报告题目：Random re-normalization an approach for accerlerating singular value decomposition and inverse computations of nearly low rank matrices
时 间：2025年5月19日（星期一）10：20
地 点：腾讯会议：491-286-145
主 办：数学与统计学院, 分析数学及应用教育部重点实验室、福建省分析数学及应用重点实验室、福建省应用数学中心（福建师范大学）、福建师范大学数学研究中心
参加对象：感兴趣的老师和研究生
报告摘要：Fast solving singular value decomposition (SVD) and inverse of a nearly low-rank matrix has a wide scope of applications in scientific computing and data analysis. When the matrix size is large, a direct computation of SVD and inverse can be computationally unaffordable. One practical strategy to save the computation resources is embedding the characteristic of matrix into the matrix computation. In this talk we propose a random re-normalization procedure for simplifying SVD and inverse computations of nearly low rank matrix. The new method is built upon random recombination of columns of matrix, which can determine the nearly rank and approximate the basis of matrix simultaneously. The decomposition and inverse are then transferred to a lower order computation problem. These merits give the new method an edge to outperform the existing methods. The promising performance of the method is supported by both theory and numerical examples.
报告人简介：徐玮玮，现任南京信息工程大学教授，博士生导师。研究方向为矩阵计算理论与技术应用；和大数据基础算法研究。学士和博士毕业于华南师范大学，博士毕业后进入中科院数学与系统科学研究院博士后流动站工作。在综合性顶刊National Science Review, 和应用数学与人工智能国际权威期刊Mathematics of Computation, SIAM J. Optim., SIAM J. Matrix Anal. Appl.，IEEE Trasctions on Neural Networks and Learning Systems 等上发表学术论文40余篇;开发了线性判别软件包，成功用于图像，基因，质谱等数据降维和分类中。主持国家自然基金面上项目，青年项目和省部级基金项目共5项；2020年入选江苏省“青蓝工程”优秀骨干教师。2022年受聘国家天元数学西北中心“天元学者”。2022年获得首届粤港澳大湾区(黄埔)国际算法算例大赛冠军。