Title :
Pruning feedforward neural networks through interval arithmetic
Author :
Drago, G.P. ; Ridella, S.
Author_Institution :
Ist. per i Circuiti Elettronici, CNR, Genova, Italy
fDate :
10/29/1998 12:00:00 AM
Abstract :
Interval arithmetic is used to detect the prunable weights of feedforward neural networks. The authors define a regularisation function that can detect irrelevant connections and which contains, as a particular case, the well known weight decay. This method is applied to the modelling of the Mackey Glass series
Keywords :
feedforward neural nets; time series; Mackey Glass series; feedforward neural networks; interval arithmetic; irrelevant connections; prunable weights; regularisation function; weight decay;
Journal_Title :
Electronics Letters
DOI :
10.1049/el:19981704
