Progressive stable interpolation (Q1370330)

In contrast to the ordinary interpolation problem, the situation where ``one datum is given after the other'' requires ``progressive'' interpolation schemes which lead to recursive algorithms. It is well known that those are numerically unstable if the spectral radius of the recursion matrix is greater than one. The authors present two approaches to construct stable numerical algorithms for the progressive interpolation problem. The first approach is based on parameter dependent basis functions. Several examples are given for \(C^1\) and \(C^2\) progressive interpolation methods using polynomials, splines, rational functions, exponentials and square root functions. In the \(C^2\) case, pairs of parameters are identified for a good smoothness of the interpolant, respectively for a good stability of the recursion. Choosing the parameters from the convex combination of such points is suggested. The second approach, which can be combined with the first, is based on parameter representations of the graph of the interpolant. Then the condition on the spectral radius of the iteration matrix defines conditions for the connection matrix which represents the geometric continuity of the graph. For various choices of basis functions, shape effects from the parameters are demonstrated numerically.