Spectral simplex method (Q263213)

This paper considers the problems of maximization and minimization of the spectral radius for nonnegative matrices with independent row uncertainties. The author proves necessary theoretical results on on-row corrections of nonnegative matrices. The spectral simplex methods for maximizing and minimizing the spectral radius are proposed. Their non-cyclicity and finite convergence is established. Statistics of the number of iterations performed by the algorithm for randomly generated uncertainty sets are presented. The author makes further an extension to general convex non-polyhedral uncertainty sets. Applications are considered to several areas including optimizing the spectral radius of a graph, constructing a productive economy in the Leontief model, stability of positive linear switching systems, and finite difference equations with nonnegative coefficients.