Model-driven Deep Learning for B5G Multiuser Beamforming Optimization

讲座名称：Model-driven Deep Learning for B5G Multiuser Beamforming Optimization

讲座人：郑淦 教授

讲座时间：12月16日16:00

讲座地点：腾讯会议直播（ID:711701511）

讲座人介绍：

郑淦博士现为英国拉夫堡大学教授，现为IEEE Fellow并担任IEEE Wireless Communications Letters 和IEEE Communications Letters编委。郑淦的研究主要关注无线通信网络的资源优化以及人工智能在通信系统中的应用，曾6次获得IEEE最佳论文奖, 包括 2013年IEEE Signal Processing Letters (该奖项引入后首次唯一获奖论文), 2015年GLOBECOM (国际通信领域两大顶级会议之一，通信信号处理 Symposium 唯一获奖论文)和2018年IEEE Technical Committee on Green Communications & Computing 最佳论文(唯一获奖论文)。他在权威国际期刊和主要国际会议分别发表学术论文87篇和50篇。学术引用总数超过8743次，单篇最高引用802次，h-index为45(Google Scholar)。

讲座内容：

Beamforming has been a key multi-antenna technique to improve the spectrum efficiency of 5G communications systems, but its optimisation is a difficult problem and therefore has not achieved its full potential. Traditional model-based numerical solutions are too complex and not effective in addressing model uncertainties. More recent data-based deep learning solutions face practical challenges such as sample efficiency, generalisation and poor performance in dynamic environments. In this talk, I will introduce our recent development in leveraging model-driven deep learning algorithms for the optimisation of beamforming. We will demonstrate that by properly incorporating available model knowledge in the neural network design, significant advantages can be achieved over state of the art for beyond 5G networks in terms of enhanced spectral efficiency, reduced complexity and channel estimation overhead, better generalisation and scalability.

主办单位：通信工程学院