题 目：A Precise Analysis of Covariance-based Activity Detection in Massive MIMO Systems
报告人：中国科学院 马俊杰副研究员
时 间：8月25日上午 10:00-11:00
地 点：交叉二号楼 B6007
摘 要：This talk is motivated by the activity detection problem for massive random access, which has attracted a lot of attention in the communications society. The activity detection problem can be formulated as a common support recovery problem for jointly sparse signals, and approximate message passing (AMP) algorithms have been proposed to tackle the problem. AMP was believed to achieve near-optimal performance for this problem but recent studies suggest that a covariance-based method can perform much better, especially when the number of antennas at the base station is large. Here we present a precise analysis of the recovery condition in the ideal exact-covariance setting. The results suggest that the number of users supported by the (ideal) covariance method is much larger than AMP. We also present some numerical comparisons of AMP and covariance-based algorithms under practical settings.
简 历：马俊杰，中国科学院数学与系统科学研究院优秀青年副研究员。2010年本科毕业于西安电子科技大学，2015年在香港城市大学取得博士学位。曾于香港城市大学、哥伦比亚大学和哈佛大学从事博士后研究。研究兴趣包括信号处理、无线通信、信息论、机器学习等，近年来主要关注无线通信中的高维信号估计问题。曾主持自然基金青年项目并参与中科院先导科技专项等科研项目，目前担任中国运筹学会青年工作委员会副秘书长。
