Stabilized recovery and model reduction for multivariate exponential polynomials (Q6543077)

This paper discusses a numerically stabilized version of Prony's method to recover multivariate exponential polynomials from discrete samples and approximation of discrete data by exponential polynomials of a given structure. In this direction, an algorithm is introduced that considers this additional information and provides a backward error estimate for the algebraic recovery method SMILE. Furthermore, the paper presents a method for approximating data with an exponential polynomial sequence, which contributes to multivariate model reduction. The methods presented in the paper are based on linear techniques and in consequence can be performed numerically stable and efficient.