Title :
Diversity/accuracy and ensemble classifier design
Author_Institution :
CVSSP, Surrey Univ., Guildford, UK
Abstract :
For an ensemble of multi-layer perceptrons (MLP), test error is shown in this paper to be well correlated with a diversity/accuracy measure computed between pairs of patterns. A weighted combination that uses the proposed measure to set the weights is shown to be less sensitive to the number of training epochs compared with majority vote.
Keywords :
multilayer perceptrons; pattern classification; diversity/accuracy measure; ensemble classifier design; multilayer perceptrons; pattern classification; training epochs; Bagging; Classification tree analysis; Computer architecture; Decision trees; Design optimization; Interconnected systems; Multilayer perceptrons; Q measurement; Testing; Voting;
Conference_Titel :
Pattern Recognition, 2004. ICPR 2004. Proceedings of the 17th International Conference on
Print_ISBN :
0-7695-2128-2
DOI :
10.1109/ICPR.2004.1334564
