UniTartu Summer School MegaData: Federated Machine Learning
Author:
Henry Narits
2nd Summer School on MegaData: Federated Machine Learning
On-site in Tartu 28 July - 10 August 2024
2nd Summer School on Federated Machine Learning
course provides an introduction to Federated Machine Learning (FL), a privacy-preserving distributed ML. The course will cover the foundational aspects of FL operation and deployment models in Edge computing. Modern FL technologies will cover various aspects, including different data distributions, aggregation algorithms, and communication efficiency approaches. The students will be introduced to state-of-the-art FL technologies and architectures and guided to investigate novel ideas in the area via lectures, practice sessions, and projects. We will also look at industry trends and discuss some innovations that have recently been developed.
The course targets MSc degree students and Ph.D. candidates looking to develop their capacity in modern computer deployment architecture at the Edge/Fog to meet the increasing demand in industry and academia. Also, the course is designed for students of joint data science and distributed system curriculum towards Edge Intelligence. We combine theory, practice sessions, and project assignments to learn about FL. After completing this course, you will learn more about designing and developing an FL solution. Some course material will be drawn from research papers, industry white papers, and technical reports.
The course can be taken on-site in Tartu, Estonia. We have a lecture and discussions in the morning session. Afternoon sessions are dedicated to practicing sessions and project work.
Application deadline 31 May
NB! Please note that every applicant must pay the application fee of 25 EUR. In the application form you must upload proof of payment. Please complete the payment on the application fee payment page.
| Focus area: | Designing and Implementing Federated Machine Learning | Coordinating unit at UT | Institute of Computer Science (Data Systems Group) |
| Study Field: | Computer Science | Course Leader | Feras Awaysheh |
| Format | Hands-on workshop | Location | Tartu, Estonia, Delta Centre |
| Course dates: | 28 July - 10 August 2024 | Apply by: | 31 May 2024 |
| ECTS: | 3 (+2 for additional assignment) | Fee: | 800 EUR |
| Study | MSc/PhD | Language | English |
Lecturers:
- Feras Awaysheh, University of Tartu, Estonia
- Sadi AlAwadi, Halmstad University, Sweden
Guest Talks:
- Afsana Khan, Maastricht University, Netherlands.
- Daniel J. Beutel, Cambridge University, UK
- Florian van Daalen, University Maastricht, Netherlands
- Mohamed Elmahallawy, Missouri University of Science and Technology, USA
- William Lindskog, Technical University of Munich, Germany
- Salman Toor, Uppsala University, Sweden
- Hossam Fakhory, Petra University, Jordan
- Hassan Eldeeb, Tartu University, Estonia
Sunday, July 28
Arrival
Monday, July 29
Arrival and welcome meeting.
Tuesday, July 30
Introduction to Machine Learning (ML pipelines).
ML Lifecycle and centralized deep learning.
Wednesday, July 31
Data privacy and Data Protection Regulation (e.g., GDPR)
Introduction to Federated Machine learning
Thursday, 1 August
FL challengers of FL
FL aggregation algorithms and applications
Horizontal and Vertical Data distribution
Friday, 2 August
Intro to FL open-source frameworks (e.g., FEDn and FLOWER)
Frameworks installation and configuration
Saturday, 3 August
Free day
Sunday, 4 August
Free day
Monday, 5 August
FL Architectures and Communication efficiency techniques
Use cases cross-silo and cross-device
Tuesday, 6 August
New trends in FL and 5.Personalized modeling
E.g., Meta Learning, Transfer Learning, Split Learning, and Interactive Learning.
Wednesday, 7 August
AutoML as a solution for FL optimization
Lightweight ML (e.g., Edge Impulse) and FL Security scenarios
Thursday, 8 August
Wrap up with real applications and FL for medical image analysis
Friday, 9 August
Participant's projects and final presentations
Saturday, 10 August
Departure
The students will be divided into groups of up to 3 to work on a group project. The group work aims to implement and develop a federated learning solution. The students will have a chance to choose one of the challenges from the project pool and propose a solution for it. The lecturers will be available for consultation during the whole course. The deliverable of the group work will be a presentation of their solution. We will extend the work with some nominated solutions for publication possibility.
Entry requirements:
Interest in designing and developing privacy-preserving ML solutions. Also, the course is designed for joint data science and distributed system curriculum students. Good Machine Learning is a mandatory prerequisite. Students are encouraged (but not necessarily required) completed Computer Networks, Distributed Systems, Cloud Computing, and Big Data Management courses.
- Online application form
- Motivation letter (up to 1 page) that demonstrates the applicant’s motivation to participate, his/her expectations about the program, how participation in the summer program relates to his/her studies and interests, and how the applicant plans to use the gained experience and knowledge in the future).
PS: Only complete applications, including all annexes submitted by the deadline will be considered for selection.
On successfully completing this course, students should be able to:
1. Demonstrate knowledge of the emerging federated machine learning (FL) deployment architecture and requirements.
2. Understand the various capabilities of advanced FL solutions and develop the ability to choose adequate systems for different problems.
3. Apply state-of-the-art FL systems to build scalable solutions for various data privacy challenges in different application domains.
4. Apply qualitative and quantitative techniques in distributed machine intelligence through lectures and design projects using leading research trends to identify the strengths and weaknesses of the various systems.
5. Develop strategic thinking and soft skills for industry and business success using cutting-edg
Last years participants' publications
Christian Lenz, Marc Leon Haller, and Robin Nachtigall
Handling Non-IID Data in Federated Learning: An Experimental Evaluation Towards Unified Metrics
Hiroki Kaminaga
MPCFL: Towards Multi-party Computation for Secure Federated Learning AggregationChristos Ntokos
Towards Accelerating the Adoption of Federated Learning for Heterogeneous Data
Leon de Franca Nascimento
Data Skew in Federated Learning: An Experimental Evaluation on Aggregation Algorithms
Mehreen Tahir
SecureFedPROM: A Zero-Trust Federated Learning Approach with Multi-Criteria Client Selection IEEE Journal on Selected Areas in Communications. Currently under review.
Two weeks prior to the start of the programme an information file will be sent to all participants. This file contains the daily schedule and relevant contact information of the programme managers.
Students are responsible for their travel, accommodation and travel insurance (visa arrangements if needed) from their home country to Tartu and back to their home country. It is recommended to visit the Tartu Welcome Centre website and arrival and housing section to find accommodation opportunities.
Image
Mehreen Tahir, a final year PhD student at Dublin City University, Ireland participated in the UniTartu Summer School in 2023.
“My journey to the UniTartu Summer School was a bit of a happy accident. There are not many courses offered in Federated Learning. Thinking that I would at least get a chance to network with like-minded people from my field, I decided to give it a shot. And it turned out to be a beautiful and welcoming place, much more than what I could have imagined!”
MegaData: Federated Machine Learning in 2023
Scholarships
Read more
Social programme
Read more
