Optimization of Bandwidth Usage in Wireless Federated Learning Systems

• Author / Creator

  Zamani, Pedram

• Machine learning (ML) and the Internet of Things (IoT) have many promises but also raise many concerns and challenges. In particular, when several users collaborate in training an ML model, preserving the privacy of their data is quite challenging. Federated learning (FL), as an ML technique based on distributed computing, has been proposed to address this and other challenges of collaborative ML training.
  While FL has gained vast popularity in both academia and industry, its deployment in practice, especially in time-sensitive and energy-limited wireless applications, is challenging. Moreover, scarce communication resources such as bandwidth must be efficiently utilized. In wireless FL systems, bandwidth allocation plays a crucial role in determining the overall performance, including training latency, model accuracy, and energy consumption. The limited availability of bandwidth, coupled with the need for synchronized communication among FL clients, makes bandwidth allocation a complex optimization problem.
  Hence, in this work, we explore the problem of minimizing the total bandwidth usage in a wireless FL system under time and energy constraints. We formulate this problem as a non-convex optimization problem that aims to minimize the total bandwidth usage of the system while respecting the mentioned practical constraints. By decomposing the problem into two subproblems, we show that it can be solved efficiently using convex optimization and iterative search techniques. Our proposed algorithm finds the optimal solution while enjoying low complexity, making it suitable for real-world implementations. Through comprehensive simulations, we demonstrate the efficiency of our approach and analyze different aspects of the problem.
  This work contributes to the ongoing research efforts in optimizing FL for wireless networks, addressing the critical challenges of bandwidth allocation along with cost constraints. The insights gained from this study can help in developing more robust and efficient FL systems for a wide range of time-sensitive and energy-constrained wireless applications.

• Subjects / Keywords

  • Federated Learning
  • Wireless
  • Bandwidth
  • Optimization

• Graduation date

  Fall 2024

• Type of Item

  Thesis

• Degree

  Master of Science

• DOI

  https://doi.org/10.7939/r3-smwj-p446

• License

  This thesis is made available by the University of Alberta Library with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.