DocumentCode :
3200891
Title :
Parallel structured network for solving balanced realization problems
Author :
Li, Zhiqiang ; Sinha, Naresh K.
Author_Institution :
Dept. of Electr. & Comput. Eng., McMaster Univ., Hamilton, Ont., Canada
fYear :
1995
fDate :
5-7Jan 1995
Firstpage :
263
Lastpage :
267
Abstract :
Given a state space description of a linear system, the balanced realization problem is to find a similarity transformation for the state variables. This transformation must transfer both the system energy matrix and the output energy matrix into diagonal form. In this paper a novel parallel structured network (PSN) is constructed to solve this problem. The three possible forms of realization, which are defined in the literature can be obtained using the PSNs. Furthermore, the fourth one, defined and found here, is new, and has not been mathematically defined in earlier work
Keywords :
backpropagation; linear systems; neural nets; realisation theory; state-space methods; balanced realization problems; energy matrix; linear system; parallel structured network; similarity transformation; state space description; Control systems; Equations; Feedforward neural networks; Linear systems; Mathematical model; Multi-layer neural network; Neural networks; Observability; Reduced order systems; Symmetric matrices;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Industrial Automation and Control, 1995 (I A & C'95), IEEE/IAS International Conference on (Cat. No.95TH8005)
Conference_Location :
Hyderabad
Print_ISBN :
0-7803-2081-6
Type :
conf
DOI :
10.1109/IACC.1995.465831
Filename :
465831
Link To Document :
بازگشت