应兰州大学数学与统计学院张国凤教授和梁兆正博士邀请,南京航空航天大学戴华教授,上海交通大学王增琦副教授将于2021年5月2日在线举办专题学术报告。
题目:Regularized Least Squares Methods for Locality Preserving Projection
时间:2021年5月2日(星期日)10:30
腾讯会议ID:303 625 584
报告摘要:Locality preserving projection (LPP), as a well-known technique for dimensionality reduction,is designed to preserve the local structure of the original samples which usually lieon a low-dimensional manifold in the real world. However, it suffers from the undersampledor small-sample-size problem, when the dimension of the features is larger than the numberof samples which causes the corresponding generalized eigenvalue problem to be ill-posed.To address this problem, in this talk we show that LPP is equivalent to a multivariate linearregression under a mild condition, and establish the connection between LPP and a LS problemwith multiple columns on the right-hand side. Based on the developed connection, wepropose two regularized least squares methods for solving LPP and discuss the applicationof these methods to image recognition.
报告人简介:戴华,1988年毕业于南京大学数学系计算数学专业,获博士学位。现任南京航空航天大学理学院教授、博士生导师。担任《高等学校计算数学学报》副主编、《Numerical Algebra, Control and Optimization》等刊物编委。长期从事数值代数、动力学反问题等方向的研究,先后承担了国家自然科学基金、江苏省自然科学基金等项目10多项,在大型线性方程组数值方法、矩阵特征值问题及其灵敏度分析、代数特征值反问题的理论与方法、矩阵方程与矩阵逼近及其在结构模型修正中的应用等方面取得了一系列研究成果,在国内外重要学术刊物上发表论文180余篇,出版著作《代数特征值反问题》和《矩阵论》。曾获江苏省科技进步奖一项。长期从事大学数学的教学,先后为本科生、硕士生和博士生讲授不同的数学课程20多门。先后承担多项教改项目,是国家双语教学示范课程“线性代数”和江苏省优秀研究生课程“矩阵论”的负责人,曾获江苏省教学成果一等奖。曾获霍英东教育基金会高等学校优秀青年教师奖、江苏省“红杉树”园丁奖,被评为江苏省高校教学名师、全国优秀教师,享受政府特殊津贴。
题目:Application of Splitting Preconditioners to the Fluid Flow Control Problems
时间:2021年5月2日(星期日)11:30
腾讯会议ID:303 625 584
报告摘要:Fluid flow control problems play an important role in industrial scientific applications.The optimal control of the Navier-Stokes equations is the most attractive one. On the basis of effective preconditioning for the Poisson control and Stoke control problems, we study the preconditioners for solving the Navier-Stokes control problem. Linearized by Picard’s iterations, Navier-Stokes control problem yields a sequence of large sparse structured linear systems. The coefficient matrices are in the two-by-two block form with square blocks. We take advantage of the block structure to exploit the preconditioners. The eigenvalue distribution of the achieved preconditioned matrix is illustrated by the norms of the submatrices. To avoid solving the saddle point subsystems in the preconditioning procedure, we propose a practical variant of the preconditioner. The numerical experiments show that the GMRES method with the proposed preconditioners is the efficient and effective solution for the Picard’s iterations with the variable viscosity coefficient. The iteration count of the preconditioned method is independent of mesh size and the regularization parameter.
报告人简介:王增琦,中国科学院数学与系统科学研究院计算数学与科学工程计算研究所博士,上海交通大学数学科学学院副教授。2007.7毕业于中国科学院数学与系统科学研究院计算数学与科学工程计算研究所,获博士学位。2007年到上海交通大学数学系工作,2011年晋升为副教授。先后获得上海交通大学晨兴青年教师奖C类与上海交通大学SMC青年教师奖B类。主要研究领域为数值线性代数,结构化大型系数方程组的迭代算法,预处理技术,最小二乘问题的数值求解。曾到日本国立情报学研究所及美国加州大学伯克利分校做访问学者。曾先后主持国家自然科学基金2项,发表论文十余篇。曾组织The Seventh China-Russia Conference on Numerical Algebra with Applications、Forum on frontiers of numerical algebra at Shanghai Jiao Tong University等学术会议。
甘肃省高校应用数学与复杂系统重点实验室
数学与统计学院
萃英学院
2021年4月30日