Quadratic and cubic spline interpolation (Q788928)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Quadratic and cubic spline interpolation
scientific article

    Statements

    Quadratic and cubic spline interpolation (English)
    0 references
    0 references
    1984
    0 references
    Let (k,m) denote the class of k degree interpolating polynomial splines whose interpolation conditions are given by mth derivative values. Here we study the (2,0) and (3,1) problems. For \(-\infty<a<b<+\infty\) and for any positive integer \(n\geq 2\), let \(\Delta:a=x_ 0<x_ 1<...<x_ n=b\) denote a partition of [a,b] with knots \(x_ i\) and steps \(h_ i=x_{i+1}-x_ i\). Let S\(p(\Delta\),k) denote the class of k degree polynomial splines on \(\Delta\). (2,0) interpolation problem: Find s(x)\(\in Sp(\Delta,2)\) such that \(s(x_ i)=f_ i (i=0,1,...,n)\), \(s'(x_ j)=f_ j\!' (j=0\) or n). (3,1) interpolation problem: Find s(x)\(\in Sp(\Delta,3)\) such that \(s'(x_ i)=f_ i\!' (i=0,1,...,n)\), \(s(x_ j)=f_ j (j=0\) or n), \(s''(x_ j)=f_ j\!'' (j=0\) or n). The interpolation problems above are well-defined. Define the functionals \[ J[f^{(k-1)}]=\sum^{n-1}_{i=0}J_ i[f^{(k-1)}]\quad(k=2,3), \] \[ J_ i[f^{(k-1)}]=\int^{x_{i+1}}_{x_ i}[f^{(k-1)}(x)+f^{(k- 1)}(x_ i+x_{i+1}-x)]^ 2dx,\quad(i=0,1,...,n-1). \] We have the first and second integration relationships and the minimization properties. Let \(\| f^{(j)}\| =\max_{x\in [a,b]}| f^{(j)}(x)|, V^ b_ a(f^{(j)})=\int^{b}_{a}| df^{(j)}|, M_ j=\| f^{(j)}\| +V^ b_ a(f^{(j)}), h=\max_{0\leq i\leq n-1}h_ i\), \(\max h_ i/\min h_ i\leq \beta_ 0, (\sum^{n-1}_{j=1}| h_ j-h_{j+1}|)/h\leq \beta_ 1\), \(\max_{0\leq m\leq n-1}(\max_{m<1}(\sum^{i}_{j=m+1}| h_ j- h_{j+1}|)/h_ m\), \(\max_{i<m}(\sum^{m}_{j=i+1}| h_ j- h_{j+1}|)/h_ m\leq \beta_ 2, R(f,x)=f(x)-s_ f(x),\) where \(s_ f(x)\in Sp(\Delta,k)\) is the solution of the (k,k-2) interpolation problem for f(x). We have the following error estimates. If \(f(x)\in C^ 3[a,b]\) with f'''(x) of bounded variation, then for the solution of the (2,0) problem and its derivatives we have \(\| R^{(j)}(f,x)\| \leq c_ jM_ 3h^{3-j} (j=0,1,2)\), where \(c_ 0=(4+3\beta_ 1)/36,\quad c_ 1=(4+3\beta_ 1)/18,\quad c_ 2=(5+\beta_ 0+2\beta_ 2)/6.\) If \(f(x)\in C^ 4[a,b]\) with \(f^{(4)}\) of bounded variation, then for the solution of the (3,1) problem and its derivatives we have \[ \| R(f,x)\| \leq c_ 0(b-a)M_ 4h^ 3,\quad \| R^{(j+1)}(f,x)\| \leq c_ jM_ 4h^{3-j}\quad(j=0,1,2), \] where \(c_ 0,c_ 1,c_ 2\) are the same as above.
    0 references
    quadratic spline
    0 references
    cubic spline
    0 references
    spline interpolation
    0 references
    variational properties
    0 references
    interpolating polynomial splines
    0 references

    Identifiers