Federated Learning

Welcome to the special course in machine learning in spring 2024: Federated Learning!

This special course provides a forum for MSc/PhD students to discuss and generate ideas on various applied research issues in Federated Machine Learning (FL). Students read FL publications in-depth, work in discussion groups, and debate. In the research seminar, the students can integrate their knowledge, skills, and practical experience gained throughout their studies. The seminars worth 3 ECTS and 3 extra ECTS can be assigned to students who decide to carry on a project in the seminars' context.

• In this main page you can find some general information about the course;
• In the Schedule tab you can find the order of presentations, recording of past sessions, links to the presentation questions, etc.;
• In the Presentations tab you can find guidelines for the seminars and evaluation methodology, amongst other relevant information;
• Check the Links tab for some useful resources.

Introduction

This course delves into the rapidly evolving field of distributed and privacy-preserving ML, aka Federated Learning (FL). FL tackles the challenge of training powerful machine learning models while preserving data privacy. It allows multiple devices or institutions to collaboratively train a model without sharing their private datasets.

Structure

As is usual, this special course in machine learning will be structured in the format of seminars, where enrolled students will pick a topic/article to present (reversed classroom approach). We will explore from the conceptual basics to the current state-of-the-art in literature through papers. Please check the tab "Schedule" to select your presentation topic. For the first session, I will be giving a talk myself on the basics of federated learning, to put everyone on a common ground for the weeks of seminars to follow. The second session will have a hands-on session on federated learning in practice.

Every student must choose 2 research papers. One to defend and one to oppose. NOTABLY: Reading a paper requires reading many references, too. Both the defender and the opponent should focus on the literature.

Learning environment

Form of study: Online learning.

The course will take place in Zoom on Wednesday at 16:15 - 18:00 in Delta room 1024 and Zoom (log into courses to see the link). Meetings shall also be recorded and made available after the sessions (but note that, if you are a registered student, attendance is required!)

NB!: the first session will take place on September 11th, 2024!

To pass the course, enrolled students will have to present their topics and come up with questions about them, which will be passed on for other students to answer. Check the tab "Presentations" for details about the methodology and what is expected from you while presenting.

Contact

Lecturer in charge: Feras Awaysheh, head of Edge Intelligence and Data Analytics group https://eida.cs.ut.ee/ feras.awaysheh@ut.ee