报告摘要: This talk has a technical part and a non-technical part.

The technical part is an overview of our new results on applying information geometry on statistical inference problems. We explain some limitations of the conventional information theoretic analysis, and present a method to go beyond such limits. This involves a new formulation of the inference problems, a new set of information metrics, and new operational meanings of these metrics in the context of data processing. We argue that our approach is a fundamental extension of the conventional approaches, and a key to put a systematic framework to the new machine learning type of problems.

In the non-technical part of this talk, we try to report some new trends in the development of the theory for information technology, and the difference between an EE-style and a CS-style research. This reflects our personal tastes in picking research problems and in driving the direction of fundamental research, with the hope of motivating more students in pursuing research results from a diverse range of directions.

 

Lizhong Zheng received the B.S and M.S. degrees, in 1994 and 1997 respectively, from the Department of Electronic Engineering, Tsinghua University, China, , and the Ph.D. degree, in 2002, from the Department of Electrical Engineering and Computer Sciences, University of California, Berkeley. Since 2002, he has been working in the Department of Electrical Engineering and Computer Sciences, where he is currently a professor of Electrical Engineering and Computer Sciences. His research interests include information  theory, wireless communications, and statistical inference. He received the IEEE Information Theory Society Paper Award in 2003, and NSF CAREER award in 2004, and the AFOSR Young Investigator Award in 2007。