User menu

Accès à distance ? S'identifier sur le proxy UCLouvain

Placing spline knots in neural networks using splines as activation functions

Primary tabs

Hlavackova, Katerina [Academy of Sciences of the Czech Republic (Prague)] Verleysen, Michel [UCL]

When using feed-forward neural networks with spline activation functions, the quality of approximation depends on the knot placement of spline functions. We demonstrate a method of choosing equidistant knots in each subdivision of the space when an arbitrary initial division is given, in order to keep the approximation error under a predefined limit.

Document type Article de périodique (Journal article) – Article de recherche
Access type Accès restreint
Publication date 1997
Language Anglais
Journal information "Neurocomputing" - Vol. 17, no. 3-4, p. 159-167 (November 1997)
Peer reviewed yes
Publisher Elsevier BV ((Netherlands) Amsterdam)
issn 0925-2312
e-issn 1872-8286
Publication status Publié
Affiliations Academy of Sciences of the Czech Republic (Prague) - Institute of Computer Science
UCL - FSA/ELEC - Département d'électricité
Keywords Cubic-spline function ; Knots of a spline function ; Approximation error ; Feed-forward neural network
Links
Bibliographic reference Hlavackova, Katerina ; Verleysen, Michel. Placing spline knots in neural networks using splines as activation functions. In: Neurocomputing, Vol. 17, no. 3-4, p. 159-167 (November 1997)
Permanent URL http://hdl.handle.net/2078.1/111331