Non-commutative extrapolation algorithms (Q1334260): Difference between revisions

The right designant [\textit{A. Heyting}, Math. Ann. 98, 465-490 (1927; JFM 53.0120.01)] of a square matrix with elements in a skew field is the last number to be obtained in a process of reduction to triangular form. If \(A\) is of dimension two, its rows being in order \(A(1,1)\), \(A(1,2)\) and \(A(2,1)\), \(A(2,2)\), its right designant is \({\mathcal R}(A)= A(2,2)- A(1,2) A(1,1)^{-1} A(2,1)\). In the paper under review, the theory of designants is applied to certain transformations of sequences of members of a skew field. The proofs of a number of auxiliary results are not given, being referred back to an author's thesis. Among them is (Property 3 in the paper) a general result concerning designants to be obtained from bordered matrices, a special case of which may be formulated as follows. \(A\) being the above matrix, let \([A,N]\) be the matrix obtained by replacing the elements in the second column by \(N(1)\), \(N(2)\), let \([A,D]\) be obtained from \(A\) and \(D(1)\), \(D(2)\) in the same way, and let \([N,D]\) be formed from the columns \(N\) and \(D\) in that order. Then \[ {\mathcal R}([A,N]){\mathcal R}([A,D])^{-1}- N(1) D(1)^{- 1}= {\mathcal R}([N,D]){\mathcal R}([A,D])^{-1}. \] As is easily verified, this relationship is not always true. In cases in which proofs are given in detail, it is assumed (see the second and third formulae from below on p. 233 and formula (21) on p. 245) that designants are invariant with respect to row permutation. However, it is again easily verified that if \(A'\) is obtained from \(A\) above by interchanging rows, the relationship \({\mathcal R}(A)= {\mathcal R}(A')\) does not always hold. The paper contains three correct results: the first concerns an analogue of Sylvester's identity for designants; the second and third concern connections between numbers derived from related sequences by means of the \(\varepsilon\)-algorithm. The first is due to Heyting (loc. cit.) as the author acknowledges; the other two are due to the reviewer [The abstract theory of the epsilon algorithm, Centre de rech. math. Université de Montréal, CRM-74 (1971) and Upon a conjecture concerning a method for solving linear equations, and certain other matters, Math. Res. Center, Univ. of Wisconsin, MRC 626 (1966)].