DocumentCode :
2867885
Title :
Adaptive learning rate and limited error signal for multilayer perceptrons with n-th order cross-entropy error
Author :
Oh, Sang-Hoon ; Lee, Soo-Young ; Shin, Sungmoon ; Lee, Hun
Author_Institution :
Mobile Protocol & Signalling Sect., Electron. & Telecommun. Res. Inst., Taejon, South Korea
Volume :
3
fYear :
1998
fDate :
4-9 May 1998
Firstpage :
2357
Abstract :
Although an n-th order cross-entropy (nCE) error function resolves the incorrect saturation problem of conventional error backpropagation algorithm, the performance of multilayer perceptrons (MLPs) using the nCE function depends heavily on the order of nCE. In this paper, we propose an adaptive learning rate to make the MLP performance insensitive to the order of nCE. Additionally, we propose a method to limit error signal values at the output nodes for stable learning with an adaptive learning rate. The effectiveness of the proposed method is demonstrated in a handwritten digit recognition task
Keywords :
adaptive systems; character recognition; entropy; learning (artificial intelligence); multilayer perceptrons; adaptive learning rate; cross-entropy error; error signal; handwritten digit recognition; multilayer perceptrons; saturation problem; Backpropagation algorithms; Entropy; Error correction; Handwriting recognition; Iterative algorithms; Multilayer perceptrons; Nonhomogeneous media; Pattern recognition; Protocols; Signal resolution;
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.1998.687230
Filename :
687230
Link To Document :
بازگشت