The 2025 edition of the course CS-E4740 - Federated Learning includes a student project. You are free to choose an application of federated learning (FL). The goal is to formulate the application as an instance of generalized total variation (GTV) minimization over an FL network.
FL networks and GTV minimization will be introduced as core mathematical structures for studying FL during the course lectures and in the course book.
Project Requirements
- Choose an FL application relevant to your interests.
- Design and analyze FL algorithms from the course to address your chosen application.
- Write a project report using the provided template tex (pdf).
- Submit your report by April 30, 2025.
- Peer-review other students’ reports until May 15, 2025. Current draft for review sheet pdf
- Submit a revised report and response letter by May 31, 2025.
Schedule
| Event | Date |
|---|---|
| First project report submission | April 30, 2025 |
| Peer grading deadline | May 15, 2025 |
| Final project submission | May 31, 2025 |
Ground Rules
As a student in this course, you must adhere to the Aalto University Code of Conduct. The two key principles for this project are:
Rule I - Be Honest
- Plagiarism is strictly forbidden – cite any sources you use.
- Randomly selected students may be asked to explain their work (including peer grading).
Rule II - Be Respectful
- The course aims to provide a safe and enjoyable learning environment.
- Any disrespectful behaviour (including on course communication platforms) will be strictly sanctioned and reported to university authorities.
Example Applications of FL
FL is used in various domains, including:
- Healthcare (e.g., personalized medical models)
- Finance (e.g., fraud detection and risk assessment)
- Smart grids (e.g., optimizing energy distribution)
- Mobile networks (e.g., predicting network congestion)
- Autonomous vehicles (e.g., improving self-driving models)
- Cybersecurity (e.g., detecting network anomalies)
FL in Healthcare
Federated Learning can enable smartphones to become personal healthcare advisors by training models using local and public health data.
Key Reference:
Rieke, N., et al. The future of digital health with federated learning. Nature Medicine, 2020.
FL in Finance
Federated Learning can improve fraud detection and credit risk assessment in financial institutions.
- Fraud detection: Aurna, N. F., et al. Federated Learning-Based Credit Card Fraud Detection. (2023)
- Risk assessment: Li, W., et al. Personal Credit Evaluation Model Based on Federated Learning. (2024)
FL at the Finnish Meteorological Institute (FMI)
- FL models can be trained individually for each FMI weather station to make local predictions.
- Python script for retrieving FMI data: click me
FL for Finnish Road Safety
- Train local models for traffic cameras to predict congestion and accidents.
- Python script for retrieving camera snapshots: click me
The Growing Role of FL in IoT
The Internet of Things (IoT) is a massive, global FL system where billions of devices communicate to improve efficiency and automation. As IoT networks grow, federated learning will play a critical role in enabling decentralized intelligence.
Ready to Start?
- Pick your FL application.
- Formulate your problem as a GTV minimization task.
- Experiment with FL algorithms from the course.
- Submit your findings and improve through peer review.
Good luck with your project!
Feel free to reach out with any questions or for further guidance!