Comparison of four algorithms accelerating the convergence of a subset of logarithmic fixed point sequences (Q1186616): Difference between revisions

The author compares four algorithms accelerating the convergence of the subset \(\text{LOGF}_1'\) of LOG. (LOG is the set of logarithmic sequences and \(\text{LOGF}_1'\) is the subset of fixed point sequences whose associated error \(e_n=S_n-S\) verifies \(e_{n+1}=e_n+a_2e^2_n+a_3e^3_n+\ldots\) where \(a_3\neq a^2_2\) and \(a_2<0\).) The algorithms are modifications of a certain version of \textit{P. Wynn's} \(\varepsilon\)-algorithm [Proc. Camb. Philos. Soc. 52, 663--671 (1956; Zbl 0072.33802)], a version of Aitken's \(\Delta^ 2\)-process, the iterated \(\theta_ 2\)-algorithm of \textit{S. Lubkin}'s \(W\)-transform [J. Res. Natl. Bur. Stand. 48, 228--254 (1952); MR 14.500)] and the \(\theta\)-algorithm of \textit{C. Brezinski} [C. R. Acad. Sci., Paris, Sér. A 273, 727--730 (1971; Zbl 0248.65004)]. Some results on the convergence rate are given and, also, several numerical examples.