*主持人：李韬 教授

*讲座内容简介：

With the development of network technology, distributed adaptive filtering (or estimation) algorithms have been widely used in many practical situations. Although the distributed filtering algorithms can be applied to a wide class of signals and requires no assumptions on the statistics of regression vectors and measurements, when it comes to stability and performance evaluation, some independency or stationarity conditions are necessarily required to carry out the theoretical analysis. In this talk, we will show that both the stability and the tracking performance bounds of a class of distributed adaptive filtering and estimation algorithms can be established under a general cooperative information condition, which needs neither independence nor stationarity of the system signal. We will further show that our information condition is actually a necessary one for a wide class of stochastic signals with decaying dependence. Furthermore, the weakest possible information condition also implies that the distributed adaptive filter can work well even if any individual filter is not stable due to lack of necessary information (i.e., the covariance matrix for each individual regressor is degenerate), which is a natural property for distributed algorithms but has not been justified rigorously in the existing literature.

*主讲人简介：

谢思宇，女，电子科技大学教授，博导，国家级青年人才计划入选者。1991年出生于四川省遂宁市，2013年获得北京航空航天大学理学学士学位，2018年获得中国科学院数学与系统科学研究院理学博士学位，导师为郭雷院士，2019-2022年在美国韦恩州立大学从事博士后研究工作，合作导师为王乐一教授。近年来以多个体网络系统为研究对象，瞄准分布式滤波、估计、优化及其在电力系统中的应用等关键科学问题展开研究，发表SCI期刊论文20余篇，包括控制方向的顶级期刊IEEE TAC、Automatica、SIAMCON，和电力系统方向的顶级期刊IEEE TSG、IEEE TITS、IJEPES等。曾获得中国科学院大学优秀博士学位论文、 IEEE CSS Beijing Chapter 青年作者奖、博士生国家奖学金等荣誉和称号。