Title :
A novel method for online training of dynamic neural networks
Author :
Chowdhury, Fahmida N.
Author_Institution :
Dept. of Electr. & Comput. Eng., Southwestern Louisiana Univ., Lafayette, LA, USA
Abstract :
A fast, efficient, and novel way of online training of dynamic neural networks is presented in this paper. The method is based on a combination of recursive least-squares and backpropagation; in a large number of cases, backpropagation can be avoided altogether. The proposed method would be suitable for real-time identification, fault detection, and control of uncertain dynamic systems
Keywords :
Kalman filters; backpropagation; fault diagnosis; identification; least squares approximations; neural nets; real-time systems; Kalman filter; backpropagation; dynamic neural networks; fault detection; identification; online training; real-time systems; recursive least-squares; uncertain systems; Autoregressive processes; Backpropagation; Equations; Fault detection; Fault diagnosis; Feedforward neural networks; Neural networks; Nonlinear systems; Real time systems; Recurrent neural networks;
Conference_Titel :
Control Applications, 2001. (CCA '01). Proceedings of the 2001 IEEE International Conference on
Conference_Location :
Mexico City
Print_ISBN :
0-7803-6733-2
DOI :
10.1109/CCA.2001.973857
