演講者：李哲榮教授

單　位：國立清華大學資訊工程系

日　期：2009年5月27日　PM 14:30

地　點：國立高雄大學理學院408室

講　題：The residual Arnoldi method for large eigenvalue problem 

摘　要：

Although there are good algorithms for solving small dense eigenvalue problems, the large-scale eigenproblem has many open issues. The major difficulty faced by existing algorithms is the tradeoff of precision and time, especially when one is looking for interior or clustered eigenvalues. In this talk, a new method, called the residual Arnoldi method, will be presented. This method has the desirable property that certain intermediate results can be computed in low precision without effecting the final accuracy of the solution. Thus we can reduce the computational cost without sacrificing accuracy.