Coefficient-free adaptations of polynomial root-finders (Q814095)

This paper deals with the classical problem of polynomial root-finders. Two popular polynomial root-finders are widely known as Durand-Kerner's and Aberth's algorithms [\textit{O. Aberth}, Math. Comput. 27, 339--344 (1973; Zbl 0282.65037); \textit{E. Durand}, Solutions numériques des équations algébriques, Tome 1: Equations du type \(F(X)=0\), Racines d'un polynome, Paris: Masson et Cie., Éditeurs, VII, 327 p. (1960), VIII, 445 p. (1961; Zbl 0099.10801); \textit{I. O. Kerner}, Numer. Math. 8, 290--294 (1966; Zbl 0202.43605)]. They have quadratic and cubic local convergence, respectively, and very rapidly converge in practice. Both algorithms recursively update the current approximations \(s_1,\ldots,s_n\) to the \(n\) roots of a monic input polynomial \(c(\lambda)\) of degree \(n\) by computing the values \(c'(s_i)\) and/or \(c(s_i)\), for \(i=1,\ldots, n.\) In this paper, the author specifies how these computations can be performed where the input polynomial is given by the scaled values \(c(s_i)\) (and possibly \(c'(s_i)\)) at the initial set \(s_1,\ldots,s_n\), and where the coefficients of \(c(\lambda)\) are never computed. He also relates the results to approximating the eigenvalues of a matrix.