System identification based on an integrated approach of sparsity, regularization and low rank approximation

Abstract

In this document applications of some theoretical and computational techniques are presented for the structured approximation of dynamic systems based on data. The research carried out in this article is focused on linear models with applications in engineering and science. Specifically, the regularization and sparsity properties are integrated into the parameter approximation with a focus on the low-rank approximation. The results are independent of a particular representation of the system and furthermore a partition of the data is not assumed as input-output. The aforementioned techniques are compared through some numerical simulations and the application in real data.