DocumentCode :
1268533
Title :
Minimising Added Classification Error Using Walsh Coefficients
Author :
Windeatt, Terry ; Zor, Cemre
Author_Institution :
Centre for Vision Speech & Signal Process., Univ. of Surrey, Guildford, UK
Volume :
22
Issue :
8
fYear :
2011
Firstpage :
1334
Lastpage :
1339
Abstract :
Two-class supervised learning in the context of a classifier ensemble may be formulated as learning an incompletely specified Boolean function, and the associated Walsh coefficients can be estimated without the knowledge of the unspecified patterns. Using an extended version of the Tumer-Ghosh model, the relationship between added classification error and second-order Walsh coefficients is established. In this brief, the ensemble is composed of multilayer perceptron base classifiers, with the number of hidden nodes and epochs systematically varied. Experiments demonstrate that the mean second-order coefficients peak at the same number of training epochs as ensemble test error reaches a minimum.
Keywords :
Boolean functions; Walsh functions; learning (artificial intelligence); multilayer perceptrons; pattern classification; Boolean function; Tumer-Ghosh model; classification error; classifier ensemble; ensemble test error; hidden epochs; hidden nodes; multilayer perceptron base classifiers; second-order Walsh coefficients; second-order coefficients; training epochs; two-class supervised learning; unspecified patterns; Accuracy; Additives; Boolean functions; Complexity theory; Correlation; Error analysis; Training; Classification algorithm; multilayer perceptrons; pattern analysis; pattern recognition; Algorithms; Databases, Factual; Pattern Recognition, Automated;
fLanguage :
English
Journal_Title :
Neural Networks, IEEE Transactions on
Publisher :
ieee
ISSN :
1045-9227
Type :
jour
DOI :
10.1109/TNN.2011.2159513
Filename :
5948414
Link To Document :
بازگشت