报告题目：Recursive Algorithms for Parameter Estimation with Adaptive Quantizer

报告人：游科友 清华大学

报告时间：2015年6月15日（周一）上午10:00

报告地点：电气馆A101学术报告厅

报告人简介

Keyou You was born in Jiangxi Province, China, in 1985. He received the B.S. degree in statistical science from Sun Yat-sen University, Guangzhou, China, in 2007 and the Ph.D. degree in electrical and electronic engineering at the Nanyang Technological University (NTU), Singapore in 2012.

From June 2011 to June 2012, he was with the Sensor Network Laboratory at NTU as a Research Fellow. Since July 2012, he has been with the Department of Automation, Tsinghua University, China as an Assistant Professor. He held visiting positions at The Hong Kong University of Science and Technology, NTU and University of Melbourne. His research interests include networked control and estimation, distributed algorithms, and sensor network.

Dr. You received the Guan Zhaozhi best paper award at the 29th Chinese Control Conference in 2010, and an IBM China Faculty Award in 2014. He was selected to the national “1000-Youth Talent Program” of China in 2013.

报告内容：

This talk studies a parameter estimation problem of networked linear systems with fixed-rate quantization. Under the MMSE criterion, we derive a recursive estimator of stochastic approximation type, and establish the necessary and sufficient condition for its asymptotic unbiasedness. This motivates to jointly design an adaptive quantizer and an estimator, whose strong consistency, asymptotic unbiasedness, and asymptotic normality are proved.

Using the Newton based and averaging technique, we obtain two accelerated recursive estimators with a fastest convergence speed of O(1/k), and exactly evaluate the quantization effect on the estimation accuracy. Under the Gaussian noise and any fixed-rate constraint, an optimal quantizer and accelerated estimators are co-designed to attain the minimum CRLB. All the estimators share identical computational complexity as the major gradient algorithms with un-quantized observations, and can be easily implemented.