站内搜索

学术报告(四)

办公室 2019-12-31 18:14

题   目:On the convergence of Jacobi-type algorithms for ICA 

报告人:李建泽 博士  (深圳市大数据研究院

时  间:141100-1140

 点:数信红楼302会议室

 

摘要:Jacobi-type algorithms for simultaneous approximate diagonalization of symmetric real tensors (or partially symmetric complex tensors) have been widely used in independent component analysis (ICA) because of the high performance. One natural way of choosing the index pairs in Jacobi-type algorithms is the classical cyclic fashion, while the other way is based on the Riemannian gradient in each iteration. In this presentation, I will mainly review our recent results in a series of papers about the weak convergence and global convergence of these Jacobi-type algorithms, under both of two pair selection rules. These results are mainly based on the Lojasiewicz gradient inequality. This is a joint work with Konstantin Usevich and Pierre Comon.

 

报告人简介:李建泽,于2013年毕业于南开大学陈省身数学研究所,获得博士学位,研究课题为算子系统的张量积理论。于20137月至20181月在天津大学数学学院工作,任职讲师。于20169月至20178月及20182月至8月在法国GrenobleGIPSA-Lab先后进行访问和博士后研究,研究课题为张量分解算法的收敛性分析。于20189月至20194月在加拿大Ryerson大学进行访问,研究课题为张量分解算法在信号处理中的应用。于20195月至今,在深圳市大数据研究院工作,任职研究科学家。