Feedback procedure neural network and its training

Author

Liang, Jiu-zhen ; ZHAO, Ming-sheng

Author_Institution

Dept. of Electron. & Eng., Tsinghua Univ., Beijing, China

Volume

1

fYear

2002

fDate

2002

Firstpage

101

Abstract

In this paper, a novel neural network model named a Feedback Procedure Neural Network (FPNN) is proposed. The FPNN can accept a functional input that resembles a procedure varying in time and its output is a vector or a scale. Information in an FPNN has two direction forms. One is straight ahead from input to output. The other is starting from input and feedback in the hidden layer, which make a cycle current. A learning algorithm for FPNN is presented which behaves as a supervised training method. Also, two test examples are given to show that FPNN can easily solve procedure problems, which are difficult for a traditional neural network.

Keywords

feedback; learning (artificial intelligence); neural nets; cycle current; feedback procedure neural network; functional input; hidden layer; information direction forms; learning algorithm; procedure problems; scale output; supervised training method; test examples; vector output; Artificial neural networks; Feedforward neural networks; Hopfield neural networks; Multi-layer neural network; Neural networks; Neurofeedback; Neurons; Recurrent neural networks; Self-organizing networks; Testing;

fLanguage

English

Publisher

ieee

Conference_Titel

Machine Learning and Cybernetics, 2002. Proceedings. 2002 International Conference on

Print_ISBN

0-7803-7508-4

Type

conf

DOI

10.1109/ICMLC.2002.1176718

Filename

1176718