FedAC: An Adaptive Clustered Federated Learning Framework for Heterogeneous Data

TL;DR

FedAC is an adaptive clustered federated learning framework that improves performance on heterogeneous data by integrating global knowledge, using an online model similarity metric, and dynamically tuning cluster numbers, leading to higher accuracy.

Contribution

It introduces a novel adaptive CFL framework that effectively combines global and intra-cluster knowledge, employs a cost-effective similarity metric, and dynamically adjusts cluster counts for better scalability.

Findings

Achieves around 1.82% and 12.67% higher accuracy on CIFAR-10 and CIFAR-100.

Enhances performance by integrating global knowledge with intra-cluster learning.

Demonstrates improved adaptability and scalability in heterogeneous environments.

Abstract

Clustered federated learning (CFL) is proposed to mitigate the performance deterioration stemming from data heterogeneity in federated learning (FL) by grouping similar clients for cluster-wise model training. However, current CFL methods struggle due to inadequate integration of global and intra-cluster knowledge and the absence of an efficient online model similarity metric, while treating the cluster count as a fixed hyperparameter limits flexibility and robustness. In this paper, we propose an adaptive CFL framework, named FedAC, which (1) efficiently integrates global knowledge into intra-cluster learning by decoupling neural networks and utilizing distinct aggregation methods for each submodule, significantly enhancing performance; (2) includes a costeffective online model similarity metric based on dimensionality reduction; (3) incorporates a cluster number fine-tuning module for improved adaptability and scalability in complex, heterogeneous environments. Extensive experiments show that FedAC achieves superior empirical performance, increasing the test accuracy by around 1.82% and 12.67% on CIFAR-10 and CIFAR-100 datasets, respectively, under different non-IID settings compared to SOTA methods.