zbMATH — the first resource for mathematics

A transformation approach to stochastic model reduction. (English) Zbl 0556.93057
This paper deals with the problem of obtaining reduced-order models for a given purely nondeterministic stationary stochastic process, generated by a minimal stochastic realization. A new approach is presented, based on an equivalence between the problem of solving Riccati equations and the canonical correlation analysis problem. This equivalence is developed using the concept of balanced stochastic realization. Also, asymptotic stability of the approximate reduced-order realization is established.
Reviewer: D.Nualart

93E03 Stochastic systems in control theory (general)
93E12 Identification in stochastic control theory
93E15 Stochastic stability in control theory
60G10 Stationary stochastic processes
93B20 Minimal systems representations
93B17 Transformations
Full Text: DOI