CONVERGENCE ANALYSIS OF NEWTON-LIKE METHODS FOR INVERSE EIGENVALUE PROBLEMS WITH MULTIPLE EIGENVALUES

摘要

We provide in the present paper a corrected proof for the classical quadratical convergence theorem (i.e., Theorem 3.3 in Friedland, Nocedal, and Overton [SIAM T. Numer. Anal., 24 (1987), pp. 634-667]) of the Newton-like method for solving inverse eigenvalue problems with possible multiple eigenvalues. Moreover, as a by-product, our approach developed here can be extended to establish a similar convergence result for an inexact version of the Newton-like method with possible multiple eigenvalues, which is an extension of the corresponding inexact Newton-like method for the distinct case in Chan, Chung, and Xu [BIT Numer. Math., 43 (2003), pp. 7-20].