Bachelor and Master Theses

In this thesis the students will have to explore the usage of Large Language Models (LLMs) to generate synthetic datasets that capture the traffic characteristics of users connected to a cellular network. The network operators are often limiting access to real-world network datasets due to privacy concerns and the proprietary nature of the data. This hinders research on network performance analysis, scheduling optimization, the utilization of AI to enhance reliability, and determinism of wireless networks.

By leveraging LLM’s generative capabilities, this work aims to produce synthetic datasets that emulate real cellular traffic behaviors. These datasets could serve as valuable resources for evaluating and  improving wireless network algorithms and architectures without requiring sensitive real-world data which are usually not available.

Goals:

• Study and review of the existing existing approaches for synthetic data generation in cellular networks
• Identify the current capabilities and limitations of LLMs for structured data generation
• Identify key features and metrics of cellular network traffic like throughput, latency, channel characteristics.
• Design and implement an LLM-based method for generating synthetic datasets for cellular traffic.
• Evaluate the realism and diversity of the generated data.

• Basic knowledge of machine learning and deep learning architectures
• Experience with Python
• Skills in data analysis and visualization
• Understanding of cellular network fundamentals is a bonus
• Two students for this project would be preferable