Training Winner-Take-All Simultaneous Recurrent Neural Networks

Author

Xindi Cai ; Prokhorov, D.V. ; Wunsch, D.C.

Author_Institution

American Power Conversion Corp., O´Fallon, MO

Volume

18

Issue

3

fYear

2007

fDate

5/1/2007 12:00:00 AM

Firstpage

674

Lastpage

684

Abstract

The winner-take-all (WTA) network is useful in database management, very large scale integration (VLSI) design, and digital processing. The synthesis procedure of WTA on single-layer fully connected architecture with sigmoid transfer function is still not fully explored. We discuss the use of simultaneous recurrent networks (SRNs) trained by Kalman filter algorithms for the task of finding the maximum among N numbers. The simulation demonstrates the effectiveness of our training approach under conditions of a shared-weight SRN architecture. A more general SRN also succeeds in solving a real classification application on car engine data

Keywords

Kalman filters; learning (artificial intelligence); recurrent neural nets; Kalman filter algorithms; car engine data; database management; digital processing; recurrent neural network training; shared-weight SRN architecture; sigmoid transfer function; single-layer fully connected architecture; very large scale integration design; winner-take-all simultaneous neural network; Databases; Engines; Management training; Network synthesis; Process design; Recurrent neural networks; Semiconductor device modeling; Transfer functions; Very large scale integration; Backpropagation through time (BPTT); extended Kalman filter (EKF); simultaneous recurrent network (SRN); winner-take-all (WTA); Algorithms; Artificial Intelligence; Computer Simulation; Decision Support Techniques; Feedback; Game Theory; Information Storage and Retrieval; Models, Theoretical; Neural Networks (Computer);

fLanguage

English

Journal_Title

Neural Networks, IEEE Transactions on

Publisher

ieee

ISSN

1045-9227

Type

jour

DOI

10.1109/TNN.2007.891685

Filename

4182370