Finite sample properties of system identification methods

IEEE Transactions on Automatic Control - Tập 47 Số 8 - Trang 1329-1334 - 2002
M.C. Campi1, E. Weyer2
1Department of Electrical Engineering and Automation, University of Brescia, Brescia, Italy
2Department of Electrical and Electronic Engineering, University of Melbourne, Parkville, Australia

Tóm tắt

In this paper we study the quality of system identification models obtained using the standard quadratic prediction error criterion for a general linear model class. The main feature of our results is that they hold true for a finite data sample and they are not asymptotic. The main theorems bound the difference between the expected value of the identification criterion evaluated at the estimated parameters and at the optimal parameters. The bound depends naturally on the model and system order, the pole locations, and the noise variance, and it shows that although these variables often do not enter in asymptotic convergence results, they do play an important role when the data sample is finite.

Từ khóa

#System identification #Predictive models #Parameter estimation #Convergence #Costs #Noise generators #Automation

Tài liệu tham khảo

10.2307/1427698 10.1016/S0167-6911(99)00022-5 10.1109/18.556602 10.1109/18.650998 söderström soderstrom, 1988, System Identification 10.1007/978-1-4757-2440-0 vidyasagar, 1997, A Theory of Learning and Generalization 10.1109/9.148343 10.1016/S0005-1098(00)00039-X 10.1109/CDC.2000.914211 cherkassky, 1998, Learning From Data chen, 2000, Control Oriented System Identification An H$_\infty$ Approach 10.1109/9.580876 10.1016/S0005-1098(97)00007-1 ljung, 1999, System Identification&#x2014 Theory for the User 10.1109/18.661510 10.1016/S0167-6911(98)00005-X bosq, 1998, Nonparametric Statistics for Stochastic Processes&#x2014 Estimation and Prediction, 10.1007/978-1-4612-1718-3 10.2307/1427698 10.1109/9.774109