Convergence of vectorial sequences. Applications (Q1338819)

A vectorial sequence of \({\mathbb{C}}^ d\), \((S_ h)_{h\geq 0}\) can be considered as the partial sum at \(t=1\) of the formal power series with coefficients in \({\mathbb{C}}^ d: {\mathbb{F}}(t)=\sum_{i\geq 0}\Gamma_ i t^ i\) with \(\Gamma_{i+1}=\Delta S_ i(=S_{i+1}-S_ i)\) for \(i\geq 0\) and \(\Gamma_ 0=S_ 0\). For such a vectorial series, vector Padé approximants have been defined and can be used for the study of the convergence of the sequence \((S_ h)_{h\geq 0}\) as ordinary Padé approximants and the epsilon algorithm for the scalar case, with the help of the Shanks transform. What gives rise to this analogous use is the expression as a ratio of two determinants of the vector Padé approximants. A generalisation of the Shanks transform is obtained. Applications to linear systems, eigenvalues, nonlinear systems are done. Algorithms, different from recursive computation of determinants, can be defined. Some numerical examples are given.