Symbolic-numeric sparse interpolation of multivariate polynomials (Q5902083)

Prony's and other methods are used for symbolic interpolation using multivariate polynomials. The work includes an error analysis and an analysis of both the stability and the sensitivity of the process with the use of bounds on generalized eigenvalues. This is all for floating-point arithmetic and fixed precision. The discussions of the sensitivity and stability, as well as the conditioning of the interpolation problem, are based on probability estimates. Algorithms and examples for the application of the analysis are presented too.