报告题目:A Class of Randomized Block Reflective Kaczmarz Method for Linear Systems with Automatic Restart

报告人：郑宁

报告时间：2022年11月22日 14:00-15:00

报告地点：腾讯会议：472-846-787

摘要：We propose novel deterministic and randomized block reflective Kaczmarz methods, which is based on the row partitions block Householder transformation, for solving linear systems. The convergence analysis on both consistent and inconsistent cases are discussed, which sheds light on why the reflective type method may converge slower than the projection type method. Subsequently, the proposed method is restarted automatically to further speed up the convergence. Numerical experiments on the synthetic random problems and image restoration problems show the efficiency of the proposed methods.

个人简历：郑宁博士，同济大学数学科学学院助理教授。同济大学应用数学本科，获同济大学数学系理学博士学位和日本综合研究大学院大学信息学博士学位，获上海市优秀博士毕业生和日本国立信息学研究所年度最佳研究等荣誉称号。曾在日本国立信息学研究所和日本理化学研究所担任特别研究员职务。主要从事约束不相容系统快速算法和美式期权定价数值方法研究，SIAM J. Matrix Anal. Appl., J. Comput. Appl. Math., J. Sci. Comput., Numer. Algorithms等国际专业杂志发表论文十余篇，且有数篇论文在投。多次受邀参加国际会议并作会议报告。

邀请人：张理评