Gestructureerde niet-lineair systeem identificatie met behulp van kernel-gebaseerde methoden ,,

摘要

There are many different model structures in the system identification literature e.g. ARX, Output-Error methods, Box-Jenkins, state space, block oriented models, etc. In addition, different parametrizations can be used e.g. linear, polynomial, piecewise linear, etc. For nonlinear system identification, support vector machines and kernel methods have been successfully applied in the past for certain classes of model structures. In general, the different options for model structure and parametrization can be used when implementing kernel methods in their primal representation. However, the situation changes when working in the dual as the resulting models are no longer parametric. Therefore, it is challenging to incorporate prior knowledge about the structure of the system within a primal-dual optimization setting of kernel methods. Given this difficulty, and considering the intrinsic advantages of the dual representation, it becomes clear that this is an important and interesting challenge. The aim of this research is to advance in this area which is at the interface between nonlinear system identification and machine learning by combining and integrating the best of both paradigms and employing both parametric and kernel-based approaches with suitable regularization schemes.