Aristotle University Of Thessaloniki (AUTH)
Design and optimization of wireless communication systems and networks with native artificial intelligence
Abstract
dc:descriptionNext generation wireless communication networks are envisioned to adopt a native artificial intelligence (AI) approach, aiming to deliver ubiquitous intelligent services, and to improve the underlying network architectures, protocols and operations through AI-driven solutions. In this context, the convergence of AI with wireless networks is driving the research in two distinct directions. Firstly, it delves into AI for wireless networks, which involves improving the performance of wireless networks through the utilization of AI and machine learning techniques. Secondly, it explores wireless networks for AI, where the goal is to enhance data-driven AI applications by jointly designing and optimizing communication and computing networks. The primary goal of this thesis is to explore the field of wireless networks for AI, while the study is not limited exclusively to this direction, and the aspect of AI for wireless networks is also addressed. Native AI is expected to heavily rely on edge and distributed intelligence, in the sense that intelligent services should be provided across the whole continuum of the network. A promising decentralized learning technique, whose primary goal is to preserve privacy, is the federated learning (FL). In FL, the participating users build collaboratively a machine learning model with the aid of a central entity, e.g., a server, which acts as a fusion center and generates the global model. Users transmit only the local training parameters to the server, and not their raw data, which alleviates privacy concerns. In the context of wireless networks, FL is termed as wireless FL (WFL) and aligns with the aspect of wireless networks for AI. WFL faces distinctive challenges owing to the unreliability of wireless links and their limited bandwidth. Therefore, efficient schemes should be employed that aim to guarantee fast convergence of FL over wireless networks and simultaneously retain increased model accuracy of the underlying task or AI application. Towards this direction, the thesis explores techniques that involve a joint optimization of communication and computation resources. This approach is prevalent throughout this dissertation, with the aim of reducing latency in WFL, while retaining high model accuracy.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Mpouzinis, Pavlos
- Μπουζίνης, Παύλος
Subjects
dc:subject × 12- Ασύρματα δίκτυα
- Βελτιστοποίηση
- Διανεμημένη τεχνητή νοημοσύνη
- Wireless networks
- Optimization
- Federated learning
- 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/56390