Thursday, November 27, 2025 | 11:00 AM
Department of Electronics, Information and Bioengineering - Politecnico di Milano
Room 3A (Building 20)
Speaker: Yongjun Kim (Korea Advanced Institute of Science and Technology)
Contacts: Prof. Marco Mezzavilla | marco.mezzavilla@polimi.it
Abstract
Artificial intelligence (AI) is rapidly reshaping the design of wireless communication systems and has given rise to new concepts such as AI-native communication. At the same time, the emergence of powerful generative models, including large language models (LLMs) and diffusion models (DMs), is pushing research beyond the traditional use of simple neural networks (NNs).As the transformation to AI models significantly affects system cost and complexity, a central question arises: which roles should be reserved for AI inside future wireless systems, to provide clear benefits over conventional signal processing?
This presentation addresses that question through a series of concrete examples. The first part introduces studies that apply simple NNs in wireless communication. These include NN-based channel state information feedback, a federated learning framework that explicitly incorporates wireless channel conditions into the model design, and the use of NN in integrated sensing and communications.
The second part discusses what generative models can uniquely contribute beyond conventional NN architectures. The role of LLMs in semantic communication and their potential to enhance system resilience are addressed.
Finally, DM for channel prediction to capture the channel dynamics across frequency is introduced. Through these examples, the talk aims to outline a view of where AI adds value in wireless communication and how to position it in the integrated systems.