A statistical learning perspective on switched linear system identification

Hybrid systems form a particularly rich class of dynamical based on the combination multiple continuous subsystems and discrete mechanism deciding which one these is active at given time. Their identification from input–output data involves nontrivial issues that were partly solved over last twenty years thanks to numerous approaches. However, despite this effort, estimating number modes (or subsystems) hybrid remains critical open issue. This paper focuses switched linear proposes an analysis their statistical learning theory. leads new theoretically sound bounds prediction error models hand, practical method for estimation other hand. The latter inspired by structural risk minimization principle developed in model selection. proposed conducted under various assumptions regularization schemes algorithms are presented. Numerical experiments also provided illustrate accuracy method.