The joint bidiagonalization method for large GSVD computations in finite precision

发布者:王丹丹发布时间:2022-12-16浏览次数:399

The joint bidiagonalization method for large GSVD computations infinite precision——江苏省应用数学(中国矿业大学)中心系列学术报告


题目:The joint bidiagonalization method for large GSVD computations infinite precision

报告人:贾仲孝 教授单位: 清华大学 数学科学系

时 间:20221218日(周日)下午 15:30-16:30

腾讯会议:426-847-526

报告内容简介:

 

Abstract:

The joint bidiagonalization (JBD) method has been used to computesome extreme generalized singular values and vectors of a largeregular matrix pair $\{A,L\}$. We make a numerical analysis of theunderlying JBD process and establish relationships between it andtwo mathematically equivalent Lanczos bidiagonalizations in finite
 precision.
Based on the results of numericalanalysis, we investigate the convergence of the approximategeneralized singular values and vectors of $\{A,L\}$. The results show that,under some mild conditions,the semiorthogonality of Lanczos type vectors suffices todeliver approximate generalized singular values with the same accuracy
 as the full orthogonality does,
meaning that it is only necessary to seek for efficient
 semiorthogonalization
strategies for the JBD process. We establish a sharp bound
 for the residual norm of an approximate generalized singular value and
corresponding
 approximate right generalized singular vectors, which can reliably
estimate the residual normwithout explicitly computing the approximate right generalized
 singular vectors before the convergence occurs.

 贾仲孝简介:

1994年获得德国比勒菲尔德大学博士学位,清华大学数学科学系二级教授,第六届国际青年数值分析家--L. Fox奖获得者 (1993),国家百千万人才工程入选者 (1999)。现任北京数学会第十三届监事会监事长(2021.12—2026.12),曾任清华大学数学科学系学术委员会副主任 (2009—2021)2010年度何梁何利奖数学力学专业组评委,中国工业与应用数学学会(CSIAM) 第五和第六届常务理事(2008.9—2016.8),第七和第八届中国计算数学学会常务理事(2006.10—2014.10),北京数学会第十一和十二届副理事长(2013.12—2021.12),中国工业与应用数学学会(CSIAM) 监事会监事(2020.1—2021.10). 主要研究领域:数值线性代数和科学计算。在代数特征值问题、奇异值分解和广义奇异值分解问题、离散不适定问题和反问题的正则化理论和数值解法等领域做出了系统性的、有国际影响的重要研究成果,所提出的精化投影方法被公认为是求解大规模矩阵特征值问题和奇异值分解问题的三类投影方法之一。Inverse Problems,Mathematics of Computation, Numerische Mathematik, SIAM Journal on Matrix Analysis and Applications, SIAM Journal on Optimization, SIAM Journal on Scientific Computing等国际著名杂志上发表论文70研究工作41个国家和地区的900多名专家与研究人员在17部经典著作、专著和教材及730多篇论文中引用1300多篇次。引用的书目包括BaiDemmelDongarraRuhevan der Vorst五人编辑的Templates for the Solution of Algebraic Eigenvalue Problems: a Practical Guide  (2000)Golub & van Loan的经典著作Matrix Computations第三、第四版(19962013)Stewart的经典著作Matrix Algorithms II: Eigensystems (2001)Bjorck的专著Numerical Methods in Matrix Computations (2015)van der Vorst的专著Computational Methods for Large Eigenvalue Problems (2002)Trefethen & M. Embree的专著Spectra and Pseudospectra, The Behavior of Nonnormal Matrices and Operators (2005)Meurant & Tebbens的专著 Krylov Methods for Nonsymmetric Linear Systems (2020)QuarteroniSacco & Saleri的专著Numerical Mathematics (2000).