Flexible and lightweight toolbox for federated learning on edge devices

Biskup, Dean

Permalink

https://hdl.handle.net/2142/117764 Copy

Description

Title

Flexible and lightweight toolbox for federated learning on edge devices

Author(s)

Biskup, Dean

Issue Date

2022-11-17

Director of Research (if dissertation) or Advisor (if thesis)

Smaragdis, Paris

Department of Study

Electrical & Computer Eng

Discipline

Electrical & Computer Engr

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• federated learning
• machine learning
• speech enhancement

Abstract

As edge devices with data collection capabilities, such as cell phones, home assistants, or autonomous vehicles become more ubiquitous, there has been a rapid increase in the amount of data collected. While this data is valuable for machine learning applications, there is an increasing demand for data privacy and effective local data processing to lower network bandwidth requirements. Federated Learning has emerged as a central paradigm of machine learning to tackle these issues, allowing for collaborative learning between many edge devices without requiring the sharing of sensitive data. To facilitate research into federated learning on real-world devices, this thesis work introduces the Federated Learning on Edge Systems (FLoES) software library, a flexible and lightweight federated learning toolbox that is targeted to run on single-board computers. The source code is available at: https://www.github.com/dbisk/floes.

Graduation Semester

2022-12

Type of Resource

Thesis

Copyright and License Information

Copyright 2022 Dean Biskup

Owning Collections