PM-MOE: Mixture of Experts on Private Model Parameters for Personalized Federated Learning

TL;DR

The paper introduces PM-MoE, a novel architecture for personalized federated learning that leverages a mixture of expert modules and energy-based denoising to improve model personalization across diverse data domains.

Contribution

It proposes the PM-MoE architecture that enhances personalized federated learning by integrating expert modules and denoising, improving performance with minimal additional training.

Findings

Significant performance improvements on six datasets.

Effective across nine model-split-based algorithms.

Validated under two heterogeneity settings.

Abstract

Federated learning (FL) has gained widespread attention for its privacy-preserving and collaborative learning capabilities. Due to significant statistical heterogeneity, traditional FL struggles to generalize a shared model across diverse data domains. Personalized federated learning addresses this issue by dividing the model into a globally shared part and a locally private part, with the local model correcting representation biases introduced by the global model. Nevertheless, locally converged parameters more accurately capture domain-specific knowledge, and current methods overlook the potential benefits of these parameters. To address these limitations, we propose PM-MoE architecture. This architecture integrates a mixture of personalized modules and an energy-based personalized modules denoising, enabling each client to select beneficial personalized parameters from other clients. We applied the PM-MoE architecture to nine recent model-split-based personalized federated learning algorithms, achieving performance improvements with minimal additional training. Extensive experiments on six widely adopted datasets and two heterogeneity settings validate the effectiveness of our approach. The source code is available at \url{https://github.com/dannis97500/PM-MOE}.