Report Title: On Two Matrix Optimization Problems: Correlation vs Euclidean Distance
Reporter: QI Houduo (School of Mathematical Sciences, University of Southampton, UK)
Report time: July 29, 2016 AM 9:00-10:00
Location: A1101#room, Innovation Park Building
Contact: Prof. ZHANG Liwei (tel: 84708351-8118)
Abstract: Matrix optimization has recently taken a new shape from a numerical perspective, where Semismooth Newton-CG has played an essential role. In this talk, we review two important matrix optimization problems of Correlation and Euclidean distance matrices. We reveal their striking similarities between them and we also emphasize their distinguishing features. We will pay a particular attention to the key problem structures that render the semi-smooth Newton-CG an efficient method for them. We then demonstrate a few of important applications, including the spherical embedding of high-dimensional data. This talk is based on the research with a number of collaborators over the past few years.
Brief introduction to the reporter:
1. 1996 PhD in Operational Research, Chinese Academy of Science, Beijing, China.
2. 1993 MSc in operational Research, Qufu normal University, China
3. 1990 BSc in Probability and Statistics, Peking University, China.
1. 2010-present. Associate professor in Operational Research, School of Mathematical Sciences, University of Southampton,UK.
2. 2004 -2009. Lecturer and then Senior Lecturer in Operational Research, School of MathematicalSciences, University of Southampton, UK.
1. Associate Editor of Asia-Pacific Journal of Operational Research
2. Associate Editor of Mathematical Programming Computation