PFL-MoE: Personalized Federated Learning Based on Mixture of Experts

TL;DR

This paper introduces PFL-MoE, a novel personalized federated learning approach that combines personalized and global models using Mixture of Experts architecture to improve performance on non-IID data.

Contribution

It proposes a flexible PFL-MoE framework that integrates existing PFL algorithms and enhances model personalization while preserving generalization.

Findings

PFL-MoE improves accuracy on non-IID datasets.

The PFL-MFE variant enhances gating network decision-making.

Experiments on Fashion-MNIST and CIFAR-10 validate effectiveness.

Abstract

Federated learning (FL) is an emerging distributed machine learning paradigm that avoids data sharing among training nodes so as to protect data privacy. Under coordination of the FL server, each client conducts model training using its own computing resource and private data set. The global model can be created by aggregating the training results of clients. To cope with highly non-IID data distributions, personalized federated learning (PFL) has been proposed to improve overall performance by allowing each client to learn a personalized model. However, one major drawback of a personalized model is the loss of generalization. To achieve model personalization while maintaining generalization, in this paper, we propose a new approach, named PFL-MoE, which mixes outputs of the personalized model and global model via the MoE architecture. PFL-MoE is a generic approach and can be instantiated by integrating existing PFL algorithms. Particularly, we propose the PFL-MF algorithm which is an instance of PFL-MoE based on the freeze-base PFL algorithm. We further improve PFL-MF by enhancing the decision-making ability of MoE gating network and propose a variant algorithm PFL-MFE. We demonstrate the effectiveness of PFL-MoE by training the LeNet-5 and VGG-16 models on the Fashion-MNIST and CIFAR-10 datasets with non-IID partitions.