Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

Hongyue Wu, Hangyu Li, Guodong Fan, Haoran Zhu, Shizhan Chen, Zhiyong Feng

TL;DR

RefProtoFL tackles the dual challenges of communication bottlenecks and non-IID data in federated learning by integrating External-Referenced Prototype Alignment (ERPA) with Adaptive Probabilistic Update Dropping (APUD). It preserves private backbones on clients while learning a lightweight shared adapter, and uses a server-held public dataset to anchor class prototypes across clients, mitigating prototype drift. APUD further reduces uplink traffic by sparsifying adapter updates with a Top-K strategy and size-weighted aggregation, maintaining convergence. Empirical results on standard benchmarks show RefProtoFL achieving higher accuracy than prototype-based baselines under severe heterogeneity while lowering communication costs, demonstrating practical gains for edge environments.

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.

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

TL;DR

RefProtoFL tackles the dual challenges of communication bottlenecks and non-IID data in federated learning by integrating External-Referenced Prototype Alignment (ERPA) with Adaptive Probabilistic Update Dropping (APUD). It preserves private backbones on clients while learning a lightweight shared adapter, and uses a server-held public dataset to anchor class prototypes across clients, mitigating prototype drift. APUD further reduces uplink traffic by sparsifying adapter updates with a Top-K strategy and size-weighted aggregation, maintaining convergence. Empirical results on standard benchmarks show RefProtoFL achieving higher accuracy than prototype-based baselines under severe heterogeneity while lowering communication costs, demonstrating practical gains for edge environments.

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.
Paper Structure (31 sections, 19 equations, 2 figures, 2 tables, 6 algorithms)

This paper contains 31 sections, 19 equations, 2 figures, 2 tables, 6 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Efficient federated optimization
  4. Prototype-based federated learning
  5. External references for federated learning
  6. Problem Statement
  7. Methodology
  8. Adaptive Probabilistic Update Dropping
  9. Update Magnitude Estimation
  10. Top-K Parameter Selection
  11. Federated Aggregation
  12. External-Referenced Prototype Alignment
  13. External-reference prototypes
  14. Local prototypes for classes missing in public data
  15. Federated aggregation of prototypes
  16. ...and 16 more sections

Figures (2)

  • Figure 1: Visualization of data partitioning in the CIFAR10 dataset.
  • Figure 2: Two-dimensional t-SNE visualization of the learned prototypes of FL methods on the FashionMNIST test set with $\alpha = 0.5$, where each point represents a sample and colors denote class labels.