Preconditioned GMRES methods for the least square problems
報告人簡介
殷俊鋒，同濟大學教授，創新創業學院副院長。主持和參與國家級和省部級以上科研項目10余項，發表論文60余篇，指導多項團隊在各類創新大賽中獲獎。學術兼職包括中國工業與應用數學學會副秘書長，中國創造學會副秘書長等。參與國家教學成果二等獎1項和上海市教學成果獎2項。
內容簡介
After reviewing the numerical solution of large sparse least square problems, we propose Kaczmarz-type preconditioned GMRES method, as well as sketching techniques. Theoretical analyses is given to guarantee the convergence and parameter tuning strategies are studied in details. Numerical experiments on least square problems further verify the efficiency of the preconditioners and show that the proposed method is superior to the existing preconditioned GMRES method.
