On Minimizing the Special Radius of a Nonsymmetric Matrix Function: Optimality Conditions and Duality Theory
- 1 October 1988
- journal article
- Published by Society for Industrial & Applied Mathematics (SIAM) in SIAM Journal on Matrix Analysis and Applications
- Vol. 9 (4), 473-498
- https://doi.org/10.1137/0609040
Abstract
Let $A( x )$ be a nonsymmetric real matrix affine function of a real parameter vector $x \in \mathcal{R}^m $, and let $\rho ( x )$ be the spectral radius of $A( x )$. The article addresses the following question: Given $x_0 \in \mathcal{R}^m $, is $\rho ( x )$ minimized locally at $x_0 $, and, if not, is it possible to find a descent direction for $\rho ( x )$ from $x_0 $? If any of the eigenvalues of $A( x_0 )$ that achieve the maximum modulus $\rho ( x_0 )$ are multiple, this question is not trivial to answer, since the eigenvalues are not differentiable at points where they coalesce. In the symmetric case, $A( x ) = A( x )^T $ for all $x,\rho ( x )$ is convex, and the question was resolved recently by Overton following work by Fletcher and using Rockafellar’s theory of subgradients. In the nonsymmetric case $\rho ( x )$ is neither convex nor Lipschitz, and neither the theory of subgradients nor Clarke’s theory of generalized gradients is applicable. A new necessary and sufficient condition is given for $\rho ( x )$ to have a first-order local minimum at $x_0 $, assuming that all multiple eigenvalues of $A( x_0 )$ that achieve the maximum modulus are nondefective. The optimality condition is computationally verifiable and involves computing “dual matrices.” If the condition does not hold, the dual matrices provide information that leads to the generation of a descent direction. The result can be extended to the case where $\rho ( x )$ is replaced by the maximum real part of the eigenvalues of $A( x )$. The authors use the eigenvalue perturbation theory of Rellich and Kato, which provides expressions for directional derivatives of $\rho ( x )$. They also derive formulas for the codimension of manifolds on which certain eigenvalue structures of $A( x )$ are maintained; these are due to Von Neumann and Wigner and to Arnold. Finally, they discuss the much more difficult question of resolving optimality when $A( x_0 )$ has a defective multiple eigenvalue achieving the maximum modulus $\rho ( x_0 )$.
Keywords
This publication has 13 references indexed in Scilit:
- Computational methods for parametric LQ problems--A surveyIEEE Transactions on Automatic Control, 1987
- THE FORMULATION AND ANALYSIS OF NUMERICAL-METHODS FOR INVERSE EIGENVALUE PROBLEMSSIAM Journal on Numerical Analysis, 1987
- Semi-Definite Matrix Constraints in OptimizationSIAM Journal on Control and Optimization, 1985
- Geometrical Methods in the Theory of Ordinary Differential EquationsGrundlehren der mathematischen Wissenschaften, 1983
- Extremal eigenvalue problemsBulletin of the Brazilian Mathematical Society, New Series, 1978
- Output feedback stabilization by minimization of a spectral radius functional†International Journal of Control, 1978
- Generalized gradients and applicationsTransactions of the American Mathematical Society, 1975
- ON MATRICES DEPENDING ON PARAMETERSRussian Mathematical Surveys, 1971
- Isoperimetric eigenvalue problems in algebrasCommunications on Pure and Applied Mathematics, 1968
- On the rank of the reduced correlational matrix in multiple-factor analysisPsychometrika, 1937