Topic：On a Procrustes-based procedure for random tomography

Guest：Dr. Christian Friedrich

Time：10月26日（周三）下午15:00-17:00

Venue：2/F Seminar Room, Business School（商学院二楼研讨室）

Host：Prof. Peihua Zhao（赵佩华教授）

Language: English

Abstract：

We revisit the Procrustes-based matching procedure of Panaretos and Konis for the consistent estimation of landmark vectors in cryo-electron microscopy of a three-dimensional particle. Due to the severe level of ill-posedness of the problem, the basic object of interest is the Gram matrix containing the inner products of the landmark vectors, with that matrix defining the ‘shape’ of the particle. A key question is how the Gram matrix constructed in the first stage, which employs relatively ‘clear’ images with all landmarks present, may be updated by using further images which have one landmark missing.As a replacement for the Frobenius norm distance that Panaretos and Konis used in this step, we propose a nonsymmetric alternative. We argue that the expression thus obtained is not only simpler, but also accounts better for certain nonstandard geometric features in this matching problem.

Short Biography：

The speaker graduated in 2003 with a Ph.D. in Mathematical Statistics at the Australian National University.After two post-doctoral periods at the ANU and Hong Kong Baptist University, he was a lecturer in Statistics at Monash University from 2008 to 2011, and an adjunct until 2013. After a one-semester stay at Southeast University in 2012, he was an Associate Professor at the Department of Mathematics at Shantou University from 2013 to July 2016. His current research interests lie at the intersection of applied probability with a geometric flavour,

spatial statistics, and image analysis, especially in problems where non-Euclidean manifolds play a major role.  

欢迎全校师生参加！

商学院

2016-10-24