Computation of the eigenvalues of convexity preserving matrices (Q1021865)

Matrices which are \(r\)-convexity preserving for \(r=0,1,\dots,k\) arise in many practical and theoretical problems. A source of many examples is provided by the collocation matrices of \(r\)-convexity preserving systems of functions. Based on results of \textit{J. M. Carnicer}, \textit{M. García-Esnaola} and \textit{J. M. Penã} [Comput. Aided Geom. Des. 13, No. 2, 179--197 (1996; Zbl 0900.68404)], in this paper, a direct method for computing the \(k\) largest eigenvalues of the \(r\)-convexity matrices is proposed. For an \(n\times n\) matrix, the computational cost is \(O(kn^2)\).