Transfer function interpolation remainder formula of rational Krylov subspace methods
share
Rational Krylov subspace projection methods are one of successful methods in MOR, mainly because some order derivatives of the approximate and original transfer functions are the same. This is the well known moments matching result. However, the properties of points which are far from the interpolating points are little known. In this paper, we obtain the error's explicit expression which involves shifts and Ritz values. The advantage of our result over than the known moments matches theory is, to some extent, similar to the one of Lagrange type remainder formula over than Peano Type remainder formula in Taylor theorem. Expect for the proof, we also provide three explanations for the error formula. One explanation shows that in the Gauss-Christoffel quadrature sense, the error is the Gauss quadrature remainder, when the Gauss quadrature formula is applied onto the resolvent function. By using the error formula, we propose some greedy algorithms for the interpolatory H_∞ norm MOR.
READ FULL TEXT
