Orthogonal polynomials, biorthogonal polynomials and spline functions (Q2659755): Difference between revisions

From MaRDI portal
RedirectionBot (talk | contribs)
Changed an Item
Changed an Item
Property / describes a project that uses
 
Property / describes a project that uses: DLMF / rank
 
Normal rank

Revision as of 22:48, 29 February 2024

scientific article
Language Label Description Also known as
English
Orthogonal polynomials, biorthogonal polynomials and spline functions
scientific article

    Statements

    Orthogonal polynomials, biorthogonal polynomials and spline functions (English)
    0 references
    0 references
    26 March 2021
    0 references
    For orthogonal and bi-orthogonal polynomials, there are several important concepts that the authors of this article succeed to generalise. Among them are so-called generating functions and the Rodrigues' formula. Concretely, the authors wish to replace the expressions \(f_m\) in the Rodrigues' formula which amount to weight-functions \(\omega\) times some simple powers (often, monomials or even constants, or at most quadratics anyway, that are then taken to \(m\)-th powers) in the classical cases, by B-splines. The \(f_m\) are assumed to be decaying exponentially, as \(O(\exp(-Bx-\epsilon x))\). As a result, their Fourier transforms become analytic in an open strip of size \(B\). For comparison, for Laguerre polynomials for example, the \(f_m\) would be multiples of an exponential weight function times \(x^m\), for Hermite weight function \(\exp(-x^2)\) (times \(1^m\)), and these are now replaced by B-splines. Both uniform B-splines (pen-ultimate section) and non-uniform knots (last section of the article) are treated, and there is even a further generalisation where the B-splines are defined by repeated convolution not starting from a characteristic function of an interval (the classical case), but a less restricted function as an initial value. The Fourier transforms of those ``generalised B-Splines'' alter in that the usual powers of sinc-functions get another factor of the Fourier transform of the starting (initial) function. The result is called a generalised Rodrigues' formula and the needed generating function is called a generalised generating function. In order to form the generalised generating function, Fourier transforms for the \(f_m\) are needed which would be simply convolutions of weight functions with some derivatives of \(\delta\)-functions in the classical case (i.e., derivatives of the weight function). This reformulated Rodrigues' formula \(\mu_m=(-1)^m f_m^{(m)}\) is equivalent to considering the \(\mu_m\)s as bi-orthogonal to a sequence of orthogonal polynomials \(Q_m\) that are one factor of the expansion coefficients of the generalised generating function; the other factor being the Fourier transforms of the \(f_m\): \[\exp(xz)=\sum\nolimits_0^\infty Q_m(x) z^m \hat f_m(\mathrm{i}z).\] The main results guarantee the existence of the \(Q_m\) polynomials and the well-definedness of the generalised generating function (these results are, in particular, Theorem~2.3 and Theorem~3.1 and -- for the most general case -- Theorem~4.3).
    0 references
    orthogonal polynomials
    0 references
    biorthogonal polynomials and spline functions
    0 references
    generalized generating functions
    0 references
    generalized Rodrigues formula
    0 references
    uniform \(B\)-splines
    0 references
    refinable functions
    0 references
    nonuniform \(B\)-splines
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references