Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Pinching Antennas-Assisted Low-Latency Federated Learning Over Multi-User Wireless Networks

Saba Asaad, Hina Tabassum, Ping Wang

TL;DR

FedPASS is a novel framework for low-latency wireless FL assisted by Pinching-antenna systems that achieves accuracy comparable to idealized FL baselines while drastically reducing the total training latency compared to conventional wireless FL.

Abstract

Federated learning (FL) over wireless networks is fundamentally constrained by unreliable communication links, particularly when uplink channels suffer from blockage, fading, or weak line-of-sight (LoS) conditions. Pinching-antenna systems (PASSs) offer a new physical-layer capability to dynamically reposition radiating points along a dielectric waveguide, enabling controllable LoS connectivity and significantly improved channel quality. This paper develops FedPASS, a novel framework for low-latency wireless FL assisted by PASS. We formulate a multi-objective optimization problem that jointly minimizes the end-to-end round latency and an upper bound on the FL optimality gap. The resulting formulation is a mixed-integer nonlinear program subject to practical constraints on scheduling, transmit power, local CPU frequency, and PA placement. To address the resulting computational challenges, we develop a two-tier iterative algorithm: an outer loop that updates scheduling, communication time allocation, and power control via block coordinate descent, and an inner loop that optimizes PA locations using a Gauss-Seidel-based coordinate update with grid search under spacing constraints. Numerical results on MNIST and CIFAR-10 demonstrate that FedPASS achieves accuracy comparable to idealized FL baselines while drastically reducing the total training latency compared to conventional wireless FL.

Pinching Antennas-Assisted Low-Latency Federated Learning Over Multi-User Wireless Networks

TL;DR

FedPASS is a novel framework for low-latency wireless FL assisted by Pinching-antenna systems that achieves accuracy comparable to idealized FL baselines while drastically reducing the total training latency compared to conventional wireless FL.

Abstract

Federated learning (FL) over wireless networks is fundamentally constrained by unreliable communication links, particularly when uplink channels suffer from blockage, fading, or weak line-of-sight (LoS) conditions. Pinching-antenna systems (PASSs) offer a new physical-layer capability to dynamically reposition radiating points along a dielectric waveguide, enabling controllable LoS connectivity and significantly improved channel quality. This paper develops FedPASS, a novel framework for low-latency wireless FL assisted by PASS. We formulate a multi-objective optimization problem that jointly minimizes the end-to-end round latency and an upper bound on the FL optimality gap. The resulting formulation is a mixed-integer nonlinear program subject to practical constraints on scheduling, transmit power, local CPU frequency, and PA placement. To address the resulting computational challenges, we develop a two-tier iterative algorithm: an outer loop that updates scheduling, communication time allocation, and power control via block coordinate descent, and an inner loop that optimizes PA locations using a Gauss-Seidel-based coordinate update with grid search under spacing constraints. Numerical results on MNIST and CIFAR-10 demonstrate that FedPASS achieves accuracy comparable to idealized FL baselines while drastically reducing the total training latency compared to conventional wireless FL.
Paper Structure (33 sections, 2 theorems, 58 equations, 10 figures, 1 algorithm)

This paper contains 33 sections, 2 theorems, 58 equations, 10 figures, 1 algorithm.

Table of Contents

  1. Introduction
  2. Background Work
  3. Related Work on PASS
  4. FL over pass
  5. Contributions
  6. Paper Organization and Notations
  7. System Model and Assumptions
  8. Over-the-Air FL Setting
  9. Communication Model
  10. Evaluation Metrics and Problem Formulation
  11. End-to-End Training Latency Model
  12. Local Computation Latency
  13. Communication latency
  14. Energy Consumption Model
  15. Energy for Local Computation
  16. ...and 18 more sections

Key Result

Lemma 1

Let assumptions (A.1)-(A.4) hold. For the learning rate $\zeta = 1/L$, the global loss in the communication round $t$ satisfies where $\mathbf{e}_{t-1}$ is defined as and defines the aggregation error induced by partial device scheduling.

Figures (10)

  • Figure 1: Graphical illustration of the considered system model for FL over pinching antenna system.
  • Figure 2: Time line of FL-assisted PASS based on TDMA.
  • Figure 3: Test accuracy vs the number of rounds for MNIST dataset
  • Figure 4: Test accuracy vs the number of rounds for CIFAR dataset
  • Figure 5: Accuracy vs total latency for different values of $K$.
  • ...and 5 more figures

Theorems & Definitions (3)

  • Lemma 1
  • Theorem 1
  • Definition 1