Tracking of time varying subspaces using neural networks

Author

Tissanayagam, P. ; Hua, Y.

Author_Institution

Dept. of Electr. & Electron. Eng., Melbourne Univ., Parkville, Vic., Australia

fYear

1994

fDate

29 Nov-2 Dec 1994

Firstpage

46

Lastpage

50

Abstract

The paper discusses the use of an artificial neural network for tracking a time varying subspace. The tracking capability of the APEX (Adaptive Principal Component Extraction) (S.Y. Kung and K.I. Diamantaras, 1990) is analyzed by evaluating an error model. By considering the amplitude of this error, the performance was measured. The simulation results show the ability of the algorithm to track a nonstationary subspace under defined conditions. A mean squared error model is also given, which shows a way of estimating the convergence time for the APEX to track a step change for different small values of learning rate parameters of the algorithm

Keywords

neural nets; signal processing; time-varying systems; tracking; APEX; Adaptive Principal Component Extraction; artificial neural network; convergence time; defined conditions; error model; learning rate parameters; mean squared error model; nonstationary subspace; step change; time varying subspace tracking; tracking capability; Array signal processing; Artificial neural networks; Convergence; Covariance matrix; Iterative algorithms; Neural networks; Neurons; Pattern recognition; Sensor arrays; Signal processing;

fLanguage

English

Publisher

ieee

Conference_Titel

Intelligent Information Systems,1994. Proceedings of the 1994 Second Australian and New Zealand Conference on

Conference_Location

Brisbane, Qld.

Print_ISBN

0-7803-2404-8

Type

conf

DOI

10.1109/ANZIIS.1994.396952

Filename

396952