Abstract：Channel estimation is an important task in wireless communications systems and many researchers have recently explored neural network solutions to this problem. In this talk, we will discuss the design of neural network systems to provide robust estimation of unseen wireless channels as part of the channel estimation process. We will discuss several distinct examples of how robustness can be provided in machine learning, particularly targeting techniques suited to battery powered mobile terminals. Our work has explored the use of reinforcement learning to track spatial channel characteristics in millimeter-wave communications channels. We have also explored the approach of selecting from several available neural networks, each of which is targetted to handle a different set of channel conditions. We have further studied novel approaches to design simple and effective training sets that provide good performance over a wide range of simulated channels.
时间：2025.07.17 09:50-10:35
地点：紫金山实验室新大楼2楼报告厅3
John S. Thompson currently holds a personal chair in Signal Processing and Communications at the School of Engineering, University of Edinburgh. He currently specializes in antenna array processing, energy-efficient wireless communications and more recently in the application of machine learning to wireless communications. To date, he has published in excess of 500 journal and conference papers on these topics. He is currently area editor for the wireless communications topic in IEEE Transactions on Green Communications and Networking. In January 2016, he was elevated to Fellow of the IEEE for research contributions to antenna arrays and multi-hop communications. He was also one of four scientists elevated to Fellow of the European Association for Signal Processing (EURASIP) in 2023 for “signal processing advances in multiple antenna and relayed wireless communication systems”.
