Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

IRS Aided Federated Learning: Multiple Access and Fundamental Tradeoff

Guangji Chen, Jun Li, Qingqing Wu, Yiyang Ni, Meng Hua

TL;DR

This work addresses reducing learning latency in wireless federated learning by introducing three IRS-aided uploading protocols (I-TDMA, I-FDMA, I-NOMA) and formulating joint optimization of IRS phase shifts, device scheduling, and communication/computation resources under accuracy constraints.It develops latency-minimization problems for each protocol and solves them with TDMA-optimal scheduling, SCA-based CVX approaches, and coordinate-descent methods, including closed-form results and Lambert W-based expressions where applicable.Key theoretical contributions include sufficient conditions where I-TDMA outperforms I-NOMA (and vice versa) in IRS-enabled settings, and a derived IRS-element count required for full device scheduling, illustrating fundamental tradeoffs between learning latency and accuracy.Extensive simulations validate the theory, showing IRS significantly improves the latency-accuracy tradeoff, enables more devices to participate under tight latency, and enhances learning performance on practical tasks such as MNIST.

Abstract

This paper investigates an intelligent reflecting surface (IRS) aided wireless federated learning (FL) system, where an access point (AP) coordinates multiple edge devices to train a machine leaning model without sharing their own raw data. During the training process, we exploit the joint channel reconfiguration via IRS and resource allocation design to reduce the latency of a FL task. Particularly, we propose three transmission protocols for assisting the local model uploading from multiple devices to an AP, namely IRS aided time division multiple access (I-TDMA), IRS aided frequency division multiple access (I-FDMA), and IRS aided non-orthogonal multiple access (INOMA), to investigate the impact of IRS on the multiple access for FL. Under the three protocols, we minimize the per-round latency subject to a given training loss by jointly optimizing the device scheduling, IRS phase-shifts, and communicationcomputation resource allocation. For the associated problem under I-TDMA, an efficient algorithm is proposed to solve it optimally by exploiting its intrinsic structure, whereas the highquality solutions of the problems under I-FDMA and I-NOMA are obtained by invoking a successive convex approximation (SCA) based approach. Then, we further develop a theoretical framework for the performance comparison of the proposed three transmission protocols. Sufficient conditions for ensuring that I-TDMA outperforms I-NOMA and those of its opposite are unveiled, which is fundamentally different from that NOMA always outperforms TDMA in the system without IRS. Simulation results validate our theoretical findings and also demonstrate the usefulness of IRS for enhancing the fundamental tradeoff between the learning latency and learning accuracy.

IRS Aided Federated Learning: Multiple Access and Fundamental Tradeoff

TL;DR

This work addresses reducing learning latency in wireless federated learning by introducing three IRS-aided uploading protocols (I-TDMA, I-FDMA, I-NOMA) and formulating joint optimization of IRS phase shifts, device scheduling, and communication/computation resources under accuracy constraints.It develops latency-minimization problems for each protocol and solves them with TDMA-optimal scheduling, SCA-based CVX approaches, and coordinate-descent methods, including closed-form results and Lambert W-based expressions where applicable.Key theoretical contributions include sufficient conditions where I-TDMA outperforms I-NOMA (and vice versa) in IRS-enabled settings, and a derived IRS-element count required for full device scheduling, illustrating fundamental tradeoffs between learning latency and accuracy.Extensive simulations validate the theory, showing IRS significantly improves the latency-accuracy tradeoff, enables more devices to participate under tight latency, and enhances learning performance on practical tasks such as MNIST.

Abstract

This paper investigates an intelligent reflecting surface (IRS) aided wireless federated learning (FL) system, where an access point (AP) coordinates multiple edge devices to train a machine leaning model without sharing their own raw data. During the training process, we exploit the joint channel reconfiguration via IRS and resource allocation design to reduce the latency of a FL task. Particularly, we propose three transmission protocols for assisting the local model uploading from multiple devices to an AP, namely IRS aided time division multiple access (I-TDMA), IRS aided frequency division multiple access (I-FDMA), and IRS aided non-orthogonal multiple access (INOMA), to investigate the impact of IRS on the multiple access for FL. Under the three protocols, we minimize the per-round latency subject to a given training loss by jointly optimizing the device scheduling, IRS phase-shifts, and communicationcomputation resource allocation. For the associated problem under I-TDMA, an efficient algorithm is proposed to solve it optimally by exploiting its intrinsic structure, whereas the highquality solutions of the problems under I-FDMA and I-NOMA are obtained by invoking a successive convex approximation (SCA) based approach. Then, we further develop a theoretical framework for the performance comparison of the proposed three transmission protocols. Sufficient conditions for ensuring that I-TDMA outperforms I-NOMA and those of its opposite are unveiled, which is fundamentally different from that NOMA always outperforms TDMA in the system without IRS. Simulation results validate our theoretical findings and also demonstrate the usefulness of IRS for enhancing the fundamental tradeoff between the learning latency and learning accuracy.

Paper Structure

This paper contains 27 sections, 9 theorems, 77 equations, 8 figures.

Table of Contents

  1. Introduction
  2. System model
  3. Computation Model of Local Computation
  4. Communication Model of Local Model Uploading
  5. I-TDMA Based Transmission
  6. I-FDMA Based Transmission
  7. I-NOMA Based Transmission
  8. Convergence Analysis and Problem Formulation
  9. Convergence Analysis
  10. Problem Formulation
  11. Proposed Solution
  12. Proposed Solution to I-TDMA
  13. Proposed Solution to I-FDMA
  14. Optimization of $\left\{ {{\bf{\Theta }},{\bf{b}},{{\bf{\tau }}^{\rm{F}}},{\bf{p}}} \right\}$ under Given ${\cal K}_a^t$
  15. Optimization of device scheduling
  16. ...and 12 more sections

Key Result

Lemma 1

With $\eta = 1/L$, we have

Figures (8)

  • Figure 1: IRS aided wireless federated learning with OMA and NOMA.
  • Figure 2: Communication latency versus communication energy under the power homogeneous setting.
  • Figure 3: Communication latency versus communication energy under the phase homogeneous setting.
  • Figure 4: Communication latency versus communication energy under the general setting.
  • Figure 5: Per-round latency versus $N$ with $E = 0.1$ Joule.
  • ...and 3 more figures

Theorems & Definitions (18)

  • Lemma 1
  • proof 1
  • Theorem 1
  • proof 2
  • Proposition 1
  • proof 3
  • Proposition 2
  • proof 4
  • Proposition 3
  • proof 5
  • ...and 8 more