A quadratically convergent local algorithm on minimizing the largest eigenvalue of a symmetric matrix (Q1260789): Difference between revisions

An algorithm is proposed for minimizing the largest eigenvalue of an affine family of symmetric matrices (matrices of the form \(A_ 0 + \Sigma x_ iA_ i\) where \(A_ 0,A_ 1,\dots,A_ m\) are real symmetric matrices and \(x_ 0,x_ 1,\dots,x_ m\) are real numbers). Under certain assumptions it is shown that, if started close enough to the minimizer, the proposed algorithm converges to it quadratically. The algorithm can readily be extended to minimizing the largest eigenvalue of an affine family of Hermitian matrices. It has also been extended to minimizing sums of the largest eigenvalues of a symmetric Hermitian matrix. Some applications of the algorithm are given as well as a discussion of its computational complexity. There are numerical examples. The paper ends with a Matlab \(M\)-file.