Outcome:
On successful completion of this course, students should be able to:
1. Demonstrate knowledge of the emerging federated learning architecture and data privacy using different frameworks.
2. Understand the various capabilities of advanced federated learning solutions.
3. Apply state-of-the-art advanced federated learning systems to build scalable solutions for data privacy in different application domains.
4. Demonstrate the ability to build complex federated learning pipelines that integrate different architectures for dealing with heterogeneous data sets and models.
5. Develop strategic thinking and soft skills for industry and business success using cutting-edge solutions that are expected to lead the next decade of research.
Teachers:
Feras Mahmoud Naji Awaysheh - Lecturer of Edge Analytics. PhD holder. 8 years of working experience in academia.
Content:
Topic |
Credit |
Lectors |
1. Introduction to Machine Learning (ML pipelines).
-ML Lifecycle
-Prepare Dataset and apply traditional ML and DL
-Compare to deep learning
-Data privacy and Data Protection Regulation (e.g., GDPR)
-Introduction to Federated Machine learning
-Simulating a vanilla FL
-Apply FedAVG algorithm
-Challengers of FL
-FL aggregation algorithms and applications
-Different FL aggregation algorithms
-Horizontal and Vertical Data distribution
-Guest presentation on vertical FL
-Intro to FL open-source frameworks (e.g., FEDn and FLOWER)
-Frameworks installation, configuration, and setup
-Guest tutorial by FLOWER framework developers
-FL Architectures and Communication efficiency techniques
-Use cases: cross-silo and cross-device
-Group read and project ideas
-One-Shot Federated Learning
-New trends in FL and Personalized modelling
-Guest session on Federated Learning for Tabular Data
-New trends e.g., Meta Learning, Transfer Learning, Split Learning, and Interactive Learning
-AutoML as a solution for FL optimization
-Guest session on AutoML
-Applying AutoML to FL |
78 |
|
Location info:
Tartu linn
Learning environment:
Studies and teaching takes place in appropriate classrooms, which have the required teaching equipment and meet the health and safety requirements.
Requirements to complete:
Active participation in lectures and practices.
Outcome method:
non-differentiated (pass, fail, not present)
Grading method:
To get a certificate, you must attend at least seven lectures and practice sessions.
Grading criteria:
Active participation in lectures and practices.
Document to be issued:
Certificate of completion/Certificate of attendance
Additional information:
Jaanika Seli, ati.taiendope@ut.ee, +372 +372 737 6426
Program code:
LTAT.TK.050