Convergence acceleration during the 20th century (Q5932843)

Methods to accelerate the convergence of sequences of numbers or of vectors are very important in numerical mathematics. The author gives a survey of some of the most important such methods and their historical background. For sequences of numbers, the processes of Richardson-Romberg and Aitken's \(\Delta^2\)-process and Wynn's \(\varepsilon\)-algorithm are prominent; the latter can also be iterated. Always the asymptotic expansion of the error of the sequences to be accelerated is of primary importance. In the vector case, one can operate componentwise, or one can use \(y^{-1}= y/(y,y)\) to replace the inverse of a number \(y\), for example in the vector version of Wynn's \(\varepsilon\)-algorithm. The article ends with a list of 155 references.