A gradient system approach for Hankel structured low-rank approximation

Rank deficient Hankel matrices are at the core of several applications. However, in practice, coefficients these noisy due to e.g. measurements errors and computational errors, so generically involved full rank. This motivates problem structured low-rank approximation. Structured approximation problems, general, do not have a global efficient solution technique. In this paper we propose local optimization approach based on two-levels iteration. Experimental results show that proposed algorithm usually achieves good accuracy shows higher robustness with respect initial approximation, compared alternative approaches.