The inexact Newton-like method for inverse eigenvalue problem (Q1397977)

The authors use an iterative method based on the inexact-Newton like method to solve a finite inverse eigenvalue problem. The approach is reminiscent of coefficients or parameter identification. Given a basis of real symmetric matrices \(A_{i}\) for \(i=1,\) .., \(n\), the problem is to find a sequence of coefficients \(c_{i}\) such that the matrix \(\sum_{i=1}^{k}c_{i}A_{i}\) has a given set of prescribed eigenvalues\( \left\{ \lambda _{1},...,\lambda _{n}\right\} \). Note that starting with different choices of matrices \(A_i\) would lead to similar matrices, since in principle they would have the same spectra. The search for the vector \(\left( c_{1}, ..., c_{n}\right) ^{T}\) involves a variation of a Newton-like algorithm which solves the nonlinear equation \(g(x)=0.\) The solution \(x\in \mathbb{R}^{n},\) is approximated by iterations \(x^{k}\) defined by \(J\left( x^{k}\right) \left( x^{k+1}-x^{k}\right) =-g\left( x^{k}\right) \) where \(J\) is the Jacobian and a suitably chosen tolerance is used as a stopping criteria. Convergence analysis and numerical experiments are provided at the end of the paper.