Quasi-linear sequence transformations and their construction (Q1612434): Difference between revisions

Algebraic properties of functions of many real variables are presented in terms of partial differential equations satisfied by the functions. \( {\mathbf x } \) is the sequence \( x(0), \ldots , x(k) \); \textbf{1} is the sequence of \( k+1 \) unit terms; with real \( a \) and \( b \), \(a{\mathbf x}, {\mathbf x}/a \) and \({\mathbf x} \pm b \) are defined in terms of operations upon successive components; with \(0 \leq i \leq k \), \([{\mathbf x} \setminus i] \) is obtained by deleting \(x(i) \) from \( {\mathbf x} \); \( ( .,.) \) denotes a scalar product; \({\mathcal D} \) being the sequence of partial difference operators with respect to the components of \( {\mathbf x } \), Aronhold's operator \(({\mathbf e}, {\mathcal D}) \) is written as \(({\mathbf e} |{\mathcal D}) \). (i) The function \(F\) is homogeneous (i.e. \( F(a{\mathbf x}) = a F({\mathbf x}) \)) if and only if \( ({\mathbf x} |{\mathcal D}) F({\mathbf x}) = F({\mathbf x}) \). (ii) \(F\) is translative (i.e. \(F({\mathbf x} + b) = F({\mathbf x}) + b \)) if and only if \(({\mathbf 1}|{\mathcal D}) F({\mathbf x}) = 1 \). (iii) \(F\) is homogeneous if and only if either \(F({\mathbf x}) = ({\mathbf e}, {\mathbf x}) \) for some constant sequence \( {\mathbf e} \) compatible with \( {\mathbf x} \) or \( F({\mathbf x}) = x(i) \phi ([{\mathbf x} \setminus i]/x(i)) \) for some \( i \) in the range \( 0 \leq i \leq k \) such that \( x(i) \neq 0 \), \( \phi \) being regular (i.e. \( C^{1} \)). (iv) \( F \) is translative if and only if \( F({\mathbf x}) - x(i) = \psi ([{\mathbf x} \setminus i] - x(i)) \) for some \( i \) within range and regular \( \psi \). (v) \( F \) is invariant with respect to translation (i.e. \( F({\mathbf x} + b) = F({\mathbf x}) \)) if and only if \( F({\mathbf x}) = \theta ([{\mathbf x} \setminus i] - x(i)) \) for similar \( i \) and \( \theta \). (vi) \( F \) is both homogeneous and translative if and only if either \( F({\mathbf x}) = ({\mathbf e},{\mathbf x}) \) with \(({\mathbf e},{\mathbf 1})\) for some \({\mathbf e}\) as in (iii) or \( F({\mathbf x}) - x(p) = \{ x(i) - x(p) \} h(\{[{\mathbf x} \setminus i] - x(p)\}/ \{x(i)-x(p)\}) \) for \( i,p \) within range with \( x(i) \neq x(p) \), \( h \) being regular. These results are extended to the case in which \( {\mathbf x} \) is partitioned into an initial subsequence \( {\mathbf y} \) and complement \( {\mathbf z} \) and translativity is expressed in terms of a relationship of the form \( F({\mathbf y}+b,{\mathbf z}) = F({\mathbf x})+b \). Illustrations of the theory are provided by functions \( F \) having convergence acceleration properties.