Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Resource-Aware Hierarchical Federated Learning for Video Caching in Wireless Networks

Md Ferdous Pervej, Andreas F Molisch

TL;DR

This work tackles privacy-preserving prediction of user content requests to optimize video caching in wireless networks under sporadic data updates and partial client participation. It introduces RawHFL, a resource-aware hierarchical federated learning framework that jointly optimizes client selection, local training rounds, and CPU frequencies within delay and energy budgets, and provides a convergence bound showing how resource decisions influence learning efficiency. The authors formulate a DC-programming-based approach to a joint optimization problem, relax binary variables, and solve it iteratively with CVX to achieve energy-efficient convergence while maintaining accuracy. Simulations demonstrate that RawHFL attains higher prediction accuracy with substantially lower or competitive energy expenditure compared to baselines, validating its practical impact for privacy-preserving, efficient video caching in wireless networks.

Abstract

Video caching can significantly improve backhaul traffic congestion by locally storing the popular content that users frequently request. A privacy-preserving method is desirable to learn how users' demands change over time. As such, this paper proposes a novel resource-aware hierarchical federated learning (RawHFL) solution to predict users' future content requests under the realistic assumptions that content requests are sporadic and users' datasets can only be updated based on the requested content's information. Considering a partial client participation case, we first derive the upper bound of the global gradient norm that depends on the clients' local training rounds and the successful reception of their accumulated gradients over the wireless links. Under delay, energy and radio resource constraints, we then optimize client selection and their local rounds and central processing unit (CPU) frequencies to minimize a weighted utility function that facilitates RawHFL's convergence in an energy-efficient way. Our simulation results show that the proposed solution significantly outperforms the considered baselines in terms of prediction accuracy and total energy expenditure.

Resource-Aware Hierarchical Federated Learning for Video Caching in Wireless Networks

TL;DR

This work tackles privacy-preserving prediction of user content requests to optimize video caching in wireless networks under sporadic data updates and partial client participation. It introduces RawHFL, a resource-aware hierarchical federated learning framework that jointly optimizes client selection, local training rounds, and CPU frequencies within delay and energy budgets, and provides a convergence bound showing how resource decisions influence learning efficiency. The authors formulate a DC-programming-based approach to a joint optimization problem, relax binary variables, and solve it iteratively with CVX to achieve energy-efficient convergence while maintaining accuracy. Simulations demonstrate that RawHFL attains higher prediction accuracy with substantially lower or competitive energy expenditure compared to baselines, validating its practical impact for privacy-preserving, efficient video caching in wireless networks.

Abstract

Video caching can significantly improve backhaul traffic congestion by locally storing the popular content that users frequently request. A privacy-preserving method is desirable to learn how users' demands change over time. As such, this paper proposes a novel resource-aware hierarchical federated learning (RawHFL) solution to predict users' future content requests under the realistic assumptions that content requests are sporadic and users' datasets can only be updated based on the requested content's information. Considering a partial client participation case, we first derive the upper bound of the global gradient norm that depends on the clients' local training rounds and the successful reception of their accumulated gradients over the wireless links. Under delay, energy and radio resource constraints, we then optimize client selection and their local rounds and central processing unit (CPU) frequencies to minimize a weighted utility function that facilitates RawHFL's convergence in an energy-efficient way. Our simulation results show that the proposed solution significantly outperforms the considered baselines in terms of prediction accuracy and total energy expenditure.
Paper Structure (14 sections, 2 theorems, 27 equations, 3 figures, 2 tables, 2 algorithms)

This paper contains 14 sections, 2 theorems, 27 equations, 3 figures, 2 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. System Model
  3. Network Model
  4. Content Request Model
  5. Hierarchical Federated Learning: Preliminaries
  6. Resource-Aware Hierarchical Federated Learning: Algorithm and Convergence
  7. Resource-Aware Hierarchical Federated Learning Model
  8. Convergence of RawHFL
  9. RawHFL: Joint Problem Formulation and Solutions
  10. Simulation Results and Discussions
  11. Simulation Setting
  12. Performance Analysis
  13. Performance Comparisons
  14. Conclusion

Key Result

Theorem 1

Suppose the above assumptions hold. When $\eta < \mathrm{min}\left\{\frac{1}{2\sqrt{5} \beta \mathrm{L}}, \frac{1}{\beta E \mathrm{L}} \right\}$, the average global gradient norm is upper-bounded as where the expectations depend on clients' randomly selected mini-batches and $\mathrm{1}_{u,\mathrm{sc}}^{k,e}$'s. Besides,$\Omega^k \coloneqq \sum_{e=0}^{E-1} \sum_{b=0}^{B-1} \alpha_b \sum_{u \in \b

Figures (3)

  • Figure 1: Global round vs average test loss and accuracy
  • Figure 2: CDF of client's total energy expense
  • Figure 3: Top-M accuracy comparisons

Theorems & Definitions (4)

  • Theorem 1
  • proof : Sketch of Proof
  • Remark 1
  • Corollary 1