报告题目1:Tensor completion and its applications
报告人1:NG, Michael Kwok-Po(吴国宝)教授
报告人1单位:香港浸会大学数学系
时间:2018年11月2日15:00-16:00
地点:仙林校区教2-314室
主办单位:350VIP浦京集团、视觉认知计算与应用研究中心、科研院
报告内容1:In this talk, we discuss the recentdevelopment of tensor singular value decompositionand its applications.
报告人1简介:吴国宝教授于1995年在香港中文大学数学系获得博士学位,现为香港浸会大学数学系主任,系首席教授(chair professor),SIAM Fellow。2017年获“冯康科学计算奖”。作为一个应用数学领域的研究者,吴教授的研究领域主要包括数字图像处理,生物信息学,数据挖掘,运筹学以及科学计算等。在这些领域,吴教授已编著了5本书,发表SCI期刊论文200余篇。吴教授目前担任Journal of Computational and Applied Mathematics, Elsevier主编,同时他还担任SIAM Journal on Scientific Computing, Numerical Linear Algebra with Applications, Wiley-Blackwell, Multidimensional Systems and Signal Processing Springer等十多个著名国际杂志的编委。
报告题目2:Lanczos Method for Large-Scale Quaternion Singular Value Decomposition
报告人2:贾志刚教授
报告人2单位:江苏师范大学
时间:2018年11月2日16:00-17:00
地点:仙林校区教2-314室
主办单位:350VIP浦京集团、视觉认知计算与应用研究中心、科研院
报告内容2:In many color image processing and recognition applications, one of the mostimportant targets is to compute the optimal low-rank approximations to colorimages, which can be reconstructed with a small number of dominant singularvalue decomposition (SVD) triplets of quaternion matrices. All existing methodsare designed to compute all SVD triplets of quaternion matrices at first andthen to select the necessary dominant ones for reconstruction. This way costsquite a lot of operational flops and CPU times to compute many superfluousSVD triplets. In this paper, we propose a Lanczos-based method of computingpartial (several dominant) SVD triplets of the large-scale quaternion matrices.The partial bidiagonalization of large-scale quaternion matrices is derived byusing the Lanczos iteration, and the reorthogonalization and thick-restart techniquesare also utilized in the implementation. An algorithm is presented tocompute the partial quaternion singular value decomposition. Numerical examples,including principal component analysis, color face recognition, videocompression and color image completion, illustrate that the performance ofthe developed Lanczos-based method for low-rank quaternion approximation isbetter than that of the state-of-the-art methods.
报告人2简介:贾志刚博士,江苏师范大学教授、硕士生导师、第一批高层次人才队伍后备人选。主要研究方向为数值代数与图像处理,至今已在 SIAM J. Matrix Anal. Appl., J. Sci. Comput., BIT, J. Comput. Appl. Math.等国际知名期刊上发表学术论文20余篇,主持和参与国家自然科学基金项目3项和省部级科研项目3项。现兼职为江苏省工业与应用数学会理事、江苏省计算数学学会理事,美国Math Review评论员,SIMAX, Automatic, JCAM等学术期刊的审稿人。