RefProtoFL: Communication-Efficient Federated Learning via External-Referenced Prototype Alignment

TL;DR

RefProtoFL introduces a communication-efficient federated learning framework that uses external-referenced prototype alignment and adaptive update dropping to improve model generalization and reduce communication costs across heterogeneous clients.

Contribution

It proposes a novel FL framework combining external-referenced prototype alignment with adaptive probabilistic update dropping, decomposing models into private backbones and lightweight shared adapters.

Findings

Achieves higher accuracy than state-of-the-art prototype-based FL methods.

Reduces communication cost significantly through adaptive update sparsification.

Effectively handles data heterogeneity with external reference prototypes.

Abstract

Federated learning (FL) enables collaborative model training without sharing raw data in edge environments, but is constrained by limited communication bandwidth and heterogeneous client data distributions. Prototype-based FL mitigates this issue by exchanging class-wise feature prototypes instead of full model parameters; however, existing methods still suffer from suboptimal generalization under severe communication constraints. In this paper, we propose RefProtoFL, a communication-efficient FL framework that integrates External-Referenced Prototype Alignment (ERPA) for representation consistency with Adaptive Probabilistic Update Dropping (APUD) for communication efficiency. Specifically, we decompose the model into a private backbone and a lightweight shared adapter, and restrict federated communication to the adapter parameters only. To further reduce uplink cost, APUD performs magnitude-aware Top-K sparsification, transmitting only the most significant adapter updates for server-side aggregation. To address representation inconsistency across heterogeneous clients, ERPA leverages a small server-held public dataset to construct external reference prototypes that serve as shared semantic anchors. For classes covered by public data, clients directly align local representations to public-induced prototypes, whereas for uncovered classes, alignment relies on server-aggregated global reference prototypes via weighted averaging. Extensive experiments on standard benchmarks demonstrate that RefProtoFL attains higher classification accuracy than state-of-the-art prototype-based FL baselines.