A new model for bidirectional associative memories

Author

Chen, Shanguang ; Jinhe Wei ; Zhang, Yonwun ; Bao, Yong

Author_Institution

Inst. of Space Medico-Eng., Beijing, China

Volume

1

fYear

1996

fDate

3-6 Jun 1996

Firstpage

594

Abstract

In this paper, the structure and performance of basic bidirectional associative memory (BAM) are examined and a BAM model with robust EBP (error backpropagation) algorithm (BAM-EBP) is proposed. In the BAM-EBP model, hidden layers are introduced in the net structure and an improved EBP algorithm is applied. The simulation of the basic BAM and BAM-EBP models in application to fault diagnosis for a control system and the comparison between the two models are carried out. The analysis and results show that both models were effective, but the BAM-EBP has better performance especially in fault tolerance, storage capacity, application adaptability. Future efforts on the theory exploration and application issues are addressed

Keywords

content-addressable storage; control systems; fault diagnosis; neural nets; application adaptability; bidirectional associative memories; control system; fault diagnosis; fault tolerance; hidden layers; robust error backpropagation; storage capacity; Artificial neural networks; Associative memory; Back; Biological neural networks; Control system synthesis; Fault diagnosis; Magnesium compounds; Robustness; Space technology; Vectors;

fLanguage

English

Publisher

ieee

Conference_Titel

Neural Networks, 1996., IEEE International Conference on

Conference_Location

Washington, DC

Print_ISBN

0-7803-3210-5

Type

conf

DOI

10.1109/ICNN.1996.548962

Filename

548962