Master projects Federated Learning

Project description

Federated Learning is a Machine Learning approach that allows multiple clients to collaboratively train a global model. This is achieved by aggregating locally trained models, eliminating the need to share local training data. Consequently, there’s no requirement for a central database to store sensitive data, preventing potential data leaks.
In FL, the responsibility of data training is shifted to each local client. Communication between the client and server occurs through parameter interaction, and not through direct data interaction. The server’s role is limited to simple parameter aggregation for global model updates. The FL framework can protect local user data while conserving the server’s computational and storage resources.

Context

The three students have different topics listed below:

Deyna Baeva: "Exploring Federated Learning for Anomaly Detection"
Sanne Joore: "Assessing the consistency of model accuracy under label flipping attacks in Federated Learning"
Kirill Smirnov: "Dynamic Grouping in Federated Unsupervised Machine Learning"

Results

Each student has different results.