Recursive algorithms for vector extrapolation methods (Q1105975): Difference between revisions

Let \(\mu^ m_ n\); \(m,n=0,1,2,..\). be given complex numbers, and let \(b_ n\), \(n=0,1,2,..\). be a sequence of either complex numbers or vectors in a vector space B over the complex number-field. Define \(G_ k^{n,m}\) to be the determinant of \(\{[\mu^ j_{\ell}]/m\leq j\leq m+k- 1\), \(n\leq \ell \leq n+k-1\}\), and \(f_ k^{n,m}(b)\) to be the determinant of the \((k+1)\)-matrix whose \(\ell\)-th \((1\leq \ell \leq k+1)\) column consists of \(\{b_{n+\ell -1}\), \(\mu^{m+j-1}_{n+\ell - 1}\}_{1\leq j\leq k}\). The authors develop recursive algorithms for computing \[ T_ k^{n,m}(b)=f_ k^{n,m}(b)/G_ k^{n+1,m}\quad and\quad S_ k^{n,m}(b)=f_ k^{n,m}(b)/f_ k^{n,m}(I)=T_ k^{n,m}(b)/T_ k^{n,m}(I), \] especially for the following three cases, where I means the sequence \(I_ n=1\) for all n. (1) Recursive computation when m is fixed. (2) Same one when m-n is fixed. (3) Same one when n is fixed and m varies. They present efficient recursive algorithms described in a Pascal-like language with the discussion on the operation counts and storage requirements. They also give the relations with the implementation of various extrapolation techniques such as the modified minimal polynomial extrapolation of the authors and \textit{D. A. Smith} [SIAM J. Numer. Anal. 23, 178-196 (1986; Zbl 0596.65016)] or the topological epsilon algorithm of \textit{C. Brezinski} [Lect. Notes Math., 584, (1977; Zbl 0352.65003)]. The recursions in the case (2) and (3) seem to be new, and efficient for the acceleration of convergence for vector-sequences.