Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Rethinking Clustered Federated Learning in NOMA Enhanced Wireless Networks

Yushen Lin, Kaidi Wang, Zhiguo Ding

TL;DR

The extensive simulation results show that the proposed cluster-based FL framework can outperform FL baselines in terms of both test accuracy and convergence rate, and jointly optimizing sub-channel and power allocation in NOMA-enhanced networks can lead to a significant improvement.

Abstract

This study explores the benefits of integrating the novel clustered federated learning (CFL) approach with non-orthogonal multiple access (NOMA) under non-independent and identically distributed (non-IID) datasets, where multiple devices participate in the aggregation with time limitations and a finite number of sub-channels. A detailed theoretical analysis of the generalization gap that measures the degree of non-IID in the data distribution is presented. Following that, solutions to address the challenges posed by non-IID conditions are proposed with the analysis of the properties. Specifically, users' data distributions are parameterized as concentration parameters and grouped using spectral clustering, with Dirichlet distribution serving as the prior. The investigation into the generalization gap and convergence rate guides the design of sub-channel assignments through the matching-based algorithm, and the power allocation is achieved by Karush-Kuhn-Tucker (KKT) conditions with the derived closed-form solution. The extensive simulation results show that the proposed cluster-based FL framework can outperform FL baselines in terms of both test accuracy and convergence rate. Moreover, jointly optimizing sub-channel and power allocation in NOMA-enhanced networks can lead to a significant improvement.

Rethinking Clustered Federated Learning in NOMA Enhanced Wireless Networks

TL;DR

The extensive simulation results show that the proposed cluster-based FL framework can outperform FL baselines in terms of both test accuracy and convergence rate, and jointly optimizing sub-channel and power allocation in NOMA-enhanced networks can lead to a significant improvement.

Abstract

This study explores the benefits of integrating the novel clustered federated learning (CFL) approach with non-orthogonal multiple access (NOMA) under non-independent and identically distributed (non-IID) datasets, where multiple devices participate in the aggregation with time limitations and a finite number of sub-channels. A detailed theoretical analysis of the generalization gap that measures the degree of non-IID in the data distribution is presented. Following that, solutions to address the challenges posed by non-IID conditions are proposed with the analysis of the properties. Specifically, users' data distributions are parameterized as concentration parameters and grouped using spectral clustering, with Dirichlet distribution serving as the prior. The investigation into the generalization gap and convergence rate guides the design of sub-channel assignments through the matching-based algorithm, and the power allocation is achieved by Karush-Kuhn-Tucker (KKT) conditions with the derived closed-form solution. The extensive simulation results show that the proposed cluster-based FL framework can outperform FL baselines in terms of both test accuracy and convergence rate. Moreover, jointly optimizing sub-channel and power allocation in NOMA-enhanced networks can lead to a significant improvement.
Paper Structure (22 sections, 4 theorems, 40 equations, 7 figures, 2 algorithms)

This paper contains 22 sections, 4 theorems, 40 equations, 7 figures, 2 algorithms.

Table of Contents

  1. Introduction
  2. Prior Works
  3. Prior Works on FL and CFL
  4. FL in Wireless Networks
  5. Motivation and Contributions
  6. System Model
  7. Clustered Federated Learning Model
  8. Hybrid NOMA Transmission Model
  9. Problem Formulation
  10. Formulation of Energy Consumption Minimization
  11. Description of Methodologies in CFL Framework
  12. BFGS Algorithm in Log-Likelihood
  13. Complexity Analysis
  14. Spectral Clustering
  15. Complexity Analysis
  16. ...and 7 more sections

Key Result

Theorem 1

The generalization gap of the empirical learned global model on unseen data $F_u\left(\widehat{\boldsymbol{w}}\right) - F_u\left({\boldsymbol{\widehat{w}}^*}\right)$ follows a probability of at least $1 - \delta$ for any $\delta \in (0,1)$:

Figures (7)

  • Figure 1: An illustration of population distribution when $N = 30$, where the distribution among classes is represented with different colors, i.e., each user has a different number of data label(s) of data under different $\alpha$ as shown in (a) and (b).
  • Figure 2: Illustration of the considered hybrid NOMA system.
  • Figure 3: The performance of the proposed framework on both datasets under different optimization strategies. $\alpha = 0.01$, $K = 10$, $T_{max}=6 s$.
  • Figure 4: Comparisons between conventional FL, random selection, and the proposed method. $\alpha = 0.01$, $K = 10$, $T_{max}=6 s$.
  • Figure 5: The impact of the different number of clusters on the performance in different datasets. $K=5$, $T_{max} = 6s$.
  • ...and 2 more figures

Theorems & Definitions (9)

  • Theorem 1
  • Remark 1
  • Theorem 2
  • Definition 1
  • Definition 2
  • Definition 3
  • Proposition 1
  • Proposition 2
  • Definition 4