Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Fedcompass: Federated Clustered and Periodic Aggregation Framework for Hybrid Classical-Quantum Models

Yueheng Wang, Xing He, Zinuo Cai, Rui Zhang, Ruhui Ma, Yuan Liu, Rajkumar Buyya

TL;DR

Federated learning faces privacy-preserving training but non-IID data harms convergence, especially when integrating quantum components. The authors propose FEDCOMPASS, a layered aggregation framework that performs cluster-based aggregation of classical feature extractors via spectral clustering, and uses circular mean aggregation plus adaptive optimization for the periodic quantum classifier. On MNIST, Fashion-MNIST, and CIFAR-10 with non-IID splits, FEDCOMPASS yields up to 10.22 percentage-point gains over FedAvg and improves convergence stability across settings. This approach demonstrates a practical path to efficient, privacy-preserving hybrid classical-quantum federated learning under data heterogeneity.

Abstract

Federated learning enables collaborative model training across decentralized clients under privacy constraints. Quantum computing offers potential for alleviating computational and communication burdens in federated learning, yet hybrid classical-quantum federated learning remains susceptible to performance degradation under non-IID data. To address this,we propose FEDCOMPASS, a layered aggregation framework for hybrid classical-quantum federated learning. FEDCOMPASS employs spectral clustering to group clients by class distribution similarity and performs cluster-wise aggregation for classical feature extractors. For quantum parameters, it uses circular mean aggregation combined with adaptive optimization to ensure stable global updates. Experiments on three benchmark datasets show that FEDCOMPASS improves test accuracy by up to 10.22% and enhances convergence stability under non-IID settings, outperforming six strong federated learning baselines.

Fedcompass: Federated Clustered and Periodic Aggregation Framework for Hybrid Classical-Quantum Models

TL;DR

Federated learning faces privacy-preserving training but non-IID data harms convergence, especially when integrating quantum components. The authors propose FEDCOMPASS, a layered aggregation framework that performs cluster-based aggregation of classical feature extractors via spectral clustering, and uses circular mean aggregation plus adaptive optimization for the periodic quantum classifier. On MNIST, Fashion-MNIST, and CIFAR-10 with non-IID splits, FEDCOMPASS yields up to 10.22 percentage-point gains over FedAvg and improves convergence stability across settings. This approach demonstrates a practical path to efficient, privacy-preserving hybrid classical-quantum federated learning under data heterogeneity.

Abstract

Federated learning enables collaborative model training across decentralized clients under privacy constraints. Quantum computing offers potential for alleviating computational and communication burdens in federated learning, yet hybrid classical-quantum federated learning remains susceptible to performance degradation under non-IID data. To address this,we propose FEDCOMPASS, a layered aggregation framework for hybrid classical-quantum federated learning. FEDCOMPASS employs spectral clustering to group clients by class distribution similarity and performs cluster-wise aggregation for classical feature extractors. For quantum parameters, it uses circular mean aggregation combined with adaptive optimization to ensure stable global updates. Experiments on three benchmark datasets show that FEDCOMPASS improves test accuracy by up to 10.22% and enhances convergence stability under non-IID settings, outperforming six strong federated learning baselines.
Paper Structure (9 sections, 6 equations, 4 figures, 2 tables, 1 algorithm)

This paper contains 9 sections, 6 equations, 4 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. The Proposed fedcompass Algorithm
  3. Overview
  4. Cluster-Based Aggregation for Classical Feature Extraction Layer
  5. Global Aggregation Optimization for Periodic Quantum Classifier
  6. PERFORMANCE EVALUATION
  7. Experiments Setup
  8. Results and Discussion
  9. Conclusion

Figures (4)

  • Figure 1: Overview of our mechanism.
  • Figure 2: Convergence curves of test accuracy versus communication rounds on MNIST under non-IID settings.
  • Figure 3: Convergence curves of test accuracy versus communication rounds on Fashion-MNIST under non-IID settings.
  • Figure 4: Convergence curves of test accuracy versus communication rounds on CIFAR-10 under non-IID settings.