Πανεπιστήμιο Μακεδονίας
Implementing a hybrid federated learning framework with neural network optimisation for improved artificial intelligence model performance
Abstract
dc:descriptionThis thesis focuses on the design and implementation of a hybrid Federated Learning (FL) framework aimed at optimizing the performance of artificial intelligence models in distributed, sensitive, and dynamic environments such as 5G networks, IoT systems, and autonomous vehicles. By combining techniques like Federated Averaging, Convolutional Neural Networks (CNN), and cloud/container technologies, it proposes a flexible three-tier architecture (client–edge–cloud) that incorporates heuristics for intelligent client selection, communication cost reduction, and enhanced data privacy. Additionally, a strategy tailoring tool (FLATT) is developed, enabling designers to configure and adapt FL strategies dynamically at runtime based on the specific requirements of the application domain. The study demonstrates that traditional FL strategies have reached their limitations in terms of performance, accuracy, and handling non-IID (non-independent and identically distributed) data. Through the integration of algorithmic approaches (FedAvg, gradient descent, evolutionary heuristics) and the use of containers and edge computing infrastructures, the proposed hybrid framework achieves significantly improved global model performance. Case study results confirm that increasing the number of participating clients, applying multi-layer aggregation, and incorporating dynamic gradient evaluation contribute to better generalization and reduced communication overhead. The FLATT tool empowers designers to formulate meta-level strategies in resource-constrained environments, establishing this approach as a solid foundation for scalable and intelligent FL applications of the future.
Degree
thesis:*- Grantor dc:publisher
- Πανεπιστήμιο Μακεδονίας
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Stergiou, Konstantinos
- Στεργίου, Κωνσταντίνος
Subjects
dc:subject × 20- Μηχανική μάθηση
- Ομοσπονδιακή μάθηση
- Μέσος Όρος Ομοσπονδιακής Μάθησης
- Υπολογισμός Αιχμής Κινητών Συσκευών
- Συνελικτικά νευρωνικά δίκτυα
- Συνεργατική μάθηση
- Διατήρηση ιδιωτικότητας
- Machine learning
- Federated learning
- Federated Averaging
- Mobile edge computing
- Convolutional neural networks
- Collaborative learning
- Privacy
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Επιστήμες Μηχανικού και Τεχνολογία
- Μηχανική και συστήματα επικοινωνιών, Τηλεπικοινωνίες
- Electrical Engineering, Electronic Engineering, Information Engineering
- Engineering and Technology
- Communication engineering and systems, Telecommunications
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/59543