Computing Hamiltonian Schur Form of Hamiltonian Matrices

发布者:系统管理员发布时间:2012-12-26浏览次数:2359

报告题目: Computing Hamiltonian Schur Form of Hamiltonian Matrices
报 告 人: Prof Delin CHU
  Department of Mathematics, National University of Singapore
报告时间: 12月27日下午3:30pm开始
报告地点: 九龙湖数学系第一报告厅
相关介绍: Let M be a 2n-by-2n Hamiltonian matrix with no eigenvalues on the imaginary
axis. Then there is an orthogonal-symplectic similarity transformation of M to 
Hamiltonian Schur form, revealing the spectrum and stable invariant subspace 
of M. This was proved by C. C. Paige and C.Van Loan in a paper published in 
1981. The proof given in that paper was nonconstructive. Ever since, the 
problem of developing a structure-preserving and backward-stable algorithm 
with complexity O(n^3) to compute the Hamiltonian Schur form of a 2n-by-2n 
Hamiltonian matrix proved difficult to solve however, so much so that it came to 
be known as Van Loans curse.
In this talk we will introduce a new method that may meet these criteria for 
computing the Hamiltonian Schur form of a 2n-by-2n Hamiltonian matrix M 
without purely imaginary eigenvalues. The new method is structure-preserving 
and is of complexity O(n^3). It is implemented using orthogonal-symplectic 
transformations only and many numerical results demonstrate that it performs 
well and is backward stable in cases where there are no eigenvalues too close
to the imaginary axis.