Multivariate polynomial interpolation with perturbed data (Q5963406)

Polynomial interpolation is an important field in numerical analysis, as we often need to find suitable functions (here: polynomials) that meet given values at prescribed points. While this problem is not hard to solve in the univariate setting, it becomes immediately difficult in several dimensions. In this paper, the authors show how to compute suitable subspaces of polynomials from all polynomials of \(n\) variables with a stable scheme using QR decompositions of matrices. Moreover, it is shown how minimal normalised polynomials are computed in a least squares sense in order to deal with rounding errors.