Bridging Model Heterogeneity in Federated Learning via Uncertainty-based Asymmetrical Reciprocity Learning

TL;DR

This paper introduces FedType, a federated learning framework that uses small proxy models and uncertainty-based learning to effectively handle heterogeneous models, enhance privacy, and reduce communication without public data.

Contribution

FedType is the first framework to bridge model heterogeneity in FL using proxy models and uncertainty-based asymmetrical reciprocity learning without public data.

Findings

Effective across diverse benchmark datasets

Reduces communication costs significantly

Maintains high model accuracy and privacy

Abstract

This paper presents FedType, a simple yet pioneering framework designed to fill research gaps in heterogeneous model aggregation within federated learning (FL). FedType introduces small identical proxy models for clients, serving as agents for information exchange, ensuring model security, and achieving efficient communication simultaneously. To transfer knowledge between large private and small proxy models on clients, we propose a novel uncertainty-based asymmetrical reciprocity learning method, eliminating the need for any public data. Comprehensive experiments conducted on benchmark datasets demonstrate the efficacy and generalization ability of FedType across diverse settings. Our approach redefines federated learning paradigms by bridging model heterogeneity, eliminating reliance on public data, prioritizing client privacy, and reducing communication costs.