Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Optimized Federated Multitask Learning in Mobile Edge Networks: A Hybrid Client Selection and Model Aggregation Approach

Moqbel Hamood, Abdullatif Albaseer, Mohamed Abdallah, Ala Al-Fuqaha, Amr Mohamed

TL;DR

This work tackles learning over hierarchical mobile edge networks with non-IID data by introducing a CFL framework that combines two-phase client selection and two-level model aggregation. By performing clustering based on gradient similarities at edge and cloud levels and by scheduling participants with fairness-aware and resource-aware mechanisms (greedy or round-robin), the approach yields specialized models per cluster while maintaining a global model. The proposed schemes reduce training time and energy consumption while improving convergence and accuracy on FEMNIST and CIFAR-10, with energy savings reaching up to about 60% and notable gains in convergence speed. The framework enables effective inter-edge collaboration, tailored modeling, and scalable resource management suitable for real-time IoT and intelligent-vehicle applications.

Abstract

We propose clustered federated multitask learning to address statistical challenges in non-independent and identically distributed data across clients. Our approach tackles complexities in hierarchical wireless networks by clustering clients based on data distribution similarities and assigning specialized models to each cluster. These complexities include slower convergence and mismatched model allocation due to hierarchical model aggregation and client selection. The proposed framework features a two-phase client selection and a two-level model aggregation scheme. It ensures fairness and effective participation using greedy and round-robin methods. Our approach significantly enhances convergence speed, reduces training time, and decreases energy consumption by up to 60%, ensuring clients receive models tailored to their specific data needs.

Optimized Federated Multitask Learning in Mobile Edge Networks: A Hybrid Client Selection and Model Aggregation Approach

TL;DR

This work tackles learning over hierarchical mobile edge networks with non-IID data by introducing a CFL framework that combines two-phase client selection and two-level model aggregation. By performing clustering based on gradient similarities at edge and cloud levels and by scheduling participants with fairness-aware and resource-aware mechanisms (greedy or round-robin), the approach yields specialized models per cluster while maintaining a global model. The proposed schemes reduce training time and energy consumption while improving convergence and accuracy on FEMNIST and CIFAR-10, with energy savings reaching up to about 60% and notable gains in convergence speed. The framework enables effective inter-edge collaboration, tailored modeling, and scalable resource management suitable for real-time IoT and intelligent-vehicle applications.

Abstract

We propose clustered federated multitask learning to address statistical challenges in non-independent and identically distributed data across clients. Our approach tackles complexities in hierarchical wireless networks by clustering clients based on data distribution similarities and assigning specialized models to each cluster. These complexities include slower convergence and mismatched model allocation due to hierarchical model aggregation and client selection. The proposed framework features a two-phase client selection and a two-level model aggregation scheme. It ensures fairness and effective participation using greedy and round-robin methods. Our approach significantly enhances convergence speed, reduces training time, and decreases energy consumption by up to 60%, ensuring clients receive models tailored to their specific data needs.
Paper Structure (23 sections, 30 equations, 11 figures, 2 tables, 2 algorithms)

This paper contains 23 sections, 30 equations, 11 figures, 2 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. SYSTEM MODEL
  4. HFL Model
  5. Clustered Federated Learning (CFL)
  6. Models for Localized Computation and Communication
  7. Problem formulation
  8. Proposed Solution
  9. Detailed Schemes
  10. Fair and Greedy Selection with Round-based Model Aggregation
  11. Fair and Greedy Selection with Split-based Model Aggregation
  12. Fair and Round-robin Selection with Round-based Model Aggregation
  13. Fair and Round-robin Selection with Split-based Model Aggregation
  14. Algorithm Complexity Analysis
  15. Performance Evaluation
  16. ...and 8 more sections

Figures (11)

  • Figure 1: The system model where the CFL is implemented at HWNs.
  • Figure 2: Illustration for CFL in HWN systematically partitions clients into several clusters.
  • Figure 3: Testing accuracy for fair and greedy selection with split-based model aggregation and HFL (FEMNIST).
  • Figure 4: Average testing accuracy for models of clusters during training rounds for both proposed approach and baselines using FEMNIST dataset
  • Figure 5: Fair and greedy selection with split-based model aggregation using CIFAR-10 dataset.
  • ...and 6 more figures