Computing derivatives of eigensystems by the topological \(\epsilon\)- algorithm (Q1095585)

The topological \(\epsilon\)-algorithm of \textit{C. Brezinski} [Lect. Notes Math. 584 (1977; Zbl 0352.65003)] is an effective algorithm for accelerating the convergence of vector-valued sequences. Let \(\{S_ k\}\) be a sequence of vectors in \(C^ n\). The basic rule of the algorithm involves to set \(\epsilon_{-1}^{(k)}=0\), \(\epsilon_ 0^{(k)}=S_ k\), \(\epsilon^{(k)}_{2m+1}=\epsilon^{(k)}_{2m-1}+y/<y,\Delta \epsilon_{2m}^{(k)}>,\epsilon^{(k)}_{2m+2}=\epsilon_{2m}^{(k)} +\Delta \epsilon_{2m}^{(k)}/<\Delta \epsilon^{(k)}_{2m+1},\quad \Delta \epsilon_{2m}^{(k)}>,\) and then \(\epsilon_{2m}^{(k)}=e_ m(S_ k)\sim s,\) where y (\(\neq 0)\) is a vector, \(<, >\) is a bilinear form such as an inner product, and \(\Delta\) is the difference with respect to k. The author applies this method for computing partial derivatives of eigenvalues and eigenvectors of parameter-dependent matrices in conjunction with an iterative procedure of \textit{C. S. Rudisill} and \textit{Y. Chu} [Numerical methods for evaluating the derivatives of eigenvalues and eigenvectors, AIAA J. 13, 834-837 (1975)]. He shows the effectiveness through two numerical examples.