Knowledge-Aware Parameter Coaching for Communication-Efficient Personalized Federated Learning in Mobile Edge Computing

Personalized Federated Learning (pFL) can improve the accuracy of local models and provide enhanced edge intelligence without exposing the raw data in Mobile Edge Computing (MEC). However, in the MEC environment with constrained communication resources, transmitting the entire model between the server and the clients in traditional pFL methods imposes substantial communication overhead, which can lead to inaccurate personalization and degraded performance of mobile clients. In response, we propose a Communication-Efficient pFL architecture to enhance the performance of personalized models while minimizing communication overhead in MEC. First, a Knowledge-Aware Parameter Coaching method (KAPC) is presented to produce a more accurate personalized model by utilizing the layer-wise parameters of other clients with adaptive aggregation weights. Then, convergence analysis of the proposed KAPC is developed in both the convex and non-convex settings. Second, a Bidirectional Layer Selection algorithm (BLS) based on self-relationship and generalization error is proposed to select the most informative layers for transmission, which reduces communication costs. Extensive experiments are conducted, and the results demonstrate that the proposed KAPC achieves superior accuracy compared to the state-of-the-art baselines, while the proposed BLS substantially improves resource utilization without sacrificing performance.

I. Introduction

In Mobile Edge Computing (MEC), an edge server instead of a cloud server can centrally explore valuable data residing on mobile clients for model training, which can reduce latency and bandwidth usage, and enhance overall system performance [1], [2]. However, transmitting raw data from mobile clients to edge servers not only consumes substantial communication resources but also introduces the risk of data leakage. Federated Learning (FL) has been emerging as a secure and efficient distributed machine learning architecture, which can train a shared global model by aggregating the local model of each client [3]. By avoiding raw data exchange, FL can prevent the exposure of sensitive client information and reduce communication overhead. Consequently, FL has gained widespread adoption in MEC, providing edge intelligence in healthcare, intelligent traffic systems, and industrial engineering [4].