Implementation improvements and extensions of an ODE-based algorithm for structured low-rank approximation
摘要
In the framework of structured matrix low-rank approximation, we propose implementation improvements and extensions of a gradient system methodology that is based on the iterative integration of a system of ODEs. The improvements are based on numerical techniques for the computation of SVDs and rank-1 matrices projection. Some extensions of the numerical method to variations of the classical structured low-rank approximation problems are then proposed.