Impact of the error function selection in neural network-based classifiers

Falas, Tasos; Stafylopatis, Andreas-Georgios

dc.contributor.author	Falas, Tasos	en
dc.contributor.author	Stafylopatis, Andreas-Georgios	en
dc.date.accessioned	2014-03-01T02:48:45Z
dc.date.available	2014-03-01T02:48:45Z
dc.date.issued	1999	en
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/34084
dc.relation.uri	http://www.scopus.com/inward/record.url?eid=2-s2.0-0033352950&partnerID=40&md5=b18460589c9a560cb4f1f1c6b87ecd0f	en
dc.subject.other	Backpropagation	en
dc.subject.other	Computational complexity	en
dc.subject.other	Error analysis	en
dc.subject.other	Functions	en
dc.subject.other	Learning algorithms	en
dc.subject.other	Maximum likelihood estimation	en
dc.subject.other	Multilayer neural networks	en
dc.subject.other	Problem solving	en
dc.subject.other	Kalman-Kwasny error functions	en
dc.subject.other	Maximum likelihood functions	en
dc.subject.other	Mean absolute error functions	en
dc.subject.other	Feedforward neural networks	en
dc.title	Impact of the error function selection in neural network-based classifiers	en
heal.type	conferenceItem	en
heal.publicationDate	1999	en
heal.abstract	This paper presents the results of a comparative study on the impact of various error functions in multi-layer feed-forward neural networks used for classification problems. The objective is the comparison of the properties of the error functions, both in terms of the training speed and the generalization ability. In an effort to avoid complexities introduced by more advanced learning algorithms, the simple back-propagation with momentum algorithm has been employed. A number of classification problems were solved with neural networks that have been trained with the usual mean square error function, the mean absolute error function, the cross-entropy or maximum likelihood function, the Kalman-Kwasny error function, as well as a novel error function designed by the authors. The results indicate that, in most problems examined, an error function other than the usual mean square gives a better performance, both in terms of the number of epochs needed for training, as well as the obtained generalization trained network.	en
heal.publisher	IEEE, United States	en
heal.journalName	Proceedings of the International Joint Conference on Neural Networks	en
dc.identifier.volume	3	en
dc.identifier.spage	1799	en
dc.identifier.epage	1804	en