Title :
Probability-based analysis and parametric synthesis of dynamic neural systems
Author_Institution :
Dept. of Inf. Technol. & Control, Belarussian State Univ. of Informatics & Radioelectronics, Minsk, Belarus
Abstract :
Dynamic neural networks are considered as specific class of nonlinear dynamic systems with neural controller in the control contour. It is explained that classical methods of analysis of nonlinear control systems are applicable for the class of systems under consideration. Specific features of application of statistical linearization method for analysis and parametrical synthesis of dynamic neural systems are considered. Analysis of system is interpreted as determination of second statistical moments of various coordinates, for example, of error signal. Synthesis is interpreted as determination of optimal parameters of neural controller, in terms of minimization of root-mean-square error.
Keywords :
control system analysis; control system synthesis; least mean squares methods; neurocontrollers; nonlinear control systems; nonlinear dynamical systems; optimal control; probability; dynamic neural systems; error signal; nonlinear control system analysis; nonlinear dynamic systems; optimal parameter determination; parametric synthesis; parametrical synthesis; probability-based analysis; root-mean-square error minimization; second statistical moment determination; statistical linearization method; Control system analysis; Control system synthesis; Control systems; Network synthesis; Neural networks; Nonlinear control systems; Nonlinear dynamical systems; Optimal control; Signal analysis; Signal synthesis;
Conference_Titel :
Neural Information Processing, 2002. ICONIP '02. Proceedings of the 9th International Conference on
Print_ISBN :
981-04-7524-1
DOI :
10.1109/ICONIP.2002.1198967
