Training trajectories by continuous recurrent multilayer networks

Author

Leistritz, Lutz ; Galicki, Miroslaw ; Witte, Herbert ; Kochs, Eberhard

Author_Institution

Inst. of Med. Statistics, Comput. Sci. & Documentation, Friedrich-Schiller-Univ., Jena, Germany

Volume

13

Issue

2

fYear

2002

fDate

3/1/2002 12:00:00 AM

Firstpage

283

Lastpage

291

Abstract

This paper addresses the problem of training trajectories by means of continuous recurrent neural networks whose feedforward parts are multilayer perceptrons. Such networks can approximate a general nonlinear dynamic system with arbitrary accuracy. The learning process is transformed into an optimal control framework where the weights are the controls to be determined. A training algorithm based upon a variational formulation of Pontryagin´s maximum principle is proposed for such networks. Computer examples demonstrating the efficiency of the given approach are also presented

Keywords

feedforward neural nets; learning (artificial intelligence); maximum principle; multilayer perceptrons; recurrent neural nets; variational techniques; Pontryagin maximum principle; approximation; dynamic multilayer neural networks; learning process; multilayer perceptron; nonlinear dynamic system; optimal control; recurrent neural networks; training trajectories; variational technique; Control systems; Multi-layer neural network; Multilayer perceptrons; Neural networks; Neurons; Nonhomogeneous media; Optimal control; Recurrent neural networks; Stochastic processes; Trajectory;

fLanguage

English

Journal_Title

Neural Networks, IEEE Transactions on

Publisher

ieee

ISSN

1045-9227

Type

jour

DOI

10.1109/72.991415

Filename

991415