A novel forward-backward smoothing based learning subspace method

Author

Huang, De-Shuang

Author_Institution

Beijing Inst. of Syst. Eng., China

Volume

2

fYear

1998

fDate

4-9 May 1998

Firstpage

1113

Abstract

This paper proposes a novel forward-backward smoothing based learning subspace method (FBSLSM), which can satisfy the requirements of being insensitive to the order of presentation of the training samples, and is of faster convergence speed. This method is applied to recognition of high resolution radar targets (three simulated ships). The computer simulation experiments show that the corresponding performance of proposed FBSLSM such as rate of correct recognition and convergence speed is satisfactory

Keywords

convergence; learning (artificial intelligence); self-organising feature maps; FBSLSM; computer simulation; convergence speed; correct recognition rate; forward-backward smoothing based learning subspace method; high-resolution radar target recognition; principal components analysis; self-organising neural network; self-supervised neural network; ships; training sample presentation order insensitivity; Computational modeling; Computer simulation; Convergence; Marine vehicles; Pattern recognition; Radar; Signal processing; Smoothing methods; Systems engineering and theory; Target recognition;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks Proceedings, 1998. IEEE World Congress on Computational Intelligence. The 1998 IEEE International Joint Conference on

Conference_Location

Anchorage, AK

ISSN

1098-7576

Print_ISBN

0-7803-4859-1

Type

conf

DOI

10.1109/IJCNN.1998.685928

Filename

685928