Geometric approach to the parallel sum of vectors and application to the vector \(\varepsilon \)-algorithm (Q457039)

Geometric interpretations of Aitken's \(\Delta^2\) algorithm and Wynn's \(\varepsilon\)-algorithm are derived. The author begins by analyzing the geometry of the parallel sum of complex numbers, where the parallel sum \(a:b\) is defined by \[ a:b = \left( a^{-1} + b^{-1}\right)^{-1}. \] Since the parallel sum is prominent in the Aitken's \(\Delta^2\) algorithm and Wynn's \(\varepsilon\)-algorithm, the geometric interpretation of the parallel sum is then used to provide a geometric interpretation of these two algorithms. For Aitken's \(\Delta^2\), if \((x_n)\) is the original real-valued sequence and \(( (Tx)_n )\) is the sequence obtained from Aitken's \(\Delta^2\), then \((Tx)_n\) is the center of the homothety transforming simultaneously \(x_n\) into \(x_{n+1}\) and \(x_{n+1}\) into \(x_{n+2}\). A result is also presented for \(\Delta^2\) applied to vector-valued sequences. In the case of Wynn's \(\varepsilon\)-algorithm, it is shown that, for a initial sequence of vectors \((s_n)\) and transformed sequence \((\varepsilon_2^{(n)} )\), \(\varepsilon_2^{(n)}\) is the focus of the Bézier parabola controlled by \(s_n, s_{n+1}\), and \(s_{n+2}\).