An inexact Cayley transform method for inverse eigenvalue problems (Q4664407)

The paper deals with the numerical analysis of inverse eigenvalue problems. An inverse eigenvalue problem can be formulated as a system of nonlinear equations. The direct application of Newton's method to this system requires the costly computation of the eigenvalues of the approximating matrix in each iteration step. Using Cayley transforms and matrix exponential techniques an approximation of the Jacobian matrix can be computed. In order to improve the efficiency the authors suggest an inexact version of the Cayley transform method to solve inverse eigenvalue problems. It is shown that this method possesses superlinear convergence. Moreover, the method is well balanced between inner and outer iterations when solving the linear equation of Newton's method with a direct solver. Finally, the performance of the proposed numerical scheme applied to an inverse Toeplitz eigenvalue problem is shown.