An algorithm for computing the numerical radius (Q4344265)

Let \(A\) be an \(n\times n\) matrix with complex entries; the numerical range \(F(A)\) is given by \[ F(A)=\{z^*Az:z\in\mathbb{C}^n,\;z^*z=1\} \] and the numerical radius is therefore defined by \[ r(A)=\max\{|w|: w\in F(A)\}. \] This is an important quantity in the analysis of convergence of numerical processes. Using a technique which is analogous to a bisection method for computing the distance to instability in control problems the authors give an iterative process which successively refines lower and upper bounds for the numerical radius. The presented method is essentially for matrices of dimension less than 300.