Premium
Neural control experiments via dynamic neural algorithms
Author(s) -
Fernández De Cañete J.,
GarcíaCerezo A.,
GarcíaMoral I.
Publication year - 1999
Publication title -
international journal of adaptive control and signal processing
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.73
H-Index - 66
eISSN - 1099-1115
pISSN - 0890-6327
DOI - 10.1002/(sici)1099-1115(199906)13:4<273::aid-acs550>3.0.co;2-x
Subject(s) - artificial neural network , backpropagation , feed forward , computer science , feedforward neural network , rprop , stability (learning theory) , algorithm , control theory (sociology) , types of artificial neural networks , time delay neural network , control (management) , artificial intelligence , engineering , machine learning , control engineering
Neural static and dynamic training algorithms have been extensively applied to the control and identification of non‐linear dynamic plants. In the present paper an extension of the static Marquardt learning algorithm, termed Dynamic Marquardt algorithm (DMA) is derived for the on‐line training of neural networks with feedforward and feedback components. The performance of the method has been demonstrated by the neural control of a highly non‐linear experimental fluid level. A stability analysis of the overall control scheme has been carried out using the conicity stability criterion. It has been found that the Dynamic Marquardt algorithm is much more efficient than Dynamic backpropagation, when the relative size of the net is bounded. Copyright © 1999 John Wiley & Sons, Ltd.