Comparison of recent methods for inference of variable influence in neural networks

Papadokonstantakis, S; Lygeros, A; Jacobsson, SP

dc.contributor.author	Papadokonstantakis, S	en
dc.contributor.author	Lygeros, A	en
dc.contributor.author	Jacobsson, SP	en
dc.date.accessioned	2014-03-01T01:23:43Z
dc.date.available	2014-03-01T01:23:43Z
dc.date.issued	2006	en
dc.identifier.issn	0893-6080	en
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/17112
dc.subject	Automatic relevance determination	en
dc.subject	General influence measure	en
dc.subject	Information theoretic approach	en
dc.subject	Sensitivity analysis	en
dc.subject	Sequential zeroing of weights	en
dc.subject	Variable influence in neural networks	en
dc.subject.classification	Computer Science, Artificial Intelligence	en
dc.subject.other	Database systems	en
dc.subject.other	Information theory	en
dc.subject.other	Mathematical models	en
dc.subject.other	Sensitivity analysis	en
dc.subject.other	Automatic relevance determination	en
dc.subject.other	General influence measure	en
dc.subject.other	Information theoretic approach	en
dc.subject.other	Sequential zeroing of weights	en
dc.subject.other	Variable influence in neural networks	en
dc.subject.other	Neural networks	en
dc.subject.other	accuracy	en
dc.subject.other	article	en
dc.subject.other	artificial neural network	en
dc.subject.other	Bayes theorem	en
dc.subject.other	information science	en
dc.subject.other	intermethod comparison	en
dc.subject.other	mathematical computing	en
dc.subject.other	mathematical model	en
dc.subject.other	noise	en
dc.subject.other	priority journal	en
dc.subject.other	Algorithms	en
dc.subject.other	Animals	en
dc.subject.other	Artificial Intelligence	en
dc.subject.other	Bayes Theorem	en
dc.subject.other	Computer Simulation	en
dc.subject.other	Humans	en
dc.subject.other	Logistic Models	en
dc.subject.other	Neural Networks (Computer)	en
dc.subject.other	Pattern Recognition, Automated	en
dc.title	Comparison of recent methods for inference of variable influence in neural networks	en
heal.type	journalArticle	en
heal.identifier.primary	10.1016/j.neunet.2005.09.002	en
heal.identifier.secondary	http://dx.doi.org/10.1016/j.neunet.2005.09.002	en
heal.language	English	en
heal.publicationDate	2006	en
heal.abstract	Neural networks (NNs) belong to 'black box' models and therefore 'suffer' from interpretation difficulties. Four recent methods inferring variable influence in NNs are compared in this paper. The methods assist the interpretation task during different phases of the modeling procedure. They belong to information theory (ITSS), the Bayesian framework (ARD), the analysis of the network's weights (GIM), and the sequential omission of the variables (SZW). The comparison is based upon artificial and real data sets of differing size, complexity and noise level. The influence of the neural network's size has also been considered. The results provide useful information about the agreement between the methods under different conditions. Generally, SZW and GIM differ from ARD regarding the variable influence, although applied to NNs with similar modeling accuracy, even when larger data sets sizes are used. ITSS produces similar results to SZW and GIM, although suffering more from the,curse of dimensionality'. (c) 2005 Elsevier Ltd. All rights reserved.	en
heal.publisher	PERGAMON-ELSEVIER SCIENCE LTD	en
heal.journalName	Neural Networks	en
dc.identifier.doi	10.1016/j.neunet.2005.09.002	en
dc.identifier.isi	ISI:000238884300013	en
dc.identifier.volume	19	en
dc.identifier.issue	4	en
dc.identifier.spage	500	en
dc.identifier.epage	513	en