, FedCS solves a client selection problem with resource constraints, which allows the server to aggregate as many client updates as possible and to accelerate performance improvement in ML models. We conducted an experimental evaluation using publicly-available large-scale image datasets to train deep neural networks on MEC environment simulations. The experimental results show that FedCS is able to complete its training process in a significantly shorter time compared to the original FL protocol.