Title :
Passivity analysis for dynamic multilayer neuro identifier
Author_Institution :
Departamento de Control Automatico, CINVESTAV-IPN, Mexico City, Mexico
Abstract :
In this work, dynamic multilayer neural networks are used for nonlinear system online identification. The passivity approach is applied to access several stability properties of the neuro identifier. The conditions for passivity, stability, asymptotic stability, and input-to-state stability are established. We conclude that the commonly-used backpropagation algorithm with a modification term which is determined by offline learning may make the neuro identification algorithm robustly stable with respect to any bounded uncertainty.
Keywords :
backpropagation; identification; matrix algebra; neural nets; stability; asymptotic stability; backpropagation algorithm; bounded uncertainty; dynamic multilayer neural networks; input-to-state stability; modification term; neuro identification algorithm; neuro identifier; nonlinear system online identification; offline learning; passivity analysis; stability properties; Asymptotic stability; Backpropagation algorithms; Circuit stability; Multi-layer neural network; Neural networks; Nonhomogeneous media; Nonlinear dynamical systems; Nonlinear systems; Stability analysis; Vehicle dynamics;
Journal_Title :
Circuits and Systems I: Fundamental Theory and Applications, IEEE Transactions on
DOI :
10.1109/TCSI.2002.807519