FedD2S: Personalized Data-Free Federated Knowledge Distillation

TL;DR

FedD2S introduces a novel personalized federated learning method using data-free knowledge distillation with layer dropping, improving client personalization, convergence speed, and fairness across diverse datasets.

Contribution

The paper proposes FedD2S, a new personalized FL approach that employs deep-to-shallow layer dropping in data-free knowledge distillation for better personalization.

Findings

FedD2S outperforms state-of-the-art FL baselines in convergence speed.

Layer dropping enhances personalized knowledge capture.

Hyperparameter tuning improves FedD2S performance.

Abstract

This paper addresses the challenge of mitigating data heterogeneity among clients within a Federated Learning (FL) framework. The model-drift issue, arising from the noniid nature of client data, often results in suboptimal personalization of a global model compared to locally trained models for each client. To tackle this challenge, we propose a novel approach named FedD2S for Personalized Federated Learning (pFL), leveraging knowledge distillation. FedD2S incorporates a deep-to-shallow layer-dropping mechanism in the data-free knowledge distillation process to enhance local model personalization. Through extensive simulations on diverse image datasets-FEMNIST, CIFAR10, CINIC0, and CIFAR100-we compare FedD2S with state-of-the-art FL baselines. The proposed approach demonstrates superior performance, characterized by accelerated convergence and improved fairness among clients. The introduced layer-dropping technique effectively captures personalized knowledge, resulting in enhanced performance compared to alternative FL models. Moreover, we investigate the impact of key hyperparameters, such as the participation ratio and layer-dropping rate, providing valuable insights into the optimal configuration for FedD2S. The findings demonstrate the efficacy of adaptive layer-dropping in the knowledge distillation process to achieve enhanced personalization and performance across diverse datasets and tasks.