Flexible and lightweight toolbox for federated learning on edge devices

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.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Electrical & Computer Engr

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2022

Author dc:creator

• Biskup, Dean

Contributors dc:contributor

• Smaragdis, Paris

• Federated Learning
• Machine Learning
• Speech Enhancement

Statement dc:rights

• Copyright 2022 Dean Biskup

Language dc:language

en, eng

Handle dc:identifier

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

OAI record oai:identifier

https://www.ideals.illinois.edu/oai-pmh?verb=GetRecord&metadataPrefix=etdms&identifier=oai%3Awww.ideals.illinois.edu%3A2142%2F117764