报告人：张绍良教授（Professor Shao-Liang Zhang，1983年吉林大学数学系计算数学专业毕业，1990年日本筑波大学工学研究科博士毕业，现为名古屋大学大学院工学研究科计算理工学专业教授。主要研究领域为数值计算方法，大规模矩阵计算的迭代方法和并行计算算法设计等。）

时间：2013年10月18日上午10:00-11:00

地点：数学系（致远楼）107室

摘要：Development of efficient solvers for solving linear systems Ax = b is one of the most significant issue on the field of computational science and engineering. The GMRES method of Saad and Schultz is a well-known solver as the typical Krylov subspace method for nonsymmetric linear systems. While the GMRES method show good convergence, it has difficulties in storage and computational costs. To remedy these difficulties, the restart technique is generally used as a practical choice. The restarted version of the GMRES method is called the GMRES(m) method. In this talk, we will focus on a relationship between the GMRES(m) method and the error equation and propose a new variant of the GMRES(m) method. Several numerical results will show that this new variant will be attractive.