ZECCHIN Matteo - Communication systems
Date: - Location: Eurecom
Abstract: The increasing demand for adaptable and easily reconfigurable wireless networks, such as OpenRAN, calls for tools that enable reliable configuration and analysis of these systems. Network operators often face the challenge of selecting the right algorithm hyperparameters to meet target key performance indicators, while also trying to understand how performance might be affected by alternative choices. This talk addresses these challenges using statistically grounded approaches. First, we introduce adaptive learn-then-test, a novel and efficient sequential hypothesis testing procedure designed to identify hyperparameters with provable reliability guarantees. Next, we apply the framework of conformal prediction to address the problem of reliable “what-if” analyses—evaluating how a system’s key performance indicators would have been affected under different network configurations. Together, the techniques presented in this talk provide tools to enable the safe configuration and deployment of wireless networks. Bio: Dr. Zecchin Matteo is a Postdoctoral Research Associate at the King's Communications, Learning & Information Processing lab at King’s College London. He obtained his PhD in Telecommunication Engineering from EURECOM in 2022. His research focuses on the intersection of wireless communication and machine learning, with a particular emphasis on the application of Bayesian methods to wireless communication systems and decentralized learning approaches.
